personal communications

Tailoring of services, servicemobility, personal mobility andterminal mobility are essentialelements in personal communi-cations. With the integration offixed and mobile services, andinterworking between differenttechnologies, personal communi-cations will soon be available.The transition from plain oldtelephony to true personal ter-minals will then become a majorstep in the history of telecom-munications. The mobileoperators may within few yearschange from being mobilenetwork operators to being pro-viders of value added services topersonal terminals.

The European Commission’sGreen Paper on mobile and per-sonal communications (1994)said: “... personal communi-cations has the ultimate potentialto reach near 80 % of the popu-lation (ie. up to one connectionper adult).” This is far beyondtoday’s penetration of wireline phones, even in industrialisedcountries. With the introduction of personal multimedia com-munications the vision of ‘the wireless’ as the customer’s firstchoice will become true.

This issue focuses on the transition from mobile telephony topersonal communications. We are used to seeing people com-municating via true pocket cellular phones in almost anysituation. Mobile communications has, during the last decades,evolved from being the businessman’s tool to becoming theconsumer’s means of receiving and making calls when on themove. This highly affects the social life of the mobile users,giving them the possibility of being available whenever wanted– avoiding ‘being held prisoner’ by the wireline phone.

The world-wide success of mobile communications, based onthe GSM standard, has led to a tremendous interest for mobilesystems. The growth of cellular services has been boosted bymobile communications being exposed to competition. Newoperators are entering the scene, and their best means to get apiece of the still growing telecommunications market is by useof radio solutions.

Until now the main cellularservice has been speech, eventhough data services have beenpresent in the operationalnetworks for several years. How-ever, over the last years the focusof the mobile society has changedfrom speech to data services dueto the tremendous growth in thedemand for information servicesand Internet access.

The next step will be integrationof the different fixed, cellularand Internet services. Mobilityfunctions in both wireless andfixed networks, and flexibleservice creation and manage-ment makes the differencesbetween fixed services, cordlesstelephony and cellular diminish.Introduction of packet switcheddata into the GSM network, likethe General Packet Radio Ser-vice (GPRS) is turning the ori-ginal circuit switched GSMnetwork into a hybrid network.

The standardisation of a third generation cellular system –Universal Mobile Telecommunications System (UMTS) hasbeen going on for several years. In addition, real broadbandmobile systems offering bit rates of up to 155 Mbit/s are beingdeveloped.

UMTS is aimed at giving access to multimedia applications.It predominantly evolves from GSM, and will interwork withGSM, possibly also picking up elements from the DECT stand-ard. In the longer term, integration between UMTS and IPnetworks offering differentiated quality of services will be thenext evolutionary step. UMTS may then give global access tomultimedia services across platforms like mobile, fixed andsatellite-based networks.

Mobility across terminals, locations and infrastructure togetherwith tailoring of high quality services will make the vision ofcommunicating anytime, with anyone, anywhere come through.The broadening of mobile communications into general per-sonal communications will substantially affect the lives ofindividual citizens as well as the functioning of the society.

1

Guest editorialR U N E H A R A L D R Æ K K E N

Telektronikk 2.1998

2 Telektronikk 2.1998

Mobile telephony – a retrospective glance

From land coverage to cellular services

Mobile communications in the form ofcommunications between a base stationand a number of mobile stations has beenavailable for several decades. Communi-cations between land and ships usingshort wave radio is an example of suchan application that has been in use in thelast century. Early communications tookplace using Morse telegraphy; later, itwas changed to voice telephony.

Land mobile applications have beenavailable since the 1930s, in the form ofclosed user groups connected to a basestation. The police were among the pilotusers of such services.

Since then, several public land mobileservices have been launched. In Norwaya service called OLT (Offentlig Landmo-bil Tjeneste = public land mobile service)was launched in the late 1960s. All sub-scribers to the OLT service listened tothe call channel and were requested byan operator to change to a traffic channelwhen a call was received. To set up a callto a mobile subscriber, one needed toknow the approximate position of themobile subscriber, in order to direct thecall to the appropriate paging area.

It soon became evident that this was notthe way to offer mobile communicationsto the mass market. Capacity problemswere envisaged due to limited availableradio spectrum. Hence, it was regardedan advantage to be able to use a cellularlayout of the base stations to allow reuseof the radio frequency spectrum to raisethe total system capacity. During thesame period, the 1970s, switchboardoperators were removed, and calls in thewireline networks were set up auto-matically. Hence, the idea of forming anautomatic international mobile networkemerged in the Nordic telecommunica-tions administrations. Among the objec-tives for the system were [1]:

• The system should automatically setup and charge a call

• It should be possible to communicatebetween any mobile subscriber andwireline subscriber, or between anytwo mobile subscribers

• The system should operate in all theNordic countries

• The use of the mobile phone should beas similar as possible to the use of thewireline phone

• The mobile subscriber should haveaccess to as many as possible of thewireline services

• Calls to the mobile should be auto-matically routed to the position of themobile, the mobile should be roamingin the network and data bases shouldkeep track of its position

• The system should be able to auto-matically switch from one base stationto another during conversation.

Such characteristics of a mobile commu-nications system seem inevitable today,but particularly the two last bullets werequite revolutionary in the late 1970s.

The first steps towards the NordicMobile Telephone (NMT) system hadbeen taken.

The analogue mobile telephone systems – first generation cellular

In 1981 the world’s first automatic cellu-lar international mobile telephone systemwas put into commercial operation in thecountries of Norway, Sweden, Denmarkand Finland. When the system was ini-tiated, it operated in the 450 MHz regionand was therefore named NMT 450.

The NMT system was mainly designedfor speech communications, and the sys-tem relies on analogue frequency modu-lation (FM). However, the signalling isdigital, with a transfer rate of 1200 bit/s.

Data communications is also possibleusing special modems. However, sincethe mobile radio channel is rapidlychanging, it may be difficult to transfererror-free data in a system with no dataerror correction methods present.

The first versions of NMT terminalsweighed more than 10 kg and weremainly manufactured for installationin cars. With today’s measures the equip-ment would be regarded as highly im-practical, expensive to buy and expensiveto use. Despite this, the launch of theNMT system was a great enhancementin making public telephony availableoutside the wireline network. The growthof traffic in the NMT system was thusmuch higher than the most optimisticprognoses, and capacity problems weresoon experienced. The NMT system wastherefore extended to the 900 MHz bandin 1986. The system is known as NMT900.

In the late 1980s hand portable mobilephones started to emerge. The first onesweighed about 750 grammes, at that timeregarded as remarkably small for a cellu-lar phone. Miniaturisation has continued,and today true pocket phones are avail-able for the NMT system.

From mobile telephony to personal communicationsJ O A R L Ø V S L E T T E N A N D R U N E H A R A L D R Æ K K E N

Figure 1 First generation cellularsystems connected to the public switched telephonynetwork (PSTN). Characteristics of first generation sys-tems: Analogue transmission, basic speech service, datacapabilities not built into systems, limited roaming capa-bilities

NETZ-C

11

11

11

11

00 12 10

TACS

11

11

11

11

00 12 10

NMT

11

11

11

11

00 12 10

PSTN

3Telektronikk 2.1998

Both NMT 450 and NMT 900 operate inparallel with GSM today, and there arestill some 3,500,000 subscribers to theNMT service throughout Europe [2].

In addition to NMT there are severalother systems belonging to what we callfirst generation cellular mobile phones,like the AMPS (Advanced Mobile PhoneServices), TACS (Total Access Commu-nication Systems), NETZ-C, MATS-Eand others, totally holding some 93 mil-lion subscribers world-wide [3].

Next step – pan-European cellular

During the 1980s several automaticmobile telephony systems had grown upthroughout Europe, mostly designed andset up to meet national needs for mobiletelephony. Parallel to this, standardisa-tion work was going on to design aharmonised mobile communications sys-tem for use throughout Western Europe.

Reservation of a common frequencyband throughout Western Europe, withthe European Post and Telecommunica-tions Union (CEPT) as a co-ordinator ofthis work, made it possible to set timepressures and national industry policiesaside and agree upon a common standardfor mobile communications for a unifiedWestern Europe.

Some objectives for the system designperformed by the Group Spécial Mobile(GSM) were [4]:

• Mobile stations should be usedthroughout Europe. The system shouldroute calls automatically to a mobile inany position inside the coverage area

• In addition to plain telephony, adaptedISDN services should also be madeavailable to the mobile subscribers

• The system should offer services to carmounted as well as hand portable terminals

• Speech quality should be at least asgood as for NMT 900 under goodconditions

• The system should allow for encryp-tion of user data over the air interface

• The system should utilise radio spec-trum in an optimum way

• The system should not demand largeupgrades of the fixed network

• It should be possible to have differentbilling structures in the different GSMsystems

• It should be possible to have severalGSM systems in one country.

The goal was that the GSM systemshould be in operation by 1991. In Nor-way a trial service was launched in 1992,and both Norwegian operators put GSMinto commercial operation in 1993. Sec-ond generation cellular was born.

There are some 82 million users of GSMworld-wide by mid-1998 [5].

Digital communications – tailored for data commu-nications?

GSM and its sister systems D-AMPS andIS-95 in USA and PDC (Personal DigitalCellular) in Japan are digital by nature,hence they are better suited to data com-munications than the first generationcellular systems. Also, error correctingmechanisms are present, making errorfree data communication more easilyavailable to the customers.

Available services in the GSM networksare for instance:

• Speech

• Emergency calls

• Facsimile

• Data services (2.4, 4.8, 9.6, 14.4kbit/s)1

• Short message service, point-to-pointas well as point-to-multipoint

• Supplementary services like call for-warding, calling line identification pre-sentation and calling line identificationrejection

• Mobile information services.

In addition, development of cordless sys-tems like the DECT (Digital EnhancedCordless Telecommunications) systemhas taken place. DECT is designed as abusiness and residential cordless systemwith higher bit rates available than GSM,but with a much less robust air interface.Hence, DECT is mainly intended forindoor wireless applications.

Being digital, the second generationmobile systems are better tailored to datacommunications than first generationcellular services. Still, a vast majorityof the connections made using GSM isfor speech telephony. For severalreasons, only the very advanced usershave started using their GSM for messag-ing and data communications:

• Terminals are not optimised for datacommunications because of the man-machine interface (too small keys anddisplays)

• High user threshold to get started usingdata services

• Users are unfamiliar with messagingservices.

1 Enhanced data services like HSCSD(High Speed Circuit Switched Dataservices) and GPRS (GeneralisedPacket Radio Service) are underspecification in what has been namedGSM phase 2+.

DECT

11

11

11

11

00 12 10

GSM

D-AMPS

11

11

11

11

00 12 10

11

11

11

11

00 12 10

IS-951

11

11

11

1

00 12 10

PDC

ISDN

Figure 2 Second generation systems connected to the integrated services digital network.Characteristics of second generation cellular systems: Digital transmission, basic speechservice, supplementary services, limited data capabilities, continental roaming


Popular – but still not the first choice

The mobile market has experienced atremendous growth in the last years,boosted by the mobile operators subsi-dising terminals to attract customers totheir network. At the moment (mid-1998),the mobile penetration in Norway hassurpassed 40 %2 of the population.

We do however note that the cellular isstill not the customers’ first choice. Mostcellular users still have both a telephone

at the office and a wireline phone athome. However, it is interesting to notethat the mobile penetration is approach-ing the fixed phone penetration (includ-ing both residential and business connec-tions) in the pioneering mobile markets.

There is, however, one major differencebetween the wireline phone and themobile phone. Experience shows thatpeople are unwilling to share mobilephone, hence mobile phones are oftenpersonal, as opposed to fixed phoneswhere a location and not a person isaddressed.

The UMTS vision

In Europe third generation cellular sys-tems have been called UMTS – Univer-sal Mobile Telecommunications System.Research on UMTS started – typicallyenough – in the late 1980s, several yearsbefore the GSM standardisation wasfinalised.

The UMTS key objectives are summar-ised as follows [7]:

• Integration of services and applicationareas. Within one system UMTS willsupport services and applicationswhich today are provided by dedi-cated systems. Target service areasare paging, mobility functions, mobiledata communications, mobile tele-phony, public cellular applications,private business applications, resi-dential cordless applications, wirelessPABX applications and mobile satel-lite access.

• Integration of fixed and mobile net-works through integration of fixed andmobile system technologies

• High quality of service, comparablewith the fixed network

• Services requiring a variety of bit rates(up to 2 Mbit/s in phase 1) and vari-able bit rate services, including multi-media services

• Global terminal roaming capability,enabling the user to access UMTS ser-vices in all regions of the world

• Satellite and terrestrial based coverage,also extending UMTS coverage toareas where it is not techno-economi-cally feasible to provide terrestrial cov-erage.

This leads to the UMTS vision of com-municating anytime, with anyone, any-where.

When research on UMTS started in thelate 1980s, UMTS was regarded as re-volutionary compared to second genera-tion systems like GSM and D-AMPS.Later, several of the UMTS great

2 By the end of 1997, the mobile pene-tration in Norway was 38.5 %, onlyexceeded by Finland with a penetra-tion of 42.1 % [6]. In Finland theysoon expect to reach a penetration of50 %.

Figure 4 UMTS is recognised as an opportunity to provide mass market wireless multimedia servicesat any location. A fully developed UMTS may give global access to multimedia services across plat-

forms as mobile, fixed and satellite-based networks

Figure 3 Terminals will increasingly communicate with servers. It is expected thatwithin few years 50 % of the mobile traffic will be data communications

Java

Webserver

Internet (IP)

Mobile

Fixed


thoughts have been implemented byenhancement of second generation sys-tems or by interworking between secondgeneration systems.

• Global coverage: GSM and its twins(DCS 1800, PCS 1900) are headingtowards global operation. 293 adminis-trations, network and satellite opera-tors from 120 countries have signedthe MoU stating that they want to pro-vide access for GSM users

• Service integration: GSM’s point-to-point short message service is anadvanced form of radio paging (withstore and forward and receipt to senderwhen message is received) integratedinto a cellular system

• Interworking between systems: stan-dardisation of interworking betweencellular and cordless systems (DECTand GSM) is going on within ETSI

• Integration of satellite components:systems for handheld satellite basedcommunications are being planned orimplemented, like ICO, Iridium andGlobalstar. All these system conceptsrely on combination terminals commu-nicating with the land based infrastruc-ture when available, otherwise relyingon satellite based communications

• Personal mobility: GSM is based onthe use of SIM cards containing thesubscriber’s identity (allowing for‘plastic roaming’ which for instancewas offered to GSM customers duringthe Nagano Winter Olympic games)

• Tailoring of services: Intelligent net-work functinality is being imple-mented in the fixed as well as in theGSM networks, giving room for tailor-ing of services and service mobility.

Hence, it seems that services and net-work aspects of UMTS is predominantlygoing to be an evolution rather than arevolution from the second generationsystems. With the ongoing upgrade ofGSM to EDGE (Enhanced Data rates forGSM evolution) to offer bit rates up to384 kbit/s, it is also a question of defini-tion if UMTS Phase 1 is going to be anevolution or a revolution compared to theenhanced GSM radio subsystem.

In later stages of the UMTS implementa-tion, integration with IP networks offer-ing differentiated quality of service willtake place. In addition, UMTS maysupport access networks, providinghigher bit rates than 2 Mbit/s. UMTSmay then give global access to multi-

media services across platforms asmobile, fixed and satellite-based net-works, turning UMTS into the real uni-versal telecommunication system.

The plan is to implement UMTS in 2002.

Mobile broadband systems

In addition to the UMTS work, realbroadband mobile systems are alsobeing developed, like in the ACTS pro-ject SAMBA (System for AdvancedMobile Broadband Applications).

At the moment work is focusing on atrial platform providing transparent ATM(Asynchronous Transfer Mode) connec-tions and supporting bearer services at upto 34 Mbit/s in a cellular radio environ-ment operating at 40 GHz.

The ultimate aim for SAMBA is to de-velop mobile broadband systems offeringbit rates of up to 155 Mbit/s to mobileusers.

Looking into the crystal ball

A general trend is that when new serviceshave been introduced in the fixed net-work, a demand has arisen for the sameservices also being available in themobile network.

Trends – mobility

One trend that has been present sincemobile telephony was introduced is the

aim of increased mobility. With the intro-duction of the S-PCN (Satellite PersonalCommunications Networks) based onLow Earth Orbit satellites the coveragewill be ubiquitous.

Four types of mobility are being referredto:

• Personal mobility: one access numberis connected to a person, regardless ofterminal and access point. One sort ofpersonal mobility is the UPT (Univer-sal Personal Telecommunications) ser-vice in the fixed network, another formof personal mobility is SIM card roam-ing (‘plastic roaming’) within GSM

• Service mobility: uniform access to thesame set of services across terminaland service platforms

• Terminal mobility: continuous mobil-ity across locations, relying on radiobased mobile terminals that can beused within the radio coverage area

• Session mobility: during a communica-tion session the user can move be-tween service platforms. One exampleis a person being alerted on her radiopager that someone wants a video con-ference with her. Then the user has thepossibility to transfer the session to aservice platform offering the requestedservice.

Trends – FMC

FMC has been a buzz word for sometime, an acronym with different inter-pretations from person to person. FMCcould be bundling of mobile and fixedservices by the service provider, or even

Figure 5 Possible evolutionary paths from today’s telecommunications solutions tothe UMTS vision. UTRA stands for UMTS Terrestrial Radio Access

UTRA

UMTS

Satellite

IPIPv6

Mobil IP ???

ISDNIN CS-1

DECTB-ISDN

IN CS-3

GSMCAMEL

HSCSD

GPRS

ICOIridium

GlobalStarTeleDesic

Mobilenetworks

Fixednetwork

Datanetwork

EDGE

W-ATM


issuing one bill for mobile and fixedsubscriptions. The service provider’sbundling of services is, however, de-pendent on the regulatory situation.

One-number services which are offeredby several service providers can be real-ised in different ways. One solution isusing DECT/GSM dual mode terminalswith support of network functionality tokeep track of the terminals both fixed andmobile networks are used.

By using GSM home base stations, whichare being standardised, a one-number ser-vice could be offered based entirely on

the mobile platform if the home basestation is connected to the MSC.

In the same way, DECT as access to thefixed network can be used, eg. the Fidoservice offered by Telecom Italia, wherea DECT terminal can be reached withinthe area of a city. For coverage beyondthe city border, a combined DECT/GSMterminal is needed.

There is a general trend that similar ser-vices can be based on a mobile or fixednetwork platform. The next generation ofIP also provides for some mobility, eg.suited for nomadic computing, and Inter-

net is an interesting platform for personaltelecommunications services.

Trends – multimedia services

Another trend is that the users are re-questing a variety of services requiringdifferent service quality and different bitrates. Examples of such services are:

• 7 kHz audio (AM quality)

• 20 kHz audio (CD quality)

• video telephony

• videoconferencing (128 – 768 kbit/s)

• messaging services

• telefax group 4

• data base access

• broadcasting services

• Internet access.

These services are made available tomobile users, either by designing systemsfor higher bit rates or by accommodatingprotocols to tailor the services to the bitrates being offered by the various radiosystems.

Another consequence of multimedia ser-vices is that the demanded bandwidthmight be different in the different linkdirections.

Figure 7 Traditional telecom operators have a different view of FMC from the pure mobile operators

Traditional teleoperators

Fixed network Mobile network

Pure mobile operators

Mobile network

Integration

Figure 6 The buzz word FMChas different meaning to dif-ferent people. To reach com-plete integrationbetween mobile, Inter-net and fixed ser-vices, thereneeds to be inte-gration atdifferentlev-

Commmon transmission/switching platforms

Commmon network intelligence platforms

Commmon management/billing system

Network independent services

Single points of contact

Bundling and packaging


Trends – wireless application protocols

The design of advanced data applicationsand services based on Internet Web tech-nology in mobile environments hasrecently started. Up to now Internet tech-nology has been aimed at powerful PCsutilising access networks like ISDN orEthernet. To fulfil the demand for mobil-ity without limiting the communicationcapabilities, future application platformsmust handle communication platformswith lower transmission capabilities,using less powerful terminals like smart-phones, microbrowsers, PDAs (PersonalDigital Assistants), palm top computers,etc.

Within year 2000 the prognosis is for 22million users world-wide using othertypes of terminals than PCs to access theWeb [8].

The first joint initiative to standardisinga protocol for wireless Web access is thework on the Wireless Application Proto-col (WAP) in the WAP Forum, driven byNokia, Unwired Planet, Motorola andEricsson. WAP Forum has recently pub-lished the draft WAP specification, iden-tifying a set of protocols and program-ming languages, which will allow furtherdevelopment of mobile phones intomicrobrowsers.

Up till now the mobile terminals havebeen much less standardised than thetraditional PCs using MS Windows asoperating system. For mobile terminalsthere is an ETSI standard for applicationson SIM (Subscriber Identity Module)cards named SIM-ToolKit. The idea isthat program code stored on the chipinside the SIM can be used to run anapplication dependent code, allowingfor remote applications downloaded bythe operator and accessed via the termi-nal. Hence, for instance user interfacecan become a competitive factor.

The introduction of the WAP is based onthe assumption that there is a demand forInternet access from portable terminals.WAP has paved the way for making avariety of terminals, including the mass-market handsets, true information app-liances.

Trends – the telecommunicationterminal as a module in amultifunctional device

Bringing the cellular phone and thepersonal organiser more or less whereverthey go is quite common among peopletoday. There is a trend towards usingelectronic personal organisers. Productscontaining both an organiser and amobile phone in one package are on themarket. As electronic commerce is be-coming more and more widespread thenext step of integration could be thateven the wallet turns electronic and be-comes part of a communication enhancedPDA. Keys and physical access controlcould also be made electronic and inte-grated in a multifunctional device.

A development in the direction of inte-grating what people usually carry in theirpockets and which can have an electronicversion into one device, will depend onthe level of security that can be obtained.Satisfactory security solutions to preventfraud, eavesdropping or unauthorised useare essential. Acceptable ways of han-dling lost or stolen devises also have tobe found.

Personal communication– what will it be?

Personal communications is an ambig-uous term, eg. in USA, Personal Commu-nications Service (PCS) is based on sec-ond generation mobile technology.In Europe the term Personal Communi-cation Networks (PCN) was used whenthe first DCS 1800 networks were estab-lished.

The basic understanding of personalcommunications implies that a person

can be reached on one number at anytime wherever they are and that there isno limitation on outgoing calls.

Using the fixed network personal com-munications can be achieved if the in-coming call is always routed to a termi-nal located close to the wanted person– personal mobility is mobility acrossterminals. Utilising personal mobility thenetwork must know where to route theincoming call. This means that some kindof action is needed from the user toinform the network of which terminalthey can be reached at. When leaving thelocation the network must also be in-formed. To register and deregister seemstoo much to ask from users.

Figure 9 When personal communications is present thebasic telecommunications device is a true pocket terminaloffering speech services, Internet access and various mul-timedia services. All services made available via the per-sonal terminal are accessed using one personal number,which can be the same during the customer’s life cycle

Figure 8WAP architecture. The idea is to offer wirelessWeb access by filtering of HTML (HyperTextMarkup Language) to WML (Wireless Markup Language) and thus adapt the web interface to the capabilities of the mobile terminals

TeleVASServer

Filter

WAP Proxy

WebServer

Filter

WAPProxy

WirelessNetwork

HTML WML

WML

WML

WML

WML

HTML


The alternative to personal mobility isterminal mobility, ie. users bring a per-sonal terminal with them. Terminalmobility offers contiguous mobilityacross locations. The penetration ofcellulars is surpassing 40 % in the pio-neer markets, indicating that terminalmobility will be the basic element inpersonal communications.

In Europe potential technologies for per-sonal communications based on terminalmobility are GSM, DECT and UMTS.Standards are made for both GSM andDECT to handle high bit rates and also tohandle packet data. Multimode terminalscapable of accessing different networksmight be an important part of personalcommunications and users will have achoice from limited local mobility to fullglobal mobility.

In the European Commission’s GreenPaper on Personal Communications [9] itsays: “With the broadening of mobilecommunications into general personalcommunications, the sector will, beforethe end of the decade, substantially affectthe lives of every citizen of the EuropeanUnion” and “– personal communicationshas the ultimate potential to reach near80 % of the population (ie. up to oneconnection per adult.)”

There are already trends in some coun-tries that mobile subscriptions are start-ing to replace fixed line subscriptions.

Rune Harald Rækken is Senior Research Scientistat Telenor R&D, where he has been employedsince 1987, working with standardisation of theGSM system, with radio propagation aspects,FMC, and mobile communications in general.He is currently in charge of the mobile and personalcommunications group at Telenor R&D.

e-mail: [email protected]

Joar Løvsletten has been with Telenor sincegraduating from the Norwegian University ofScience and Technology in 1976. He worked inthe Network Department, dealing with radio relaysystems, until 1990, when he started a group work-ing on radio access systems. Løvsletten came toTelenor R&D in 1995 and is now Senior Engineerworking with personal communications.


The first phase of personal communica-tions seems to be a change from POTS(plain old telephony system) to personalterminals.

Personal communications is presentwhen the basic telecommunicationsdevice has become a personal pocket ter-minal instead of a wireline terminal andthe standard subscription implies using apersonal pocket terminal instead of POTS.

A next phase, following the transitionfrom POTS to personal terminals, iswhen services beyond speech, like Inter-net access and multimedia services, willbe available over pocket terminals.

The technology will be able to handlethe development outlined. The userdemand for personal communications isof course uncertain, but it seems thatusing personal terminals will be a re-quirement.

The introduction of personal pocket ter-minals creates new problems. For ex-ample, nobody wants to be available atall times. The situation and the surround-ings have to be taken into account beforemaking or receiving calls. In aircraft, forexample, the use of cellular phones isprohibited and few will use the phoneduring a dinner party. Silence modes formonitoring incoming calls already exist,but users have to learn to behave whenusing pocket terminals. In general,

flexible ways of handling received callswhen not available will be a requirementwhen personal pocket terminals are used.

Will users want to use only one terminal,their personal pocket terminal, both atwork and private? Will the differentiationbetween working and non-working hoursdecrease? These, among others, are ques-tions that might be raised regarding theintroduction of personal communications.

The transition from POTS to personalphones carried at all times, will be amajor step in the history of telecommu-nication and will of course influencepeople’s everyday life and also have animpact on the functioning of society.

References

1 Løken, B et al. NMT-systemet : etfellesnordisk automatisk mobiltele-fonsystem. Telektronikk, 75, (2),1979.

2 European analogue & digital cellularsubscriber up-date. Mobile Commu-nications International, 50, April1998.

3 Helme, S. There’s still life in the olddog yet. Mobile CommunicationsInternational, 50, April 1998.

4 Audestad, J A. Det felleseuropeiskelandmobile system : status og fram-drift. (The Pan European land mobilesystem. Status and drive.) Guest lec-ture in the course “Radio systems” atthe Norwegian University of Tech-nology, 1986. Paper in Norwegian.

5 The GSM MoU Association’s website (http://www.gsmworld.com/)

6 Mobile Europe, February 1998.

7 Bjørnland, D F, Lauritzen, G O.UMTS : the Universal MobileTelecommunications System. Telek-tronikk, 91, (4), 127–132, 1995.

8 ComputerWorld Norge, 37, 1997.

9 Commission of the European Com-munities. Towards the PersonalCommunications Environment.Green Paper on a common approachin the field of mobile and personalcommunications in the EuropeanUnion, 1994.

This article considers the next genera-tion mobile communications systems.The characteristics of these systemsare described by outlining the systemrequirements and the services en-visaged. Further, a short overview ofEuropean research activities is given,followed by an overview of standardis-ation activities within ETSI and ITU.The work on next generation mobilecommunications systems is now in anintense stage, aiming at finalising thefirst sets of specifications by the end ofthe century. Details of the system con-cept are thus subject to continuouschange, and in order to avoid publish-ing outdated results, this article there-fore takes a rather general approach.

1 Introduction

In Europe and in the Nordic countries inspecific, the history of automatic landmobile systems [1] starts in 1969 withthe development of NMT (Nordic MobileTelephone). The analogue NMT systembecame a success, partly because of itsrobustness and simple mobile stations.Basic telephony with roaming possibilitywas the service offered. NMT hit themarket at the right time, and is con-sidered to be a first generation cellularsystem. Other first generation systemsemerged, notably TACS and Netz-C.

Second generation systems, based ondigital technology all the way from theinterconnection point to the mobilestation, emerged with the introduction ofGSM. The development of GSM startedalready in 1982 within CEPT (EuropeanPost and Telecommunication Confer-ence), and was carried on in ETSI from1988. As for the analogue systems, thedominant service of GSM has until nowbeen basic telephony. GSM introducedthe separation of terminal and user iden-tity, which enabled enhanced securityand user roaming across terminals.

GSM was launched for the first time forcommercial operation in 1991; in 1997more than 100 countries had at least oneGSM network. GSM is a success, nowdeployed in almost every corner of theworld.

The research society moved onto thirdgeneration studies many years beforeGSM was launched for commercialoperation. The idea that a third genera-tion system should consist of differentaccess components, including a domi-

nating terrestrial access component pro-viding at most 2 Mbit/s, was establishedrelatively early. Regarding terminals andservices, however, the focus of attentionin the early days centred on the conceptof a personalised lightweight hand port-able. This is now commonplace also insecond generation systems, and the focusfor third generation has been widened toinclude a wide range of terminals tosupport various services, including multi-media.

On the standardisation arena, an earlystep towards the standardisation of athird generation mobile system was takenin 1985 by the establishment of CCITTInterim Working Party 8/13 (IWP 8/13),now known as ITU-R Task Group 8/1.The group was responsible for the stan-dardisation of Future Public Land MobileTelecommunications Systems (FPLMTS),now known as IMT-2000. ETSI followedthe ITU in 1991 with the formation of aSub Technical Committee called SpecialMobile Group 5 (SMG 5) responsible forthe standardisation of a third generationmobile system called Universal MobileTelecommunications System (UMTS).1

Perhaps the most important result fromthe first years was the contribution toWARC 92 [2], resulting in an allocationof 230 GHz for future mobile communi-cations in the 2 GHz band. Being domi-nated by operators and academia, therewere strong forces within standardisationto repeat the GSM success by standard-ising a new system from scratch. Later,the view has changed to take moreaccount of existing technologies. This isparticularly pronounced in ETSI, wherethe legacy from GSM is strong. Thedevelopment takes into account the factthat the UMTS technology should comeabout from an evolution of GSM, or atleast the interoperation with GSM shouldbe eased. This ensure that the vastamount already invested in GSM technol-ogy pays off also by the introduction ofUMTS. Further, this enables the accessto UMTS or the provision of UMTS-likeservices based on the GSM access, whichis an important goal for GSM operators.

The development of third generationstandards is gradually becoming moremarket driven. Large markets are stilluncovered by mobile services, at thesame time as more advanced markets

demand new and better services. Twomega trends which are important en-ablers are the ‘Internet explosion’ andthe huge growth in cordless and mobileterminals. Indeed, due to this marketprospect, large resources are alreadyspent on development, both from indu-stry and operators/service providers. Itis easy to forecast that the rate of invest-ment in third generation developmentwill increase even further in the nearestcoming years.

However, the development is not solelymarket driven. Today’s extremely highcompetence and know-how withinmobile communications compared toprevious years represents in itself adriving force. On the one hand, operatorsand service providers have developedadvanced skills for network deploymentand operation as well as for servicedesign and provision. On the other hand,equipment manufacturers make continu-ing technological improvements support-ing the possibility of increasing systemcomplexity and capability, as well asreducing the equipment size and price.Developments contributing to this arefound within microelectronics, modula-tion and coding, battery technology,antenna design, MMI, etc.

As opposed to this good basis for de-velopment, there is, however, a possi-bility that the development is sloweddown or threatened by diverse interestsin new technology. This may be experi-enced from industry or operators aimingto exploit beneficial positions as pro-ducers and owners of existing technologyor systems. Having said that, a situationrepresenting a serious threat to the de-velopment is not anticipated; the ideasto evolve the development from existingtechnology have rather contributed tocost-effective and viable solutions. An-other situation that may seem morethreatening to the growth of third genera-tion systems is the availability of spec-trum. The bands identified by WARC-92are shared with other radio communica-tion systems and services, and are in manycases already in use, and surveys show thatadditional spectrum is required [2, 17].

2 Services and System Requirements

2.1 Overview

First generation systems offer a switchedtelephony service to users who may

9

Third Generation Mobile CommunicationsJ A N E R I K S E N , G E I R O L A V L A U R I T Z E N , B J Ø R N H A R A L D P E D E R S E N A N D S T E I N S V A E T

Telektronikk 2.1998

1 The UMTS concept was strongly in-fluenced by the RACE Mobile (R-1043)project at the end of the 1980s.

move freely within the service area. Thismay be referred to as a plain old tele-phony system (POTS) with roamingcapabilities. In second generation sys-tems, the service repertoire has beenextended to include not only POTS, butalso a limited set of data services.Further, the mobility management func-tions are more powerful with more intel-ligent signalling. The concept of mobilityhas also been widened introducingmobility of users across terminals byseparating the user and terminal identity.Seen from the user, however, there isnot much difference in the functionalityoffered by first and second generationsystems. Both the bandwidth limitation

and the rigidly standardised set of ser-vices are obstacles for introducing ad-vanced data services. The radio interfaceis designed for narrowband transmission,and the systems rely on switched narrow-band connections in the fixed infra-structure.

Major goals of third generation systemsare to overcome the limitations in band-width and to provide flexibility for ser-vice provision. This paves the way for awide range of services, including multi-media, and supports personification ofthe services. A goal of third generationsystems is also to provide a uniformaccess to personal communication based

on different access networks, both wiredand wireless, in different operating en-vironments. The aim is to provide per-sonal communication anytime and any-where.

2.2 Service Provision inDifferent Environments

Third generation systems concentratecurrently on new radio access networksthat can provide services up to 2 Mbit/s.Although this type of access network willbe a major component, the third genera-tion systems also intend to provideaccess through enhancements of existingradio access networks (eg. GSM), as wellas through wired access networks. At alater stage, microwave radio access net-works may be introduced which providebearer capabilities far beyond 2 Mbit/s.As will be discussed below, the servicesprovided will not only depend on thetype of access network, but also on theenvironment in which the network ope-rates.

It is foreseen that wireless will graduallybecome the access of choice for all per-sonal communications. Third generationsystems aim to provide a quality of ser-vice for wireless access that is com-parable to that of wired access. Servicemobility is supported, so that the serviceis available to the user in a uniform wayregardless of the environment. Althoughthe services are similar both in qualityand appearance in different environ-ments, the bit rate of the bearer servicesprovided will depend on the type ofaccess network and the operating en-vironment. Services requiring the highestbit rates will therefore not be supportedin certain access environments. However,many services are expected to be de-signed in a flexible way, so that they canbe carried on bearers with different bitrates depending on the capabilities of theaccess network.

It is required that third generation sys-tems accommodate different types ofterminals which may have different capa-bilities for supporting the various ser-vices. The actual services provided willthus also be dependent on the capabilitiesof the terminal.

As discussed above, the interdependencebetween terminal capabilities, environ-ments and the services provided haveimplications on the system requirements.This is illustrated in Figure 2.


NMT:

POTS+

roamingpossibilities

GSM(DECT):

POTS+

Limited dataservices,

pan-Europeanroaming

UMTSIMT-2000

widebandservices,

globalroaming

and personalcommunication

1st generation 2nd generation 3rd generation

Figure 1 The different generations of mobile communications systems

Services

Environments

Terminalequipment

Systemrequirements

Figure 2 Impact on system requirements by services, environments and terminal equipment

2.3 Flexibility in Services and Applications

The flexibility in the design of services inthird generation systems does not onlyfacilitate the personification of services.It also enables service providers to distin-guish their services and it eases the ser-vice creation by minimising the re-engi-neering required to provide new services.In order to provide this flexibility, theservice components (service capabilities)are standardised rather than the servicesthemselves, an approach already knownfrom Intelligent Networks (INs) [3].

Service capabilities consist of bearersdefined by QoS (Quality of Service)parameters and mechanisms to enableservice providers and operators to createtheir own supplementary services, tele-services and end-user applications. Thus,to give a list of services for third genera-tion would just be based on speculation.However, some services may be fore-seen, such as telephony, different datatransmission, video and multimediaapplications.

Third generation systems will be de-signed to be flexible not only for serviceprovision, but also for deployment indifferent environments, and easily ex-pandable. This is dealt with in the chap-ters considering radio and network re-quirements.

The versatility of third generation sys-tems makes them suitable for manyapplications, such as office applicationsto access services normally provided byPBXs and LANs, for cordless applica-tions in the domestic domain, and forcellular operation for access to publictelecommunications. Further, the thirdgeneration systems are candidates forradio in the local loop, which is a viablesolution for rapid deployment of fixedtelecommunications, particularly applic-able in developing countries. The marketis believed to be at least as large as themarket for mobile communications.

2.4 Requirements for the Radio Interface

The requirements to the radio interfacefor third generation systems will ob-viously be quite different from earliergeneration systems. In NMT the offeredservices were speech and low rate data.Even in GSM and other digital secondgeneration systems the services are

limited to speech and low rate data,though there has been some enhancementwithin GSM to offer rates in the order ofmagnitudes of tenths of kbit/s, andrecently even packet switched services.

For third generation systems the require-ments to services will be very different.These systems are required to offer awide range of services from traditionalspeech to packet or circuit switched datahaving transfer rates in the order of a fewmillion bits per second. However, all ser-vices will not necessarily be offered in allkinds of environments. As discussed pre-viously, there is an interdependence be-tween terminal capabilities, environ-ments and the services provided.

The environments may be classified intothe outdoor vehicular macro cell environ-ment, the outdoor vehicular suburbanenvironment, the urban environment withlow mobility pedestrian users, and theindoor very low mobility environment[4,5]. It is obvious that the requirementsto the radio interface will be quite differ-ent in these different environments. Atone end we have the vehicular macro cellenvironment where the terminals willhave high velocity and rather low den-sity. In this environment deployment willbe focused on coverage, ie. covering alarge area, and giving some minimumservice to all users within the area ofcoverage. At the other end we have theindoor pico cell environment where theterminals will have very low or no ve-locity, but some mobility. In this en-vironment deployment will be focusedon capacity, and giving high data rate ser-vices to the users in the area of coverage.

In the remainder of this section require-ments for UTRA (UMTS TerrestrialRadio Access) are considered [6], thoughsimilar requirements within ITU-R arereferenced [5]. Thus the term third gener-ation system in most cases means UMTSin this context.

The requirements on the radio interfaceare focused on

• Bearer capabilities

• Operational capabilities

• Spectrum usage

• Complexity and cost.

2.4.1 Bearer Capabilities

The radio interface should support arange of maximum user bit rates that

depends upon the environment the useris currently located in as follows:

• Rural Outdoor (Vehicular Macro cell):at least 144 kbit/s (goal to achieve384 kbit/s) at speed of 120 km/h (max-imum speed 500 km/h)

• Suburban Outdoor (Vehicular sub-urban): at least 384 kbit/s (goal toachieve 512 kbit/s) at speed of120 km/h

• Indoor / Short range outdoor (Micro /Pico cell): at least 2 Mbit/s with mo-bility.

It is desirable that the radio interfaceshould allow evolution to higher bit rates.

The radio interface should be flexible inallowing a range of user bit rates andtypes of services in different user en-vironments:

• Parallel bearer services (service mix),real-time / non-real-time

• Circuit and packet switched bearers

• Scheduling of bearers according topriority

• Adoption of link to quality, traffic andnetwork load, and radio conditions

• Wide range of bit rates with suffi-ciently low granularity

• Variable bit rate real time capabilities.

The radio interface should support usermobility within and between radio ope-rating environments:

• Provide handover between cells of oneoperator

• Provide handover between differentdeployment environments

• Allow handover between differentoperators or access networks

• Allow efficient handover to secondgeneration systems, eg. GSM.

2.4.2 Operational Capabilities

The radio interface should provide com-patibility with services that are providedby present core networks:

• ATM bearer services• GSM services• Internet Protocol based services• ISDN services.

On a global basis numerous core net-works exist. The modes of operation andthe range of services and bearers that


third generation is foreseen to provideare very different. Therefore, the radiointerface of these systems must exhibitgreat flexibility in order to be able tomeet requirements given by these differ-ent aspects. This implies that the radiointerface should be capable of meetingrequirements given by various radio ope-rating environments, operation acrossmultiple operators, multiple types of ope-rators, operation across different regula-tory regimes, a variety of terminal types,and a variety of services and bit rates.

It is seen as a mandatory requirement thatthe operator should be excepted fromexhausting frequency planning after re-configuration of his network due to ser-vice or traffic requirements. Whethersuch a requirement is feasible withoutlosing in other ways, however, is yet tobe investigated. Therefore, the radiointerface should allow for flexible use ofthe allocated frequency bands with mini-mum required spectrum planning afterre-configuration of the network due tochanging traffic or service requirements.

It is envisaged that public, private andresidential operators will operate systemswithin the third generation concept. Pub-lic operators will operate systems cover-ing a wide range of coverage area andservices with minimum guaranteed gradeof service in wide coverage areas andwith very high capacity and user datarates in environments with limited cover-age. Private operators will operate in lim-ited areas and give service to a limitedgroup of users, possibly with some limit-ation on services and mobility. Residen-tial operators will operate within verysmall cells (pico cells), probably with avery limited range of service capabilities.The radio interface should allow forthese three types of operators, but withpredetermined and guaranteed levels onquality of service for the public operatorwhen in the presence of other operators.

2.4.3 Spectrum Usage

The radio interface should have highspectrum efficiency for typical mixturesof different bearer services, and for lowrate speech services the efficiency shouldbe equally good as that for GSM. Ofcourse it is difficult, or more likely im-possible, to tell what the typical mixturesof bearer services will be. However, it isrequested that the chosen technology islikely to have high spectrum efficiency inorder to exploit the given frequency bandin the most efficient way.

It is likely that some of the provided ser-vices, for instance Internet connectionwill require higher capabilities at thedownlink compared to the uplink. Thisimplies that the radio interface shouldprovide for asymmetric use of the radioresources on up- and downlink.

Within the bands that are allocated toUMTS multiple operators should beallowed without co-ordination. It isassumed that different operators willbe given different frequency bands todeploy their services within. Given suchan assumption, any specific operator isfree to deploy his third generation net-work without mandatory co-ordinationwith operators using an adjacent band.

Some of the frequency bands that arecurrently used by first and second gener-ation systems will undoubtedly be suit-able for services which first will be pro-vided by third generation. To allow forsuch services to be deployed even furtherin an efficient manner it is preferablethat the third generation technology canmigrate into these frequency bands in aneasy and non-prohibited manner. There-fore, it is necessary that the interface iscapable of this migration.

2.4.4 Complexity and Cost

It is likely that third generation terminalswill come in several types. This willallow for a wide range of features tobe included within the different types,giving different terminal classes. It isrequired that terminals are availablewithin the range of classes, both simpleand inexpensive ones for low end usersand more complex and thereby expensiveterminals for high end users.

For the network it is desirable to haveone radio interface which can adapt todifferent modes of operation in order tomeet the predefined needs of the radiooperating environment and service en-vironment.

The overall cost of third generation isinfluenced by the technology chosen andthe desired quality and reliability. Themain elements that contribute to theoverall cost of UMTS are:

• Development cost

• Equipment cost

• Deployment cost

• Operational cost.

2.4.5 Other Requirements

It is desirable that the radio interfacetechnology that is chosen within ETSImeets at least the technical requirementsas a candidate technology for IMT-2000within ITU.

It should be possible to deploy and oper-ate a third generation network within alimited bandwidth in order to meet re-gional regulatory requirements.

A third generation system should becapable of co-existing with other systemswithin the same or neighbouring fre-quency bands depending on systemsand regulations.

2.5 Network Functions and Requirements

This chapter deals with the requirementsimposed on third generation systemsregarding the functionality and designof the networks.

2.5.1 Network Modularity

One of the basic requirements on thirdgeneration network architecture is themodular system concept where the accessnetworks are clearly separated from corenetworks at specified interfaces. Further,the standards will open for a many-to-many relation between access and corenetworks.

This concept is supported by the need fordifferent access networks due to differentenvironments (eg. satellite, wide-areacellular, cordless, fixed) and differentcore networks. The concept also enablescompetition between different standardsand implementations of access networksand core networks.

Within ITU, these principles are referredto as the IMT-2000 Family of SystemsConcept (IFS). Within ETSI, similarideas have been developed, based on thework on Global Multimedia Mobility(GMM) [7]. The GMM is a frameworkthat allows different access networks tointeract with different core networksbased on a common architectural con-cept. This is illustrated in Figure 3 (takenfrom [7]).

This modular approach invokes the needfor standardised interfaces between dif-ferent system components.


2.5.2 Network Functionality

The required network functions are de-duced from considerations of operationalrequirements. The network functions aregrouped into functional entities, andfunctional architectures are created byidentifying the functional entities andtheir interactions. Functional architec-tures are under continuous developmentboth for IMT-2000 and UMTS [8,9].These architectures include functionalentities for (the list is not exhaustive):

• Radio resource management

• Mobility management

• Call-, connection- and bearer control

• Interworking

• Access- and service control

• Security control

• Network management.

Functions for radio resource managementare necessary for allocation and controlof radio communication resources. Thisincludes allocation of radio resources,set-up and release of radio channels,power control, modulation, coding andcompression. In third generation systems,the radio resources must be shared be-tween circuit switched and packetswitched communication. Additionalfunctions for packet data transfer capa-bility are required, including access con-trol, multiplexing, error and flow control.

Functions for mobility managementinclude location registration, paging,functions for supplying routing informa-tion, and functions for handover. Regard-ing the handover functions, it is neces-sary to have functions for radio channelquality estimation, handover decision andexecution. Macro diversity may be pro-vided in the access network, and special

functions are required in order to controlthe recombination of information duringa handover.

Call-, connection- and bearer controlfunctions are required to perform the set-up and release of calls. For connectionoriented services, the functionality iswell developed and described. For con-nectionless services, the notion of a callis absent, and work is progressing inorder to specify necessary functionalityfor end-to-end control.

The call-, connection- and bearer controlfunctions may act in a hierarchical man-ner, supervised by the call control, whichperforms end-to-end QoS negotiationsand set-up or release of the requirednumber of connections associated withthe call. Connections are end-to-endassociations of bearers and the separationbetween call and connection gives extra


MobileTE

AccessNetworks

Examples:• GSM

BSS• DECT• S-PCN• ISDN• B-ISDN• UMTS• LAN• WAN• CATV• MBS

UIMMobile

TE

FixedTE

UIMFixed

TE

Core TransportNetworks(including

intelligence)

Examples:• GSM NSS+IN• ISDN/IN based• B-ISDN

+IN based• B-ISDN

+TINA based• TCP/IP based

ApplicationServicesDomain

ApplicationServicesDomain

Terminal Equipment Domain Access Network Domain Core Transport NetworksDomain (including intelligence)

Figure 3 The GMM architectural framework [7]

freedom in allocation of the functionalityto different network elements, whichmay be useful eg. for multimedia calls.The connection control performs the set-up and release of the connections, basedon performance requirement such asBER and time delay. Finally, the bearercontrol functions allocate the requirednumber of bearers to each connection,and triggers the establishment of theactual physical channels. Bearer controlfunctions are required both for the airinterface and the interface between theaccess- and core networks, and will bedifferent according to the characteristicsof the physical channels they control.

The interworking functions are necessaryin order to connect different access net-works to different core networks. Func-tions will be specified for interworkingwith ISDN, B-ISDN, X.25 PDN (publicdata network) and IP data traffic. Inter-working with existing second generationmobile systems is also important.

Access- and service control is concernedwith the control and access to services,and handles the service related process-ing. In an IN approach, as eg. taken forIMT-2000 [8,10], the service controlfunctionality is triggered by the call con-trol, and interacts with functions for ser-vice access and switching, as well asspecialised resource functions. Serviceaccess requires the provision of terminalcapability information, user- and sub-scription information and triggering ofauthentication. The specialised resourcefunctions are responsible for variouskinds of user interaction, eg. receivingdigits and playing announcements, bothfor IN-provided services, multimedia ser-vices and packet data transfer.

Security control is becoming increasinglyimportant in communications. Mobileand personal telecommunications ser-vices impose additional security require-ments, since mobile communication sys-tems inherently involves wireless access,and third generation systems must allowusers to roam across network and nation-al boundaries. The security control in-cludes functions for [11,12]:

• Authentication of user, mobile termi-nal and service provider

• Privacy, anonymity and confidentialityof user location, user identity and userdata

• Authorisation and access control tolimit the access to user and serviceprofile data

• Denial of access to particular services.

Functions for network management arenecessary to ease and support tasks suchas planning, installation, provisioning,operation, maintenance, administration,and customer service. Efficient networkmanagement functions are important forthe ease of operation and thus the cost-effectiveness of the system. IMT-2000bases the network management on theconcept of a Telecommunications Man-agement Network (TMN), which is beingstudied by ITU-T.

2.5.3 Flexibility and Evolution

For the mapping of functional entities tonetwork elements, it is essential that ahigh degree of flexibility be supported.In other words, the mapping should givefreedom to create a number of differentreference configurations, so that the sys-tem may be optimised regarding per-formance and cost to different applica-tions and environments. Both stand aloneconfigurations with gateway to the fixednetworks and integrated solutions withshared functionality should be possible.

The modular structure and the freedom toadd extra functionality also support an-other aspect requiring flexibility, namelythat the system should be easily expandableto allow growth in size and complexity.

3 Research and Stan-dardisation Activities

This chapter considers the research andstandardisation activities on third genera-tion mobile communications systems.Research on third generation technologyis performed at universities, withinindustry, and by telecommunicationsoperators. This chapter considers someresearch activities performed in Euro-pean organisations, and gives by nomeans a complete or representativepicture of the research carried out ina global scale.

Standardisation of the next generationcellular system is done in all regions ofthe world, the Far East, America andEurope, by amongst others, the Inter-national Telecommunications Union(ITU) and the European Telecommunica-tions Standardisation Institute (ETSI).

3.1 Research within ACTS

ACTS (Advanced CommunicationsTechnology and Services) is one of thespecific programmes of the ‘FourthFramework Programme of EuropeanCommunity activities in the field of re-search and technological developmentand demonstration (1994 to 1998)’.ACTS builds on the work of the earlierRACE programmes (Research and De-velopment in Advanced CommunicationsTechnologies for Europe). Within ACTSa large number of projects study issuesrelated to third generation mobile com-munication. One of the most importantgoals is to develop and demonstrate theviability of technology and solutions.Many results will be forwarded to stan-dardisation bodies.

Some of the issues studied in ACTS thatare applicable for third generation mobilecommunication are:

• Wireless ATMIn particular low range high bit ratesystems, typically 25 – 155 Mbit/s,normally with limited mobility.

• Satellite component of UMTSAddressing performance parameters,economic and technical feasibility.

• Migration paths towards third generationNetwork architectures, interworking,mobility management, service plat-form technology.

• Radio transmission technologyAddressing adaptive antennas, multi-mode operation and software radio,amongst others

• Mobile terminals and applications.

3.2 Research within COST

COST (European Co-operation in thefield of Science and Technical Research)is a framework for R&D co-operation inEurope, and involves participants from25 European countries.

COST has research activities in manydifferent domains, the largest of thesebeing the telecommunications domain.Within this domain, around five ongoingprojects address issues within mobilecommunications. Topics addressedinclude:

• Antenna technology• Mobile satellite communications • Radio technology • Radio wave propagation • Network aspects.


3.3 Research withinEURESCOM

EURESCOM (European Institute forResearch and Strategic Studies withinTelecommunications) has several pro-jects on third generation mobile commu-nications issues. This includes projectson:

• Mobility management for cordlessterminals realised by IN

• Support of terminal mobility based onTINA

• Radio technology for broadbandaccess

• Wireless ATM including mobility man-agement.

3.4 Standardisation within ITU

As mentioned previously, ITU is respon-sible for the work on IMT-2000. Thestandardisation of IMT-2000 (at that timecalled FPLMTS) was initiated in 1985.

3.4.1 The IMT Vision and IMT-2000 System

The standardisation of future mobilecommunications within ITU is based onthe International Mobile Telecommuni-cations (IMT) vision, which aims to pro-vide seamless operation of mobile termi-nals throughout the world, where Any-where – Anytime communicationsrequire coverage by both terrestrial andsatellite networks. The satellite compo-nent, denoted Global Mobile PersonalCommunications by Satellite (GMPCS),comprises a wide range of global satellitesystems providing fixed and mobile ser-vices in different frequency bands.

The current work concentrates primarilyon IMT capabilities that will be availablearound the year 2000, with the target sys-tem designated IMT-2000. IMT-2000represents the satellite and terrestrialcomponent of IMT that will be availablearound the year 2000 and operate in the 2GHz frequency band. Standardisation ofIMT-2000 is one of the strategic priori-ties of the ITU.

One of the important early results wasthe contribution to the allocation of fre-quencies in the 2 GHz band, whichwas resolved by WARC 92.

3.4.2 Organisation

The work on IMT-2000 within ITU issplit between ITU-R and ITU-T [10]. Aspecial group (Intersector Co-ordinationGroup – ICG) takes care of the co-ordi-nation of the work. ITU-R Task Group8/1 does all the radio related work onIMT-2000/FPLMTS. Within ITU-T, thework is split in several study groups.Study Group 11, responsible for the sig-nalling and protocols, is the lead StudyGroup.

3.4.3 Time Schedule

A phased approach is adopted for thedefinition of IMT-2000. According to thetime schedule the specifications forPhase 1 will be ready by the end of thiscentury [13]. The finalisation of the spec-ifications will be preceded by a period ofproposals for standardisation, evaluationand consensus building, as illustrated inFigure 4.

3.4.4 Services and Applications

It is the intention that IMT-2000 shallmeet the service requirements as givenin chapter 2 based on bit rates up to2 Mbit/s. A list of capabilities is definedin [14] for the initial phase (called capa-bility set 1), and a short summary isgiven in Table 1. Phase 2 will include bitrates higher than 2 Mbit/s, and provideservices such as support of high data rateneeds of portable computing users andsupport of enhanced multimedia commu-nications requirements.

Services will be available virtually every-where, including land, maritime, andaeronautical situations. The user of thepersonal terminal will be able to access atleast a minimum set of services in anysituation. This set of services includesvoice telephony, a selection of data ser-vices, access to UPT, and an indicationof other services available [15].

The applications in developing countriesare considered important in the work onIMT-2000. The wireless local loop appli-


1997 1998 1999 2000

PROPOSALS

EVALUATION

CONSENSUS BUILDING

SPECIFICATIONS

Time schedulefor ITU-RIMT-2000 standards

development activities

Figure 4 Time schedule for IMT-2000 standards development, Phase 1 (from ITU homepage, May 98, ref: http:// www.itu.int/imt/2-radio-dev/time/time.html)

cation is important in this context, andthe market is likely to be at least as big asthe mobile market. Likewise, the satellitecomponent of IMT-2000, together withother global satellite systems, will likelyprovide the first telephone in many ruralvillages. The terrestrial infrastructure willthen follow as demand increases. Thus,integration of future terrestrial and satel-lite systems is highly desirable to facili-tate seamless evolution.

3.4.5 System DesignConsiderations

IMT-2000 will meet the requirementsdiscussed in chapter 2. It is stressed thatit is important to have a high degree ofcommonality in the IMT-2000 family ofsystems. Commonality is important toreduce development costs, obtain cheaper

equipment through economy of scale andlimit the need for interworking functions.Commonality of radio interfaces isstressed in particular, to simplify the taskof building multimode mobile terminalscovering more than one operating en-vironment. Within ITU, and especiallyfrom Japan and Europe, it is seen asmandatory that the radio interface shouldbe able to operate in two different modes,FDD (frequency division duplex) andTDD (time division duplex), or rather,there will be different types of terminalsfor each of these modes. Depending onthe specific technological solution that ischosen the deployment of the systemswith these two modes will differ some-what. At this point it is not possible totell how the two modes should worktogether and be compatible with eachother.

On a global basis there are some differ-ences on how spectrum is used in differ-ent countries. This will obviously haveimpact on the specific implementation ofany system and technology. In additionto this there are quite different existingtechnological platforms and regulatoryregimes, and it is obvious that these dif-ferences will have an impact on thechoice of system technology, systemdesign and deployment. Because specificsolutions have not yet been chosen, it israther difficult to tell what the considera-tions and impact on design will be.

The IMT-2000 functional architecture isbased on the IN concept, and the evolu-tion of its functional capabilities is there-fore closely linked to the evolution of IN.It must therefore be anticipated that IMT-2000 will take on board later versions ofIN, switching and signalling standards,where radio and mobility management isa natural part of the protocols. A descrip-tion of the functional model is given inAppendix 1.

3.5 Standardisation in ETSI

The ETSI General Assembly resolveda strategy for UMTS standardisation in1996, stating that the work should pro-ceed based on GMM recommendations.The work responsibilities are allocated tothe Technical Committees SMG (SpecialMobile Group), NA (Network Aspects)and SES (Satellite Earth Stations andSystems).

3.5.1 Organisation andResponsibilities

SMG is responsible for the UMTSGeneric Radio Access Network, and forthe GSM Core Network Evolution. NA6UMTS is responsible for the ISDN CoreNetwork Evolution. The UMTS satellitecomponent is developed in a co-opera-tion between SMG and SES, under theleadership of SMG.

Within SMG, the following Sub-Techni-cal Committees are involved:

• SMG1 – responsible for specificationof services

• SMG2 (UMTS Working Party) –responsible for the physical layer ofthe radio interface and the study of allradio engineering aspects

• SMG4 – responsible for UMTS dataservice requirements


Services

• At least 144 kbit/s in vehicular environment, at least 384 kbit/s for pedestrianenvironments and at least 2 Mbit/s for office environments

• High quality speech

• Support of advanced and existing addressing mechanisms

• Virtual home environment for service portability between IMT-2000 networks

• Intra-Family handover, seamless handover between cells of one IMT-2000 network

• Connection oriented and connectionless services

• Roaming among IMT-2000 family members

Terminals

• Range of terminals with different capabilities, from voice to multimedia, multibandand multimode terminals

Access networks

• New BSSs in the UMTS/IMT-2000 spectrum

• Support of packet services

Core networks

• Possible evolution from existing core networks

• Support of packet switched and circuit switched operation, based on BroadbandNetwork Architecture (PDH/SDH/ATM)

• Interworking with ISDN, B-ISDN, X.25 PDN and IP data traffic

Table 1 Key features in IMT-2000 Phase 1

• SMG 6 (WPA UMTS) – responsiblefor the network management functionsand interfaces

• SMG 9 (UMTS Working Party) – responsible for the specificationof the UMTS SIM-card / MobileEquipment interface

• SMG10 (Working Party C) – respon-sible for UMTS security

• SMG11 – responsible for Speech Cod-ing

• SMG12 – responsible for the specifica-tion of the network architecture, func-tions and signalling.

3.5.2 Time Schedule

The time schedule for UMTS develop-ment is subdivided into the followingphases [4]:

• UMTS Phase 0 is the implementationof UMTS services in evolved GSMnetworks (using the evolved GSMradio access).

• UMTS Phase 1 will use the UMTS radioaccess. Operation is planned for 2002.

• After year 2002 UMTS will evolve inannual phases.

To reach the goals in Phase 1, the moredetailed milestones have been identified.The UMTS concept is by now relativelyclear, and the specifications for Phase 1should be finalised before the year 2000.After a period of trials and industrialdevelopment, it is hoped that UMTS canbe launched in 2002. This is illustrated inFigure 5.

3.5.3 Content of UMTS Phase 1

The overall content of UMTS has not yetbeen decided because the future develop-ment of UMTS will be dependent on theoverall technology development and can-not be forecast in detail.

The content of the first phase of UMTS,however, has been agreed upon [16] andis described in Table 2 in terms of ser-vices, terminals, access networks andcore transport.

3.5.4 Services and Applications

UMTS, like IMT-2000, will meet the ser-vice requirements as given in chapter 2.Although the services are similar to thoseof IMT-2000, there are some natural dif-ferences in the importance placed on

different applications. In particular, it isseen that UMTS places less importanceon wireless local loop applications and

the satellite access component, which areparticularly interesting applications indeveloping countries.


UMTS Phase 1 standards

Pre-operational trials

UMTS Phase 1 commercial operation

Figure 5 Time schedule for UMTS development, Phase 1

1998 1999 2000 2001 2002

Services

• Up to 144 kbit/s with high mobility, 144–512 kbit/s for wide area mobility, 2 Mbit/s for restricted mobility

• High quality speech using low bit rates

• Advanced addressing mechanisms

• Virtual home environment service portability

• Seamless indoor, outdoor and far outdoor

• Dual mode / band of operation of GSM/UMTS in one network

• Roaming between GSM and UMTS networks

Terminals

• Dual mode/band GSM/UMTS

• Multimedia terminals

• Adaptive terminals

Access networks

• New BSS in the UMTS/IMT-2000 spectrum

• Flexible, adaptive radio interface

Core transport

• Evolution from GSM and ISDN

• Support of variable bit rates and mixed traffic types

• Support of packet data by Internet protocols

• Mobile/fixed convergence elements

Table 2 Key features of UMTS Phase 1

3.5.5 System Design Considerations

UMTS will meet the system designrequirements given in chapter 2. Theimportance of modularity, flexibility andcommonality is equally important as forIMT-2000.

Even within Europe it is regardedmandatory to have both TDD and FDDmode terminals. The specific solutions ata technical level are discussed at thismoment. However, within Europe andcountries that choose the UMTS system,the basis from second generation is prob-ably not very different. These countriesprobably have GSM-900 and/or GSM-1800, and some of them even DECT. Inaddition, they have some Wireless LocalLoop (WLL) systems based on DECT orother technology. System design willtherefore very much depend on the spe-cific need of different services in differ-ent deployment scenarios. In additionand very much in connection with this,the regulatory framework within theregion (countries) will put some frame-work on operation of second and thirdgeneration systems. This framework willeven create new requirements and initiatenew development within second genera-tion systems.

The original idea of UMTS was a newnetwork that integrated all existing andnew services in one universal network.This view has changed, and the currentidea is to support services in differentnetworks based on interworking, in par-


GSMBSS

N-ISDN

UTRAN

HL R

MSC G-MSC

SGSN GGSNIP-

networks

X.25IWUGbIWUA

ISUPISUP

IPIP

MAP

X,25

Gb

Iu

A

UMTSCore

network

Figure 6 Evolution of UMTS. Blue part refers to UMTS Phase 1, red part to later phases

ticular interworking with GSM. Thisclose link to GSM is one of the diffe-rences between UMTS and IMT-2000.It is the intention that UMTS will evolvefrom the GSM as illustrated in Figure 6.

UMTS Phase 1 will consist of a UMTSTerrestrial Radio Access Network(UTRAN). The UMTS Core Networkwill be introduced in later phases. Thusin Phase 1, UTRAN will use the GSMtechnology as the core network. As seen,this includes both the circuit and packetswitched infrastructure. In order to inter-act with the GSM infrastructure, inter-working units are required at theUTRAN interface.

The functional architecture proposed forUMTS is not so closely linked to the INarchitecture as is seen for IMT-2000.One particular case is the fact that UMTSintends to employ the protocol developedfor GSM for interrogation with the HLR,ie. the Mobile Application Part – MAP.The corresponding protocol in IMT-2000is defined to be the IN Application Part –INAP. This difference is more politicallythan technically motivated, and more-over, it is not crucial, since the conver-sion between the protocols is easily per-formed.

Whereas ATM is specified as an impor-tant transport mechanism in IMT-2000,there is not yet full consensus on thisissue in UMTS. It is anticipated, how-ever, that ATM will become one ofseveral optional transport mechanisms,

as for IMT-2000. Originally, the ideawas that UMTS should be a mobileextension of B-ISDN services. However,the fast growth of the Internet has madethis an equally important partner net-work.

Among the system design considerationsit is important to mention that an agree-ment on the UMTS Terrestrial RadioAccess interface is reached. This agree-ment is a compromise between the con-cepts W-CDMA and CD-TDMA, bothtechniques being supported and allocatedto different parts of the frequency band.

Appendix 1 – IMT-2000functional architecture

A suitable functional model based on theservice and feature requirements forIMT-2000 is essential for developing thenecessary signalling protocols and forspecifying the appropriate physical inter-faces. The functional model specified forIMT-2000 is implementation-independ-ent and captures the basic functional enti-ties (FEs), and specifies relationships be-tween the FEs. The physical interfacesto be specified are identified by usingsuitable mapping of various FEs in thefunctional model to physical entities (eg.base station (BS), mobile switchingcentre (MSC), home location register(HLR), visited location register (VLR),and service control point (SCP). FigureA1 shows an example of a mapping intophysical entities based on the FEs de-scribed in the following.

The functional architecture for IMT-2000as described in [10] is based on the mod-elling principles used for developing thesignalling protocols for IN (IntelligentNetworks) and is shown in Figure A2.

The management and control of radioresources will be carried out by proce-dures that operate in parallel with andrather independent of the overall commu-nication control procedures. Two planeshave therefore been defined:

• Radio Resource Control (RRC) plane

• Communication Control (CC) plane.

The basic management of radio resourcesresides in the RRC plane, while the over-all bearer and connection control residesin the CC plane. Special plane interactionprocesses perform the synchronisationbetween the two planes.


SDF

Transport and signaling

Signaling only

Fixed network (including wireline users)

SCF

SACF

SCP

SDF

SDP

SCF

SACF

MSC

CCF´

MSC

CCF´

BS

CCF´

SSF

TACF

BCFr

RRC

RFTR

BS

TACF

BCFr

RRC

RFTR

SSF SSF

BS: Base stationMSC: Mobile services switching centerSCP: Service control pointSDP: Service data point

Figure A1 A standalone IMT-2000network and allocation of functional

entities to physical nodes. The functionalentities are described in Table A1

BCF

SRF

SCF

SDF

LRDF

BCAF

TACAF

CCAP

MCFUIMF

SACF

ACF

ADF

CCF

TACF

LRCF

MRTR RFTR

RRCMRRC

BCFr

Planeinteraction

Planeinteraction

Radio resource control plane

Communication control plane

Signaling relationship

Bearer control relationship

SSF

Figure A2 The IMT-2000 functional architecture


RRC plane functions

These functions are in charge of assigning and supervising radioresources as required in the radio access subsystems.

• RRC – Radio Resource ControlHandles overall control of the radio resources, such as selection andreservation of radio resources, handover decision, RF power control,and system information broadcasting.

• MRRC – Mobile Radio Resource ControlHandles the mobile side of RRC.

• RFTR – Radio Frequency Transmission and ReceptionHandles the radio transmission and reception of control and userinformation at the fixed side of the radio interface, including radiochannel and error protection coding and decoding.

• MRTR – Mobile Radio Transmission and ReceptionHandles the radio transmission and reception of control and userinformation at the mobile side of the radio interface. This includesradio channel and error protection coding and decoding.

CC plane functions

These functions are in charge of overall access, service, call, bearerand connection control.

• SDF – Service Data FunctionFE that handles storage and access to service- and network-relateddata, and provides consistency checks on data (eg. service profileand mobile multimedia attributes)

• SCF – Service Control FunctionFE that contains the overall service control logic and handles servicerelated processing activity

• LRDF – Location Registration Data FunctionHandles storage and access to subscriber identity and location data,and provides consistency checks on data (eg. location information,identity data, and active/inactive status)

• LRCF – Location Registration Control FunctionContains the mobility control logic for location registration. Thefunctionalities include location management, identity management,user verification, paging control, and providing routing data.

• ADF – Authentication Data FunctionHandles storage and access to authentication data (eg. security-related parameters) and provides consistency checks on data.

• ACF – Authentication Control FunctionContains the mobility control logic for authentication. The function-alities include user authentication, authentication processing, andconfidentiality control (eg. ciphering management)

• SSF – Service Switching FunctionFE associated with the CCF’. It provides the set of functions requiredfor the interaction between CCF and SCF.

• CCF’ – Call Control Function (enhanced)FE that provides call/connection processing control (eg. establishes,maintains and releases call instances); provides trigger mechanismsto access IN functionalities; establishes, maintains, and releasesbearer connections in the network, and so on.

• SRF – Specialised Resource FunctionFE that provides the specialised resources required for the execu-tion of IN-provided services (eg. user signalling receivers, announce-ments, conference bridges), mobile multimedia services and packetdata transfer services.

• SACF – Service Access Control FunctionProvides non-call- and non-bearer-associated processing and con-trol (eg. in relation to mobility management).

• TACF – Terminal Access Control FunctionProvides the overall control of the association and connection(s)between a mobile terminal and the network. Examples of function-alities include terminal paging execution, paging response detectionand handling, handover decision and handover execution.

• BCF – Bearer Control FunctionControls the interconnection of bearers.

• BCFr – Bearer Control Function (radio bearer associated)Controls the interconnection and adaptation of a radio bearer to thecorresponding fixed line bearer.

• MCF – Mobile Control FunctionProvides the overall service access control logic and processing atthe mobile side of the radio interface. Specifically, it interacts withthe network for non-call/non-bearer-associated services (eg. mobilitymanagement).

• CCAF’ – Call Control Agent Function (enhanced)FE that provides service access for users. It is the interface betweenuser and network CCFs.

• TACAF – Terminal Access Control Agent FunctionProvides access for the mobile terminal. Examples of functionalitiesinclude terminal paging detection and paging response initiation.

• BCAF – Bearer Control Agent FunctionControls the interconnection and adaptation of a radio bearer to therest of the mobile terminal.

• UIMF – User Identification Management FunctionProvides the means to identify both the user and the mobile terminal(or termination) to the network and/or to the service provider. It in-cludes functionalities to store user-related information such as useridentity and security- and privacy-related information, provide foruser authentication, ciphering key generation, and so on.

Table A1

References

1 Grimstveit, L, Hans Myhre, H. Thehistory of mobile communications inNorway. Telektronikk, 91, (4), 1995,15–20.

2 WARC’92. The world administrativeradio conference for dealing withfrequency allocations in certain partsof the spectrum. Malaga-Torremoli-nos, 1992.

3 TS 22.05 ver. 3.0.0 1998-03. Univer-sal Mobile Telecommunications Sys-tem (UMTS). Services and ServiceCapabilities, UMTS 22.05 ver. 3.0.0.

4 Universal Mobile Telecommunica-tions System (UMTS). Selection pro-cedures for the choice of radio trans-mission technologies of the UMTS.(UMTS 30.03).

6 UMTS Baseline document. Positionson UMTS agreed by SMG. (UMTS30.01).

5 ITU-R. Guidelines for evaluation ofradio transmission : Technologies forIMT-2000. (Recommendation ITU-RM.1225.)

7 ETSI programme Advisory Commit-tee (PAC). Global MultimediaMobility (GMM) : a standardisationframework. ETSI/PAC16(96)16,1996.

8 ITU-T SG11. Draft new recommen-dation Q.FNA, network functionalmodel for FPLMTS, Version 6.1.0.Geneva, January 1997.

9 ETSI. Special Mobile Group :UMTS. Network Principles. SophiaAntipolis, 1997. (Draft TechnicalSpecification, UMTS 23.05 v.0.10.0.)

10 Pandya, R et al. IMT-2000 Standards: Network Aspects. IEEE PersonalCommunications, August 1997.

11 ITU-R M.1078.

12 TS 33.20 ver. 3.0.1, 1997-07, Uni-versal Mobile TelecommunicationsSystem (UMTS). Security Principlesfor UMTS. UMTS 33.20 ver. 3.0.1.

13 ITU home page, time schedule http://www.itu.int/imt/2-radio-dev/time/time.html

14 ITU-T Study Group 2. Liason State-ment to Task Group 8/1 on IMT-2000Family Concept for Members of theIMT-2000 Family Phase 1. Docu-ment 8-1/6-E, 27 November 1997.

15 ITU-R. Framework for services sup-ported on International MobileTelecommunications-1000 (IMT-2000). Rec. M.816-1.

16 UMTS Baseline document, UMTS30.01 V3.1.0 (1997-10). Positions onUMTS agreed by SMG.

17 Spectrum for IMT-2000, UMTSForum Report, October 1997.


Stein Svaet is Research Scientist at Telenor R&D,Kjeller, where he has been employed since 1988.He is working in the field of mobile and personalcommunications, with a special interest in networkarchitectures for UMTS and mobile broadbandsystems.


Bjørn Harald Pedersen is Research Scientist atTelenor R&D, Kjeller, in the Mobile Communicationgroup. He is currently working with network andservice aspects, functional architecture and re-quirements for third generation mobile communica-tion systems. He has also been participating in thestandardisation of UMTS in ETSI.


Geir Olav Lauritzen is Research Scientist at TelenorR&D, Kjeller. He is working in the field of mobilecommunication with special interests in networkfunctionality and capacity aspects for land mobilesystems.


Jan Eriksen is Research Scientist at Telenor R&D,Kjeller, where he is working with issues related tocellular systems and personal communications.His main involvement is within the physical layerof the radio interface and wave propagation. He iscurrently also involved in the ongoing standardisa-tion work within ETSI SMG2 and ITU-R RG 8/1 on3rd generation systems.



With the forthcoming introduction ofsystems for Global Mobile PersonalCommunications by Satellite (GMPCS),mobile satellite services will finallybecome truly personal. Already avail-able is Mobiq, a notebook PC sizedterminal based on the Inmarsat mini-M standard, heating up the market-place for Iridium, Globalstar and ICOannouncing introduction of theirhandheld services from late 1998onwards.

Iridium will undoubtedly be first tomarket, emphasized by fifteen success-ful launches in twelve months complet-ing the venture’s 66 satellite constella-tion. However, the challenge of provid-ing a continuous global service is highdue to the fact that these systems relyon large constellations of medium- orlow earth orbiting satellites.

The GMPCS service concept is almostidentical to GSM, including voice, faxand low speed data services. The sys-tems even integrate with GSM andother cellular standards by the pro-vision of multi-mode terminals.

The global pocket phone represents amilestone in telecommunications and itis expected that future evolution willfocus on increasing the transmissionspeed rather than shrinking the termi-nal size. Along this evolutionary path,satellite platforms represented byTeledesic and Inmarsat Horizons,may very well take the lead in therace where cellular systems so farhave been the front runners.

1 Introduction

For some years several systems havebeen under development for provision ofvoice- and data services to multi-modehandsets world-wide using low ormedium earth orbiting satellites. Thesesystems have been referred to underacronyms such as S-PCS (Satellite Per-sonal Communication Services) and BigLEOs (Low Earth Orbits). In 1996, theInternational Telecommunications Union(ITU) World Telecommunications PolicyForum labelled them GMPCS – GlobalMobile Personal Communications bySatellite.

The adventure began back in 1985, whenMotorola engineer Bary Bertiger con-ceived of the Iridium system, as a resultof seeking a means to enhance cellular1

coverage. Bertiger shared his idea with

colleagues Ken Peterson and RayLeopold, and the trio considered variousoptions – such as high-flying balloons –before settling on LEO satellites. Twoyears later, research and development gotunder way and in 1990 the Iridium sys-tem was announced at simultaneous pressconferences in Beijing, London, Mel-bourne and New York [1].

Shortly after Iridium’s announcements,several other concepts and playersentered the arena, and today Iridium,Globalstar and ICO are seen as the maincontenders. The expectations are highand so are the risk factors, both techno-logically and economically. In addition,there are complex issues of political andregulatory issues to be solved beforethese multi-billion dollar systems canoffer their services as promised from late1998.

In this article we will discuss some of thetechnical aspects and parameters of theGMPCS systems and illustrate theplanned constellations. The marketopportunities will also be briefly touchedupon as will recent developments withinthe cellular and satellite environment,including ongoing regulatory activities.Finally, some thoughts on the expectedfuture evolution of personal satellitecommunications are presented.

2 GMPCS – technicaldescription

The planned GMPCS systems have muchin common. They all target similar mar-kets with similar services and they intendto complement the coverage of cellularsystems. However, their technical char-acteristics such as choice of orbit, mul-tiple access method and frequency band,differ.

In this chapter we will discuss servicesand technical aspects for GMPCS sys-tems.

2.1 Service portfolio

The GMPCS service portfolio is almostidentical to what we are familiar with inGSM and typically includes:

• voice – quality comparable to GSM

• 2.4 – 7.2 kbps data and fax2

• paging

• positioning3

• GSM compatible SMS (Short Message Service)

• ISDN (Integrated Services DigitalNetwork) supplementary services.

The end-user charges are not settled, butthe potential operators have indicatedthat they will be in the $1–3 range. Land-line charges may be added for long dis-tance calls, making the end-user chargeseven higher.

2.2 Space segment

Orbit alternatives

There is no such thing as an optimumorbit. This is clearly illustrated by thefact that none of the system operatorshave landed on the same orbital charac-teristics, resulting in four definitions ofthe perfect orbit for delivery of commu-nications to pocket-sized terminals.

The high altitude (36,000 km) of tra-ditional communication satellites in thegeostationary orbit (GEO) results in timedelays of the signal that could make themdisadvantageous for GMPCS. Addition-ally, powerful transmitters and receivers,difficult to incorporate into a pocket-sized unit, are required to compensate forthe large signal attenuation. Also, usersin higher latitudes can experience serviceproblems due to the low elevation angle,which can result in signals being blockedby buildings, trees and mountains.

To overcome these problems (but pos-sibly introducing others), the GMPCSproposals are based on highly inclinedLEO and MEO (Medium Earth Orbit)orbits made possible by advances insatellite technology, typically operatingaround 1,000 km and 10,000 km abovethe earth surface respectively.

As a consequence of the lower satellitealtitude, the number of satellites needs to

Global Mobile Personal Communications by SatelliteA R I L D F L Y S T V E I T A N D A R V I D B E R T H E A U J O H A N N E S S E N

1 In this article, the term ‘cellular’means terrestrial Public Land MobileNetworks (PLMNs), although GMPCSsystems also apply cell structured cov-erage.

2 Higher data rates are offered for fixedinstallations.

3 The GMPCS positioning services arenot designed to compete with the farmore accurate GPS (Global Position-ing System).


be increased to achieve continuous globalcoverage. Typically, a global LEO con-stellation consists of 40–80 satellites,whereas a MEO constellation requiresapproximately 10–15 satellites to coverthe earth.

Inter-satellite links

The concept of Inter Satellite communi-cation Links (ISLs) makes it possible toswitch traffic between satellites in thesame or adjacent planes. Traffic may becarried over long distances independentlyof ground networks, thereby minimisingthe required number of gateways forglobal coverage.

ISLs are seen as one of the major tech-nological challenges for new satellitesystems, such as GMPCS.

Handovers

While a GEO satellite covers a fixed areaof the earth, LEO and MEO satellitesmove relative to the earth with the cover-age area of a specific satellite varyingover time as its beams move across theearth. Because of this satellite motion,GMPCS systems require a more complexground segment, tracking satellites andperforming handovers to avoid losingcalls as satellites move over the horizon.

Two types of handovers exist, compar-able to intra-MSC and inter-MSC hand-overs in GSM. Intra-satellite handover,which occurs in systems where a satel-lite’s coverage area moves continuouslyas the satellite moves, means that the callis handed over to another beam (cell) onthe same satellite. In systems where thesatellites redirect their antenna to main-tain a fixed coverage area on the groundas the satellites move, intra-satellite han-dovers are rare. Handovers betweensatellites are called inter-satellite hand-overs.

For systems with ISLs handovers aremore complicated, since the whole chainof crosslinks must be handed over to thenext chain of crosslinks simultaneously.Even though interworking with a cellularsystem will be implemented, function-ality for handling of inter-system hand-overs between satellite systems andcellular systems for multi-mode termi-nals will not be present.

The average time between inter-satellitehandovers is approximately 4–8 minutesfor LEO constellations depending on alti-tude, but in special situations handovers

can occur as frequently as every 10 sec-onds. For MEO systems, inter-satellitehandovers occur every hour on average.Of course, intra-satellite handovers occurmore frequently and depend on the num-ber of spot beams provided by the satel-lite [2].

End-to-end communication delay

As the transmission path between userand satellite is much shorter for LEO andMEO systems than for GEO systems, thepropagation delay is reduced dramati-cally. Isolated, this affects voice quality,which will be perceived as better. Fordata communications, the reduced delayhas positive impacts on protocol com-patibility.

Table 1 illustrates typical minimum andmaximum one-way propagation delaysbetween user and satellite for a LEO(Iridium) and a MEO system (ICO)[2][3]. As a reference, the minimum one-way propagation delay for a GEO satel-lite is approximately 120 ms.

• In LEO systems with polar orbits, thehighest concentration of satellites isabove the North and South poles,resulting in excellent coverage of thepolar areas. At the equator, the dis-tance between the orbits is at its maxi-mum. Decreasing the inclinationresults in reduced coverage at high lat-itudes and increased ‘satellite density’at equator.

• Due to the large coverage area of aMEO satellite, decreasing the inclina-tion does not affect the coverage of thepolar area in the same way as it doesfor LEO systems.

Unlike GSM, the GMPCS handheld ter-minals cannot be used indoors except atwindows with line of sight to at least onesatellite. Obstacles like trees, mountainsand buildings may cause blockage, andthe higher the minimum elevation anglefor which a system can guarantee theavailability of at least one satellite at anytime, the better the quality of the system.None of the systems have been designedto provide communication in cities.

Multiple access methods

The two most frequently used methodsfor shared access on the user link in mod-ern cellular systems are CDMA (CodeDivision Multiple Access) and TDMA(Time Division Multiple Access). As anexample, the GSM cellular standard usesTDMA, whereas the IS-95 standard inthe US is based on CDMA. GMPCS sys-tems will be based on either of these twoprinciples.

Frequency bands for MSS

In 1992 WARC (World AdministrativeRadio Conference) allocated frequencybands in L-band (16.5 MHz) and S-band(2 x 30 MHz plus 16.5 MHz) to MSS(Mobile Satellite Systems) systems. Fig-ure 1 illustrates the frequencies to beused on the mobile link for the variousGMPCS operators.

The sharing of the 16.5 MHz wide L-band spectrum to be used by the IridiumTDMA system and the two CDMA sys-tems Globalstar and Odyssey has beenan issue for negotiations and discussionsin various forums over the last years.Currently, the agreement is that Iridiumwill use the band 1610–1615.15 MHz,whereas the CDMA systems will sharethe remaining 11.35 MHz of the band[4]. However, if some of the systemsnever go operational, the band will be

LEO MEO

Minimum [ms] 2.60 34.5

Maximum [ms] 8.22 48.0

Table 1 One-way propagation delayuser-satellite

In addition to propagation delay, the end-to-end communication delay is influ-enced by several other parameters. Sig-nificant contributors are voice codecs,switching/buffering operations andassembling/disassembling processes.[2] indicates that the delay in a typicalTDMA system like Iridium is in therange of 100 ms, excluding propagationdelay. Similar delays are expected also inCDMA based systems.

Coverage – truly global?

The GMPCS operators claim to offerglobal coverage. For provision of ser-vices to the global population, this is vir-tually true. However, regarding globalgeographic coverage, the systems areslightly different. These differences aremainly due to the space segment configu-ration chosen and in particular the incli-nation of the orbital planes:


shared equally among the remainingoperators.

C-band and Ka-band frequencies will beused for feeder links.

2.3 Ground and user segment

Cellular systems integration

The GMPCS systems are designed to becompatible with cellular systems andcomplement their coverage. GSM net-work elements and functionality likemobility management will be reused with

minor modifications, and multi-modehandsets are offered for various cellularstandards. Typically, the terminal willoperate in its cellular mode when it iswithin cellular coverage and only beswitched to satellite mode (manually orautomatically) when the user is out ofcellular coverage or in an area only offer-ing an incompatible standard. Figure 2illustrates this concept.

Gateways

All systems use a set of gateways tointerconnect with terrestrial networks,fixed and mobile. The exact number of

gateways depends on the orbital heightof the satellites, use of ISLs and theoperational strategy.

The operation of a gateway could, from arevenue perspective, be desirable or be ameans to avoid bypassing of the nationalfixed networks. Therefore, the number ofgateways in a system could be higherthan strictly required.

GMPCS gateways are interconnectedthrough a global network used for sig-nalling, exchange of administrative dataand in some cases even for routing oftraffic.

Figure 1 Mobile link frequency assignments

Satellitecoverage

Cellularcoverage

1610 MHz 1626,5 MHz

Iridium (uplink)

Iridium (downlink)

Globalstar (uplink)

Odyssey (uplink)

ICO (downlink)

1980 MHz 2010 MHz

ICO (downlink)

2170 MHz 2200 MHz

Globalstar (uplink)

Odyssey (uplink)

2483,5 MHz 2500MHz

L-bandL-band S-band

Figure 2 GMPCS complements cellular coverage


All systems aim to reuse existing GSMtechnology and interfaces in their gate-ways.

Terminal equipment

The handheld, portable phones used forGMPCS are slightly larger and heavierthan ordinary single mode cellular hand-sets. Terminals will be available as singlemode satellite only and as dual (triple)mode cellular/satellite. Available proto-types of such dual mode handsets weighapproximately 400 g. The average trans-mit power of the terminals is in the samerange as for cellular only handsets, whichis a necessity in order to comply with thetype approval conditions for handheldterminal equipment.

Terminals are expected to be moreexpensive than existing cellular handsets.Information available indicates pricesvarying between $500 and $3,000 perunit.

Pagers and fixed terminal types, mainlydesigned for rural and remote areas, willalso be available. These include tele-phone booths and satellite radio-antennaunits for standard PABX equipment.

2.4 Numbering and addressing

One number

The ITU has assigned the shared countrycode 881 to GMPCS operators on ashared basis. One digit Network Identifi-cation Codes following the country codeidentifies each GMPCS operator as oneinterconnected global entity and enablesrouting of GMPCS calls to the nearestgateway. Single mode GMPCS terminalswill be assigned numbers from this num-bering plan.

Dual-mode GMPCS subscribers will alsobe addressed on one number, indepen-dent of whether she/he is communicatingin cellular or satellite mode. This singlenumber could either be a national cellularnumber or an 881 satellite number de-pending on the profile and needs of eachsubscriber.

Apart from the user-convenience of hav-ing one address, the choice of numberingplan impacts routing and charging, espe-cially in the to-mobile direction, as dis-cussed below.

National cellular numbers

A call to a dual-mode GMPCS subscriberwith a national cellular number willalways be routed via his/her home cellu-lar network (country), irrespective of thecommunication mode used, becausecellular numbers are part of the nationalnumbering plan and not recognised asGMPCS/cellular numbers until they areanalysed in the home network.

If the call originates outside the sub-scriber’s home country, it will always becharged as any other international call tothat country. If the call originates withinthe subscriber’s home country, it will becharged as a cellular call. When the dual-mode GMPCS subscriber is operating insatellite-mode, this charge is not ex-pected to be sufficient to cover theexpenses. As a result, the called GMPCSsubscriber will need to be charged forincoming calls (split billing).

The question then is whether a dual-mode GMPCS subscriber finds it attrac-tive to pay for incoming calls when oper-ating in satellite mode.

Figure 3 Prototype of the Kyocera and Motorola dual-mode handsets for Iridium

ICO Odyssey Globalstar Iridium0

500

1000

1500

2000

2500

3000

3500

4000

4500

Figure 4 Capacity per satellite in number of voice channels


Dedicated 881 country code

A call to a dual-mode GMPCS subscriberwith an 881 number will always berouted to a satellite gateway. If the ter-minal is in cellular mode, the call will bere-routed to the destination cellular net-work.

Independent on where in the world thecall originates such calls will always becharged at the correct GMPCS charge,irrespective of whether the terminal isoperating in cellular or satellite mode.This charge is expected to be higher thanfor cellular calls.

The question in this case is whether thecalling party finds it attractive always topay GMPCS charge to reach a dual modeGMPCS subscriber, even when this sub-scriber is within GSM coverage andcommunicating in GSM mode.

Table 2 The Iridium satellite launch history

Launch Date Launch Vehicle Launch Provider # of Launch SiteSatellites

5 May 1997 Delta II Boeing 5 Vandenberg, California

18 June 1997 Proton Krunichev 7 Baikonur, Kazakhstan

9 July 1997 Delta II Boeing 5 Vandenberg, California

20 August 1997 Delta II Boeing 5 Vandenberg, California

14 September 1997 Proton Krunichev 7 Baikonur, Kazakhstan

26 September 1997 Delta II Boeing 5 Vandenberg, California

8 November 1997 Delta II Boeing 5 Vandenberg, California

8 December 1997 Long March China Great Wall 2 Taiyuan, China2C/2D Industry

20 December 1997 Delta II Boeing 5 Vandenberg, California

18 February 1998 Delta II Boeing 5 Vandenberg, California

25 March 1998 Long March China Great Wall 2 Taiyuan, China2C/2D Industry

30 March 1998 Proton Krunichev 7 Baikonur, Kazakhstan

6 April 1998 Delta II Boeing 5 Vandenberg, California

2 May 1998 Long March China Great Wall 2 Taiyuan, China2C/2D Industry

17 May 1998 Delta II Boeing 5 Vandenberg, California

2.5 System capacity

The GMPCS systems promote them-selves as an extension to cellular systemsrather than direct competitors. The abilityto compete is in fact rather low, as thetotal number of voice channels for allfour systems combined is below 200,000.

The total frequency allocations for MSSin L- and S-band are 93 MHz. For com-parison, one channel for analogue satel-lite TV broadcast requires approximately27 MHz. This illustrates the limitation inthe total capacity these systems can offer,and the limitation in number of systems.A high degree of frequency reuse isrequired to offer an acceptable capacity.

Satellite capacity is for Iridium andGlobalstar satellites limited by powerconstraints. For replacement satellites thecapacity may be higher, as this constraintis expected to be overcome.

The number of voice channels persatellite is higher for the MEO systems,which is natural due to the lower numberof satellites and the instant area to becovered by one satellite. Since satellitecapacity influences satellite cost, usingMEO type satellites in a LEO configura-tion to increase capacity would raise thetotal system cost to an unacceptablelevel.

3 Systems overview

3.1 Iridium

Motorola announced the ambitiousIridium plan eight years ago, as the firstof nearly a dozen quite similar proposalsfrom various vendors and operators. Thesystem was named after the element with77 electrons, since the original proposalassumed a 77 satellite constellation. In1992 Motorola announced system en-hancements enabling a reduction in sizeof the constellation to 66 satellites. The66 satellites are located 780 km abovethe earth’s surface in a LEO orbit [5].11 satellites are equally spaced in each ofthe 6 near polar orbits (inclination 86.4°)together providing global coverage. Eachsatellite completes a full orbit cycleevery 100 minutes. An illustration of theIridium satellite constellation is given inFigure 5.

Each of the 66 satellites weighs approxi-mately 700 kg and provides coverage inan area of 600 km in diameter. The 48spot beams provide a cellular footprintstructure. The satellite lifetime is ex-pected to be 5–8 years. Iridium satelliteshave been launched using Delta II (Boe-ing, USA), Long March 2C/2D (ChinaGreat Wall Industry, China) and Proton(Khrunichev State Research and Produc-tion Space Centre, Russia) rockets, eachcarrying five, two and seven satellitesrespectively. Fifteen successful launchessince May 1997 have placed 67 opera-tional satellites in orbit. The last fivewere added on 17 May 1998 and markedthe technical completion of the venture’s66 satellite constellation. Five of the 72satellites which have been launched aresuffering from mechanical or communi-cation problems, reducing the number ofspare satellites to one. Table 2 lists theIridium launches.

Each satellite is linked with two satellitesin the same plane (one in front and onebehind) and with one in the adjacentplane on both sides via ISLs. None of


Iridium’s competitors use this technol-ogy.

The Iridium ground network consists of11 gateways interconnecting the Iridiumsatellite network with the PSTN (PublicSwitched Telephone Network) and otherterrestrial cellular networks. The smallnumber of gateways is possible sinceISLs are used.

Iridium has signed a $3,450 million con-tract with Motorola for building, launch-ing and operating the space segment,excluding gateways. In addition, Iridiumhas executed a five year $2,880 millioncontract with Motorola for operation andmaintenance of the satellite network,including satellite launches for replenish-ment. Raising $4,715 to fund the ongoinginfrastructure build-out and commercial-isation, Iridium has completed its antici-pated funding needs.

The system is owned by Iridium LLC,an international consortium of 19 equitypartners from 15 countries. Motorola isthe largest investor (19 %), but the list ofinvestors also includes other large manu-facturers as Lockheed Martin and strongtelecom operators as Korea Mobile Tele-com.

Iridium will launch their service 23September 1998.

3.2 Globalstar

The Globalstar system [6] was originallyproposed in 1986 by Ford Aerospace toprovide mobile satellite communicationsservices to automobiles. Globalstar LP,which was founded by Loral Space &Communications and Qualcomm in1991, has now evolved into a world-wideconsortium of 12 strategic partners in-cluding Alcatel, France Telecom, AirTouch and Vodafone.

48 LEO satellites organised in eight cir-cular orbits inclined 52° relative to theequatorial plane with six equally spacedsatellites per orbit constitute the Global-star space segment. Each satellite com-pletes an orbit cycle every 113 minutes atan orbital height of 1,414 km. The con-stellation covers the area between 70°Sand 70°N latitude with the ‘satellite den-sity’ on each hemisphere at its highest ina band centred approximately at ±40° lat-itude. The non-coverage area over thepoles can be recognised in Figure 7.

Figure 6 The Iridium satellite

Figure 5 The Iridium satellite configuration


The 450 kg Globalstar satellites areunlike the Iridium satellites of the bent-pipe type. Each satellite has 16 beamsand is designed for a minimum lifetimeof 7.5 years. The Globalstar constellationwill be lifted into space using Delta II(Boeing, USA), Zenit-2 (NPO Yuzhnoye,Ukraine) and Soyuz (Starsem, France)launch vehicles carrying eight, twelveand four satellites respectively. In Febru-ary and April 1998 eight Globalstar satel-lites were placed in orbit using two DeltaII launch vehicles. The Globalstar launchschedule is presented in Table 3.

Initially, Globalstar planned to build upto 210 gateways to minimise land-basedlong-distance charges and to respectnational boundaries and thereby over-come potential problems with thenational regulatory authorities. However,Globalstar recently announced that it willrely on only 50–75 gateways since somecountries have decided to use gatewaysin neighbouring countries rather than gothrough the expense of building theirown.

The total accumulation of capital isapproximately $2,875 million includingequity, bank financing and resale of gate-ways. Loral Space & Communications isthe largest investor with approximately33 % ownership. The initial Globalstarsystem cost is estimated to be $2,600million.

Globalstar plans to launch a service for‘premium customers’ with 44 satellitesearly 1999 [6] [7].

3.3 ICO Global Communications

ICO Global Communications [8] wasestablished in 1995. The original pro-posal came from Inmarsat, which still isa large investor. The ICO owners com-prises 60 investors from 51 countries,including Hughes, TRW4, BT and T-Mobil. In addition, all Inmarsat membersare indirect investors through Inmarsat’scollective contribution to ICO.

The 10 ICO satellites are arranged in twocircular MEO orbits 10,355 km above theearth surface. Each of the orbital planesare inclined 45° relative to the equatorialplane as Figure 8 illustrates. A satellite

Figure 7 The Globalstar satellite configuration

Launch Date Launch Vehicle Launch Provider # of Launch SiteSatellites

14 February 1998 Delta II Boeing 4 Cape Canaveral, Florida

24 April 1998 Delta II Boeing 4 Cape Canaveral, Florida

3Q98 Zenit NPO Yuzhnoya 12 Baikonur, Kazakhstan



1Q99 Soyuz Starsem 4 Baikonur, Kazakhstan

1Q99 Soyuz Starsem 4 (1) Baikonur, Kazakhstan

2Q99 Soyuz Starsem 4 (1) Baikonur, Kazakhstan

Table 3 Globalstar calendar of completed and scheduled launches

4 TRW joined forces with ICO inDecember 1997 and terminated theOdyssey project due to lach of financ-ing.


completes one cycle around the earth insix hours. Each satellite will coverapproximately 30 per cent of the earth’ssurface at a given time ensuring a signifi-cant coverage overlap.

MEO satellites are heavier than LEOsatellites as a consequence of the higherpower requirements. The 2,600 kg heavysatellites are manufactured by Hughes, adesign based on the HS601 geostationarysatellite bus. Communication through the163 beam satellite is transparent. The lifespan of ICO satellites is expected to beapproximately 10–12 years. The satelliteswill be launched by a variety of vehicles:Atlas IIAS (Lockheed Martin, USA),Delta III (McDonnell Douglas, USA),Proton (Khrunichev State Research andProduction Space Centre, Russia) andZenit/Sea Launch (NPO Yuzhnoye,Ukraine and a Boeing Consortium) willlaunch one, five, three and three satellitesrespectively. The first launch is sche-duled for late 1998.

The 12 ICO gateways, called SANs(Satellite Access Nodes), are intercon-nected via the ICONET ground network.The SANs will be the primary interfacebetween the satellites and the terrestrialnetworks.

ICO has awarded Hughes Space andCommunications two contracts worth$2,200 million to build and launch the12 satellites. In addition, an NEC ledconsortium is building the ground net-work in a contract initially worth $616million. Including other minor contracts,ICO is investing more than $3,000 intheir global satellite system. ICO hasmanaged to raise approximately $1,500million in equity from investors.

Current plans assume that the ICOsystem will be operational third quarter2000.

3.4 Odyssey

Odyssey was proposed by TRW Inc. [9]with Teleglobe Canada as the only addi-tional investor. In January 1997, Odysseysigned a service provider agreement withChinaSat and on 11 March 1997, Dae-woo Corp. of South Korea and KumhoGroup announced that they had in prin-ciple agreed with TRW to join the pro-ject.

17 December 1997 TRW announced thatit was acquiring an equity interest in ICO

Figure 8 The ICO satellite configuration

Figure 9 The Odyssey satellite configuration


and that the Odyssey project thereby wasterminated. Odyssey as envisaged is de-scribed below.

The Odyssey satellite network was simi-lar to ICO. 12 satellites circumferencedthe earth in three circular MEO orbits at10,354 km altitude. The orbits were in-clined 50° relative to the equatorialplane. It took a satellite six hours to com-plete a full cycle. The planned constella-tion is illustrated in Figure 9.

The Odyssey satellite footprint consti-tuted a 61 cell structure. The expectedsatellite lifetime was 15 years. Onlyseven gateways were used to connectwith terrestrial networks.

The cost figure of $3,200 million in-cluded gateway costs. Odyssey hadserious problems with the financing ofthe system with only $150 million com-mitted to the project at the time of termi-nation.

The operational target date for Odysseywas 2001.

4 Recent developments

It is almost ten years since the firstGMPCS systems were announced, andnone of them are yet in service. Duringthose ten years there has been a signifi-cant change of scenery, affecting thoseassumptions initially made by thefounders of the GMPCS systems. Thesedevelopments will inevitably affect thesize and characteristics of the GMPCSmarket as well as the conditions foroperating such services globally. Wewill discuss some of these developments,namely the rapid penetration of terrestrialcellular services, the introduction ofregional satellite-based systems, thedemand for broadband services andthe regulatory barriers and possiblemeasures.

4.1 The cellular explosion

AT&T projected early on that the totalworld market for cellular phones wouldbe about 900,000. With more than 165million cellphones in use world-widetoday, we know that estimate was off byat least a factor of 180, and still growing.With such working assumptions, theoutlooks for a global handheld satellite-based phone, nearly independent of localinfrastructure, were of course differentthen than they are today. GSM and other

cellular standards are present in morethan 140 countries [10], and this numberis rapidly increasing. Market forecasts5

indicate that the global market forGMPCS services will reach 12 millionsubscribers in 2005. This represents2–3 % of the expected global cellularmarket. However, large areas remainunserved by cellular and fixed infrastruc-ture and a patchwork of cellular networkstandards around the world makes travel-ling with a cellular phone difficult. Thisdevelopment has caused the GMPCSoperators to seek new business opportu-nities in 1997. Iridium announced theconcept ‘Iridium Cellular Roaming Ser-vice’. By the use of multi-mode (dual-/triple-mode) terminals, and translatingdiffering cellular protocols, Iridium willact as an interworking system acrossincompatible cellular standards. In prin-ciple, a dual mode terminal supportingtwo cellular standards (without satellitemode) may be used, and Iridium, as athird party, will only provide the inter-working and billing for all calls [11].

4.2 Regional satellite systems

Another category of systems expected tooffer the GMPCS systems strong compe-tition is the growing number of GEO-based regional systems. These systemsare established with significantly lowerinvestments than the GMPCS systems.

One such system is the Asia CellularSatellite System (ACeS). This systemwill provide voice, facsimile and pagingservices via GEO satellites to hand-heldmobile and fixed terminals throughoutSoutheast Asia, India and China. Launchof the first ACeS satellite is planned for1998, with service beginning a fewmonths later.

4.3 The demand for morebandwidth

The GMPCS systems, like the cellularsystems, provide data communicationsat low bitrates. There is an increasingdemand in the fixed networks for morebandwidth, partly fuelled by the rapidgrowth of Internet services. Mobile userswill also require higher bitrates for moreefficient data communication and Inter-net access. The GSM operators are al-ready planning the implementation ofboth circuit switched and packetswitched higher rate enhancements totheir GSM data services. HSCSD (HighSpeed Circuit Switched Data) will pro-vide data rates up to 57.6 kb/s and GPRS(General Packet Radio Service) will offerdata rates up to 115 kb/s to GSM users.Even before the service introduction ofthe GMPCS systems, one might claimthat GMPCS, in certain market segments,will provide less satisfactory data servicecapabilities. In light of these develop-ments, both Iridium and Globalstar,among others, have announced next gen-eration systems targeting broadband andmultimedia service distribution. This willbe further discussed in chapter 5.

5 Performed for Telenor by Gemini C4

Labs.

Iridium Globalstar ICO Odyssey

Orbit LEO LEO MEO MEO

Altitude [km] 780 1,414 10,355 10,354

Number of satellites 66 48 10 12

Number of planes 6 8 2 3

Inclination 86.4 ° 52 ° 45 ° 50 °

Orbital period 100 minutes 113 minutes 6 hours 6 hours

Number of gateways 11 50–75 12 7

System cost [billion $] 3.4 2.6 3.0 3.2

Commercial operation 23 Sep 1998 Early 1999 3Q 2000 –

Table 4 Summary of GMPCS systems characteristics


4.4 Regulatory restrictions,measures and deliberationinitiatives

Regulatory restrictions are recognised asa key problem for global and regionalsatellite systems seeking to provide trans-border communications. These problemsare often less predictable and may easilybe underestimated when forecasting theglobal market potential. There may beseveral reasons for maintaining regula-tory barriers:

• Concern about bypass of local net-works, of prime importance for de-veloping countries with poor infras-tructures

• Security (especially a concern forcountries in conflict areas and incountries with drug dealer problems)

• Lack of adequate legislation for thehandling of a more liberalised environ-ment

• Protection of national telecom opera-tors and service providers

• Radio frequency interference con-siderations

• Lack of qualified staff and in handlinglicence requests etc.

The regulatory measures may take differ-ent forms, and among such measures are:

• Prohibitions on use of satellite telecomequipment

• High custom duties and licence fees.Custom duties exceeding 50–60 % ofthe equipment cost, and annual licencefees of several thousand US dollars arenot uncommon, especially in somedeveloping regions outside Europe andNorth America. This may apply to sys-tem and gateway operators as well asservice providers

• Incomplete and badly co-ordinatedprocedures, creating bureaucraticdelays in handling applications andcommissioning of terminals

• Neglect of a country to respect typeapprovals of another country, even incases where both countries have signedagreements of mutual recognition oftype approvals

• Jamming of satellite systems not com-plying with national requirements.This is an extreme measure that hasnot so far been applied in practice, butwhich is talked about by representa-

tives of certain African countries as apossible ultimate reaction to enforcenational policies.

However, several initiatives haverecently been made to progress theliberalisation of the regulatory regimesfor satellite communications, and mobilesatellite communications in particular.The introduction of the GMPCS systemsin the near future has been one of thedriving factors for this progress. On theglobal scale we should first of all takenotice of the work done by the twoUnited Nations (UN) organs, ITU andWTO (World Trade Organisation), inparticular of the following initiatives:

GMPCS MoU

A Memorandum of Understanding(MoU) has been worked out by theITU World Telecommunications PolicyForum in order to facilitate GMPCS.The document was opened for signingin March 1997.

The signatories to the MoU are admini-strations, GMPCS operators, serviceproviders and manufacturers. Withintheir respective roles they agree to co-operate to stimulate free circulation ofequipment, by addressing topics such astype approval, licensing and marketing ofterminals, custom arrangements, andaccess to traffic data. Both Telenor andPT (Norwegian Post andTelecommunications Authority) havesigned this ITU document.

Activities within the WTO

Ministers from 28 countries (includingNorway), representing 84 % of the tradein information technology, issued a Dec-laration in November 1996 on Trade inInformation Technology (ITA) that willcome into effect once the subscriptionrepresents 90 % of the trade. A principalgoal is to have the customs duties onequipment reduced to zero by year 2000.

The work of WTO’s Group on BasicTelecommunications (GBT) to negotiatecommitments to liberalise basic telecommarkets was successfully concluded on15 February 1997. The commitmentsrelate to issues such as licensing, inter-connection, transparency, independenceof regulatory bodies, spectrum and uni-versal service obligations. 69 countriesmade commitments by the Februarydeadline. The formal entry into force ofthe commitments is scheduled to be

1 January 1998. But where an individualparticipant’s commitment for a specialservice is to be phased in, the actualimplementation will take place at thedate specified in the schedule.

Additional activities to alterregulations in Europe

The EU (European Union) Commissionand the European Radio Communica-tions Committee (ERC), an organ withinCEPT (Conférence Européenne desPostes et Télécommunications), haveboth taken steps to expedite the develop-ment towards a more liberalised telecommarket in Europe, through a series ofCommission Directives, Council resolu-tions, recommendations and action plans,addressing all the essential issues in-volved in removing exclusivity rights andopening the market for free competition.

5 Future Evolution ofPersonal SatelliteCommunications

As already mentioned, a major trend incommunications is the demand for morebandwidth to cater for higher bitraterequirements for the provision of Internetand multimedia type services. The trendis well established in the fixed networks,the terrestrial cellular systems, like GSM,are already planning enhancements forhigher bitrate data services, and the satel-lite community is of course also wellawake.

The next generation mobile communica-tion has to a large extent been synony-mous to the standardisation work goingon in ETSI (UMTS) and ITU (IMT-2000). Also in UMTS (Universal MobileTelecommunications System), focus isput on the bandwidth trends, and in arecent report from the UMTS Forum[12], multimedia services are noted as akey driver for UMTS. UMTS will needto provide both narrow and widebandservices. Part of the UMTS concept is asatellite component, providing satellitebased services, both narrowband andwideband. The UMTS Forum states thatthese will have a place in the market intheir own right, and in providing bothearly and temporary coverage before andas terrestrial networks are rolled out.Integration of the terrestrial and satellitecomponents of UMTS is highly desirableto provide actual global coverage for atleast a subset of UMTS services. Suchcommonalities are also expected to


reduce the costs of the satellite terminals,and thereby make them more attractive.

The work within ETSI (EuropeanTelecommunications Standards Institute)of specifying a UMTS-specific satellitesystem to take the role as satellite com-ponent has not been successful and is notlikely to happen. We regard it as morelikely that proprietary systems, such asthe GMPCS systems and other futuresystems under development, will beadopted as part of the UMTS ‘family’in the long term. Among wide- andbroadband satellite system conceptsunder development are Microsoft’sTeledesic, Motorola’s Celestri andInmarsat’s Horizons.

These initiatives represent new chal-lenges in the history of satellite commu-nications. Teledesic’s original proposalincluded a network in the sky of no lessthan 840 LEO satellites providing Inter-net services to companies and end-usersglobally [13]. The concept has now beenscaled down to 288 satellites. The Tele-desic project is backed by initiatorMicrosoft along with the Boeing corpora-tion, Matra Marconi Space, and sinceMay 1998, also Motorola. After Moto-rola joined Teledesic the faith of Celestri[14], a hybrid LEO/GEO system (63LEO satellites) for broadband services isannounced. Inmarsat’s Project Horizonsplans a GEO system providing widebandservices to portable terminal units thesize of a notebook PC. These are only afew of several plans and concepts re-vealed. The use of higher frequencybands, such as the 20/30 GHz Ka-bandis a common feature for many of theseplans opening up for much more capacitythan what is available in the currentlyused L- and S-bands.

A certain degree of commonality acrossservices provided by the fixed networks,terrestrial cellular networks (current andenhanced versions), UMTS/IMT-2000(International Mobile Telecommunica-tions System) and the satellite systems ismost desirable, especially from the end-user point of view. Commonality in userinterfaces, consistent user profiles andsimple numbering and addressingschemes are among the aspects expectedto be strong future user requirements inthe potential jungle of systems and termi-nal types. This fits well into the trend ofFixed Mobile Convergence (FMC) andshould be kept in mind when developingand introducing new systems and ser-vices, both terrestrial- and satellite-based.

In light of this, personal communicationstakes on a broader meaning. Multimediaservices may in nature not be well suitedfor pocket sized terminals and the aspectof Terminal Mobility isolated, wellknown from GSM, will not be the mainfocus. Personal Mobility and ServiceMobility, the ability for a user to be in-dependent of a specific terminal whenaccessing his/hers services, and theability to access those personalised ser-vices independent of terminal type andserving network, will be importantaspects. In developing the future satellitesystems for multimedia services, theseaspects must be taken into considerationto secure the position of satellite commu-nications into the exciting next century oftelecommunications.

Acknowledgements

The authors would like to thank Mr.Kristen Folkestad, Telenor Satellite Ser-vices, who has contributed to this articlewith his experience and knowledge onregulatory matters. Our appreciation alsogoes to Mr. Håkon Lundamo, formerlyTelenor Satellite Services, and Mr. ScottJacobs, Gemini Consulting, who workedwith us on the GMPCS issue for nearlysix months.

References

1 Iridium Today, 3 (4), July 1997.

2 Werner et al. Analysis of SystemParameters for LEO/ICO-SatelliteCommunication Networks. IEEE J.Selected Areas Commun., 13 (2),371–381, 1995.

3 Ramesh, R. Availability Calculationsfor Mobile Satellite CommunicationsSystems. IEEE Proc. of Vehiculartechnology Conference, April 1995.

4 Ananasso, F, Priscoli, F D. The Roleof Satellite in Personal Communica-tions Services. IEEE Journal onSelected Areas in Communications,13 (2), 1995.

5 Iridium official Internet web-site:http://www.iridium.com.

6 Globalstar official Internet web-site:http://www.globalstar.com.

7 Kiernan, V. Globalstar at the startinggate. Satellite Communications,36–39, July 1997.

8 ICO official Internet web-site:http://www.i-co.co.uk.

9 TRW official Internet web-site:http:// www.trw.com/seg/sats/ODY.html.

10 Mobile Communications Interna-tional, 43, 67, 1997.

11 Iridium Today, 3 (3), 1997.

12 UMTS Forum. A Regulatory Frame-work for UMTS. 25 June 1997.(Report no. 1.)

13 Teledesic official Internet web-site:http://www.teledesic.com.

14 Celestri official Internet web-site:http://www.mot.com/GSS/SSTG/pro-jects/celestri.

Abbreviations

ACeS Asia Cellular Satellite System

CDMA Code Division MultipleAccess

CEPT Conférence Européenne desPostes et Télécommunications

ERC European Radio Communi-cations Commission

ETSI European Telecommuni-cations Standards Institute

EU European Union

FMC Fixed Mobile Convergence

GBT Group on Basic Telecom

GEO Geostationary Orbit

GMPCS Global Mobile PersonalCommunication Services

GPRS General Packet Radio Service

GPS Global Positioning System

GSM Global System for MobileCommunication

HSCSD High Speed Circuit SwitchedData

ICO Intermediate Circular Orbit


IMT-2000 International MobileTelecommunications System

Inmarsat International Mobile SatelliteOrganization

ISDN Integrated Services DigitalNetwork

ISL Inter Satellite Link

ITU International Telecommuni-cations Union

kbps kilobit per second

LEO Low Earth Orbit

LLC Limited Liability Company

LP Limited Partnership

MEO Medium Earth Orbit

MoU Memorandum of Under-standing

MSC Mobile Switching Centre

MSS Mobile Satellite Systems

PABX Private Automatic BranchExchange

PLMN Public Land Mobile Network

PSTN Public Switched TelephoneNetwork

PT Norwegian Post and Tele-communications Authority

SAN Satellite Access Node

SMS Short Message Service

S-PCS Satellite Personal Communi-cation Services

TDMA Time Division MultipleAccess

UMTS Universal Mobile Tele-communications System

UN United Nations

WARC World Administrative RadioConference

WTO World Trade Organization

Arvid Bertheau Johannessen was, at the time of writing thisarticle, Research Scientist at Telenor R&D, Kjeller. He hasbeen with Telenor since 1990, working with system- andnetwork aspects of mobile communication systems in gen-eral and satellite based systems in particular. Since Feb-ruary 1998 he has been employed by Telenor Satellite Ser-vices AS, Mobile Division, with focus on development andoperation of the Inmarsat services as well as future strate-gies for mobile satelite services within Telenor.


Arild Flystveit was, at the time of writing this article, Re-search Scientist at Telenor R&D, Kjeller, Personal Commu-nications Group. He has been employed by Telenor since1995, most of the time working with system- and networkaspects of mobile satellite communication systems. SinceMarch 1998 he has been employed by Telenor Mobil, Net-work Division, working with product development in generaland implementation of Intelligen Networks in particular.



This article presents the developmentof CTM within ETSI (EuropeanTelecommunications Standards Insti-tute) and EURESCOM (EuropeanInstitute for Research and StrategicStudies in Telecommunications), witha focus on the work done inEURESCOM Project 606 CTM (Cord-less Terminal Mobility) Phase 2. Viewsand statements presented in this articleare Telenor’s only, and do not neces-sarily represent the views of otherEURESCOM P606 partners. Partnersin the EURESCOM P606 are BritishTelecom, CSELT, FINNET Group,France Telecom, Swiss Telecom PTT,Telefonica, Telenor and Telia.

1 Introduction

The mobile phone seems today to be con-sidered by most customers as an additionto their fixed line subscription, and notyet as a substitute. A reason for thiscould be that users prefer the higherspeech quality and quality of service offixed telephony, but more important isprobably that the cost of using a mobilephone is considerably higher than using afixed line. One should, however, expectto see a rapid decrease in the price differ-ence over the next few years.

Thus, in light of the rapid growth in themobile communications market, onemight argue that introducing CTM willhelp the fixed network operators toprevent a massive loss of customers tomobile operators. At the same time,CTM could be looked at as an importantvalue-added service in the fixed network,representing a potential source of incomefor the fixed network operator. Anotheraspect that could prove to be an advan-tage for CTM is the higher capacity ofDECT compared to GSM, for example.The increased capacity could be used tooffer higher data rates and higher trafficcapacity in certain areas. As the GSMoperators are expecting an increase in thedata traffic in the years to come, higherdata rates could prove to be a powerfulfeature for CTM. Also, the developmentof packed data will be an importantfeature in order to support efficient solu-tions for Internet access, e-mail, filetransfer, etc. Finally, with the develop-ment of CTM/GSM roaming, the end-users will gain a higher quality of ser-vice. Such a combination of technologiesmay represent a first step towards a thirdgeneration mobile system.

ETSI CTM Phase 1 will support thefollowing core requirements [2]:

• outgoing calls

• incoming calls

• location handling

• authentication and ciphering

• emergency calls.

In addition, the optional requirementsdefined for CTM phase 1 are [2]:

• service profile modification

• service profile interrogation

• call forwarding not reachable

• subscription registration /deregistration

• location registration suggest

• network authentication.

CTM phase 1 is defined for the singlepublic network case, for roaming be-tween public IN based networks, and forroaming between private networks andpublic IN networks. For the single publicnetwork, the functional architectureshown in Figure 1 has been developed.

In phase 1, the radio access part will besupported by DECT GAP (Generic

Cordless Terminal Mobility (CTM)– Support of cordless mobility within the fixed network

I V A R O L D E R V I K , K N U T B A L T Z E R S E N , T O R I L N A T V I G A N D G I S L E P E D E R S E N

SCUAF

CCAF

CUSF

CCF

SSF

SCFmm SCFsl

SDFmm SDFsl

INAP

DSS1+

ISDN access(BA or PRA)

mm = mobility managementsl = service IogicSDF = Service Data FunctionSCF = Service Control FunctionSSF = Service Switching Function

CCF = Call Control FunctionCUSF = Call Unrelated Service FunctionSCUAF = Service Control User Agent FunctionCCAF = Call Control Agent Function

Figure 1 CTM phase 1 architecture

2 History andbackground

2.1 ETSI work on CTM

The mobile communication market hasshown a great increase and has become abig success in later years. The Scandina-vian countries have a mobile penetrationaround 40 %, and there are still no signsof saturation in the market. In order to beable to offer mobility even in the fixednetwork, the work on CTM has beencarried out within both ETSI andEURESCOM since 1994. The concept ofCTM is to offer IN supported terminaland service mobility, the cordless accesspart being based on DECT (Digital En-hanced Cordless Telecommunications)or CT2 (Cordless Telecommunicationsgeneration 2). ETSI have divided thework on CTM into a phased approach:CTM phase 1, CTM phase 2 and CTMphase 2+. The work is co-ordinated bythe ETSI Project CTM (EP CTM), andinvolves several bodies within ETSI.NA1/2 is responsible for the servicedescription, NA6 is responsible for archi-tecture and functionality, the DECT pro-ject is responsible for the radio accesspart and SPS is responsible for the net-work protocols. The CTM phase 1 stan-dard was finalised in 1997. CTM phase 2will probably be finalised in 1998, andCTM phase 2+ is divided into differentfeature packages (eg. CTM Phase 2+ FP1). So far, no estimate for finalisation ofthe FPs is given. A short description ofthe different phases will be contained inthe following.


Access Profile), or alternatively by CT-2.The access will be based on ISDN accessand DSS1+. The signalling within the INnetwork will be based on ETSI Core-INAP CS1 and CS2.

To further develop the concept of CTM,ETSI is currently also undertaking workon CTM Phase 2. This phase is addingsome new aspects to the phase 1 defini-tion of CTM, like several CTM supple-mentary services and the possibility toindicate for a CTM user his or her mail-box status (message waiting indication)[6]. Another important aspect of phase 2is the work on external handover (seeclause 3.5). The radio access part willin phase 2 be supported by DECT CAP(CTM Access Part).

In order to be able to introduce newfeatures and services in a flexiblemanner, the ETSI CTM project hasagreed that CTM Phase 2+ should bedivided into feature packages (FPs). Sofar, the CTM project has initiated workon four new FPs:

• CTM Phase 2+ FP 1: Circuit Switcheddata Services; 32 and 64 kbps Un-restricted Digital Information

• CTM Phase 2+ FP 2: CTM ShortMessage Service (SMS)

• CTM Phase 2+ FP 3: CTM/GSMRoaming (feasibility study)

• CTM Phase 2+ FP 4: Point-to-PointProtocol (PPP) interworking for Inter-net access and general multi-protocoldatagram transport (suggested).

New FPs could be added in the furtherdevelopment of CTM.

2.2 EURESCOM work on CTM

Eurescom P606 CTM Phase 2 work wasinitiated in June 96, and was completedin January 98. The main focus of thework has been to develop the CTMfeature further, contributing to the ETSIstandardisation work on CTM with focuson ETSI CTM Phase 2+. P606 has alsoundertaken work on the migrationtowards UMTS (Universal MobileTelecommunications System) by study-ing what in EURESCOM has beencalled CTM pre-UMTS.

In EURESCOM, earlier projects havealso had a focus on personal mobility /terminal mobility within IN. The firstEURESCOM project addressing mobility

aspects within IN was P230 ‘Enablingpan-European services by co-operationbetween PNO’s IN platforms’. P230mainly focused on interworking betweenIN platforms, but also aspects of DECTused as radio access and IN mobilitycontrol were treated. To continue thework on mobility in IN, EURESCOMP507 ‘Mobility Applications Integrationin IN’ was initiated, focusing on thesupport of mobile services on IN archi-tectures, considering the CTM service asa basis for short-term scenarios [1]. TheEURESCOM P606 was a follow up ofearlier projects, but with a stronger focuson developing features and services forCTM. The pre-UMTS part of P606 wasseeking an evolution path for CTMtowards the UMTS, contributing toUMTS work in ETSI, among others.

3 EURESCOM P606

EURESCOM P606 CTM Phase 2 wasdivided into three different tasks, moreor less following the three stage method:

• Task 1 focused on service features anddetailed service description both forCTM and for CTM pre-UMTS

• Task 2 focused on functional descrip-tion, architecture and informationflows for CTM and CTM pre-UMTS

• Task 3 focused on the additions neces-sary on the protocol level to supportthe new services / features developedfor CTM and CTM pre-UMTS.

Work done in P606 has also been usedas a basis for contributions towards ETSICTM standardisation, with focus onETSI CTM Phase 2 and ETSI CTMPhase 2+. It should be mentioned that thephasing of CTM within EURESCOMwas not equal to the phasing of CTMwithin ETSI. Main work areas in P606have dealt with features like CTM/GSMroaming, SMS, handover, on-air sub-scription, data-services, Virtual Home

Environment, etc. EURESCOM P606was also looking into the developmentwithin standardisation towards UMTS.

This chapter will describe some of thework done in EURESCOM P606.

3.1 64 kbps UnrestrictedDigital Information (UDI)

An important set of applications forCTM data services in the consumermarket will probably be several forms ofInternet access (eg. www browsers andftp). For the business user, teleworkingmay represent a group of importantapplications. File transfer (eg. ftp) ande-mail should be supported, but wwwbrowsers would also be interesting tothese users.

These applications need to be supportedin an optimised manner. One possibleway to meet these demands is the64 kbps UDI. The 64 kbps UDI is a fixedbandwidth symmetric ISDN bearer ser-vice functioning more or less as a ‘bitpipe’, which makes it very flexible. It isalready being utilised in fixed networksfor Internet access today.

The major advantage to the users is thatprices may be comparable with standardanalogue services, while transmissionspeed will be well over that of standardmodems. The set-up time will also bemuch shorter than for modems, whichtake some 20 seconds to do their negotia-tion process.

3.1.1 DECT/ISDN InterworkingProfiles

Within the DECT community there isalready considerable progress made indeveloping standards for ISDN overDECT. Two different DECT/ISDN inter-working profiles described below mighthelp cater for the user needs mentionedabove:

EURESCOMP230

EURESCOMP507

EURESCOMP606

ETSI

CTMPhase 1

CTMPhase 2

CTMPhase 2+

Figure 2 Relations between EURESCOM projects and ETSI


In April 1996 the first edition of ‘DECT/ISDN Interworking for end system con-figuration’, ETS 300 434 (also calledISDN Access Profile – IAP), was pub-lished.

The IAP applies when FP (Fixed Part)and PP (Portable Part) together constitutean ISDN terminal. The ISDN applica-tions are located in the PP. There is noIWU in the PP. The FP maps the re-ceived layer 3 messages at the ISDNinterface to DECT layer 3 messages andvice versa.

The bearer services being covered are[ETS 300 434-1]:

• circuit mode speech

• circuit mode 3.1 kHz audio

• circuit-mode 64 kbps unrestricteddigital information.

A work program has been started withinETSI Project DECT on ‘DECT/ISDNInterworking for Intermediate System

Configuration’ DE/RES 03039 (alsocalled Intermediate ISDN access Profile– IIP) [DE/RES-03039].

The IIP applies when FP and PP togetherconstitute a gateway between an ISDNnetwork and an ISDN terminal. The FPand the PP have an IWU each, that mapsthe messages between the ISDN interfaceand the DECT air interface. This impliesthat an ISDN terminal can be connectedto a DECT PP supporting IIP via theISDN S interface.

The bearer services being supported arethe same as for the IAP1.

The profile best supporting the CTM userneeds is the IAP, since the IIP is mostlytailor made for Radio in the Local Loop(RLL) solutions. Nevertheless, the IIPcan be useful in supporting a standardISDN terminal connected to a DECT PPsupporting ISDN as described above.

3.1.2 Impact on existing standards

One important aspect of this service isthat the implementation is only foreseento impact on the access network. Thismeans that only the DECT FP and PPneed to be enhanced to support 64 kbpsUDI. In particular, there is a need tosupport ‘double slot’ (two 32 kbps timeslots concatenated) to be able to reach thedemanded data rate of 64 kbps. This willprobably mean that a hardware upgradeof all FPs in a voice only DECT networkis necessary. It is also important to noticethat the DECT/ISDN interworking pro-file adds extra error correction and ARQmechanisms to the 64 kbps bit stream, tothe cost of a need for new buffers.

The FP will also have to be able to inter-work with ISDN networks.

With regard to standardisation, theupgrade can probably be made by incor-porating parts of the DECT/ISDN pro-file(s) mentioned earlier with the CAP(CTM Access Profile). The only specialdemand 64 kbps UDI will have on mobil-ity aspects is for handover. For voiceconnections bit errors during handovermight be acceptable to the user. Thismight not be the case for a data connec-tion as bit errors will lower effectivethroughput and in the worst case evenlead to loss of connection. This is highlydependent on the type of application. Forexample, Internet applications will bene-fit from strong error recovery proceduresin the TCP/IP protocols.

Another issue is the type of user equip-ment and user interface. Most of today’sdata terminals expect the user to bestationary, which limits the need forsupport of handover.

3.2 Short Message Service

3.2.1 SMS description

The Short Message Service (SMS) in-vestigated in P606 has been modelled onthe basis of the GSM Short Message Ser-vice, preferably to be perceived by theuser as virtually equivalent to the SMSservice in GSM.

This implies that we may define threedifferent SMS services, as in GSM:

1. Mobile (portable part) originatedpoint-to-point SMS

2. Mobile (portable part) terminatedpoint-to-point SMS

3. Cell broadcast SMS.

The point-to-point services enable CTMusers to send and receive short messagesto and from other CTM users or in prin-ciple any other suitable kind of parties(eg. GSM users). The cell broadcast shortmessage service enables the CTM opera-tor to broadcast unaddressed and un-enciphered messages at regular intervalsto all CTM users located within a givengeographical area. The cell broadcastSMS was not investigated by P606.

3.2.2 A possible implementationmethod for the point-to-pointSMS

In EURESCOM P606, CTM SMS is de-scribed as a call unrelated service, mak-ing use of the SCUAF (Service ControlUser Agent Function) and CUSF (Call-unrelated User Service Function)as defined in IN CS-2.

When transmitting a point-to-point shortmessage, the message will be relayed bya short message service centre (SM-SC).Unlike GSM, the information flows be-tween this service centre and the rest ofthe CTM network are considered to bewithin the scope of the CTM specifica-tions, ie. making use of a standardisedprotocol.

As in GSM, the CTM network shouldcontain mechanisms to ensure that a mes-sage is stored in the short message centreuntil the message has been successfullydelivered to the CTM user (or is timed

1 In addition, packet-mode (X-31 caseB) D-channel (packet data) andpacket-mode (X-31 case B) B-channel(packet data) may be supported.

FT PT

ISDNS/T or S

DECTD

DFS DPS

Figure 3 ISDN Access Profile

FT

DIFS = DECT Intermediate Fixed SystemDIPS = DECT Intermediate Portable SystemDIS = DECT Intermediate SystemIWU = Interworking UnitIA = Interface Adapter

PT

ISDNV, T and P

DECTD

DIFS DIPS

IWU

IA

IWU

ISDNS

DIS

Figure 4 Intermediate ISDN Access Profile

DFS = DECT Fixed SystemDPS = DECT Portable SystemFT = Fixed TerminationPT = Portable Termination


Figure 5 Architecture for SMS in a CTM network

out), as well as to report back to the orig-inating CTM user when the message hasbeen received at the SM-SC.

A key driver in the method describedbelow is to implement SMS with a mini-mum of impact on existing protocols.Hence, the most suitable method forinterworking against the SM-SC is con-sidered to be the one described in GSM03.47, basing the interworking uponITU-T Signalling System No. 7.

3.2.3 Implementation of SMS inCTM network

Figure 5 shows the architecture for CTMSMS, where the user is located in his orher home network, or in a visited net-work. The interface between the IWFand the SMSC is based on GSM 03.47description of SMS-MAP. The interfacebetween the SCFsl and the IWF is basedon INAP. Communication across net-work boundaries is supported by a SCF-SCF relationship.

3.3 CTM/GSM roaming

CTM and GSM are technologies withdifferent characteristics which could becombined to provide a more powerfulsolution. CTM focuses on being a fixednetwork solution for the support of cord-less mobility, which gives the user goodspeech quality, advanced services andhigh capacity. However, CTM couldexperience difficulties providing efficientcoverage in rural areas, as well as con-tinuing coverage in suburban areas. Onthe other hand, GSM is attractive becauseof its wide area coverage, both in Europeand in other parts of the world, as well assupporting a radio interface adapted tomobile environments.

In a scenario consisting of both CTM andGSM in combination, the two systemswould behave complementary to eachother and the scenario would appearvery attractive in the evolutionary pathtowards UMTS utilising the best of twoblessings.

The co-operation between GSM opera-tors and CTM operators represents apowerful combination as it is possible tocombine the two systems in the future.A possible way of integration is to com-bine the IN logic and database in theCTM network with the IN logic anddatabase in the GSM network, thusdeploying a common IN platform for thesupport of CTM/GSM. In EURESCOM

P606, several scenarios have been devel-oped, looking at solutions based on GSMnetworks without any IN capabilities aswell as GSM networks with IN capabili-ties.

3.4 Virtual Home Environment(VHE)

3.4.1 Background

Within telecommunications in generalthere is a shift in focus from standardisingservices to standardising service capa-bilities. This means that instead of havingstandards describing a service in detail,core features are being standardised.These features can be put together to forma wide spectrum of different services, thusallowing for product differentiationbetween competing service providers.

Service capabilities are also drivers fornew services within mobile networkssuch as CTM and GSM. The question iswhat to do when a user roams out of hisor her home network. To facilitate roam-ing and still keep your ‘home services’,the concept of virtual home environment(VHE) has been introduced.

3.4.2 VHE Implementationscenarios

To be able to support VHE the visitednetwork has to, as a minimum, supportcertain triggers. These triggers will forinstance be set off when there is a requestfor a change to the CTM user’s serviceprofile. Then, service control can be car-ried out in principle in one of two ways:

1 Control of the service is released to thehome network

2 Relevant data are sent from the homenetwork to the visited network so thatcontrol of the service can be carriedout from the visited network.

There are also a number of subsets ofthese two scenarios such that not onlycontrol of the service is released to thehome network, but also the connectionitself is being redirected via the homenetwork.

In general, the choice of implementationcomes back to which network should bein control; the home network or thevisited network.

Several aspects such as security andsignalling load have to be investigatedbefore a solution can be recommended.

SCUAF CUSF

SCFmm SCFsl

SDFmm SDFsl

mm = mobility managementsl = service IogicSDF = Service Data FunctionSCF = Service Control FunctionCUSF = Call Unrelated Service Function

SCUAF = Service Control User Agent FunctionINAP = Intelligent Network Application ProtocolSMS = Short Message ServiceMAP = Mobile Application PartSMSC = Short Message Service Centre

IWF SMSCINAP SMS-

MAP

Possible network boundary


3.5 Handover

3.5.1 Definitions

Handover is defined as the process ofswitching a call in progress from onephysical channel to another. This processis initiated when there is a degradation ofthe quality in the original channel causedby for example movement of the portableor some incidental radio disturbance.

Handover may be either internal or ex-ternal. Internal handover processes areinternal to one fixed termination only,whereas in the external case there is ahandover between two different fixedterminations. This means that externaldiffers from internal handover in thatexternal handover occurs between twoindependent systems, where each systemhas its own lower layers of protocols andan independent set of network layer ser-vice access points. For CTM the aim is tosupport some forms of external handoverby providing a common managemententity above the respective logical enti-ties in the two fixed terminations.

Figure 6 Handover scenarios. The fixed termination (FT) may represent a single radio base stationdirectly connected to the fixed network, or a cluster of base stations connected via a cluster controller

to the fixed network

3.5.2 Why implement external han-dover?

With the DECT systems on the markettoday it is possible to provide continuousoutdoor radio coverage with internal han-dover in a geographical area of about 1–2km2 in urban environments. However, ifeither indoor radio coverage or more traf-fic capacity is to be added, the area withinternal handover must be reduced corre-spondingly. Even though future DECTsystems may be capable of providinginternal handover in larger geographicalareas, external handover obviously stillrepresents a solution which is a lot moredesirable from the point of view of theCTM service provider as well as the net-work planner.

3.5.3 External handover scenarios

Figure 6 shows three different scenariosfor external handover, with increasingdegree of complexity;

• Scenario 1 – Intra-switch handover

• Scenario 2 – Inter-switch handoverwith the SSF/CCF in the respectiveswitches connected to the sameSCF/SDF

• Scenario 3 – Inter-switch handoverwith the SSF/CCF in the respectiveswitches connected to the differentSCF/SDF.

All of these scenarios have been investi-gated in the CTM work of EURESCOM.

4 CTM evolution towardsUMTS

UMTS (Universal Mobile Telecommuni-cations System) is the name of the ETSIinitiative within SMG (Special MobileGroup) to migrate today’s second genera-tion mobile system (GSM) towards thirdgeneration mobile communication. Theexisting work program suggests commer-cial operation in 2002.

There is at the present no exact definitionof what UMTS (Universal MobileTelecommunications System) really is.However, the basic idea behind UMTS isto have a personal telecommunicationsscenario where the user can access agiven set of services anywhere and atany time. The technology behind UMTSshould be irrelevant to the user, andcould be a mix of, among others, PSTN/ISDN, GSM, CTM, satellite, and so on.Such a scenario requires advanced inter-working procedures between differentnetworks, utilising the best from eachdifferent technology, to achieve thedesired integration of independentservices of today.

CTM will have a natural role within themigration toward UMTS, and togetherwith cellular and satellite systems sup-port multimedia applications any time,anywhere with a focus of attention onsupporting business and residential users’needs for high bitrate mobile communi-cation. To be able to support such ser-vices in an efficient manner, packet datacommunication will be important. Thismeans that in the evolution towardsUMTS, data services in CTM willbecome increasingly important involvingthe strong data communications possibil-ities within the DECT specifications.

Today, the maximum data rate supportedby DECT is 552 kbit/s. There is workgoing on within the DECT communityon refining the radio interface so thatdata rates up to 2 Mbit/s can be sup-ported. This will also be highly interest-ing in a UMTS environment. Otherimportant aspects of CTM in the migra-tion towards UMTS is higher speech

Scenario 1

SSF/CCF-1

FT-1 FT-2

SSF/CCF-2

FT-3 FT-4

SSF/CCF-3

FT-5 FT-6

SDF-2

SCF-2

SDF-1

SCF-1

Scenario 2 Scenario 3

CCF = Call Control FunctionSDF = Service Data FunctionSCF = Service Control Function

SSF = Service Switching FunctionFT = Fixed Termination


quality and a higher capacity, for ex-ample compared to today’s GSM net-works. This does not imply that CTMand GSM need to be competing tech-nologies. The co-operation betweenGSM operators and CTM operators rep-resents a powerful combination, becauseit is possible to deploy a common INplatform for the support of both CTMand GSM. The scenario could be en-hanced further by incorporating a newdedicated UMTS radio interface in addi-tion to CTM and GSM, or even by in-cluding a satellite component. Co-opera-tion between CTM operators and satelliteoperators would help the mixed scenarioto provide an even better coverage andtogether with GSM operators, offercommunication any time anywhere.

5 Summary andconcluding remarks

This article has focused on the technicalaspects of CTM, referring to work withinETSI and EURESCOM. It has also beenpointed out that CTM enables the fixednetwork operators to introduce servicesthat are highly attractive, and that CTMmight contribute to wiping out the linebetween mobile and fixed network com-munications as seen from the end user(Fixed Mobile Convergence – FMC). ACTM-like service has recently beenlaunched in Italy, and the next 1–2 yearswill bring some highly interesting feed-back on the market prospects for CTM.

6 References

1 Lauritzen, G O. Mobility applicationsintegration in Intelligent Networks.Telektronikk, 91 (4), 182–184, 1995.

2 ETSI. IN Architecture and Function-ality for the support of CTM. NA061302-1 version 7.0.0.

3 ETSI. IN Architecture and Function-ality for the support of CTM; Part 2:CTM IN interworking aspect. NA061302-2 version 1.1.10.

4 ETSI. IN Architecture and Function-ality for the support of CTM; Part 3:CTM Interworking between privatenetworks and public Intelligent Net-works. NA 061302-3 version 3.0.

5 ETSI. Cordless Terminal Mobility(CTM)-Phase 1; Service Description.DE/NA-010039.

6 ETSI. Cordless Terminal Mobility(CTM)-Phase 2; Service Description.DE/NA-010061.

7 EURESCOM P606 Deliverable 1.

8 DRAFT EURESCOM P606 Deliver-able 2.

Knut Baltzersen holds an M.Sc in physics andworks as Chief Engineer at Telenor R&D, Kjeller.He is currently involved in work on personalcommunications, with a special focus on valueadded services.


Ivar Oldervik worked as Research Scientist atTelenor R&D until October 1997 within areas likeIntelligent Networks, GSM Phase 2+/CAMEL, aswell as Cordless Terminal Mobility. He is nowemployed by Telenor Mobil working mainly withinthe area of Intelligent Networks.


Toril Natvig is Research Scientist at Telenor R&D,Kjeller, working in the field of mobile communicationwith special interests in network functionality forland mobile systems.


Gisle Pedersen was, at the time of writing thisarticle, working at Telenor R&D, Kjeller, in thefield of mobile and personal communications. He is currently employed by Telenor Mobil , wherehe is in charge of the GSM network developmentgroup.



Based on available equipment andstandards which are finalized or underpreparation, the potential of DECT(Digital Enhanced Cordless Telecom-munications) as a part of personalcommunication / third generation isexamined. Non-technical factors in-fluencing the future for DECT are alsoaddressed. This article mainly add-resses DECT as an access technologywhile the necessary support from thenetwork behind is not stressed.

Personal communicationscenario

The term personal communication isambiguous. In an evaluation of howDECT can be a part of personal commu-nication a common understanding of theterm is necessary. To gain this commonperception a scenario is described andalso some consequences of introducingpersonal communication are included.Two main trends are foreseen as indi-cated in Figure 1:

• Change from plain old telephony topersonal pocket terminals

• Development of pocket terminalsranging from only speech to multi-media capability.

In the Nordic countries with their highpenetration of cellulars the use of cellularphones for low bitrate services instead ofwired phones is emerging. Also resultsfrom a field trial performed by Telenorusing DECT to provide local mobility ina small town indicate a willingness toreplace the fixed phone with personalterminals. The trial customers statedclearly, however, that small pocketphones had to be personal. If a cordlessterminal should replace one fixed phone

in a household it gave little added valueas the terminal had to be kept stationary.In this trial we also experienced thatusers of mobile services always want anextended area of coverage.

Both existing and new operators havingcellular licences will compete with thefixed network services in a liberalisedtelecommunication market. Their ser-vices will to a large extent be based onthe use of pocket terminals and moveusers from wired phones to personalterminals.

The introduction of pocket phones led toa change in telecommunication behaviour.As long as only fixed phones are avail-able the attitude is that as soon as you getaccess to a phone, ie. when you get toyour desk or when you are back home,you have to remember to make that orthose calls. Sometimes you have to hurryto find a phone to make the call at anappropriate time. To deal with incomingcalls when you are not at your desk or athome, answering machines, switchboardoperators or call forwarding can help.

By carrying a pocket phone you canmake calls whenever you want almostwherever you are, but the circumstanceshave to be taken into account. The samegoes for incoming calls. In countries withhigh penetration of cellulars it can beobserved that people use the pocketphone in many different environments.If there is a need to make a call they do itthere and then. There is a discussiongoing about which situations or circum-stances use of the phone disturbs orannoys the surroundings and thereforeshould not be accepted.

For more and more people one personalpocket phone will be the basic telecom-munication terminal and use of wiredphones will gradually decrease.

Multimedia pocket terminalsreplacing personal phones

There is a general trend toward servicesbeing offered in the fixed network alsobeing wanted in mobile communications.Also the quality of the service should becomparable.

The emerging pattern of using cellularswhenever and wherever needed cancontribute to the creation of a marketdemand for advanced terminals capableof handling data, video and so on. Mak-ing and receiving calls and having accessto message services wherever you arewill not be sufficient. Users would like tohave access to the same services usingtheir pocket terminal as they have using afixed terminal, for example, being able toaccess Internet from a pocket terminalwill be a feature required by users. Adevelopment towards cellular handsetsbeing advanced PDA-like terminals withpersonal information stored indicates thatpersonal pocket terminals and therebyterminal mobility, will be an essentialpart of personal communication.

As more and more PC functionality isintroduced in terminals the stored infor-mation and user profiles will make theterminals an important part of everydaylife both at work and for private use. As aconsequence of this the same terminalsmay be used both in job situations andoutside working hours. Also the tendencyto work from different locations andtowards less differentiation betweenworking and non-working hours enhancethe need to use only one terminal.

When advanced terminals capable ofhandling multimedia-like services areavailable it can be assumed that userpatterns seen today among advancedcellular users will be even more apparent.If there is a need to have some kind ofinformation, to do a transaction, ordersome goods, watch a sports event orremotely control some device, you do itthere and then.

Given that personal communicationbased on pocket terminals becomes thecommon telecommunication subscrip-tion, new markets will emerge. People onthe move with their pocket terminals willoften be in need of peripheral equipmentlike printers, large screens and key-boards. From such stand-alone equip-ment which can be available at con-venient points for travellers, access tothe home server can be obtained using

DECT as a part of personal communicationJ O A R L Ø V S L E T T E N

F i le Ed i t View Element Type

Figure 1 Personal terminals replacing plain old telephones


pocket terminals. The connection can beobtained by use of a standard interfacebetween the pocket terminal and theequipment in use, eg. some kind of dock-ing station. Paper prints can be taken,documents or other files can be preparedor video watched. The invoice for the useof peripheral equipment and the connec-tion will be related to the pocket termi-nal, and only one invoice is received.

Demand for capacity

A differentiated demand for capacitydependent on location can be expected.A span in demanded bitrates is antici-pated to vary from relatively low bit ratesup to 2 Mbit/s.

Users are expected to use their pocketterminals for narrowband services like

speech wherever they are. But in someareas more than others, users will be in aposition to use their pocket terminals formore advanced services. This will mainlybe indoors or where some special activi-ties are taking place outdoors, eg.markets, outdoor restaurants, recreationareas, sports arenas.

Telephony

Evolutionaryservice

Images Data

Fax

ISDN

Video

Multiple services

DECT

Robust self plannedreal time radio

channel selection

Cost effective Coexistence

Mobility

Multiple accessrights

Seamlesshandover

Highcapacity

Inter-operability

Security Quality voice

Features

Residential

Business

Public

RLL

Multipleenvironments

PSTN

X.25D AMPS

LAN

NMT IEEE802

TACS

GSM

ISDN

Multiplenetwork access

Residential

FP

PP

Office

FP

PP

PP

PPs

WRSFP

PP

RLL/PCS

Multipleconfigurations

Figure 2 Overview of DECT applications and features (from [1])

FP = Fixed PartPP = Portable PartWRS = Wireless Relay

Station


Generally, people moving from one placeto another cannot be regarded as users ofadvanced services, amongst others due toreduced bit rate when on the move.

DECT applications andfunctionality

DECT is a flexible radio access technol-ogy that can support a wide range of ser-vices and applications to different kindsof networks as indicated in Figure 2. TheDECT standard is based upon TDMA/TDD and includes a mandatory real timeDynamic Channel Allocation whichmakes frequency planning unnecessary.It also secures that different uncoordi-nated DECT systems can coexist in thedesignated DECT frequency band withan efficient use of the frequency spec-trum.

A DECT handset can access differentnetworks, eg. public and private, andbase stations can be shared between dif-ferent operators. One part of a public net-work can host a private user group, or aprivate system can provide public accessfor visitors. The same base station canalso handle different services, eg. speechand data.

The DECT standard is prepared for thefuture by the provision of escape codesand a multitude of reserved codes andmessages in every layer of the standardso that evolutionary applications can bespecified.

A generic diagram of a DECT system isshown in Figure 3. The main standard,DECT CI (Common Interface), inprinciple covers the air interface betweenthe Fixed radio Termination (FT) and thePortable Part (PP). The CI can be re-

garded as a tool-box from which selec-tions are made when services and appli-cations shall be specified. DECT is trans-parent to services provided by the con-nected network and offers in itself onlycordless capability and mobility. TheInterworking Unit (IWU) which is net-work specific and the End System (ES),which may be a microphone, speaker,keyboard or display, are not parts of theDECT CI standard but are subject tospecific attachment requirements forthe chosen network to secure end-to-endcompatibility.

A DECT profile is a subset of the DECTCI that gives an unambiguous descriptionof the air interface for specified servicesand applications. In Figure 4 differentservices and applications based on DECTprofiles are indicated.

The profiles include all requirements forinteroperability of equipment from dif-ferent manufacturers.

The DECT profiles are:

• Generic Access Profile (GAP)The basic DECT profile for any appli-cation supporting 3.1 kHz telephony.

• Data ProfilesA family of profiles exploit the lowerlayer data services of DECT, speci-fically oriented towards LAN, multi-media and serial data capability. Eachmember of the family is optimised fora different kind of user services.Widely varying bandwidths are avail-able, and DECT can support net datathroughput of n x 24 kbit/s up to amaximum of 552 kbit/s.

• DECT/GSM Interworking Profile (GIP)Describes how DECT FP is connectedvia an IWU to the GSM networkwhich will then see a DECT user as aGSM subscriber.

• DECT/ISDN Interworking ProfilesTwo profiles are described:

The End System (ES) allows a suitablePP to access the ISDN network usingDECT signalling. The FP and the PPtogether appear to the ISDN networkas an ISDN terminal.

The Intermediate System (IS) providesfor a wireless link between an ISDNnetwork and ISDN terminals con-nected to an S interface. The ISDNterminals have transparent access toall network defined services basedupon the basic channel structure2B+D.

Network

PSTNISDNGSMPBX

:

IWU RC

DECT FT

DECTPT

ESCI

DECT FP DECT PP

Figure 3 A general DECT-system

Network

PSTNISDNGSMPBX

:

IWU RC

DECTInfrastructure

CI

DECTPT

ISDN

S-BUS

Multimediaterminal

Telephony32 Kbit/s ADPCM

Figure 4 DECT services and applications

CI : Common InterfaceIWU : Interworking UnitRC : Radio Controller or CCFP (Central Control Fixed Part)PT : Portable TerminationFT : Fixed TerminationFP : Fixed PartPP : Portable PartES : End System


• CTM Access Profile (CAP)The profile will secure interworkingwith CTM (Cordless Terminal Mobil-ity) networks. CTM will make widemobility in DECT networks possible.

• RLL Access Profile (RAP)The profile is divided into two parts:

Part 1, ‘Basic Services’, includes tele-phony services, a 64 kbit/s PCM bearerservice and over-the-air OA&M ser-vices.

Part 2, ‘Advanced Services’, specifiesISDN services (2B+D and 30B+D)and a data port for broadband packetdata services up to 552 kbit/s.

Also Wireless Relay Stations (WRS) arestandardised for DECT. These can beused to provide extended coverage whenuse of a base station is inappropriate.

In the ETSI Project DECT there is workgoing on

• to standardise advanced dual mode ter-minals DECT/GSM where both modescan be active at the same time

• to standardise implementation ofGPRS (General Packet Radio Service)over DECT

• to standardise 2 Mbit/s over DECT.

DECT’s potential as a partof personal communica-tion / third generationmobile

Deployment of DECT network

It is assumed that dual mode terminals,DECT/cellular, having one number andbeing capable of handling advanced ser-vices and applications are used. Offeringservices over such a terminal is a goodexample of FMC – fixed mobile con-vergence.

DECT access is provided as a public ser-vice in areas with high densities of userswanting to use their pocket terminals formore advanced services than speech.This will be mainly indoors, wherepeople stay for some time. DECT net-works may be a mixture of private net-works providing access for the publicand public networks. A part of the publicnetworks can be virtually private, eg.inside the premises of a company.

As DECT is not designed for velocitycoverage, passengers in cars or on publictransport may have a problem, eventhough trials have proved that goodspeech quality can be obtained in carsexceeding 80 km/h.

In rural areas DECT access can be pro-vided indoors by private DECT systems.There will be a multitude of DECTislands and when moving between theseislands the cellular mode of the pocketterminal has to be used. In public DECTislands any DECT terminal can be used.The standard opens for private systems toappear as public systems to visitingusers. Thus, the DECT mode of the ter-minal can in principle be used whereverDECT coverage is provided.

Services

When DECT provides speech 32 kbit/sADPCM is used and the quality is nearthe quality of wired phones. A stand-alone DECT system provides seamlessmobility in an area covered by base sta-tions connected to one radio controller.With the implementation of CTM in thefixed network handover between basestations belonging to different radio con-trollers is also obtained. If a DECT sys-tem is connected to a cellular network themobility management is handled by thecellular network.

One terminal can have access rights toseveral different kinds of DECT net-works and roam between them. The pro-file standards will ensure interoperabilitybetween equipment from different manu-facturers. For example, if a terminal iscapable of handling voice and a specificdata profile used for Internet access,GAP will ensure that the terminal can beused towards any DECT system withvoice capability. The data profile securesaccess to any system with the specificdata profile implemented, independent ofmanufacturers.

ISDN solutions for DECT are on themarket. Data profile standards for552 kbit/s over DECT exist and there iswork in progress to standardise 2 Mbit/s.Hence, DECT has a potential to handlemultimedia-like services.

However, there will be limitations in thefunctionality of pocket terminals basedon any technology. Terminals should besuited to being carried, eg. in a shirtpocket, hence the size has to be limited.

User interfaces like screen and keyboardwill suffer. Even if solutions like speechrecognition may help to overcome suchproblems, one can expect the user friend-liness of stationary terminals to be better.Especially the need for a larger screenthan can be implemented in the pocketterminals will be apparent when graphicsor video are presented.

Capacity

A main problem when offering servicesover radio is the limited bandwidth dueto limitation in available frequencyresources. If use of a personal terminalcapable of handling multimedia-like ser-vices is becoming the basic telecommu-nication utilisation for each individualthere will be a bandwidth problem usingexisting cellular technologies.

DECT may help to solve a capacityproblem. Due to short range and thedynamic channel selection feature DECTis capable of handling advanced servicesin areas with high user densities.

An uncertainty regarding DECT andcapacity is that a multitude of DECTsystems, both private and public,supporting different kinds of services cancoexist, eg. in the same building. Thiscan make the planning of a DECT systemregarding traffic capacity somewhattricky.

Factors influencingDECT’s success as part ofthird generation

DECT products on the market

The DECT CI standard was finalised in1992, but DECT has still not ‘taken off’.One reason may be that it took some timebefore the DECT profile standards werein place. It seems that DECT residentialsolutions are popular and DECT businessapplications has an increasing marketshare. Products for these applications aremade by many manufacturers, and smallDECT systems handling ISDN are nowavailable for residential and businessapplications.

Lately, there has been an increasinginterest in DECT as RLL solutions.World-wide there is a huge demand forsubscriber lines, and DECT as a stan-dardised access technology is an interest-


ing choice for the access network. Coun-tries with poor telecommunication infras-tructure is the main market for DECTRLL, and mainly fixed telephony is pro-vided.

Only a few manufacturers are offeringequipment for public DECT networksincluding mobility. To provide widemobility in DECT networks a supportfrom the network behind is required. ForPSTN/ISDN this means a CTM solution.A standard for CTM is expected to beready in the near future.

Time gap for DECT

DECT CTM solutions and the GSM/DCS1800 might be competing techno-logies. DECT has the greatest potentialfor third generation like services andapplications, thus an advantage overGSM/DCS1800. However, in 2002UMTS products are expected to be avail-able, and one important question is if thistime gap is sufficient for building DECTCTM networks to introduce advancedservices. Also, one must bear in mindthat the GSM technology includingDCS1800 will be enhanced to handlemore advanced services in the sameperiod, eg. GPRS (General Packet RadioServices). The bitrate of the GPRS tech-nology will be approximately 115 kbit/swhich is well below the capability ofDECT.

If DECT is going to be part of third gen-eration solutions advanced products haveto achieve a certain penetration beforecompeting, advanced GSM and UMTSproducts are available. The time limit foradvanced DECT products to penetratethe market can be discussed. However,there is no doubt that roll-out has tohappen soon if DECT shall be a partof third generation.

Also, there is a time gap for substantiallylower pricing of services based on DECTcompared to GSM, as the cost of usingmobile services is steadily falling. How-ever, advanced GSM and UMTS equip-ment and services can be expected tohave a relatively high price in an earlyphase.

DECT and operators

There has been an interest among Euro-pean operators for DECT CTM solutions,and Telecom Italia has built DECT net-works in 28 cities mainly providing out-door coverage. The networks are basedon proprietary CTM solutions. The ser-vice was opened on 1 January 1998, andthe experiences made by Telecom Italiacan have an important impact on DECT’sfuture as a part of public networks.

Operators that started public DECT trials3 – 4 years ago were quite enthusiastic,but time has passed without flexiblesolutions for public DECT becomingavailable. Hence, the interest has fadedamong many of them.

It seems that several operators who pre-viously had an interest in public DECTare going for other technologies likeDCS1800, but some still go for DECT.

Manufacturers and DECT

Manufacturers making DECT equipmentalso make GSM/DCS1800, a technologywhich is a commercial success.

There is still a demand from operatorsand the corporate market for public andlarge business systems of handlingadvanced multimedia-like services, ie.utilizing the full potential of DECT. Ifthis demand is sufficient to convince themanufacturers there is an increasedpotential for profit when advanced DECT

equipment is included in the productportfolio. There is therefore reason tobelieve that DECT will be part of a thirdgeneration solution.

A general impression is that manu-facturers have been reluctant to imple-ment functionality beyond speech intheir DECT products.

It seems unlikely that manufacturersbeing big in the GSM industry will pro-mote or be drivers for advanced DECTsolutions at their own risk. They find thesafest commercial way ahead towards3rd generation to be enhanced GSMtechnology and UMTS.

There will be many DECT products forthe residential and business market, butthe most interesting application seems tobe DECT RLL. When operators imple-menting DECT RLL ask for more ad-vanced services the market can be suffi-cient for manufacturers to come up withadvanced solutions.

Conclusions

Technically, DECT has a potential tobecome a part of third generation.Especially dual mode DECT/GSMsolutions capable of handling multi-media-like services would have qualitiessuited for being part of a personal com-munication platform.

However, it seems that other factors thantechnical ones will influence the future ofadvanced DECT networks. There is atwo-step time gap for advanced DECT,first before enhanced GSM solutionsbecome available and then before UMTSenters the market. DECT needs a certainpenetration before this happens. For thetime being it seems that the process ofimplementing advanced functionality inDECT products is rather slow. If DECTis going to be a part of personal commu-nication, the manufacture of advancedDECT products and implementation ofthese into DECT systems and networkshave to happen in the near future.

Reference

1 ETSI. ETR 178 First edition: RadioEquipment and Systems (RES); Digi-tal European Cordless Telecommuni-cations (DECT); A high level guideto the DECT standardisation. ETSI,Sophia Antipolis, October 1995.

Joar Løvsletten has been with Telenor sincegraduating from the Norwegian University ofScience and Technology in 1976. He worked in theNetwork Department, dealing with radio relay sys-tems, until 1990, when he started a group workngon radio access systems. Løvsletten came toTelenor R&D in 1995 and is now Senior Engineerworking with personal communications.


1 Introduction andhistorical perspective

Mobile radio has given the Europeanindustry its most important success in thelast decades. Crucial to this success werethe standards developed by ETSI (Euro-pean Telecommunications StandardsInstitute) that have been adopted not onlyby European countries, but by many oth-ers as well. The best-known example isGSM (Global System for Mobile com-munications), the most widespread sys-tem for cellular communications. It hasreplaced a multitude of analogue stan-dards throughout Europe, and seems tobecome the de-facto world-wide standardfor cellular communications. The DECT(Digital Enhanced Cordless Telecommu-nications) standard [1] promises tobecome a similar success story. It wasoriginally designed in 1991 in order toreplace the several analogue standardsfor cordless telephones that existed inEurope at that time. The first commercialproducts went on sale in 1994, whenabout half a million terminals were sold.In 1997, about 8 million terminals werein use, and this number is expected toincrease to 30 million by the year 2000.

Since the early 1990s, the market forDECT equipment has increased bothgeographically and in scope. Geographi-cally, many extra-European countrieshave accepted the DECT standard (some-times with slight modifications); this isalso reflected in the change of the inter-pretation of DECT: while the ‘E’ inDECT was originally translated asEuropean, this has been changed toEnhanced, to do away with any geo-graphical restrictions. Another importantchange has occurred in the scope of cord-less system. In the analogue days, ‘cord-less’ meant just that the microphone andearpiece were attached to the ‘base sta-tion’ (i.e. the usual phone), not by wirebut by a radio link. With the DECT sys-tem, many new applications becamepossible: one base station can serve sev-eral mobile terminals, and it is possibleto combine several base stations to a ‘cel-lular’ system, where the mobile terminalcan connect to various base stations dur-ing a talk and make handovers. Thisallows companies to build up a wirelessprivate branch-exchange system (PBX),for example. When we take this featureto its extreme and combine it with capa-bilities of wired intelligent networks, italso becomes possible to build a low-mobility cellular system based on DECT.

Field trials for such systems have beendone in Norway [2] and other countries.Finally, DECT is intended to be used inwireless-local-loop (WLL) applications.There, the DECT link replaces the high-installation-cost connection from the sub-scriber to a central office (CO), or, e.g.,to a village centre connected by ordinarycable to the CO.

The new applications of DECT pose newchallenges both to the network and to theradio link. The networking aspects havebeen treated e.g. in [3] and another con-tribution in this special issue [4]. Thepresent paper will concentrate on theradio link. The quality of the transmis-sion can be limited either by noise, by thetime dispersion of the mobile radio chan-nel, or by co- and adjacent channel inter-ference. Various methods have been pro-posed to mitigate those effects, and wewill discuss them below. Section 2 givesthe specifications of the air interface asprescribed by the DECT standard. Sec-tion 3 describes the performance in noiseand co-channel interference. The effectsof the time dispersion are analyzed inSection 4. The next two sections discussdiversity and equalizers, two means toovercome the basic limitations on theperformance. Section 7 compares DECTto the rival system PHS from Japan.The next two sections deal with practicalimplementation issues: Section 8 de-scribes a testbed for assessment of theperformance of DECT terminals in vari-ous environments, and Section 9 com-pares the popular heterodyne receiver to

more advanced direct-conversion re-ceivers. A summary wraps up this paper.

2 The DECT air interfacespecifications

DECT, like GSM, uses a combination ofTDMA and FDMA for multiple access.The 20 MHz band that has been allocatedto DECT (1880 – 1900 MHz) is dividedinto 10 frequency bands; the carrierfrequencies are actually separated by1.728 MHz. On each carrier, 12 users canbe accommodated in a TDMA scheme:the time axis is divided into frameswhich last 10 ms each, and are dividedinto 24 timeslots. Each user is assignedtwo timeslots that are separated by 5 ms:the first timeslot is used for the down-link, the second slot for the uplink. Thisimplies that time division duplexing isused for separating uplink and downlink(in contrast to GSM, where frequencydivision duplexing is applied). We willsee later that this has important conse-quences for diversity, and also for smart-antenna applications. The timeslots areseparated from each other by guardtimes, which are necessary, since the run-times from the different terminals to thebase station may be different. The 52 µsguard time prescribed in DECT cantheoretically deal with a 15 km distancebetween base station and mobile terminal– this is far more than actually needed;even when directional antennas are used,the ‘cell radius’ is usually less than 3 km.The rest of the guard time is used to give

45

The DECT radio linkA N D R E A S F . M O L I S C H , H E I N Z N O V A K A N D E R N S T B O N E K

Telektronikk 2.1998

DECT-Frame, 11520 bit (=10ms)

Fix → Portable Portable → Fix

0 1 11 12 22 23

Full Slot, 480 bit (=417us)

S − Field D − F i e l d

Guard-Time,60 bit

(52 us)

Pre-ample,16 bit

(14 us)

PacketSynchWord,16 bit

(14 us)

SignalisationA-Field64 bit

(56 us)

DataB-Field324 bit

(281 us)

Figure 1 DECT frame and DECT slot structures

the transmitter amplifiers time to powerup – the faster these amplifiers have tobe, the more expensive they become.

Each timeslot begins with a 16 bit pre-amble and a 16 bit packet synchroniza-tion word (the 32 bit S-field). This isfollowed by the D-field. This D-fieldcontains 64 bits of signalling (A-field)and 324 bits of data. The A-field includesa Cyclic Redundancy Check (CRC) forthe detection of packet reception failures;it allows transmission of signalling in-formation with a net rate of 4.8 kbit/s.Finally, the B-field contains 324 bitsof data, enabling the transmission of32 kbit/s ADPCM coded speech in oneslot. The overall transmission data rate istherefore 1.152 Mbit/s. This frame/slotstructure suggests that DECT terminalscan be designed to be far less complex,and thus less expensive, than GSM.DECT does not use a complicated speechcoder, which explains the fact that itrequires 32 kbit/s instead of the 6.5 or13 kbit/s prescribed by GSM. It also con-tains no channel coding for the speechdata, so that the bit error rate (BER) for

tolerable transmission quality has therather low value of 10-3 (as compared to10-1 for GSM). By combining two orseveral timeslots, DECT can providemuch higher bitrates, up to 552 kbit/s,and carry asymmetric traffic, makingDECT attractive for data transmission.The modulation format is GFSK (Gaus-sian Frequency Shift Keying) with nomi-nal modulation index hmod = 0.5, whichis equivalent to Gaussian Minimum ShiftKeying. However, the tolerances in themodulation index are quite large.

The time-bandwidth product of the Gaus-sian filter BGT is 0.5, where BG is thebandwidth of the Gaussian filter, and T isthe symbol period. This provides a com-promise between emitted spectrum andintersymbol interference (see Figure 2).The resulting RF-channel bandwidth isabout 1.56 MHz (for a 99 % powercriterion). With the specified channelspacing of 1.728 MHz, adjacent channelemission up to -40 dBc is allowed. Table1 shows a summary of important DECTparameters.

3 Performance in noiseand co-channel inter-ference

The structure of the receiver is not speci-fied in the standard. Three realizationshave been at the centre of attention:coherent reception, differential phasedetection, and limiter-discriminatordetection. Coherent reception is difficultto implement, because the tolerances forthe transmitted signal as foreseen in thestandard are very large, so that carrier

recovery and tracking is difficult. In dif-ferential phase detection, we compute thedifference of the phase between two sub-sequent samples that are separated byone bit length. It is thus not necessary torecover the carrier phase, because anyoffset (that remains unchanged duringone bit duration) will be eliminated whenwe take the difference of two phases.Finally, limiter-discriminator detectionuses the fact that MSK can be seen as akind of frequency-shift keying: we justdetect the instantaneous frequency withan FM/AM converter and a thresholddetector.

Extensive investigations have been doneto find out the performance of MSK in aflat Rayleigh-fading channel with noise.Adachi and Parsons [5] found that fordifferential phase detection and for fre-quency-discriminator detection, the errorprobability is about 0.7/SNR (where thesignal-to-noise ratio SNR is defined asbit energy over noise power density)when the receiver filter bandwidth ischosen in an optimum way. For differen-tial phase detection, the optimum filterbandwidth is about BrT = 0.75, forlimiter-discriminator detection,BrT = 0.55.

Another problem is co-channel inter-ference (CCI). For a system like DECT,cell planning in advance of operation isneither feasible nor desirable. The capac-ity would be exploited very inefficientlyand a complete redesign of the cell struc-ture would be necessary every timesomeone installs a new base station. So-called dynamic channel allocation avoidsthese problems. The mobile part listenson the time-slots it does not use whetherthere is a channel with low CCI avail-able, and moves the user there if there isa need to do so. Also, carrier frequencychanges are possible. Of course, such ascheme cannot eliminate all CCI, butonly minimize it. It is common to modelthe CCI (and also adjacent channel inter-ference ACI) as Gaussian (see e.g. [6]),so that it affects the BER in the sameway that noise does; the BER is about0.7/CIR, where CIR is the carrier-to-interference ratio. For CCI, the CIR isnot influenced by the receiver filter,while the ACI is influenced.

4 Performance in time-dispersive channels

Mobile radio channels are not only time-variant, i.e. fading, but also time-dis-


-4 -2 0 2 4

0

-20

-40

(f-f0)T

S(f)/dB

Figure 2 Spectrum of a DECT signal

Transmission format FDM/TDD/TDMA

Frequency band 1880 – 1900 MHz

Centre frequency for channel N (0..9) 1 897 344 – N * 1728 kHz

Number of frequency channels 10

Number of users per frequency channel 12

Channel spacing 1.728 MHz

Frame length (24 slots) 10 ms

Data rate 1.152 Mbit/s

Modulation method GFSK, BGT = 0.5

Tolerated bit error rate < 10-3

Table 1 Important parameters of the DECT standard

persive [7]. When we send a delta pulsefrom the transmitter, we will receive aseries of delta pulses with different delaytimes, attenuation factors, and phaseshifts. Two quantities characterize thistime dispersion in a rough approxima-tion: the maximum excess delay, i.e. thedifference between the delays of the firstand the last echo, and the rms delayspread, i.e. the square root of the secondcentral moment of the square of theabsolute value of the impulse response.At the receiver, the different echoes are‘smeared’ by filtering and the integrationprocedures inherent in the receiver. Ifnow the rms delay spread is very muchsmaller than the bit length, then thereceiver does not ‘notice’ the time dis-persion of the channel, and we can de-scribe the channel as ‘flat fading’, theway we did in the previous section. Ifthis condition is not fulfilled, however,then the time dispersion leads to inter-symbol interference, and thus to errorsthat cannot be eliminated by increasingtransmitter power. These errors are thuscalled ‘irreducible errors’, although thisname must not be misinterpreted – theycan actually be reduced by equalizers ordiversity. The BER associated with theseerrors is called ‘error floor’.

The time dispersion of the channel alsomakes the synchronization much moredifficult. In FDMA systems, the usualway to synchronize a receiver is to per-form an eye-pattern analysis, and put thesampling time at the time instant wherethe eye has its widest opening. Changesin the optimum sampling time are moni-tored and adjusted by a control loop witha rather low bandwidth (on the order ofthe Doppler frequency, say, 100 Hz orless). However, this type of synchroniza-tion is both difficult and inefficient inTDMA systems with preambles. Thedesign of the S-field in DECT rathersuggests a different approach, whichworks in three steps:

(i) The ‘power up’ ramp for the rele-vant slot tells the receiver veryroughly (within several bits) that anew timeslot is about to start.

(ii) The sampling time is determinedfrom the preamble (i.e. the first 16bits of the S-field). The preambleconsists of alternating +1/-1 infor-mation bits. This means that thereceived phase difference has theshapes depicted in Figure 3. Thesampling time is then put into themiddle between two zero-crossingsof the received phase difference.

(iii) The word (or packet) synchroniza-tion is done with the help of thepacket synchronization word, i.e.bits 17–32 of the S-field. Due tothe group delay in a time-dispersivechannel, the received sequence canbe shifted effectively by one bit for-ward or backward (as compared to aflat-fading channel). By comparingthe (known) packet synchronizationword with the received sequence, wecan detect such a shift, and adjustthe receiver timing appropriately.

Actually, the synchronization proceduredescribed above is not optimum. Im-provements have been proposed by us[8]; the additional gain in BER is up to afactor of 4. It is important to note thatthis burst-by-burst synchronization (alsocalled ‘adaptive sampling’ or ‘training-sequence-based determination of thesampling time’) differs fundamentallyfrom the pure eye-pattern analysis men-tioned above. Eye-pattern analysis is atype of ‘blind’ synchronization, i.e. canbe used when we have no knownsequence in our frame structure. It is thusmore general, but gives worse results.For DECT, however, the knowledge ofthe S-field should be used to improveperformance.


b) Two-delay channel withdestructive interference.

a) Two-delay channel withconstructive interference.

Figure 3 ‘Eye pattern’ (decisionvariable as function of the sampling

time) of the preamble in channels withweak and strong group delay distortions

10-2

10-3

10-4

10-5

fixed sampling, experiments[Crohn et. al. 1993]

adaptive sampling,experiments[Schultes et. al. 1994]

fixed

sam

pling

GMSK, B

T=0.50

0

GMSK, B

T=0.62

5

GM

SK, BT=

0.75

0G

MSK, n

o fil

ter

MS

K,

no f

ilter

adap

tive

sam

plin

gBit

erro

r pr

obab

ility

.01 0.1 1

Figure 4 Error floor as a function of the rms delay spread for unfilteredand filtered MSK

Delay spread normalised to bit interval (S/T)

For GMSK without any receiver filter-ing, the error floor obtained with (opti-mum) adaptive sampling is K • (S/T)2,where K is about 10-3. Any receiverfiltering increases K: for BrT = 0.625,K is about 0.05. Figure 4 shows the errorfloor for various filter widths. This isalso an important design parameter forDECT systems: we have seen above thata small receiver bandwidth decreases theeffects of noise and adjacent channelinterference, while it has little influencein the co-channel interference perfor-mance. Now we see that it increases theeffects of the channel time dispersion.For various environments, differentreceiver bandwidths will be optimum.This suggests the implementation of vari-ous filters in the receiver, with the possi-bility to (automatically) choose the opti-mum. Of course, such a strategy is onlyfeasible if the filters are implementeddigitally.

In a first approximation, errors due tonoise, CCI and time dispersion are in-dependent and their associated BERs canbe simply added to give the total BER.This is not strictly true, because thegroup delay distortions (i.e. what causesISI-induced errors) are strongest in thefading dips, where also the noise has itsstrongest effect. However, these are al-ready rather fine points that need not beconsidered in a first system design.

5 Diversity

DECT demands a BER of 10-3 withoutchannel coding in order to give accept-able speech quality. As we have seen, thenoise-induced BER is about 1/SNR in aRayleigh-fading environment, so that theaverage SNIR must be better than 30 dBin order to fulfil this requirement on theaverage. Furthermore, the indoor en-vironment is often stationary: when auser is sitting in a fading dip (i.e. a pointof low received signal strength), he willnot move out of it for the duration of thewhole conversation. These problems canbe greatly reduced by the application ofdiversity. Usually, DECT systems havetwo antennas at the base station, and weselect the signal of one of them or com-bine them. Furthermore, we can alsohave two antennas at the mobile terminal.In the following, we will first considerthe uplink, with one antenna at themobile and two antennas at the basestation. Another configuration will betreated later.

The simplest approach to using thesignals from antenna branches A and Bis the so-called switching diversity.When the transmission starts, we alwaysuse branch A. If the transmission qualityachieved with that branch is not accep-table (i.e. either the received signalstrength is too low, or the CRC check inthe A-field tells us that we have a toolarge BER), then we use antenna B forthe reception of the next timeslot. If thereception quality achieved with thatantenna is also insufficient, we switchback to branch A, and so on. The ad-vantage of this approach is that we needonly one additional antenna and oneswitch compared to the no-diversity case.This is offset, however, by two seriousdrawbacks: first, we do not use con-sistently the ‘better’ signal: if, for ex-ample, both antennas give insufficienttransmission quality, then we might beswitching back and forth between a badand an even worse signal. This alsomakes the definition of the threshold(what is ‘sufficient transmission qual-ity’?) difficult. If the threshold is toohigh, then we are constantly switchingback and forth; if it is too low, then weaccept signals that sound rather bad tothe user. The second drawback of theswitching diversity is associated with theinherent time delay. The switching be-comes effective only 10 ms after themeasurement. In the meantime, the usermay have moved so much that the signalfrom the branch that was formerly bad isnow good, and vice versa.

A more involved approach is selectiondiversity. In this case, we monitor thequality of the signals from both diversitybranches, either the received signalstrength, or the BER, and choose thebranch that gives the better results.This avoids all problems with defining athreshold, and it also becomes possibleto do ‘instantaneous switching’. Thedrawback of this method is that we needmore hardware. For RSSI-driven diver-sity (received-signal-strength-indicator),we additionally need (compared to theswitched-diversity case) two sensors forthe instantaneous power. For BER-drivendiversity, we need additionally a com-plete receiver (we have to receive anddemodulate the signals from bothbranches). Finally, we could also com-bine the signals from the two branches(not just select from them). However,this approach gives little improvementover the selection diversity.

An additional requirement for the diver-sity to work properly is that the signals atthe two antenna branches are uncorre-lated. If the signals are impinging uponthe base station from all directions, thisis not a big problem: we just require thatthe antennas are spaced a fraction of awavelength, e.g. 4 cm – this can easilybe done in a base station. If, however,the directions-of-arrival (DOAs) of theincoming waves are all concentratedwithin a small range, then the necessarydistance between the antennas increasestremendously. Such concentrated DOAsusually occur only if the base station issituated above the rooftops – a situationthat does not occur in the usual cordlessapplications, but which might becomerelevant for WLL links.

Up to now, we have only consideredthe uplink, but of course we need goodspeech quality also on the downlink. Ifwe have only one antenna on the mobileterminal, then we must use the basestation antennas in such a way that thetransmission quality for the downlink isenhanced. This is facilitated by the factthat up- and downlink are on the samefrequency. In a stationary environment,we can therefore say that if it is better touse the signal from antenna A in the up-link, it is also better to transmit fromantenna A in the downlink (reciprocitytheorem). Note that this is not true inGSM, because it uses frequency divisionduplexing, so that the channels for uplinkand downlink are in principle always dif-ferent. Problems for the branch selectionin DECT can only occur if the channelchanges strongly within 5 ms, the timebetween the uplink and downlink trans-missions. Such a quick change in thechannel can for instance occur in the caseof fast-moving scatterers (cars passingby) – the mobile terminal is assumed tobe slow-moving anyway. An alternativewould be to install two antennas also atthe mobile terminal. As mentionedabove, the antennas have to be spacedonly a few centimetres apart, and even ifthe distance is smaller, inherent patterndiversity usually leads to good decorrela-tion of the received signals.

Using diversity considerably improvesthe performance in noisy or interference-limited environments. The required SNR(for 10-3 BER) is decreased by some10 dB, depending on the exact diversityscheme, if two branches are used.Adding a third branch, however, de-creases the required SNR only by anadditional 3 dB. This is why only two


diversity branches are used in mostequipment. Diversity also decreases theeffects of time dispersion, see Figure 5(note that these are theoretical errorprobabilities under ‘ideal’ circumstances;measured BERs in actual systems aresomewhat higher). However, the toler-able delay spread is still limited; even inthe best implementations, it must remainbelow about 300–400 ns. For environ-ments with larger time dispersion, al-ternative methods are required, namelyequalizers, which will be treated in thenext section.

6 Equalizers

The DECT specifications do not foreseean equalizer, because its implementationwas deemed to be too expensive at thetime the specifications were written.However, for some new applications,delay spreads of 500 ns or larger occur,which cannot be dealt with by ‘normal’receivers even when diversity is applied.Thus, an equalizer is required. Since noequalizer was foreseen in the standard,one might think that a ‘blind’ equaliza-tion (adjusting the equalizer coefficientswithout knowledge of either the channelor the data) is required. Fortunately,however, we can use the simpler andmore efficient ‘training-sequence-based’equalization by employing the (known)S-field as a training sequence. There iseven a subsequence in the S-field (bits17–27) that has very good autocorrela-tion properties (the sidelobes have onlyheight 2). This sequence can thus be usedfor measuring the impulse response ofthe channel, and for the training of theequalizer.

Essentially, two types of equalizer havebeen proposed for DECT: Viterbiequalizers [10] [11] [12], and DFEs(decision feedback equalizers) [13]. Ina Viterbi equalizer, we first measure theimpulse response of the channel. Wethen determine for each possible inputsequence what the output from thischannel would be. When the data arereceived, we analyze what data sequencehas been most probably sent. One im-portant parameter for such an equalizeris the so-called ‘constraint length’, i.e.the anticipated length of the impulseresponse of the channel. If we anticipatelong impulse responses, then the equal-izer becomes very complex. On the otherhand, if the anticipated length is shorterthan the actual length of the impulseresponse, then the results will become

inaccurate. Up to now, only very shortconstraint lengths have been proposed(two-state Viterbi equalizers), whichwere shown by simulations to handledelay spreads up to about 500 ns.

In a DFE, we have essentially a (non-lin-ear) filter, i.e. a tapped delay line with afeedback section. The filter coefficientsare adjusted in such a way that during thetraining phase, the mean squared errorbetween (known) training sequence andthe actual received sequence is mini-mized. Also in this equalizer, we have atrade-off between complexity and per-formance. One implementation proposedin [13] is a DFE with 3 taps in the feedforward and 2 taps in the feed backwardsection; its performance (simulated) isshown in Figure 6.

Equalizers have a twofold beneficialeffect in time-dispersive channels. Theequalizer does away with the error floor,indeed. On the other hand, they alsoreduce the noise-induced errors in suchenvironments. Some time dispersion inthe channel will actually lower the BERas compared to the non-dispersive case.This is due to the fact that the time dis-persion provides inherent time diversity,which can be used to combat fading.Echoes that arrive at the receiver at dif-ferent times suffer from different fading:


Ave

rage

BE

R

.01 0.1 1

Normalized delay spread S/T

10-2

10-3

10-4

10-5

10-6

10-7

10-8

10-9

no diversity

RSSI-driv

en d

ivers

ity

BER-driv

en d

ivers

ityFigure 5 Reduction of error floor by

diversity (from Ref. [9])

Bit

erro

r pr

obab

ility

- B

ER

0

Signal-to-noise ratio dB

1

10-1

10-2

10-3

10-4

10-5

10-6

S/Tbit=0 (flat fading)S/Tbit=0.1 without equalizerS/Tbit=0.1 with (3,2) DFES/Tbit=0.2 without equalizerS/Tbit=0.2 with (3,2) DFE

non - fading

10 20 30 40 50

Figure 6 BER as a function of the SNR for a (3,2) DFE for various delayspreads. From [7]

it is very improbable that the receiver isin a fading dip with respect to all the dif-ferent paths from transmitter to receiver.Thus, the high error probabilities in thefading dips are eliminated. This is alsoreflected in Figure 6. As the time disper-sion increases, the performance with theequalizer becomes more similar to theperformance of ‘non-fading’.

7 The competition: PHS

While the DECT has found wide accep-tance in many countries, Japan has de-vised its own rival standard, which alsohas achieved considerable success. Thisstandard is called PHS (Personal Handy-phone System), and is mostly identical toDECT in its system design. It is also anFDMA/TDMA/TDD system, and uses32 kbit/s ADPCM without error correc-tion. However, there are some importantdifferences in the radio link: the modula-tion format is π/4-DQPSK, so that twobits are sent with each symbol. The pulseshapes used are pulses with a square-rootraised-cosine spectrum with α = 0.5.There are only eight timeslots (fourusers) per carrier frequency. The radiochannels are only 300 kHz wide, the bitrate is 384 kbit/s.

The change in the modulation formatalso leads to important changes in theperformance of the radio link. The sensi-tivity of π/4-DQPSK to noise in aRayleigh-fading channel is about thesame as for GMSK for the same Eb/No.For co-channel interference, π/4-DQPSKperforms worse by about 3 dB. The biterror rate due to time dispersion of themobile radio channel is about 0.5(S/Tsymbol)2. With the data rates used in

PHS, this implies that a delay spread of230 ns is admissible. This is quite similarto the values that occur for the DECTsystem; there, we had between 120 ns(measured in a suboptimum system) and250 ns (computed for an optimum sys-tem). At first glance, this similarity isastonishing: PHS uses a symbol durationthat is larger by a factor 6 than in DECT,so that we would expect a much smallersensitivity to time dispersion. However,it can be shown that π/4-DQPSK has aninherently larger susceptibility to timedispersion errors, because adaptive sam-pling does not reduce ISI-errors asstrongly as it does in GMSK [14].

The major advantage of PHS is the largespectral efficiency. While DECT can put120 users into 20 MHz, PHS has 300users in 23 MHz. This is caused mainlyby the fact that PHS uses a four-levelmodulation format, which has inherentlyhigher spectral efficiency. On the otherhand, the higher sensitivity to co-channelinterference requires a larger separationbetween base stations; this reduces thecapacity (compared to DECT) by a factorthat lies between 1.4 and 2, so that from apure ‘radio link’ point of view, thecapacity of a DECT system should bevery similar to that of a PHS system. Fur-thermore, DECT has some capacityadvantages due to various system aspects,like handovers.

In Japan, the 77 frequency channels havebeen divided into the 37 lower channelsthat are used for home and office use (i.e.real cordless and PBX applications) andthe 40 upper channels, which are used fora low-mobility PCS system. This systemhas been very successful in Japan; italready has several million users. On

the other hand, this system has alsoshown the enormous investments that arenecessary to build up such a system:the metropolitan area of Tokyo is nowcovered by several hundred thousandbase stations. We finally note that in theUSA, there is the so-called PWT systemthat combines the system specificationsof DECT (including multiple-accessspecifications) with the π/4-DQPSKmodulation format; this results in a per-formance that is very similar to DECT.

In any case, the competition betweenDECT and PHS will probably not bedecided in the technical sector, but byregulatory and marketing aspects. Willthe Europeans be ready to provide thespectrum from 1900–1920 MHz for asystem competing with DECT, and willthe Japanese provide the spectrum from1880–1900 MHz?

8 Testing the radio link:the testbed of TU Wien

While most research into the radio linkperformance of DECT is done by simu-lations on the computer, there is no sub-stitute for real measurements. The perfor-mance of hardware can be best assessedin a testbed, where certain prescribed testprocedures can be done under controlledcircumstances. At the TU Wien wedesigned and built such a testbed for theevaluation of the DECT standard’s physi-cal and medium access layer specifica-tions and to find the limitations of theDECT link for indoor and outdoor appli-cations especially in dispersive radiochannels.

The first generation of the testbed imple-mented the TDMA and FDMA structureof DECT, and was restricted to a simplexlink [15]. The basic arrangement of thetestbed is shown in Figure 7. It consistsof a single direction DECT RF-link, setup by one desired transmitter (TX1), oneinterfering transmitter (TX2), the re-ceiver (DRX), and an indoor channelsimulator (CAN). The return link isestablished by a wired low data ratesignalization connection from the re-ceiver via a controlling personal com-puter (DTC) back to the desired trans-mitter (TX1). Transmitted data, receiveddata and timing as well as synchroniza-tion information is transported by a highdata rate connection to the error counter(ERC), which calculates the bit error aswell as the synchronization error rate.The latter one gives the rate of slots


Transmission format FDM/TDD/TDMA

Frequency band 1895 – 1918 MHz

Number of frequency channels 77

Number of users per frequency channel 4

Channel spacing 300 kHz

Frame length (8 slots) 5 ms

Data rate 384 kbit/s

Modulation method π/4-DQPSK, square-root raised-cosinewith α = 0.5

Peak mobile transmitter power 80 mW

Table 2 Important parameters of the PHS standard

which were missed because the receivercould not synchronize. These basic build-ing blocks of the testbed are completedby extensive diagnostics and measure-ment hardware, which give full hardwareand system transparency.

The receiver is an incoherent direct con-version receiver (see section 9). Gain andfrequency control may be operated withdifferent algorithms and any combinationof internal/external frame, slot, or bitsynchronization may be chosen foradvanced performance diagnostics.The internal bit synchronization of thereceiver uses the algorithm presented insection 4.

The second generation of the testbedcurrently completed focuses on exten-sions which may be added to the DECTstandard in the future. It also has evenmore thorough diagnostic possibilities:not only does it record the number oferrors which occurred, but the completereceived signal in its complex basebandrepresentation (to the computer hard-disk). This gives the opportunity to applyoff-line postprocessing to the recordeddata. To study the effect of variableantenna patterns we implemented aswitched beam antenna with a Butlermatrix, which allows us to select oneof eight antenna beams for transmissionand reception at the base station and so toinvestigate the feasibility and usefulnessof switched beam antennas in futureDECT systems.

9 Receiver hardware

When looking at cordless communica-tions equipment today, we find mainlytwo competing transceiver architectures.Heterodyne transceivers have been wellestablished for a long time, but needquite a number of expensive and hardlyintegrateable RF- and IF-components.Direct conversion architectures, althoughknown in theory for a long time, are rela-tively new in real implementations but

Figure 7 The DECT testbed


CAN

TX2

Forward RF Link

Wired Return Link

DTC

DRX

ERC

TX1

show a high potential for integration andare thus relevant candidates for theimplementation of DECT phones [16].

9.1 Heterodyne architecture

The principle of a heterodyne DECTtransceiver with a single IF is wellknown (Figure 8). The receiver RF partconsists of a discrete high performanceimage rejection filter followed by apreamplifier and a simple receiver IFmixer. The receiver IF part operates at anIF of e.g. 110 MHz and consists of a dis-crete, high performance IF filter followedby an IF amplifier and a discriminatordemodulator. The IF filter is one of themost intricate points in performance andcosts in the heterodyne transceiver solu-tion. Taking advantage of the constantenvelope of GMSK, the IF amplifier is anon-linear limiter amplifier without gaincontrol.

The transmitter IF part consists of an FMmodulator on IF level which is controlledby the Gaussian filtered transmit datafollowed by an IF amplifier and a simpleIF low-pass for IF harmonics suppres-sion. The transmitter RF part is builtfrom a simple IF mixer, a discrete, high-performance image rejection filter, apower efficient class C medium poweramplifier and a harmonics-low-pass. Thesynthesizer consists of a voltage con-trolled oscillator, a dual modulus pre-scaler and low frequency synthesizercomponents.

image rejectfilter

pre-amplifier

IFfilter

IFamplifer

demodu-lator

ante

nna

switch synthesizer

harmonicslow pass

filter

imagerejectfilter

IFfilter

IFamplifer

modu-lator

mediumpower

amplifier

IFmixer

IFmixer

RX-

data

TX-

data

Figure 8 Heterodyne architecture

9.2 Direct conversionarchitecture

Figure 9 shows a direct conversion re-ceiver, where the signal is transformedfrom baseband to RF and vice versa in asingle step without using an intermediatefrequency. Here, image frequencies donot appear.

The RF-part of a direct conversion re-ceiver consists of a simple and cheap pre-filter for noise and interference reductionfollowed by a low-distortion preamplifierwith switchable gain to reduce thedynamic range. The down-conversionfrom RF to baseband has to be performedin a more complex quadrature down-con-verter. This quadrature converter suppliesin-phase and quadrature components ofthe received signal in baseband.

The receiver baseband part performschannel filtering, level control and A/Dconversion by baseband signal process-ing followed by digital demodulation. Toaccomplish all these tasks, which wereshared between the RF and IF parts in theheterodyne architecture, baseband signalprocessing can be done in two ways.

The first possibility is to filter the com-plex baseband signal by two analoguelow-pass filters for channel selection,then to amplify the signals in a variablegain amplifier with accurate gain control,so that the dynamic range is highly re-duced. Finally, A/D conversion is per-

formed by a low resolution ADC. Asa second possibility the signal can befiltered and level adjusted only coarselyby a low-order filter and a simple gaincontrol, which means that both channelselection and gain control have to bedone in the digital domain. Thus, a high-resolution ADC must be used for A/D-conversion, and channel filtering is donedigitally.

The transmitter baseband-part convertsthe transmit data stream into complexsymbols and performs the shaping of in-phase and quadrature component of theGMSK signal. The transmitter RF-partshifts the complex baseband signals in aquadrature up-converter to RF. Amplifi-cation is again performed in a powerefficient class C medium power amplifierand a subsequent low-pass removes car-rier harmonics. The synthesizer is equalto that in the heterodyne solution.

9.3 Comparison

Both architectures are well suited forDECT transceivers. The differencesoccur mainly in the receivers, and bothimpose different challenges for imple-mentation.

Filtering by passive elements with highdynamic range makes the heterodynereceiver uncritical in intermodulation andnoise performance. The limiter-amplifierin the IF part and the discriminator-

demodulator are simple and low-powersolutions and avoid the necessity for anautomatic gain control (AGC) by theirinherent high dynamic range. The FMmodulator in the heterodyne transmittersupports a GFSK signal with a naturallylow accuracy of modulation index, butmatches quite well to the heterodyne dis-criminator demodulator. Due to the in-accuracy in modulation index and thenon-linear limiting, equalization of delayspread is not possible with the limiterdiscriminator receiver.

A well known critical point of the RF-part in direct conversion receivers is DC-blocking due to second order non-lineari-ties in the quadrature down-converter atreception of strong signals and self-reception of the local oscillator. This canbe removed by AC coupling with care-fully selected timing constants. In thereceiver baseband part, the direct conver-sion principle requires linear signal pro-cessing from the output of the quadraturedown-converter to the demodulator. Theaccuracy of the automatic gain controlcan be traded against the resolution of thefollowing A/D converter, both requiringpower consuming circuitry.

We have seen that the major drawback ofthe heterodyne solution is that it uses anumber of discrete high-performancefilters, which cannot be integrated andmake up for a considerable part of thecosts of a DECT telephone. The directconversion DECT transceiver completelyavoids costly discrete elements by mov-ing signal processing mainly to the base-band, but also to more complex but inte-grateable RF circuitry. So far, companiesseem to have chosen the low-risk-tech-nology heterodyne handset with its inher-ently higher production costs. If theychoose a direct conversion solution, itwill need a higher development effort,but will eventually yield a very low-costconsumer-electronic product.

10 Conclusion

DECT has been designed as a low-costsystem which has important conse-quences for the radio link. Due to theabsence of channel coding, the bit errorrate in the channel must be quite low, i.e.10-3. This requires high signal-to-noiseratio, low co-channel interference, andlow time dispersion. This last conditionis aggravated by the fact that no equalizeris foreseen and that DECT is a high-data-rate system. However, diversity is widely


Figure 9 Direct conversion architecture

pre-filter

pre-amplifier

basebandsignal processing

digitaldemodulation

ante

nna

switch synthesizer

harmonicslow pass

filter

datastream

conversion

mediumpower

amplifier

quadraturedown

converter

quadratureup

converter

RX-

data

TX-

data

IQ shaping

and successfully applied to improve BERand tolerance to time dispersion. Formost PCS and WLL applications, equal-ization will be necessary to obtain satis-factory performance, but co-existence ofhand-helds with or without equalizers isno problem.

On the other hand, the very simplicity ofthe DECT system allows a straightfor-ward and therefore cheap construction ofthe hardware. Tolerances in the standardare not very tight, which allows the useof cheap components and even single-chip implementations of complete hand-sets. These low production costs havebeen the prerequisite for the phenomenalgrowth of the DECT market, and allowto anticipate an even brighter future.

References

1 ETSI. Radio Equipment and SystemsDigital European Cordless Telecom-munications Common Interface.DECT Specification, Part 1 – 3, ver.02.01, 1991.

2 Pettersen, M, Rækken, R H. DECToffering mobility on the local level :experiences from a field trial in amultipath environment. Proc. 46thIEEE Vehicular Techn. Conf.,Atlanta, 1996, 829–833.

3 Eskedal, B E. The DECT system.Telektronikk, 91 (4), 25–29, 1995.

4 Other DECT paper in this issue.

5 Adachi, F, Parsons, J D. Error rateperformance of digital FM mobileradio with postdetection diversity.IEEE Trans. Comm., 37, 200–210,1989.

6. Liu, C L, Feher, K. Bit error rate per-formance of π/4-DQPSK in a fre-quency-selective fast Rayleigh fadingchannel. IEEE Trans. VehicularTechn., 40, 558–568, 1991.

7 Parsons, J. The mobile radio channel.New York, Wiley, 1992.

8 Paier, M, Molisch, A F, Bonek, E.Determination of the optimum sam-pling time in DECT-like systems.Proc. EPMCC’97, 459–465, 1997.

9 Molisch, A F et al. Error floor ofunequalized wireless personal com-munications systems with MSK mod-

ulation and training-sequence-basedadaptive sampling. IEEE Trans.Comm., 45, 554–562, 1997.

10 Kadel, G. Performance of the DECTsystem with a 2-state Viterbi equal-izer. COST 231, TD(93)100.

11 Lopes, L B. A non-coherent equal-izer receiver for DECT-type systems.COST 231, TD(93)116.

12 Mogensen, P E et al. Evaluation ofan advanced receiver concept forDECT. Proc. 45th IEEE VTC,514–519, 1995.

Safavi, S et al. An advanced basestation receiver concept for DECT.Proc. 45th IEEE VTC, 150–154,1995.

13 Fuhl, J, Schultes, G, Kozek, W.Adaptive equalization for DECT sys-tems operating in low time-dispersivechannels. Proc. 44th IEEE VTC,714–718, 1994.

14 Molisch, A F, Bonek, E. Reductionof error floor of differential PSK inmobile radio channels by adaptivesampling. IEEE Trans. VehicularTechn., in press.

15 Schultes, G et al. A Testbed ForDECT Physical and Medium Access-Layer. Proceedings PIMRC’92,349–356, 1992.

16 Schultes, G, Kreuzgruber, P, Scholtz,A L. DECT Transceiver Architec-tures : Superheterodyne or DirectConversion? Proceedings 43rd IEEEVTC, 953–956, 1993.


Ernst Bonek is Professor of radio frequency en-gineering at the Technische Universität Wien. Heis interested in radio propagation for cellular mobileradio, ‘smart’ antennas and mobile radio networks,and serves as a working group chairman for ‘An-tennas and Propagation’ in the COST Action 259‘Wireless Flexible Personalised Communications’.He is also Member of the Board of Post undTelekom Austria.


Heinz Novak is Assistant Professor at the radio fre-quency department of the Technische UniversitätWien. He is engaged in the research on transceiverarchitectures for mobile and cordless communica-tions systems with a strong emphasis on DECT.His present research interests include the imple-mentation of a demonstrator for a third generationcommunication system, which includes a switchedbeam adaptive antenna.


Andreas F. Molisch is Assistant Professor at theradio frequency engineering department of theTechnische Universität Wien. His current researchinterests are bit error probabilities of cordless tele-phones, modelling of the mobile radio channel, andOFDM. He participated in COST 259 ‘WirelessFlexible Personalised Communications’. Earlier, heworked on surface acoustic wave filters, radiativetransfer in atomic vapours, and smart antennas.



1 Introduction

Personal and mobile communicationssystems have experienced an overwhelm-ing increase in number of users in the

recent years. With an increasing numberof users and also additional services likedata services being introduced there is agrowing need for capacity.

Capacity can be increased by higherspectrum efficiency or an extended fre-quency band. Since the frequency bandallocated to personal communicationssystems is limited, a number of tech-niques is being developed to achievehigher spectrum efficiency. In a cellularsystem the radio communication is be-tween the user and a base station whichprovides radio coverage within a certainarea, which is called a cell. In this casethe capacity is limited by the cell density,the frequency reuse distance and thenumber of users which can be servedsimultaneously by each base station.

The traditional ways of separating userscommunicating with one base station isby frequency, FDMA – Frequency Di-vision Multiple Access; by time, TDMA– Time Division Multiple Access; or bycode, CDMA – Code Division MultipleAccess. Traditional base station antennasare omnidirectional or sectored, but thisis a ‘waste’ of power because most of itwill be transmitted in other directionsthan towards the user. In addition, thepower radiated in other directions will beexperienced as interference by otherusers. Figure 1 illustrates the concept of acell and multiple users communicatingwith one base station.

Spectrum efficiency can be increased byusing smaller cells, so-called microcells,or by using frequency hopping, a tech-nique which disperses interference. Allthese techniques are used in current digi-tal mobile communications systems (eg.GSM) and will be employed also infuture systems. However, in addition tothese, the most promising technique issmart or adaptive antennas. This tech-nique adds a new way of separating userson one base station, namely by space,SDMA – Space Division MultipleAccess. An additional capacity gain isachieved by reduced interference levels,making lower frequency reuse distancespossible. This will be explained further inchapter 4.

The idea of smart antennas is to useadaptive, directional antennas at the basestation. By directing the antenna beamtowards the communication partner only,more users can be assigned to the samebase station. In this way, a much moreefficient usage of the power and the spec-trum is achieved, than in the traditionalcase of omnidirectional or fixed beamantennas. Figure 2 illustrates the conceptof SDMA with multiple users communi-cating with a base station employing

Smart antennas– The answer to the demand for higher spectrum efficiency in personal communications systems

M A G N E P E T T E R S E N A N D P E R H J A L M A R L E H N E

BaseStation

BaseStation

Figure 1 Multiple users communicatingwith one base station; users separated byfrequency, time or code

Figure 2 Base station employing smartantennas; users separated by direction


smart antennas. In this case the users canuse the same physical radio communica-tion channel simultaneously if they aresufficiently separated by direction. Aswill be clear in chapter 2, SDMA is thefinal step in a development towardsincreasingly ‘smarter’ antennas.

In the last couple of years there has beenlarge activity in the field of smart anten-nas for mobile and personal communica-tions. So why is the interest appearingnow, and not five or ten years ago? Theanswer probably lies with the fact thatuntil now there has not been much reasonto worry about spectrum efficiency. Also,if a base station is to track a large numberof users simultaneously, the computa-tional cost will be large, making power-ful, expensive signal processors neces-sary. Only recently have such processorsreached a reasonably low cost.

Smart antennas have a number of ad-vantages over traditional omnidirectionalor sectored base station antennas in addi-tion to increased capacity, includingincreased range, higher level of securityand possibility for new services. This willbe discussed in more detail in chapter 4.

Although this article focuses on smartantennas for land based mobile commu-nications, one should keep in mind thatthe technology can and will be exploitedalso for other types of radio systems.

2 Basic principles

What do we mean by the term ‘smartantenna’? The theory behind smart an-tennas is not new. The technique has formany years been used in electronic war-fare (EWF) as a countermeasure to elec-tronic jamming. In military radar systemssimilar techniques were used as early asWorld War II. There is in principle anumber of ways in which an adaptivelyadjustable antenna beam can be gene-rated, for instance by mechanicallysteered antennas. However, the techniquealmost exclusively suggested for landbased mobile and personal communica-tions systems is arrays, or group an-tennas, giving directivity. This tech-nology will be explained in some detailin chapter 3.

The main philosophy is that interferersrarely have the same geographical loca-tion as the user. By maximising theantenna gain in the wanted direction andsimultaneously placing radiation patternminima in the directions of the inter-ferers, the quality of the communicationlink can be significantly improved. Inpersonal and mobile communications, theinterferers are other users than the userbeing addressed.

2.1 The concept of a ‘smart’ antenna

Several different definitions for smartantennas are used in the literature. One

useful and consistent definition can bethat the difference between a smart/adap-tive antenna and a ‘dumb’/fixed antennais the property of having a steerable andfixed lobe-pattern, respectively. This isthe definition to be used in this article.Figure 3 illustrates the concept of a smartantenna. Normally, the term ‘antenna’comprises only the mechanical construc-tion transforming free electromagnetic(EM) waves into radio frequency (RF)signals travelling on a shielded cable andvice versa. We may call it the radiatingelement. In the context of smart an-tennas, the term ‘antenna’ has an ex-tended meaning. It consists of the radiat-ing element and a control unit. We candenote the control unit the smart an-tenna’s intelligence and it is illustrated inFigure 3 by the ‘brain’, normally realisedusing a digital signal processor (DSP).The processor controls feeder parametersof the antenna, based on several inputs,in order to optimise the communicationslink. The optimisation criteria can differas will be explained in the next section.This shows that smart antennas are morethan just the ‘antenna’, but rather a com-plete transceiver concept.

2.2 Levels of intelligence

Based on the definition above, one can de-fine ‘levels of intelligence’. These are de-scribed below and illustrated in Figure 4.

• Switched lobeThis is the simplest level of intelli-gence, and comprises only a basic

"Intelligence"

Control

RF in/out

To/from radiosub system

Steerable lobe

Figure 3 Principle of a smart antenna


switching function between separatedirective antennas. The antenna lobe ischosen which has its angle of directionclosest to the direction towards theuser. Because of the higher directivitycompared to a conventional antenna,some gain is achieved. Such anantenna will be easier to implement inexisting cell-structures than the moresophisticated adaptive arrays, but itgives a limited improvement.

• Dynamically phased arrayBy including a direction of arrival(DOA) algorithm for the signal re-ceived from the user, a continuoustracking can be achieved. The an-tenna’s main lobe is directed towardsthe user, and thus the signal level ismaximised.

• Adaptive antennaIn addition to a DOA algorithm, ifalgorithms for determining the direc-tion towards interference sources (e.g.other users) are added, the radiationpattern can be adjusted to null out theinterferers. In addition, by usingspecial algorithms and space diversitytechniques, the radiation pattern can beadapted to receive multipath signalswhich can be combined. These tech-niques will maximise the signal-to-interference-and-noise ratio (SINR). A

complementary gain is achieved in thedownlink direction by using the sameradiation pattern as found for recep-tion. This requires that the radio chan-nel does not change very rapidly. Amore comprehensive description of thetechnology used is found in chapter 3.

Conventional mobile systems usuallyemploy some sort of antenna diversity(e.g. space or polarisation diversity).Adaptive antennas can be regarded as anextended diversity scheme, using morethan two diversity antennas. In this con-text, phased arrays will have a greatergain potential than switched lobe anten-nas because all elements can be used fordiversity combining [1] [2].

2.3 An evolutionary path forsmart antennas

All the described levels of intelligence inthe previous section are technologicallyfeasible today. However, in the domainof personal and mobile communications,an evolution can be foreseen in the utili-sation of smart antennas towards gradu-ally more advanced solutions. The evolu-tion can be divided into three phases [3]:

1. Smart antennas are used on uplinkonly. (Uplink means that the user is

transmitting and the base station isreceiving.) By using a smart antenna toincrease the gain at the base station,both the sensitivity and range are in-creased. This concept is called HighSensitivity Receiver (HSR) and is inprinciple not different from the diver-sity techniques implemented in today’smobile communications systems.

2. In the second phase, directed antennabeams are used in the downlink (basestation transmitting and user receiving)direction in addition to HSR. In thisway, the antenna gain is increased bothon uplink and downlink, which impliesa spatial filtering in both directions.Frequencies can be more closely re-used, thus the system capacity in-creases. The method is called SpatialFiltering for Interference Reduction(SFIR). It is possible to introduce thisin second generation systems, like forexample GSM.

3. The last stage in the development willbe full Space Division Multiple Access(SDMA). This implies that more thanone user can be allocated to the samephysical channel simultaneously in thesame cell, only divided by direction.In a TDMA system, two users will beallocated the same time slot and carrierfrequency at the same time and in the

Switched lobe Dynamically phased array Adaptive antenna

Signal

Interference

Figure 4 Levels of intelligence


same cell. In phase 2, the capacity isincreased due to tighter frequencyreuse. In phase 3, an additional in-crease in capacity is achieved locallyper base station. Introducing SDMA insecond generation TDMA systems willbe difficult, but it will be a naturalcomponent in third generation systems,like UMTS.

The ‘levels of intelligence’ in the pre-vious subsection describes the level oftechnological development, while thesteps described here can be regarded aspart of a system evolution.

3 Technology for Smart Antennas

This chapter explains some of the under-lying technology employed in order toimplement smart antennas. The treatmentgiven here is far from complete, butshould give an impression of the basicideas. The technology is based on arrayantennas where the radiation pattern isadjusted by altering the amplitude andrelative phase on the different array ele-ments. Even though the reception andtransmission parts are integrated, andmuch of the same hardware normallywill be used, they will be explained sepa-rately. This is because conceptually theuplink and downlink are quite differentin terms of smart antennas.

3.1 Reception (uplink)

Figure 5 shows schematically the imple-mentation of the reception part of a smartantenna. The antenna array contains Nelements. The N signals are being com-bined into one signal, which is the inputto the rest of the receiver (channel de-coding, etc.).

As Figure 5 shows, the smart antennareception part consists of four units. In

addition to the antenna itself it contains aradio unit, a lobe forming unit and asignal processing unit [4].

The array will normally have less than 10elements in order to avoid too high com-plexity of the signal processing. The ele-ments could be positioned linearly (lineararray) if a sector is to be covered or in acircle (circular array) if symmetrical cov-erage around the base station is desired.Often, each of the array elements will

w1

w2

w3

wN

SignalProcessing

Unit

Antenna

1

2

3

N

Lobe Forming Unit

•••

•••

RadioUnit

Figure 5 Smart antennas, reception part

LFULFULFU

Ver

tical

(pa

ssiv

e)co

mbi

ning

LFU - Lobe Forming Unit

a) b) c)

Figure 6 a) Linear array, bird’s eye view; b) Circular array, bird’s eye view; c) linear array with multiple antenna elements perarray element


consist of a number of antenna elementsseparated vertically which are combinedpassively to achieve a narrower verticallobe [5]. Each of the array elements willmore or less resemble a traditional basestation antenna. Figure 6 shows a circularand a linear array, and also a linear arraywith multiple antenna elements per arrayelement.

The radio unit consists of down conver-sion chains and (complex) analogue todigital conversion (A/D). There must beN down conversion chains, one for eachof the array elements.

The signal processing unit will, based onthe received signal, calculate the com-plex weights w1-wN with which the re-ceived signal from each of the array ele-ments is multiplied. These weights willdecide the antenna pattern in the uplinkdirection. The weights can be optimisedfrom two main types of criteria; maxi-misation of received signal from the de-sired user (bullet 2 from sec. 2.2) or max-imisation of the SINR by suppressing thesignal from interference sources (bullet 3from sec. 2.2). In theory, with M antennaelements one can ‘null out’ M-1 interfer-ence sources, but due to multipath propa-gation this number will normally be

lower. Figure 4 (chapter 2) shows a casewhere the SINR is maximised by placingnulls in the direction of the interferencesources.

The methods which maximise SINR doin principle require knowledge of theinstantaneous channel response in bothtime and space from both the desired userand all the interference sources. Commonfor all the methods which suppress inter-ference is therefore that they require areference or training sequence, i.e. aknown bit sequence must be transmittedperiodically in order for the receiver tobe able to estimate the channel response.This sequence should be unique for eachuser.

In the lobe forming unit the actualweighting of the received signal fromeach of the array elements is performed.In the most advanced case this unit willbe an integrated channel equaliser/smartantenna. In this case N•D weights will berequired, where D is the number of sym-bol periods (depth) in the channel equal-iser. This is called a spatio-temporalfilter [6] [7], because it removes the un-desired signal components and keeps thedesired ones both in time and spacedomains.

When the lobe forming is performed dig-itally (after A/D) the lobe formingand signal processing units will normallybe integrated in the same unit, namelythe DSP. The separation in Figure 5 isdone to clarify the functionality. It is alsopossible to perform the lobe forming inhardware at radio frequency (RF) orintermediate frequency (IF) [8].

3.2 Transmission (downlink)

The transmission part of the smartantenna will be schematically very simi-lar to the reception part. The transmissionpart is shown in Figure 7. The transmis-sion signal is split into N branches, eachbeing weighted by (complex) weightsz1-zN in the lobe forming unit. Theweights are calculated by the signal pro-cessing unit. The weighting decides theradiation pattern in the downlink direc-tion. The radio unit consists of the upconversion chains. In practise, some com-ponents, like the antenna itself and theDSP, will be the same as on reception.

The principle difference between uplinkand downlink is that since there will beno smart antennas employed to the userterminals (mobile stations), on downlinkthere is no way of knowing the instan-taneous channel response. One couldthink that the weights calculated onreception (w1-wN) would be optimal alsoon transmission, but this will normallynot be the case. If time division duplex isused (TDD, uplink and downlink trans-mission separated in time), the uplink-weights will be valid on downlink only ifthe channel does not change during theperiod from uplink to downlink trans-mission, i.e. if the user is moving veryslowly. This can rarely be assumed ingeneral. If frequency division duplex isused (FDD, uplink and downlink trans-mission separated in frequency) theweights will generally not be the same,because of the channel response depen-dency on frequency.

The technique most frequently suggestedto solve this problem is the estimation ofDOA (chapter 2). On uplink the DOA forthe desired user is estimated, i.e. thedirection from which the main part of theuser signal is received. This direction isused in downlink transmission. This isdone by choosing the weights z1-zN sothat the radiation pattern is a lobe di-rected towards the desired user, as illus-trated for instance in Figure 2.

z1

z2

z3

zN

DOA-estimatefrom uplink

Antenna

1

2

3

N

Lobe Forming Unit

•••

•••

SignalProcessing

Unit

Spl

itter

RadioUnit

Figure 7 Smart antenna, transmission part


Figure 8 The uplink transmission experiences much angular spread due to multipath propagation. Downlink transmission based ondirecting a single lobe in the direction of maximum reception (DOA) is not optimum in this case

The DOA can be estimated by a numberof well known algorithms, for instanceMUSIC, ESPRIT or COSSA [3].

3.3 Technical challenges andcritical factors

There are a number of technical chal-lenges which must be overcome in orderfor smart antennas to achieve their fullpotential in mobile and personal commu-nications. These include both modemimplementation aspects and network

aspects. The aspects mentioned in thissubsection are only some of the areaswhere research is done to find propersolutions.

The downlink is one of the major chal-lenges. As described in the previous sub-section the lack of knowledge about theinstantaneous channel response on down-link can to some extent be compensatedfor by estimating the DOA and transmitin that direction. But this makes thedownlink performance very dependent

on the channel response behaviour inspace. If there is much multipath propa-gation, so that the received signal hasmuch angular spread, directing a singlelobe in the direction of maximum recep-tion is far from optimum in terms ofmaximising signal power transferred tothe user. Interference suppression, tomaximise SINR, is also difficult ondownlink for the same reason.

In addition, when the channel is changingrapidly, the optimum direction calculated


on uplink may not be valid on downlinktransmission. Figure 8 illustrates the casewith much multipath propagation andangular spread. Much attention is at themoment given to this problem, and a lotof effort is put into finding possible solu-tions.

The linearity in reception and trans-mission chains is another critical factor.If the transfer functions for the up- anddown-converter chains are not perfectlyknown the actual weights will be differ-ent from the calculated ones, and theantenna radiation pattern may changesignificantly because of this. Since thetransfer functions of the up- and down-converter chains are temperature depen-dent and also may change with time,frequent on-line calibration of the radiounits of the smart antennas is necessary[8].

One of the network aspects deservingrenewed attention when smart antennasare being introduced is connection set-up. When the user initially establishescontact with the base station, no angularinformation is available and some meansfor the base station to ‘find’ the user isnecessary. A possible solution is that thebase station scans the cell for new usersby continuously sweeping through thecell with a ‘search’ beam. A similarlycritical point is handover. When the con-nection is to be handed over to anotherbase station some means must be foundto provide the network with informationof which base station to switch to.A possible solution is to use an externalsystem for positioning, for instance GPS.Information of the location could enablethe network to make an ‘educated guess’about which cell to hand the connectionover to. Another possibility is a sweeping‘handover beam’ mapping the user signalquality from different base stations [10].

Upon introduction of full SDMA (chap-ter 2) there will be new demands on radioresource allocation. For instance in aTDMA system, each user must be allo-cated not only a carrier frequency and atime slot, but also an angle. When angu-lar collisions occur for two users usingthe same physical radio channel, one ofthem must switch to another channel.This must be done in a way so that theconnection is not broken. This indicatesthat a system using SDMA must employsome sort of dynamic channel allocation.Figure 9 shows the occurrence of anangular collision.

When using smart antennas a ‘spatialisolation’ between the desired user andinterferers is achieved by directing theantenna lobe in the direction of the de-sired user [11]. However, if the inter-ference source is much closer to the basestation than the desired user, the isolationis lost and communication may breakdown. This is similar to a ‘near-far-prob-lem’ and indicates that some sort ofpower control is necessary, so that thetransmit power from the user terminal isadjusted according to the distance fromthe base station.

4 Implications onpersonal and mobilecommunications

Introducing smart antenna technologyinto personal and mobile communica-tions is perhaps the most important newtechnique to increase capacity in mobilenetworks since the cellular technologywas introduced. It has impact on radioperformance and stretches several limitsbeyond today’s feasibility level. In thiscase, as in most others, there are alsodrawbacks and cost factors. In addition,it will have impact on radio planning.Also, new demands are put on networkfunctionality. All of these aspects are dis-cussed in this chapter.

4.1 Potential improvements

There are potentially great improvementsin using smart antennas in mobile com-

BaseStation

Figure 9 Angular collision

a)

b)

Figure 10 Reduced frequency reuse dis-tance when employing smart antennascan be attained by using smaller cell

clusters. a) Traditional 7-cell cluster asused in e.g. GSM; b) Possible 3-cellcluster when using smart antennas


munications. They have a number ofadvantages over traditional omnidirec-tional or sectored base station antennas.These include:

• Increased capacity. By transmittingelectromagnetic power only in thedirection of the communication part-ner, the interference experienced byother users will be reduced, the fre-quencies can thus be reused more fre-quently as Figure 10 shows an ex-ample of, and capacity is increased.

• Increased range. Because the basestation antennas will be more directivethan traditional ones, the range willincrease if the same amount of poweris transmitted. This is useful in sparse-ly populated areas where the base sta-tions can be positioned further apartbecause capacity is not dictating basestation deployment.

• New services. The spatial knowledgeof the position of the user makes newservices possible. These services canfor instance include emergency calllocation or location specific billing.

• Security. It is more difficult to tap aconnection when smart antennas areused. To successfully tap a connectionthe intruder must be positioned in thesame direction from the base station asthe user.

• Reduced multipath propagation. Whena narrow antenna beam is used, theeffect of multipath propagation, due tosignal reflections, will be greatly re-duced. This could ease the require-ments on future modem design.

The most important of the advantagesmentioned above is increased capacity.A capacity increase of 3 to 5 times isforeseen for GSM and similar TDMAsystems [12]. For CDMA systems theincrease is expected to be even higher.As described above, the reduced fre-quency reuse distance will lead to in-creased capacity. In future systems, theintroduction of SDMA will give an addi-tional capacity gain, because more userscan communicate simultaneously via onebase station.

Because interference suppression is themost important single factor increasingcapacity, let us give an example of thepotential. Given a system with N mutu-ally interfering users and an adaptiveantenna with N + K antenna elements.For each of the N users, it is then pos-sible to eliminate N - 1 interference

sources and simultaneously exploit K + 1multipath signal components [2].

4.2 Drawbacks and cost factors

There is always a price to pay. Increasedcomplexity leads to higher equipmentcosts. Besides, smart antennas will notbe equally suitable in all geographic anddemographic scenarios. It is a questionwhether the higher cost will be com-pensated for by the gain in all scenarios.

One drawback is the physical size. Toobtain a reasonable gain, arrays of sev-eral elements are used, having an elementspacing of 0.4 – 0.5 wavelengths [9][13]. (The wavelength for 900 MHz is34 cm, while it is 17 cm for 1800 MHz.)Thus, these are large antenna systemsand for 900 MHz the construction can beimpractically large. For example, with 10elements spaced 0.5 wavelengths apart,the total array width is 1.7 m on900 MHz. Constraints on the size canalso be a deciding factor for the choice ofdiversity technique (i.e. space vs. polari-sation). Smart antenna systems are alsocomputationally intensive. In order totake advantage of the potential benefits,there is a need for powerful and rapidnumeric processors and control systems.There is also a need for efficient algo-rithms for real-time optimising and signaltracking.

4.3 Radio planning

It is necessary to develop a new way ofthinking when it comes to choosing goodsites for base stations with smart antennasystems. The current rule in placing basestations along a highway, a railway lineand so on, is optimised on the basis of afixed radiation pattern. Angle dependentaccess methods can, to a small degree,only be exploited in this way. By puttingthe base stations further away from theroad or line, the beam can be swept alongthe coverage area. In this way, the spatialdimension is much better exploited. Thisis illustrated in Figure 11. New radioaccess protocols are necessary to takeadvantage of the new dimension in e.g.connection set-up and handover.

4.4 Network aspects

Even though smart antennas basically isa radio sub-system technology, it hassome implications on the networking

functions as discussed in chapter 3. Inconnection set-up, the base station needssome means to locate the user, and estab-lishing directional information. Alsowhen handover is to be performed, thenetwork needs some positioning informa-tion about the user in order to make theright decision. This will require addi-tional signalling information comparedto current methods.

4.5 What about GSM?

Smart antennas will be fully exploited inthird generation systems. However, partsof the technology can be integrated incurrent second generation systems likeGSM and DCS1800. It is possible to usesmart antennas in GSM without makingchanges to any other components thanthe Base Transceiver Station1 (BTS)[14]. It is difficult to implement fullSDMA in GSM because it is not possibleto allocate more than one user per time-slot and frequency per BTS.

In ongoing trials using DCS1800 (chap-ter 5), smart antennas will only be used atthe BTS. Consequently, the gain ishighest on the uplink direction. TheSINR is expected to be raised by up to30 dB! [6] Other sources [16] indicate animprovement of 4 to 8 dB for downlinkand 14 dB for uplink. One of the conse-quences of the improved signal qualitycan be a capacity increase of 3 to 5 times.

5 Current situation

This chapter contains a brief overview ofthe current situation in the area of smartantennas with regards to ongoing re-search and state-of-the-art.

5.1 Availability and state-of-the-art

As mentioned earlier, in addition to be-coming an integral part of third genera-tion mobile communications, smartantenna technologies have the potentialto significantly improve the capacity insecond generation systems, like GSM/DCS1800. Therefore, research is focused

1 In GSM, the base station subsystem(BSS) is organised in a controller part(BSC – base station controller) and theradio part (BTS – Base TransceiverStation). The BTS consists of one ormore transmitter-receiver units (TRX).


on current systems as well as on futuresystems, like UMTS.

Already, simple implementations ofsmart antennas designed for DCS1800are available commercially. These areswitched lobe solutions (similar to bullet1, section 2.2), with a small number of

lobes and switching based on signal levelonly [17] [18]. Most likely, many manu-facturers are working on more advancedsmart antenna solutions for GSM andDCS1800, but for reasons of competitionit is difficult to get information about thiskind of product development.

Apart from simple switched solutions,the closest smart antennas have come toimplementation is testbeds which havebeen demonstrated on a stand-alonebasis, and which are being implementedin a running network at a later stage. TheEuropean ACTS-project TSUNAMI[8][11][15] has successfully demon-

Figure 11 Different radio planning principles must be employed when smart antennas are to be used. a) Conventional highway coverage technique;

b) Possible method of highway coverage using smart antennas

b)

BaseStation

BaseStation

a)

BaseStation

BaseStation


strated a testbed consisting of one basestation and two mobile stations [19]. Thetestbed is based on DCS1800 technology,and will be implemented into the DCS1800network of Orange PCS in the UK.

Ericsson have reported impressive resultsfrom a similar testbed [14], also based onDCS1800 technology. This testbed is tobe implemented into the DCS1800 net-work of Mannesmann Mobilfunk in Ger-many.

5.2 Ongoing research

A lot of international research is beingdone to develop smart antenna solutions.The research ranges from algorithms forDOA-estimation and channel characte-risation, investigation of the implicationson network aspects like handover andcall-set-up, estimations of capacity andrange improvements to development ofhardware like antennas and receivers. Inaddition to land based mobile communi-cations there is also research focusing onsmart antennas for satellite communica-tions.

A major drive in European research with-in this area is the aforementionedTSUNAMI-project, which has its mainfocus on smart antennas for UMTS.Some of the European institutions whichhave published results are the Universi-ties of Bristol and Bradford in the UK,Aalborg University in Denmark, RWTHAachen in Germany and also the RoyalInstitute of Technology in Sweden.

In the United States, the research onsmart antennas focuses on benefits forCDMA-systems to a larger extent than isthe case for Europe. Some of the institu-tions reporting results are Lucent Tech-nologies and AT&T Labs Research,Telecom Solutions Inc. (Virginia) in co-operation with universities of Virginiaand Texas, MIT Lincoln Lab and HughesNetwork Systems and NASA.

In Asia, Kohno Laboratories of theYokohama National University, Mit-subishi Electric Corporation and the Uni-versity of Singapore, among others, havereported results. However, the European,American and Asian institutions men-tioned above only make up a small per-centage of the universities and companiescurrently performing research within thearea of smart antennas.

6 Summary andconclusions

The concept of smart or adaptive an-tennas have the potential to significantlyimprove the capacity in mobile and per-sonal communications systems, almostto the degree where it can be called arevolution. Additional benefits includeincreased range, new services, increasedlevel of security and protection againstmultipath propagation.

The main idea of smart antennas is thatthe base station antenna radiation patternis adjusted adaptively in such a way thatthe antenna beam is directed towards thedesired user only, and nulls are directedtowards the interference sources. In thisway, one base station can separate userson the same physical communicationchannel by angle. This is achieved byusing array antennas on which the radia-tion pattern can be adjusted by alteringthe amplitude and relative phase on eachof the array elements.

Field trials have shown that smart an-tennas do work. However, the full po-tential of the technology remains yet tobe demonstrated. There is still a numberof technical challenges which must besolved. In addition, there are some draw-backs compared to current systems, likelarger antenna constellations and moreexpensive hardware.

Smart antenna solutions will most likelybe an integral part of third generationmobile communications systems, but lesspowerful solutions can also be imple-mented in second generation systems,like GSM, to give a significant gain.Smart antennas will influence mostaspects of the systems, also networkaspects and radio planning.

Few experts seem to doubt that smartantennas will have a large impact onfuture mobile and personal communica-tions systems, and much research is cur-rently taking place to find proper solu-tions.

7 References

1 Winters, J H, Gans, M J. The rangeincrease of adaptive versus phasedarrays in mobile radio systems. 28thIsilomar Conference on Signals, Sys-tems and Computers, Pacific Grove,CA, USA, 31 Oct – 2 Nov 1994,109–115.

2 Winters, J H, et al. The impact ofantenna diversity on the capacity ofwireless communications systems.IEEE Transactions on Communica-tions, 42 (2/3/4), 1994, 1740–1751.

3 Bull, T, et al. Technology in smartantennas for universal advancedmobile infrastructure (TSUNAMIR2108) : overview. RACE MobileTelecommunications Summit 1995,Cascais, Portugal, 22–24 Nov 1995,88–97.

4 Barrett, M, Arnott, R. Adaptiveantennas for mobile communications.Electronics and CommunicationsEngineering Journal, August 1994,203–214.

5 Johannison, B. Active and adaptiveantennas for mobile communications.Seminar and Demonstration onAdaptive Antennas and AntennaDiversity for Mobile Communica-tions, Aalborg, Denmark, 19 Jun1997.

6 Öberg, T et al. Adaptiva antenner förmobil kommunikation. NordiskRadioseminarium 1996, Lund,Sverige, 27–28 Aug 1996, 44–47.

7 Ogawa, Y et al. Spatial-domain path-diversity using an adaptive array formobile communications. Proceed-ings of ICUPC’95 – 4th IEEE Inter-national Conference on UniversalPersonal Communications, Tokyo,Japan, 6–10 Nov 1995, 600–604.

8 Mogensen, P et al. A Hardwaretestbed for evaluation of adaptiveantennas in GSM/UMTS. 7th IEEEInternational Symposium on Per-sonal, Indoor and Mobile RadioCommunications – PIMRC’96,Taipei, Taiwan, 15-18 Oct 1996,540–544.

9 Forssén, U et al. Adaptive antennaarrays for GSM900/DCS1800. 44thIEEE Vehicular Technology Confer-ence – VTC’94, Stockholm, Sweden,8–10 Jun 1994, 605–609.

10 Rheinschmitt, R et al. Networkaspects of the introduction of adap-tive arrays in mobile communicationssystems. RACE Mobile Telecommu-nications Summit 1995, Cascais,Portugal, 22–24 Nov 1995, 367–371.


11 Clarkson, J W. An advanced antennasystem – unplugged. ACTS MobileTelecommunications Summit 1996,Granada, Spain, 27–29 Nov 1996,302–307.

12 Dam, H. Smart antennas in the GSMsystem. TSUNAMI Seminar andDemonstration of Adaptive Antennas,Aalborg, Denmark, 18 Jan 1996.

13 Wells, M C. Adaptive antennas forfrequency reuse in every cell of aGSM network. IEE Colloquium on‘Mobile Communications towardsthe Year 2000’, London, 17 Oct1994, 11/1–11/6.

14 Forssén, U et al. Ericsson GSM field-trial with adaptive antennas. NordiskRadioseminarium 1996, Lund, Swe-den, 27–28 Aug 1996, 48–51.

15 Mogensen, P et al. TSUNAMI (II)stand alone testbed. ACTS MobileTelecommunications Summit,Granada, Spain, 27–29 Nov 1996,517–527.

16 Zetterberg, P. Signal processing algo-rithms for base station antenna arraysin GSM. KTH Endagsseminariumöver temat ‘Avancerade Grupp-antenner för kommunikation och

radar’, Stockholm, Sweden, 5 Nov1996, 1–3.

17 Beckman, C. Multibeam antennas.KTH Endagsseminarium över temat‘Avancerade Gruppantenner förkommunikation och radar’, Stock-holm, Sweden, 5 Nov 1996, 1–3.

18 Northern Telecom. Smart antennas.Product sheet, Devon, England,1992.

19 Mogensen, P. Overview of CPKwork in TSUNAMI II & systemaspects of adaptive antennas forGSM. Seminar and Demonstrationon Adaptive Antennas and AntennaDiversity for Mobile Communica-tions, Aalborg, Denmark, 19 Jun,1997.

Abbreviations

A/D Analogue to digital (conversion)

ACTS Advanced CommunicationsTechnologies and Services,4th European Union frame-work programme in tele-communications researchin Europe, 1995 – 1998

BTS Base Transceiver Station(GSM term)

CDMA Code Division MultipleAccess

COSSA Direction of arrival (DOA)algorithm

DCA Dynamic Channel Allo-cation

DCS Digital Cellular System

DOA Direction of Arrival

DSP Digital Signal Processor

EM Electromagnetic (waves)

ESPRIT Estimation of Signal Param-eters via Rotational Invari-ance Techniques, Directionof arrival (DOA) algorithm

EWF Electronic Warfare

FDD Frequency Division Duplex

FDMA Frequency Division MultipleAccess

GSM Global System for Mobilecommunications

HSR High Sensitivity Receiver

IF Intermediate Frequency

MUSIC Multiple Signal Classifica-tion, DOA algorithm

NASA National Aeronautics andSpace Agency (USA)

PCS Personal CommunicationsSystem

RACE Research in Advanced Com-munications in Europe, 3rdEuropean Unions frameworkprogramme in telecommuni-cations research in Europe.Ended in 1995

RF Radio Frequency

SDMA Space Division MultipleAccess

SFIR Spatial Filtering for Inter-ference Reduction

SINR Signal to Interference andNoise Ratio

TDD Time Division Duplex

TDMA Time Division MultipleAccess

TSUNAMI Technology in Smart An-tennas for Universal MobileInfrastructure, RACE IIproject R2108 and ACTSproject AC020

UMTS Universal Mobile Tele-communications System

Magne Pettersen is Research Scientist at TelenorResearch & Development, Kjeller. He is workingin the field of personal communications, and hisspecial interest is in modelling and characterisationof the mobile radio channel, an area in which he iscurrently pursuing a PhD at the Norwegian Univer-sity of Science and Technology.


Per Hjalmar Lehne is Research Scientist atTelenor Research & Development, Kjeller. He isworking in the field of personal communications,with a special interest in antennas and radio wavepropagation for land mobile communications.


Evil communications corrupt goodmanners – I Corinthians 15:33.

A man by nothing is so well betrayed,as by his manners – Edmund Spenser.

The use of mobile telephones invarious situations has become anelement in the definition of sociallyappropriate/inappropriate behavior.They are causing us to reconsider howwe construct our social worlds. Theiruse demands a reevaluation of thetaken-for-granted assumptions ofeveryday life. This paper examineshow people deal with inappropriatemobile telephones use, particularly inrestaurants. The data, that is the talk,was collected in focus groups andthrough participation in electronicdiscussions on Usenet. The partici-pants from the focus groups included50 persons – 34 men and 16 women.Of the respondents, 30 reported ex-perience with a mobile telephone whilethe remaining 20 reported only limitedexperience. The paper has also drawnon Goffman’s notion of drama andstaging. From this basis it examinesthe reasons that restaurants are partic-ularly sensitive to the use of mobiletelephones. There is a discussion of,among other things, parallel frontstages and coerced eavesdropping.Finally, there is a discussion of themanagement strategies availablewhen ‘threatening’ situations arise.

1 Introduction

Not long ago I attended a church servicein my wife’s hometown, a small villageon the West Coast of Norway. Thechurch is a beautiful old stone buildingthat was built around the year 1000. Theinside of the building has white wallswith the remains of drawings from thefirst part of the millennium. In addition,there are beautiful pews and an altar thatreflect a long tradition of intricate wood-working. I was sitting near the back ofthe church not really listening to thesingsong of the preaching, only enjoyingthe sunshine through the windows. I wastrying to take in the scene when sud-denly, on the other side of the aisle Iheard the telltale peeping of a mobilephone. The offender sprang out of thepew in which he was sitting and rushedout of the church. After exchangingglances with several others sitting nearme, things settled down again and Iretreated back into my inner thoughts.

By now, most of us have been in a bus,an elevator or a restaurant and have heardor seen somebody talking on a mobiletelephone.1 In addition to the locationsmentioned above, I have seen, or ex-perienced, people using mobile tele-phones in places ranging from artgalleries to toilet stalls. In a 1937Readers Digest article Smith noted that,“There is no room in the house so privatethat he (sic.) cannot crash it by tele-phone” (Smith cited in Fischer 1992,225). This lament now seems to havebeen extended to the far corners of theknown world. These sightings show justhow widespread the technology hasbecome while the awkwardness of thesesituations points to the ways in which weare having to re-examine, or perhaps re-exert our ideas of propriety.

In this paper I am interested in examininghow it is that people are making sense ofthese changes and where people place theboundary between the appropriate andthe inappropriate use of mobile tele-phones. I focus particular attention onthe use of mobile phones in restaurantssince it seems to have become a meta-phor for vulgarity. The development ofthis metaphor is, however, not that sur-prising from a sociological perspective.After all, it is in restaurants that one seesquite clearly people dealing with theissues of self-presentation, boundaryissues, etiquette, and ‘stage manage-ment’.

This paper is a continuation of my workon the social boundary issues associatedwith the introduction of video telephony(Ling 1996). The mobile telephone, inmany respects, represents the opposite ofvideo telephones. Where videophonesrequire one to be fixed to a specific loca-tion, one can roam with a mobile phone.Where one is forced to pay attention totheir conversation partner with a videotelephone, one is freer to carry out paral-lel activities with a mobile telephone.Where the video telephone conversationis a well-bounded event, the mobile tele-phone call is less well defined and canintrude while one is on the bus, in arestaurant ... or in church.

65

“One can talk about common manners!”:

The use of mobile telephones in inappropriate situationsR I C H L I N G

Telektronikk 2.1998

1 In this article I use the term mobiletelephone. This is a translation of theNorwegian term mobiltelefon andrefers to all radio based telephonedevices aside from the cordless tele-phones whose coverage is usually lessthan 3-400 meters. In Norway the termmobile telephone refers to both devicesthat are mounted in cars and thosecarried about. In the US, the similargeneric term is cellular telephone.

Figure 1 The advent of the mobile telephone has meant that we are experiencing theuse of the device in new and unexpected places. As a result we are forced to reexamineour sense of propriety

LAR

S B

AH

L / 2

MA

J / S

AM

FO

TO

The work analyses data from a series offocus groups held in the Oslo area inaddition to data collected in otherforums. The data was analyzed usingqualitative analysis techniques and theresults have been further illuminatedthrough the use of Goffman’s dramatur-gical approach to small group interaction.The methods of the analysis will be dis-cussed in the next section followed by anexamination of Goffman’s applicabilityto this work, an analysis of the socialmeaning of mobile telephones and finallyby the examination of the inappropriateuse of mobile phones.

2 Method and theoreticalbackground

2.1 Method

The work here represents an analysis ofand reflections upon the information pro-vided by informants in two different fora.The data, that is the talk, was collected infocus groups and through participation inelectronic discussions on Usenet. Theparticipants from the focus groups in-cluded 50 persons – 34 men and 16women.2 Of the respondents, 30 reportedexperience with a mobile telephone,while the remaining 20 reported onlylimited experience. It is worth noting thatmen were far more likely to have re-ported experience with mobile tele-phones. The preponderance of men, ca.75 %, reported having used the deviceswhile only slightly more than 30 % ofthe women said the same.

The inclusion of information fromUsenet is somewhat outside the tradi-tional approach to social science and thusis worthy of special comment. In a way,this form of information gathering is atype of telephone survey in that the inter-action is of an anonymous nature. None-the-less, this approach is useful in thecollection of natural talk. The data usedhere was collected by posting a commentin the forum on restaurants asking for theopinions of other participants. The origi-nal comment clearly indicated that I wasinterested in a sociological analysis ofthe responses. Those who responded,four individuals, were thanked for their

input and their comments were added tothe database included here. The advan-tage of this approach is that one can eas-ily take preliminary hunches andconcepts, form them into additionalquestions and gather additional informa-tion on these specific questions. Thisinteraction with the data follows in thetradition of qualitative research, thoughthrough a new medium (Glazer andStraus 1967; Lofland and Lofland 1984;Spradley 1979).

It is important to be aware of the differ-ences between the data gathered throughthe two approaches. The first is thedegree to which informants felt pressureto respond. In the case of the electroni-cally gathered information, the forum inwhich the text was produced is open toall with Usenet access. Thus, there is nospecific responsibility to respond. Bycontrast, those who participate in focusgroups and in other interview situationsare participating in a somewhat unnaturalsetting where there is perhaps a certainsocial pressure to contribute. Thus, in thecomments from Usenet one might get theloudmouths where in focus groups onemay pressure people to formulate ideasthat they otherwise would not have con-sidered. Another aspect of the interactionon Usenet is that it is textual, not verbal.While this is a barrier to spontaneousremarks – that are often the most telling– it also means that the remarks are com-posed. That is, there is an element ofreflection in the messages (though somethat have read flame wars may be led towonder about the quality of the reflec-tion). A final difference is that the focusgroups were Norwegian where theUsenet input was more global, thoughdominated by the US. I mention thesecontrasts to clarify the data, not to assertone source as superior to another.

These two sources of data were exam-ined using standard qualitative analysistechniques. That is, the text was ex-amined, classified, re-examined andfurther classifications and conceptsformed the basis for the material reportedbelow (Spradley 1979; Glaser andStrauss 1967). It will become obvious tothe reader that the work of Goffman isdeeply integrated in this paper. I must bequick to point out, however, that this isan inductive analysis. The concepts andthe taxonomy presented below arose outof the data. This paper is not an attemptto prove Goffman, an impossible tasksince he left no hypotheses to prove ordisprove. Rather, Goffman’s insights,

along with others from the interactionistschool, have shown themselves to be use-ful for illuminating the concepts and forsetting them into relief.

2.2 Theoretical background:the contribution of Goffman

This paper makes use of Erving Goff-man’s contribution to sociology. It isGoffman, in the tradition of Mead andBlumer, who has focused on the drama-turgical aspects of every-day life. Thisapproach is particularly interesting in thathe often gained insight into the structureof society by examining the situationwhen things went wrong or when therewas stress. It is when our ability to main-tain façade comes into serious disarraythat we see people move to defend issuesof importance. In addition, we also see thestrategies used to deal with the situation.The results of the focus groups and otherinputs have underscored that the use ofmobile telephones presents us with astressing of the normal situation in society.

Before accepting Goffman’s work with-out question, however, one should alsoconsider its shortcomings. At the mostobvious level, many have criticized Goff-man for what is considered a lack ofscientific rigor. In the words of Giddens,Goffman’s popularity “derives morefrom a combination of an acute intelli-gence and a playful style than from a co-ordinated approach to social analysis”(1984, 68). Meyrowitz notes that Goff-man’s work is considered by some to be“a stylistic merger of the scholarly mono-graph and the novel” (1985, 32).3 Byway of a response, his work seems tohave been so evocative that it has beenand continues to be a rich source ofinspiration for others. Thus, in spite ofthe fact that his methodology is undocu-mented and his work is inaccessible, ithas inspired others. This speaks to hissuccess in observing and describingsocial interaction.

Another criticism is that Goffman fo-cused his analysis on only face-to-faceinteraction. There was little or no ana-lysis of mediated interaction in his work(Giddens 1984, 68-9; Meyrowitz 1985,32, 345 n. 53). Like the previous point,


3 Both of these theorists draw on Goff-man in their own work. Thus, thiscriticism seems to be more of a strawhorse upon which one can beat withimmunity.

2 The focus groups were conducted inNorwegian and the analysis of the talkwas largely undertaken before thematerial was translated into Englishfor inclusion in this paper.

this limitation has not prevented the re-application of his insights to mediatedsituations, or as in this case, hybrid situa-tions. His work has served as a point ofdeparture for examining of various formsof electronically mediated interaction.

Looking beyond the issue of style and thefailure to examine mediated interactionthere are two other criticisms leveled atGoffman. These include the sense Goff-man, and the interactionist school towhich he belongs, do not adequately de-scribe the reproduction of social struc-ture. In addition, these theorists tend toignore the idea of motivation by paintinga picture of society as populated by cyni-cal manipulators (Turner 1986, 459).

In relation to the first critique there is inGoffman an often implicit – but some-times explicit – assumption of society’sreflexive nature. Reflexivity refers to themutual assumptions one carries intointeraction that are built up and institu-tionalized over time. This is how inter-actionists, and to an even greater degreethose using the closely related sociologyof knowledge, describe the developmentand maintenance of social structure(Berger and Luckmann 1967, 47–128).For those using these perspectives,institutions, such as the way in which weeat dinner or the way we speak on thetelephone, are events that include thereciprocal typification of habitual actionsthat have built up over time. This mayseem like a ‘lite’ version of analysis forthose who see the structure of society interms of large institutions such as thechurch, capitalism or kinship. None-the-less, the Goffmanian approach does notpreclude the inclusion of these types offactors. Rather, it is more concerned withevery-day, micro-social issues.

The second point is well taken. To readGoffman is to read about petty schemes.One reads about people who try topresent the best face but are continuallyhampered by masks that do not fit cor-rectly, by irreverent children or naïveidiots who ask the wrong question or bythe mobile telephone that rings at thewrong time. According to this reading ofGoffman, if we were to take away thesecontrivances and face the naked ape as itwere, there would not be much to takestock of.

There really is more to us than just pettyfaçade management. After all, our veryinterest in reading Goffman, the empathythat we feel towards the ruses and char-

acters he describes, speaks to a morecomplete individual than the simple cyni-cal manipulator. In his defense, manydisciplines have their more or less com-plete stick figures (or one might callthem crash dummies) upon which theybuild their theories. There is the eco-nomic man of the economists, theSkinnerian response machine, etc. Eachis an invention to make a point. Goff-man’s creation is the same.

While accepting the criticisms outlinedhere, I find in Goffman a useful and anevocative access point to the analysis ofthe data that is presented below. At thesame time, I have been taken in. It is dif-ficult not to be. Reading his analysis isone of those exciting adventures asso-ciated with being social scientists. Theinsight, the wit, the analysis and the con-ceptualization all come together in a richand insightful analysis.

2.3 The dual nature of mobilephones

One last stop before the main analysiswill be to consider the social meaning ofthe mobile telephone. This device hassprung onto the scene over the lastdecade. In Norway, 40 % of the citizensuse a mobile telephone. The combinationof access to the devices, the developmentof an adequate network and various othermarket related aspects have been impor-tant in the growth of this industry (Bakke1996, 83; see also Haddon 1996b). Inaddition to the more traditional analysesof market forces, one can also examinethe rise of mobile telephony from theperspective of its social meaning. Whatdoes the mobile telephone symbolize? Inthis connection, the work of researchersat the University of Sussex in the UK isof interest. That of Silverstone and Had-don (1996) and Silverstone, Hirsch andMorley (1992) is particularly insightful.

When examining the social meaning ofnew technology, it is useful to considerhow the various devices and servicesrelate to pre-existing social structures. Sil-verstone et al. (1992) describe the dualnature of communication technologies. Onthe one hand, they are objects that occupyspace and can be displayed as an exampleof the user’s status and position in society.In addition to their status as objects of dis-play, they also have a use value. It is theformer in which I am particularly inte-rested. Silverstone et al. note that:

All technologies have the potential tobe appropriated into an aesthetic en-vironment (and all environments have,in some sense, an aesthetic). And manyare purchased as much for theirappearance and their compatibilitywith the dominant aesthetic rationalityof the home as for their functional sig-nificance (1992, 23).

The introduction of a new technologyinto what Silverstone and Haddon callthe moral economy of the home alsomeans that they must find their func-tional role. This role is not necessarilythat which the designers had intended.Rather, it is through the combined dis-play and use of the objects that the userscome to understand the devices and it isthrough this process that the users weavethem into a part of their own identity(Silverstone and Haddon 1996, 58, seealso Marx 1994).

One might, for the moment, consider themobile telephone a type of jewelry like abelt buckle, a watch, or a broach. That is,the device can have functional features aswell as being a type of display. It can beappreciated or scorned. It can be theobject of desire and interest or it can beseen as a symbol of vulgarity and badtaste. In order for the display to besuccessful, however, it must be done atthe appropriate time, in the appropriateway with the appropriate élan.4 Thedifference between a mobile telephoneand a necklace, however, is that theformer can be set into life by otherswho call in at the wrong time.

There are two final points to be madehere. The first is that the adoption of newtechnology is a conservative process inthat there is a prerequisite to incorporatethe object into users’ daily routines.There is in this process, however, a sensein that one can not go back home. Uponthe domestication of the object, lifechanges, the new object and theirattached routines nudge their way intothe daily life of the user in both obviousand in not so obvious ways (Silverstoneand Haddon 1996, 60; Stone 1994).


4 These comments are reminiscent of thedevelopment of a social norm pro-scribing against the use of digitalwatches that ‘peep’ each hour. Whenfirst on the market, many people usedthe peep function causing a discussionaround their use during films andother public gatherings.

Finally, there is a focus on the domesti-cation of devices such as the TV or thePC within the home in the work outlinedabove. A mobile telephone is slightlydifferent since the device is, almost bydefinition, individual and not attached toa physical location. Thus, some of theaspects of the moral economy of thehousehold need to be seen in a slightlybroader form. It may be that the deviceis only used internally within the family,at the cabin or as an internal messagingsystem. Apart from these alternatives,however, mobile telephones are not nec-essarily the same type of display or useobject as, for example, a TV or a PC.This is expressed in both the use of theobject but also in the reorientation of theuser to the object itself.

3 Inappropriate use ofmobile telephones

Now it is time to turn to the use and mis-use of mobile telephones. The data fromthe focus groups indicated an almostvisceral reaction to the inappropriate useof mobile telephones. The respondentswere able to offer clearly formulated andwell-rehearsed verbal sanctions againstthose who used mobile telephones in-appropriately. At a broad level, the factthat these formulations were readily athand indicates that the provocation is not

at the level of the individual rather it is atthe social level. In addition, it is evidenceof the debate at a somewhat broadersocial level over the norms of telephoneuse. It indicates that we are in the processof sorting out our collective sense of howto deal with this issue (Haddon 1996a).

The respondents in the focus groups werequick to point out various situationswhere they felt it was inappropriate touse mobile telephones. These includedairports, stores, meetings, on trains andbusses, at various social functions and intheaters.

It irritates me to go to airports and Ithink it is embarrassing and very pre-tentious to sit there with that tele-phone. Sitting there with a cup ofcoffee with a man beside you with aPC in his lap. It simply irritates me.I think it is unreasonable.

Once [my husband] had [a mobiletelephone] in to town and we were inthe town and were shopping and rightas we came out of a store the tele-phone rang. I thought it was very dis-gusting to stand there outside the storeand talk and he thought he was goingto be so big and because and I sawthat. I think that was very disgusting.

Last week I was in a meeting and therewere 10 or 12 in the meeting. Four of

them had a mobile telephone. Therewas shuttle traffic out to the hallway.That is unacceptable.

I was at a Christmas party the otherevening and there was a lady who flewin and out of the ring [where childrensing Christmas carols] 3–4 times.I cannot understand why one will havea mobile telephone at a Christmasparty. And this with meetings, thereare five men each with their ownmobile telephones. They ring justabout continually, there should be arule that one leaves their mobile tele-phone and pager in the reception area.

I think you have to be considerate whenyou have a mobile telephone. Forexample, that you don’t have it onwhen you are at the movies. It is notso nice when they begin to ring in themovies, it is not so pleasant and theaterand several other places where youshould not call. I think that people needto be considerate and turn them off.

In their defense, many users of mobiletelephones understand that its use is notappropriate in all situations. Somerespondents noted that it was embarrass-ing if the telephone rang in certain situa-tions. Others were sensitive to the needto turn off the device when they wereotherwise engaged.

If I had it I would have turned it offnow. It is not that important. In cine-mas and such places as now, I wouldhave turned it off. One can say thatthere are taboo areas.

The discussion here illustrates the de-velopment of a social norm proscribingthe areas in which it is appropriate to usemobile telephones.

3.1 Restaurants as a specialsocial situation

In addition to the situations outlinedabove, a common theme in among therespondents in the focus groups was theirritation over the use of mobile tele-phones in restaurants. To understand thisirritation it is perhaps useful to examinethe social meaning of eating at a restau-rant. There are two elements that are ofinterest here: the territorial nature ofrestaurants and Goffman’s idea of face.

3.1.1 Boundaries in restaurants

Some of the things that make a restaurantspecial from a social perspective are that


Figure 2 The fact that mobile telephones allow for personal communication from andto many locations mean that third parties have audio access to our conversations. Theconsequent breaching of social boundaries has a resonance for Norwegians who havedeveloped a sense of ‘managed unreachability’ as a basic cultural trait

LAR

S B

AH

L / 2

MA

J / S

AM

FO

TO

one’s use is temporary and it has ele-ments of being both a public and aprivate space (Lipman 1967; Mars andNicod 1984, 66-69; Silverstone 1996).The tables, booths and the portions of thelunch counter occupied by the patronsbecome theirs for the period of occu-pancy.5 To claim a territory, however,we must go through the rituals of estab-lishing and agreeing upon illusoryperimeters or ‘symbolic fences’ with ourfellow patrons (Gullestad 1994).

Some [territories] are ‘situational’;they are part of the fixed equipment inthe setting (whether publicly or pri-vately owned), but are made availableto the populace in the form of claimedgoods while-in-use. Temporary ten-ancy is perceived to be involved, mea-sured in seconds, minutes or hours,informally exerted, raising constantquestions as to when it terminates.Park benches and restaurant tablesare examples (Goffman 1971, 29).

We become quite accomplished at ignor-ing others who are in quite close proxim-ity through the use of a fictive curtainbetween tables that are, in reality, quiteclose to each other or even the very sametable. These barriers allow each diningparty to maintain the notion of a type ofprivacy that is, more or less, open for allto see and hear.

Goffman goes on to note that there areoften demarcations, based on architectureand the placement of furniture that stabi-lize claims to space. These conventionsallow for the erection of easily observ-able boundaries between individualsand parties on an ad hoc basis (Goffman1971, 33–34). In this way, an individualor group can temporarily identify whereit is that they belong and also determinetheir expectations as to where otherpatrons have the right to intrude. Thewaiters and waitresses, of course, enjoya certain freedom of movement withinthrough this complex of fiefdoms.

Within the stability there is also aflexible nature to these settings. Tablescan be divided or combined in order toaccommodate varying numbers of partic-ipants, groups can occupy adjoiningtables or counter space and commandeer

the area in between, etc. This notion offlexibility varies, however, with the classof the restaurant (Mars and Nicod 1987,49). In exclusive establishments thetables are only adjusted ‘back stage’, thatis before the arrival of the customers.This minimizes the need for the patronsto readjust their ‘borders’ with theentrance of other claimants. In other,more proletarian restaurants the cus-tomers often take the initiative to divideand re-divide the space as the needarises. While this allows for a more con-centrated complex of groups – somethingin which the proprietor is interested – italso means that there is the need to adjustclaims to territory, carry out cross borderraids when the need arises for extrachairs or condiments, and in the worstcase, it may mean wars of attrition whentwo parties lay claim to the same terri-tory.6

The notion of social boundaries has aparticular resonance for Norwegians.Social anthropologists have identifiedwhat they call “managed unreachability”as a basic cultural trait. While seen withforeign eyes, behavior may seem coldand aloof. When judged from a Nor-wegian perspective, however, this is seenin a positive light. The management ofcontact with others in public situations,and more privately, is seen as a way toprotect a sense of self (Gullestad 1994,167; see also Haugen 1983).

3.1.2 Maintenance of face

The second aspect of interest here is thatof façade or ‘face’. Goffman suggeststhat one is maintaining their face whentheir presentation and management ofself is internally consistent. The develop-ment and maintenance of face is not asolitary activity, rather it is a social pro-duction. The face one presents is only asgood as the willingness of the audienceto treat it as real. Thus, façades are acrystallization of our willingness to beintegrated into society (Goffman 1967,7). As Goffman notes:

A person’s performance of face-work,extended by his tacit agreement to helpothers perform theirs, represents hiswillingness to abide by the groundrules of social interaction. Here is thehallmark of socialization. If he andothers were not socialized in this way,interaction in most societies and mostsituations would be a much more haz-ardous thing for feelings and faces(1967, 31).

Face helps us to integrate ourselves insociety, since we base and adjust our pre-sentations on our perceptions of the situ-ation. Without this, behavior would spinoff in unpredictable directions and com-mon intersubjective understandings ofthe interaction would be impossible.

If one considers a restaurant in this con-text, it can be seen as a dynamic stageupon which one’s façade is displayed. Itis a special situation where one is askedto combine etiquette and social finesse. Itis where there is often a demand to cele-brate social unity with one’s family,friends or colleagues. At the same time,there are many unexpected turns andtwists that a restaurant visit can take.There is a well prescribed set of rulesand rituals that must be observed – thecorrect use of utensils, the way in whichone eats, the topics that are available forconversation, etc. There are many possi-bilities for personal adventure – meetingthat special someone in a singles’ bar,closing an important deal, savoring aspecially cooked meal, celebrating abirthday or other significant transition,etc. Finally, there are many potential haz-ards – spilling the wine or the soup, mak-ing inappropriate remarks, the need tocover over the fact that one received apoorly cooked meal, having to deal withUncle Fred after he has had too much todrink, meeting others who one had hopedto avoid in that situation, using the wrong


5 For long term patrons, certain posi-tions may become ‘theirs’ wheneverthey are present. To dispute this claimcan result in severe sanctioning (Lip-man 1967).

6 Singles’ bars and restaurants whereone eats in order ‘to be seen’ are in aclass for themselves. In the singles’bar, the barriers are there to beopened and closed as the attentions ofothers are sought or are to be avoided(Collas 1995; Giuffre and Williams1994; Haavio-Mannila and Snicker1980; Parker 1988). Common openingploys – the purchase of a drink by aremote admirer, the use of the line“haven’t we met somewhere before”,etc. – are both inquiries as to the open-ness of the local landscape. In the caseof restaurants where one appears to beseen, one may even use strategies tounderscore their presence, that is theyopen themselves for observation, i.e.the grand entrance or sending amessage to the management to pagethemselves over the PA system. Thislatter strategy seems somewhat similarto the characterization of the patronwho uses the mobile telephone only asa way to draw attention to themself.

knife and thus indicating unfamiliaritywith ‘correct behavior’, etc. (Chen1990–91; Fine 1995; Giuffre and Wil-liams 1994; Parker 1988). All of thesedemands require that one becomes fluentin the maintenance of face.

One can see that eating at a restaurant isan important social performance by ex-amining the rules of etiquette (Jackson1952, 325; Duncan 1970, 266–69). Inevery culture, the rules of eating decorumare extensive and often include quite pre-cise orderings of events and prescriptionsof the exact placement of eating equip-ment, drink and food (Mars and Nicod1987, 28–29; Rombauer and Becker1964, 9–25). Manners also indicate statusand hierarchy. It is through our use ofmanners that we indicate the way inwhich we expect to be treated. Thus,holding the teacup with the little fingerextended in the proper way is in thewords of Duncan, a “dramatization ofthe self” and a message to others that weexpect a certain type of treatment (1970,266). Geertz also comments on the wayin which manners indicate status. Henotes that etiquette is a recripocially builtbarrier that surrounds the individual.Etiquette protects the inner stability ofthe occupant and, as one goes up thesocial ladder, “the thicker the wall ofetiquette protecting the emotional life”(1972, 290; see also Gullestad 1992,165). All of this attention to the processof eating, particularly when it is done inpublic, speaks to the importance of theevent, as well as to the potential for prob-lems. It also speaks to the importance ofreaffirming the social order (Cahill 1990,391).

The final aspect of face in restaurants isthat a restaurant visit is an event with aclearly visible price tag. All participantsknow, with some clarity, the cost of eat-ing out; it is listed for all to see on themenu. Implied in this is the assertion ofstatus and the idea of gifting. Thus, to eatout is to give evidence as to one’s posi-tion in society. To invite one to a restau-rant is, in some respects, to set a price ona social relationship. In a similar way, toaccept such an invitation has the implica-tion of indebting one to their host. Thus,untoward things that disturb the experi-ence, such as the ringing of a mobile tele-phone, not only disrupts the work ofmaintaining façades, but can also de-preciate the exchange between the diningpartners. This sentiment was quite com-mon among the informants who noted,for example “I have paid a lot of money

for that meal” and “[Hearing a mobilephone in a restaurant] can make you feellike you’ve wasted money and made abad choice about what you did thatevening.”

Thus, the development and maintenanceof face in a restaurant is a delicate pro-cess. While there are benefits to beenjoyed by its successful achievement,there are also perils. The importance ofthe event is underscored by the elaboraterituals of etiquette associated which havedeveloped and by the clearly visible eco-nomic aspect of the event.

3.2 Use of telephones inrestaurants

Against this backdrop it is not surprisingthat respondents in the focus groups citedthe use of a mobile telephone in restau-rants as a particularly galling social vio-lation.

It is awful to see those who go around,I could throw up! They use it even inrestaurants, it looks so dumb.

One can talk about common manners!When I was out at Mat & Vinhus [alocal restaurant], it costs a lot ofmoney to eat there, and the mobiletelephone rang for somebody sittingbehind us that had a lot of time to talkand talk. That is not good manners,I mean he could have said, just onesecond and gone out.

I think it is repulsive to sit there andtalk with people on a mobile telephonein a restaurant.

With a point of departure in Goffman’sconcept of ‘face’ and that of temporary‘territoriality’ one can begin to examinewhy the use of mobile telephones is sounpalatable. The most basic objection tomobile telephones has to do with thesounds associated with their use. Whendiscussing the ‘modalities of violation’of personal space, Goffman brings up‘sound interference’. In this case theviolator fills up his or her accorded spaceand then some. One can violate the terri-tory of others by carrying out an en-counter over a longer than proper dis-tance and thus ‘obtruding’ into the socialspace of others (1971, 33–34, 51). Anaspect of this discussion that Goffmanneglects is the type of sound one makes.There are a whole family of sounds thatare inappropriate, particularly in restau-rants. The unwilling belch or flatulenceare only the tip of the iceberg. Any par-

ent or member of a fraternity knows thevariety and imaginativeness with whichsome diners can willingly create ‘inap-propriate’ noises. The ringing of a mobiletelephone fits into this family of inappro-priate sounds, almost regardless of itsvolume.7

In restaurants where it is the most diffi-cult to make a long term claim to space,i.e. cafeterias, it is the easiest to use amobile telephone. That is, a certain clashof personal space is taken for grantedand people are called upon to managebarriers most actively. In addition, thebackground noise may be of sufficientvolume to cover over the mobile tele-phone conversation. On the other hand,in those restaurants where the boundariesbetween parties are the easiest to define,i.e. more formal restaurants where theindividual is able to claim a table for awhole evening at the cost of a large bill,there is the assumption that the restau-rant, and the other patrons, take responsi-bility for barrier maintenance.

3.2.1 Ringing

Several respondents noted the abruptnesswith which the mobile telephones rang.One person said that “They sit on bussesand trams and on street corners and inrestaurants and the telephones peep andpeep”. The most abrupt sound producedby the mobile telephone is often theirringing. The peeping or ringing is bynature an intrusive sound not unlike thatof an alarm clock. One informant noted


7 It needs to be noted that the accept-ability of the mobile telephone is some-what place dependent. In a restaurantwhere higher levels of noise are part ofthe setting and where one is notexpected to treat their attendance as aparticularly special occasion, a tele-phone is more acceptable because itsuse is covered by other activities. Onecan include children’s restaurants, fastfood restaurants and informal cafesand eating locales in, for example,shopping centers where there is whatone might call active backgroundsound, i.e. fountains or musak. On theother hand, in situations where one isexpected to attend to the here and now,i.e. the vintage of the wine, the topic ofconversation, the quality of the food,the look in her or his eye, etc., then thedirection of attention to a telephonicpartner would be a greater intrusion.

“The ring-ring of a phone disturbs themood. Distracts the diner.” While satis-fying the demand to alert the attention ofthe user, the ringing can test one’s abilityto manage the social situation.

Many other restaurant patrons havedeveloped a set of responses with whichthey can cover over unexpected sounds.The crashing of glasses or plates isparried with the use of a smile; the inter-ruption of a child is dealt with eitherthrough forbearance or a short scolding;a belch is either ignored or laughed at;the abrupt arrival of the waiter is tole-rated and the ringing of a traditional tele-phone in a restaurant is ignored since it isnot a signal of importance to theparticipants. The adroitness with whichone can smooth over such social roughspots was called face work by Goffman.

Each person, subculture, and societyseems to have its own repertoire offace saving practices. It is to thisrepertoire that people partly referwhen they ask what a person or a cul-ture is ‘really’ like. And yet the partic-ular set of practices stressed by partic-ular persons or groups seem to bedrawn from a single logically coherentframework of possible practices. It isas though face, by its vary nature canonly be saved in a certain number ofways. (Goffman 1967, 12–13).

Through the use of poise and savoir-faireone minimizes the damage of embarrass-ing incidents to the social situation. Itseems, however, that the ringing ofmobile telephones is not adequatelyroutinized such that their ringing causesdifficulties in maintaining face. There is,in addition, a complicating factor withmobile telephones. The ringing is oftensuch that it is not just the intended calledwho must interrupt their conversation toanswer the phone. Rather, the growingpervasiveness of mobile telephonesmeans that the ringing occasions theneed for all of those with a mobile phoneto check if theirs is ringing.

Finally, the ringing of a phone in an in-appropriate situation draws attention tothat which follows, namely a telephoneconversation. The phone conversationwill, in effect, take the user out of thesocial context in which he or she was andplace them into another. Thus, the ring-ing alerts others that the discussionwhich follows will be different in toneand subject.

3.2.2 Loud talk

The second type of sound that was irri-tating to the respondents was that of theperson talking on the phone. It was com-mon for the informants report that thosewho used a mobile phone employed‘loud talk’. One noted that: “People talkloud on telephones. Louder than usual, atleast. That is as annoying as having aloud party next to your table.” Thisbrashness violates territories and makesit difficult to maintain the face:

When you have a loud person talkinginto a [mobile] phone and you (can’thelp but) hear only half of the conver-sation, that disrupts the coziness of therestaurant feeling.

While many attribute the volume of talkto a desire for display or a sign of vul-garity, one can also examine this issuefrom the perspective of how the technol-ogy effects our form and style of interac-tion. At one level, when speaking on thephone there is the need to replace thevisual gesturing of face-to-face commu-nication with a form of verbal gesturing.The telephone conversation is a speechevent with distinguishing characteristics.Several of these break with the style andtone of the speech events that are com-mon in face-to-face communication. Inface-to-face conversation quite nuancedbody language has several functions.Through our use of nods, glances, smallsounds and other gestures we indicateattention, the desire to speak, the desireto retain the floor and indicate pauses.We also use these devices to impartmeaning and emphasis. All of thesegestures are changed in a normal tele-phone conversation. Visual gesturesare replaced by intonation and linguisticstructure in ‘grounding’ the conversation(Rutter 1987, 105, 126; Martin 1991,95–97; see also Duncan 1972, and Saks,Shegloff and Jefferson, 1974). Instead ofrelying on body language to control turntaking, pauses, emphasis, etc., these aredone with what one might call verbalgestures. We use tones such as ‘uh’ toreplace the lack of eye-contact that con-trols turn taking, phrases such as ‘ah ha’replace nodding and other signals of con-tinued attention on the part of the lis-tener, etc.

In addition to the different style of talk,one is often prompted to speak louderwhen using the phone. This is because ofthe need to be heard over backgroundnoise on the line and the general social-

ization that one speaks up on the tele-phone (Martin 1991, 95–97).8

Thus, the need to include verbal gesturesand the habit of speaking up on the phone,in addition to the fact that the wholeepisode is announced by the ringing of thedevice, makes it easy for others to identifythe talk as a telephone conversation. If thelistener has preconceived problems withthe use of telephones in restaurants theease of identifying the talk allows them toplay out their prejudices.

3.2.3 The management of parallelfront stages

Respondents talked about considerationsthat went beyond the directly audibleaspects of mobile telephony. The firstof these is that the combination of theprivate telephone call and the privatedinner conversation means that one mustjuggle two parallel interactions. Whenone begins to talk on a telephone, theyhave, at least partially, removed them-selves form the ongoing interaction intheir physical location (Clark 1995). Oneinformant in the focus groups expressedan annoyance with this potential. Thisperson noted that “It is a little irritatingwhen ... you sit together with someoneand they cannot set [the mobile phone]down.”

While it is possible to carry on parallelactivity while eating at a restaurant orwhile speaking, these possibilities arelimited (Ling 1994; Lohan 1996a). Theintrusion of a mobile telephone callthreatens the pre-existing situation. Atthe first level, one must choose whichconversation takes precedence. Goffmancalls this accreditation.

Messages that are not part of the offi-cially accredited flow are modulatedso as to not interfere seriously with theaccredited messages. Nearby personswho are not participants visibly desistin some way from exploiting their com-munication position and so modifytheir own communication, if any, so asto provide interference (Goffman1967, 35).


8 According to Bakke (1996) we expect acertain level of background noise fromthe phone. If this is not to be heard onefeels that the phone is dead and itreduces the sense of co-presence. Thishas led to the use of ‘comfort noise’ indigital systems that are, by nature,more free from background noise.

If the telephone call receives accredita-tion the mobile phone user has in effecttwo front stages,9 that in the physicalrestaurant and the telephone conversa-tion.10 Quite often, the behavior appro-priate in the one situation is not appropri-ate in the other. The different relationshipthat one has with one’s dining partnersand one’s telephone partner means thatthe negotiation of topics, the depth, pas-sion and emotion with which one canaddress the one party and the range ofcommon understandings will probablynot be the same for the other (Garfinkle1967, 35–75). The shifting front stagesfor the phone user makes it possible forothers to see the user of the mobile tele-phone in a type of verbal cubism. Whilethe face-to-face restaurant talk may be,for example, cozy, intimate and integra-tive, the talk on the mobile phone may beof power relations, fast deals and officepolitics. Another example of this that hasbeen seen for its comic value is whenone’s intimate dinner with a lover isinterrupted by a call from an irate spouse.The stage management can become quitecomplex. Like a cubist painting, thespeaker on the mobile phone is seen fromtwo perspectives. There is a certain dis-sonance when judged from the perspec-tive of classical notions of talk, dialogueand narration that present a single per-spective to the audience.

In order to manage the stage one canerect various types of partitions. One can,for example, temporarily move from the‘front region’ of the restaurant to a recep-tion area or other more removed locationin order to allow the harmonious comple-tion of the call. This strategy is, however,only partially successful if the offense tothe decorum was done by the ringing ofthe telephone in the first place. A secondstrategy discussed by Goffman is that theoriginal audience be granted a temporaryback stage status (Goffman 1959, 139).Thus, they will be allowed insight intothe staging of another performance. Inthe short term, this allows the completionof the telephone call: however, oncebelief in the first performance is sus-pended, rebuilding its façade may be adifficult proposition, particularly if theoriginal audience is only begrudginglywilling to accept its back stage status.

During the period of the call the diningpartners are left in a particularly stressfulsort of suspended status in that they areasked to wait. They are not dismissed,rather, they are left hanging. Only afterthe conclusion of the call can theyresume of their earlier status. Whilewaiting they must engage themselvesin some type of waiting strategy that iseasily discarded when the other summonsthem back. This is a particularly difficultsocial juxtaposition.

One possibility is to somehow acknowl-edge their presence to the caller – i.e.the mobile telephone user indicates thepresence of the third person and mayeven make open side comments to themor, in some cases, the third person mayeven make open side comments to thecaller and the called. Another possibilityis that the third person may be called onto indicate that they are not engaged inthe conversation at all (Lohan 1996a).An informant described this latter situa-tion for a daily restaurant patron and afrequent mobile telephone user. “Healways has a companion with him who isleft to peruse the alphabetized beer bottlecollection behind the bar for five or tenminutes.” Here, the companion is left tomaintain his own illusion of indiffe-rence.11 This strategy for dealing withthe situation will be discussed imme-diately below.

3.2.4 Coerced eavesdropping12

Another aspect of boundary managementis the issue of eavesdropping. The com-mon understanding of this problem is

that those inside a conversation fear thatthose outside the conversation canbecome privy to secrets. There was aslightly different perspective reportedhere in that the informants feared thatthey would hear too much, a sort of co-erced eavesdropping. The need to guardagainst this was seen as a problem.

Several of the respondents noted that theaudience to a mobile phone conversationin the restaurant is not only one’s imme-diate dining partners, but also those whoare within hearing distance of the con-versation. The placeless nature of thecall (from the perspective of the remotecaller) means that the talk may take upprivate issues not intended for others tohear. The information exchanged can be,or perhaps more importantly, can appearto be revealing. Thus, the audience isprovided involuntary access to portionsof the phone user’s life, a clear violationof Gullestad’s managed unreachability(Gullestad 1994). Many of the respon-dents talked about unwittingly eaves-dropping in on other’s telephone conver-sations. In the words of one respondent:

I think it is fine that the telephone isstationary at home. A conversationshould be between two persons. I thinkit is unfortunate that others are there.

[Hearing a mobile telephone at arestaurant] breaks up the enclosedsocial circle that a restaurant offers.Seeing people communicate and besocial at different tables is part of therestaurant experience.

In other cases the topic may be judged tobe superfluous as in the case reported byone informant “[The user of a mobilephone] yaks away about (obviously)trivial matters.” Regardless, the need toguard against hearing too much wasreported to have been disruptive.


11There are echoes of Simmel’s discus-sion of the stranger in this situation(Simmel 1971, 143–149; Bogard,1996).

12The word eavesdrop refers to theground upon which rainwater dripsfrom the eaves of a house. Presumablyif one were to stand inside the eaves-drop they become privy to the commu-nication inside the house.

9 A later formulation was developed byGoffman in which he presents the ideaof frames (1974). Framing refers tothat set of conventions that define‘what is going on here’. While thereare many alternative possibilities,social interaction requires that theparticipants in an interaction agree, atleast generally on the same notion ofan interaction’s content and form. Thisagreement frames the activity (Gid-dens 1984, 89–90).

10This is particularly true if the mobiletelephone user has received the call. Ifhe or she has initiated the call they canat least prepare their dining partnersfor the event, and they also avoid thenoise of the ringing telephone. It isalso worth noting that these two frontstages are not necessarily mutuallyexclusive. It is, in fact quite possiblefor others to participate quasi-activelyin a telephone conversation. This isparticularly true if the caller, thecallee and the third person are allfamiliar with each other and are per-haps planning a common occasion(Lohan 1996a).

3.2.5 The mobile phone as avulgar metaphor

The fact that one carries on a telephoneconversation may have become an in-appropriate gesture in its entirety. Againand again the informants went beyond ananalysis of the component elements andencapsulated the whole event by de-scribing it as rude or disconcerting.

Talking on a cell phone at dinner, inmovies, (or at a seminar (yes, I’ve seenit)) is just plain rude. My family neverallowed telephone calls at the dinnerhour in my home, because when peopleare sharing a meal, it is the time tocommunicate and socialize. (emphasisadded)

I think it is a little impolite whenpeople stand there and bellow in amobile telephone ... or me it is notgood manners. It is like standing thereand whispering to somebody. That isnot polite either.

There was not a discussion of the mecha-nics of body language or the non-con-textual nature of the conversation, rather,there was a reaction to the event as awhole. Almost as drinking coffee from asaucer, holding a wine glass from thebase, eating from one’s knife, or in somecultures eating with one’s left hand havebecome a gloss or a characterization of acertain type of person, the use of amobile telephone in a restaurant hasbecome a metaphor for the vulgarbreaching of manners.

3.3 Management strategies

Once the breach has been made, how doother patrons respond? When face hasbeen threatened, Goffman refers to theneed to undertake ‘face work’.

By face-work I mean to designate theactions taken by a person to makewhatever he is doing consistent withface. Face work serves to counteract‘incidents’ – that is, events whoseeffective symbolic implicationsthreaten face. Thus poise is oneimportant type of face work, forthrough poise the person controlshis embarrassment and hence theembarrassment that he and othersmight have over his embarrassment(Goffman 1967, 12–13).

The mutual management of embarrass-ment illustrates the integrative functionof face work. This is what Berger and

Luckmann see as the reciprocal nature ofsociety (1967, 47–91). In this way “onehas an interest in a defensive orientationtowards saving his own face and a pro-tective orientation saving the face of oth-ers” (Goffman 1967, 14). Thus, we arelikely to ignore minor infractions in thespirit of the occasion.

There are other situations, however, thatgo beyond the simple management ofembarrassment. Setting this type of a sit-uation aright is an important issue sinceits continuation can threaten the situationand can insult our sense of identity. Dun-can, in fact, asserts that the flagrantignoring of manners can be seen as aserious threat to order.

Anger over ill manners of others arisesout of the belief that not following ourmanners is a way of telling us that weare not really important in the eyes ofthe transgressor. We excuse a faux pasmade out of ignorance (and soon cor-rected) because we still feel the im-portance of our manners as a socialbond. We laugh at comic depictionsof vulgarity so long as the majesty ofwhat we hold important is notthreatened. But we do not laugh atsavage ridicule or continued vulgarity,because they endanger the socialprinciple upon which our mannersare based (Duncan 1970, 267).

When a breach has been made and thereis the need for face work, the audiencehas several responses available to them.At the simplest level Goffman suggeststhat one employs studied non-obser-vance. Beyond this one might challengethe offender or vilify the offender inabstentia.

3.3.1 Civil inattention

One of the most common strategies fordealing with a relatively minor offense,such as talking on a mobile telephone, isto engage in what Goffman calls ‘civilinattention’ (Goffman 1963, 85–86). Thisis described as the willingness to turn ablind eye towards behavior that repre-sents a potential threat to face. It is anattempt to gloss over the disturbanceand to carry on.

Via socialization in manners and courte-sy, we understand the intrusiveness pay-ing attention to the conversations of oth-ers. Thus, the public use of a mobile tele-phone tests the courtesy of patrons in arestaurant. As discussed in the section oneavesdropping, those outside the conver-sation may fear that they will hear toomuch. The reaction is to become em-barrassed for the potential embarrass-ment of the telephone user. Thus, wemust parry our own interest and replacethat with a set of staged stances that aresocially defined, e.g. making a display of


Figure 3 The use of the mobile telephone in restaurants by employees, but moreimportantly by patrons, has become a metaphor for the vulgar use of the device

LAR

S B

AH

L / 2

MA

J / S

AM

FO

TO

reading the menu, looking at the decor,suddenly becoming engrossed in thepapers one is carrying in their hand or, asreported above, looking at beer bottlecollections. This is studied non-obser-vance. As the mobile telephone becomesnormalized we will likely need to de-velop a repertoire of suitable inattentivepostures which we can assume.

The goal of this is to not disturb thescene.

The combined effect of the rule of self-respect and the rule of consideratenessis that the person tends to conducthimself during an encounter so as tomaintain both his own face and theface of the other participants ... A statewhere everybody temporarily acceptseveryone else’s line is established.This kind of mutual acceptance seemsto be a basic structural feature ofinteraction, especially the interactionof face-to-face talk (Goffman 1967,11).

The fact that we generally give no out-ward sign of reacting to the use of mobiletelephones indicates that the offense isnot as significant as others.

3.3.2 Challenges and responses

Beyond civil inattention one has theability to sanction others using varioustypes of challenges. In the case of“Extreme impropriety ... is likely toresult in his being stared at or studiouslynot seen” (Goffman 1963, 87).

In these cases one may make a transitionfrom civil inattention to the slightlystronger studied non-observance (1967,17–18). The differences between the twoseem to be the degree to which the ob-server acknowledges the existence of theobserved. Civil inattention includes theidea of a token recognition followed byfocusing on other matters. By contrast,studied non-observance seems to be mak-ing a great display of refusing toacknowledge that the observed exists.The former seems to be more of astrategy of interaction for those who aremore or less in agreement on the agendawhile the latter may indicate more of adisagreement as the nature of the setting.Civil inattention is a courtesy and studiednon-observance is a negative sanction.

Beyond these relatively discrete sanc-tions one can also employ strategies suchas glaring, sighs, poorly disguised com-

ments etc. Further steps include directconfrontations, ask the waiter or otherauthority to intercede, and the like. Thesestrategies may or may not succeed. If theoffender refuses to curtail their behaviorin the face of the sanctions, then the ballmoves back to the person who issued thesanction. It may be that their face hasbeen put into question. This means thateither they are left to bluster, they canretaliate or they “can withdraw from theundertaking in a visible huff-righteouslyindignant, outraged, but confident of ulti-mate vindication” (Goffman 1967,22–23).

There is obviously a balancing point inall of this. If one begins to make more ofa disturbance in their sanctioning thanthe situation allows, they may becomethe object of sanctions themselves.

Improper conduct, however, does notautomatically release others from theobligation of extending civil inatten-tion to the offender, although it oftenweakens it. In any case, civil inatten-tion may be extended in the face ofoffensiveness simply as an act of tact-fulness, to keep an orderly appearancein the situation in spite of what ishappening (Goffman 1963, 87).

3.3.3 Attribution

Judging from the responses of infor-mants, a final strategy for dealing withthe use of mobile telephones in restau-rants and other public places seems to beex post facto stereotyping of the offender.While not addressing the situation on thespot, these stereotypes help to form ageneral attitude toward mobile telephoneuse. These are the building blocks in therise of new social norms. They are theformulation and rehearsal of preventativesanctions. It is as if to say “If you do this,people like me will have a poor opinionof you”. This follows in the tradition ofGluckmann (1963) who notes that infor-mal discussion and gossip can functionto hold a group together.

The stereotypes or attributions used byinformants included the notion of over-important status display, the pointless-ness of the talk and finally, illusions as tothe character of the offender. In termsof the first, it was common to hear thatthose who use mobile phones inappropri-ately were showing off or that they weresome type of a status seeking yuppie(Haddon 1996a, 3).13

I know a patron at my lunch time spotwho apparently hires a school kid tocall him at set intervals during lunchevery time he is there, lest we be (bliss-fully!) unaware of his being the proudowner of a mobile phone.

The second type of attribution had to dowith the non-instrumental nature of thediscussions. This type of an assertion iswell-grounded in commentaries from anearlier era of the telephone. When systemcapacity was limited and calls wereexpensive social talk was discouraged(Ling 1994, 13–14). Those who used thetelephone for inappropriate calls werelikely to be seen as gossips or wags. Sim-ilar glosses were applied to those whomade open use of mobile phones. In thewords of one informant.

One sees people standing around onthe street and there is a lot of baloneythat is said there. Unnecessary discus-sions so I don’t see the point of it.

Finally, some respondents went beyondrather neutral descriptions to the asser-tion that the mobile telephone was adevice used to facilitate illicit activity.This is seen in the following sequenceof comments.

Person 1: It is a fact that more joblesshave a mobile telephone now thanthose who have a job. And a largeportion of these are as a matter of factimmigrants.

Person 2: But how is it that you saythat the jobless teens and youth havemobile telephones?

Person 1: Jobless on paper. I wouldnot say they are.

Person 3: No.

Person 1: I would suggest that thereare many pushers that have mobiletelephones ... Criminals, real crimi-nals.

Person 3: Oh, that doesn’t sound toogood.

Person 1: No, but it is true. I comefrom Oppsal and there is big timecrime and there are all the worsttypes and they go around with mobiletelephones. They call all the time andthere is no let up.


13In a brochure published by Telenordescribing the use of mobile tele-phones in restaurants, there is the title“It is only dumb to be a Yuppie sevenyears too late.”

Here the use of a mobile telephone is notjust a small social faux pas, rather it hasbecome a symptom of, in the speaker’smind, dangerous social trends, i.e. immi-gration and drug use.

Regardless of the stereotype applied tothe offender, the application of a glossis an attempt to make sense of the wayin which technology is being used. Thenecessity for this is being pressed uponus by the growing use of these techno-logies. In many respects, one can notremain neutral. They must take up a posi-tion on one or the other side of the barri-cade. It is here that the social bond is indanger of becoming unraveled, and it ishere that the most intense strategies ofsocial maintenance need to be deployed.

4 Conclusion

In this paper, based on the examinationof focus group data, I have examinedhow people make sense of the inappro-priate use of mobile telephones, particu-larly in restaurants. I have looked at thesocial meaning of mobile phones andrestaurants. In addition, I have lookedat how the use of the former in the latterpresents both the user and other patronswith a difficult social situation. The anal-ysis has drawn on the work of Goffman’snotion of drama and staging in order toplace the analysis in a larger framework.Goffman’s concepts of social boundaries,face, front stage / back stage, civil inat-tention, studied non-observance andsanctioning via attribution have beenused.

This paper is a further examination of theways in which technology has shiftedsocial boundaries. Devices such as thevideo telephone and the Internet havemeant that we need to reconsider how itis that we construct our social worlds.They have made demands on the taken-for-granted assumptions of everyday life.These developments mean that we cancommunicate in new ways at new timesin new places. While there are un-imagined possibilities there are also un-imagined complications. In sorting theseout we will start to see technologies’effects on power relations, gender andage differences.

5 References

Bakke, J W. 1996. Competition in mobiletelephony. Telektronikk, 92 (1), 83–88.

Bakke, J W. Forthcoming. Technologiesand interpretations : the case of the tele-phone. Research in science and technol-ogy studies : material culture. S. Goren-stein (ed.). Research in science and tech-nology studies, vol. 10. Greenwich, CN,JAI Press.

Berg, A J. 1996. A mobile phone is amobile phone, is a mobile phone, is amobile phone? Presented at the COST248 meeting 9.7.96, Brussels, Belgium.

Berger, P, Luckmann, T. 1967. Thesocial construction of reality : a treatisein the sociology of knowledge. NewYork, Anchor.

Bogard, W C. 1996. Simmel incyberspace : Distance and strangeness inpostmodern communications. Paperpresented at the Amercian SociologicalAssociation meeting 19 Aug. 1996.

Cahill, S E. 1990. Childhood and publiclife : reaffirming biographical divisions.Social Problems, 37 (3), 390–402.

Chen, V. 1990–91. Mien Tze at the Chi-nese Dinner Table : A study of the inter-actional accomplishment of face. Re-search on Language and Social Inter-action, 24, 109–140.

Clark, J. 1995. Personal discussion.Boston, MA.

Collas, S F. 1995. Some of my bestfriends are ... : constructing borders andboundaries between self and other, usand them. Presented at the 1995 Meetingof the American Sociological Associa-tion.

Duncan, H D. 1970. Communication andthe social order. London, Oxford.

Duncan, S. 1972. Some signals and rulesfor taking speaking turns in conversa-tions. Journal of personality and socialpsychology, 23 (2), 283–292.

Fine, G A. 1995. Wittgenstein’s kitchen :Sharing meaning in restaurant work.Theory and Society, 24, 245–269.

Fischer, C. 1992. America calling : asocial history of the telephone to 1940.Berkeley, CA., University of California.

Garfinkle, H. 1967. Studies in ethno-methodology. Oxford, Polity.

Geertz, C. 1972. Linguistic etiquette. In:Readings in the sociology of language.Fishman, J A (ed.). The Hague, Mouton,282-295.

Geertz, C. 1973. The interpretation ofculture. New York, Basic.

Giddens, A. 1984. The constitution ofsociety : Outline of the theory of struc-turation. Berkeley, Univ. of California.

Giuffre, P A, Williams, C L. 1994.Boundary lines : labeling sexual harass-ment in restaurants. Gender and society,8 (3), 378–401.

Glaser, B, Strauss, A. 1967. The dis-covery of grounded theory. Chicago,Aldine.

Gluckman, M. 1963 Gossip and scandal.Current Anthropology, 4, 307–316.

Goffman, E. 1959. The presentation ofself in everyday life. New York, Double-day.

Goffman, E. 1963. Behavior in publicplaces : note on the social organizationof gatherings. New York, Free Press.

Goffman, E. 1967. Ritual Interaction :Essays on face-to-face behavior. NewYork, Pantheon.

Goffman, E. 1974. Frame analysis. NewYork, Harper.

Gullestad, M. 1992. The art of socialrelations : essays on culture, socialaction and everyday life in modernNorway. Oslo, Universitetetsforlaget.

Haavio-Mannila, E, Snicker, R. 1980.Traditional sex norms and the innovativefunction of afternoon dancing. Researchin the Interweave of Social Roles, 1,221–245.

Haddon, L. 1996a. Mobile telephonyissues : discussion paper for COST 248,Mobile sub-group. Presented at COST248 meeting 11.04.96, University of Sus-sex, Brighton, UK.

Haddon, L. 1996b. The development ofthe UK mobile telephone market. Pre-sented at the COST 248 meeting09.07.96, Brussels, Belgium.


Haugen, I. 1983. Tilgjengelighet ogsosiale selv. Arbeidsnotat, Institutt forsamfunnsforskning, Oslo.

Jackson, T. 1952. Some variables in roleconflict analysis. Social forces, 30,323–327.

Ling, R. 1994. A taxonomy of telephoneconversations based on the relationshipof the callers : preliminary results of aqualitative examination. Kjeller, TelenorR&D. (TF.R 33/94.)

Ling, R. 1996. The technological defini-tion of social boundaries : Video tele-phony and the constitution of groupmembership. Telektronikk, 92, (1),61–73.

Lippman, A. 1967. Chairs as territory.New Society, 20, 564–6.

Lofland, J, Lofland, L H. 1984. Ana-lyzing social settings. Belmont,Wadsworth.

Lohan, E M. 1996a. The domestic phonespace : the telephone space in the homeand the domestic space in telephony : aparticipant observation case study of thetelephone in three Dublin households.COST 248.

Lohan, E M. 1996b. Key points for COST248 project on home and work : theincorporation of new technologies intothe private sphere. Presented at theCOST 248 meeting 09.07.96, Brussels,Belgium.

Mars, G, Nicod, M. 1984. The world ofwaiters. London, George Allen &Unwin.

Martin, M. 1991. Hello, Central? Gen-der, technology and culture in the forma-tion of telephone systems. Montreal,McGill-Queens.

Marx, G. 1994. New telecommunicationstechnologies require new manners.Telecommunications policy, 18 (7),538–551.

Meyrowitz, J. 1985. No sense of place :The impact of electronic media on socialbehavior. Oxford, New York.

Miller, H. The presentation of self inelectronic life : Goffman in the Internet.http://www.ntu.ac.uk/ soc/psych/miller/goffman.htm

Parker, S. 1988. Rituals of gender : astudy of etiquette, public symbols andcognition. American anthropologist, 90(2), 372–384.

Rombauer, I, Becker, M. 1964. The joyof cooking. New York, Signet.

Rutter, D. 1987. Communicating by tele-phone. Oxford, Pergamon.

Saks, H, Schegloff, E A, Jefferson, G.1974. The simplest systematics for theorganization of turn-taking for conversa-tions. Language, 50 (4), 696–735.

Silverstone, R. 1996. Personal conversa-tion. 15.07.1996, Brussels, Belgium.

Silverstone, R, Haddon, L. 1996. Designand domestication of information andcommunication technologies : technicalchange and everyday life. Communica-tion by design : the politics of informa-tion and communication technologies.

Mansell, R, Silverstone, R (eds.). Lon-don, Oxford.

Silverstone, R, Hirsch, E, Morley, D.1992. Information and communicationtechnologies and the moral economy ofthe household. Consuming technologies :media and communication in domesticspaces. Silverstone, R, Hersch, E (eds.).London, Routledge.

Simmel, G. 1971. Georg Simmel : onindividuality and social forms. Levine, DN (ed.). Chicago, Chicago.

Spradley, J P. 1979. The ethnographicinterview. New York, Holt, Rinehart andWinston.

Stone, A. 1994. The Eyes of the Vampire: Desire, Bodies, and Legibility inCyberspace. At:Virtual Reality Oslo1994. Oslo, Norway, 18 – 20 Aug.

Turner, J H. 1986. The structure of socialtheory. Chicago, Dorsey.


Rich Ling is a sociologist working at Telenor R&D,Kjeller. He received his Ph.D. in sociology from theUniversity of Colorado, Boulder. Since coming toNorway he has worked at The Resource StudyGroup founded by Jergan Randers and has been apartner in the consulting firm Ressurskonsult. For thepast five years he has worked with Telenor and hasbeen active in researching issues associated withnew information technology and society.


Status


International Research and Standardization Activities in Telecommunications

Editor: Per Hjalmar Lehne

This issue of the Status section contains four papers, rangingfrom standardisation and research on broadband access net-works, quality of service (QoS) and ETSI’s 10th anniversary,1988 – 1998.

EURESCOM stands for ‘European Institute for Research andStrategic Studies in Telecommunications’. It is organised as aGerman GmbH, and has its headquarter in Heidelberg, Ger-many. Telenor is currently participating in 15 projects which areshown in the table on the right. EURESCOM originally organ-ised only the Public Network Operators (PNOs) in Europe, butin the last year membership has been opened for all fixed andmobile network operators as well as service providers. It isinteresting, though, that no new members have applied so far.

First, Mr. Umberto Ferrero presents project P614 – ‘Implement-ation strategies for advanced access networks’. Mr. Ferrero isproject leader of P614 which started in October 1996 and willend in November 1998. In his paper he explains the goals of theproject and different broadband access technologies and archi-tectures. The project aims at providing guidelines for the PublicNetwork Operators (PNOs) for the implemenation of broadbandaccess technologies.

The second paper addresses P806-GI – ‘A common frameworkfor QoS/Network performance in a multi-provider environ-ment’, and is written by Mr. Ola Espvik, Mr. Terje Jensen andMr. Bjarne Helvik. Providing a defined quality of service (QoS)to the end-user has become more complicated after the liberal-isation of the telecommunications market because several net-work operators and service providers may be involved in pro-viding the final service. The P806-GI project tries to investigatethe challenges and define a common framework for QoS uponwhich the PNOs can agree.

The third paper also adresses QoS, but from a user point ofview. Mr. Magnus Krampell discusses the efforts done in theETSI project TIPHON – ‘Telecommunications and Internet Pro-tocol Harmonization Over Networks’. EP TIPHON has definedso-called Quality Classes, which are used as examples of ageneric quality model. Mr. Krampell’s paper introduces a gen-eral methodology for a user-oriented quality model and pro-poses an extended set of Quality Classes.

The European Telecommunications Standards Institute – ETSI– was founded in 1988 from the reorganisation of CEPT (Con-férence des Administration Européennes de Postes et Télécom-munications). In conjunction with the General Assembly, ETSIarranged a one day seminar to celebrate its 10th anniversaryunder the title ‘The State of the Art in European Telecommuni-cations Standardisation’. The seminar was held in Nice, Franceon 25 March 1998. Telektronikk was present and a report fromthe seminar is the last paper in this issue. A strong Norwegiandelegation, counting 10 people, was present.

79

IntroductionP E R H J A L M A R L E H N E

Telektronikk 2.1998

Project no Duration Name

P 605 Feb 96 – Jupiter – Joint Usability, Performability andFeb 98 Interoperability Tests in Europe

P 608 Aug 96 – TINA concepts for 3rd generation mobileApr 98 systems

P 614 Oct 96 – Implementation strategies for advancedNov 98 access networks

P 616 Nov 96 – Enhanced ATM Implementation IssuesMar 98

P 617 Mar 96 – Evaluation of High PerformanceMay 98 Databases for Interactive Services

P 619 Feb 96 – PNO – Suppliers Technical InterfacesMar 98

P 710 Feb 97 – Security for the TMN X-interfaceDec 98

P 714 Feb 97 – CSCW Tools for TelecommutingApr 98

P 801 Mar 98 – Local Information Infrastructure toJun 99 Support Community Communication

Networks (CCN)

P 802 Jan 98 – Strategic study into sustainability withJun 98 respect to social & economic impacts

of telecom services

P 806 Mar 98 – A common framework for QoS/NetworkSep 99 performance in a multi-provider environment

P 807 Mar 98 – Jupiter II – Usability, performability &Mar 99 interoperability trials in Europe

P 813 Apr 98 – Technical Development and Support forJun 99 European ATM Service Introduction

P 816 Mar 98 – Implementation frameworks forSep 99 integrated wireless-optical networks

Table 1 EURESCOM projects in which Telenor is currently participating (April 1998)

Per Hjalmar Lehne is Research Scientist atTelenor Research & Development, Kjeller. Heis working in the field of personal communica-tions, with a special interest in antennas andradio wave propagation for land mobile com-munications.

e-mail:[email protected]


The access network is the portion of the telecommunicationinfrastructure linking every single customer to the closestlocal exchange: traditionally dimensioned to deliver tele-phony, it now calls for a thorough upgrade to accommodatenew midband and broadband services, exploiting the ex-tensive range of available technologies. EURESCOM P614represents the third generation of access network relatedEURESCOM Projects, after P306 ‘Access network evolu-tion and preparation for implementation’ [1,2] and P413‘Optical networking’ [3].

1 Introduction: what is EURESCOM P614?

The Project started in October 1996 and involves over 50experts from 12 European Operators. P614 addresses the accessnetwork dilemma with a more holistic view, and reflects threemain ideas developed during the previous Projects and lessonslearned. First, the future access network will feature a numberof alternative implementations and several architectures, andwill extend its domain towards longer reach allowing for nodeconsolidation: European Operators need to understand and copewith a number of architectures and systems with a single, com-prehensive overall network perspective to assure the effectiveexploitation of the heavy investment involved. Second, theaccess network evolution is quickly moving towards realdeployment, and co-operative projects have to focus on prac-tical issues, such as specifications, outside plant technologiesand interoperability of new and old technologies. Third, themonitoring and the contribution to standardisation bodies, to-gether with the techno-economic appraisal of technical imple-mentation need to be carried out on an ongoing basis byEURESCOM members, to quickly and effectively react to thechanging access network scenario.

2 Broadband access networksrequirement and specifications

New standardization bodies and groups of interest are sproutingevery so often, and it is increasingly difficult to monitor all ofthem, particularly for small operators. P614 provides memberswith the accurate coverage of the work going on within all therelevant study groups. Moreover, sensitive topics can be jointlyaddressed and supported, elaborating common papers to speedup subsequent processing of established bodies. P614 membersare identifying areas where standards are urgently required, withthe purpose of contributing as appropriate in a co-ordinated way(Deliverable 9).

Significant cost reduction is achieved by the mass market of one(or few) solutions. EURESCOM access network related Projectsidentified ATM-based passive optical networks (ATM-PON)complemented by short-range high-speed copper based trans-mission as the preferred option. This choice, which dates backto mid 1994, facilitated the quick convergence of FSAN (FullServices Access Network) members to this solution, enablingthe successful development of the specification requirementsand its downstreaming to established standardisation bodies.Specific targets are selected as areas requiring further and quickstudy or areas showing inconsistencies. So far, the first area forcommon effort is the Service Node Interface requirements iden-tification, analysing advantages and disadvantages of the differ-

ent architectures incorporating such interfaces (Deliverable 1).Another critical study area is the identification of requirementsfor the broadband home networks (Deliverable 11).

3 Evolution towards broadband: the end of a single architecture

The dramatic technological evolution and the regulatory change-out are leading to the installation of a variety of broadbandaccess systems. On the one hand, Operators look at target solu-tions: a wider time frame scenario enables the investigation offibre rich implementations, taking advantage of potential bene-fits in terms of global network rationalisation and optimisation.The overall access network optimisation potentially enablessignificant cost savings, improved service quality and eventualintegration of services for business customers, provided thatspecifications and standard solutions are being developed.

The target solutions encompass both ATM-PON and ATMpoint-to-point links; one or several evolution paths from a set ofexisting access networks architectures to the identified targetarchitecture is being addressed (Deliverable 2). On the otherhand, the compelling need to enter new markets and providenew services requires the full exploitation of existing infra-structures, re-using twisted pairs, coaxial cables and, undercertain circumstances, even powerlines. The broad diversity oftechnical implementation is able to match the diversity of ser-vice offering and acceptance, existing infrastructures and otherlocal constraints, enabling the investment optimisation (Deliver-able 6). The results offer Operators the possibility to identify therelevant parameters to be considered when planning an upgradeof their own access network, understanding the capability andlimits of the existing initial architectures and giving full knowl-edge of the established target solution (including performanceand reliability figures).

4 Going wireless: broadband radio access

Broadband radio technologies are becoming more and moreclose to real field deployment, and are offering very promisingapplications for both incumbent operators and new competitors.But the big, potential, cost savings stimulate a big hype of inter-est, boosted by an aggressive marketing campaign. P614 re-views the radio access network architectures and technologiesgiving B-ISDN/ATM access to residential and business cus-tomers; a Request For Information (RFI) was issued. P614 isalso reviewing ongoing research and standardisation activitiesand estimate the type of technologies available in the long term.

P614 highlights the truth on time for available systems and tech-nical capabilities. Many systems are in their infancy and the widediversity of technical implementation and regulatory constraintsare making the successful exploitation of such technologies moredifficult. The need for common air interface and radio-to-fibreinterfaces have been identified as a key enabler for a real andextensive use of broadband radio solutions; the visions elaboratedon are now being downstreamed in the appropriate bodies (Deliv-erable 4). A promising development is the possibility to exploitbroadband radio technologies on top of fixed, fibre-rich accessnetworks as an ideal complement to by-pass costly final droptechnologies and home network related problems.

EURESCOM Project P614– Implementation strategies for advanced access networks

U M B E R T O F E R R E R O


5 Basic technologies and practicalimplementation aspects

It is widely known that 70 % of the access network costs is dueto infrastructure – how can this burden be reduced? Re-use ofcopper wires is well known, but other solutions include adopt-ing innovative and relatively little known installation techniquesthat make it possible to re-use ducts, simplify installation, makethe infrastructure more flexible, or simply adopting ultimatesolutions whenever possible.

The deployment of new broadband technologies in the accessnetwork implies a number of practical implementation issuessuch as powering, testing, cabling and outside plant technology.P614 promotes the understanding of these technologies with theorganisation of the Workshop on Access Network Cabling andInstallation Techniques (ANCIT, 30-31 March 1998). P614focuses on feasibility evaluation and cost estimates, supportingOperators choices with the necessary technological understand-ing (Deliverable 5).

6 Optical technologies in the accesstoday and tomorrow

New optical low cost technologies are emerging, such as cheappoint-to-point optical transmission sub-systems, allowing forseveral applications as drop, home cabling or alternative opticaldistribution network. Moreover, the increasing availability ofoptical fibres in the access enables the fruitful exploitation oftechnologies that have already proved their effectiveness inother network segments, such as wavelength multiplexing andsophisticated measurements and maintenance techniques(Deliverable 7).

FTTH has been discussed since the early 1980s. Now, with thereal deployment of hybrid systems and major technologicalbreakthrough, the exploitation of the FTTH concept becomesreally viable, at least for some applications and in some en-vironments. P614 members are willing to stimulate the costreduction of FTTH solutions in such a way that it can be usedas early as possible for avoiding intermediate evolutionary stepsat least in green-field areas (Deliverable 8). The idea is to iden-tify function limitations and Operator requirements that couldmake FTTH configuration cheaper, and to justify from an eco-nomic point of view a service offer corresponding to the FTTH

MDF: main distribution frame OM: optical monitoring

FTTC/Cab

FTTB

ONU

ONU

ONU

ONU

ONU

ONU

O

L

T

OM MDF

L E X

FTTH

FTTH/A

Manhole

Duct Duct/direct buried

Cables: Connectors:

Splices in enclosure: Splitter:

Figure 1 FTTX infrastructures


configuration: a cheap broadband FTTH access might be a suit-able complement to the narrowband mobile service offering(Deliverables 12, 13 and 14).

7 The price of broadband evolution

EURESCOM Projects P306 ‘Access network evolution andpreparation for implementation’ and P413 ‘Optical networking’elaborated on extensive techno-economic evaluations whichproved very useful as commonly agreed guidelines for thedefinition of network evolution strategies. So, one of the P614objectives is to identify economically viable implementationstrategies for an effective broadband access network introduc-tion, maintaining a comprehensive and updated cost evaluationof system and architectural alternatives. P614 periodically up-dates results evaluating new service assumptions and includingnew technological developments (Deliverable 3).

Marketing people tend to propose extremely low cost figureswhen the technology is being developed, and much higherprices when the contract is approaching ... So, from readingreports and magazines the latest technology always seems tobe the cheapest, and drawbacks seem to be very minor. P614reports bring back to earth the (easy) enthusiasm, recommend-ing under which circumstances solutions can really be cost-effective. This activity represents an ongoing concern forEURESCOM, supporting Shareholders with independent andcritically assessed estimations of the economic implications ofthe broadband access network deployment. Studies are made insuch a way as to make the application of the results by the

Figure 2 Line cost: 2 Mbit/s asymmetric connection [5]

2.000

1.800

1.600

1.400

1.200

1.000

800

600

400

200

00 20 40 60 80 100

Broadband connection demand (%)

Line

cos

t (E

CU

) (%

)

BPON-8

BPON-32

BPON-64

BPON-128

HFC, Cable modem

ADSL

HFC, ASB

ADSL, feeder inc.

HFC, no ret. path

100% d.a0% d.a

100%d.a

0%d.a 100%

d.a0%

d.a

100%d.a

0%d.a

Operators easy, taking into account country and area specificparameters, such as social profile, service offering, civil workcosts, duct availability, labour costs etc.

8 Relationship outside the EURESCOMcommunity

P614 is becoming increasingly popular as a reference source ofindependent broadband access network recommendations, aswitnessed by the number of hits collected by the web-site andthe request for reprints and documents from operators and man-ufacturers all over the world: access network experts from US,Brazil, Korea, Taiwan, Indonesia and New Zealand, as well assome European suppliers and consultants, have alreadyrequested P614 reports. P614 has established links with severalaccess network related Projects. A bilateral agreement is estab-lished with ACTS OPTIMUM (Optimized Network Architec-tures for Multimedia Services) Project to use the OPTIMUMtool and exchange cost estimates.

9 Dissemination of results

EURESCOM Project P614 is committed to open disseminationof results, reflected in the choice to make all the deliverablespublic (i.e. accessible by people outside EURESCOM Share-holders). The effort to raise consensus on the Project results isconfirmed by the plan to favour co-operation with other bodies(international standardisation organisations and manufacturers)and with the preparation and running of periodical seminars andworkshops. P614 presents its results at several internationalconferences and maintains a public web-site where a full rangeof information on the Project is made available; papers and De-liverables can be easily downloaded and P614 members can becontacted for further information.

10 Conclusions and future work

The Operators tend to swing between the idea of doing nothingor everything as far as broadband access network is concerned.P614 results recommend what to do (something), when to do it(with specific phasing, following the maturity of different tech-nologies), and where (in areas of well defined technical andservice characteristics).

P614 tries to spot, and tackle, some areas that have been out offocus so far, increasing the awareness of their potential and giv-ing answers to ‘hot’, much discussed questions, or to counteractsome ‘myths’ on technical capabilities and trends.

P614 is becoming a recognised independent and reliable sourceof information on broadband access network evolution. Thenext challenge can only be the actual field deployment of someof the available systems.

References

1 Warzanskyj, W, Ferrero, U. Access network evolution inEurope : a view from EURESCOM. In: Proc. ECOC 1994,Florence, Italy, 25–29 September 1994.

2 Ims, L Aa et al. Multiservice access network upgrading inEurope: a techno-economic analysis. IEEE CommunicationsMagazine, December 1996.

3 Ferrero, U. Broadband optical access network : cooperativework among European PNOs. In: Proc. of ECOC 1996,Oslo, Norway, 15–19 September 1996.

4 Ims, L Aa et al. Key factors influencing investment strate-gies of broadband access network upgrades. In: Proc. ISSLS98, Venezia, Italy, 22–27 March 1998

All the papers published by access network relatedEURESCOM Projects, together with all the P614 Deliverables(as listed in Table 1) and information on the workshops, areavailable from the Internet at:

http://www.cselt.it/Cselt/euresc/P614/welcome.html

Acknowledgement

This document is based on results achieved in a EURESCOMproject; this does not imply that it reflects the common technicalposition of all the EURESCOM Shareholders/Parties. Theauthor gratefully acknowledges the support of EURESCOMfor carrying out this work. The author wishes to express aspecial thanks to all P614 participants, since this paper reflectsthe results of their work.

EURESCOM P614 members are: BT, CNET-France Telecom,CSELT, DTAG, Hungarian Telecom Company, OTE, PortugalTelecom, Swisscom, Telecom Ireland, Sonera, Telefónica I+D,Telenor.


Title Planned issue date

D1 An analysis of the relative benefits of proposed SNI standards April 1997

D2 Target B-ISDN access network architectures December 1997

D3 Techno-economic analysis of major factors of B-ISDN/ATM upgrades September 1997

D4 Opportunities for broadband radio technologies in the access network January 1998

D5 Optical technologies for advanced access networks: early results December 1997

D6 Evolution paths towards target B-ISDN access networks July 1998

D7 Optical technologies for advanced access networks: final results September 1998

D8 Elaboration of common FTTH guidelines July 1998

D9 Contributing to standardisation bodies and other fora April 1998

D10 Techno-economic evaluation of B-ISDN access networks September 1998architectures, scenarios and business cases

D11 Evaluation of Broadband Home Networks for residential and small September 1998business

D12 FTTH: Definition of the suitable powering architectures * February 1998

D13 FTTH: Definition of access network quality and cost * February 1998

D14 FTTH: Definition of related service offer strategies * February 1998

Table 1 EURESCOM P614 Deliverables

* Deliverables to be produced following approval by the Board of Governors.

Umberto Ferrero graduated in Telecommu-nication Engineering from the Politecnico diTorino in February 1992. He joined CSELTin 1991, where he now works as Senior En-gineer. He is EURESCOM P614 ProjectLeader and he is involved in access networkarchitecture design and economic evaluation.Ha has published more than 40 papers onaccess networks related topics.



The multitude of operators/providers together with the widerange of services emerging in telecommunication networks,request for general descriptions dealing with quality of ser-vice issues. Such descriptions should be established betweenthe end-user and the provider as well as between the pro-viders. The latter is more pronounced as a number of pro-viders could co-operate in order to fulfil a service request.

1 Introduction

It is essential to observe that the quality of a service is basicallydifferent from the attributes of the service. The quality (definedby the value of a set of quality parameters) of a service tells thecustomer whether an ordered service is delivered within specifi-cations of an agreement. The point of having parameters de-scribing Quality of Service (QoS) is that they – when givennumerical values – enable the customer to have a well definedopinion of what he is paying for, and allow him to make com-parisons in the market. This requires a common understandingof QoS and how it is measured.

Typically, for an Internet type of service, the customer (and ser-vice provider) has to rely on services from different networkoperators, to fulfil his mission. For the years to come it is fore-seen that more customers will require more explicit guaranteesof QoS for the Internet services. QoS is therefore becomingincreasingly more important to competing public network oper-ators (PNOs) / service providers and customers.

A majority of PNOs apply the ITU QoS framework (ITU-TRec. E-800) [1] in their quality work. However, there are vari-ous other international organisations like ISO, ETSI etc thataddress the issue of QoS, each having to some extent their owncollection/system of parameters. Having many collections ofQoS parameters in use – some of them even being a mixture ofthe others together with some ‘droplets’ of attributes and ‘per-sonal inventions’ – makes life difficult both for the customer/user and the service provider – as well as creating inefficiencyin research programs addressing aspects of QoS.

Imperative to the future is to avoid solutions to frameworks andparameters dependent upon the ever increasing number of ser-vices and actors in the telecommunication market. TheEURESCOM project P806-GI: A Common Framework forQoS/Network Performance in a multi-Provider Environment isgoing to investigate the challenges of QoS in a multi-providerservice provision with the aim of reaching an agreement amongthe EURESCOM shareholders.

The challenges are further elaborated in Section 2. Section 3then deals with existing frameworks and approaches, prior tooutlining the steps undertaken in Section 4.

2 The multi-provider service agreementinterface challenges

Operating telecommunication networks have long traditions forinterworking. That is, it should be possible to transfer informa-tion between users in networks managed by different organisa-tions. Interconnect agreements must be described correspond-ingly. Traditionally, this is done among a few actors of the samekind (the national PNOs) and where ‘the QoS impairmentbudget’ is subdivided between the actors according to ITU stan-dardised reference connections.

Considering the increasing number of actors as well as the dif-ferent types of actors, (eg. platform operators, access providers,transport provider, content hosts and users/customers) and thecorresponding identification of network domains, multiple com-binations of competition and co-operation are expected. Furtheremphasis on regulations concerning interconnect, in particularlooking at dominating operators, is also anticipated.

It is foreseen that QoS/performance agreements should be estab-lished related to most of the interfaces between actors. In partic-ular, any requirements for unbundling the network elements andfunctional groups could lead to a considerable number of non-traditional interfaces.

Ensuring availability and successful calls/information transfersoriginated in one network domain and destined for otherdomains is also an issue. In particular, the users may requireexplanations and possible compensations by the operator/pro-vider dealing with the ‘originating’ side (of which he is a cus-tomer). Therefore, interconnect agreements have to incorporatemeans that the relevant operator/provider can apply in order toavoid deteriorating situations. That is, on the transfer/call levelboth outgoing and incoming situations must be considered andthe views of both parties of an interface must be taken on. Thisimplies that the contents of such agreements should adequatelycover issues arising for the purpose of protection and guaranteebetween actors, see Figure 1. Guarantee issues include state-ments from relevant actors promising a sufficient high level ofQoS such that the end-users are believed to be satisfied. Protec-tion issues include precautions against receiving traffic flowsseverely reducing the QoS within the domain considered. Theprotection aspects may, again, be seen from more parties; eitherthese could be stated in order to limit the load from neighbour-ing domains (as seen from the left domain in Figure 1 b)) orthese could be stated in order to avoid degradation of QoS dueto load generated from connected subscribers (as seen from theright domain in Figure 1 c)).

EURESCOM project P806-GI:– A Common Framework for QoS / Network Performancein a multi-Provider Environment

O L A E S P V I K , T E R J E J E N S E N A N D B J A R N E H E L V I K

a)

b)

c)

Figure 1 Interconnect agreements must consider aspects of: a) guarantee for originating part; b) protection for destinating party;

c) protection for originating party


An agreement should include aspects likethe conditions for operation, how to assessthese conditions and actions to be under-taken in case agreed levels on any of theconditions are exceeded. Naturally, theparticularities on interfaces have to be con-sidered when describing these aspects. For instance, delays maybe more essential on packet switched relationships while block-ing probabilities are used on circuit switched connections. In thegeneral case, variables describing delays as well as variablesdescribing blocking are relevant.

The approach to be followed may begin by describing thepotential roles present in the provisioning of services. Thismay also look at the different phases of the service life cycle.Contributions from regulatory bodies should also be taken intoaccount as a motivation for identifying interfaces betweenactors.

Such interfaces may be located between horizontally separateddomains (domains with similar functionality) and between verti-cally separated strata (strata where the lower one provides cer-tain services for the one above). That is, horizontal means inter-connecting network domains at the same functional level, eg.,between two exchanges. Vertical means connecting differentlevels in a functional view or network architectural view, theinterface between a content host and a platform operator, orbetween a platform operator and a transport provider. An ex-ample of the latter, related to Intelligent Networks, is connec-tions between Service Control Points and Service Data Points.Functional relationships would exist between the different net-work elements, like when service logic/data are executed on aplatform provided by another actor. That is, an interface maybe more diffuse than a physical link between network elements.Naturally, when interfaces are incorporated within an element,assessing the conditions may become more involved unlessexternal equipment can be used.

In addition, an actor would also like to know the characteristicsof the loads at the ingress points. For instance, load with highervariability or having severe correlation may influence the net-work such that the performance is significantly degraded com-pared to situations where such effects are not present. There-fore, descriptions of traffic characteristics should also be given.Alternatively, counteracting measures could be taken if theoffered traffic load deviates from the characteristics. It mightbe tempting to either avoid such situations or make the relevanttraffic flow confirm with the better behaving characteristics(eg., by applying shaping). Activating management mecha-nisms, like call gapping, in order to smoothen the impact on thenetwork conditions would lower the QoS for the service con-cerned. Therefore, the way management procedures affect QoSmust be taken into account. Naturally, if the situation is suchthat no significant reduction of performance is foreseen, thetraffic flow could be treated as it is. This may, in one way, seemsimilar to a ‘best effort’ manner for treating the traffic load andcould be stated as such in a contract. In that way the thresholdsin the agreement may be regarded as minimum values which arecommonly met and where better values can be found when thenetwork state allows for it. However, multiple thresholds couldbe given where different reaction patterns are described for eachinterval.

Based on the description of roles and interfaces, a frameworkfor interconnect agreements related to QoS is defined. Otheraspects, like legal, economical, technical protocol specific etc,should also be considered; however, these are not treated inP806-GI. In principle, the contents of an agreement looked atmay include, ref. Figure 2:

• Interface description

• Traffic patterns expected (with relevant parameters in order tocharacterise the traffic flows)

• QoS parameters and thresholds

• Evaluation schemes including procedures for carrying outmeasurements

• Reaction patterns.

Varying aspects of these topics may be pronounced for thedifferent interfaces. For this project we point out the need of ageneral framework to be applied. As part of the evaluation pro-cedures, exchange of data between interconnected actors willbe referenced.

Naturally, the network operator will usually be the only instancehaving a complete view of the network state. In order to limitany disturbances following a network condition, it is a soundprinciple to choke the relevant traffic flows at the edges of thenetwork. Considering interconnect, this means that suitablemechanisms should be present in the network elements asso-ciated with the ingress points.

Often, having described the conditions to appear on the inter-face, belonging measurement schemes are identified. Values forvariables of the arrival processes, service mixture and the result-ing performance are to be captured. As for every measurement,decisions of when, where and what to measure must be made.That is, topics like time and duration, interface/location andevents have to be specified. These are measurements that maybe carried out by both parties of an interface. It must also bedecided whether or not continuous measurements are to be per-formed. As measuring could be regarded as sampling, it is alsoto be agreed upon if terms in a contract can be questioned basedon a single measurement period or if a number of measurementperiods have to be done of which several indicate that the termscan be questioned before the contact is renegotiated or othermeans are applied. This is also seen as a trade-off between thetime for reactions (responsiveness of a scheme) and the effortneeded for preparing and carrying out the reactions.

In addition, measures treating sudden changes in the arrivalprocesses must be present. Load control is an example of suchquick response measures. Which measures to apply for an inter-face should also be stated in the interconnect contract. A num-ber of measures, operating on a range of time scales, may bethought of.

Load control implying rejecting or delaying traffic is consideredas a feature utilised during operation. One of the purposes may

Interfacedescription

Performancevariables

Trafficpatterns

Performancethresholds

Measurementschemes

Reactionpatterns

Figure 2 Possible structure of an interconnect agreement


be to avoid that a single group of services/call types seizes toolarge a fraction of the capacity leading to degradation of thequality for other services/call types. This may be particularlyrelevant when mass calling services are introduced.

Another example where load control between operators may berequested is when one operator chooses to reroute trafficthrough network domains1 that are not prepared for this. Thismay happen when the more direct connection is not available,eg. due to link failure. Although accounting rules may treat thissituation by introducing financial compensations, using mea-sures for not degrading the QoS for the remaining servicescould be more fruitful in order to keep one’s reputation.

3 Existing QoS frameworks andapproaches

Various international bodies have made efforts towards makinggeneral QoS frameworks, eg. see [4]. Some of these are:

• ITU-T E-800 where part of this standard is adopted by IEC asterminology standard IEV 191

• ETSI framework, [2]. This framework is based on the work ofthe FITCE Study Commission [3]

• The service failure concept of EURESCOM P307. Theattributes of a service failure is regarded independent of theservice it affects and its cause in the network [7]

• QoS in layered and distributed architectures

- The ISO/OSI QoS framework [6]

- The Telecommunication Information Networking Architec-ture Consortium (TINA-C) QoS framework [14]

• ETNO Working Group 07/95 on QoS. This group is workingtoward a consistently defined set of common European QoSparameters (QoS indicators). The aim is harmonised Euro-pean QoS definitions and possibly performance targets forpan-European services, in order to facilitate comparison ofthe results of the measurements. The work is based on theapproach of the FITCE Study Commission and ETSI. Up tonow, the work has concentrated on voice telephony.

In addition, EURESCOM P603 [8] has studied the bearer ser-vice part of some actual QoS frameworks. It was concluded, [4],that the frameworks were partially overlapping but essentiallysupplementary, since they focus on different aspects. Although

concepts may be different, and there are considerable differ-ences in terminology, a unification seems feasible. None ofthese are primarily intended as a basis for measurements, norare they intended for a quantitative comparison of QoS aspectsacross different types of bearer networks.

Aagesen [5][12][13] made a thorough discussion on QoS in opendistributed processing (ODP) systems. One conclusion is thatthere are no fundamental contradictions between the OSI, ITU-ISDN, ODP, and TINA QoS frameworks. Common features are:i) objects offering services and performing QoS handling func-tions, ii) QoS parameters/attributes, and iii) contracts betweenobjects. However, the focus and the use of concepts differ.

A brief description of a collection of EURESCOM projects hav-ing addressed or QoS/NP aspects is given in [11].

4 How to proceed

From the studies so far is seems feasible to reach a commonunderstanding as regards a first common framework applicablein a multi-provider environment, where the providers may oper-ate in different

• domains and/or• strata.

The most important issue is to reach a core structure and mini-mum set of parameters that the various actors can agree upon.Based upon the experiences gained in P307 [15] and P619 [9]together with the recommendations of P611 [10], see Figure 3,EURESCOM – in its project P806-GI during 1998–1999 –addresses the following topics:

• A preliminary common end-user – provider QoS interface isto be established. This includes a well defined mapping of theend-user’s terminology/parameters into ‘top level’ parametersof a common harmonised QoS/network performance frame-work to be chosen/constructed. A QoS agreement betweenend-user and provider is to be structured and defined. Theinterface, framework and agreement should be independentof specific services and network/provision technologies. Theusefulness of the framework in connection with servicedegradation analysis is also observed.

• After having investigated the above mentioned scheme in anIN and Internet scenario, a generic service structure, customer– provider QoS interface, QoS parameter structure, and struc-ture for service level agreement applicable at the interfacebetween provider(s) and end-user will be proposed. A Memo-randum of Understanding (MoU) related to the structure anddefinitions of the previous objective and items is also to beprepared.

• Including the experience gathered in EURESCOM projectsP307 and P619 as regards data dissemination and commontargets, adequate procedures for collecting and exchangingQoS data in a multi-provider environment will be investigatedand structured.

P806-GI is organised as three main tasks. In the first of thesethe common framework is to be elaborated. As previously de-scribed, this is done in two phases. After the first phase, in the

1 Under the control of one or several other operators.

TINA-CETNO

TINA-C

ETSI EURESCOM

P611 P307

P603

P806-GI

Frameworks

Projects

ITU

ISO

Figure 3 Input and basis for elaborating generic framework for multi-providerenvironments


second task, the framework is to be applied for IN and Internetconfigurations. For these applications the structure of intercon-nect agreements as depicted in Figure 2 is to be examined to-gether with the framework. A third task deals with managementof measurement and data handling. During the second phase ofthe first task a (possible) refined framework is described to-gether with a draft MoU.

The project started in March 1998 and is scheduled to end inSeptember 1999. Currently, there are nine PNOs involved,sharing the 100 man-months allocated. Two workshops/semi-nars are planned, the first one around November 1998 and thesecond one around October 1999. Three deliverables are to beissued, edited by the three main tasks.

5 Conclusion

The public network operators have for a number of years usedITU-T E.800 as their conceptual framework with respect to QoSand network performance. In addition, the ETSI and ETNOframeworks are in use. Although important aspects of existingframeworks are still valid, the challenge is to agree upon a com-mon harmonised QoS framework sufficient in an ODP/multi-provision handling of services. The EURESCOM project P806-GI suggests to take an initiative towards a more widely appli-cable QoS framework aiming at a consensus among all majororganisations and bodies dealing with QoS issues. In the workoutlined above, the project P806-GI team will also communi-cate with a wide range of actors, both from the provider sideand from the customer side.

6 References

1 ITU-T. Terms and definitions related to Quality of Serviceand Network Performance including Dependability. ITU-TRecommendation E.800, August 1994.

2 ETSI TC-NA. Network Aspects : general aspects of Qualityof Service and Network Performance. ETSI TechnicalReport ETR 003 (ref. RTR/NA-042102), October 1994.

3 FITCE Study Commission. The study of Network Perfor-mance considering Customer Requirements, Final Report,1993.

4 Helvik, B. Elements of a Measurement Framework for aBearer Services. Telektronikk, 93, (1), 7–25, 1997.

5 Aagesen, F A. QoS frameworks for open distributed pro-cessing systems. Telektronikk, 93, (1), 26–41, 1997.

6 ISO/IEC JTC1-SC21. Quality of Service Framework. Work-ing Draft No. 2, SC21-WG1 meeting, Yokohama, 1993.

7 EURESCOM. Project P307 Reliability Engineering; De-liverable D3. An approach to dependability modelling oftelecommunication networks and services, November 1995.

8 EURESCOM. Project P603 Quality of Service : Measure-ment Method Selection; Deliverable D1. QoS overview,services and measurement methods-selected services andmethods, October 1997.

9 EURESCOM. Project P619 PNO-Supplier Technical Inter-faces; Deliverable D6. Recommendations to Improve theTechnical Interfaces between PNO and Suppliers, 1998.

10 EURESCOM. Project P611 Overall Strategic Studies; Task3.11 Report. Quality of Service and Network Performance,1997.

11 Groen, H et al. EURESCOM and QoS/NP related Projects.Telektronikk, 93, (1), 184–195, 1997.

12 Aagesen, F A. Quality of Service in layered architecture.Trondheim, SINTEF, 1994. (SINTEF report no. STF40F94088.)

13 Aagesen, F A. Architectures for the modelling of QoS func-tionality, Telektronikk, 91, (2/3), 56–68, 1995.

14 TINA-C (Telecommunication Information NetworkingArchitecture Consortium). Quality of Service framework.Doc. no. TR_MRK.001_1.0_94, TINA consortium, 1994.

15 EURESCOM. Project P307 Reliability Engineering; De-liverable D5. European Quality Measurements, August1995.

Ola Espvik is Senior Research Scientist andeditor of Telektronikk. He has been withTelenor R&D since 1970 doing research in traf-fic engineering, simulation, reliability and mea-surements. His present research focus ison Quality of Service and Quality Assurancemainly related to various EURESOMC pro-jects.


Terje Jensen is Senior Research Scientist atTelenor R&D, Kjeller. He is mainly engaged inactivities related to performance analysis,dimensioning and service quality for a numberof networks and systems. He is Task Leaderof P806-GI.


Bjarne Helvik, Dr. Techn., is Professor at theNorwegian University of Science and Tech-nology (NTNU), Department of Telematics.His fields of interest are QoS, Dependability(modelling, analysis, fault-tolerance, fault-management), Teletraffic (modelling, analysis,simulation, measurements) and Telematicssystem architecture.



Quality of Service (QoS) is an ambiguous term. Often it isregarded as equivalent to reserved/guaranteed bandwidth.This reflects a technology oriented view of telecommunica-tions and data networks. The end users often knows nothingabout bandwidth, jitter, packet loss and other terms whichthey are confronted with.

What the average user would like to discuss is Quality, usingterms which they are familiar with. A generic Quality Modelis thus very much needed. Especially when considering afuture scenario when users daily need to assess and controlthe Quality of the abundance of services available.

The Quality Classes defined in the ETSI project TIPHONare used as a good example. This paper introduces a genera-lised methodology to define a user-oriented Quality Modeland also proposes an extended set of Quality Classes.

1 Introduction

1.1 Definitions of QoS and Service Quality

QoS is a term that is often misused. The ITU-T Rec. E.800 [1]defines QoS as: “The collective effect of service performancewhich determines the degree of satisfaction of a user of the ser-vice”.

This definition can be understood as covering all aspects ofservice provisioning that can affect the users’ perception ofQuality, eg. provision, repair, billing, help desk, etc.

However, the term QoS is sometimes being used for describingthe ATM transfer capabilities (eg. CBR, VBR, UBR, ABR) [2].QoS is also sometimes used for describing the Resource Reser-vation Protocol for IP, RSVP [3].

Apparently, there is a need to structure the discussion on QoSand make a separation between what the user perceives as Qual-ity, what the user’s application implements (and the require-ments it puts to the network) and what the network implements.(The latter is often referred to as Network Performance.)

The ETSI document ETR 003, ‘General aspects of QoS andNetwork performance’ [4] gives generic and particular defini-tions of QoS, Network Performance and their interrelationships.The document describes a method for the management of net-work performance related to QoS and it describes a genericframework for capturing the user’s QoS requirements, mappingthem to QoS service performance parameters, and consequentlyto Network Performance parameters. The process of managingthe feed-back which results in the QoS perceived is also de-scribed.

This paper discusses a simpler methodology with a similarobjective but focusing on the Information transfer phase (asdefined in [10]) of a service only. The method can also beused to define Quality Classes for other phases.

1.1.1 Components of QoS

QoS depends on:

• the Support, ie. the secondary services such as information,provisioning, billing, account management, etc.

• the Security, ie. the security aspects related to bothnetwork/services and customers’ expectations

• the Operability, ie. the ease of use, etc.

• the Serviceability, ie. the access and retention of the serviceand transmission aspects, etc.

The user of a service generally identifies two ‘bodies’ whenusing a service:

• the Administration, ie. the ‘human related aspects’

• the Network, ie. the ‘technical related aspects’.

The ‘human related’ aspects are concerned with ‘Support’ and‘Security’, while ‘technical related’ aspects are mostly con-cerned with ‘Operability’ and ‘Serviceability’ (includingAccessibility, Retainability and Integrity (transmission))(from [1] and [5]). We will in this paper concentrate onOperability and Serviceability.

1.1.2 User-oriented concepts of Service Quality

The concepts presently provided by the standards bodies are notsuitable for end-users. For example, the ETSI ETR 138 [6] de-fines 18 parameters (for PSTN and ISDN only!). The numberof possible combinations of these is far beyond what a user canhandle (even if the individual parameters themselves wereunderstandable to the average user!).

When users use (or are about to use) a service, it would be bene-ficial if the ‘Quality’ (user perceived Quality related to networkperformance) were selected from a limited set of QualityClasses, all well known to the user (this implies the need fora generic Quality Model where the same Quality Classes applyto all services). What is needed is:

• A small set of Quality Classes, (7–9 is seen as a target num-ber. This would allow users to keep a good understanding ofall Classes).

• Quality Classes that relate to what the users perceive whenthey use a service.

• Quality Classes defined in terms of parameters that have avalue domain understandable by the user.

The parameters should relate to human ‘limitations’ (whichchange slowly, if at all ...) rather than to technological develop-ment (which changes rapidly). Some examples are Delay (ETSIETR 250 [9] contains a division of delays into sub-domainsbased on what humans can distinguish), Lip-synch (how wellvideo and sound must be synchronised), video resolution, andvideo frame rate.

Of course, there may also be other parameters (eg. bit-errorrate) to cover other aspects (eg. correctness).

1.1.3 Provider-oriented concepts of Service Quality

The provider is in a different situation. For the provider it isvital that the Quality provided can be measured, so that it can beproven that the Quality agreed with the customer was provided.Parameters defined by standard bodies do fit the providers well.

The ETSI project TIPHON as a way towards a Harmonised User-Oriented Quality Model?M A G N U S K R A M P E L L


What is needed is thus, apart from the user-oriented QualityClasses, a mapping to provider (or network) oriented para-meters.

1.2 Problems of having many models

A likely future scenario is that users will have a multitude ofservices and would prefer one Quality Model that applies to allservices rather than being faced with a different model for eachservice (or provider).

Different models for different services (and providers) wouldmean that users would have to remember the context in whicha certain Quality Class is used. In the worst case, one QualityClass may mean different things, depending on where it is used!

Obviously, users would be the losing party in such a situation.

1.3 Why is a harmonised user-oriented modelneeded?

In PSTN the goal of the network planning used to be to providethe highest possible Quality within the cost limits set up. Thiswas helped by the fact that most Public Network Operatorsenjoyed a monopoly situation.

The present situation with the opening up of the market fortelecommunications warrants a different strategy. Customerswill be provided with what they pay for, no more and no less.This will shift the responsibility to the customer – what are theyreally paying for?

Advanced users (mostly big companies) will always be able todefine their needs and negotiate the relevant Quality required.In fact, most operators already have SLAs (Service LevelAgreements) with their major customers.

However, the average residential user and SME (Small andMedium Enterprise) users cannot afford to have the competenceand equipment needed to define and negotiate their needs.

If a commonly known Quality Model existed, the users couldselect a Quality Class, confident that they know what to expect.The provider would also know what is required, since measure-ment methods and best practices would soon be created for the(relatively few) Quality Classes concerned. For each new ser-vice that is introduced, the provider would have to define howthe different Quality Classes would be supported. Compare thisto the ad-hoc situation (largely existing today) where eachprovider defines the Quality independently for each service!

Figure 1 Would it not be convenient if this question was easily answered?

4 (Best) 3 (High) 2 (Medium) 1 (Best Effort)

Speech Quality Equivalent or Equivalent or Equivalent or(one way, better than better than better thannon-interactive G.711 for all G.726 for all GSM-FR for allmeasurement) types of signals types of signals types of signals

End-to-End Delay < 150 ms < 250 ms < 450 ms(mouth to ear)

Call Set- Direct IP < 1.5 s < 4 s < 7 sup time

E.164 Number < 2 s < 5 s < 10 s

E.164 Number < 3 s < 8 s < 15 svia clearing houseor roaming

Email alias lookup < 4 s < 13 s < 25 s

Table 1 Proposed model from the ETSI TIPHON Project(NB. All delay parameters represent an upper bound for 90 % of the connections over the TIPHON System)The exact look of the table is still under discussion as this is written

76

543

21 8

CHOOSE QUALITY


So, the successful introduction of a Quality Model with alimited number of Quality Classes would lead to:

• Increased customer satisfaction (since users would knowwhat they agree to and may have a way to verify that theyget what was promised)

• Saved cost for the providers (since a Quality assurance modelwith measurements etc. that is reusable can be developed.Also, there would be thousands of customers per QualityClass rather than individual settings – this would makeplanning easier).

2 Existing user-oriented Quality Models

Some attempts at creating user-oriented Quality Models exist.Two such models are presented here.

2.1 ETSI Project TIPHON as an example

The ETSI project TIPHON [7] has proposed four (4) QualityClasses for voice services (over IP), see Table 1.

It can be noted from the table that TIPHON has chosen only afew ‘technical’ parameters as affecting the Quality perceived bythe user. Other parameters may be important, but do not affectQuality directly.

2.2 Other examples

The Multimedia Communications Forum (MMCF) has pro-posed three (3) Quality Classes for multimedia applications [8],see Table 2.

The MMCF has chosen to define their Quality Classes usingmore parameters. The disadvantage of this approach is that theQuality Classes are less ‘service independent’.

3 How can a harmonised user-orientedQuality Model be reached?

There are similarities between the models described above.However, few Generic Quality Models exist that can be appliedto any service. [11] contains one such model; it does not, how-ever, contain any concrete Quality Classes.

3.1 A methodology

This paper proposes a methodology for developing a user-oriented Quality Model that can be applied to any service.The methodology presented has been applied (partly) in theETSI TIPHON project [7] presented above.

The methodology is based on the following steps, that areiterated until a reasonably stable set of Quality Classes hasbeen reached (see also Figure 2):

1 Create a table.

2 Find relevant Network Parameters. Make these rows in thetable.

3 Find the value domain of each Network Parameter.

4 Divide each Network Parameter value domain into a numberof discrete sub-domains. The division into sub-domainsshould be based on human aspects, eg. reaction times.

QoS Parameter QoS Class 3 QoS Class 2 QoS Class 1

Audio transfer delay with echo control < 150 ms < 400 ms < 400 ms

Audio frequency range > 0.05 – 6.8 kHz > 0.3 – 3.4 kHz > 0.3 – 3.4 kHz

Audio level (typical) – 20 dBm0 – 20 dBm0 – 20 dBm0

Audio error free interval > 30 min. > 15 min. > 5 min.

Video transfer delay < 250 ms < 600 ms < 10 s (still image only)

Video/audio differential delay > 150 and < 100 ms > 400 and < 200 ms N/A

Video frame rate > 25 frames/s > 5 frames/s N/A

Video resolution > 352 x 288 > 176 x 144 N/A

Video error free interval > 30 min. > 15 min. N/A

Audio differential delay < 100 ms < 200 ms < 1s

Error free interval > 30 min. > 15 min. > 5 min.

Data rate > 500 kb/s > 50 kb/s > 5 kb/s

Table 2 The model proposed by MMCF


Best

High

Medium

Best Effort

Best

High

Medium

Best Effort

High Urgency

High Urgency

High Urgency

High Urgency

Low Urgency

Low Urgency

Low Urgency

Low Urgency

5 Define 7–9 reasonably disjoint user-oriented Quality Classes.Make these columns in the table. These are really ‘profiles’,and mnemonic names are recommended to make them moretangible.

6 Define a mapping between each Quality Class and the differentNetwork Parameter value sub-domains by grouping value setsfrom different Network Parameters together. The mapping ofeach Quality Class to Network Parameters is seen when look-ing down the corresponding column. (Several sub-domainsmay be placed in one Quality Class if needed, see Figure 2.)

3.2 Applying the methodology

In order for the results to be generic, the methodology should beapplied in a service-independent way. The assumption here isthat all application specific data is at some point transported asbits. The differentiating factors can thus be expressed in termsof bandwidth, bit error rate, delay, jitter, etc. This is the kind ofNetwork Parameters foreseen.

The value domains of such Network Parameters are often welldefined. Concerning the division of the Network Parametervalue domains into sub-domains there is much material in thePsychology and Sociology disciplines that should be helpful.A reasonable set of Quality Classes should not be very difficultto define and agree on. (This step should, however, include dis-cussions with users and user groups.) The difficult part is decid-ing which Network Parameter sub-domains to group together inthe different Quality Classes. Several iterations and trial imple-mentations may be needed.

4 Possible extensions to the TIPHONmodel

The four Quality Classes proposed byTIPHON (Best, High, Medium and BestEffort) may be generic enough to beapplicable to many services. However,sometimes you need to indicate howQuality should be changed. Table 3shows a possible way to segment therequirements into Importance andUrgency.

For some applications the choice of seg-ment could be automatic (eg. Voice OverIP favours Urgency before Importance;File transfer using FTP would favourImportance before Urgency). In othercases the user will have to give someindication on how Important/Urgent thedata being transferred is (eg. whenbrowsing the web, it is sometimes de-sirable to get the web page quickly justto see if it contains anything important;in other cases the web page contains criti-cal information that must not be lost).

The application of the four segments tothe four Quality classes (from TIPHON)would lead to 16 Classes, see Table 4.

Low Importance High Importance

High Urgency Quick but not Quick and Safenecessarily Safe

Low Urgency Best Effort type Safe but notnecessarily Quick

Table 3 Segmentation of Quality classes

Table 4 A combined table with 7 classes

Best

High

Medium

Best Effort

Best

High

Medium

Best Effort

3 7

6

5

4

2

1

Low Importance High Importance

QualityClass 1

P1 Valuesubdomain 1

P2 Valuesubdomain 1

Pn-1 Valuesubdomain 1


Param 1

Param 2

Param n-1

Param n

•••

QualityClass 2

P1 Valuesubdomain 2

P2 Valuesubdomain 2

QualityClass 7

P1 Valuesubdomain 3

P2 Valuesubdomain 3

P2 Valuesubdomain 4


QualityClass 8

P1 Valuesubdomain 4


Pn Valuesubdomain 1

•••

Figure 2 The principles of the methodology presented


If some of the resulting 16 classes are combined, we could endup with 7 Classes (see also Table 4):

1 Best-Effort ‘Best Effort’

2 Quick-but-not-so-Safe ‘xxx’

3 Quickest-but-not-so-Safe ‘yyy’

4 Safe-but-not-so-Quick ‘zzz’

5 Safest-but-not-so-Quick ‘mmm’

6 Quick-and-Safe ‘nnn’

7 Quickest-and-Safest ‘Best’

It remains to put better mnemonics to the Classes (the xxx, yyy,etc.) in order to make them easily understandable and easy toremember. It also remains to be tested whether the serialisationused here is suitable (another possible sequence would be 1-4-5-2-3-6-7). Finally, it remains to try the Quality Classes out on aset of services and see whether relevant mappings to networkparameters can be made.

5 Conclusions

This paper points out the irregular use of the term Quality ofService (QoS) and identifies the need for a harmonised QualityModel which is understandable to users and at the same timemappable to Network Parameters that providers use to ensurethe Quality delivered. (Using Quality Classes does in no wayrestrict the provider from offering a fine granularity of Qualitylevels if it so chooses.)

The ETSI Project TIPHON is used as an example of an attemptto create such a model. The proposed Quality Classes aregeneric and easily understood. However, they are perhaps toogeneric to be applicable to all kinds of services.

The TIPHON Quality Model is also used as a basis for someinitial ideas on how the model can be extended.

Magnus Krampell is Project Supervisor atEURESCOM GmbH , supervising projects in theareas of IP/Internet and QoS. He is presently onleave from Telia AB, where he has been employedsince 1990. Experiences include: Development ofTMN systems, TINA work at Bellcore in USA and re-search and development of broadband systems andservices. Mr. Krampell received his B.Sc in Mathe-matics in 1984 and his Lic.Sc in Computer Sciencein 1988.e-mail: [email protected]

6 References

1 ITU. Terms and Definitions related to Quality of Serviceand Network Performance including Dependability. ITU-TRecommendation E.800, Geneva 1994.

2 EURESCOM. European switched VC-based ATM NetworkStudies – stage 2, Deliverable 1 : Specifications, Volume 1of 11 : Outline of the Specification. EURESCOM ProjectP515, www.eurescom.de.

3 IETF draft-ietf-rsvp-spec-11, Resource Reservation Proto-col, RSVP – Version 1, Functional Specification, February,1996.

4 ETSI. Network Aspects (NA) : General Aspects of Quality ofService (QoS) and Network Performance (NP). ETSI Tech-nical Report ETR 003, 1994.

5 EURESCOM. Quality of Service : measurement methodselection, Deliverable D1 : QoS overview, services andmeasurement methods – selected services and methods.EURESCOM Project P603, www.eurescom.de, 1997.

6 ETSI. Network Aspects (NA) : Quality of Service indicatorsfor Open Network Provisioning (ONP) of voice telephonyand Integrated Services Digital Network (ISDN). ETSITechnical Report ETR 138, 1994.

7 ETSI Project TIPHON, Working Group 5, DTR/TIPHON-05001, version 1.2.2, www.etsi.fr, 1998.

8 Multimedia Communications Forum (MMCF). MultimediaCommunications Quality of Service, Part I : Framework;Part II : Multimedia Desktop Collaboration Requirements.1995 (www.mmcf.org).

9 ETSI. Transmission and Multiplexing (TM); Speech com-munication quality from mouth to ear for 3.1 kHz handsettelephony across networks. ETSI Technical Report ETR250, 1996.

10 ITU-T. Integrated Services Digital Network (ISDN). Over-all network aspects and functions. General aspects of Qual-ity of Service and Network Performance in digital networks,including ISDNs. ITU-T recommendation I.350, Geneva,March 1993.

11 Oodan, A, Ward, K, Mullee, T. Quality of service intelecommunications. IEE Publishing (UK), 1998. ISBN0-85296-919-8.

The European Telecommunications Standards Institute – ETSIwas established in 1988. The first General Assembly (GA) washeld in March the same year. On the occasion of the 10thAnniversary, a special one-day conference was held in Nice,France on 25 March 1998 in conjunction with the General As-sembly. The conference subtitle was raising expectations as tothe contents: “The State of the Art in European Telecommuni-cations Standardization”.

The conference was opened with a speech on ETSI’s historyand role by Dr. Karl Heinz Rosenbrock, Director of ETSI since1990. Then there were seven speeches on specific topics:

• Universal Mobile Telecommunications System (UMTS)• Broadband Radio Access Networks (BRAN)• Terrestrial Trunked Radio Access (TETRA)• Voice Telephony on the Internet (TIPHON)• Multimedia Applications & Communications• European ATM Services Interoperability (EASI)• Telecommunications Security.

The seminar was closed by the chairman of the General As-sembly, Antonio Castillo, who summed up the status in onesentence: “What is state of the art?” The answer was: “Every-thing is changing.”

The seminar was clearly divided into two parts, one coveringradio based communications and the other fixed network ser-vices, systems and applications. Thus the ‘hottest’ subjects werecovered: Mobile communications, Internet and Multimedia.

Introduction

Dr. Karl Heinz Rosenbrock has been Director of ETSI since1990. He made the opening speech by summarizing the story ofEuropean telecommunications standardisation since the found-ing of CEPT in 1959. The title of the speech reflected his viewson the importance of ETSI: “ETSI’s role in standardisation – thepast, the present & the future”. He summed up the facts in num-bers, which are shown in Table 1. ETSI’s technical organisationas per October 1997 is shown in Figure 1.

93

European Telecommunications Standardisation– ETSI’s 10th Anniversary 1998

P E R H J A L M A R L E H N E

Telektronikk 2.1998

• 485 full membership organisations

• Members from 34 European countries

• 81 observing organisations

• 48 associated member organisations

• 1471 Standards (ETS) produced since 1988

• 432 Technical Reports (ETR)

Table 1 Some facts about ETSI as per April 1998

Figure 1 ETSI technical organisation as per October 1997

Rosenbrock’s ‘message’ was about the connection betweenETSI’s main goals and the inner market of Europe. This isclearly stated in one of ETSI’s original objectives:

“To produce the technical standards which are necessaryto achieve a large & unified European telecommunicationsmarket.”

Later, this point has been emphasised in June 1995 to “Makinginternational standards happen first in Europe”.

The ‘efficiency’ of ETSI has improved during the years, at leastmeasured in relation to the number of published reports per yearor the time needed to bring forward a new standard. Rosenbrockspent some time emphasising this point with graphs and figures.For example, the number of published deliverables per year hasincreased from approximately 30 in 1990 to over 600 in 1997,while the cost per deliverable has decreased from over 500 kECUin 1990 to 30 kECU in 1997. Some people would probablyargue that such figures are not the only relevant measure ofsuccess.

The past and present are easy to fill with facts and figures, whilethe future is definitely more ‘in the mist’. The issue is the roleof ETSI in the future. Will ETSI be the main driving force infuture European telecom standardisation? This part of thespeech was as misty as the future might seem. Anyone whocan create a critical mass around a technology is in the positionof producing a standard, and the key to success is inventivenessand creativity. The word ‘convergence’ was much used. WhileRosenbrock’s postulate was that convergence of technologieswill come by itself, what is needed now is the ‘convergence ofpeople’, whatever is meant by that.

Radio based communications

The fact that there were three speeches on radio communica-tions standards shows what a high priority this has.

Friedhelm Hillebrand, chairman of Technical CommitteeSpecial Mobile Group (TC SMG) went through the status for thework towards next generation mobile communication, in ETSIterms UMTS – Universal Mobile Telecommunications System.From the ‘global’ GSM to UMTS ‘covering the Universe’, washis opening words. The framework for UMTS was set up in July1996 with the conclusions from the work with Global Multi-media Mobility – GMM. The focus was shifted for UMTS, andreference is often made to ‘old’ and ‘new’ UMTS. The changesare summed up in Table 2.

Most important to notice is the focus on Internet and IP, as wellas UMTS now being an evolution from GSM instead of a ‘new’network. The first phase of UMTS is the so-called ‘Release 99’having elements from GSM Phase 2+.

Fixed radio systems are also having a great momentum. JanKruys is chairman of the ETSI project Broadband Radio AccessNetworks (EP BRAN), which started up in April 1997. Fixedradio access technology is directly competing with for examplehigh-speed photonic networks, so why spend so much resourceson this? Radio technology has its force in a high degree of flexi-bility, fast and cheap installation and low cost in general. Thefocus of the work is very much on deregulation and multimediaservices. User service demands vary greatly over a very shortperiod of time leading to large variations in bandwidth demand.Additionally, users are very sensitive towards the responsetimes. EP BRAN has taken over the responsibility for HIPER-LAN/1, ETSI’s high speed wireless LAN, from previous TCRES (Radio Equipment and Systems) and this is now part of aseries of specifications for broadband radio access standards.Table 3 shows the three types of broadband radio access net-works that EP BRAN is working towards. The first specifica-tions are expected to be ready in June 1999, while the wholeproject is scheduled to end in June 2002. It is all about radiotechnology with high performance and an open-ended structure.

The work on TETRA – TErrestrial TRunked Radio – started asearly as 1989 in RES6. TETRA will eventually become a familyof standards, and the work is performed in four phases, wherePhases 1 and 2 are finished. Phase 3 is expected to finish in thethird quarter of 1998, while the work on Phase 4 has just started.Brian Oliver is chairman of EP TETRA and among other thingshe focused on the question of success for TETRA. The Memo-randum of Understanding (MoU) has been signed by 66 organ-isations from 19 countries. One of the results from EP TETRA


“Old” UMTS “New” UMTS

Core ideas Integration of all existing 1) Focus on innovative newand new services into servicesone universal network 2) Support of GSM services

Partner-networks Broadband ISDN Intranets and Internet

Introduction Migration from existing Evolution from GSM andnetworks ISDN networks

Roaming New development, Evolution of GSM roaming,INAP based MAP based

Standardization FPLMTS as one mono- IMT 2000 family in ITU andlithic standard in ITU ANSI, ETSI, ARIB/TTC

Table 2 Change of focus for UMTS [1]

HIPERLAN/2 A complement to HIPERLAN/1 providing communicationbetween portable computing devices and broadband corenetworks. User mobility is supported within the local ser-vice area, 50 to a few hundred metres. Up to 25 Mb/s

HIPERACCESS An outdoor, high speed radio access network, providingfixed radio connections to customer premises. Up to 25 Mb/s

HIPERLINK A very high speed radio network for infrastructure-likeapplications; a typical use is the interconnection of HIPER-ACCESS networks and/or HIPERLAN Access points into afully wireless network. Up to 155 Mb/s.

Table 3 EP BRAN types of broadband radio networks

is the specification of the Digital Advanced Wireless Service(DAWS). DAWS is based upon TETRA’s advanced PDO(Packet Data Optimised) protocol. It is possible to obtain fullATM speed of 155 Mb/s with full mobility. EP TETRA is co-operating closely with EP BRAN and EP TIPHON (see nextsection), as well as TC SMG.

Fixed network standards, systems and services

The second main session of the seminar focused on future highspeed networks and services.

Internet Telephony, or Voice over IP, is a hot topic. ETSI hasrecognised this fact and started EP TIPHON – Telecommuni-cations and Internet Protocol Harmonization Over Networks.Helmut Schink is chairman for TIPHON which started in April1997. The focus of the project is on voice- and voice-band com-munications over IP and the interoperability with PSTN/ISDN/GSM. Voice over IP is based on ITU recommendation H.323from ITU-T SG16 [2], and there are strong relations to VoIP-Forum/IMTC (International Multimedia Teleconferencing Con-sortium), IETF (The Internet Engineering Task Force) and ITU-T. Several possible scenarios were presented, eg. ‘IP to PSTN/ISDN/GSM’ and ‘PSTN/ISDN/GSM to PSTN/ISDN/GSM overIP’, as shown in Figure 2. One of the problems currently beingfocused is the addressing of TIPHON IPs. Traditional IP add-resses are both complex and dynamic, and not very useful froma standard telephone keyboard. The solution is to use ITU’sE.164 numbering scheme. There is a close co-operation withTC NA (Network Aspects) on this.

Rolf Rüggeberg, chairman of ETSI Project Multimedia Ter-minals and Applications (EP MTA) opened his speech bysetting up a definition of Multimedia:

"Integration of text, data, graphic, audio, voice and video ina uniform user platform by simultaneous usage of more thanone of these elements."

He went on to show a list of different bodies working with mul-timedia standardisation: 11 global, 9 European and 3 USregional, and the list was probably not exhaustive. The point inshowing this was probably to focus on the interest for the topic,but also show how difficult it can be to work on standards formultimedia. The main focus of this speech was to show somepossible future visions (or horror-scenes?), as for example:

It was pointed out that this is not necessarily the future we want,but it could be possible.

The Interoperability of ATM networks and services are dealtwith by the ETSI Project European ATM Services Interopera-bility (EP EASI). The chairman, Mike Bexon, focused in hisspeech very much on ATM MoU. EP EASI started as late asOctober 1997. The ATM MoU Working Group (WG) was setup by ETNO (European Public Telecommunications NetworksOperators’ Association) in January 1997, and in co-operationwith EURESCOM, ETSI and the ACTS project JAMES (JointATM Experiment on European Services), the MoU was ready inDecember 1997. The goal of EP EASI is to work out two Euro-pean Norms (ENs) and two so-called guideline documents. TheENs are meant to be ‘Supe-ICSs’ (Implementation Confor-mance Statements).

Security in future telecommunications

Technical Committee Security (TC SEC) works on securitymatters in modern, digital telecommunications. TC SEC ischaired by Michael Walker, and his speech addressed thedifferent aspects and the importance of security. Security isimportant to avoid eavesdropping and fraud. Previously, this hasbeen solved by using a special hardware module and crypto-chips to ‘secure’ protocols and firewall technologies. The bigchange came with the migration from analogue to digital mobiletelephony where security also encompasses authentication andencryption, and not least, the Subscriber Identity Module (SIM).The encryption also functions to authorise the use of the radiochannel. In electronic commerce, security is very important, andthere are great challenges on secure electronic transactions frommobile and/or personal terminals. Today the payment terminalsare fixed installations in shops, at petrol stations and so on. Alot of the standards work is carried out outside ETSI, as for ex-ample SET – Secure Electronic Transaction, and a lot of thebasics on encryption techniques have been put into the X.400and X.500 recommendations from ITU.


The Networked Home:

11 March 2009:

You are approaching your home, the Home Area Network(HAM) reports:

All domestic appliances o.k.No unusual occurrences.Call up security report.You have 5 items of mail waiting, including aregistered letter.What would you like for dinner?Please enter your mealtime.

IP Network

IWFLocal or distributedfunction IWF Local or distributed

function

PSTN/ISDN/GSM

PSTN/ISDN/GSM

Figure 2 Scenario ‘PSTN/ISDN/GSM to PSTN/ISDN/GSM over IP’

Final words

Antonio Castello, chairman of ETSI General Assembly (GS)closed the seminar by, among other things, postulating that thereal ‘State of the Art’ is that everything is changing. What holdstoday, does not necessarily hold tomorrow, and ETSI as anorganisation must relate itself to that fact.

References

1 Hillebrand, F. Universal Mobile Telephone System. In:ETSI 10th Anniversary Conference : The State of the Art inEuropean Telecommunications Standardization. Nice,France, 25 March 1998.

2 Ulseth, T. ITU-T multimedia standardisation activities.Telektronikk, 93, (2), 87–89, 1997.

Abbreviations

ANSI American National Standards Institute

ARIB Association of Radio Industries and Businesses

ATM Asynchronous Transfer Mode

BRAN Broadband Radio Access Network

DAWS Digital Advanced Wireless Service

EASI European ATM Services Interoperability

EN European Norm

EP ETSI Project

ETNO European Public Telecommunications NetworksOperators’ Association

ETR ETSI Technical Report

ETS ETSI Technical Specification

FPLMTS Future Public Land Mobile TelecommunicationsSystem

GMM Global Multimedia Mobility

GSM Global System for Mobile Communications

HIPERLAN High Performance Radio Local Area Network

IETF Internet Engineering Task Force

IMT-2000 Integrated Mobile Telecommunications 2000

IMTC International Multimedia TeleconferencingConsortium

INAP Intelligent Networks Application Protocol

IP Internet Protocol

ISDN Integrated Services Digital Network

ITU International Telecommunications Union

MAP Mobile Application Protocol

MTA Multimedia Terminals and Applications

PSTN Public Switched Telephone Network

RES Radio Equipment and Systems

SIM Subscriber Identity Module

SMG Special Mobile Group

TC Technical Committee

TETRA TErrestrial TRunked Radio Access

TIPHON Telecommunications and Internet ProtocolHarmonization Over Networks

UMTS Universal Mobile Telecommunications System


Per Hjalmar Lehne is Research Scientist atTelenor Research & Development, Kjeller. Heis working in the field of personal communica-tions, with a special interest in antennas andradio wave propagation for land mobile com-munications.

e-mail:[email protected]

Kaleidoscope


Contents

I Introduction

II Determination of the single subscriberconnection probability

III General calculation of efficiency and degree ofhindrance

IV The first simplification of the calculations

V The second simplification. Calculation of thedegree of hindrance

VI Other procedures for elucidation of the likelyefficiency. Comparison of the different ex-pressions for the allowable hindrance

VII Objections to the applicability of the calcula-tions. Calculations on a different basis

VIII Calculation of the degree of hindrance for thedaily traffic

IX Calculation of the degree of hindrance in sys-tems with skewing of circuits (slip multiple)

X Calculation of the degree of hindrance in sys-tems with spare equipment

XI Calculation of the degree of hindrance withcombination of skewing and spare equipment

XII Summary

Annex I Table showing the degree of hindrance forvarious values of the number of circuits andtraffic intensity according to calculations basedon Bortkewitsch’s tables

Annex II Table showing the degree of hindrance forvarious values of the number of circuits andthe traffic intensity according to calculationsbased on simplified formulas as presented inthe report

Annex III Presentation of traffic intensity and efficiencyvalues corresponding to various values of thenumber of circuits, given a degree of hind-rance of 0.001

Annex IV Diagram showing the likely relation betweentraffic and circuits in an ordinary automatictelephone system

Annex V Diagram showing the likely efficiency of circuitsin an ordinary automatic telephone system

I Introduction

1. We will try to apply probability calculation to determine thenecessary number of switches in a switch group and within asection of co-operating switches in a group.

The mathematical concept of ‘probability’ is commonly definedin the following way:

The probability of an event E (Ereignis, événement) is the ratiobetween the number of cases that are favourable for the occ-urrence of this event and the total number of possible events,given that all cases are equally likely and mutually independent.

Given, for example, an urn with N balls, of which there are mwhite and n black, m + n = N, and if those balls by a suitableaction are mixed so that they are mutually non-orderly located,then the probability that a single ball picked is a white ball= m/N. Likewise, the probability of picking a black ball = n/N.

In the following we shall apply the two main theorems ofprobability theory.

The first theorem states that among different events E1, E2, E3,..., En, that depend on the probabilities p1, p2, p3, ..., pn, theprobability P(1+2+3+ ... +i) that either E1 or E2 or E3 ... or Ei willoccur = p1 + p2 + p3 + ... + pi, i.e. equal to the sum of theprobabilities of the single events.

If this is applied to an urn with balls, e.g. n1 white, n2 red, n3green, n4 black, altogether n1 + n2 + n3 + n4 = N balls, and if weset n1/N = p1, n2/N = p2, n3/N = p3, n4/N = p4, then the proba-bility of picking either one white or one green ball (irrespectiveof which) is: P(1+3) = p1 + p3.

The second main theorem states that if different events E1, E2,E3, ..., En, depend on, respectively, p1, p2, p3, ..., pn, then theprobability P1.2.3 ... i that both E1 and E2 and E3 ... and Ei willoccur – i.e. a combination of those events will occur –= p1p2p3 ... pi, being the product of the single event proba-bilities.

At trials with the last mentioned urn this means that the proba-bility of picking a red ball and, after replacing it, picking a greenball (those two pickings being considered a combination orcommonality) will be: P2.3 = p2p3.

We will also have to consider complementary probabilities, i.e.probabilities that exclude one another, i.e. the probability p thatan event will occur and the probability (1 – p) that it will notoccur.

In the first mentioned example of white and black balls m/N andn/N are complementary probabilities Their sum is (m+n)/N = 1.In the second example among others (p1 + p3) and (p2 + p4) arecomplementary probabilities. Similarly, at the combined pick-ings, p2p3 and (1 – p2p3) are complementary probabilities.

99

On the calculation of switches in an automatic telephone systemAn investigation regarding some points in the basis for the application of probability

theory on the determination of the amount of automatic exchange equipment

B Y T . E N G S E T , K R I S T I A N I A 1 9 1 5

T R A N S L A T E D B Y A R N E M Y S K J A

Telektronikk 2.1998

Some comments on the translation are found onpage 44.

II Determination of the connectionprobability of single subscribers

2. When one intends to apply probability theory to the localtelephone correspondence, one may proceed in various ways.One may, for instance, put the main weight on a presentationtaking into account that calls from different subscribers to theexchange in general occur at different time instants, and onemay try to calculate the probability that a certain number ofcalls arrive during a certain interval, corresponding to the meanconversation time within some specified period, for instance thebusiest hour of an ordinary workday. This approach, however,will incur substantial difficulties, unless arbitrary simplifica-tions are made, the influence of which hardly can be estimatedin an obvious way.

As the intention is to obtain an insight into the process of theaggregated telephone traffic at any instant within working hours– be it for the whole subscriber network or for certain larger orsmaller groups of subscribers – we have preferred to direct ourattention to what happens or can be assumed to happen at anarbitrary epoch within a specified work period. In other words,we direct the focus of our investigation at simultaneous eventsat each moment of the period under consideration.

We name this period the selected time Tc. Within Tc, which maybe the busiest hour or the busiest quarter-hour or some otherspecified work time, traffic is assumed to flow fairly evenly,and for a longer time space Tt show, for a given exchange, acertain mean number of calls r per Tc with a certain meanduration t per call.

Tt is measured in days, Tc and t in seconds. The product rt thusindicates the mean aggregate call (or conversation) secondswithin the period Tc, and rt / Tc = y indicates the mean numberof simultaneous local telephone connections at the exchangeduring the period Tc. It is assumed that r and t are found by avery large number of observations within the telephone networkconcerned under current, satisfactory manual telephoneoperation during a lengthy period.

3. We will assume that n subscribers are connected to theexchange. If throughout Tc there were exactly y connections,then according to the given probability definition above, thegeneral mathematical probability pg of finding a subscriber busywith an own initiated call at a random epoch within Tc would beexactly y / n. Since y is only an average number, as the realnumber of simultaneous connections in Tc can vary between 0and substantially more than y (eg. 2y or even more), one cannotgive any exact, real probability pr for finding a calling sub-scriber by a single sample. One must be satisfied by an appro-ximate indication by setting pr ≈ y / n.

Thus pr ≈ y / n in a way indicates the average likelihood duringtime Tc to find any particular subscriber in the course of a con-versation initiated by himself. We can, however, not acceptright away the quantity y / n as a measure of our subscribers’proneness to make calls. We know there are great differencesbetween different subscribers’ use of the telephone. We musttherefore make a judgement as to whether the estimate of y / n(or some differing value, in any case a small proper fraction lessthan one) implies any harmful error, ie. an error leading to und-

underestimation of traffic requirements by calculating a toosmall number of switches.

By our considerations we will remember the great regularitythat is apparent in the values of r and t at one particular tele-phone network through a lengthy period, and the value r’sdependence on the number of subscribers and the time of day(the concentration). These average values emerge as a sum or asan extract of the total calling activity of the whole set of sub-scribers. The single subscribers act in the calling sense quiteuncoordinated. In the long run, however, each subscriber mustsatisfy a certain requirement that is particular for him.

A particular subscriber will be denoted Ai. The total time spentby this subscriber during the selected period Tc of each daythrough a lengthy time space of Tt = D days shall be denoted Ti,measured in seconds. Ti / (D · Tc) = pi denotes the ratio betweenthe time spent on calling and the total time available for thissubscriber.

One may also view the case in such a way as to assume the totaltime under observation, D · Tc, to be subdivided in a very largenumber (N) of very short time intervals ∆t. Thus D · Tc = N · ∆t.For subscriber Ai’s telephone calling there will, out of the totaltime N · ∆t, be mi favourable intervals and (N – mi) = ni un-favourable intervals ∆t. Here we set mi · ∆t = Ti and consequ-ently ni · ∆t = (DTc – Ti).

From the above follows

(1a)

(1b)

4. We will now assume that we take momentary test samples atthe exchange whether Ai’s line is occupied by an outgoing call,as we suppose such a sample technically realised in a suitableway. We will let the instant of any sample be completelyrandom, with no pre-knowledge whether the line is occupied ornot. According to the nature of the sampling one can assumethat all cases are equally likely.

Furthermore, as we can assume that any start and close of aconnection are independent relative to the sampling points, wemust be allowed to consider the sample results as independent,in spite of the fact that a very large number of ‘connectioncases’ or intervals with connection (corresponding to a singlecall) will follow immediately after one another.

5. In order to make sure that we do not expose ourselves to anyself-delusion by these considerations, we shall make a compari-son with the previously treated example of white and blackballs. Instead of placing these balls in an urn where they bysome suitable movement are mixed, we will assume the ballsfastened along the rim of a circular disk that can be turnedaround a vertical axis in the centre of the disk. The disk may beset in an even rotation by some mechanism. The balls, all of thesame size, are densely positioned along a circle line with centrein the axis of the disk. They are hidden behind a screen with asmall slot in front of the ball ring, through which a single ballmay be observed as it passes the slot during rotation. The slot isnormally closed by a slide. By a quick opening of the slide at

1− pi( ) =1−miN

=N − mi

N=

niN

.

pi =Ti

D ⋅Tc=

mi ⋅ ∆tN ⋅ ∆t

=miN

,


any moment a white or a black ball may be observed throughthe slot. At a simultaneous view of the adjacent parts of twoballs it is an established rule that the ball leaving the slot is notconsidered. The slide is only kept open a short moment, so thatonly a part of a ball can pass the slot during opening.

When now, as in the first mentioned example, the number ofwhite balls is m, the number of black balls is n and the jointnumber m + n = N, then the probability of observing a whiteball in an arbitrary moment is p = m / N and the probability ofobserving a black ball is (1 – p) = n / N, exactly like theprobability of picking a white or a black ball from the beforementioned urn. By repeated samples with the rotating ball ringthe choice of observation moments must be made withoutknowledge of the speed of rotation, so that the conditions of theprobability definition, concerning even likelihood of occurrencefor all cases and their mutual independence, are fulfilled. Underthis condition we may consider the ball urn and the hiddenrotating ball ring as fully equivalent mechanisms for illustrationof the concept of mathematical probability.

6. The analogy between the ball ring with m white and n blackballs, altogether N balls, and the case of total time N · ∆t with miconnection intervals and ni non-connection intervals, altogetherN intervals ∆t is obvious. There is, however, the difference thata traversed time space N · ∆t with its telephone calls never re-peats itself, whereas the ball ring for each round presents thesame order, unless it is changed by some particular arrange-ment. Thus telephone traffic presents a picture of more disorderand randomness, implying that there is less need to fear that theconditions of the probability definition are or will be unfulfilled,when the events of connection are observed. On the other handone should be aware that there is no constant probability relatedto the single subscriber line such that that of the urn or ball ring.More about this later.

7. It appears from the discussion that the quantity of pi = mi / Nfor subscriber Ai’s call activity, in analogy with the quantityp = m / N for the ball trial, has the character of a probability,namely the probability that subscriber Ai at an arbitrary instantis busy with a self-initiated call, and that the quantity (1 – pi) =ni / N for the same line – corresponding to the ball trial’s(1 – p) = n / N – should be considered as the complementaryprobability, ie. the probability of finding the same line not busywith a self-initiated call. (Any incoming call from other partiesto Ai are neglected).

8. According to the preceding arguments we feel justified topresent the following definition: A subscriber Ai’s connectionprobability pi is the ratio between his assumed time Ti of outgo-ing call occupancy and the total time D · Tc.

In order to avoid misunderstanding we have here used the term‘assumed’ rather than ‘favourable’1), that is usually applied inthe general definition of probability.

III General calculation of efficiency anddegree of hindrance

9. We will now consider a group of subscribers A1, A2, A3, ..., Ai, ...,An, whose connection probabilities in the selected period are p1, p2,p3, ..., pi, ..., pn, respectively, and which have jointly available forconnection through the exchange a certain number x of adjacentcircuits with the necessary equipment.

Assuming that p1, p2, p3, ..., pi, ..., pn are known and exactly givenconstant quantities, we shall calculate the probabilities P0, P1, P2,..., Pr, ..., Px of finding, respectively, 0, 1, 2, 3, ..., r, ..., x out of theavailable circuits busy with calls at an arbitrary moment within theselected interval Tc on a similarly arbitrary day within Tt. Further-more we shall calculate the probabilities Px+1, Px+2, ..., Px+(n–x) offinding x + 1, x + 2, ..., n circuits busy in case one had x + 1 or x + 2or ..., or n circuits available at the exchange instead of just x.

In order to facilitate the process of this calculation we shall apply acouple of easily understandable product symbols. We shall establisha product of r out of the n given quantities p1, ..., pn

only once in any product. According to combination theory

products. Each of these products can be identified by a number

behind a parenthesis in front of the product expression:

no ambiguity can occur. When we have formed a product of rprobabilities p, eg. p1p2p3 ... pr, we will form another product(1 – pr+1) (1 – pr+2) ... (1 – pn) of the complementary probabilitiesof the remaining (n – r) probabilities p. In general we

product of r probabilities there will be a product of complementaryprobabilities corresponding to the remaining (n – r) probabilities.We will furnish each of the such given complementary productswith the same number as the primary product.

omit the latter number indication as superfluous when we presentthe product of both those products, thus writing:

Thus to the primary product ( j pi(n)

(r)

∏ corresponds the

counterproduct ( j 1− pk( )k≠i(n)

(n−r)

∏ . We will, however,

such counterproducts will likewise be nr

since for each

shall denote such a product by 1− pk( )k≠i

(n−r)

∏ . The number

( j pi(n)

(r)

∏ . The notation (n) under ∏ may be omitted when

within the series 1,2,K , j,K nr

by putting the number

there will be altogethern n −1( )K n − r +1( )

1⋅2 ⋅3K r= n

r

such

and denote this by pi(n)

(r)

∏ , where any quantity pi appears


1) The Norwegian text applies the terms ‘paaregnelig’ and‘gunstig’.

A sum of all products formed this way for a given r will bedenoted in the usual way:

10. We will now calculate P0, the probability that all x circuitsare found to be free by a sample test, ie. the probability of noconnection.

As subscriber Ai’s connection probability is pi, then his non-connection or idleness probability is (1 – pi). Likewise Aj’s idl-eness probability is (1 – pj). As the matter in question here is theprobability that all of A1 and A2 and A3 and ... and Ai and Aj and... and An are without a telephone connection, the secondprobability theorem implies that the requested probability isgiven by the product of all the idleness probabilities:

(2,0)

In order to find the probability of observing only one connectiona more compound operation is needed. First we seek theprobability of finding A1 busy and all the other subscribers idle.This probability will be denoted p(1). According to the sametheorem we have:

p(1) = p1(1 – p2) (1 – p3) ... (1 – pn)

In the same way are found:

p(2) = (1 – p1)p2 (1 – p3) ... (1 – pn)

...

p(k) = (1 – p1) (1 – p2) ... pk ... (1 – pn)

...

p(n) = (1 – p1) (1 – p2) ... (1 – pk) ... (1 – pn–1)pn

As the question is whether the first or the second or the third or... or the nth case will be observed, one will according to thefirst theorem have:

(2,1)

In order to find the probability of observation of two connec-tions we will first seek the probability p(1,2) of finding A1 and A2in connection and simultaneously all the others without. Thenwe get:

In the same way we obtain:

p(1,2) = p1p2 1− p3( ) 1− p4( )K 1− pn( )= (1Π

(2)pi Π

(n−2)1− pk( )k≠i

P1 = p(1) + p(2) +K + p( j)K + p(n)

= p( j)j=1

j=n

∑ =j=1

j=n

∑ ( j Π(1)

pi Π(n−1)

1− pk( )k≠i

P0 = 1− p1( ) 1− p2( ) 1− p3( )K 1− pk( )K 1− pn( ) = Π(n)

1− pk( )

j=1

j= nr

∑ ( j pi(n)

(r)

∏ ⋅ 1− pk( )k≠i(n)

(n−r)

∏

( j pi(n)

(r)

∏ ⋅ 1− pk( )k≠i(n)

(n−r)

∏

... ...

Furthermore:

... ...

... ...

... ...

Hereby we obtain:

(2,2)

... ...

In a similar way is obtained:

(2,3)P3 =

j=1

j= n3

∑ ( j Π(3)

pi Π(n−3)

1− pk( )k≠i

P2 = p(1,2) + p(1,3) +K + p(1,n) + p(2,3) +K + p(2,n) + p( j,n)

+K + p(n−1,n) =α=1

α=n−1

∑β=2

β=n

∑ p(α,β )α<β

=j=1

j= n2

∑ ( j Π(2)

pi Π(n−2)

1− pk( )k≠i

p(n−1,n) = 1− p1( ) 1− p2( )K 1− pn−2( ) pn−1pn

= (n n −1( )

1⋅2 Π(2)

pi Π(n−2)

1− pk( )

p( j,n) = 1− p1( ) 1− p2( )K 1− pj−1( ) pj 1− pj+1( )K pn

= (q Π(2)

pi Π(n−2)

1− pk( )

p(2,n) = 1− p1( ) p2 1− p3( ) 1− p4( )K 1− pn−1( ) pn

= ( 2n − 2( ) Π(2)

pi Π(n−2)

1− pk( )

p(2,3) = 1− p1( ) p2 p3 1− p4( ) 1− p5( )K 1− pn( )= (n Π

(2)pi Π

(n−2)1− pk( )

p(1,n) = p1 1− p2( ) 1− p3( ) 1− p4( )K 1− pn−1( ) pn

= ( n −1( ) Π(2)

pi Π(n−2)

1− pk( )

p(1,4) = p1 1− p2( ) 1− p3( ) p4 1− p5( ) 1− p6( )K 1− pn( )= (3 Π

(2)pi Π

(n−2)1− pk( )

p(1,3) = p1 1− p2( ) p3 1− p4( ) 1− p5( )K 1− pn( )= (2 Π

(2)pi Π

(n−2)1− pk( )


(2,4)

... ...

(2,x)

(2,x+1)

... ...

(2,n)

The last equation expresses the probability Pn of finding at anarbitrary moment all n subscribers occupied by self-initiatedcalls on a set n of exchange circuits, irrespective of possible callcompletion due to the state of called subscribers. This proba-bility is equal to the product of the connection probabilities ofall subscribers, a result that also follows directly from thesecond theorem.

11. With the probabilities P0, P1, P2, ..., Px, ..., Pn all possi-bilities with n internal connections are exhausted, since any testwill find either 0 or 1 or 2 or 3 or ... or n circuits occupied, incase there are altogether n circuits. Thus the sum of all thoseprobabilities must express certainty:

P0 + P1 + P2 + ... + Px + ... + Pn = 1.

If there are only x circuits available, then the alternativeprobability of finding either 0 or 1 or 2 or 3 or ... or x circuitsoccupied in an arbitrarily chosen moment is P0 + P1 + P2 + ... +Px. The probability of finding all x circuits busy and simul-taneously 1 or 2 or ... or (n – x) ‘standing’ calls without possi-bility of progress is Px+1 + Px+2 + ... + Px+(n–x) = 1 – (P0 + P1 +P2+ ... + Px), where P0, P1, P2, ..., Px, ..., Pn are determined bythe expressions above, given as functions of p1, p2, p3, ..., pn,which quantities are assumed known, and indicate the indiv-idual probabilities for each subscriber of being connectedthrough the exchange of an ideal telephone system with norestrictions. In the determination of p1, p2, p3, ..., pn enter allcalls and all telephone use, even calls that encounter busy or no-answer subscribers.

As can be seen, by the applied method under the conditionmentioned, we have been able to calculate an expression of sys-tem efficiency, given by the probability P(0x) = P0 + P1 + P2 +... + Px, with full mathematical exactness without anyhypothesis concerning the single call or connection duration orconcerning the average duration of calls or connections.

Pn =j=1

j= nn

∑ ( j Π(n)

pi Π(0)

1− pk( )k≠i

= pi(n)

(n)

∏ = p1p2 p3K piK pn

Px+1 =j=1

j= nx+1( )

∑ ( j Π(x+1)

pi Πn− x+1( )( )

1− pk( )k≠i

Px =j=1

j= nx

∑ ( j Π(x)

pi Π(n−x)

1− pk( )k≠i

P4 =j=1

j= n4

∑ ( j Π(4)

pi Π(n−4)

1− pk( )k≠i

12. We shall now proceed to apply the obtained exact and gen-eral results in an appropriate manner. First of all we notice thatthe quantities P0, P1, P2, ..., Px, ..., Pn indicate the probabilitiesof occurrence of certain states. We shall, therefore, call themstate probabilities. P0, for instance, gives a measure of howoften we can expect to find all x (or n) circuits free. If we take avery large number of samples, N, and if the number of ‘favour-able’ cases for total non-occupancy is q0, then according to our

words: among the N samples we expect to find q0 cases withoutany connection. In reality the result will differ to a larger orsmaller extent from that, however, if we gradually increase thenumber of samples, then the fraction q0/N will gradually

other cases will behave:

q1 / N will approach the calculated value of P1

q2 / N will approach the calculated value of P2

...

qx / N will approach the calculated value of Px

...

qn / N will approach the calculated value of Pn

We have not, however, by this obtained any direct expressionfor the relative amount of telephone calls that pass unimpededwith the application of x circuits for a set of n subscribers. Butthis can now theoretically be carried out by means of the foundexpressions.

13. We will again consider the total time D · Tc = N · ∆t, whereN is a very large or extremely large number and ∆t is a verysmall or extremely small time interval. We assume that samplesare taken momentarily. If a test falls exactly between two inter-vals, we will assume, like in the ball ring test, that it falls in thesecond interval. The intervals are assumed to be so short thatwithin one interval there will not be more than one state change.If a test falls exactly between two different states, (it occurssimultaneously with a state change), we will consider it belong-ing to the second state. We choose N so large, and hence theinterval ∆t so small that two adjacent tests will not fall withinthe same interval, ie. we let ∆t converge towards the negligibletime that is necessary to record a certain state. With these con-ditions a non-ambiguous marking of the state of any intervalwill be ascertained.

As will be remembered, for the total time N · ∆t there will foreach subscriber Ai be a favourable or presumed connection timeof mi · ∆t seconds, as mi · ∆t = pi · N · ∆t. The sum of presumedconnection times of all subscribers during the total time isTs = (p1 + p2 + p3 + ... + pi + ... + pn) · N · ∆t seconds.

is the average connection time per subscriber and

TS / n = pi1

n

∑ ⋅ N ⋅ ∆t / n

approach 1− pk( )(n)

(n)

∏ . In a similar way sample tests of

calculations and definitions P0 = 1− pk( ) = q0 / N(n)

∏ . In other


gives the average number of simultaneous connections per inter-val ∆t.

On the other hand we will now study how the variousconnection states are distributed over the total time N · ∆t. Asindicated, the number of cases or intervals that are favourable orpresumed for the number of (i) simultaneous connections is qi,where qi/N = Pi, ie. qi = N · Pi . Consequently, the total durationof corresponding states is ti = qi · ∆t = Pi · N · ∆t. Hereby thetotal time N · ∆t will be subdivided in time sections t0, t1, t2, ...,ti, ..., tn, corresponding to the n + 1 different connection or callstates P0, P1, P2, ..., Pi, ..., Pn. In other words, in an ideal systemone can presume that out of N · ∆t seconds there will be:

t0 = P0 · N · ∆t seconds with no connection,

t1 = P1 · N · ∆t seconds with one connection,

t2 = P2 · N · ∆t seconds with 2 connections, and so on up to

tn = Pn · N · ∆t seconds with n connections.

From this follows the presumption that there in the time

t0 must be carried 0 · P0 · N · ∆t connection seconds




...

ti must be carried i · Pi · N · ∆t connection seconds

...

tx must be carried x · Px · N · ∆t connection seconds

...

tn must be carried n · Pn · N · ∆t connection seconds__ ___________________________Sum:N · ∆t Ts = (p1 + p2 + p3 + ... + pi + ... + pn)

· N· ∆t seconds

Hereby we have obtained to get the whole presumed ‘traffic’subdivided on n + 1 sections of the total time. Each time sectionti is of course in reality composed of a number of disjoint timeintervals, which should not prevent a compound view of allintervals characterised by the same connection state.

14. With the results thus obtained we are now able to indicatethe amount of presumed traffic that we are likely to carry in anideal system with x internal circuits, and consequently the likelyamount that will be prevented because of the lack of circuits orswitches.

The sum of the required traffic, measured in connectionseconds, that is carried during the time t0 + t1 + t2 + ... + txamounts to (1 · P1 + 2 · P2 + 3 · P3 + ... + x · Px) · N· ∆tconnection seconds. During this time there are no hindrances, asthe number of simultaneous connections do not surpass x. In theremaining time tx+1 + tx+2 + ... + tn–1 + tn a traffic of (x · Px+1 +x · Px+2 + x · Px+3 + ... + x · Pn–1 + x · Pn) · N· ∆t will be carried,since all x circuits are busy with traffic during this time, where-as (1 · Px+1 + 2 · Px+2 + 3 · Px+3 + ... + (n – x) Pn) · N · ∆t

TSN ⋅ ∆t

= pi1

n

∑


connection seconds during this time will be prevented frombeing carried.

The ratio between the part of traffic that can be carried by x co-ordinate exchange circuits and the entire presumed traffic fromn subscribers will be designated the efficiency of x internal cir-

cuits for n subscribers and denoted . The ratio between the

part of the traffic that is not carried by x such circuits and theentire presumed traffic from n subscribers will be designated thedegree of hindrance of x internal circuits for n subscribers and

denoted .

From the presentation above now follows:

(3)

where P1, P2, ..., Px, ..., Pn are given the values calculated aboveas functions of p1, p2, ..., pn.

Likewise is obtained:

(4)

where Px+1, Px+2, ..., Pn are given the values calculated above asfunctions of p1, p2, ..., pn.

15. Hence the problem that we have presented is theoreticallysolved under the conditions given, according to which theprobabilities p1, p2, p3, ..., pn are known, constant quantities andthus during the traffic process are assumed to remain non-in-fluenced by hindrances because of lack of circuits. The laststatement, however, is virtually the case only when the numberof such hindrances is very small compared with the number of‘busy’ messages.

As it appears from the above, the quantities and are

proper fractions such that + = 1. If one wants

efficiency and hindrance expressed in percent or per thousand,one can use the notations

ϕoonx

ϕooonx

, and ψ oonx

or ψ ooonx

, where we have

ϕoonx

=100 ⋅ϕnx , ϕooonx

=1000 ⋅ϕnx

ψ oonx

=100 ⋅ ψ nx , ψ ooonx

=1000 ⋅ ψ nx .

ψ nxϕnx

ψ nxϕnx

ψ nx = (1⋅ Px+1 + 2 ⋅ Px+2 + 3 ⋅ Px+3 +K +(n − x)Pn )

⋅ 1( p1 + p2 +K + pn )

,

ϕnx = (1⋅ P1 + 2 ⋅ P2 + 3 ⋅ P3 +K +(x −1)Px−1

+ x(Px + Px+1 + Px+2 +K +Pn )) N∆tTs

= (1⋅ P1 + 2 ⋅ P2 + 3 ⋅ P3 +K +(x −1)Px−1

+ x(Px + Px+1 + Px+2 +K +Pn )) 1( p1 + p2 +K pn )

,

ψ nx

ϕnx

IV The first simplification of thecalculations

16. When we next will try to bring the found general expres-sions to application on actual problems of a practical nature,we will, as is easily seen, encounter a multitude of difficulties.These can only be bypassed by sacrificing a fully exact cal-culation.

If in a network there is a large number of subscribers, it will bepractically impossible to evaluate each single subscriber’sconnection probability. Even if we were able to, by anenormous amount of counting and measurement, to determinethe quantities pi, (i = 1, 2, 3, ..., n), then by the calculation of thestate probabilities Pi (i = 0, 1, 2, 3, ..., x, ..., n) even for a limitedsubscriber number such as n = 100, the number of multipli-cation and summation operations would be so huge as to makethe calculation virtually impossible.

Since the single subscribers’ connection probabilities, for rea-sons unnecessary to discuss, will not stay constant, whereas thetraffic intensity of the collective set of subscribers by experi-ence is seen to stay nearly constant through years, save periodicvariations over day and season, we will direct our attention atthe average traffic intensity per subscriber during the selectedtime (eg. the busiest hour). This is a quantity that can relativelyeasily be determined with rather great exactness.

17. Let us assume that we have found the average connectiontime of T seconds per subscriber during the selected time of Tcseconds in a system with no restrictions. The averageconnection time per subscriber then is

p = T / Tc.

Thus p is exactly determined.

Instead of the real set of n subscribers with in general differentconnection probabilities we consider an imagined set of n sub-scribers with each the same connection probability p. We will

now calculate and for this subscriber set. That is

done by setting p1 = p2 = p3 = ... = pn = p. In order to indicatethe altered character of the calculated quantities we will put abar over the letters that indicate these quantities. Hereby we get:

(5,0-n)

P0 = 1− p( )n

P1 = n1

1− p( )n−1 ⋅ p

P2 = n(n −1)1⋅2 1− p( )n−2 ⋅ p

2

P3 = n(n −1)(n − 2)1⋅2 ⋅3 1− p( )n−3 ⋅ p

3

K

Px =n(n −1)K n − x +1( )

1⋅2K x1− p( )n−x ⋅ p

x

K

Pn = pn

.

ψ nxϕnx

Consequently we obtain:

(6)

(7)

It should be remembered here that p denotes the fraction of theselected time (the busiest hour) that on average occupies eachsubscriber during outgoing calls. np thus denotes the averagenumber of existing connections over that time.

this time ‘gets lost’ by being conveyed on x circuits, while thedifferent subscribers are assumed to have the same average useof the telephone, but otherwise use the telephone randomly.

and we must try to get an idea whether it by a suitable degree of

and also whether the quite essential simplification done willlead to the target of obtaining practically useful formulas.

It is thus a matter of clarifying how the quantities P0, P1, P2, ...,Px, ..., Pn, considered as functions of p1, p2, ..., pn, change byequalising p1, p2, ..., pn, given that the sum p1 + p2 + ... + pn =np is constant. By the application of the theory for maxima andminima on the n + 1 equations (2, 0-n) that determine P0, P1,P2, ..., Px, ..., Pn as functions of p1, p2, ..., pn, that are consideredvariable under the condition p1 + p2 + ... + pn = np = constant,one finds that each of the quantities P0, P1, P2, ..., Px, ..., Pnhave either maximum or minimum when p1 = p2 = p3 = ... = pn= p (apart from certain exceptions to be discussed below). Tothe extent that the quantities Px+1, Px+2, ..., Pn have theirmaxima for p1 = p2 = p3 = ... = pn = p, then in any case

probabilities leads to no harmful error for the system operation,since the calculated hindrance is greater than the one obtainedby exact calculation. As we need to be certain not to miscal-

ψ nx > ψ nx , and the equalization of the connection

accuracy can replace the presumed exact quantity of ψ nx ,

18. At present the quantity ψ nx is our primary interest,

ψ nx denotes the fraction of telephone calling that during

ψ nx = (1⋅n n −1( )K n − x +1( ) +1( )

1⋅2 ⋅3K x +1( ) 1− p( )n− x+1( ) ⋅ px+1( )

+2 ⋅n n −1( )K n − x + 2( ) +1( )

1⋅2 ⋅3K x + 2( ) 1− p( )n− x+2( ) ⋅ px+2( )

+3 ⋅n n −1( )K n − x + 3( ) +1( )

1⋅2 ⋅3K x + 3( ) 1− p( )n− x+3( ) ⋅ px+3( )

+K + n − x( ) pn

) 1n ⋅ p

.

ϕ nx = (1⋅ n1

1− p( )n−1 ⋅ p + 2 ⋅n n −1( )

1⋅2 1− p( )n−2 ⋅ p2

+3 ⋅n n −1( ) n − 2( )

1⋅2 ⋅3 1− p( )n−3 ⋅ p3 +K

+ x −1( )n n −1( )K n − x −1( ) +1( )

1⋅2 ⋅3K x −1( ) 1− p( )n− x−1( ) px−1( )

+ x(n n −1( )K n − x +1( )

1⋅2 ⋅3K x1− p( )n−x ⋅ p

x +K + pn

)) 1n ⋅ p


culate, we must try to clarify when the quantities P0, P1, P2, ...,Px, Px+1, ..., Pn reach their maximum and minimum values.

A more detailed analysis indicates that the necessary and suf-ficient condition to make Pr a maximum when p1 = p2 = p3 = ...= pn = p is:

(8)

Here r denotes any number from 0 up to n.

for certain binomial coefficients.

The condition for a minimum value of Pr under the givenassumptions is:

(9)

The conditions for maximum or minimum, respectively can alsobe expressed by

where the upper inequality sign indicates maximum and thelower one indicates minimum. By setting equality sign theabove mentioned exception case (‘undetermined’) is obtained.

By solving the resulting quadratic equation we obtain:

(10)

−2 n − 2r − 2

+ n − 2

r −1

± 4 n − 2

r − 2

+ n − 2

r −1

2

− 4 n − 2r − 2

+ 2 n − 2

r −1

+ n − 2

r

n − 2r − 2

−2 n − 2r − 2

+ 2 n − 2

r −1

+ n − 2

r

= rn

1m 1−r −1( )

r⋅ n

n −1( )

= r

n1m n − r

r n −1( )

> p for maximum< p for minimum

− n − 2r − 2

+ 2 n − 2

r −1

+ n − 2

r

p

2(8’)

+2 n − 2r − 2

+ n − 2

r −1

p − n − 2

r − 2

<>0, (9’)

− n − 2r − 2

1− p( )2 + 2 n − 2

r −1

1− p( ) p − n − 2

r

p

2 > 0.

n − 2r − 2

, n − 2

r −1

and n − 2

r

are the known expressions

− n − 2r − 2

1− p( )2 + 2 n − 2

r −1

1− p( ) p − n − 2

r

p

2 < 0.

We remark that the lower sign in front of the square root corre-sponds to an interchange of complementary probabilities (wherep is replaced by (1 – p) and (1 – p) by p).

Furthermore, by a closer look one recognises that the mentionedexceptions, that appear as undetermined, in reality correspond to

the probability in case.

If, as a rule, we apply the upper sign, we will be on the safe sideas long as we make sure to have

(11)

The equality sign is only applied for the lowest r that appears inthe calculation. If this condition is fulfilled for r = x + 1, then itis fulfilled for all higher values of r, and thus the required safety

for n subscribers with an average connection probability of pduring the selected time (ie. on average np simultaneous ‘speechconnections’ in the period) the condition for the simplified cal-culation method is:

(12)

At this point we will assume this condition to be fulfilled.Should there be a question of choosing x below the given limit,one would have to check by closer scrutiny whether that wouldlead to any harmful error in the whole calculation. We havebefore the development of this general formula by anexamination of all conceivable combinations for sets of 4, 5, 6,7, 8 subscribers calculated the values that np for above-givenreasons must not exceed for r = 2, 3, ..., n–2 (n = 4, 5, 6, 7, 8)and have found the results shown in the table below left, whereC denotes the critical value of np.

As is easily seen, there is full agreement with the aboveformula, in which the value under the square root sign is nevernegative.

19. It is now possible for certain small subscriber sets to cal-culate the degree of hindrance for given values of p = y/n, wherey as repeatedly mentioned is the average number of simul-taneously existing telephone calls, and for given values of x =r – 1, where x is always a positive number. The results can for agiven number n (n = number of subscribers) be presented in atable where the left column contains the values of x and the toprow

is given, and we want to find the values of y corresponding tocertain values of x, then this can be done by interpolation, using

gives the values of y. If a value of ψ , say ψ =1 / 1000

x +1( ) 1− 1− xx+1( ) ⋅ n

n−1( )

= x +1( ) 1− n−x−1

x+1( ) n−1( )

≥ np

is obtained for all terms in ψ nx . For x circuits available

r 1− 1− r−1( )r ⋅ n

n−1( )

= r 1− n−rr n−1( )

≥ np

Pr = Pr , ie. the equalization neither increases nor decreases


r = 2 r = 3 r = 4 r = 5 r = 6

n = 4 C =

n = 5 C =

n = 6 C =

n = 7 C =

n = 8 C =

2 − 2 13

2 − 32

2 − 2 25

2 − 53

2 − 2 37

3 − 32

3 − 3 15

3 − 2

3 − 157

4 − 2 25

4 − 2

4 − 167

5 − 53

5 − 157 6 − 2 3

7


(14)

where n0 denotes the real number of subscribers in the con-sidered set. By this we have obtained a further and veryessential simplification of the calculated degree of hindrance.

22. However, we must even here try to assess whether thischange of calculation has any harmful effect or not, ie. whether

we will once more study the expression

From this is obtained:

(15)

Furthermore (log P denotes here log nat P):

log Pnr = logy

r

r!+ log 1− 1

n

+ log 1− 2

n

log 1− 3

n

+K + log 1− r −1n

+ n 1− r

n

log 1− y

n

Pnr = nr

p

r1− p( ) n−r( )

= 1r!

n n −1( ) n − 2( )K n − r +1( )n ⋅n ⋅nK n

yr

1− yn

n 1− rn( )

= yr

r!⋅ 1− 1

n

1− 2

n

K 1− r −1

n

1− y

n

n 1− rn( )

ψ ∞x is greater or not greater than ψ n0 x . For this purpose

ψ ∞x = 1x +1( )! y

x+1( )e

−y(1+ 2y

x + 2( ) + 3 ⋅ y2

x + 2( ) x + 3( )

+ 4 ⋅ y3

x + 2( ) x + 3( ) x + 4( ) +K

n0 − x( )x + 2( ) x + 3( )K n0( ) y

n0 −x−1( )) 1

y

V The second simplification.Calculation of the degree of hindrance

calculation of a such table will be an increasingly long-lastingand tedious work with increasing n, and one will not reach veryhigh numbers before the task becomes insuperable. Despite thesimplified formula that only gives approximate results, thepractical demands will necessitate further simplifications of theexpression for the degree of hindrance. Here we will apply theprocedure of Bortkewitsch.

Here we will set np = y, ie. p = y / n. Then we will imagine thaty is given a constant value equal to the given average number ofsimultaneous connections in busy time, however so that n ex-ceeds any limit while p approaches 0. In other words, weimagine that the given set of n subscribers with a certainaverage requirement p of telephone use is replaced by a gradu-ally increasing set of subscribers with a gradually decreasingrequirement each, until we approach an infinite number of sub-scribers with an infinitely small requirement of telephone useeach, however so that the average collective traffic is exactlythe same as that of the real set of subscribers. Then we get:

where e is the base of the natural logarithm. The present ex-pression:

(13)

is Bortkewitsch’ formula.

values of P∞(x+1), P∞(x+2), ..., P∞(n–x) we get for n = ∞:

By introducing in the formula for ψ nx the thus obtained

pn→∞

= 1r!

yre

−y = p∞r

Pnrn → ∞ =

n n −1( )K n − r +1( )1⋅2K r

yn

r

1− yn

n−r( )

= 1r!

n → ∞⋅ nn

⋅n −1( )

n⋅

n − 2( )n

K

n − r +1( )n

⋅ yr ⋅ 1− y

n

n−r( )−y( ) −y( )

= 1r!

⋅ yr ⋅ 1− y

n

− ny( ) −y( )

= 1r!

yre

−y

Pnr = nr

p

r1− p( )n−r

21. Let us consider the first terms of ψ nx , namely

20. It appears obvious from the formula for ψ that the

the values of ψ in the table closest to the given value of

ψ in the table closest to the given value of ψ for each x.

(15’)

where in the expression after ‘symb.’ the notation Bs, Bs–1, Bs–2,..., B1 should be replaced by Bs, Bs–1, Bs–2, ..., B1, which is thenotation for Bernoulli numbers. If n in the expression for

ddn

log Pnr

= 1

n2

1

1− 1n( ) + 2

1− 2n( ) + 3

1− 3n( ) +K +

r −1( )1− r−1

n( )

+ yn

1− rn( )

1− yn

+ log 1− yn

= 1

n2 (1+ 1

n+ 1

n2 + 1

n3 +K

+2 1+ 2n

+ 22

n2 + 2

3

n3 +K

+3 1+ 3n

+ 32

n2 + 3

3

n3 +K

K

+ r −1( ) 1+ r −1n

+r −1( )2

n2 +

r −1( )3

n3 +K

)

+ yn

1+ yn

+ y2

n2 + y

3

n3 +K − r

n1+ y

n+ y

2

n2 + y

3

n3 +K

− yn

+ y2

2n2 + y

3

3n3 +K

= 1

n2 1+ 2 + 3+K + r −1( ) + y

2 − ry − y2

2

+ 1

n3 1+ 2

2 + 32 +K + r −1( )2 + y

3 − ry2 − y

3

3

+ 1

n4 1+ 2

3 + 33 +K + r −1( )3 + y

4 − ry3 − y

4

4

+K

+ 1

ns+1 1+ 2

s + 3s +K + r −1( )s + y

s+1 − rys − y

s+1

s +1

+K

= 1

n2

r −1( )r1⋅2 + 1

2y

2 − ry

+ 1

n3

r −1( )r 2r −1( )1⋅2 ⋅3 + 2

3y

3 − ry2

+ 1

n4

r −1( )r1⋅2

2

+ 34

y4 − ry

3

+K

+ 1

ns+1 symb.

r −1( ) + B( )s+1 − Bs+1

s +1+ s

s +1y

s+1 − rys

+K ,

expression will be essentially determined by the first term

This term is positive when

y2 – 2ry + (r – 1)r > 0.

By resolving the equation

y2 – 2ry = –r2 + r

is obtained

According to the nature of the problem we have y < r, andhence we must choose the solution

(16)

Thus for these values of y and r the term mentioned is equal to 0.

when r > (4/3)2. In a similar way one can find correspondingconditions for the following terms to be positive. There is, how-ever, no need for a closer study of this, since the character of thetask implies that n >> r, and besides, the following terms can begiven the form (r/n)4 · Θ4, (r/n)5 · Θ5, ..., (r/n)s+1 · Θs+1, ...,where Θ4, Θ5, ..., Θs+1, ... are proper fractions, so that the sumof these terms can be neglected in comparison with the first twoterms. Thus we content ourselves by establishing as a sufficientcondition for

(17)

increasing n, and under the given conditions we have

the stated conditions are fulfilled we can be sure that the simpli-fied calculation does not lead to any harmful error.

From this follows immediately ψ ∞x > ψ n0 x , and when

P∞r > Pnr .

However, if ddn

log Pnr > 0, then Pnr increases with

y < r − r , or

r > y + 12 + y + 1

4 , and r << n

ddn

log Pnr > 0, that

If in the second term y < r − r , then this term is positive

If y < r − r or r > y + 12

+ y + 14

, then the term is positive;

if y > r − r or r < y + 12

+ y + 14

, then the term is negative.

y = r − r , or

r = y + 12

+ y + 14

y = r ± r .

1

n2

r −1( )r1⋅2 + 1

2y

2 − ry

.

ddn

log Pnr is made sufficiently large, then the value of the


23. It now remains to add the series that is contained in

n0 – x normally is a number of more than one digit, the seriesmay contain a substantial number of terms. Also, in some likelypractical cases the initial part of the series may consist of in-creasing terms from left to right. Thus, in general the calculationmust include a considerable number of terms in order to obtainsufficient exactness. We must, therefore, find a way of appro-ximate calculation of this series in order to determine the degreeof hindrance for commonly occurring (greater) values of n0, yand x.

To this end we will study the infinite series:

(18)

1+ 2yx + 2( ) + 3y

2

x + 2( ) x + 3( ) + 4y3

x + 2( ) x + 3( ) x + 4( )

+n0 − x( )

x + 2( ) x + 3( )K n0( ) ⋅ yn0 −x−1

+K

n0

x + 2( ) x + 3( )K x + n0( ) yn0 −1

+K ∞

=ψ ∞x x +1( )!ey

yx

=1+ 2yx + k( ) + 3y

2

x + k( )2 + 4y3

x + k( )3 +K

+n0 y

n0 −1

x + k( )n0 −1 +n0 +1( )y

n0

x + k( )n0+K ∞

=1+ yx + k( ) + y

2

x + k( )2 + y3

x + k( )3 + y4

x + k( )4 +K ∞

+ yx + k( ) + y

2

x + k( )2 + y3

x + k( )3 + y4

x + k( )4 +K ∞

+ y2

x + k( )2 + y3

x + k( )3 + y4

x + k( )4 +K ∞

+ y3

x + k( )3 + y4

x + k( )4 +K ∞

+ y4

x + k( )4 +K ∞

+K ∞

= 1

1− yx+k( )

+ yx + k( ) ⋅ 1

1− yx+k( )

+ y2

x + k( )2 ⋅ 1

1− yx+k( )

+ y3

x + k( )3 ⋅ 1

1− yx+k( )

+K ∞

= 1

1− yx+k( )

2 , where

ψ ∞x . As can be seen it concerns a finite series. Since (19)

Thus we have:

(20)

By choosing k such that equation (18) is satisfied we get:

where n1 = n0 – x – 1.

Thus we have:

(21)

We can always imagine a number h determined so that:

(22)

It is evident that the number h must be greater than k, andconsequently from the previous h > 2. The number 2 may thusbe considered as a ‘lower limit’ for h. It is of course alsopossible to find an ‘upper limit’ for h by means of the pre-viously given limitation of k. We will, however, be content with alimitation that can be immediately put forward by studying theexpression for

values of n1 and x, (y < x << n1), h will be less than 2k, so

h <n0 − x

2+ 3

ψ ∞x as a function of k, as at least for considerable

ψ ∞x = 1x +1( )! ⋅ y

x

ey ⋅ 1

1− yx+h( )2

ψ ∞x = 1x +1( )! ⋅ y

x

ey ⋅ 1

1− yx+k( )2

1− yn1 −1

x + k( )n1 −1 ⋅n1

1− yx+k( )2 1− y

x + k1− 1

n1

ψ ∞x = 1x +1( ) ⋅ y

x

ey 1+ 2y

x + 2( ) + 3y2

x + 2( ) x + 3( ) +K +n0 − x( )y

n0 −x−1

x + 2( )K n0( )

= ψ ∞x 1−n1y

n1 −1

x + k( )n1 −1 +n1 +1( )

x + k( )n1y

n1 +n1 + 2( )

x + k( )n1 +1 yn1 +1

+K ∞

= ψ ∞x 1− yn1 −1

x + k( )n1 −1 ⋅n1

1− yx+k( )2 1− y

x + k1− 1

n1

.

ψ ∞x = 1x +1( )! ⋅ y

x

ey ⋅ 1

1− yx + k( )

2

asn0 − x

4+ 3

2is the arithmetic mean of the n0 − x −1 values

21

, 2 + 3

2

, 2 + 3 + 4

3

,K ,

2 + 3 + 4+K + n0 − x( )n0 − x −1

.

2 < k <n0 − x

4+ 3

2,


logarithms, as we have:

(25’)

(27’)

If one wants to know the number of correct decimals in the cal-culated degree of hindrance, one can apply the given expressionwhere Θ is set equal to 1. This control may with sufficient use-fulness for practical applications be limited to values for thedegree of hindrance in the vicinity of the prescribed magnitude.

By means of formula (14) in the present exposition and Bort-kewitsch’ tables2) we have calculated a table (Annex I) over thedegree of hindrance for all values of x from 1 to 19 with appli-cation of the values of y = 1/2, 1, 1.5, ..., 10 that imply a degreeof hindrance close to that prescribed. Since Bortkewitsch onlyallows for 4 decimals, and because of that only a few terms ofthe series in question can be expressed by means of his tables,and since, as explained, the series has only a weak convergence,our table can not present the degree of hindrance with greatexactness. In order to further elucidate the matter we have,

formula (25) for x = 4, 7, 10, 13, 16, 19, 22 and some of thementioned values of y.

As can be seen from the thus produced table (Annex II) thedeviations between corresponding values in the two tables arenot significant. From this we can conclude that we do not causeany great error when producing a 1 ‰ curve by applyingformula (25), that is deduced from the previously given moreexact expressions by setting h = k = 2.

VI Other procedures for elucidation ofthe likely efficiency. Comparison ofthe different expressions for theallowable hindrance

26. We have in the foregoing presented a method, by whichone can express the ability or lack of ability of an automatictelephone exchange to carry a certain amount of traffic. Thereexist also other methods that can be applied for this purpose.

Instead of considering the part of traffic that can be expected tobe carried or hindered when offered to a certain number ofparallel central exchange circuits or switches, one can forinstance attempt to find the probability of all those circuitsbeing occupied, ie. the probability of full utilisation3), or onecan, if preferred, seek the probability that under full utilisationthere are requirements that can not be satisfied, ie. the proba-

therefore, calculated ψ xy with 7 decimals by means of

logψ xy = − 12

log π − 12

log2 − 12

log x +1( )− x +1( ) log x +1( ) + x log y + x +1− y( ) loge

−2 log x + 2 − y( ) + 2 log x + 2( )

logψ xy = − log2 − log3−K − log x +1( ) + x log y

−y loge − 2 log x + 2 − y( ) + 2 log x + 2( ),

bility of overfill. For both of those probabilities apply, althoughnot to the same extent, that the smaller they are, the greater isthe carrying capacity of the system.

The first mentioned full utilisation probability with x circuits fortraffic y in busy hour will be denoted P(xy) , and the last men-tioned overflow probability with x circuits for traffic y in busyhour Pxy.

(28). Let us regard the first of these probabilities. As we keep towhat is stated in the sections of the first and the second simplifi-cations we will have:

(28)

This series is strongly converging, and it will therefore be ofpractically no significance whether the addition within the par-enthesis is done up to and inclusive the term

number, or one regards the series as infinite and seeks thecorresponding sum. One can even in many cases obtain suffici-ent accuracy by only adding a limited number of terms.

As stated, P(xy) gives us the probability that all x circuits orswitches are occupied. Such a state can be caused by either x or(x + 1) or (x + 2) or ... or n0 simultaneous conversations or calls.P(xy) does not directly give a measure of the ratio of the tele-phone activity that is hindered. This appears from the fact that

gives the probability of x calls, of which none are hindered,since there are precisely the sufficient number of x circuits,while the terms following the second do not indicate the numberof calls that are hindered under the probability states given bythose terms. The first term indicates the probability, ie. the rela-tive frequency, of a state where the utilisation limit is reached,which may not be exceeded without traffic hindrances, and thefollowing terms indicate the probabilities of the states that areassociated with such hindrances. In combination these proba-bilities represent a measure of a certain admissible shortcomingof the system, and the expression for P(xy) may thus very wellbe applied for evaluation of the usefulness of the system, whenone fully realises what it expresses. P(xy) = 0.001 thus does notindicate an average likely loss of 1 call out of 1000, but thelikelihood that all x circuits or switches are fully occupied anaverage of 1 second out of 1000. In a similar way as that pre-viously dealt with one can construct 1 ‰, 2 ‰, 3 ‰, ..., r ‰curves for the probability or risk of full load P(xy) for deter-mination of the number of switches x for a certain given trafficy. This is already known to have been carried out by MasterErlang and engineer Christensen in Copenhagen4).

As an annex to the present account a table is given, by which a1 ‰ curve is drawn for ψxy and a 1 ‰ and a 4 ‰ curve are

the first and largest term of the expression, 1x!

⋅ yx

ey ,

yn0 −x( )

(x +1)(x + 2)K n0, where n0 denotes the real subscriber

P(xy) =

1x!

yx

ey 1+ y

(x +1)+ y

2

(x +1)(x + 2)+ y

3

(x +1)K (x + 3)+K


2) See ‘Das Gesetz der kleinen Zahlen’.3) In XII Summary called ‘blocking’. 4) See ‘Elektrotechnische Zeitschrift’, H. 46, 1913.

drawn for P(xy). It can be seen that the curves for ψxy and P(xy)have the same character, however to a given x and a given ‰value corresponds a significantly larger y for ψxy than for P(xy).

It appears from the foregoing that there is no mutual disagree-ment between the expressions for ψxy and P(xy). They have boththeir justification, their theoretical and practical significance.The difference rests in the fact that they are measures of differ-ent things. It is a matter of opinion what measure to choose andwhat degree of hindrance or full load risk one will allow for theestablishment of a telephone system.

28. We will now consider the overload probability:

(29)

exactly x simultaneous telephone calls (conversations). We nowassume that in the course of these x conversations a new callarrives. Then one of two cases will occur: Either will this call,when it gradually reaches the x circuits end tests if they arebusy, happen to replace a conversation just being concluded, inwhich case no hindrance will occur. Or there will not occur sucha conclusion, as all x conversations continue and thus preventsthe new call from obtaining a connection. In the latter case wesay that the call is lost. In order to succeed a new call must beinitiated, provided there is no remedy, by which new tests canbe made without a new call. More about this below. However,in the case of x conversations in progress and one callunsuccessfully searching for a free circuit or switch, then thematter is no longer the state, whose probability of existence is

but a state whose probability of existence is

This is the state that implies the first case of hindrance, Thesecond and somewhat less frequent case is implied by theprobability

The third case:

The sum of these probabilities gives the alternative or totalprobability of overflow or for the occurrence of hindrances:

1(x + 3)!

⋅ yx+3

ey Etc.

1(x + 2)!

⋅ yx+2

ey

1(x +1)!

⋅ yx+1

ey

1x!

⋅ yx

ey

As already mentioned, 1x!

⋅ yx

ey denotes the probability of

Pxy = 1(x +1)!

yx+1

ey 1+ y

(x + 2)+ y

2

(x + 2)(x + 3)+ y

3

(x + 2)K (x + 4)+K

= P(xy) − 1x!

yx

ey

(29)

This quantity thus gives a measure of the likely frequency ofreal hindrances. As a concept it is therefore to be preferred forP(xy). The two expressions are also different because they indi-cate different things. The difference between them is the firstterm of P(xy).

29. It is interesting to observe that

The state probability of x simultaneous telephone conversationsfor a traffic intensity y:

is thus generative with respect to the whole overflow proba-bility. Assume a rectangular co-ordinate system with axis x andordinate pxy and construct the curve corresponding to x! · pxy =fx(y) = e-y yx, (x denotes a constant, positive integer). Thenaccording to the given integral Pxy = 1/x! times the area that islimited by the curve, the abscissa and the ordinate pxy for thegiven value of y. This is assumed not to exceed the earlier givenlimits. Having constructed a set of such curves on graph paper,corresponding to relevant x values, one can easily, by countingthe squares related to the given values of y obtain a fairly exactindication of the concerned values of Pxy.

We have derived Pxy independently of this integration formula.It is, however, a question, that need not to be answered here,whether the same clarity of the relevant phenomena can beobtained by this integration, with traffic intensity varying be-tween 0 and y, as with the previously presented operationsbased on general considerations of various state probabilitiesunder a given average traffic intensity y.

As Pxy can be formed by integration from 0 to y of Bort-kewitsch’ formula, so also integration of P(xy) dy between the

the total number of telephone conversations that on average arehindered or lost in any moment of the busy hour. This can easilybe verified by differentiation:

same limits leads to y ⋅ ψ ∞xy , where ψ ∞xy has the

previously indicated meaning, and y ⋅ ψ ∞xy thus indicates

pxy = 1x!

e−y

yx

, pxy = p∞x( )

Pxy = 1(x +1)!

yx+1

ey 1+ y

(x + 2)+ y

2

(x + 2)(x + 3+K

=1− e−y

1+ y1

+ y2

1⋅2 + y3

1⋅2 ⋅3 +K + yx

1⋅2 ⋅3K x

= 1x!

e−y

yxdy

0

y

∫5)

Pxy = 1(x +1)!

yx+1

ey 1+ y

(x + 2)+ y

2

(x + 2)(x + 3)+K


5) Remark: The author of the present report has been madeaware of this relation by a representative of Western ElectricCo., after having derived Pxy by general probability con-siderations.

Thus

integration of P(xy) dy = Px–1,y dy between 0 and y, which in amathematical sense can be said to illustrate the significance of P(xy).

According to the preceding we have:

Thus we have the following remarkable relation:

This integral can be given a certain geometric interpretation

times the volume of a spatial region, the form and dimensions of

region is for instance limited by the three planes of a three-dimensional rectangular co-ordinate system and a curved sur-face). The state probability given by this expression, namely theprobability that (x – 1) circuits are occupied by the traffic y, thushas a remarkable mathematical significance in relation to thephenomena dealt with here. We will, however, not go any furtherinto this matter. By means of tools for mechanical integration thisintegral formula might be of practical use for derivation of tablesand curves that we need. Thus one could also avoid the errors inthe calculation due to the fact that one cannot in advance set anexact value for the quantities h and k discussed earlier. Thesehave been set equal to 2, which results in a calculated degree ofhindrance somewhat larger than the probable value.

which are determined by the expressiony

x−1

(x −1)!e

−y. (The

such that the degree of hindrance ψ ∞xy is depicted as 1y

1y

dy0

y

∫y

(x−1)

(x −1)!0

y

∫ e−y

dy = ψ ∞xy ;

d2

dt2 y ⋅ ψ ∞xy( ) = y

x−1

(x −1)!e

−y.

y(x−1)

(x −1)!0

y

∫ e−y

dy = Px−1,y

The quantity y ⋅ ψ ∞xy is thus generated by definite

P(xy)0

y

∫ dy = y ⋅ ψ ∞xy

ddy

y ⋅ ψ ∞xy( )= 1

(x +1)!ddy

yx+1 + 2

x + 2( ) yx+2 + 3

x + 2( ) x + 3( ) yx+3 +K

e−y

= 1x +1( )! e

−y( x +1( )y

x +2 x + 2( )

x + 2( ) yx+1 +

3 x + 3( )x + 2( ) x + 3( ) y

x+2

+4 x + 4( )

x + 2( )K x + 4( ) yx+3 +K −y

x+1 − 2x + 2( ) y

x+2

− 3x + 2( ) x + 3( ) y

x+3 −K )

= 1x!

e−y

yx + y

x+1

x +1( ) + yx+2

x +1( ) x + 2( ) + yx+3

x +1( )K x + 3( ) +K

= P(xy)

VII Objections to the applicability of thecalculations.Calculations on a different basis

30. Until now we have during our discussion of the topic athand mainly had in view the demands that the total traffic froma set of subscribers can be expected to make at any moment ofthe selected time or the busy hour. We have with the degree ofexactness that we have been able to obtain tried to find expres-sions for determination of the part of the traffic that due to thelack of circuits can not be expected to be carried. We have notset forth any conditions with respect to the technical arrange-ment that should make access possible, or which in case ofinsufficient equipment for the total traffic should turn away thehindered traffic. Furthermore we have not, as already stated,made any hypothesis about the duration of each telephoneconnection, nor have we distinguished between the single timeintervals within each connection. There might be a questionwhether we during our efforts of treating the topic in a mostgeneral way have defined the task so that the derived solutionalso corresponds to what the technical equipment can attain inreality. It cannot be denied that some objections may be made.We shall explain this in more detail.

We have calculated the probability, approximately, that allavailable circuits are occupied by traffic from the set of sub-scribers. It is clear that as long as this state prevails all excesstraffic is hindered. If, however, the system is arranged in a waythat a subscriber who is not given access at once is allowed towait for access without making a new call, then the hindrancewill cease as soon as a circuit becomes free, and the hindrancewill not lead to a ‘lost’ call. Considering the calculated degreeof hindrance as a measure of what is lost and thus must becompensated for by new calls, then our calculation will not, ifinterpreted as an anticipated loss, correspond to the performanceof a system that permits such waiting. A system of this type caneven be arranged so that within certain traffic limits practicallyno calls are lost. This does not mean that there are no restric-tions in such a system. On the contrary, the restrictions may incertain cases be serious enough for subscribers that have to waitfor access. In that sense the calculated degree of hindrance maybe applied to illustrate the efficiency of these systems.

We have on the other hand also calculated the probability thatnot all common circuits are occupied by traffic, and for the timethat such a state prevails there is not assumed any hindrance.However, does this in general correspond to what happensunder the actual traffic in busy hour? – We shall consider asimple case.

A subscriber sends a call request at a moment when all commoncircuits are occupied. If the system is not arranged to permitwaiting, then the call and the connection that in an ideal manualsystem would follow immediately would get lost. Let us nowimagine the case when a circuit becomes available immediatelyafter the rejection of the call. Then there is in our calculation nolonger any hindrance for the following portion of the call, whichnevertheless in the actual system gets lost as a whole. Thus itseems like our calculation involves waiting, and even for aduration similar to that of an ordinary communication.

It should, however, be noted that with regard to this disagree-ment with the assumptions for the present system a certain com-


The number n, being a definite integer, is chosen so that certainconditions for the intervals formed are fulfilled. There isnothing to prevent the choice of a higher number than the first

unaffected by this. On the contrary, under certain conditions

increase with increasing n and approach a definite limit for n = ∞.

We will keep m unchanged and let n approach ∞. This leads to:

where e is the base of the natural logarithm.

This is again Bortkewitsch’ formula, that we now have en-countered a third time after having dealt with the subject fromwidely different aspects.

average traffic intensity, ie. we have m = y. Consequently, weobtain:

This value of the state probability in question is derived withoutany consideration of the number of subscribers or the individualdemands of each subscriber. We have only presumed theknowledge of or the consideration of the average number ofcalls in the selected time or the busy hour as well as the averageduration of connections. Those two averages determine theaverage traffic intensity.

The probability of receiving more than x calls in the period

about the number of subscribers, it may be imagined to be avery great number):

In order to determine the degree of hindrance only on the basisof this development all connections must be considered to be ofequal duration.

P̃xy = yx+1

(x +1)!e

−y1+ y

(x + 2)+ y

2

(x + 2)(x + 3)+K

t is also on this basis given by (nothing is presumed

36. We have thus again arrived at the expression 1x!

yxe

e−y

for the probability of finding x calls during the period t. 7)

P̃x = 1x!

yxe

e−y

n → ∞

As indicated above m = N ⋅ t / Tc and thus equal to the

P̃x = 1x!

nn

⋅ n −1n

⋅ n − 2n

K n − x +1n

⋅mx1− m

n

− nm + x

m( ) −m( )

= 1x!

mxe

−m, lim n = ∞( ),

(see the section ‘The second simplification’) P̃x will

adopted. However, in general P̃x does not remain

P̃x = n(n −1)(n − 2)K (n − x +1)1⋅2 ⋅3K x

mn

x1− m

n

n−x Hence the degree of hindrance is obtained:

by dividing the hindered calls by the total number of calls.

We have put the mark ~ over the letters denoting the givenprobabilities and the degree of hindrance in order to emphasisethe difference between the present and the previous develop-ments. As one can observe, the final results are identical for thethree procedures in question.

VIII Calculation of the degree ofhindrance for the daily traffic

37. So far we have been engaged in the obstacles for the trafficprocess that occur within the selected time. We will now try toapply the obtained results for a calculation of the traffic volumethat on average can be expected to be hindered over twenty-fourhours, and the corresponding degree of hindrance. We willdenote this the general or daily degree of hindrance, contrary tothat previously discussed, which suitably may be denoted theparticular or momentary (short term), or, as far as it is referredto the busiest time, the maximum degree of hindrance.

As can be remembered, when traffic intensity is y and thenumber of circuits or switches is x, we have found a usableexpression of the traffic volume that can be expected to behindered in the time unit (one second) within the selected time:

ψ̃ xy = yx+1

(x +1)!e

−y1+ 2y

(x + 2)+ 3y

2

(x + 2)(x + 3)+K

1y

,


7) Remark: Dr. Spiecker assumes a constant number n of callsper hour and thus arrives at a result for the state probability(formula (9)):

which with the notation used by us can be expressed asfollows (by Spiecker n/m corresponds to our y and a to x):

The value of this expression increases with increasing n and

Thus for any actually appearing traffic intensity his calcu-

derived by assuming a number of calls only given as anaverage value in the busy hour.

lated probability Wa = Px is smaller than P̃x , that is

is only transformed to 1x!

yxe 1

ey = P̃x for n = ∞.

Wa = na

1m

a1− 1

m

n−a=

Px = nx

yn

x

1− yn

n−x

=

1x

n(n −1)K (n − x +1)

(n − y)x y

x 1

1− yn( ) − n

y

y

Wa = na

1m

a1− 1

m

n−a,

Here, according to our first representation, Hc denotes thenumber of hindered traffic seconds (‘conversation seconds’,‘phone seconds’) that can be expected in one second. Thenumber can equally well denote the number of hindered traffichours (‘conversation hours’, ‘phone hours’) that can beexpected per hour. According to the representation in sectionVII, Hc can even

As the period Tc is arbitrarily chosen, the expression above maybe considered valid for any period of the day as long as thatperiod is short enough to grant only very small changes in thetraffic relations.

As the hindered traffic in one second thus is Hc connectionseconds, then in an infinitesimal time interval dt the hinderedconnection seconds are Hcdt, and in the period from t = 0 tot = h within one day:

(31)

This gives an average per second:

(32)

Dividing this by the average second-traffic in the period h,which we will denote by yh connection seconds, then the gen-eral degree of hindrance for the period h will be:

(33)

If y could be given as a known time function of such form thatthe integration could be carried out, then the last equation would

In case y could be considered a linear function of t, this integra-tion would be a simple matter. However, as well known thetraffic rises and falls in a way that only approximately can beexpressed by a function of a more complicated nature. Bymeans of a large number of observations in a telephoneexchange one might possibly propose a mathematical express-ion for y’s dependence on t on average over the day, or onemight, without any consideration of such a possible appro-ximate mathematical relation construct a twenty-four hour tra-ffic curve. The integration could then be carried out either byseries expansion or by subdivision of the curve into a certainnumber of approximately linear parts. Both procedures will,apart from the carrying out of the observations, demand quite a

determine the general degree of hindrance ψ xyh.

ψ xyh= H h

yh= 1

hyhHcdt

0

h

∫ = 1hyh

e−y

0

h

∫y

x+1

(x +1)!+ 2

yx+2

(x + 2)!+K

dt

H h = 1h

Hcdt0

h

∫ connection seconds

Hh = Hcdt0

h

∫ connection seconds

denote the number of hindered calls in the period t.

Hc = yψ xy = e−y y

x+1

(x +1)!+ 2

yx+2

(x + 2)!+ 3

yx+3

(x + 3)+K

= e−y y

x+1

(x +1)1

1− yx+2( )2 (approximately)

substantial work. And the result of such work would hardly suf-fice for the estimation of the circumstances at other telephoneplants.

38. Since we in the present report have set ourselves the task oftreating the subject at issue in a most general way, we will evenhere try to find a solution that presumably will satisfy thedemand of a wide applicability.

The problem may be expressed as follows, as we still considerthe traffic of one day (twenty-four hours), regarded as theaverage of the traffic of many days:

When the traffic intensity y at time t = 0 is 0 and at time t = h is

arbitrary time in between?

We assume for better overview that t is measured in hours.

It is sensible to search for the answer by means of Gauss’ so-

when t = 0. However, by closer study it turns out that if p ischosen to give a value close to zero during the night, then theconcentration to the busiest hour of the day (ie. the ratio be-tween the traffic of this hour and the total twenty-four hourtraffic) will be too large.

Furthermore, one can proceed tentatively with Bortkewitsch’

for t = h, as indicated in our question. The curve that accordingto this expression describes y as a function of t in a rectangularco-ordinate system has its peak at t = h, its zero points at t = 0

it ought to have characteristics to make it useful for the plannedcalculations. However, when carried out it appears that it hassimilar drawbacks as Gauss’ error curve. If h is determined soas to obtain a concentration according to normal traffic condi-tions, then the curve shows too high traffic at the end of day.Setting h so that traffic at the end of day is practically zero leadsto an unreasonably high concentration.

These discrepancies are of course connected with the fact thatthe functions in question are derived or can be derived frompure probability considerations pertaining to a collection ofitems or events around a maximum point, without introducingin the formulas any coefficient taking into account the distri-bution of phenomena effected by certain unknown causes. Wewill in the modified Bortkewitsch’ formula introduce such a co-efficient, writing:

(34a)y = ye

−cttch e

h

ch

and t = +∞, and its two inflexion points at t = h ± h . Thus

function by setting y = yqe−t

th 1

h!. Utilizing Stirling’s formula

one can write y = ye−t

th e

h

h. Here y = 0 for t = 0 and y = y

called error curve y = ke− p

2x

2

, and thus set y = ye− p

2t

2

.

Naturally, we must set the time so that we have y = y

y = ymax , how large is the traffic intensity y = f (t) at an


To simplify we set:

c = k/h, leading to

(34b)

We will denote k the concentration coefficient.

39. It can be seen from the given expression for y that y = 0 for

It is just by this displacement of the inflexion points effectedby the variation of k that concentration can be increased or

so that the curve can be adapted to the existing or assumedconditions, at least for the time interval t = 0 to t = h. Thus onecan set k = 1, 2, 3, 4, etc. and get the inflexion points for t = h ±h,

We will now imagine the working day subdivided in two parts,the first part from t = 0 to t = h with increasing traffic, and thesecond part from t = h to t = 3h with decreasing traffic.

Let us set k = 3. The traffic volume for the day according to the

(35)

Assuming h = 5, corresponding to a workday of 15 hours, we

a concentration of 13.7 % which is a sensible ratio. If weassume h = 42/3, corresponding to a workday of 14 hours, theconcentration will be 14.7 %, which is higher than the usuallevel. If we assume a workday of 15 hours, then at the end ofthis time we will have a traffic intensity:

The curve thus shows that it has decreased to approx. 6.7 % ofthe maximum intensity. This last number must, however, beconsidered too high under normal circumstances. The curvedoes not fall fast enough for the evening traffic.

y = y e5

3e

−93

35

3 = ye−6 ⋅33 = y ⋅0.00247875K ⋅27

= y ⋅0.0669K

obtain M = y ⋅7.282K , so y = M7.28K = M ⋅0.137K , giving

M = y0

3h

∫ eh

3e

− 3h t

t3dt = yh 2e

3

27− e

−69 + 3 + 2

3+ 2

33

= yh ⋅1.45623K

curve y = y eh

3e

− 3h t

t3

will be:

t = h 1± 12

, t = h 1± 1

3

, t = h ± h

2, etc.

decreased (without changing the other parameters y and h)

for t = 0, y = y for t = h, y = ymax for t = h, while by

settingd

2y

dt2 = 0 the inflexion points of the curve are found

at t = h 1± 1k

.

y = y eh

ke

− kh t

tk

If, on the contrary, a concentration coefficient of 4, then we willget a lower final traffic, namely:

On the other hand will the concentration be far too high. Thusthe curve will not give a fully satisfactory illustration of the fad-ing of the traffic in the evening and at night.

Because of this we will only utilise the first part of the curvefrom t = 0 to t = h. For the afternoon traffic we will reverse the

time t = 0 indicate the end of the day traffic and consideringtime as positive backward towards t = h. Furthermore we willconsider h twice as big with the latter curve as the former.

morning and afternoon traffic volumes respectively. We willassume k to be equal to 4.

We then get:

Thus we have here a concentration of approx. 14.2 %. Outsidethe 15 working hours, which may be assumed to be from 7.30a.m. until 10.30 p.m., the traffic is considered to be equal to 0.

From the expression for M we see that it may be set equal to

Thus we can apply one and the same curve

for an approximate calculation of the average day traffic duringcertain fractions of the working time as well as the whole time,when we know the average maximum intensity (the number oftelephone hours per busy hour), the duration ph of the ordinarydaily working hours and the average concentration coefficient.

y = y eph

k

e− k

ph ttk

M3h = y e3h

4e

− 43h t

t4dt = y ⋅3h ⋅0.475

0

3h

∫

Mh2= y

0

h2

∫ eh2

4

e− 4

h2tt4dt = yh2 ⋅0.475

M = y h1 + h2( ) ⋅0.475K = y ⋅15 ⋅0.475 = y ⋅7.125 = y 10014.2

Mh1= y

0

h1

∫ eh1

4

e− 4

h1tt

4dt

= y ⋅h16e

4

256− 1

4+ 1

4+ 3

16+ 6

64+ 6

256

= y ⋅h1 ⋅0.475

8)

We now set M = Mh1+ Mh2

, where Mh1and Mh2

denote

curve (by turning it around the ordinate y), letting the

y = y e5

4e

−123

45

4 = ye−8 ⋅34 = y ⋅0.00033546K ⋅81

= approx. 2.7% of y.


8) The number is rounded upwards.

40. Hereby means are made available for calculating theaggregate traffic hindrance or the daily degree of hindrance. Itwould, however, lead too far to carry out this calculation withminute exactness, since the integral to be solved

(36)

where the relation between y and t is given by

series expansion. We will therefore make a detour and let thisbe as short and straightforward as possible, while we make sureto be on the safe side.

In the rectangular co-ordinate system where we assume y in apractical way given as a function of t we will through the zeropoint draw a straight line in the plane of the traffic curve. Letthe equation of this line be:

(37)

where a = ph and r denotes a constant of such property that thestraight line does not intersect the curve outside the zero point,but either is a tangent to the curve between its first inflexionpoint and the end point given by t = a, or that it remains outsideor on the upper side of the curve.

In a well known way we find that r must be such that:

(38)

If we stick to a = 15 and k = 4, then we have to choose r so that

Hence we set r = 2.

between the intersection point and the end point mentioned isequal to 2. Instead of the found curve we will now employ the

simpler calculation when carrying out the necessary integration.Since the broken line entirely lies outside the curve between itsend points, it is obvious that the traffic volume and the corre-sponding amount of hindrances that we obtain are not too small.On the other hand, it is evident that since the greatest deviationbetween the line and the curve occurs around the lower part ofthe curve, while around the upper part the deviation is relativelyquite small, the result will not be essentially too large.

41. Let us now determine the hindered communication H(a–r)

thus constructed line (0,0)→ (a − 2, y)→ (a, y) to obtain a

Through the endpoint (a, y) of the curve we draw a parallel

to the t − axis until it intersects the line y = ya − 2

t. The distance

115 − r

≥ 2.7182815

⋅ 2764

, or

15 − r ≤ 3202.71828 ⋅9 , r ≥1.919K

1a − r

≥ ea

k −1k

k−1

y = ya − r

t

y = y ⋅ eph

k

e− k

ph ttk

, only can be found by laborious

0ph

∫ e−y y

x+1

(x +1)!+ 2

yx+2

(x + 2)!+ 3

yx+3

(x + 3)!+K

dt

Hence we get:

Thus we obtain:

The traffic volume that is hindered during working hours a – r,ie. outside the r busiest hours, thus on average per hour amountsto:

(39)

As the traffic volume that the calculation is based on amounts

H(a−r)

(a − r)< y

x+1

(x + 2)!⋅e−y ⋅ 1

1− yx+3

3 telephone hours.

H(a−r) < a − ry

[e−y

(y

x+2

(x + 2)!1

1− yx+3( ) + 2

yx+3

(x + 3)!1

1− yx+4( )

+3y

x+4

(x + 4)!1

1− yx+5( ) + 4

yx+5

(x + 5)!1

1− yx+6( )

+5y

x+6

(x + 6)!1

1− yx+7( ) +K )]y

0

< [(a − r)e

−y

y 1− y(x+3)( )

yx+2

(x + 2)!+ 2

yx+3

(x + 3)!+ 3

yx+4

(x + 4)!+ 4

yx+5

(x + 5)!+K

]y

0

< [(a − r)e

−y

y 1− yx+3( )3 ⋅ y

x+2

(x + 2)!]y0

= (a − r)

1− yx+3

3 ⋅ yx+1

(x + 2)!e

−y

H(a−r) = e

−y

0

y

∫y

x+1

(x +1)!+ 2

yx+2

(x + 2)!+ 3

yx+3

(x + 3)!+K

(a − r)y

dy =

(a − r)y

[e−y

(y

x+2

(x + 2)!+ y

x+3

(x + 3)!+ y

x+4

(x + 4)!+ y

x+5

(x + 5)!+ y

x+6

(x + 6)!+K

+ 2y

x+3

(x + 3)!+ 2

yx+4

(x + 4)!+ 2

yx+5

(x + 5)!+ 2

yx+6

(x + 6)!+K

+ 3y

x+4

(x + 4)!+ 3

yx+5

(x + 5)!+ 3

yx+6

(x + 6)!+K

+ 4y

x+5

(x + 5)!+ 4

yx+6

(x + 6)!+K

+ 5y

x+6

(x + 6)!+K

+K )]y0

At time t = 0, y = 0 and at t = a − r, y = y.

H(a−r) =0

a−r

∫ e−y y

x+1

(x +1)!+ 2

yx+2

(x + 2)!+ 3

yx+3

(x + 3)!+K

dt, and

dt = a − ry

dy.

during the time a − r, when y = ya − r

t. We have:


Hence we get:

By means of one of these equations and the equations

The curve, the t-axis and the ordinate y0 enclose an area:

The area that is enclosed by the same axis and ordinate plus theslanting side of the trapeze is:

The difference between these two areas, with the here given

difference between the segment of the curve area above theinflexion point that is cut off by the slanting side of the trapeze,and the triangular area that is enclosed by the previously ment-ioned parallel with the t-axis, the same trapeze side and theupper part of the curve. The segment is thus approx.

influence of this area on the degree of hindrance greater thanthat of the segment, namely because of its position and becauseof its size that is substantial compared to the mentioned differ-ence. The difference between the two lower areas also con-tribute, if only to a minor degree, to increasing the efficiency,and here the effect of the difference is in the same direction asthe position.

What is said here can of course be proved mathematically bymeans of the equations and a more detailed discussion for thegiven curve, essentially determined by the concentration co-efficient k = 4. However, we consider it unnecessary to gofurther into the matter, as anybody by construction and closerexamination of the curve and the trapeze, and by considerationof the expression for the degree of hindrance, may convincehimself of the correctness.

Consequently, we are on the safe side when we for calculationof the degree of hindrance apply the straight line through theinflexion point given by:

4y100

greater than the latter area. Nevertheless is the

length of a is approx. y ⋅0.04. Exactly that size is the

12

12

a − r0

y0 = y ⋅1.05K

y0

a2

∫ ey

4

e− 4t

a t4dt = y ⋅1.01K

r0 =2mu − a

2 (3u − v)v − u

= 2.94K

r1 =2mv − a

2v − u

= 2.19K

a =15, a − r0 + r1( ) = 2 My

= 2m =14.248K , we get:

r0 =r1u + a

2 (v − u)v

; r1 =r0v + a

2 (u − v)u

.

a2 − r0a2 − r1

= uv

(42)

By means of the equation of the former line we get:

is hindered under the conveyance of the traffic that is re-modelled to increase linearly from y = 0 at t = r0 to

In a similar way as before we obtain:

we get the degree of hindrance

Furthermore:

Hence we get quite the same values as before, which must beexpected, as we in both cases have dealt with a linearly in-

It is not, however, in order to get this self-evident confirmationof the calculation of the partial degree of hindrance that we havecarried out this new calculation, but rather to obtain a satis-factory determination of the hindered traffic volume, so as toindicate with certainty an upper limit for the average daily

creasing traffic from 0 to y.

ψ xy(a−r0 −r1)< ψ xy(r)

⋅ 2(x + 2 − y)

ψ xy(a−r0 −r1)< y

x

(x + 2)!⋅e−y ⋅ 2

1− yx+3

3

Dividing by the traffic volume concerned,y2

a − r0 − r1( ),

H (a−r0 −r1 ) =

e−y

0

y

∫y

x+1

(x +1)!+ 2

yx+2

(x + 2)!+ 3

yx+3

(x + 3)!+K

(a − r0 − r1 )

ydy

<(a − r0 − r1 )

1− yx+3

3y

x+1

(x + 2)!e

−y

y = y at t = a − r1.

We will now determine the traffic volume H (a−r0 −r1 ) that

dt =(a − r0 − r1 )dy

y=

a2 − r0( )dy

y0

as well as the line y = y.

y = y(a − r0 − r1 )

t −r0 y0a2 − r0( )

=y0

a2 − r0( ) t −

r0 y0a2 − r0( )


Consequently:

Setting now 1) x = 17, y = 9 and 2) x = 10 and y = 4, and insert-ing the stipulated value of a, we get the ratios:

1)

2)

As earlier suggested we cannot assume any certainty that this cal-culation is on the safe side. On the other hand we know that the

respectively, are greater than the likely values. If, on the otherhand, we set r = 1, which for the applied curve must be ex-pected to give too small ratios, then we get the values respec-tively:

1)

2)

Altogether we have after the previous reason to consider the ratio

good agreement with the real or likely expression for usefulvalues of x and y. We can bring the value closer to that calcu-lated by means of the quantities r0 and r1 by increasing the

In the latter case we get the ratio

This gives for

2(a − 2)(a + 2)(x + 2 − y)

+ 4(a + 2)

= 117

26(x + 2 − y)

+ 4

.

factor(x + 2 − y)

2(x + 3)

3

(x + 3 − y)3

(x + 2)3 , which can be replaced by

1(x + 3- y)

or by 1(x + 2 − y)

.

2(a − 2)(a + 2)

⋅ (x + 2 − y)2

(x + 3 − y)3 ⋅ (x + 3)

3

(x + 2)3 + 4

(a + 2)

to be in very

2 ⋅14 ⋅82 ⋅133

16 ⋅93 ⋅123 + 2

16= 0.320K = approx. 32

100

2 ⋅14 ⋅102 ⋅20

3

16 ⋅113 ⋅19

3 + 216

= 0.278K = approx. 28100

previously found ratios under 1) and 2), 43100

and 46100

2 ⋅13 ⋅82 ⋅133

17 ⋅93 ⋅123 + 4

17= 0.405K = approx. 41

100

2(a − 2)(x + 2 − y)2

(x + 3)3

(a + 2)(x + 3 − y3

(x + 2)3 + 4

a + 2=

2 ⋅13 ⋅102 ⋅20

3

17 ⋅113 ⋅19

3 + 417

= 0.369K = approx. 37100

HM

< e−y y

x

(x +1)!1

1− yx+2

2

2 a − 2( ) x + 2 − y( )2x + 3( )3

(a + 2)(x + 3 − y)3

(x + 2)3 + 4

(a + 2)

M = 12

a + 2( )y

These ratios are, respectively, 4/100 and 3/100 less than theratios that give full certainty that we do not calculate too smallvalues. The resulting differences with a degree of hindrance of1 ‰ in the busiest hour thus amount to 4 hundredths and 3hundredths of 1 ‰, ie. 4/100000 and 3/100000 of the total daytraffic, which must be considered sufficient exactness.

For a normal course of traffic (with k = 4 or a concentration of14.2 %) we can in accordance with the above approximatelycalculate the daily degree of hindrance by means of the alreadycalculated degree of hindrance for the busiest hour by setting:

(45)

44. If we want to construct a curve for a useful daily degreeof hindrance, say for 4/10 ‰, then according to the presentdevelopment one can apply the formula:

(46)

For values of x greater than 10 the ordinates of the former curvewill be higher than that of the latter, and the differences willincrease the more x is increased. That implies: For a givenconcentration the curve of a given daily degree of hindrance,say 4/10 ‰, will for a certain higher value of x = number ofcircuits or switches, say x > 10, allow a greater maximum traffic

degree of hindrance for the busiest hour, say 1 ‰, will allow forthe same value of x. By the expression ‘the correspondingdegree of hindrance for the busiest hour’ is understood the

intensity y = y than the curve for the corresponding

For ψ xy(d,4)= 4 / 10 ‰ this curve will approach the curve

for ψ xy(r)= ψ xy not far from the point x =10, y = 4.

ψ xy(d,4)= e

−y yx

(x +1)!⋅ 1

1− yx+2( )2 ⋅ 1

1726

(x + 2 − y)+ 4

By increasing the corresponding values of x and y in a

way that x + 2 − y upwards approaches 26, then the daily

degree of hindrance decreases towards 517

per mill.

ψ xy(d,4)= 1

1726

(x + 2 − y)+ 4

‰

If ψ xy(r)=1 ‰, then straightforward we get:

ψ xy(d,4)< ψ xy(r)

⋅ 117

26(x + 2 − y)

+ 4

2) 117

26(x + 2 − y)

+ 4

= 117

26(10 + 2 − 4)

+ 4

= 0.426K = approx. 43100

1) 117

26(x + 2 − y)

+ 4

= 117

26(17 + 2 − 9)

+ 4

= 0.388K = approx. 39100


degree of hindrance in the busiest hour that for a number ofcircuits x = 10 results from the same traffic that would becarried by 10 circuits with the same daily degree of hindrance.

There is thus for the two curves a difference for higher values ofx that should not be left out of consideration when evaluatingthe traffic requirements.

IX Calculation of the degree ofhindrance in systems with skewing of circuits (slip multiple)

45. So far we have only considered the cases where the xcircuits or switches that are arranged for common application bya certain number of subscribers or a certain amount of traffic areall equally accessible for different calls demanding service.

There exist, however, arrangements where two or more bundlesof x circuits work together in order to reduce the hindrances,partly by a peculiar skewing of the circuit arrangements, partlyby placement of a certain number of common spare circuits, orby a combination of such skewing and spare arrangements.

46. We will first consider the skewing system, and particularlywith regard to co-operating groups of preselectors.

Without skewing one will usually have that each group of x2

forward searching primary preselectors will be able to searchfor access by x fully co-ordinate secondary preselectors (or

primary group selectors), so that each group has its own com-plete system of common circuits. With skewing, on the otherhand, certain circuits will be accessible for two or more groupsof x2 preselectors. We will initially stick to the simplest andmost surveyable type of skewing, characterised by one stepforward for each circuit. By the combination of two groups of x2

primary preselectors in this way, numbering these from 1 to 2x2

and numbering the circuits from 1 to 2x, we will get:

The system, consisting of two mutually skewed groups, by thearrangement of the multiple connections in this manner forms aclosed ring.

In a similar way one can combine 3 or 4 or 5 ... or r groups.

47. We will now try to show how one can calculate appro-ximately hindrances and degree of hindrance for such anarrangement, consisting of r groups, each of x2 primary pre-selectors with rx skewed circuits or secondary preselectors, aswell as with an average traffic intensity y for each group in thebusy time.

Let us initially set r = 2.

We will denote by Rs the risk that because of the systematicarrangement a hunting selector bypasses all of the s free out ofthe existing 2x circuits. This risk is equal to the ratio of thenumber of fully occupied circuits to the total number of circuitsthat can be hunted for. Rs is thus a proper fraction and may bedenoted the systematic probability of hindrance. If s ≥ x + 1,then this probability is equal to 0. Thus we have R2x = R2x–1 =R2x–2 = ... = Rx+1 = 0. If s < x + 1, then there is always a possi-bility that the hunting occurs on a fully occupied set of circuitsand thus that the call ‘gets lost’. By a detailed examination ofthe combination problem at hand we find the following valuesof Rs for a skewed system as described above for a combinationof two groups:


Preselector number hasunlimitedaccess tockt. no.

on and after 1 up to and including x2 1

on and after x + 1 up to and including x2 + x 2

on and after 2x + 1 up to and including x2 + 2x 3

on and after 3x + 1 up to and including x2 + 3x 4

...

on and after (x – 1)x + 1 up to and including x2 + (x – 1)x x

on and after x2 + 1 up to and including 2x2 x + 1

on and after x2 + x + 1 up to and including 2x2 and 1 to x x + 2

on and after x2 + 2x + 1 up to and including 2x2 and 1 to 2x x + 3

on and after x2 + 3x + 1 up to and including 2x2 and 1 to 3x x + 4

...

on and after x2 + (x – 1)x + 1 up to and including 2x2 and 1 to (x – 1)x 2x

These expressions, where the numerator is greater than 1, arecompressed to give:

47a1K x

Rx =1⋅ 12xx

Rx−1 = 2 + x − 21

1

12x

x −1

Rx−2 = 3 + x − 31

2 + x − 2

2

1

12x

x − 2

Rx−3 = 4 + x − 41

3 + x − 3

2

2 + x − 2

3

1

12x

x − 3

K

Rx−i = ( i +1( ) + x − i +1( )1

i + x − i

2

i −1( )

+ x − i −1( )3

i − 2( )+K + x − i − p + 2( )

p

i − p +1( )

+K + x − 2i

⋅1) 1

2xx − i

K

R3 = ( x − 2( ) + 2 x − 3( ) + 3 x − 4( )+K

+ p x − p +1( )+K + x − 3( )2 + x − 2( ) ⋅1) 12x3

R2 = x −1( ) + x − 2( ) + x − 3( )+K +2 +1( ) 12x2

R1 = 12

We now set:

These sums can be expressed in the following way:

48a1K x

S1 = Rx

S2 = Rx + Rx−1

S3 = Rx + Rx−1 + Rx−2

KSi+1 = Rx + Rx−1 + Rx−2 +K + Rx−i

Sx−2 = Rx + Rx−1 + Rx−2 +K + R3

Sx−1 = Rx + Rx−1 + Rx−2 +K + R3 + R2

Sx = Rx + Rx−1 + Rx−2 +K + R3 + R2 + R1

47b1K x

Rx =1⋅ 12xx

Rx−1 = x +11

12xx

Rx−2 = x + 22

12xx

Rx−3 = x + 3x

12xx

K

Rx−i = x + ii

12xx

K

R3 = 2x − 3x − 3

12xx

= (x − 2)(x −1)2(2x − 2)(2x −1)

R2 = 2x − 2x − 2

12xx

= x −12(2x −1)

R1 = 2x −1x −1

12xx

= 12


The calculated quantities R and S apply as mentioned to a sys-tem of 2 groups with each x2 primary preselectors and together2x secondary preselectors.

For a triple system of corresponding type we obtain:

48b1K x

S1 =1⋅ 12xx

S2 = x + 21

12xx

S3 = x + 32

12xx

S4 = x + 43

12xx

S5 = x + 54

12xx

K

Si+1 = x + i +1i

12xx

K

Sx−2 = 2x − 2x − 3

12xx

Sx−1 = 2x −1x − 2

12xx

Sx = 2xx −1

12xx

= xx +1 The corresponding sums S take the following form, as we indi-

cate the triple system by means of a separate index:

For an r-tuple system we get:

501K 2x

S31

=1⋅ 13xx

S32

= x + 21

13xx

S33

= x + 32

13xx

K

S 32x−2

= 3x − 22x − 3

13xx

S 32x−1

= 3x −12x − 2

13xx

S 32x

= 3x2x −1

13xx

= 2xx +1

491K 2x

R2x =1 13xx

R2x−1 = x +11

13xx

R2x−2 = x + 22

13xx

R2x−3 = x + 3x

13xx

K

R2x−i = x + ii

13xx

K

R3 = 3x − 32x − 3

13xx

R2 = 3x − 22x − 2

13xx

R1 = 3x −12x −1

13xx

= 23


48. Hereby we have got means available for calculation ofhindrances and degree of hindrance in an ordinary skewing sys-tem consisting of any number of equally large groups of pre-selectors under the condition that the arrangement constitutes a‘ring’, and that the skewing for each group is also fully cyclic.

We shall now keep to the representation of traffic based onconsideration of a sequence of calls. The traffic intensity is stilldenoted y, which here indicates the average number of callsfrom x2 primary preselectors in the course of the average con-

be on average approximately equally large in each group.

In order to be able to carry out the calculations in a fully generalway we will assume that the system in question is so equippedin relation to the traffic that the hindrances in any case are very

nection duration t in the busy hour. y is assumed to

521K r −1( )x

Sr1

=1⋅ 1rxx

Sr2

= x + 21

1rxx

Sr3

= x + 32

1rxx

K

S r(r−1)x−2

= rx − 2r −1( )x − 3

1rxx

S r(r−1)x−1

= rx −1r −1( )x − 2

1rxx

S r(r−1)x

= rxr −1( )x −1

1rxx

=r −1( )xx +1

511K (r −1)x

R(r−1)x =1 1rxx

R(r−1)x−1 = x +11

1rxx

R(r−1)x−2 = x + 22

1rxx

K

R3 = rx − 3r −1( )x − 3

1rxx

R2 = rx − 2r −1( )x − 2

1rxx

R1 = rx −1r −1( )x −1

1rxx

=r −1( )

r

small. We will, therefore, in general not let the probability ofone particular call being hindered be of any advantage to suc-ceeding calls. It is easily seen that this simplification of the cal-culation does not cause any harmful error.

We have already calculated the probability of the various call

sponding to each state. The probability of x, x + 1, x + 2, ..., rx,

intensity ry is:

when x + 1 calls arrive, of which the x first occupy one circuiteach in the system. There is then 1 call with a risk of passing the

The loss that one can estimate will be caused by this state andrisk thus amounts to:

(54,1)

arrive, the risk for the first call after the xth is R(r–1)x and for thefollowing R(r–1)x–1. The likely loss then becomes:

(54,2)

In the same way we get, respectively, for 3, 4, ..., (r – 1)x calls

in excess of x in t the following likely losses:

q2 = Px+2 R(r−1)x + R(r−1)x−1( ) = Px+2 ⋅ Sr2

= e−ry

(ry)x+2

(x + 2)!⋅

x + 21

rxx

Under the state with probability Px+2 , when x + 2 calls

q1 = Px+1 ⋅ R(r−1)x = Px+1 ⋅ Sr1

= e−ry

(ry)x+1

(x +1)!⋅ 1

rxx

free circuits and thus get lost. This risk is: R(r−1)x = 1rxx

.

Under the state probability Px there are no traffic

hindrances. They only begin under Px+1 = e−ry

(ry)x+1

(x +1)!,

531K rxK

Px = e

−ry(ry)

x

x!

Px+1 = e−ry

(ry)x+1

(x +1)!

Px+2 = e−ry

(ry)x+2

(x + 2)!K

Prx = ery

(ry)rx

(rx)!K

... calls, respectively, in t under an aggregate traffic

states as well as the number of calls during t corre -


+ e−ry

(ry)rx+1

(rx +1)!(r −1)x(x +1)

⋅ 1

1− ry(rx+2)( ) + 1

1− ry(rx+2)( )2

< e−ry

rxx

⋅ (ry)x+1

(x +1)!(e

ry − (ry)(r−1)x

r −1( )x( )! (1+ ry

rx − x−12( )

+ (ry)2

rx − x−12( )2 + (ry)

3

rx − x−12( )3 +K + (ry)

2x−1

rx − x−12( )2x−1 ))

+ e−ry

(ry)rx+1

(rx +1)!(r −1)x(x +1)

⋅ 1

1− ry(rx+2)( ) + 1

1− ry(rx+2)( )2

= e−ry

rxx

⋅ (ry)x+1

(x +1)!e

ry − (ry)(r−1)x

r −1( )x( )! ⋅1− (ry)

2x

rx− x−12( )2x

1− ry

rx− x−12( )

= e

−ry

rxx

⋅ (ry)x+1

(x +1)!⋅(e

ry − (ry)(r−1)x

r −1( )x( )! (1+ ry

r −1( )x +1( )

+ (ry)2

r −1( )x +1( ) r −1( )x + 2( ) +K

+ (ry)2x−1

r −1( )x +1( ) r −1( )x + 2( )K r −1( )x + 2x −1( ) +K ))

+ e−ry

(ry)rx+1

(rx +1)!⋅ (r −1)x

(x +1)⋅ 1

1− ry(rx+2+k1 )

+ e−ry

(ry)rx+1

(rx +1)!⋅ 1

1− ry(rx+2+k2 )

2

< e−ry

rxx

(ry)x+1

(r +1)!⋅(e

ry − (ry)(r−1)x

r −1( )x( )! (1+ ry

r −1( )x +1( )

+ (ry)2

r −1( )x + 32( )2 + (ry)

3

r −1( )x + 42( )3 + (ry)

4

r −1( )x + 52( )4 +K

+ (ry)2x−1

r −1( )x + 2x2( )2x−1 ))


Under the state corresponding to this last equation there willarrive exactly as many calls as there are circuits. Despite thefact that some of these calls will be blocked, so that one or morecircuits will be left free, we will still, in consistency with whathas been said above, not take advantage of this circumstancewhen deciding the loss for higher number of calls than rx in

to 1 for each call exceeding rx. Thus we get:

Q rxy

= q1 + q2 + q3 + q4K +q(r−1)x + q(r−1)x+1

+q(r−1)x+2 + q(r−1)x+3 +K

= e−ry

rxx

(ry)x+1

(x +1)!⋅(1+ ry

1+ (ry)

2

2!+ (ry)

3

3!+K

+ (ry)(r−1)x−1

r −1( )x −1( )!)

+ e−ry

(ry)rx+1

(rx +1)!(r −1)x(x +1)

⋅(1+ ry(rx + 2)

+ (ry)2

(rx + 2)(rx + 3)

+ (ry)3

(rx + 2)(rx + 3)(rx + 4)+K ) + e

−ry(ry)

rx+1

(rx +1)!

⋅(1+ 2ry

(rx + 2)+ 3

(ry)2

(rx + 2)(rx + 3)

+4(ry)

3

(rx + 2)(rx + 3)(rx + 4)+K )

The aggregate loss in t thus amounts to:

551,2,3K

q(r−1)x+1 = Prx+1 ⋅ S

r(r−1)x

+1

= e−ry

(ry)rx+1

(rx +1)!(r −1)x

x +1+1

q(r−1)x+2 = Prx+2 S r(r−1)x

+ 2

= e

−ry(ry)

rx+2

(rx + 2)!r −1( )xx +1

+ 2

q(r−1)x+3 = Prx+3 S r(r−1)x

+ 3

= e

−ry(ry)

rx+3

(rx + 3)!(r −1)x

x +1+ 3

K

t. So we assume the risk of hindrance to be equal

543K (r −1)

q3 = Px+3 R(r−1)x + R(r−1)x−1 + R(r−1)x−2( ) = Px+3Sr

3

= e−ry

(ry)x+3

(x + 3)!

x + 32

rxx

q4 = Px+4 ⋅ Sr4

= e−ry

(ry)x+4

(x + 4)!

x + 43

rxx

K

q(r−1)x = Prx ⋅ S r(r−1)x

= e−ry

(ry)rx

(rx)!(r −1)x

x +1

When we load the ordinary bundle only moderately, ie. we letx0 be substantially larger than y, then we may consider theexpression behind the inequality sign as approximately equal toy1. The traffic that resort to the spare section from all r groups isequal to ry1. The number of circuits that are for common use bythis traffic is as mentioned x1. The amount that is hinderedbecause of busy or lacking circuits in the spare section is there-fore:

(58)

The degree of hindrance becomes:

(59)

52. On this basis we will calculate the degree of hindrance for asystem with x0 = 8, x1 = 2, y = 6, r = 2. This gives:

log y1 = 9 log 3 + log 128 + log 25 – 6 log e – log 9!= 9 ⋅ 0.4771213 + 2.1072100 + 1.3979400– 6 ⋅ 0.4342945 – 5.5597631 =

4.2940917+ 2.1072100+ 1.3979400 7.7992417

+ 1. – 1– 2.6057670– 5.5597631 – 8.1655301

= 0.6337116 – 1

Num. log y1 = 0.43024 = y1.

One must remark here that because of the small difference be-tween x0 and y we get a somewhat too large value of y1causedby setting the previously discussed quantity k equal to 2 in the

fraction 1

1− yx+k( )2 . It should correctly be somewhat

y1 = e−6

69

9!10

2

42 = e

−6

9!3

9 ⋅29 ⋅52 ⋅22

24

= e−6

9!3

92

75

2

1

ψ r(x1y1 ) =Qr(x1y1 )

ry=

e−ry1 ry1( )x1 +1

ry x1 +1( )!1

1− ry1

x1 +2

2

Qr(x1y1 ) =e

−ry1 ry1( )x1 +1

x1 +1( )!1

1− ry1

x1 +2

2

y1 = e−y

yx0 +1

x0 +1( )! 1+ 2y

x0 + 2( ) + 3y

2

x0 + 2( ) x0 + 3( ) +K

< e−y

yx0 +1

x0 +1( )! 1− yx0 +2

2

greater than 2, how much depends on x and y, but is otherwise aquestion that is not easily answered. It may be noted that arecord of traffic hindrances based on Bortkewitsch’ table showsa total hindrance of 0.3192 for x = 8 and y = 6. But this numberis determined by summation of only a limited number of termsof a series that is not strongly converging, and should only beused with care. If we take the mean value of the two numbers0.43024 and 0.31290 we get a value that does not differ muchfrom the likely one. Thus for further treatment of the matter we

have ry1 = 2 · 0.37157 = 0.74314. We round this numberupwards to 0.75 and set ry1 = 3/4. Furthermore we will for therest of the calculation set k = 3, which is more in agreementwith the mentioned table. Thus we write

– log 6 – log 64 – log 289 =

1.4313638+ 2.6020600

4.0334238+ 2. – 2

– 0.3257209– 0.7781513– 1.8061800– 2.4608978

– 5.3709500= 0.6624738 – 2

The degree of hindrance for the system in question with a trafficintensity of y = 6 thus is near 4 ‰, which is more than recomm-ended.

If we allow for y = 5.5 we get for the same system, when we inboth cases set k = 3:

y1 =e

−112 11

2( )9

9!⋅ 22

2

112 = e

−112 11

9 ⋅22 ⋅112

9!⋅29 ⋅112

= e−11

2 119

9!⋅27

We thus getQr(x1y1 )

ry= 0.04597

12= 0.00383.

Num. log Qr(x1y1 ) = 0.04597 = Qr(x1y1 )

logQr(x1y1 ) = log 27 + log 400 − 3 / 4 log e

Qr(x1y1 ) =e

−ry1 (ry1 )x1 +1

(x1 +1)!1

1− ry1

x1 +3

2

= e− 3

434( )3

3!1

1−34

5

2

= e− 3

4 ⋅33 ⋅202

3!⋅ 43 ⋅17

2 = e− 3

4 ⋅27 ⋅ 4006 ⋅64 ⋅289

set y1 = 0.43024 + 0.312902

= 0.37157. Thus we have


work, and presumably ought to be left for closer investigationby means of pure experience. Experience is also what offers thenecessary guidance with respect to the application of thetheoretically based results, and what will serve to deepen theunderstanding of, and thereby also provide an increasingly exactmathematical treatment of the technical problems.

XII Summary

58. On the basis of pure probability considerations and througha series of simplifications of the derived general mathematicalexpressions we have found that the likely degree of hindranceψxy of an automatic telephone system for a certain number x offully co-ordinate and co-operating central exchange circuits thatare supplied to carry a traffic of intensity y in the selected time(a busy hour or a busy quarter-hour), with sufficient accuracycan be expressed by:

By the degree of hindrance is here understood the ratio betweenthe part of telephone traffic that suffers hindrance because of theinsufficiency of the x circuits, and the total traffic that searchesfor access over the same circuits. By traffic intensity is under-stood the average number of simultaneous telephone connec-tions over the selected time. This number is equal to the averagenumber of telephone calls that search for access over the xcircuits in the course of the average duration of a connection inthe selected time.

We have shown that the simplifications done are always allow-able under certain further stated conditions, given in mathe-matical form, and we have thereby been able to make certainthat we are on the safe side with our calculations.

We have shown that, by including a limited number of termsfrom the given infinite series, we get a result very close to thatof the expression behind the inequality sign when doing cal-culations for values of x and y that lead to small hindrances,corresponding to those that will be allowed in practice. As anallowable degree of hindrance we have in accordance with curr-ent practice set forth one per mill, and we have constructed acurve for a such degree of hindrance on the basis of calculations

shows the value of the traffic y that is allowed for a given x orthe number x needed to carry a traffic y. This curve is presentedbeside two other curves that have been utilised by telephoneengineers as guidance for determining x in relation to y or of yin relation to x, and which have been constructed on the basis ofcalculation of the probability of blocking, ie. the probability ofall x circuits being busy. These curves are denoted, respectively,1 ‰ and 4 ‰ curves for full load.

59. Furthermore, we have carried out calculation of the relationof the average daily degree of hindrance to that of the busyhour, and we have found that with the concentration of traffic in

where we have set ψ xy = e−y

yx

(x +1)!⋅ (x + 2)

2

(x + 2 − y)2 . The curve

ψ xy = e−y

yx

(x +1)!1+ 2y

(x + 2)+ 3y

2

(x + 2)(x + 3)+ 4y

3

(x + 2)(x + 3)(x + 4)+K

< e−y

yx

(x +1)!⋅ 1

1− yx+2( )2

the busy hour such as usually found in larger cities one maystate that a degree of hindrance of 1 ‰ in the busiest hourcorresponds to an average degree of hindrance of 4/10 ‰ forthe daily traffic.

60. Next we have calculated the degree of hindrance for what isnamed a skewing system (slip multiple system), where the co-operating central exchange circuits are not fully co-ordinate, butskewed in relation to one another. As we combine in thelongitudinal direction of the system r groups of such skewedcentral exchange circuits, each group consisting of x circuitswith attached selectors, and make sure that the combination isarranged as a complete cycle and that in each group there areequally many skewing steps as there are skewed circuits, thenfor a traffic of y for each group we can determine the likelydegree of hindrance for the skewing system by:

By a detailed calculation we have by means of this shown that askewing system consisting of 4 groups of ordinary 10-contactselectors result in an increased accessibility of at least 37 1/2 %compared to a system of the same selectors without skewing.

61. Furthermore we have calculated the likely degree ofhindrance for a non-skewing system with spare circuits, ie. asystem where a certain number of fully co-ordinate centralexchange circuits with attached secondary selectors are com-mon for several groups of primary selectors. The calculation isbased on the expression in this report for the likely hinderedpart of the traffic, which is the degree of hindrance multipliedby the total or non-hindered traffic. We have found that byletting 2 secondary selector groups with 8 circuits or selectorsin each group be supplemented by 2 spare circuits with attachedselectors, which can receive for forwarding the calls that arehindered in the 2 groups, we can also obtain an improvement ofaccessibility of 37 1/2 % (or a saving in the number of concernedselectors of 27 3/11 %), not considering the saving of secondaryselectors immediately resulting from the spare arrangement. We

justified this modification when the ratio y/x by a reduction of xreaches a certain magnitude, which in a system with spareselectors will be the case for the secondary selector groups thatreceive the majority of calls.

62. Finally, we have by application of the derived equationswith corresponding use of the last mentioned factor calculatedthe degree of hindrance for a couple of combined skewing andspare systems consisting of 10-contact selectors arranged in asimplest possible manner. In this way we have found an im-

have under this instead of the factor 1

1− yx+2( ) in the

relevant expressions applied 1

1− yx+3( ) , as we have

ψ rxy

=

e−ry

(ry)x

(rx − x)!(rx)!(x +1)

1+ ry1!

+ (ry)2

2!+ (ry)

3

3!+K + (ry)

rx−x−1

(rx − x −1)!

+ e−ry

(ry)rx

(rx − x)(rx +1)!(x +1)

⋅ 1

1− ryrx+2( ) + e

−ry(ry)

rx

(rx +1)!⋅ 1

1− ryrx+2( )2


provement of the access possibilities of 50 % beside a saving inthe number of secondary selectors of 12 1/2 to 13 1/3 % obtainedby the spare arrangement. This results in a total increase in theefficiency of a certain larger number of secondary selectors of73 10/13 % (or with a given traffic a total reduction of 42 2/9 % ofthe number of these selectors), all under the assumption that theconditions for the application of the calculations are satisfied.

63. The last mentioned arrangements allow such a reduction ofthe x circuits needed for the traffic y, while retaining the per-mitted degree of hindrance, that we in part exceed theconditions that guarantee the calculations to always be on the

safe side. Possible calculations of what can be obtained byextended skewing arrangements supplemented by spareselectors for further increasing the accessibility seem, therefore,to be linked with special difficulties on the present basis, whenthe calculations are required to give sufficiently certain results.Consequently, they probably ought to be carried out by appli-cation of empirical formulas to be derived by means of experi-ments or experience, in case more complicated arrangements ofthe types mentioned can be regarded as useful in spite of themore complicated circuitry that will be involved.


y = 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0

x =

0

1

2 0.0328 0.1037

3 .0040 .0234 .0570

4 .0004 .0044 .0159 .0375

5 .0007 .0036 .0112 .0246

6 .0001 .0007 .0029 .0078 .0164

7 .0001 .0007 .0022 .0053 .0118

8 .0001 .0005 .0015 .0042 .0084

9 .0001 .0003 .0013 .0041 .0062

10 - .0004 .0010 .0024 .0044

11 .0001 .0003 .0008 .0017 .0032 .0057

12 - .0001 .0003 .0006 .0012 .0023 .0042

13 - .0001 .0002 .0004 .0009 .0018 .0031

14 - - .0001 .0003 .0007 .0013 .0024

15 - .0001 .0002 .0005 .0010 .0018

16 - .0001 .0002 .0004 .0008 .0014 .0023

17 - .0001 .0002 .0003 .0006 .0010 .0017

18 - . .0001 .0002 .0004 .0008 .0013

19 - .0001 .0002 .0003 .0006

20 - - .0001 .0003

21 - - - .0001

22 - -

23 -

24

Annex I

Table showing the degree of hindrance for different values of x = the number of circuits and y = the traffic intensity according to calculationsbased on Bortkewitsch’ tables.


y = 10.0 10.5 11.0 11.5 12.0 12.5 13.0 13.5 14.0 14.5 15.0

x =

18

19 .000680 .001144 .001852

20

21

22 .000871 .001338 .002041

25

30

40

50

60

70

80

90

100

y = 19.0 19.5 20.0 35.0 36.0 37.0 52.0 53.0 54.0 79.0 80.0 81.0

x =

18

19

20

21

22

25

30 .000954 .001369 .001919

40

50 .000609 .001036 .001703

60

70 .000525 .000812 .001232

80

90

100 .000593 .000839 .001173


Annex III

Presentation of values of traffic intensity y and efficiency α with a degree of hindrance ψxy = 0.001 corresponding to circuit numbersof x = 4, 5, 6, ..., 22, 30, 40, ..., 100. (For the underlined values of x the calculation of y is done by means of the formulas given inthe report. For the others interpolation is applied.)

x y αα x y αα

4 0.70 0.175 18 9.62 0.534

5 1.12 0.224 19 10.37 0.546

6 1.60 0.266 20 11.13 0.556

7 2.12 0.303 21 11.89 0.567

8 2.71 0.339 22 12.67 0.576

9 3.33 0.370 30 19.07 0.636

10 3.96 0.396 40 27.40 0.685

11 4.62 0.420 50 35.92 0.718

12 5.29 0.441 60 44.61 0.743

13 5.98 0.460 70 53.46 0.764

14 6.69 0.478 80 62.40 0.780

15 7.41 0.494 90 71.42 0.794

16 8.13 0.508 100 80.53 0.805

17 8.87 0.522

Annex IV

Diagrams showing the likely relation between traffic and circuits in an ordinary automatictelephone system.

Y

0 10 20 30 40 50 60X

45

40

30

20

10

0

80

70

60

50

40

35

Y

60 70 80 90 100X

Annex V

Diagram showing the likely efficiency ofcircuits in an ordinary automatic telephone sys-

0 20 40 100X

0,80

0,60

0,40

0,20

0,00

α

60 80

Remarks on the translation

General

The present document is a translation from Norwegian toEnglish of a report written in 1915. The time span from the timeof writing to the time of translation is thus more than 80 years.This fact has naturally influenced the translation on severalcounts and in several ways. To the general development comes,possibly, personal traits of author and translator as well.

As the translator in case I want to point out the following:

• As most languages of modern societies undergo continuouschange, this has probably been the case to a larger extent forNorwegian than for most other languages because of itsparticular history of the last century.

• The author appears to have been a very accurate person whoall along wanted to make sure that all relevant conditions andreservations at any point were clearly expressed. That seemsto have led to a meticulous accumulation of characterisingwords and many subordinate clauses.

• The development of underlying technology during the giventime span has been formidable, and even a present day trans-lation should not presume this development.

• The concepts and terminology of teletraffic theory and that ofprobability theory in general has undergone a similarly for-midable development, and should not be fully anticipated inthe translation of a text originated in 1915.

• In veneration of the author and with a wish to keep the timecolouring of the original report, not much effort has beendone to modernise the text, but for some cases when a longsingle period has been cut in two or more, while avoiding anyalteration of the meaning.

Specifics

In the report there are some quite central terms that may betranslated in different ways, and a choice has to be made inorder of correct understanding and in view of the statementsabove. Some of these terms have even been replaced by newterms in modern Norwegian. Some examples will be listed andcommented.

• velger (orig.: vælger) The most direct translation is selector,which is also used in chapters IX – XI about gradings. Inthese chapters the pre-selection function (in two stages) istreated. As a generic term in the rest of the report we haveused switch. That includes different functional devices, likeselector, preselector, finder, connector, etc.

• hindring, hindringsgrad: hindrance, degree of hindrance.This is the one most central term in the report. Today thecommonly used Norwegian term is ‘sperr’ = blocking orcongestion. In the main traffic model of this report a com-bined delay and blocking concept is more appropriate, andthe proposed terms seem very suitable. It is worth mentioningthat in the only known publication by Engset, written inGerman, the terms ‘Hinderung’ and ‘Hinderungsgrad’ areused.

• blokkering, overfylling: blocking, overfill. Those aredescriptive terms for the probabilities of at least x (= numberof servers) and at least x + 1 positions occupied. Here there isno weighting by the number in queue.

• forskyvning, reserveledninger: skewing, spare circuits. Theterm grading (Norw.: ‘gradering’) is not applied and wasprobably not in common use. The report treats two types ofgrading, one of a slip multiple type and the other progressive.The first has a skewed structure and the other straight.

A R N E M Y S K J A


A typed report of 130 pages written by Tore Olaus Engsetin 1915 indicates a much more extensive pioneer work inthe teletraffic field by Engset than has been publicly known.The well known Engset formula is solely based on the onlypublished teletraffic work by Engset, a two-page articlefrom 1918.

The 1915 report was discovered by Villy Bæk Iversen in thefiles of Copenhagen Telephone Company, and was first pre-sented in a guest lecture at the University of Trondheim in1995 and in a short paper at the 13th Nordic TeletrafficSeminar (NTS-13) in Trondheim in 1996. The assumed ori-ginal of the report was later found at Norsk Telemuseum,Oslo.

The Engset report contains original modelling of teletrafficaccording to a model that was taken up by Conny Palm in amore general form more than twenty years later. It is aqueue system with departure from the queue. Engset evenallows an arbitrary number of individual traffic sources. Heincludes aspects that are only treated much later by Palm,Cohen, Dartois and others. Even daily variations aremodelled, and simple cases of progressive and slippedmultiple gradings and combinations of both are treated.

Throughout, Engset aims at practical useful modelling. Hetherefore introduces simple approximations, and he is keento assure that his approximations are on the safe side. Withthe same objective he even presents dimensioning diagramsfor practical use.

1 Introduction

In the teletraffic community Engset is a household nameconnected with the mathematical formula that carries his name.Until recently the only known teletraffic work by Engset is thearticle that presents his formula: ‘Die Wahrscheinlichkeits-rechnung zur Bestimmung der Wählerzahl in automatischenFernsprechämtern’. The article, which is a concentrated workof merely two pages, appeared in Elektrotechnische Zeitschrift,1918, Heft 31 [1] and in a French translation [2] in 1921. AnEnglish translation by Eliot Jensen appeared in Telektronikk No.1, 1992 [3], with the title ‘The Probability Calculation to Deter-mine the Number of Switches in Automatic Telephone Ex-changes’. In the same issue of Telektronikk Eliot Jensen alsopresented a commentary [4], discussing Engset’s paper in thecontext of later developments.

In 1995 a copy of a report from 1915 by Engset [5] was dis-covered in the files of Copenhagen Telephone Company(KTAS). A survey of the report was presented in a guest lectureat the Norwegian Institute of Technology by Villy Bæk Iversenthe same year, and a brief written presentation appeared in therecords of The Thirteenth Nordic Teletraffic Seminar of August1996 [6]. In the abstract of this paper Iversen states: “Based onan unknown report by Engset from 1915 and a review of thehistory of traffic engineering in the period 1900 – 1925 it isshown that Engset’s contributions to traffic theory are morefundamental than so far acknowledged”.

On this background, and after getting acquainted with a copy ofthe report, the question was brought up with editor Ola Espvikof the Telenor (Norwegian Telecom) technical journal Telek-tronikk, whether there was an interest in publishing informationabout Engset’s more extensive work. First a search in the Tele-

nor files was initiated, and the presumed original of the KTASreport copy was found.1 Telektronikk’s editor encouraged atranslation into English [7] of the report, as it was assumed thatthere might be some interest outside Scandinavia, and as it wasin line with the present general policy of Telektronikk to addressitself to the international telecommunication community.

The main interest in the report today will be from a historicalpoint of view, as there are no aspects of the report that have notbeen studied and reported during the ensuing 83 years. It isworth noting, though, that the report was finished in 1915, whileErlang’s B-formula [8] came in 1917 and Engset’s formula [1]in 1918. Since also the modelling procedure is different fromthat of Erlang, and several aspects of the report were firsttreated in publications many years later, there is good reason forthe above quote by Iversen. It is fair to say, though, that Engsetin his 1918 paper gives due credit to Erlang and his colleaguesin Copenhagen, and it is my opinion that even if the 1915 reporthad been published, it would not have reduced the originalvalue of Erlang’s pioneering work.

143

The Engset report of 1915Summary and comments

A R N E M Y S K J A

Telektronikk 2.1998

1 There is a number (some 25 noted) of minor differencesbetween the two versions, mostly due to manual insertions inthe typewritten text. Generally, the Telenor version is themore complete/correct one.

Tore Olaus Engset, 1865 – 1943

Pho

to:N

orw

egia

nT

elec

omm

unic

atio

nsM

useu

m


A page from the original Engset report of 1915

Pho

to:S

vein

EF

aste

ng/S

AM

FO

TO

he specifies n + 1 different states, which in fact is descriptive ofa system with x servers and q ≥ n – x queue positions. Theimplication of this is a queuing system where departure from thesystem is independent of whether it is from a server or from awaiting position. From the outside the system looks like that ofan infinite server group with no ordering of positions, witharrivals and departures according to some given distributions.From the point of view of an arbitrary customer arriving he mayfind a free server and immediately occupy it. Otherwise, he maystay waiting until a server becomes free (he may have to com-pete), or he loses patience and leaves. Only in the latter case isthe call lost.

The outcome of these assumptions is that within a given timeperiod (T = Ts + Tw) there will be two distinct traffic volumes,one given by the total time (Ts) spent in the server group and theother given by the time (Tw) spent waiting. The calls willconsist of three categories, those only being served, those onlywaiting, and those first waiting and then being served. Engsetdefines two relative quantities: the ‘degree of efficiency’ϕ = Ts / (Ts + Tw) and the ‘degree of hindrance’ ψ = Tw /(Ts + Tw). Consequently, ϕ + ψ = 1. There is thus no defined‘loss’ or ‘congestion’. There are of course lost calls, namelythose that leave the system (for loss of patience) without beingserved. This part is not calculated, and neither are waiting times.

One may note here that this Engset model is neither the ‘lostcalls cleared’ (Erlang) nor the ‘lost calls held’ (Molina) model.A comprehensive treatment of what in this context may betermed an extended Engset model by Conny Palm is given in[10] and [11] (K. Telegr.-Styr. Tekn. Med. No 7-9, 1937 andTele No. 1, 1957). The Palm model, which naturally has noreference to Engset’s unpublished work, treats a system withunlimited queue with exponential service and waiting, howeverwith separate service and waiting parameters. Engset’s case is aspecial case of Palm’s general model, with a common holdingtime parameter.

Apart from this preliminary discussion of the main model weshall point out a few particulars of the report:

1 A limited number of identical traffic devices serve a larger,still limited, number of individual traffic sources. A callingsource gets immediate service if there is at least one freedevice, otherwise it is allowed to wait until it gets delayedservice, or it leaves the system voluntarily.

2 All traffic sources are mutually independent, calls arrive atrandom and there are no dependencies between call arrivalsand holding times.

3 No assumption is made about holding time distributions.Individual traffic characteristics are according to the generali-sed Engset formula as later introduced by Cohen (1957) [12].

4 Engset presents a proof that identical sources is the leastfavourable case. A publication on this matter was first pre-sented in 1970 by Dartois [14].

5 Several simplifications are made in order to be able to cal-culate particular cases. Foremost is the assumption of equaloffered traffic from all sources, the assumption of an un-limited number of sources (Erlang), and introduction ofapproximations. Engset is consistent in proving that appro-ximations are on the safe side, ie. the approximation is lessfavourable than the exact solution.


2 The name of Bortkewitsch is according to his book inGerman, edited in Leipzig 1898. In some library referencesthe name is spelled Bortkiewicz.

One can only speculate why Engset did not go on to publish hisfindings in the 1915 report. First of all, there was no traditionfor scientific publishing in his working environment. Inconnection with system studies for a possible introduction of anautomatic telephone system in Norway a national committeehad been established, with Engset as one of the committeemembers. The work led to close contact with Danish teletrafficexperts, and the complete 1915 report was made available inCopenhagen (and probably in Stockholm). The report containsreferences to Erlang and Christensen (Denmark), Grinsted (Eng-land) and Spiecker (Germany). The Grinsted/Spiecker referenceis applied by Engset in one of three different approaches toobtain the binomial (Bernoulli) formula with the Poissonformula as a limit case. The Erlang/Christensen reference is to a1913-publication with the so-called Christensen formula, thatapplies the Gaussian error curve for dimensioning.

Engset’s approach seems to be his own original one, based onindividual traffic demands. It permits waiting for service, andthe subscriber holding time is independent of the fate of the call.Engset may have been discouraged from further pursuit of hismodel, simply by becoming convinced that Erlang’s limitedserver group truncation was a better model. This truncation isadopted in the 1918 paper, which otherwise applies the generalapproach of the 1915 report.

In the following we shall mainly stay with the 1915 report andonly by particular mention refer to the 1918 paper.

Engset’s formal education may look rather modest. After hisjunior high school exam and telegraph course in 1883 he nevercarried out any full time studies. However, based on spare timework he graduated as a student in 1892 and as M.Sc. inmathematics and physics at university in 1894. In itself thisindicates a very great capacity. His teletraffic work shows thathe was well versed in mathematics at a medium to high level,and his works later in theoretical physics (not to be discussedfurther here) indicate a notorious capability in advancedmathematics. Engset’s approach to teletraffic analysis is clearlycoloured by the deep insight in the practical aspects that he hadacquired as a telephone engineer with many years of experience.The main reference in the report is Bortkewitsch: ‘Das Gesetzder kleinen Zahlen’ [9].2

Engset is very conscious of making his assumptions clear. Heapplies mainly two basic theorems of probability calculation,that of the probability sum of disjunct events and that of theprobability product of mutually independent events. Further-more, he starts with the basic assumption of a limited number ofindividual traffic sources, each of which asks for a certain ser-vice time per time unit. This traffic demand may be different foreach source. No particular holding time distribution is assumed,and neither is any interarrival time distribution prescribed, onlyfull independence between sources.

Engset applies the concept of ‘statistical equilibrium’. His mainmodel assumes a limited number of n sources and a limitednumber of x servers. There is no mention of a queue. However,

6 The independence condition between arrivals is shown to leadto the binomial (Bernoulli) distribution and in the limit toPoisson distribution (referred to as Bortkewitsch’ formula).

7 Engset discusses at length the highest one hour concentration(busy hour) of the full 24 hour traffic, and introduces a geo-metrical model for this, in order to find the total 24 hour loss,given the busy hour loss.

8 Engset carries out detailed studies of slip multiple gradings,progressive gradings and combinations of both, with examplecalculations. Like other early authors on the matter he seemsnot to be aware of the special overflow characteristics ofrandom traffic.

9 Calculations according to two approximations are carried outand the results tabulated. Diagrams for dimensioning pur-poses are drawn.

2 Preliminaries

In the introductory Chapter I Engset introduces the classicalmodel of random drawing of black and white balls from an urn,along with the two ‘main theorems of probability’: 1) The sumtheorem of mutually disjunct events, and 2) The producttheorem of mutually independent events.

In Chapter II a detailed discussion of practical telephone trafficaspects is carried out. A telephone exchange with a set of xidentical connecting devices (lines, selectors) is assumed. Thesedevices serve a set of n individual traffic sources (subscribers).An observation time Tc is assumed to be a period of homo-geneous traffic. A number r of calls in Tc with an averageduration of t defines an average traffic intensity of y = rt / Tc.The traffic is individualised by assuming a subscriber Ai beingbusy with outgoing calls an accumulated time Ti out of totalobservation time Tt = D · Tc, where D is a (great) number ofdays in order to get enough observations for a single subscriber.An alternative to measuring the single intervals that make up Tiis to assume that Tt consists of a very large number N of veryshort intervals ∆t and counting the number of intervals mi whenAi is busy. Then the expression Ti / Tt = mi / N = pi may be inter-preted as the relative time that Ai is busy = the probability thatAi is busy = connection probability of Ai = the traffic load of Ai.(Throughout, only outgoing calls are considered.)

3 State probabilities, efficiency and hindrance

In Chapter III the combined states of the n subscribers arestudied. Only the number r of busy subscribers is of interest,and any state is completely described by this number. Thus,with n subscribers there can be n + 1 different states: 0, 1, 2, ...,r, ..., n. For the x ≤ n serving devices there can be only x + 1different states: 0, 1, 2, ..., x. The state x includes n – x + 1 totalstates with 0, 1, ..., n – x subscribers that are busy, but not beingserved. For the total number n + 1 of states there are as manyprobabilities defined: P0, P1, P2, ..., Px, Px+1, ..., Pn.

This system is identical to a queuing system with x servers andn – x (or an unlimited number of) queue positions. It is implicit-ly assumed that the subscriber holding time is independent of itsfate, if it is immediately served, if it is served after waiting inqueue, or if it leaves the queue without service.

An arbitrary state r is defined by r subscribers being busy andthe remaining n – r subscribers free. There are

individuals. Each subscriber i has its individual busy probabilitypi and the complementary free probability (1 – pi). The probability

free subscribers can be expressed on the form

is the jth product of (n – r) individual probabilities (1 – pk). Thetotal probability of state r will then be:

(1)

The normalising condition is:

P0 + P1 + P2 + ··· + Px + ··· + Pn = 1 (2)

It should be noted that Pr is the probability of r busy sub-scribers, whereas the probability of all x servers being busy is

Px + Px+1 + Px+2 + ··· + Pn

At this stage Engset introduces the concept of ‘state probabil-ity’, and he notes that the above calculation is valid irrespectiveof any hypothesis concerning the duration of the single connec-tions or the average duration of connections.

With these assumptions the total traffic intensity can be ex-pressed in two ways:

(3)

where each Pr is a function of p1, p2, ... , pn, as given above.

The traffic intensity y consists of two parts, one effective partthat can be measured directly on the x servers, and one waitingpart that can be measured on the calling subscribers while theyare waiting for service. There is no traffic rejected by the sys-tem. All subscribers waiting for service either leave the systemvoluntarily without being served, or they get delayed servicewhen finding a free device. The same call may contribute toboth parts. Engset considers the first part as the effective trafficand the second part as the hindered traffic.

In order to operate with relative quantities he defines a degreeof efficiency ϕnx and a degree of hindrance ψnx. These are in acompact way defined by

(4)ϕnx = 1

y⋅ r

r=1

x−1

∑ ⋅Pr + x⋅ Prr=x

n

∑

y= pii=1

n

∑ = rr=0

n

∑ ⋅Pr

Pr =j=1

j= nr

∑ Π j

(r)pi ⋅ Π j

(n−r)1− pk( )k≠i

of r individual probabilities pi . Similarly, Π j

(n−r)1− pk( )k≠i

Π(r)

j pi ⋅ Π j

(n−r)1− pk( )k≠i

. Here Π j

(r)pi means the j

thproduct

of each of the nr

combinations with r busy and n − r

nr

= n

n − r

= n!

r! n − r( )! ways of picking r out of n


(5)

ϕnx + ψnx = 1 (6)

Engset states that these are exact expressions, given that all pican be considered unaffected by the degree of hindrance. Hepoints out, however, that that may only be the case for smalldegrees of hindrance.

4 The first simplification of calculations

It may be noted that the assumption of individual trafficdemands, expressed by pi for any subscriber Ai, is in fact theassumption made for what is later called the generalised Engsetformula. The solution given above is thus in fact the generalisedsolution, however with the difference due to the different handl-ing of calls that are not given immediate service.

It is pointed out by Engset that this general assumption leads tovery complicated calculation even for limited numbers of sub-scribers of, say, n = 100. Because of this in Chapter IV hemakes the simplifying assumption of setting:

p1 = p2 = p3 = ... = pn = p (7)

This condition leads to the simple expressions for the stateprobabilities of

(8)

which is in fact the binomial (Bernoulli) distribution.

By introducing equation (8) in (4) and (5) and setting y = np weget the special expressions for the degree of efficiency and thedegree of hindrance (using the dashed symbols to indicate thedifference):

(9)

(10)

Engset investigates in his report how the condition (7) influences

the values of and in particular the

values for r > x. He shows that if for r = x + 1, then the

relation applies to all values r > x + 1, and then also

Essentially, this is the same conclusion as that reached byDartois [13] in 1970. Thus, under those conditions, it is safe touse the simplifying assumption of (7). The critical values of y =np is tabulated as a function of n and r (= x + 1) for some smallvalues. The conclusion is that the simplification of assuming alltraffic sources to be identical leads to maximum hindrance andthus is justified under normal circumstances, ie. for practicalvalues of traffic and degree of hindrance.

ψ nx >ψ nx .

Pr >Pr

Pr as compared to Pr ,

ψ nx = 1np

⋅r=x+1

n

∑ r−x( )⋅ nr

⋅pr ⋅ 1− p( )n−r

ϕ nx = 1np

⋅ rr=1

x−1

∑ ⋅ nr

⋅pr ⋅ 1− p( )n−r + x⋅

r=x

n

∑ nr

⋅pr ⋅ 1− p( )n−r

Pr = nr

⋅pr ⋅ 1− p( )n−r ,

Ψnx = 1y

⋅r=x+1

n

∑ r−x( )⋅Pr , where5 The second simplification of

calculations

A further simplification of expressions is obtained in Chapter Vby letting the number of traffic sources approach infinity, whilethe total traffic is kept constant. This is the basic Erlangassumption. Engset applies the elementary mathematical mani-pulation leading to

(11)

ie. the Poisson formula, which he denotes Bortkewitsch’formula.

Since Engset wants to stick to the assumption of a limitednumber n = n0 of sources, he intends to prove that above a cer-tain value of r all the Poisson terms are greater than thecorresponding Bernoulli terms. This is done by taking thederivative of (the log of) the Bernoulli terms with respect to n,and finding the condition that makes this derivative positive.This leads to the condition

(12)

When condition (12) is satisfied, the Poisson terms are greaterthan the corresponding Bernoulli terms, and the calculation ison the safe side when the Bernoulli terms in (10) are replacedby Poisson terms from r = x + 1 to r = n = n0:

(13)

The corresponding non-truncated, weighted sum is

(14)

where k can be chosen to satisfy the equation.

After some manipulation a more compact expression of thetruncated sum in (13) is obtained:

(15)

Again, in order to be on the safe side, the value h = 2 is chosen

(Later in the report, under the study of gradings, h = 3 isassumed to be a better choice.)

For practical calculations Stirling’s formula for the factorialfunction is proposed, to give:

(16)

ψ xy = 1

2π x +1( )⋅ y

x

x +1( )x+1 ⋅ ex+1−y

1− yx+2( )2 > ψ xy

and the notation is then replaced by ψ xy > ψ ∞x .

ψ xy = yx

x +1( )! ⋅e−y ⋅ 1

1− yx+h( )2 , where 2 < h <

n0 − x2

+ 3

ψ ∞x = e−y

y⋅

r − x( ) ⋅ yr

r!r=x+1

∞∑ = y

x

(x +1)!⋅e−y ⋅ 1

1− yx+k( )2

ψ ∞x = e−y

y⋅

r − x( ) ⋅ yr

r!r=x+1

n0

∑

r > y + 12

+ y + 14

n→∞lim Pnr =

n→∞

lim nr

⋅ p

r ⋅ 1− p( )n−r = yr

r!e

−y


(28)

The combinations of occupation states in this grading systembecome rather complicated, whereas according to the assump-tions it is possible to define occupation states based on busysubscribers only, expressed by the Poisson distribution:

(29)

Hindrance of calls (calls held in waiting) only occurs for k > x,and the probability that a call does not get access, given thatthere are x + 1 calls in the system is given by Rx = S1.

(See remarks below for the general case).

For the more general case of r groups (27) and (28) take theforms of

(27a)

and

(28a)

Under Engset’s general assumption that no calls are lost, onlyhindered, and here also the source number is infinite, the stateprobabilities for a total r-group system can be expressed exactlyby

(29a)

Only the states of k > x will imply hindered calls, and by similarreasoning as in the case of r = 2 we get the values of hinderedtraffic for each state from k = x + 1 to k = rx:

(30)

For states above rx all new calls are put on waiting, and we getcorrespondingly:

(31)

q(r−1)x+h = Prx+h S(r−1)x + h( )=

e−ry

ry( )rx+h

rx + h( )! ⋅r −1( )xx +1

+ h

, h =1,2,3,K ,

qi = Px+1 ⋅ Si =e

−ryry( )x+i

x + i( )! ⋅

x + ii −1

rxx

, i =1,2,3,K ,(r −1)x

Pk =e

−ryry( )k

k!

Si = R(r−1)x− jj=0

i−1

∑ =

x + ii −1

rxx

Rs = R(r−1)x−i =

x + ix

rxx

=

x + ii

rxx

, i = 0,1,2,K ,(r −1)x −1

Similarly for x + 2 we get q2 = Px+2 ⋅ S2 , q3 = Px+3 ⋅ S3 etc.

The contribution to the hindrance thus becomes q1 = Px+1 ⋅ S1.

Pk =e

−2y2y( )k

k!

Si = Rx− jj=0

i−1

∑ =

x + ii −1

2xx

The total hindrance can be expressed by

(32)

and the degree of hindrance by

(33)

In order to be able to do practical calculations some appro-ximations are introduced like before. Two examples are cal-culated with r = 4 groups, x = 10 selector outlets and a trafficper group of y = 5 and y = 5.5 respectively. The calculation indi-cates that y = 5.5 is permitted within the requirement of 1 ‰hindrance. The non-graded case permits only y = 4, and thatmeans a traffic increase of 37.5 %, or alternatively, a saving of≈ 27 % in the amount of selectors for the same grade of service.

Remark: The assumption of equations (28) and (28a) of the S-parameter given as a sum of R-parameters is a simplification,as only first order probabilities are included. This implies thatevents are mutually disjunct, which is not the case. For ex-ample, when r = 2, the first case, S1 = Rx is correct, whereas thesecond, S2 = Rx + Rx–1 should be replaced by S2 = Rx + Rx–1 –Rx · Rx–1. Engset seems to be aware that his solution is notexact. However, as elsewhere, he is satisfied that the error ison the safe side, and that the error is small for low values of thedegree of hindrance.

10 Progressive and combined gradings

As a first step Engset in Chapter X describes a straight gradingwhere each of r groups is served by a set of secondary devicesx0 that is offered a traffic y. The traffic carried on x0 is y0, andthe remaining total traffic of r · y1 = r · (y – y0) is offered to acommon group of x1 devices. The calculation is straightforwardaccording to his previous method for non-graded systems. Twoexample calculations are carried out for r = 2, x0 = 8, x1 = 2 andy = 6 and y = 5.5 respectively. The case of y = 5.5 leads toslightly less than 1 ‰ hindrance. The obtained advantage of thegrading is thus very close to that found for the slip multiplegrading, 37.5 % traffic increase or ≈ 27 % equipment saving.

Engset also considers the possibility of having several stages inthe grading, as a more general progressive grading, withoutcarrying out any calculations. The method is, however, easilyapplied to such cases.

At the time there was no understanding of the effect of the over-flow character of the traffic, and Engset does not take that intoaccount. He only considers the first moment of the traffic distri-bution, which of course leads to too optimistic results.

Finally, in Chapter XI a combination of a slip multiple gradingas a first stage, followed by progression to another slip multiplestage, is studied. The calculation method does not differ fromthat used for each of the two separate methods. According to thecalculation a saving in the amount of selectors of ≈ 42 % isobtained, while maintaining the grade of service. As an alterna-tive the equipment could be kept non-reduced in order to carry

ψ rxy

=Qrxy

ry

Qrxy

= qii=1

∞∑


more traffic. It is concluded that, while maintaining the grade ofservice, the grading would permit a traffic increase of ≈ 74 %.

A discussion with some coarse estimates is carried out with theaim of adapting the slip multiple grading to varying trafficrequirements. Among other the multiple slips may be modifiedto double steps and to overcomplete cycles.

11 Concluding remarks

The above discussion attempts to give a brief survey of Engset’squite voluminous report, without missing any of his majorpoints. The report goes a long way in showing details of themathematical development as well as the approximations feltnecessary to obtain practical usable results. Also detailednumerical calculations are carried out. Much of this is of courseof less interest today, given the tools available for numericalcalculations.

The report demonstrates very clearly that the primary objectivewas to find ways of dimensioning the new automatic telephonesystems that were being studied for introduction in theNorwegian telephone network. Engset had a leading position inthese efforts, as he was member of a committee established in1910 to give advice on the matter for the General Director ofTelegrafverket (Norwegian Telecom Administration).

This committee was doing extensive studies by travelling inEurope and America as well, and came out with a report in1913. The implementation of the committee recommendationswas delayed because of the war, and the system offers wereready for evaluation in 1916. For the purpose of this evaluationthe committee was supplemented by Johannsen and Christensenfrom Copenhagen and Hultman from Stockholm, all verynotable people in the business. This is very interesting in viewof the Engset report being completed in 1915.

One may also note that from 1917 on there were regular meet-ings between the Nordic telecom administrations. Engsetparticipated in ten such meetings from 1917 to 1935, whereasthe teletraffic community in Denmark was primarily within theprivate company KTAS, and thus not participants in these meet-ings. (Also in Norway there were many private telephone com-panies at that time.) It is this author’s impression that there wasa lively exchange of information between the Nordic telecom-munication communities during the transition from manual toautomatic systems. Traffic matters were of great operational andeconomic importance.

Engset’s 1915 model fell in a way between Erlang’s two modelsof 1917, that of loss (congestion) system and that of waitingsystem. Erlang applied the birth-and-death process in an explicitway, with emphasis on the distribution types. Engset, in a moreimplicit way, showed that the arrival independence lead toBernoulli and Poisson distributions, and argued that his modelset no conditions on holding time distributions. However, with alimited number of traffic sources a bias would exist for uneventraffic demand per source.

References

3 Engset, T. The Probability Calculation to Determine theNumber of Switches in Automatic Telephone Exchanges.Translation by E Jensen. Telektronikk, 88, (1), 90–93, 1992.

4 Jensen, E. On the Engset Loss Formula. Telektronikk, 88,(1), 93–95, 1992.

5 Engset, T. Om beregningen av vælgere i et automatisk tele-fonsystem. Kristiania, 1915. Unpublished report.

6 Iversen, V B. Thore O. Engset : a pioneer in teletraffic en-gineering. NTS13, Trondheim, 1996.

7 Engset, T. On the Calculation of Switches in an AutomaticTelephone System. Translation of [5] by A Myskja. Telek-tronikk, 94, (2), 1998 (this issue).

8 Erlang, A K. Solution of some problems in the theory ofprobabilities of significance in automatic telephoneexchanges. In: Brockmeyer, Halstrøm, Jensen. The Life andWorks of A.K. Erlang. Copenhagen, 1948. First published inDanish 1917.

9 Bortkewitsch. Das Gesetz der kleinen Zahlen. Leipzig. B.G.Teubner, 1898.

10 Palm, C. Some investigations into waiting times in tele-phone plants. Tekniska Meddelanden från Kungl. Telegraf-styrelsen, 7–9, 1937. (In Swedish.)

11 Palm, C. Research on telephone traffic carried by full avail-ability groups. TELE, 1, 1957. Edited by E. Malmgren.

12 Cohen, J W. The generalised Engset formula. PhilipsTelecommunication Review, 18, 158–170, 1957.

13 Dartois, J P. Lost call cleared systems with unbalanced tra-ffic sources. Proc. of the 6th International Teletraffic Con-gress, Munich, 215/1-7, 1970.

Appendix

Queue with departures.Palm’s and Engset’s models

1 Introduction

Conny Palm published in 1937 a study of a queueing systemwhere customers waiting in a queue may leave the queue with-out being served.

The study was first published in Swedish [10] and later inFrench. A revised version appeared in a collection of articles byPalm published in 1957 [11].

It turns out that Engset’s main model in his report from 1915 [5]under certain conditions is identical to this model by Palm.Palm’s model assumes the following:

• The number n of servers is limited

• The number N of customers is unlimited


(9)

As there are y calls per time unit, the waiting traffic can beexpressed by

(10)

3 The special case of b = s.Comparison with Engset’s result

In the special case of b = s, which is in fact Engset’sassumption, we get the congestion

(11)

Engset’s definition of ‘hindrance’ or ‘hindered traffic’ isidentical to that of equation (10). The degree of hindrance, ψ, isgiven by the ratio between the hindered traffic and the offeredtraffic. In the general case we get from (6) and (10)

(12)

In the given case with b = s we have from (10) and (11)

(13)

In this case α · β = A, β = n and α = A/n. From equation (2) we get

(14)

With substitution in (13) we obtain

(15)

A simple manipulation leads to

(16)

This is identical to equation (14) in Engset’s report [5], exceptthat Engset truncated the infinite series at N = the limitednumber of traffic sources. (In Engset’s notation N = n0.) Thelast term in the parenthesis thus becomes

By omitting this truncation it is demonstrated that Engset’smodel leads to a result identical to the Palm model with an in-finite number of sources, a number n of servers and an un-

N − n( )AN−n−1

n + 2( ) n + 3( )K N

ψ = An

n +1( )! ⋅e−A1+ 2A

n + 2+ 3A

2

n + 2( ) n + 3( ) +K

=v − n( )A

v−1

v!v=n+1

∞∑ e

−A

ψ = An−1

n −1( )! ⋅e−A1− W 1− A

n

W = n!

An

Av

v!v=n

∞∑

ψ = 1α

1W

+ α −1

⋅ A

v

v!v=n

∞∑ e

−A

ψ =AwA

=W ⋅ E1,n

1+ W −1( )E1,n⋅ n ⋅b

A ⋅ s1W

+ α −1

Cn A,n( ) = Av

v!v=n

∞∑ e

−A

Aw = τw ⋅ y = Cn ⋅β 1W

+ α −1

τw = Cn ⋅ βy

1W

+ α −1

limited queue, where customers leave the queue according to anexponential ‘patience’ function with a waiting time parameterequal to that of service time.

The relative value ψ is the ratio between the traffic load in thequeue and the offered traffic A.

Engset developed the expression (16) by weighting the Poisson

since these are the numbers of subscribers in queue for thecorresponding terms.

In order to justify using the Poisson distribution instead of theBernoulli (binomial) distribution for a limited number ofsources, he shows by taking the derivative of the logarithm ofthe binomial terms with respect to the number of terms, thatincreasing the number of terms towards Poisson leads to an

increasing queue traffic when For practical

dimensioning that will normally be the case. By applyingPoisson instead of Bernoulli when there is a limited number ofsources, the calculation will thus be on the safe side, ie. waitingin queue will be less than that calculated.

Under the given conditions one will have

(17)

The common factor (ν – n) represents the weighting by thenumber in queue, while the three expressions without weightingrepresent, respectively, the sum of the Bernoulli terms fromn + 1 to N, the truncated sum of the Poisson terms from n + 1to N, and the sum of the Poisson terms from n + 1 to ∞.

Engset has calculated values with four decimals according tothe middle expression of (17). The choice of N-value is notstated and most value sets are such that the value of the ex-pression will be < 0.01. In those cases the differences, if any,between corresponding values only turn up in the last (fourth)decimal.

ν − n( )v=n+1

N

∑ Nν

An

ν1− A

n

N−ν

<ν − n( )A

ν−1

ν!ν=n+1

N

∑ e−A

<ν − n( )A

ν−1

ν!ν=n+1

∞∑ e

−A

A < n +1− n +1.

terms Ax

x!e

−afrom x = n +1 upwards by 1, 2, 3 etc.,


Arne Myskja is Professor Emeritus at the NorwegianUniversity of Science and Technology, where he hasbeen employed since 1960. He was formerly with theNorwegian Telecommunications Administration, andhas had teaching/research assignments in the US,Denmark, Sweden, Australia and France. His mainareas have been switching systems and teletraffictheory. Most recently he has devoted his time tostudies related to T. Engset.



Tore Olaus Engset is known in the teletraffic communityfor the formula that carries his name. To most others he israther anonymous, or he is known as the Director Generalof the Norwegian Telecommunications Administration from1930 to 1935, after having served the same institution hiswhole adult life from 1883 on.

New light is cast on Engset by a newly discovered extensivereport on teletraffic modelling, a report that was never pub-lished, but which contains original aspects, some of whichwere first treated much later by other authors.

Even less has been known of the impressive scientific workby Engset on atomic theory, as demonstrated by a report inGerman finished in the last year of his life, 1943. This againled to the discovery of a 50 page series of articles on thesame subject in the reputed journal Annalen der Physik in1926 – 1927, which – among other things – discusses thethen recently published Schrödinger wave equation.

Who was this man? Very little has been publicly known, andvery few people now can remember him. The time is dueto supplement this first publishing of some of his unknownwritten work by a broader personal presentation. Maybe hisvery modesty limits the possibilities of following his track,but there is something that is much more than nothing.

1 Introduction

Engset’s formula is well known and appears in most presentday textbooks on teletraffic theory. His only publication [1] ona traffic theory topic is still being referred to 80 years after itsappearance in the German journal Elektrotechnische Zeitschriftin August 1918.

The Danish scientist and engineer A.K. Erlang is rightly called‘the father of teletraffic theory’. His publications on the subjectand related matters span from 1909 until his death in 1929. Hisprincipal paper [2] containing the B- and C(D)-formulas camein 1917.

Engset’s formula has been – and may be – considered as anappendix to or an extension of Erlang’s B-formula. The differ-ence in principle is that while Erlang assumed an unlimitednumber of traffic sources the Engset model permits any arbi-trary number of sources. Thus, one may as well turn the matteraround and say that Erlang’s formula is a special case ofEngset’s formula, obtained by letting the number of sources in-crease to infinity. However, the sequence of the publicationsshould not be neglected.

In spite of the naming of the formula, Engset as a teletraffic pio-neer has not been a focus of interest within the scientific com-munity, which may be seen as natural in view of his very limitedknown production. The reason for a renewed interest is the dis-covery of a voluminous report by Engset of 130 typewrittenpages from 1915 – two years before Erlang’s main paper – onteletraffic matters [3]. The report is an original work with manyaspects, some of which were taken up by others only manyyears later.

When Engset’s 1915 report now appears in an English translation[4], along with a summary article [5] on the report, it is an oppor-tune occasion to supplement all this by some biographical notes.

T.O. Engset is a publicly known name on two counts: One isthat of the teletraffic expert as focused here, and the other is thatof the man serving Telegrafverket (later: Televerket/Telenor,the public telecommunication operator in Norway) for 52 years,the final five years as its Director General.

There is no extensive biography of Engset written up till now.However, there are bits and pieces several places. In T. Rafto’sTelegrafverkets historie 1855–1955 [6] there are many refe-rences to Engset.

The best separate presentation is offered by L.A. Joys in Telev-erket’s internal periodical Verk og Virke no. 4, 1967 [7]. Thepresentation contains the main biographical data, some citationsfrom ref. [6], and it has also an anecdotal character. Joys, by theway, was probably the one person who devoted most effort instudying Engset’s formula and its significance. He wrote severalpapers on the formula, and at the age of 70 earned his doctorateon the subject. Joys’ recursion formula has been applied for cal-culation of Engset tables [8]. In the introduction to his Engsetpresentation Joys regrets the lack of sources. The 1915 reportwas not known to him, and would certainly have supplementedhis knowledge and elevated his already high esteem of Engset.

An interesting short biography is presented by Eliot Jensen in acommentary article [9] in 1992 to his English translation [10] ofEngset’s 1918 paper. E. Jensen states the following:

“Engset was very interested in mathematics. This, in additionto more than 20 years of experience in practical teletrafficmatters, forms the background for the theoretical work hepublished in 1918. The main material of the paper had beenestablished already by 1915, but Engset at that time did notconsider it of sufficient scientific substance to have it pub-lished. However, subsequent contact with prominent peopleof the Copenhagen Telephone Company, bringing him im-pulses from the traffic research there, appears to havechanged his mind”.

It is not clear how E. Jensen knew that “the main material hadbeen established already by 1915”. There is no other indicationthat he had any closer knowledge of the 1915 report. The im-pulses from the CTC people to publish his 1918 paper are verylikely, but he may as well have been discouraged from pub-lishing his somewhat different model in the 1915 report.

The editor of Telektronikk, Ola Espvik, requested me to collectinformation for a more comprehensive biographic presentation.For reasons of time and other limitations it has not been pos-sible to write a full biography. That would require extensivesearch in public and private archives, and would also require adifferent frame of publication. It is even uncertain if much morecan be found than what is presented here.

The intent is not to draw a picture of the professional man only,but rather to give a broader picture, including the human aspects,in a narrative of an in many ways remarkable man. Since it is toappear in a technical journal, it is only fair to advise those read-ers who are solely interested in the scientific aspects to leave atthis point and rather go to references [1], [4] and [5].

The man behind the formulaBiographical notes on Tore Olaus Engset1

A R N E M Y S K J A

1 In the baptism protocol the name is Thore Olaus Engeseth.Himself, he always wrote Tore Engset. Locally, the normalform today in writing is Engeset, while pronounced more likeEnset.


2 The background

Tore Olaus Engset was born 8 May 1865 at Stranda in thecounty of Sunnmøre in western Norway. Stranda lies in theheart of the fjord country that is famous for its breathtakingnature views, with nearby Geiranger, for decades the target oflarge tourist cruisers. T. Engset, in retrospect after retirement,reviews some of the great panoramas that he had seen in his life,notably the Grand Canyon and Niagara Falls, “But of all I haveseen of great and beautiful from the Creator’s hand nothing hasfor me been like the view from the top of Roaldshorn that Sun-day in my 19th year”.

Stranda in 1865 was a mainly agricultural community with verymodest – many would say poor – living conditions, but hassince then developed into a prosperous industrial town, manu-facturing furniture, foodstuffs, clothing, mechanics, electronicsetc., with foodstuffs having taken the lead over furniture. As anillustration the industrial turnover increased from 95 millionkroner in 1967 to 750 in 1987, a factor of eight in nominalvalue. Since then the figure has more than doubled. Tourism isimportant, and even national championships in alpine skiing hasbeen hosted by the town. The present unemployment rate is lessthan one percent.

Engset (or rather Engeset) is a small community with a littlegroup of small farms along a rather steep slope. Tore grew upon the farm named after his great-grandfather Kornelius (today,the ‘Karneles farm’). The owners of the farm are known in anunbroken line since 1540. The earliest of Tore’s forefathers onthe farm was born in 1655.

Tore was the second youngest among ten children, four ofwhom died at one year or less. There were also two older half-brothers and one half-sister. Tore’s mother died when he wasnine, and her last words to her then 22 year old daughter was:“Don’t forget Tore”. And she did not. She stepped into hermother’s place and delayed her own marriage for several years.Elen Regine was the only of Tore’s sisters to grow up. In spiteof her own ten children, she told in a letter that he always kept aspecial room in her heart. She admired her younger brother verymuch, but there were limits: When he told her that in the futureshe would be able to speak over long distances without tele-phone wires, she replied that he might tell her whatever hewanted, she did not believe a word.

The school records from 1875 show that 10-year old Tore nevermissed one day. He was later sent to junior high school in thetown of Aalesund, from where he graduated at the age of 18,receiving a prize from a local legacy in the form of two bookvolumes with golden letters on the back, with a hand-writtennote by the school principal, for “diligence, progress and goodconduct”.

3 Professional career

T. Engset was admitted to the telegraph school in Stavanger in1883, and that turned out to be decisive for his future career. Hegot his exam at the end of the year, and from January 1884 hewas employed as a telegraph assistant at Lødingen, Arendal andTrondheim, until in January 1890 he got a position as juniorsecretary at the Director General’s office in Kristiania (Oslo).

Since he had only his junior high school exam from 1883, hewent on with studies in his spare time. He graduated as a stu-dent in 1892 and went on with university studies in mathematicsand physics and got his M.Sc. degree 1894.

Gradually he advanced to secretary, office manager and headof the traffic section of Telegrafverket. As such he was also thedeputy for the DG.

From 1897 there were two departments, the administrative andthe technical, with the administrative department as the more‘political’ branch, which became Engset’s responsibility. In1920 a further subdivision was done, with five departments andwith Engset as head of the traffic department. Then in 1926there was a reversal back to two divisions, one traffic divisionand one technical division. In fact it was a grouping of depart-ments on a higher level, according to the growth of the institu-tion. Engset kept his position as second in command.

When Rasmussen retired from the position of Director Generalin 1905 Engset was one of five applicants for the vacant posi-tion. However, the new DG, Heftye, was chosen from outsideranks, and he kept the position until he died in an accident in

The ‘Karneles farm’ at Engset, Stranda, where Tore Olaus Engset grew up


1921. Engset acted in the position for parts of 1921 and 1922and applied a second time with no more success than before.The new man was Nickelsen, also an outsider, who at the end of1929 left the position because of political differences with theministry about the independent authority of his institution.

Engset was now 65 years old, and he admits in a private letterthat he was very much in doubt whether to apply for the topposition a third time. A couple of colleagues suggested that hemight simply neglect to apply this time, as he would probablybe asked anyway. Realising that such procedure would be toorisky he sent his application at the latest possible time. Therewere nine applicants. Engset was appointed on 22 August 1930and received a telegram of congratulations from the Minister.

Before this Engset had to go through a hard fight with the all-male union within the organisation. In his interim position asacting DG he had appointed a female head of staff. The unionleader was also editor of the newsletter of his organisation, andfrom this position he fought for the male candidate, claimingthat the female candidate was given her previous position in thestaff office in order to take care of women’s interests. He evenpaid a visit to the Minister trying to influence him on the matter,and to get a speed-up of the process of appointing the DG.Engset claimed that his candidate had responsibility for thewhole staff, and was clearly best qualified. He put the matterforward to the Minister, who admitted that he generally pre-ferred men for such high positions, but he did not want toreverse the appointment. Even formally, Engset claimed that hewas on safe ground. A 1924 decision by the National Assemblyhad given equal rights of higher positions, given proper educa-tional background.

Engset feared that the episode might weaken his possibility toobtain the top position, so he felt a great relief when he receivedthe telegram from the Minister. The particular background ofhaving been second in command for so many years, and now atthe age of 65 was acting DG for a second time, made him par-ticularly sensitive to the risk of being bypassed once more. Hewas certainly aware that there had never been an insider in theposition. He turned out to become the first. He had asked theMinister to avoid a long interregnum like that in 1921–22,“when they let me hang in the position a full year, and then cutme down to let me lie in the gutter (that was exactly what Isaid)”2. Those were strong words by a man of Engset’scharacter.

Engset’s lifelong career in telecommunication administrationcoincides with the first fifty years of telephone in Norway.From the beginning there was competition between many localprivate companies that were established for the promotion of thenew medium and its basic technology on one side, and the pub-licly owned and operated telegraph company on the other. Theprivate companies had the edge in most cities because of localbusiness initiatives and a more flexible operation. There were,however, problems of parallel network construction and lackinginterconnection. The companies were solely locally based,whereas the public telegraph company had a national coverage.

The matter became one of national political interest. The ruralrepresentation in the National Assembly was strong, and theidea of equal opportunities for all citizens, irrespective of loca-

tion and operational cost, became decisive. After much delay anew law was passed by the National Assembly in 1899, givingexclusive rights of telephone establishment and operation to thestate, and the public telegraph company was authorised to carryout the task.

The monopoly law of 1899 remained virtually unchanged formore than eighty years, and almost as long as it took to carryout the process of transferring all private telephones to thestate(!). Much less time was needed to reintroduce full competi-tion 99 years after the initial law enactment. The long delay inenforcing the law was due to the lack of public funding and tolocal resistance. In concentrated local communities it was pos-sible to establish and operate a telephone service at cheaperrates than could be done nationwide. Financing of expansionaccording to demand by loans and stock emission was straight-forward in private companies, whereas the authorities did notpermit that for the public state owned system.

Increasingly, though, also local telephone came under stateownership, but not without fierce fights in many places. A veryinformative recount is found in [6]. According to this sourceeven notable people like Fridtjof Nansen and Erik Werenskioldwere involved in the ‘telephone war’. In the ‘violent dispute’there were demands that “the director general as well as hisexpert advisor, head of office Engset, must be removed”. (ErikWerenskiold, by the way, later made the official portrait paint-ing of Director General T. Engset, 53 years after he made thecorresponding portrait of Tank-Nilsen, the first DG of Tele-grafverket).

No doubt Engset was in the front line of the telephone develop-ment. He had to stand against the public dissatisfaction withsupply being far behind demand, and to fight for better fundingfrom the government. In fact, even in the National Assemblylengthy discussions were carried out on what would today beconsidered petty details. At the same time his primary task wasthe technical planning and accomplishment.

A particularly challenging task was that of introducing a newgeneration of telephone exchanges. In spite of the first auto-matic exchange having been put in operation in 1892 in USA,manual operation was predominant for another 20–30 years. In1910 a three-man expert committee was established to evaluatethe development of telephone switching and give recommenda-tions for the choice of a future system. An immediate cause wasthe need for expansion of telephones in the capital city.

The committee took its task very seriously. They visited ninecities in Sweden, Denmark, Germany, Austria and Holland, andeight cities, from New York to Los Angeles and San Francisco,in the United States. They spent 48 days travelling in Europeand 71 days on the US trip. The final committee recommenda-tion [11] was put forward in March/April 1913, and contained67 printed pages and 13 appendices. The committee proposedconversion to a fully automatic system with primary andsecondary exchanges. The imminent world war led to changesand delays, and the first automatic exchange was not put inoperation until 1921.3 As this committee work represents a very

2 In a private letter September 1930.

3 The first public automatic telephone exchange in Norway wasput in operation in 1920 by the private telephone company inthe town of Skien.


important part of Engset’s career, we shall quote from ref. [6] intranslation:

“It was a great, all new and difficult task the study committeeof 1910 had embarked upon. It should make a judgement onthe three main solutions for telephone exchanges: the man-ual, the semi-automatic, and the fully automatic. It shouldfurther estimate the actual and future demand and inviteoffers from the telephone manufacturers. Chief engineerAbild, office head Engset and telephone director Iversen eachhad insight, experience and detailed knowledge in telephonematters. They now investigated thoroughly everything con-cerning modern telephone operation and carried out excur-sions in European countries and through America all the wayto the Pacific coast. The result was a thorough report thatrecommended transition to a fully automatic system withprimary and secondary exchanges.

The fully automatic solution would among other things solvethe difficult problem of large exchanges, as it permitted asubdivision in several satellite exchanges and establishmentof the single exchanges according to subscriber density. Indesign the automatic system represented a decentralisation,in operation a centralisation. The new City Centre exchangein Kristiania (Oslo) was intended to assemble the large busi-ness subscribers, that were the greatest users of telephoneand had the most mutual correspondence. For connection tothe main exchange were planned four sub-exchanges in thecity for private and small business telephones. All exchangesshould have connection to four suburban exchanges belong-ing to the Kristiania telephone area. The system was aimed at30,000 subscribers, but could be adapted to as many as90,000 numbers. The construction period was set to fouryears, and the cost estimated at 8.5 million kroner.

On the basis of the report an initial grant of 980,000 kronerwas given by the National Assembly for the start of recon-struction works. A disappointment was that ‘for budgetaryreasons’ the 4-year plan could not be kept and constructiontime had to be extended to 6 years. A factor that nobody couldhave foreseen came up: the war. As the war erupted, all theongoing works had to cease for a while. The delays that fol-lowed could not be regained. On the contrary, new interrup-tions arose incessantly. The bids on system installations couldnot be brought to a decision until 1916. For examination ofthe bids the committee was enforced by three experts fromabroad in order to guarantee best knowledge and fair choice:Telephone director Fr. Johannsen had headed the reconstruc-tion and reorganisation of the Copenhagen telephone, so thatit was made ‘a gold mine for experts wanting to study moderntelephone operation’. Telephone director Axel Hultman fromStockholm was an outstanding telephone authority with inter-national reputation, and he had himself designed a telephoneselector; while the Danish telephone engineer P.V. Chris-tensen, as a specialist on the mathematical side of the matter,assisted with calculation of the traffic equipment.”

Engset had already at this stage had some contact with theCopenhagen people, and that may even have instigated the moreactive co-operation around the evaluation of system proposals.What is clear, however, is that he must have been working onhis traffic modelling for some substantial time before the 1916co-operation. His comprehensive 130 page report, dated 1915,must have taken considerable time, at least months, moreprobably years. It was developed from scratch as a left-hand

activity beside his work in a demanding administrative positionas second in command in a big national public service organisa-tion. This point will be considered closer in the next chapter. Itis mentioned here in order to highlight the very close relationbetween Engset’s traffic studies and the great task of intro-ducing a radically new technical solution for large-scale tele-phone switching in the national network. Practical solutionson a safe theoretical basis is the governing idea of the report.

World War I was a time of great expansion in telegraph andtelephone traffic, but also of scarcity of material and equipment.For the first two years the available funding of expansion was infact reduced, and serious crises arose in traffic performance, asdemands far exceeded capacity. It led to outcries, and an emer-gency plan was issued to improve the situation. It did help, butinflation sapped some of its intended improvement. However,the post-war period was a constructive one with substantialexpansion, until the 1921 crisis hit hard. Engset was in charge,until the new DG Nickelsen took over in 1922. The followingyears were marked by struggle for more independence fromdirect government control with more freedom of financing andoperation. The limitations were felt as a straitjacket that pre-vented expansion according to market needs. When Nickelsenleft in protest and Engset was finally appointed, the new world-wide economic crisis was already in progress.

In a review of Engset’s career at the time he was appointed,T. Rafto in [6] writes: “From 1894 he worked as secretary atthe ‘Main office’ in Kristiania. From now on and all throughthe years he was head of traffic and operational matters. Thisman of administration ‘par excellence’ was absorbed in hiswork with ardent interest. Prudent, knowledgeable and withimpressive energy he became the best support of the director.He expressed openly his gratitude of having been given a ‘placein the sun’, and he enjoyed his work.”

The newly appointed Engset, with his experience from the crisisin 1921–22 had to handle a similar situation from 1930 on-wards. There were restrictions instead of expansion in mostareas. “As Engset in 1930 got the full responsibility, a similarsituation emerged as that of ten years earlier when he acted asdirector during the vacancy. While then the post-war depressionreigned, now a new world crisis was under development. Theboom in world markets in the latter half of the twenties gave theillusion of continuous progress of prosperity. But during sum-mer and autumn 1929 the radiant hope of better times wascrushed. ... It was helpful that Engset was accustomed to deal-ing with tight budgets and wage reductions.”

National coverage by long distance telephone had been obtainedby 1920. While landline telegraph gradually stagnated, long dis-tance telephone increased continually. Radio telegraph and –from the twenties – radio telephone were also growing. Engset,with his radio expert colleague Hermod Petersen, worked onexpansion of the radio network and on creating a nationalbroadcast system. Also picture telegraph came on the agenda,and the first unit was ordered in 1930.

During Engset’s five year period on top, the fight for more in-dependence continued. A committee was established to take anew look at the question. A favourable recommendation got partsupport and part opposition from different ministries. It wassupported by the parliament committee, but the proposal fellin the plenary session.


The economic crisis gradually eased during the five year periodand the matter of integrating the whole national telephone sys-tem under one administration emerged again. Engset’s politicalviews in general are not known, but already in 1930 he ex-pressed the opinion that “if the country got a labour govern-ment, it would probably bring in order the telephones in Nor-way like in many other countries.” He was confident that hisinstitution would be capable of taking over, but it must be donein a lenient way, should the transition be for the better. Whenthe labour government took office in 1935, an initiative in thedirection was taken. Limited funding, however, kept the processat a slow pace, and the take-over was not finalised until 1972(!).

Engset’s leading position in the organisation of the nationalcommunication system, with an ever increasing interconnectionwith a global system, naturally brought him in touch with inter-national work.

Already in 1901–1902 he stayed with his wife for an extendedperiod in Copenhagen. The main reason was poor health. Hisdoctor ordered him to quit work for a while to recover. Al-though it is not known, it is fair to assume that he got in touchwith colleagues in Copenhagen already at this stage. The com-mittee work from 1910 onwards gave him a broad view of thewhole western world, and later in close co-operation with Dan-ish, and to a lesser extent, Swedish colleagues.

From 1917 on a more regular form of contact was establishedbetween the telecom administrations in the Nordic countries. Asurvey of 80 years of Nordic telecom co-operation is given in[12]. Altogether 96 conferences are listed. Out of 14 such con-ferences in the period 1917 – 1934 Engset took part in ten. Hewas also a delegate to the very important international telecom-munication congress held in Madrid in 1932. This was the con-gress where the International Telecommunication Union (UIT)was founded on the basis of the three separate committeesCCIF, CCIT and CCIR.

A Norwegian Encyclopaedia from around 1950 (Norsk All-kunnebok) [13] has a mention of Engset, where his main careerdata are given. The further information is as follows: “He hadlittle schooling, but made a lot of studying on his own, collect-ing great knowledge, even outside his own profession. He wasNorwegian delegate to several international conferences on tele-graph and telephone matters and gained much honour at homeas well as abroad.”

Tore Engset retired at the age of 70 in 1935. An outstandingman of duty and independence had finished his 52 year career.At that point the international and national economic conditionswere on the rise. Optimism was returning, until – alas – WorldWar II erupted. But that is another story.

4 Science for service

There is this dual picture of Tore Engset: The lifelong profes-sional career man – modest and ambitious at the same time – atthe front of a team creating a many-faceted national telecommu-nications system; and the introvert amateur academic who qui-etly developed his mathematical traffic model. Which one of thetwo was the most influential? In Rafto’s words [6]: “Beside hisprimary function, all through the years he was occupied withscientific investigations”.

In his own country and among non-specialists of traffic theoryhe is the professional man who only at the end of his careerreached the top, and therefore was less visible as a man behindthe scenes. In the international teletraffic community he is theman behind the Engset formula. However, his basis is only thetwo-page article from 1918, appearing in the wake of Erlang’smore famous work of 1917.

The newly discovered report from 1915 changes the picture rad-ically. Engset created his own traffic model from basics as anoriginal work. His only assumption is that of mutually in-dependent sources that leads to Bernoulli/Poisson distribution.Otherwise he does not make any assumptions of distributions.From this basic model he arrives at results that are only pub-lished much later: Departure from queues (Palm 1937), the gen-eralised Engset formula (Cohen 1957), most unfavourable casewhen identical sources (Dartois 1970), total loss with day-varia-tions of traffic, calculation of gradings, etc.

Engset was not satisfied with only theoretical solutions. For himthe application of scientific methods was not a target in itself,but rather a tool to obtain improved telephone service. His veryclear intention was to create practical dimensioning tools thatcould be applied within the new automatic telephone systemsthat were being planned by himself and his colleagues. With thelimited means that were available for numerical calculations hewas very conscious of two conditions: Approximations had tobe introduced, and these approximations had to be on the safeside. The second condition implied that the real service shouldbe better than that calculated. Even with the simplified calcula-tions available, he found it important to establish dimensioningdiagrams that allowed direct reading of value sets, and he con-cluded his report with calculated tables and diagrams.

As the 1915 report was never published, and the Erlang modelwas accepted as a better proposal, also by Engset himself, thereport was put aside and kept in the files without being utilised.The Engset formula of 1918, however, applied the Erlang trunc-ation (1917) and was accepted as the solution for loss systemswith limited number of sources. The Erlang and Engset for-mulas are both formulas for congestion probabilities due toservice system limitation, the former given by a truncation ofthe Poisson distribution and the latter by a truncation of theBernoulli distribution.

As the scientific teletraffic work of Engset is discussed in anaccompanying article [5], we shall simply refer to those articlesand omit any further treatment at this point.

This, in the present context of teletraffic, ends the story ofEngset’s scientific work. However, in the final section of thisarticle a quote from Joys’ biographical notes indicates thatEngset after retirement worked on issues of atomic theory, andthat he was very interested in the utilisation of nuclear power.

5 The physicist

The information in Joys’ notes led me to a further search thatbrought me in contact with Engset’s step-grandson in Oslo,Willi Urne Fonahn. It turned out that he kept some papers ofEngset’s in a safe deposit box. By a recent inspection of thecontents such a dissertation was in fact found. It is dated springand summer 1943, a few months before Engset’s death in Octo-


Page from Annalen der Physik, 1927 [16]


ber the same year. The dissertation is written in German and isof 147 typewritten pages [14]. It is a heavily mathematicalwork, with the mathematics expressions mainly written in byhand. It takes special knowledge and probably a lot of time toevaluate the quality of the work. This is not done so far, as sev-eral physics professors, after a brief look, have indicated that itwould take more time than they have available to come upwith a serious evaluation.

However, professor Iver Brevik of NTNU took time off over aweekend to look into the matter, and he has permitted me toquote his comments after that limited survey:

“This is a very extensive work, almost 150 typewritten pagesin all. The author focuses on the fundamental building blocksof the hydrogen atom, namely the proton and the electron,and tries to describe the physical properties of these ‘Kor-puskeln’ and their interaction within the framework of theclassical theory, and partly also the elder version of quantummechanical theory.

The work may roughly be divided into three different parts.The first part describes classical models of the proton and theelectron, as particles that have finite extensions and are keptin mechanical equilibrium by means of mechanical stresses.This physical picture of the elementary particles appears out-dated today, although to a certain extent it has become re-vived again in recent years in connection with the so-calledCasimir effect (ie. the quantum mechanical theory of zeropoint oscillations). The second part of the work considers theelectron in its orbit around the proton in the hydrogen atom(elliptic orbit), according to the first version of the quantumtheory as constructed by Bohr, Sommerfeld and othersaround 1920. The last part treats the theory of electricallycharged particles moving at high velocities, and one mayconsider this as a study of the scientific discipline which isnow called relativistic electrodynamics. The special theoryof relativity here plays an important role.

On the whole, the present work seems to be very carefullydone. It is clear that Engset has made a thorough study of thebackground material belonging to the classical theory, andpartly also to the quantum theory in its elder version. It isquite striking how Engset goes a long way to test his theorywith experiment, by carrying out detailed numerical calcula-tions of the theoretical predictions.

One may ask oneself: is this large work of scientific impor-tance today, or is it mainly of historical interest? I wouldthink that the second answer is the more correct. At the endof the 1930s the new version of quantum mechanics wasdeveloped, implying that the motion of the electron orbit inthe hydrogen atom ought properly to be described in terms ofwave theory, constructed in accordance with the Schrödingerequation. On the basis of this, it lies at hand to conclude thatthe physical picture of this work was essentially outdatedalready at the time it was written. However, the situation isactually more complicated: In fact, Engset published someyears earlier several short papers in the well known journalAnnalen der Physik (Leipzig), showing that he was wellaware of Schrödinger’s work. The first of these papers waseven published almost immediately after Schrödinger’spaper, in the same year (1926). Engset developed a waveequation which, however, is stated by him to containsome differences from Schrödinger’s equation in the concretesituation discussed.

There are some strong statements in Engset’s treatise whichcan be criticised on the basis of present day’s knowledge ofphysics. However, leaving such points aside, one cannot helpbeing impressed by this major work made by a man in hisretiring age. After all, this field of research lay outsideEngset’s main activity in his younger years. One may ofcourse wonder about what could have been Engset’s motiva-tion for undertaking such a ‘tour de force’. Was it to continuehis earlier studies, aiming at further publications in Annalender Physik? Or was the motivation to submit the work as adoctoral thesis? It seems very natural to suggest that the lat-ter option is correct, although apparently we do not havewritten sources helping us to draw a more definite conclusionat this point.”

Brevik, by his browsing through the report, discovered refe-rences to the publications by Engset in Annalen der Physik in1926 and 1927. A further search led to two publications [16, 17]in seven parts, altogether about 50 journal pages. Those articlesput the 1943 report in a new light, as they indicate a seriousinterest and capability in fundamental physics already in thetwenties, while Engset was still in his most active period ofmanaging and expanding the national telecommunications net-work. It is interesting to note that the starting point of the firstpart of [15], dated June 1926, is a reference to Schrödinger’spublication of his famous wave equation in the same journalearlier the same year(!).

Knowing Engset’s seriousness in all matters, it is hard to be-lieve that the dissertation is not also a serious piece of work.He is not known to have done any physical experiments, so hissources of knowledge in the field must have been confined toliterature sources, but, as Brevik points out, he did detailednumerical calculations to check his results. The report, howeverimpressive, will most probably be of historic interest today. Thevalue of the 1926/27 publications in Annalen der Physik is notfurther assessed so far.

There is a straightforward explanation to the eight year longempty interval from 1927 to 1935 of Engset’s active research inphysics. In 1927 his wife became seriously ill and died later inthe year after her husband had taken leave from his duties totend to her. It meant a serious break in the private life of a manin his sixties. Then there was a turbulent time with politicalimplications that led to the resignation of DG Nickelsen and thetake-over of the top position by Engset, until he resigned at 70in 1935.

Engset’s two reports, the 1915 teletraffic report and the 1943atomic physics report, are now safely deposited at Norsk Tele-museum, Oslo, where also other information, like contemporarynewspaper clips from Engset’s time, is available.

6 The private man

In his biographical notes [7] Joys recounts the following anec-dote:

“I never met Engset myself. True enough I had read aboutErlang and Engset as a young engineer while employed in theStandard corporation, first in Antwerp and then in Madrid.Since ‘our’ traffic theoretician at the time was M. Merker,who possessed the absolute truth in the best of all concerns,I remember that I pitied my slightly confused compatriot. The


fact that Engset was Director General was unknown to meuntil the author and ‘tramp’ Erling Winsnes turned up inMadrid in 1932. He said that it would ease his personal, andthereby his friends’, financial situation, if he were able tosend to Dagbladet an interview with Engset, who just thenattended an international telegraph and radio congress inMadrid.4

Neither Winsnes nor any of us other Norwegians knew whatEngset looked like, but Winsnes found him. Winsnes ex-plained that it happened in the way that he went to the con-gress building at the end of a session. People poured out,chatting and gesticulating. Finally, a lonely fellow with per-plexed looks came out. Winsnes went straight over to him,saying: ‘How do you do, Mr. Engset?’, getting the answer:‘Just fine, and how are you?’.”

As Joys’ recount of Engset is probably the most complete oneup till now, it is worth continuing the quote:

“Engset was undoubtedly introvert and absentminded attimes, as his close colleagues could confirm. Sleepless he wasalso. Even as office head in the central branch he might turnup in the long distance operator hall in the middle of thenight for inspection. He was not always welcome, as I under-stand.

Engset’s grandchild (in fact step-grandchild)5 recounts thathe was easily allured to take him to the movies, as he wouldfall asleep immediately.

He seems to have been something of a pedant and very mucha perfectionist. Among his colleagues Engset was said tohave acquired enough knowledge for a law degree at the Uni-versity, but the fear of not getting top grades held him back.The truth of this is of course doubtful, but it does say some-thing about Engset’s character.

When the editor of Telegrafbladet (the newsletter of the maleunion)6 asked for a picture for the newsletter on the occasionof Engset’s appointment to Director General, Engset repliedthat he would have a photograph taken the same day. A sug-gestion of using an already available picture was rejectedwith a remark that he wanted his staff newsletter to have apicture of the Director General, not of the office manager.

Engset was, however, a complex character. He was sober andobjective as administrator and scientist, friendly and obligingamong colleagues and at home. At the same time he was emo-tional and poetic. He gave elegant versified talks, once evenin German at a major event with German participants.

He was deeply religious, and even wrote a hymn for his ownfuneral.

It is said that Engset and some of his colleagues once werechatting at random. One of them brought up the question ofwhat they ought to bring out with them if a fire broke out inthe building. Engset immediately declared that the mostimportant things to save were his personal papers.

As I heard about those papers, I thought that something ofinterest might be found in the main files of the institution or

elsewhere. It turned out that kept in the family there werepapers containing an extensive dissertation about the theoryof nuclear fission, by which Engset had occupied himself foryears. He assumed already before World War II that nuclearpower would be the future form of energy to the advantage ofmankind. The subject engaged him very strongly. His dis-sertation is dated 1943, the year of his death, two years beforeHiroshima, which he certainly could not have dreamt of.

With the rapid development of nuclear theory after Engset’stime one must assume that his dissertation is obsolete. Maybeat his time it represented pioneer work, maybe not. Anyway,Engset had obviously found a new hobby to replace traffictheory.

It is said that Engset in his retirement had his daily walk to aplace where there was good sand for writing with his stick.Most people have their hobbies. Some have sports, a few havemathematics as their pastime. Hobbies are as a rule a form ofrelaxation from daily toil. Its goal is limited and most oftenrelated to private life. Whether the hobby even happens toleave a mark in society is ‘pure chance’ – to stay within theterminology of traffic theory.

Sometimes the foremost of amateurs reach the top of profes-sional level. This, however, presupposes a particular environ-ment. Engset was such an amateur. The group around Tele-phone Director Johannsen was the environment.”

So far L.A. Joys. It is fair to point out that the only teletrafficwork by Engset available to Joys was the two-page article from1918. That the 1915 report, which actually existed in the centralfiles, was not found, is of course a pity, considering Joys’ keeninterest in Engset’s teletraffic work.

The recount by Joys on Engset’s life as a private man can besupplemented as an extension to the description of his back-ground as given in the beginning of this article.

Tore Engset had a career of more than 13 years in Telegraf-verket behind him when he married in 1897 at the age of 32.His wife was Marie Amalie Nord from Kristiania who was one

4 See previous mention of the 1932 Madrid congress.5 Author’s remark.6 Author’s remark. Tore Olaus Engset with his wife Marie Amalie


a letter to his niece Anna Saxegaard Gjerding, with whomMarie had had regular correspondence for years, Tore paints avery touching picture of his loving wife for 30 years after shedied in 1927. “My office occupied me at all times, and in myscarce leisure time I jumped to studies and problems that stoleher husband away from her. And when I broke down fromoverexertion or for other reasons was thrown on the sickbed,she completely gave herself to nursing me. Such was her love,sincere, rich, great, with no thought of herself. ... We alwaysloved to be together. She was, as you know, so nice and cheer-ful and sociable. So youthful and brave”. She also, in her privatenotes, expresses her sincere love for her husband. He, on theother hand, took leave from his work to stay with her when herillness became more severe.

Tore Engset’s generosity is another quality which is clear fromprivate correspondence between two of his nieces, talking aboutthe Christmas presents received from their uncle in Kristiania in1900. “What a joy there was in the home that evening. Such ajoy was never seen in our crowd of children, my mother said.”There were ten children. Later, during the war with foodscarcity in the capital, when the retired widower Engset was liv-ing with his stepdaughter, the generosity was returned with farmproducts being sent from his relatives at Stranda, for which heexpressed great gratitude.

When Engset was finally appointed to the top position after hav-ing applied for it on three occasions, he writes to Anna, in replyto her congratulations, about all the received attention: “... andall this has pleased me much, not least the kind lines from you,

of a family of 12 children. She was then a widow with twochildren, a boy, Fritz Gude, and a girl, Julie Kathrine (Lilli),after her first husband Anton Julius Bolivar Kreutz, a ship-owner and captain who died on a trip to Canada. She was 41when she married Tore. They had no children, but Tore con-sidered his stepchildren as his own. The boy went to Americaquite early, and the family lost track of him, in spite of manyefforts to find him. Lilli was married to Adolf Fonahn, later pro-fessor at the University of Kristiania. They had one boy, WilliUrne, who always considered Tore Engset as his grandfather.

Willi Urne Fonahn (81) is probably the only person today withany closeness to Tore Engset, who still remembers him. He hadand still has great admiration for his grandfather, whom heremembers as very hardworking, honest and at the same time anaffectionate family man. He describes “an intelligent, endearingperson and a wonderful grandfather”. The boy Willi had onceavoided to pay his fare on the tramway, and he was kind ofproud of his achievement when he told his grandfather, whoimmediately gave him a coin and sent him back to pay andexcuse himself.

His working day was often extended far into the night, andsometimes he had two bowls, one with hot and one with coldwater to put his feet into in order to stay awake.

Engset was a shy person, in particular towards women in hisyouth. How he got the courage to propose to the nine year olderwidow seems a mystery, unless she helped him very much onthe way. Anyway, it turned out to be a very happy marriage. In

Tore Engset (seated at the head of the table) with heads of department in 1930

Photo: Norwegian Telecommunications Museum

dean Saxegaard’s daughter, who always maintained the corre-spondence with my beloved Marie! Oh, if only she and Elenhad lived to see that their dear Tore received an open and honestrecognition!” His sister Elen died only a short time before hisappointment.

Tore Engset’s poetic nature is demonstrated in his funeralhymn, but even more in an invited memorial article in theChristmas edition 1941 of the regional newspaper Sunnmør-sposten from his home region. In this article he writes about aglorious Sunday in August 1883. He had just received a tele-gram granting his admittance to the telegraph school in Sta-vanger.

“I was jubilant to send a telegram of acceptance, and on my wayhome I saw the mountain Roaldshornet flooded with sunlight.”He decided to climb to the top, as he had heard of the fascinat-ing view, and he was well used to mountain hiking. “I wassurprised and thrilled – yes, overwhelmed”. He goes on to de-scribe the view in poetic terms, naming more than fifteen wellknown mountain tops, then likewise the many islands lining thecoast.

“The great contrasts in the environment of nature and, in spiteof these, the wonderful harmony of the panorama under the glo-rious sun of the summer day had a fascinating effect on my sen-sitive soul: I lay down on my back on the highest point, lookingtowards the sky, and wept like a child”.

From this he goes on to describe other places in Norway, wherehe had travelled extensively during his work, continuing withfamous views of Germany, Switzerland and Italy, and the blue,blue Mediterranean, to end up with the Grand Canyon andNiagara falls. At Niagara Falls he claims to have spoken to theman who had jumped the Falls in a steel barrel, Bob Leach.Arms and legs had been crushed, but he was now all right.Worse was the fate of the woman who wanted to repeat thedeed. She turned mad, proving that women should not alwaystry to be as capable as men(!).

He reflects on the contrast between the extreme happiness thatSunday on the home mountain, when he had got admittance tothe national telecom organisation, and his time on the top of thatorganisation, “... it was rather stormy, candidly spoken a ratherdisgusting weather, with little sun ...”.

In an illustrated biographic dictionary [17] from 1916 ToreEngset, then department head, is presented with family relationsand personal vita. The editor had put questions to all personspresented in the dictionary: 1) Which persons, institutions,events have had the greatest influence on your development andlife? 2) Tell some little anecdote that characterises the environ-ment of your adolescence and yourself.

Engset’s replies were: 1) Dean Saxegaard of Stranda, withwhom I went to confirmation in 1880. 2) I have never felt moreproud and happy and more amply rewarded than when, as a lit-tle boy, I received my first pay of 2.00 kroner (20 pence) foreight days as a shepherd boy in the mountains of Sunnmøre.

Dean Saxegaard, by the way, happened to become Engset’sbrother-in-law 17 years after that confirmation.

The complexity of Tore Engset’s character also implied a gen-eral scepticism towards humankind. Maybe he felt that his ex-perience called for some carefulness. He is quoted to have said:Watch out for complaisant people!

The event of Engset’s retirement from his position as DirectorGeneral on his seventieth birthday was of course one of generalpublic interest. After a broad newspaper presentation of theachievements of his career he was asked for an interview. Herefused, saying: “It is not that I have done anything that shouldbe concealed for the world. However, it pleases me more to dis-appear quietly.” And the newspaper continues: “Quietly andwithout great gestures Engset has worked during his 52 yearsof service in Telegrafverket. He is, however, a man of insight,determination and the courage of his convictions, faithfullycarrying out his work within his administration, where he hasbeen of great influence in the inner circles. And thus even forthe whole country”.

One might reflect further on Tore Engset’s dual life: That of avery demanding position, normally requiring the full mental andphysical capacity of a very robust man, which he was definitelynot; and that of very capable scientific work, carried out mainlyduring evenings and nights, and in two very separate areas. Itbrings to mind the duality of Robert Louis Stevenson’s ‘Dr.Jekyll and Mr. Hyde’, though certainly without the sinister traitsof the latter.

His achievements are the more admirable, as his health wasgenerally not very good. He suffered from asthma and wasprone to catching colds. He had a full year sick leave in 1901–1902, when he stayed with his wife for several months in a pen-sion in Copenhagen. In private letters his wife expresses con-cern about his physical health as well as his nerves and hissleeplessness. Himself he refers to periods of overexertion.However, after his sick leave the situation improved, and hewas capable of carrying the full burden of his responsibilitiesfor many years to come.

Engset was made Commander of the Second Order of Dan-nebrog and Knight of the Legion of Honour.

After his retirement Engset moved to stay with the family of hisstep-daughter Lilli Fonahn, and it seems that they were on verygood terms, as indicated in private letters and also according torecounts by his step-grandson. This was the time when heworked on his dissertation on atomic physics.

We now write 1998, 115 years after Tore Engset began hiscareer as a boy of 18 and 55 years after his death. With theincreased knowledge that we now have of this outstanding manit is appropriate to give him the honour he deserves by issuingthe present publications. In addition to this, The NorwegianSociety of Chartered Engineers (NIF) has decided to establishthe Engset Award for young candidates in telecommunications.The award is supported by Telenor and some major telecomcompanies in Norway and will be presented at an annual con-ference arranged in co-operation by NIF and NTNU (Norwe-gian University of Science and Technology), the first timein January 1999.


Acknowledgement

In addition to the written sources of information, as given in thebibliography below, several people have contributed with bitsand pieces to the mosaic of Tore Engset’s life.

I was received with enthusiasm at Stranda by members of TheHistorical Society of Stranda (Stranda Sogelag), Harald Bergeand Signy Mork (granddaughter of Anna Saxegaard Gjerding,as mentioned above), together with the mayor of Stranda, AnneLise Lunde, and the editor of the local newspaper, Helge Søvik.They have furnished valuable information. Later informationhas been added by Lina and Ole Gjerde, who are writing amemorial article on Tore Engset for the 1998 Annual ofStranda Sogelag. They all admit to having practically no pre-vious knowledge of the scientific side of Engset’s career.

Willi Urne Fonahn and his daughter Wenche Fonahn wereextremely helpful in adding to the information, and by trans-ferring the atomic physics dissertation to Norsk Telemuseum,where Arve Nordsveen and Anne Solberg took care of the trans-fer, and supplied further information, in particular newspaperclippings.

Other sources of information that I have utilised are Statsarkivet(the National Archives) in Trondheim and Riksarkivet (the Na-tional Archives of Norway) in Oslo.

References

1 Engset, T. Die Wahrscheinlichkeitsrechnung zur Bestim-mung der Wählerzahl in automatischen Fernsprechämtern.Elektrotechnische Zeitschrift, Heft 31, 1918.

2 Erlang, A K. Solution of some problems in the theory ofprobabilities of significance in automatic telephoneexchanges. In: Brockmeyer, Halstrøm, Jensen. The Life andWorks of A.K. Erlang. Copenhagen, 1948. First published inDanish 1917.

3 Engset, T. Om beregningen av vælgere i et automatisk tele-fonsystem. Kristiania, 1915. Unpublished report.

4 Engset, T. On the calculation of switches in an automatictelephone system. Translation of [3] by A. Myskja. Telek-tronikk, 94, (2), 1998 (this issue).

5 Myskja, A. The Engset report of 1915 : summary and com-ments. Telektronikk, 94, (2), 1998 (this issue).

6 Rafto, T. Telegrafverkets Historie, 1855 – 1955.

7 Joys, L A. Til minne om T. Engset, Kristiania. Verk ogVirke, 4, 1967.

8 Rubas, J. Table of Engset Loss Formula. Melbourne, Post-master-General’s Department Headquarters, 1969.

9 Jensen, E. On the Engset Loss Formula. Telektronikk, 88,(1), 93–95, 1992.

10 Engset, T. The Probability Calculation to Determine theNumber of Switches in Automatic Telephone Exchanges.Translation of [1] by E. Jensen. Telektronikk, 88, (1), 90–93,1992.

11 Iversen, Abild, Engset. Indstilling fra Komiteen for telefon-centralstationsordningen i Kristiania. Kristiania, 1913.

12 Toft, Utvik. 80 års nordisk telesamarbejde 1917 – 1997.NTOB/NORDTEL. København, Fihl-Jensens Bogtrykkeri/Offset AS, 1997.

13 Norsk Allkunnebok. Fonna Forlag 1950 –.

14 Engset, T. Die Wasserstoffkorpuskeln, ihre Strahlung unddie fundamentalen physikalischen Grössen. Unpublishedreport, 1943. (German)

15 Engset, T. Die Bahnen und die Lichtstrahlung der Wasser-stoffelektronen (4 parts). Annalen der Physik, 80 (1926)823; 81 (1926) 572; 82 (1927) 143 and 184.

16 Engset, T. Die Bahnen und die Lichtstrahlung der Wasser-stoffelektronen. Ergänzende Betrachtungen über Bahn-formen und Strahlungsfrequenzen (3 parts). Annalen derPhysik, 82 (1927) 1017; 83 (1927) 903; 84 (1927) 880.

17 With, N. Illustrert Biografisk Leksikon. Kristiania, A/S With& Co.s Trykkeri, 1916.


Arne Myskja is Professor Emeritus at the NorwegianUniversity of Science and Technology, where he hasbeen employed since 1960. He was formerly with theNorwegian Telecommunications Administration, andhas had teaching/research assignments in the US,Denmark, Sweden, Australia and France. His mainareas have been switching systems and teletraffictheory. Most recently he has devoted his time tostudies related to T. Engset.


personal communications

Documents