code-division multiplexing for in-service out-of-band monitoring of live ftth-pons

Vol. 6, No. 7 / July 2007 / JOURNAL OF OPTICAL NETWORKING 819

Code-division multiplexing forin-service out-of-band monitoring

of live FTTH-PONs

Habib Fathallah* and Leslie A. Rusch*

Centre d’Optique Photonique et Laser, Laval University, Laval, Québec,G1K 7P4, Canada

*Corresponding authors: [email protected]; [email protected]

Received November 2, 2006; revised January 30, 2007; accepted May 1, 2007;published June 5, 2007 �Doc. ID 76713�

We propose, to the best of our knowledge, a novel in-service live fiber-to-the-home (FTTH) passive optical networks (PONs) management solution. Our so-lution uses a modified direct-sequence (DS) optical code-division multiplexing(OCDM) system and overcomes the optical time-domain reflectometry (OTDR)point-to-multipoint shortcomings. Our solution addresses various service pro-visioning and network maintenance challenges in PONs, alleviates their com-plexity, and reduces their operational cost. In addition, our system exploitspassive devices (or encoders) to demark service provider ownership and re-sponsibility from that of customers. Our OCDM-based management solutioneasily scales up from FTTH time-division multiplexing (TDM)-PON to WDM-PON and TDM/WDM-PON to support as many as a thousand customers, allusing only one monitoring wavelength. We address the coding system and de-velop capacity curves for different PON scenarios. © 2007 Optical Society ofAmerica

OCIS codes: 060.2330, 060.4510.

1. IntroductionFiber-to-the-home (FTTH) passive optical networks (PONs) seem to be the ultimatewinning solution for tomorrow’s last–first mile bottleneck. Important FTTH deploy-ments have been carried out in North America, Europe, and Japan over the lastdecade. Starting from 1:1 (one fiber to one customer) in the early 1990s, passive split-ters (PS) together with time-division multiplexing (TDM) technologies have enabledup to 1:128 for the GPON standard (ITU G. 984) with forward error correction (FEC).Recently in [1], the authors report a testbed with 1:256 PS, and future extralargeXL-PON systems are aimed at splitting factor of up to 1024 [2].

FTTH managers, however, are still missing an efficient monitoring technologyappropriate for TDM-PON, even for 1:32 BPON standard (ITU G.983) [3,4]. Standardout-of-band optical time-domain reflectometry (OTDR), based on Rayleigh back-scattering and power reflections and used to monitor point-to-point links, is ineffectivein point-to-multipoint TDM-PONs [3–6]. The OTDR trace at the central office (CO) isa linear sum of the backscattered and reflected power from all the network branches.It is difficult and even impossible for the CO manager to distinguish the events in onebranch from others.

Many FTTH management problems result from the very little information availableto the CO manager about the network. This directly affects the quality of service anddramatically increases the administration, the maintenance, and the provisioningcosts. For example, when two branches experience equidistant failure events they areindistinguishable. Even when a fiber fault results in an unambiguous event, thefaulty branch is not identified, requiring a truck-roll tour and outside intervention oftechnicians. Every branch must be checked separately from its end by means ofupstream power meter and/or OTDR transmission in order to identify the faultyone [4]. Moreover, when an optical network terminal (ONT) is not communicatingwith the CO, the manager cannot make a remote diagnosis and determine the cause;whether this is due to a fiber cut or simply because the ONT is disconnected or turnedOFF.

Few solutions for in-service TDM/PON management have been proposed [5–11]; allof them are impractical mainly because their capacity is limited to a few tens of cus-

1536-5379/07/070819-11/$15.00 © 2007 Optical Society of America


tomers. For instance, a unique signature (or identifier) may be assigned to each leg ofthe network, before the ONT. A discrete Bragg grating with a unique wavelength isplaced at each ONT. At the CO side, different interrogation techniques have been pro-posed, including a broadband source, multiwavelength laser, and tunable laser and fil-ters, etc. [5–11]. These systems are impractical for high numbers of subscribers due tothe very large spectrum to be sliced, i.e., one slice for every network leg. For 32 homecustomers (128 in GPON with FEC), and using narrow slice width of 0.8 nm, a totalbandwidth of 25.6 nm �102.4 nm� is required [8,11].

This lack of in-service live monitoring solution becomes more critical with the com-ing advanced and prospective PON architectures such as wavelength division multi-plexing (WDM) PON or hybrid TDM/WDM-PON (where a TDM-PON-like system isbuilt over every wavelength in a passive WDM network). These architectures promisemany hundreds of FTTH customers per fiber. This calls to an extremely high capacityPON monitoring technology.

In this paper, we focus on some of the in-service management problems and pro-pose, for the first time to the best of our knowledge, a simple optical code-divisionmultiplexing (OCDM) architecture that overcomes the problems of high-capacityTDM-PONs using only one monitoring wavelength. We illustrate how our OCDMmonitoring system could support hundreds to thousands of customers. A single wave-length, i.e., standard out-of-band 1650 nm (in compliance with ITU-T L.41) or the lessexpensive 1625 nm laser, could be used. Note that in our application we use theOCDM codes to carry management information and not the data itself. This makesthe proposed OCDM scheme easy to implement, as low transmission rates suffice. Wealso develop architectural solutions allowing our management system to monitorfuture WDM-PON and hybrid TDM/WDM-PON networks.

In Section 2, we provide a detailed description of numerous TDM-PON manage-ment challenges. These include problems related to OTDRs, FTTH installation, ser-vice provisioning, maintenance, and control/ownership, thus responsibility demarca-tion. We explain also how our OCDM system addresses all these issues and providespromising solutions. In Section 3, we introduce the OCDM system and coding engi-neering issues. In Section 4, we develop architectural solutions to expand our solutionto future, more complex passive optical networks including WDM-PON andTDM/WDM-PON. In Section 5, we address the coding system and develop capacitycurves for different PON scenarios.

2. TDM-PON Management ChallengesFTTH network managers face numerous challenges to obtain the necessary informa-tion to reduce installation, provisioning and maintenance complexity, and cost. More-over, service quality could not be guaranteed for the customers and service levelagreements (SLA) could not be contracted or respected without in-service and in-timefull knowledge of the state of the network. The ITU-T M.60 recommendations definethe telecom infrastructure requirements concerning the operation, administration,maintenance, and provisioning issues. In this section, we describe some of these prob-lems and show how our OCDM technique offers a promising solution.

2.A. OTDR DrawbacksPoint-to-multipoint problem. Standard OTDR based on Rayleigh backscattering andpower reflections, used to monitor point-to-point links, are ineffective in point-to-multipoint TDM-PONs [3–6]. Each branch termination connected to an ONT, as wellas every splice, connector, and fiber default or break, contributes to a reflection peakand/or power loss step a so-called event. The OTDR receiver observes the power that isthe sum of Rayleigh backscattered and reflected light powers from all individualbranches. As the number of braches in the network increases, the OTDR trace com-plexity increases proportionally. The network manager still has no efficient tools torecognize which event corresponds to which connector, splice, fusion, or branch end.The manager task is even more difficult when an observed peak event corresponds tosimultaneous or close-in-time events in the network. Our OCDM encoders will intro-duce codes inside the reflected powers from every branch end, allowing the COreceiver to decode reflections from each branch separately. Identifying the code in areflected train of pulses is the equivalent of identifying a specific branch.


OTDR dynamic range. PSs with a high number of ports (32, 128, and, recently, 512)induce very high loss (15, 21, and 27 dB), rapidly calling for much higher dynamicrange in future OTDRs. In [10], the authors recommended the use of a reflective ele-ment at every fiber end to increase the amount of power returning to the OTDR. Thisreduces the requirement of very high dynamic range since the OTDR receiver mea-sures the reflected power to detect the end of a branch instead of the Rayleigh back-scattering power that is typically tens of decibels lower (�−40 dB in [12]). In our man-agement system, the OCDM encoder already performs this reflection role, in additionto introducing signatures inside the reflected power. Note also that our monitoringsystem considers the reflected powers only, in contrast to the standard OTDR thatmeasures the reflected powers and the Rayleigh backscattered powers as well. Thisgreatly alleviates the loss–power budget constraints of the system.

2.B. Limits of the ONT IntelligenceIn currently deployed FFTH networks, the ONTs are intelligent enough to cyclicallycommunicate with the optical line terminal (OLT) at the CO. Status cells areexchanged almost every 100 ms to inform the network manager about the quality ofthe received signal at the ONT. When the quality of this signal degrades, the ONTinformation is not sufficient for the network manager to localize the origin of the deg-radation.

The quality of the signal delivered to the customer depends on three different inde-pendent elements: (i) the transmission quality at the central office, (ii) the fiber linkquality, and (iii) the receiver quality at the ONT (including detection and all signalprocessing steps). Constant communication between the OLT and the ONT is essen-tial for the network manager to estimate the total quality of the transmission link andmany other aspects. When the transmission quality degrades, it is not possible tolocalize the impairment, not to determine if the impairment is due to (i), (ii), or (iii).

The network manager needs information about every segment of the network.Information about the OLT transmission quality is already available to the networkmanager through integrated tapped detection at the OLT laser. Information about thetotal quality of the system is available through direct communication between theOLT and the ONT. The missing information is that of the fiber link between the OLTand the ONT. Recall that the focus of this paper, as this is the cases of the authorsof [8–11], is to provide a technique for the network manager to obtain this missinginformation. Our technique allows the network manager to identify the faulty branchand to in-service monitor the total loss exhibited by any working fiber.

2.C. FTTH Service ProvisioningFixing the remote node (RN) to a customer’s distance. During FTTH installation, someof the OTDR suppliers recommend that no customers be connected to the RN withequal length fibers. This recommendation is made to ensure that every networkbranch will contribute to a discrete, distinguishable event in the OTDR trace. Notethat events coming from equidistant customers overlap in the OTDR trace and con-fuse the network manager. This installation criterion reduces the chance that equidis-tant events overlap in an OTDR trace, but it cannot guarantee that other reflectionsources such as connectors, splices, fusions, etc. cause ambiguity in the OTDR trace.Note that this recommendation is not a part of the FTTH standard. The use of simpleOCDM encoders at the ends of the network branches allows multiple reflections tooverlay in a pseudoorthogonal way. The distance between the customers and theremote node is no longer an issue, and no longer needs to be controlled or determinedin advance of installation, as recommended by various OTDR suppliers.

Unpredictable customer demand. Customer demand to FTTH services is unpredict-able, thus the network ramification and distances to clients is unpredictable. Theabove fiber length recommendation increases the complexity and the cost of provision-ing new customers. This problem becomes even more evident when the number ofbranches in a TDM-PON increases. This increases the number of discrete and over-lapping events in the CO OTDR trace, dramatically aggravating the task of the net-work manager.

Unused RN ports. In typical PON roll-out, some unused capacity remains at thesplitters, resulting in unused ports, as illustrated in Fig. 1. The unconnected ports inthe RN simultaneously reflect back a power and contribute to a high peak event that


is an accumulation of many overlapping, indiscernible events. The OCDM codes couldbe installed at the RN output ports, preventing the CO manager to be confused by theevents occurring from the unused ports of the RN. When a new customer is connectedto one port, its specific encoder is removed and placed at the fiber end in the front ofthe customer ONT.

2.D. FTTH MaintenanceFiber cut identification. When a fiber cut occurs in a distance that coincides or is closeto an event of another branch, the CO manager will observe no event in the OTDRtrace. For an isolated fiber break, a new peak event will be observed, but no indicationthat helps determining the faulty branch. Extensive prior knowledge of all elements ofthe network could help the manager guess the nature of the event; however, this is aninefficient control of the network. In [3], the authors propose that the network man-ager wait until receiving a call from the disconnected customer. If the number of cus-tomers claiming service interruption increases, the manager will understand that thefault occurred in the feeder instead of the distribution segment. This is intuitive, butprovides no industry solution to the PON management issues. In our system, when afiber branch is cut, the monitoring equipment will miss the encoded signal that is spe-cific to the cut branch. Note that when one customer calls for an interruption, themanager cannot determine if the problem is coming from the fiber or from the ONTitself. In that case, the OTDR suppliers recommend that a technician go to the cus-tomer premise and start testing by checking the ONT first. We will see in the end ofthis section how the OCDM passive encoding device delimits the service providerresponsibility from that of the customer. In effect, in all cases, the service provider hasdirect control and responsibility of the OLT and the fiber network. However, the ONTis sometimes, depending on the service provider’s market choices, under the customercontrol.

Quality of service. The service provider cannot rely on customer calls to assure itsquality of service. It is crucial for any telecom system to be able to guarantee a levelof service that is defined by the number of interruptions per year and the mean dura-tion for these interruptions. Efficient monitoring requires automated fiber fault loca-tion and preventive in-Service and live network measurements. In addition, custom-ers will request SLAs with service providers to ensure an acceptable quality of service.Loopback information including an estimate of the total loss, received end power, theONT state, etc., is necessary to ensure adequate management of the network, etc. Ourencoding system could be used to have an accurate estimate of the total loss exhibitedthrough the network for every termination.

2.E. Demarcation PointAs shown in Fig. 1, the demarcation point is the location in a network where control(or sometimes ownership), and thus responsibility, switches from the service providerto the customer. The carrier maintenance responsibility starts from the central officeand stops at these so called demarcation points. The demarcation point is adjacent tothe customer premises, in the front of the ONT connector. The demarcation devicecould be installed outside the customer premise, this accessible to the technicianswithout the customer presence. The customer controls (or sometimes owns), hencetakes care of the ONT that is installed inside his premises or home (see Fig. 1). When

Fig. 1. Typical FTTH PON showing the demarcation points and different distribution–drop fiber recommended having unequal lengths.


no communication exists between the CO and the ONT, the manager cannot deter-mine if this is caused by a fiber cut that is his maintenance responsibility or if simplythe ONT is disconnected or turned OFF, which is the customer responsibility. PONsneed a device, preferably passive, to be installed at every ONT outside the premisesand that identifies exactly the demarcation point for that branch. Note that existingU.S. service providers own the ONTs, and place them outside the home to ensureunconstrained access to them. The demarcation issue, as defined in this paper, doesnot apply for this case. We will see in Section 3 how our OCDM encoder fulfills thisneed and serves as the demarcation point device.

3. OCDM Solution for In-Service TDM-PON ManagementOur OCDM management technique requires only one wavelength, preferably in the1650 nm wavelength already reserved for standard FTTH TDM-PON monitoring, orthe nonstandard but inexpensive 1625 nm wavelength [3]. Our technique is also scal-able for more advanced and complex PON networks, always using a single wave-length. As illustrated in Fig. 2, every network branch is terminated by a standard pas-sive wavelength selector (WS), widely used in PONs, isolating the standardmonitoring U band from the other data bands at the front of every ONT and at the COas well.

3.A. System DescriptionThe transmission section of the OCDM monitoring equipment at the CO (Fig. 2) con-sists of a U-band pulsed laser driven by a processor to transmit short pulses with apredetermined low frequency rate (few megahertz or lower) and adequate power tosupport the back and forth cumulative loss. Every pulse propagates through the treenetwork, is split at the PS, is coded by encoders Enci, i=1 to N, and is then reflectedback to the CO. An encoder consists of a passive device that fragments an incomingpulse into a number of p subpulses distributed in time according to a specific code, i.e.,direct sequence coding in the time domain. Every encoder at the branch terminationimplements a unique code.

Our proposed network management system is a modified form of direct sequenceDS-OCDM, and our codes are also the so-called optical orthogonal codes (OOC). ThePS (coupler) combines the upstream encoded pulses together as in standard OCDM.In the CO, a tunable DS decoder, consisting in Fig. 2 of an optical switch and a bankof fixed decoders (similar to encoders but introducing delays in reverse order), dis-criminates responses coming from different branches of the tree network. Everyhealthy branch in the network contributes an autocorrelation peak. A missing auto-correlation peak indicates the corresponding network branch is broken or exhibitsabnormal power loss. Furthermore, the height of the detected autocorrelation peak inthe normally working branches, indicates the cumulative end-to-end loss of the fiberlink, including that attributed by the fiber, connectors, splitters, splices, fiber bending,etc. Recall that our technique provides for the network manager the currently missinginformation about the loss specifically involved by the fiber link. As explained in Sub-section 2.B, cyclic communication alone between the OLT and ONTs is not sufficientto provide the network manager with information specific to the fiber link.

Fig. 2. OCDM-based TDM-PON monitoring system, every network branch (with arbi-trary unconstrained length) is assigned an encoder, one tunable decoder at the CO.


3.B. System BenefitsWe show here that our coding system addresses many of the management challengesdescribed in Section 2. System benefits are described qualitatively in this section andthe quantitative capacity advantage is studied separately in Section 4.

Overcomes OTDR shortcomings. The receiver of our monitoring equipment alsoacquires a cumulative reflections (or impulse responses) coming from all branches, inaddition to that of the feeder. However, all these impulse responses are discernable bymeans of the decoder and detection process, i.e., the use of orthogonal codes results inthe individual impulse responses being overlaid orthogonally. Furthermore, the corre-lation process is made using the reflected powers and not the Rayleigh backscatteredpowers. Hence, the dynamic range requirement is very much alleviated, as the back-scattered power is �40 dB lower than that reflected [3]. In Subsection 3.C, we will seehow the reflectivity of an encoder could influence its choice and design.

Helps service provisioning. New customers are no longer required to be connectedwith different fiber lengths from those previously installed. This alleviates installa-tion complexity for technicians. Network expansion can continues naturally and isbetter adapted to unpredictable customer demand. Even for the unused ports in theRN, we recommend the placement of our encoders, as this avoids simultaneous reflec-tions to contribute to a undesirable high peak event, which could mask other faultsclose to the ports connectors.

Reduces maintenance complexity and cost. Our system allows the CO manager tohave real-time, full information about all network elements. It does not rely on cus-tomer help or calls to diagnose the network state. The manager will no longer be con-fused between fiber and ONT faults. Recall that when a fiber fault occurs the carrieris responsible and our proposed technique allows troubleshooting without involvingthe customer, however an ONT fault depends in most cases on the customer himself.The latter fault is not considered a service interruption. Without complete andin-service, live PON monitoring, an outside intervention by technicians is necessary ineither case: fiber and ONT faults [4].

Passive demarcation solution. Our encoders are passive components that can beplaced outside the customer premises (i.e., home); the ONT however could be locatedinside the customer home, i.e., it is under his control, (sometimes his property), and ishis total responsibility. The encoder delimits the service provider control (or owner-ship) and responsibility from that of the customer.

3.C. Encoder DesignOur DS-OCDM network management technique is different from standardDS-OCDMA in various aspects: (1) the required transmission rate is low, dramaticallyalleviating the known implementation challenges of OCDMA; (2) no reception, notransmission or modulation is needed at the ONT side, only the coding operation isneeded; and (3) while standard DS encoders as in Fig. 3(a) work in transmission, inour case, reflection is sufficient so no circulator is required.

We propose four designs for the modified DS encoder, all appropriate for our appli-cation, in Figs. 3(b)–3(e). All exploit splitters and delay lines to fragment the incident

Fig. 3. Encoding designs: (a) standard encoder, (b) modified encoder, (c) 2q+1 splitterencoder, (d) 2q splitter encoder, and (e) MBG encoder.


pulse into subpulses, disperse them, gather them back to the fiber and return them tothe CO. Each of these encoder designs has its own advantages and drawbacks. Thefirst, Fig. 3(b), is the straightforward modification of the standard one of Fig. 3(a),entails 2�1:2q splitters and one circulator. The second implementation in Fig. 3(c)requires one 1:2q+1 splitter and eliminates the circulator. The implementation in Fig.3(d) requires only one 1:2q splitter. Finally, Fig. 3(e) is an implementation using amultiple Bragg grating (MBG), i.e., a series of discrete gratings at the same wave-length but with different reflection intensities and physical locations [13].

We define three key design parameters for our system: (1) high coding capacity (thenumber of different codes should be a minimum of 128 for standard GPON with FEC);(2) the modified DS encoders to be located at network terminations should be low costand low component count; and (3) the power loss between the incident pulse andreflected coded signal fed back to the fiber should be minimized in order to reduce therequired dynamic range of the monitoring system.

The encoder in Fig. 3(b) is the most expensive because of the circulator (with 1.5 dBloss per pass) and exhibits the highest power loss. The encoder in Fig. 3(d) is lessexpensive, however it exhibits very high loss due to the reflection coefficient �, typi-cally �4% �13 dB� [12]. Increasing � to 50%, using special mirror at the end of thefiber, will make this setting attractive. The encoder in Fig. 3(c) is less expensive thanthat in Fig. 3(e) because of the Bragg gratings compared to PSs, however MBGs hasless insertion loss. In Table 1, we outline the components count and derive the powerloss equations of each of the proposed DS-OCDM encoders. The coding loss illustratedin Table 1, is defined as the ratio between the individual powers of a returned and anincident pulses; however, the coding loss is the ratio between the sum of powers of allthe returned pulses and the power of the incident pulse. The coding loss is very impor-tant in our analysis since this directly affects the correlation and detection perfor-mance.

The MBG could be inserted directly in the data fiber without WS because it oper-ates out-of-band. Note however, that this MBG could not be written with high reflec-tivity, 40%–60% was reported in [13]. The encoders of Figs. 3(c) and 3(e) look verycompetitive in terms of cost and reflectivity.

3.D. Loss BudgetRecall that our correlation process uses the reflected power instead of the Rayleighbackscattered power, thus alleviating the dynamic range requirement, compared withthe standard OTDR system, for �40 dB [3]. Furthermore, the existence of an encoder,acting as a mirror at a fiber end increases the reflected power, thus additionally alle-viates the dynamic range [14]. The decoder introduces an additional loss, but thisremains quite limited compared with the gain we achieve in the dynamic range. Theimportance of the loss/power budget in our system should be considered in a monitor-ing (or OTDR) context rather than in data communication context. OTDR and moni-toring researchers usually use special photodetectors with extremely high photo-sensitivity and large dynamic rage. Commercial OTDRs currently work with receivedpower that ranges from −90 to −100 dBm and supports a dynamic range of�120 dB [14].

3.E. System ModelingLet � be a collection of binary n-tuplets generated by [15] from the specification (n, w,�a, �c) where n is the code length, w is the Hamming weight, �a and �c are constraintsdefined as follows. For vectors X, Y��, the expressions

Table 1. Characteristics of the Encoding Designs

Standard[Fig. 3(a)]

Modified[Fig. 3(b)]

2q+1 Splitter[Fig. 3(c)]

2q Splitter[Fig. 3(d)]

MBG[Fig. 3(e)]

Components 2 PSs of 1:2q 2 PSs of 1:2q,1 circulator

1 PS of 1:2q+1 1 PS of 1:2q 2q gratings

Pulse loss (dB) 3+6q 3+6q 6+6q 6q+� —Number of pulses 2q 2q 2q+1 2q 2q

Coding loss (dB) 3+3q 3+3q 3+3q 3q+� 7 to 10


RX�� = �i=1

n

X�i�X�i + �� a,

RXY�� = �i=1

n

X�i�Y�i + �� c

define, respectively, the autocorrelation and the cross-correlation functions, both for−n+1��n−1. Note that the code weight w (i.e., the number of logical ones per code)is equal to the number of delay lines in the encoding devices of Figs. 3(b) and 3(d), andthe number of Bragg gratings in the MBG encoder of Fig. 3(e).

OOCs could never be made truly orthogonal, especially in our asynchronous case.The autocorrelation will generally be perturbed by interference coming from otherbranches. It is known that the performance of most OCDMA systems is limited by theinterference. In this application, the interference problem is greatly lower than that intraditional OCDMA system, because of various aspects: (i) the repetition rate of theincident monitoring pulse can easily be reduced, avoiding interference between reflec-tions resulting from subsequent pulses; (ii) the distance between customers will affectthe interference statistics since only close customers will cause appreciable interfer-ence.

It is not our goal in this paper to develop the methodology of dealing with interfer-ence rejection and detection threshold, but it is clear that our system suffers less frominterference problem, and requires an analysis different from traditional OCDMA. Forinstance, encoders do not transmit data, i.e., every pulse will contribute to an encodedsignal without any modulation; i.e., the activity factor is always 1. In addition, theperformance study of our system should focus in the future on the probability of falsealarm rather than the bit error rate or the probability of error. This is particularlynew as analysis method and tool of an OCDMA system. It is, however, not in the pur-pose of this paper to develop advanced and deep theoretical analysis of the system.

4. OCDM for Complex PONs Management4.A. WDM-PON ApplicationFigure 4 shows a WDM-PON network architecture based on WDM multiplexer(upstream) and demultiplexer (downstream). In WDM-PON, every customer is servedby a dedicated wavelength [5,6], this allows fixed high bandwidth delivery to all users.Previously proposed PON management uses half of the available wavelengths for dataand the other half for monitoring, i.e., every customer is assigned two wavelengths,one for data and another for monitoring. In Fig. 4, we also show an architectural solu-tion that allows our technique to manage this WDM-PONs with much fewer wave-length resources. In our system, we place a WS before the 1:K WDM demultiplexerinput in order to isolate the monitoring U band from the data bands. A 1:K PS splitsthe monitoring signal into K copies that are coupled again with data fibers at thedemultiplexer K outputs. Upstream and downstream data are assumed here to sharethe same fiber; other variants could be similarly derived. Our technique eliminates all

Fig. 4. Implementation of OCDM management in advanced WDM/PON architecture.


the monitoring wavelengths and makes them available for data, hence doubling thenetwork capacity. Only a single monitoring wavelength is split and distributed to allclients.

4.B. TDM–WDM-PON ApplicationFigure 5 shows typical TDM/WDM-PON network architecture where a separate TDMsystem is built over every WDM wavelength. This takes advantages from TDMA andWDM technology in order to dramatically increase the number of users servedalthough per client bandwidth is reduced. Similar to the WDM/PON case, we use�1+K� wavelength selectors and a 1:K PS to create an alternate path for the OCDMmonitoring signal. It should be noted that this method does not control 100% of the

Fig. 5. Implementation of OCDM management in advanced TDM/WDM-PON.

Fig. 6. Coding system design tool: (a) TDM-PON systems, (b) TDM/WDM-PONsystems.


fiber paths since the monitoring signals avoids going through the demultiplexers. Ourarchitecture, however, controls most of the sensitive segments of these PONs.

4.C. Capacity AnalysisFor all of the three architectures, the OCDM system is unchanged, only the capacityrequirement changes. In this section, we concentrate on the coding capabilities of aDS-OCDM system. We look to the key parameters that maximize the different opticalorthogonal codes that could be assigned in our application.

In [15], the authors recently developed an OOC generating technique, the so calledouter-product matrix based algorithm, that maximizes the number of codes for a givenweight w, length n, and unit cross-correlation and autocorrelation sidelobes ��a=�c=1�. We used this algorithm to develop a coding system. In Figs. 6(a) and 6(b) we showthe number of codes (i.e., clients) versus code length n for different code weights w.From Fig. 6(a), a designer could determine the appropriate (w, n) pairs to supportEPON, GPON, etc. For w=4, four delay lines are required, the encoder of Fig. 3(c)uses one 1:8 PS (i.e., 1 :23 PS with q=2) and the MBG encoder of Fig. 3(e) requiresfour gratings. Figure 6(b) presents codes that could support TDM/WDM-PON ofFig. 4(b). In our analysis here we assumed the optical pulses sent from the monitoringunit were normalized in power and duration. More extensive analysis would includethe detector sensitivity, the absolute pulse power, and duration.

5. ConclusionsWe proposed a novel in-service live PON management solution using a modifiedDS-OCDM system that overcomes the OTDR point-to-multipoint shortcomings, aswell as various service provisioning and network maintenance challenges. Our solu-tion alleviates the provisioning and maintenance complexity of PONs and reducestheir operational cost. In addition, our system provides passive devices to demark ser-vice provider control (or ownership) and responsibility from that of customers. OurOCDM-based management solution easily scales up from FTTH TDM-PON to WDM-PON and TDM/WDM-PON, supporting close to a thousand customers, all using onlyone monitoring wavelength. We developed different encoding settings for our systemand developed capacity curves using appropriate codes.

References1. R. P. Davey and D. B. Payne, “The future of optical transmission in access and metro

networks—an operator’s view,” in Proceedings of the 31st European Conference on OpticalCommunication (ECOC 2005), (IEEE, 2005), paper We 2.1.3.

2. G. Talli, C. W. Chow, E. K. MacHale, and P. D. Townsend, “High split ratio 116 km reachhybrid DWDM-TDM 10Gb/s PON employing R-ONUs,” Presented at the EuropeanConference on Optical Communication, Cannes, France, 24–28 September 2006, paperMo4.5.2.

3. N. Gagnon, A. Girard, and M. Leblanc, “Considerations and recommendations for in serviceout-of-band testing on live FTTH networks,” in Optical Fiber Communication Conference(OFC 2006) (Optical Society of America, 2006), paper NWA3.

4. A. Girard, FTTx PON Technology and Testing (EXFO Electro-Optical Engineering, 2005).5. Y. Chien-Hung and C. Sien, “Optical fiber-fault surveillance for passive optical networks in

S-band operation window,” Opt. Express 13, (2005).6. T. Pfeiffer, “Monitoring and protecting the optical layer in FTTH networks,” in Proceedings

of the FTTH Conference and Expo, Las Vegas, Nevada, 3–6 October 2005.7. H. Schmuck, J. Hehmann, M. Straub, and Th. Pfeiffer, “Embedded OTDR techniques for

cost-efficient fibre monitoring in optical access networks,” Presented at the EuropeanConference on Optical Communication, Cannes, France, 24–28 September 2006, paperMo3.5.4.

8. C.-K. Chan, F. Tong, L.-K. Chen, K.-P. Ho, and D. Lam, “Fiber-fault identification forbranched access networks networks using a wavelength-sweeping monitoring source,” IEEEPhoton. Technol. Lett. 11, (1999).

9. S.-B. Park, D. K. Jung, H. S. Shin, D. J. Shin, S. Hwang, Y. Oh, and C. Shim, “Optical faultmonitoring method using broadband light source in WDM-PON,” Electron. Lett. 42, (2006).

10. S. Hann, J. Yoo, and C.-S. Park, “Monitoring technique for hybrid PS/WDM-PON by using atunable OTDR and FBGs,” Meas. Sci. Technol. 17, 1070–1074 (2006).

11. J.-H. Park, J.-S. Baik, and C.-H. Lee, “Fault-localization in WDM-PONs,” in Optical FiberCommunication Conference (OFC 2006) (Optical Society of America, 2006), paper JThB79.

12. D. Derickson, Fiber Optic Test and Measurement (Prentice Hall PTR, 1997).13. J. M. Castro, I. B. Djordjevic, and D. F. Geraghty, “Novel super structured Bragg gratings

for optical encryption,” J. Lightwave Technol. 24, 1875–1885 (2006).


14. N. J. Frigo, P. P. Iannone, K. C. Reichmann, X. Zhou, and M. W. Stodden, “Centralizedin-service OTDR testing in a CWDM business access network,” J. Lightwave Technol. 22,2641–2652 (2004).

15. H. Charmchi and J. A. Salehi, “Outer-product matrix representation of optical orthogonalcodes,” IEEE Trans. Commun. 54, 983–989 (2006).

code-division multiplexing for in-service out-of-band monitoring of live ftth-pons

Documents