“coverage area management...

84
IST-2000-25390 OBANET D3.1 “Coverage area management specifications” Contractual Date of Delivery to the CEC: 08/31/2001 (M09) Actual Date of Delivery to the CEC: 08/31/2001 (M09), revised 28/05/2004 Author(s): José F. Kukielka, Gregorio Núñez (ALC) Javier Marti, V. Casares, Juan L. Corral, Valentín Polo, Javier Herrera, Pablo Sanchis, José M. Martínez (UPVLC) Gerd Grosskopf (HHI) Participant(s): ALC, UPVLC, HHI Workpackage: WP3 Est. person months: 21 Security: PU Nature: Report Version: v 1.1 Total number of pages: 84 Abstract: In this document, all the functionalities of the optically beam-formed base-stations (BS) are defined and the techniques to carry them out described. The problems and effects that have to be considered as well as the limits of operation are studied. A description of the algorithms that will be used to implement each functionality (beam-switching, beam-shaping) and related strategies (user tracking/location or cell reconfiguration) in the Beam Adaptation Protocols (BAP) is provided. These BAP will be implemented within the simulation tool to be developed in D3.2. Keyword list: Coverage area management strategies, Beam adaptation protocols, Beamforming functionalities

Upload: others

Post on 09-Dec-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

IST-2000-25390 OBANET

D3.1 “Coverage area management specifications”

Contractual Date of Delivery to the CEC: 08/31/2001 (M09)

Actual Date of Delivery to the CEC: 08/31/2001 (M09), revised 28/05/2004

Author(s):

José F. Kukielka, Gregorio Núñez (ALC) Javier Marti, V. Casares, Juan L. Corral, Valentín Polo, Javier Herrera, Pablo Sanchis, José M. Martínez (UPVLC) Gerd Grosskopf (HHI)

Participant(s): ALC, UPVLC, HHI

Workpackage: WP3

Est. person months: 21

Security: PU

Nature: Report

Version: v 1.1

Total number of pages: 84

Abstract: In this document, all the functionalities of the optically beam-formed base-stations (BS) are defined and the techniques to carry them out described. The problems and effects that have to be considered as well as the limits of operation are studied. A description of the algorithms that will be used to implement each functionality (beam-switching, beam-shaping) and related strategies (user tracking/location or cell reconfiguration) in the Beam Adaptation Protocols (BAP) is provided. These BAP will be implemented within the simulation tool to be developed in D3.2. Keyword list: Coverage area management strategies, Beam adaptation protocols, Beamforming functionalities

Page 2: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 2 of 84

Executive Summary In this deliverable, the Base Station (BS) and Subscriber Station (SS) features that will be demonstrated using the simulation tools developed in OBANET are specified. All the functionalities of the optically beam-formed BS are defined and the techniques to carry them out described. The problems and effects to take into account as well as the limits of operation are outlined. A description of the algorithms that will be used to implement each technique is provided. Section 1 provides an overview of the state-of-the art of broadband wireless access networks, in particular focused on the on-going standards for broadband wireless access neetworks, ETSI BRAN HIPERACCESS and IEEE 802.16.1 WirelessMAN. The specifications and requirements, such as frequency plan, frequency reuse pattern, SNR requirements, and so on, that have to be satisfied by the beamforming networks that are investigated in OBANET are based in both standards. In section 2 a detailed description of the MAC protocol that is going to be tested employing the simulation tools to be developed in WP3 is provided. Modifications of that protocol must be implemented to incorporate advanced beamforming capabilities to a BS. Section 3 describes different models that will be employed to simulate the OBANET scenario. Different traffic models have been tested for the considered broadband data services: voice, video and data. Simplified but realistic antenna models have been studied, including BS antenna elements, antenna arrays and SS directive antennas. Also a radio propagation model has been studied including specific statistical effects such as multipath or fading at the 40 GHz propagation. Finally, mobility models for the mobile scenario were also considered. A description of the simulation, specification and programming techniques to evaluate the performance of the proposed protocol modifications as well as the structure of the simulation software that will be developed in WP3 are presented in section 4. The programming platform design was chosen to be as event oriented, and the main processes required for such simulations were identified. The SDL specification language was chosen to describe the OBANET scenario protocols and functionalities. Finally, the different modules, libraries and interfaces of the simulation tools are proposed. In section 5, the beamforming functionalities that will be tested using the simulation software and the techniques to implement them are specified, emphasising the implementation algorithms and constraints, and making connections with the experimental work to be developed in other workpackages. As beam-switching comprises the capability of scanning the BS antenna to the desired direction, the required control data to manage the switching control has been specified. The beam-switching scenario enhancing possibilities were studied, mainly antenna gain control and dynamic cell reconfiguration. Also, the mandatory modifications in the system protocols to support the beam-switching technique were identified and solutions were proposed. Finally, the implementation of a direction-of-arrival estimation algorithm using beam-switching was considered. Performance evaluation studies will be further carried out in WP3 in order to demonstrate how the system capacity can be increased employing such adaptive beamforming/beam-switching strategies. Finally, conclusions and remarks are provided in section 6 and several underlying mathematical models for statistical are included as an annex.

Page 3: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 3 of 84

Table of Contents List of Abbreviations List of Tables List of Figures

5

7

8

1. Introduction 1.1. Broadband Wireless Access Networks 1.2. Optical Beamforming for Wireless Systems 1.3. Objectives of this Deliverable

10 10 12 16

2. Broadband Wireless Protocols 2.1. Physical Layer 2.2. Medium Access Control Layer 2.3. Network Layer Aspects

17 17 19 22

3. System Models 3.1. Traffic Models 3.1.1. Voice Traffic 3.1.2. Video Traffic 3.1.3. Data Traffic 3.1.4. Ethernet Traffic 3.1.5. TCP/IP Traffic 3.2. Antenna Models 3.2.1. BS Antenna 3.2.2. SS Antenna 3.3. Radio Channel Models 3.3.1. Radio Propagation at 40 GHz 3.3.2. Multipath and Shadowing Fading 3.3.3. Statistical Radio Channel Model 3.4. Mobility Models 3.4.1. Out-door Mobility 3.4.2. In-door Mobility

24 24 24 25 26 27 28 31 32 34 34 35 35 36 37 38 39

4. Protocols Simulation and Specification 4.1. Discrete Event Simulation 4.2. Protocols Specification 4.3. Simulations Structure

42 42 44 45

5. Beamforming Functionalities and Strategies 5.1. Beam Switching 5.1.1. Switching and Steering 5.1.2. Beamwidth Control 5.1.3. BWA Protocols Modifications

47 47 47 49 55

Page 4: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 4 of 84

5.1.4. Beam Switching Algorithm 5.2. Tracking and Locating 5.3. Beam Shaping 5.3.1. Gain Control 5.3.2. Beam Shaping Algorithm 5.4. Cell Reconfiguration 5.5. System Monitoring

56 57 61 62 64 64 65

6. Conclusions and Remarks

67

Appendix. Statistical Distributions A.1. Probability Distributions A.1.1. Geometric A.1.2. Poisson A.1.3. Exponential A.1.4. Gamma A.1.5. Pareto A.2. Fading Distributions A.2.1. Lognormal A.2.2. Rayleigh A.2.3. Rice

69 69 69 69 70 70 70 71 71 72 73

References

74

Page 5: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 5 of 84

List of Abbreviations ATM Asynchronous Transfer Mode BAP Beam Adaptation Protocols BE Best Effort BER Bit-Error Rate BWA Broadband Wireless Access BS Base Station CCI Co-Channel Interference CIR Carrier-to-Interference Ratio CS Control Station DCA Dynamic Channel Allocation DECT Digital Enhanced Cordless Telephone F/B Front-Back Ratio FEC Forward Error Control FER Frame-Error Ratio GPRS General Packet Radio Services GSM Global System for Mobile IPP Interrupted Poisson Process LMDS Local Multipoint Distribution System LMS Land Mobile Satellite LoS Line-of-Sight MAC Medium Access Control MMDS Microwave Multipoint Distribution System MMPP Markov Modulated Poisson Process MPEG Moving Pictures Experts Group MVDS Multipoint Video Distribution System n-IRP n-Interrupted Renew Process ntr-PS Non Real Time Polling Services PDU Protocol Data Unit PIB Photonic Integrated Beamformer QoS Quality of Service rt-PS Real Time Polling Service SDL Specification and Description Language SLL Secondary-Lobe Level SNR Signal to Noise Ratio SONET/SDH Synchronous Optical NETwork/Synchronous Digital Hierarchy SS Subscriber Station TCP/IP Transmission Control Protocol/Internet Protocol TDD Time-Division TDM Time-Division Multiplex TDMA Time-Division Multiple Access UTP Unshielded Twisted Pair UGS Unsolicited Grant Service UGS-AD Unsolicited Grant Service with Activity Detection UMTS Universal Mobile Telecommunications System WAP Wireless Access Protocol

Page 6: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 6 of 84

WLAN Wireless Local Area Networks WLL Wireless Local Loop VoIP Voice Over IP

Page 7: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 7 of 84

List of tables Table 2.1 Recommended frame parameters Table 3.1 Voice source model parameters Table 3.2 Video source model parameters Table 3.3 Data source model parameters Table 3.4 Equivalence of Markov chain and exponential dwell time Table 3.5 TCP/IP traffic model parameters Table 3.6 100 Kbps scaled TCP/IP parameters Table 3.7 Current distributions for a four-elements array Table 3.8 Rice K factor in several environments Table 3.9 Pedestrian user parameter Table 3.10 In-door motion mean times Table 3.11 In-door motion transition probabilities Table 5.1. Estimated approximations for Ap and ∆θ–3dB (@ 90º) Table 5.2. Approximations for beamwidth and scan angle Table 5.3. Optimal beamwidth values Table 5.4. Phases and delays for the scan angle fine-tuning functionality Table 5.5. Tracking scan times values Table 5.6. Maximum error in the tracking algorithm Table 5.7. Beamwidth ranges for the –5 to 5 dB current variation Table 5.8. Beamwidth ranges for the –20 to 6 dB current variation

19 24 25 26 28 30 30 33 36 39 41 41 52 52 54 54 60 61 63 63

Page 8: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 8 of 84

List of Figures Figure 1.1 OBANET scenario: switched beam instead of a sectorial antenna Figure 1.2 An OBANET proposed capability: cell reconfiguration Figure 2.1 TDD/TDMA frame structure Figure 2.2 Downlink subframe structure Figure 2.3 Uplink subframe structure Figure 2.4 Block Diagram of the Physical Layer Figure 2.5 MAC PDU generic format Figure 2.6 MAC data PDU format Figure 2.7 ATM based BS to CS interconnection Figure 3.1 A two-states Markov Chain Figure 3.2 Trace of a voice packets source Figure 3.3 Trace of video packets sources: 128 kbps and 512 kbps Figure 3.4 Trace of a data packets source Figure 3.5 A n-states Markov chain Figure 3.6 Short and Long term IPP components Figure 3.7 MMPP philosophy of the TCP/IP traffic model Figure 3.8 Trace of an 100 Kbps internet traffic source Figure 3.9 Packet length with a=1 and b=0.9 Pareto distribution Figure 3.10 Binary rate for an internet traffic source Figure 3.11 Parabolic radiation pattern of an element Figure 3.12 Fixed SS antenna radiation pattern Figure 3.13 Mobile SS antenna radiation pattern Figure 3.14 Two-states Markov chain radio channel model Figure 3.15 Two dimensional random motion Figure 3.16 Street grid movement and turn probability Figure 3.17 In-door motions Figure 3.18 In-door motion states diagram Figure 4.1. Flow control for the next-event time-advance approach. Figure 4.2. Simulation Tool Structure Figure 5.1. Beam switching angular positions Figure 5.2. Beam-steering functionality Figure 5.3. Radiation pattern modification due to pointing Figure 5.4. Beamwidth and SLL for an 4-elements antenna array Figure 5.5. Beamwidth and SLL for an 8-elements antenna array Figure 5.6. Beamwidth and SLL for an 16-elements antenna array Figure 5.7. Worst case pointing angle error due to element antenna Figure 5.8. Worst case gain reduction due to element antenna Figure 5.9. Beamwidth estimation error and SLL for an 4-elements array Figure 5.10. Beamwidth estimation error and SLL for an 8-elements array Figure 5.11. Beamwidth estimation error and SLL for an 16-elements array

13 15 17 18 18 18 20 21 23 24 25 26 26 27 29 30 30 31 31 32 34 34 37 38 39 40 40 44 46 48 48 49 50 50 51 51 52 53 53 53 54 55 56 58 58

Page 9: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 9 of 84

Figure 5.12. Beamwidth error for an extended currents range Figure 5.13. Slot based beam-switching approach Figure 5.14. Efficiency loss due to switching time Figure 5.15. Pilot tone in the data burst preamble Figure 5.16. Monopulse tracking algorithm Figure 5.17. Tracking scanning time worst case Figure 5.18. Tracking scan time for 4, 8 and 16 beams Figure 5.19. Tracking estimation error worst case Figure 5.20. Tracking algorithm with overlapping estimation Figure 5.21. Gain control acting over the beamwidth Figure 5.22. Beamwidth control acting over the current distribution Figure 5.23. Cell reconfiguration Figure A.1. Geometric mass function Figure A.2. Poisson mass function Figure A.3. Exponential density distribution Figure A.4. Gamma density distribution Figure A.5. Pareto density distribution Figure A.6. Lognormal density distribution Figure A.7. Rayleigh density distribution Figure A.8. Rice density distribution

59 60 60 61 62 63 65 69 69 70 70 71 72 72 73

Page 10: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 10 of 84

1. Introduction

A tremendous growth of broadband multimedia interactive services is expected for next five years. The provision of these services requires high-performance and high-reliable communication networks for both the access part and the transport part. In principle, wired and wireless terrestrial access networks can meet stringent specifications to provide broadband interactive services. Despite of the huge potential offered by wired access networks, wireless access is sometimes preferred in supporting broadband services due to: (i) mobility of users is possible (mobile communication networks), (ii) fast deployment, (iii) very good bandwidth over cost ratio and data rate scalability, and (iv) low maintenance and high flexibility in maintenance and upgrading costs.

Broadband Wireless Access (BWA) networks operating at millimetre-wave frequency bands are expected to fulfil the bandwidth requirements of such broadband multimedia services. At these frequencies, temporary obstacles and adverse meteorological conditions may drastically impair on the system performance by intercepting the required line-of-sight radio wave propagation, which results in degrading the system performance, and even more, making the system unavailable.

OBANET project aims at studying, proposing, implementing and testing specific strategies as well as their associated technologies to optimise adaptive BWA networks performance and to increase the service reliability in two scenarios: fixed and mobile radio accesses. In particular, the project focuses on how such strategies may be implemented using optically beamformed distribution antennas in fixed and mobile wireless networks in the 40 GHz band. Specific network management strategies must be implemented in order to make profit of the beamforming advantages. These strategies have to dynamically address changes in the network features and resources in order to optimise the network performance. 1.1. Broadband Wireless Access Networks

Recently, there has been an huge increase in the deployment of fixed BWA networks. Such systems have been deployed world wide and many vendors are offering them. And telecommunications operators are offering voice, data, and video services through them. Several frequency bands have been allocated for this systems, mainly in the 28 GHz (e.g., LMDS) and 40 GHz (e.g., MVDS) but also below 5 GHz (e.g., MMDS or WLL), though with reduced performance.

Page 11: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 11 of 84

In general, fixed BWA is a point to multi-point communication system operating above 10 GHz (depending on country of licensing) that can be used to provide high-speed Internet access, real-time multimedia, file transfer, remote access to corporate local area networks, interactive video, video-on-demand, video conferencing, and telephony among other potential applications. A definition of broadband access is the one that provides at least 5 Mbps rate per user in the downstream direction and 500 Kbps in the upstream [Khu99]. Next generation systems may require higher speeds, up to 155 Mbps.

A typical fixed BWA system distribute low power, high frequency signals over relatively short distances through a direct line-of-sight path. The architecture of these networks is cellular, taking profit of the short transmission distances imposed by the high attenuation arising from rain and foliage loses at millimetre-wave frequencies [Xu00, Xu00a, Cha99] in order to use efficient frequency and polarisation reuse patterns. Cells are typically designed as squares with four 90º sectors and spaced 4-6 Km. apart.

The network architecture uses radio hubs called base station (BS) to serve a group of users with high roof-top antennas [Cra99]. The idealistic circular coverage area around the cell site is divided into 4, 8, 12, 16, or 24 sectors and is covered using sectorial antennas with 90, 45, 30, 22.5, or 15º beam-width. The customer premises equipment or subscriber station (SS), consist of the user digital equipment (in-door unit) which connects to the network infrastructure, and the microwave equipment (out-door unit) mounted on the rooftop and pointing to the base station using a directional antenna.

There are many manufacturers that offer such systems in the millimetre-wave band (ALCATEL, MARCONI, NORTEL, HARRIS, ERICSSON, etc.) and all their systems have in common:

cellular schemes up to 12 sectors per cell, QPSK and 16 and 64-QAM modulation formats, up to 155 Mbps bit rate per sector, BS output powers up to 18 dBm (at 40 GHz), and variable coverage range depending on the modulation and

carriers.

Different fixed BWA trials in the millimetre-wave band have been deployed among Europe: for instance, ALCATEL, HKNET or PSINET are some vendors and operators involved in such developments. A landmark in such initiatives has been the Cellular Radio Access for Broadband Services (CRABS) Project 215 in the European ACTS program. CRABS is based on the possibility of extending a DBV-S digital TV system to include interactive services (Internet access). The project has found its extension

Page 12: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 12 of 84

in the EMBRACE IST project [Emb00] which looks inside introducing BWA mass market and low cost equipment. Different recommendations related with digital TV and multimedia distribution have included interactive channels in order to provide fixed BWA such as DAVIC 1.2 or the close alternative DVB-RCL. In response to growing market pressure for low-cost equipment, ETSI and IEEE have established standardisation projects for broadband radio access networks: BRAN HIPERACCESS and 802.16.1 WIRELESSMAN. Most of the system parameters for the OBANET project such as the cellular pattern, duplex and access methods, frame duration and scheme, medium-access procedures, etc. are directly based on those-proposed in these emergent standards.

There has been also an increasing interest on developing a broadband and global mobile communication system. These third-generation (3G) mobile systems are being studied and specified under the name of UMTS by the ETSI and its dedicated sub-technical committee SMG5 [Rap95]. The aim of these systems is to provide users with advanced communication services with broadband capabilities and using a single standard [God97]. Many applications can be provided using UMTS, but its maximum capacity of 2 Mbps may be a limiting factor for users of high data rate multimedia services, especially if high-definition video is required. Progress on video coding is a strong factor in the bit rate requirements for users, which would eventually relax system bandwidth requirements. However, recent studies have shown that there is an increasing demand for capacity [Mbs96].

The millimetre-wave band is very promising for these high data rate mobile applications because of its wide frequency spectrum, compact and light equipment and ease of interference-free system configurations. The Mobile Broadband System (MBS) is a wireless cellular network operating at this band. It has been developed in the European collaborative RACE II project R2067, addressing techniques, technology as well as system concepts. The MBS has been described as the fourth-generation mobile system but it is more an enhancement to 3G systems [Fer93]. It will allow to transport ATM cells over the air interface from 2 up to 34 Mbps. To provide an easy interference free frequency reuse every few kilometres, the 60 GHz band, with a maximum of propagation attenuation due to oxygen absorption, has been proposed by CEPT for MBS. These networks are commonly formed by picocells from 100 up to 500 m However, the successful deployment of BWA technologies faces a number of critical challenges [Xu99]:

Page 13: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 13 of 84

Spectrum Efficiency. The fast increase in capacity and the limited radio spectrum requires of networks with very high spectrum efficiency. Several techniques for increasing capacity have included cell reduction (micro-cells) or sectorisation, Dynamic Channel Allocation (DCA), aggressive frequency reutilization or higher-order modulations. Also a good demand scalability (load density can increase by two or three orders of magnitude in the different stages of the network deployment) of the system bandwidth is a must.

Coverage. Wireless systems in the millimetre-wave band suffers

significant attenuation from rain, vegetation, and shadowing. A good coverage can be a changeling task specially in the initial stages of the system deployment. BWA networks should provide better than 80% coverage in the service area.

Throughput and QoS. Efficient system protocols, scheduling and

queuing tasks and multimedia and data convergence become critical to ensure high throughput and good QoS features.

1.2. Optical Beamforming for Wireless Systems Smart Antenna technology exploits multiple antennas in transmitting and receiving modes, with associated coding, modulation and processing to improve the performance of a radio system in terms of capacity and coverage. The main philosophy is that interferers rarely have the same location as the users. By maximising the antenna gain in the user direction, if permitted, and simultaneously minimising it in the interference direction the signal quality can be significantly improved.

Usually, smart antennas are electronically controlled acting over the amplitude and phase of each element. However these systems are bandwidth limited because electronic shifting is intrinsically narrow-band. Optical beam control networks provide fast beam steering of antenna array and large bandwidth [And96]. Different kind of smart antennas can be defined depending on the beamforming capability and signal processing:

Switched lobe. Comprises a basic switching between separate directive antennas or lobes of an array (Buttler matrix). The best performance lobe is chosen in terms of received power.

Switched beam. Including a direction-of-arrival algorithm,

continuous or discrete tracking of a single directive beam can be achieved, so the received power is maximised.

Page 14: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 14 of 84

Adaptive array. Special signal processing algorithms can be used to implement user tracking, interference reduction, spatial multiplexing, etc.

In order to introduce beamforming functionalities in BWA networks

the OBANET project proposes an scenario with a space-switched single-beam antenna covering the whole sector from the base station, instead of the traditional omnidirectional or sectorial antenna, depicted in figure 1.1. This new approach differs from the others in that the BS covers the service region with a group of narrow beams, which can be rapidly scanned to a number of different positions synchronised to the sequences of the allocated time intervals. This beam-switching approach matches perfectly with a time based system as the BS antenna is illuminating only in the desired direction and just in the assigned slot of time [Aca91].

t1

t2

t3

Figure 1.1. OBANET scenario: switched beam instead of a sectorial antenna.

The introduction of smart antennas in BWA networks will have a

large impact in their performance in many aspects [Leh99]:

Range Increase. Multiple antennas coherently combine the signal energy improving the available gain. Owing to smart antennas will be more directive than traditional sectorial or omnidirectional antennas, a range increase is obtained. This allows to place the base stations further apart leading to a lower cost deployment. Also spatial diversity of smart antennas combats the channel fading.

Interference Reduction. Multiple antennas can be adaptively

combined to cancel or minimise interference due to spatial diversity. As cellular systems are typically interference limited smart antennas can dramatically increase the system CIR.

Page 15: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 15 of 84

Capacity Increase. The implication of a higher CIR is the possibility for reduced frequency reuse distances. System capacity increases of 300% or higher are expected.

Other aspects such as security considerations, spatial

multiplexing, new services related with spatial information, weight and size of millimetre-wave antennas, etc.

There are several companies that offer electronically controlled

adaptive antenna arrays (PARATEK, WAVEBAND, WIRELESS-ONLINE, ARRAYCOMM, METAWAVE, etc.) for fixed wireless local loop, but only at the physical layer. Neither MAC nor network control is provided.

Although the benefits of the OBANET scenario introducing smart antennas are many, there are also drawbacks and cost factors:

Mobile location and tracking becomes necessary [Hel97]. This restriction could be easily saved as some tracking and location algorithms can be implemented with no or little protocol overhead.

Traditional medium-access control (MAC) protocols commonly

assume that broadcast messages can be sent to all terminals, and in some periods all terminals can access simultaneously (in contention) to the channel. This is not feasible for directive antennas so those protocols must be specially adapted. This is always done by having more protocol overhead (i. e. sending the broadcast messages to all subsectors) [Hor95]. As the efficiency of the protocols is a key parameter for a broadband network, a good design of the beamfoming-adapted protocols is a must.

Also the time used to switch between the different positions

during a frame means a wasted bandwidth and must be taken into account in the protocol design (guard times). As an example of this requirement, the slot based switching approximation in [Oba01] imposes a switching time in the order of tens of nanoseconds.

The implementation of beamformed antenna BS allows the

compensation of high scattering man-made or natural impairments. But it also permits the specification of more sophisticated coverage strategies such as cell reconfiguration and cell reshaping or interference reduction due to least-interfering beam-switching [Blo01]. Figure 2 schematises the cell reconfiguration capability.

Page 16: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 16 of 84

Figure 1.2. An OBANET proposed capability: cell reconfiguration.

Finally, some control is needed to perform the beam switching. For instance the spatial location of the users is needed. This estimation can be readily done at the registering stage of the connection by measuring the received power in different positions of the beam. Two main parameters will be considered as inputs for the beamformer control algorithms:

Detected power. It is directly related to the channel status and can be used to determine possible obstacles, rain conditions, etc. It can be estimated in both BS and SS.

The error control and packet-loss ratio. The number of errors

corrected after the FEC decoding and the number of damaged packets are determined by the error probability. As cellular systems are typically limited by the interference they can be used as an estimation of the co-channel interference levels.

The channel monitoring measurements can be made in the SSs

which should then report to the BS and jointed with the BS measures. Also a centralised monitoring of all BSs and SSs can be done in a control station (CS), instead of a distributed control in which a BS only takes into account its own cell information. The communications needs between CS and BSs must be evaluated in terms of physical medium (capacity, availability, etc) and the real-time requirements (hardware/software, process, etc). In the OBANET project, only centralised monitoring will be considered when introducing beam-switching functionalities.

Therefore our main objectives and design challenges are:

To make minimal modifications needed in the standardised BWA protocols in order to support the proposed beam-switching scenario.

Page 17: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 17 of 84

Developing efficient algorithms for tracking (mobile) and

locating (fixed), coverage management, interference reduction and system performance monitoring, and to incorporate them as system capabilities in the modified BWA protocols.

Evaluating the subscriber SS to BS and the BS to CS

communications requirements and providing an efficient and feasible solution.

1.3. Objectives of this Deliverable This document is intended to specify the OBANET system protocols or Beam Adaptation Protocols (BAP), which are the objectives of work-package 3. Its structure corresponds to the following aims:

To give a simplified but realistic overview of the standardised BWA protocols and their main features and mechanisms. To be compliant with those standards is a must.

Specify the simulation and designing techniques that are being

used to evaluate the modified protocols.

Specify the proposed beamforming strategies that will be implemented in such protocols taking into account the system monitoring, coverage management, beam switching, etc.

To present the models for data traffic generation, radio channel

simulation, users mobility, antenna diagrams and other models that will be further used in the system simulations.

Page 18: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 18 of 84

2. Broadband Wireless Protocols Next, a brief description of the standards for BWA networks which are under development by IEEE 802.16.1 and ETSI BRAN: WirelessMAN [Iee01] and HIPERACCESS [Hyp01] is provided. The main primitives and mechanisms have been identified. All the protocol schemes, procedures and mechanisms considered in the BAP developed in OBANET WP3 are intended to be compliant with this emergent standards. 2.1. Physical Layer The physical layer is a combination of Time Division Duplex (TDD), Time Division Multiplex (TDM) in the downlink and on-demand Time Division Multiple Access (TDMA) in the uplink. Therefore, the channel bandwidth is time-slotted and users are required to provide explicit or implicit information regarding their needs for bandwidth. As a TDD frame, the uplink and downlink transmissions share the same frequency, but are separated in time. The frame has a fixed length, but the length of each sub-period is a system parameter that may be dynamically modified. Allowable frame times are 0.5, 1 and 2 ms. The available bandwidth in both directions is defined with a granularity of one time-slot, which is a multiple of 4 modulation symbols. The modulation rate is selected in order to obtain an integer number of time-slots within each frame. Figure 2.1 depicts the frame structure described above.

frame j-1 frame j frame j+1 · ··· ··

downstream subframe upstream subframe

adaptivettime-slot

Figure 2.1. TDD/TDMA frame structure.

The downlink sub-frame starts with a frame header, which contains a preamble (based on 16 symbols CAZAC sequences) used for synchronisation and equalisation. This preamble contains also control sections in order to ensure the system interoperability. This control sections are used basically to notify the terminals of the bandwidth assignments in both uplink and downlink. The downlink data transmissions are organised in robustness order. Downlink transmissions are governed by the BS in a contention-free broadcast mode. Figure 2.2. depicts the downlink subframe structure.

Page 19: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 19 of 84

prea

mbl

e

Control QPSKdata

16-QAMdata

64-QAMdata

broadcast

DowlinkMap

UplinkMap

Figure 2.2. Downlink subframe structure.

There is a time gap between downlink and uplink subframes, which allows time for the BS to switch from transmit to receive mode. The uplink sub-frame is subdivided into request access and data transmission time intervals. In the request access interval, the SS transmits registering or bandwidth request packets. This period is a contention access mode so a collision resolving algorithm, such as ALOHA, must be provided. The BS will identify the successful users requests and allocate them bandwidth in the following frames. Next, the data transmission period in which the SS transmits their data as scheduled by the BS. In order to obtain a good protocol efficiency, request access packets have to be smaller than data packets and therefore their time intervals are usually called minislots. Figure 2.3. shows the uplink subframe structure.

REG BW SS-1 SS- SS-N

contention

access collision

gap

preamble Figure 2.3. Uplink subframe structure.

The physical layer coding and modulation functional blocks are depicted in figure 2.4. The decoding and demodulation processes have a reciprocal block diagram at the receiver side.

Page 20: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 20 of 84

Randomisation FECEncoder

ConvolutionalEncoder

MappingShaping andModulationg.

Preambledata to RF

Figure 2.4. Block Diagram of the Physical Layer.

First, the data payload is segmented into blocks of data designed to fit the proper FEC codeword size (256 bytes) after a convergence layer pointer is added. This pointer is used to identify the first MAC Protocol Data Unit (PDU) which starts in the present burst. Afterwards, a randomisation process is used to minimise the possibility of transmission of unmodulated carrier and to ensure adequate number of bit transitions to support clock recovery. Several error correction schemes have been proposed. Usually a Reed-Solomon RS(255,239) FEC is used as inner code while a 2/3 convolutional encoder with 2 tail-bits is used as outer code. Finally a burst preamble based in 16 byte CAZAC sequences is added in the uplink data burst. To maximise the utilisation of the air-link, the physical layer uses a multi-level modulation scheme. The modulation used shall be QPSK, 16-QAM or 64-QAM. The sequence of modulation bits shall be mapped onto a sequence of modulation symbols using Gray constellations. Prior to modulation, the I and Q signals shall be filtered by square-root raised cosine filters. The excess bandwidth factor shall be either 0.15, 0.25 and 0.35. Due to the large amount of spectrum available in the millimetre-wave band, several baud rate options and channel bandwidth have been chosen. Those baud rates are chosen to provide an integer number of time-slots per frame. The frame size choice is a trade-off between transport efficiency and latency. Table 2.1. summarizes the recommended frame parameters. Channel Size Roll-off

Factor Baud Rates Frame Size Time Slots

0.15 10.8 Mbaud 2 ms 5400 0.25 10 Mbaud 2 ms 5000 12.5 MHz 0.35 9.2 Mbaud 2 ms 4600 0.15 24.2 Mbaud 1 ms 6050 0.25 22.4 Mbaud 1 ms 5600 28 MHz 0.35 20.6 Mbaud 1 ms 5150 0.15 43.4 Mbaud 0.5 ms 5425 0.25 40 Mbaud 0.5 ms 5000 50 MHz 0.35 37 Mbaud 0.5 ms 4625

Table 2.1 Recommended frame parameters.

Page 21: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 21 of 84

2.2. Medium Access Control Layer As a network that employs a shared medium, there must be a mechanism to provide an efficient algorithm to implement the medium sharing. Within each sector, users must adhere to a transmission protocol which minimises contention between them and which enables the service to be tailored to the delay and bandwidth requirements of each user application. The considered MAC protocols support the following service primitives: MAC_CONNECT.request MAC_CONNECT.indication MAC_CONNECT.response MAC_CONNECT.confirm MAC_CHANGE.request MAC_CHANGE.indication MAC_CHANGE.response MAC_CHANGE.confirm MAC_DATA.request MAC_DATA.indication MAC_DISCONNECT.request MAC_DISCONNECT.indication MAC_DISCONNECT.response MAC_DISCONNECT.confirm The connection primitives are used by a BS or SS to request a connection activation. The disconnection primitives are used to request the termination of an active connection. The data primitives define the transfer of data between the MAC layers. Existing connections may be dynamically modified in their characteristics with the MAC_CHANGE primitives, for example, to reflect variable bandwidth requirements. As can be seen the MAC is connection-oriented, where a connection defines both the mapping between the BS and the SS and the Quality of Services (QoS) for the data exchanged during the connection. Every MAC PDU has the form of figure 2.5. Every PDU starts with a common fixed-length header, which mainly identifies the connection and informs about the PDU type and length. The header may be followed by a payload, which consists of zero or more sub-headers (for packing, fragmentation and bandwidth management purposes) and zero or higher-layer packets or protocol management messages. Finally,

Page 22: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 22 of 84

the MAC PDU may be CRC protected. The generic MAC PDU format is shown in figure 2.5.

MAC PDU common header

payload(optional)

CRC (optional)

Figure 2.5. MAC PDU generic format.

Several management messages are used in the MAC protocol. Only the maps and channel descriptors, which are sent in the control interval of the physical frame are described, as they are the more interesting for simulation purposes. Both message types must be transmitted by the BS periodically to define the channel characteristics and access maps. The uplink and downlink channel descriptors provide information to define the characteristics of the physical channel, including the following parameters: channel identifier, symbol rate, frequency, transmission power, preamble pattern, transition gaps, roll-off factor, modulation type, error-control information, etc. The uplink and downlink maps allocate the time-slots to access to the channel information for the SS. Two different high-layer PDU are allowed: fixed length packets such as ATM cells or MPEG2 frames, and variable length packets such as Ethernet or Internet packets. The subconvergence service may have a payload header suppression which reduces or completely deletes the packet header. Figure 2.6 shows the structure for a data MAC PDU

MAC PDU common header

packet header (w or w/o PHS)

packet payload (fixed or variable length)

Figure 2.6. MAC data PDU format.

Different data service flows are defined in order to improve the efficiency of the bandwidth management services. The basic services comprise:

Unsolicited Grant Service (UGS). UGS is designed to support real-time service flows that generate fixed size data packets on a periodic basis, such as T1/E1 and VoIP. The BS shall provide fixed size data grant burst types at periodic intervals.

Real Time Polling Service (rt-PS). It is designed to support real-time

service flows that generate variable size data packets on a periodic basis, such as MPEG video. The service must offer real-time, periodic, unicast request opportunities, which meet the flow’s real-time needs and allow the SS to specify the size of the desired grant.

Page 23: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 23 of 84

Unsolicited Grant Service with Activity Detection (UGS-AD). UGS-AD is designed to support UGS flows that may become inactive for substantial portions of time (i.e. tens of milliseconds or more) such as VoIP with silence suppression. The service provides data grant burst when the flow is active and unicast polls when the flow is inactive. This combines the low overhead and low latency of UGS with the efficiency of rt-PS. Though USG-AD combines UGS and rt-PS, only one scheduling service is active at a time.

Non Real Time Polling Service (nrt-PS).This service is designed to

support non real-time service flows that require variable size data grant burst types on a regular basis, such as FTP. The service offers unicast polls on a regular basis which assures that the flow receives request opportunities even during network congestion.

Best Effort (BE). It is intent to provide efficient service to best effort

traffic. In order for this service to work correctly, the request policy setting should be such that the SS is allowed to use contention request opportunities. This will result in the SS using contention request opportunities as well as unicast request opportunities and unsolicited data grant burst types.

There are several mechanism in which the SS can request for its

bandwidth needs to the BS:

Requests refers to the mechanism that a SS uses to indicate the BS that it needs upstream bandwidth allocation. A request may come as a stand alone header (specific for bandwidth request) or as a piggyback request. Because the adaptive modulation, all request shall be made in terms of the number of bytes needed to carry the MAC PDU. The BS may grant bandwidth explicitly to each connection or to all the connections belonging to the SS.

Polling is the process by which the BS allocates to the SS bandwidth

specifically for the purpose of making bandwidth requests. These allocations may be for an individual SS or for a group of them. A SS with a current UGS connection may stimulate the BS to poll it by sending a poll-me indication in the MAC PDU header.

The BS controls assignments on the uplink channel through the

uplink maps and determines which time-slots are subject to collisions. Collisions may occur during registering or bandwidth management periods. The method of contention resolution is based on a truncated binary exponential back-off algorithm, with the algorithm parameters controlled by the BS.

Page 24: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 24 of 84

The procedure for initialising a SS can be divided in several phases. On initialisation or after a signal loss the SS shall acquire a downlink channel, scanning through the system channels, using several non-volatile operational parameters. Once the SS has achieved synchronisation, it will acquire the channel control parameters and the downlink and uplink channel descriptors and maps. Finally, the BS will perform the registration, transfer the operational parameters and set-up the connections. For this purpose, all BS and SS have an unique MAC address such as the IEEE 48-bit directions. 2.3. Network Layer Aspects

Finally, there are some communication requirements between the different BS in the cellular network. These communications are needed to perform the handover process in the mobile scenario and also the data interchange for the centralised interference reduction algorithm. There are several standard solutions in order to communicate several BS to a CS. They are all based in the set-up of a digital traffic path over a wired network, providing the required bandwidth.

Acceptable solutions for this communications are based on ATM

switches connecting the BS and CS over User Network Interfaces (UNI) such as SONET/SDH, or UTP-3 links [Pol96] as shown in figure 2.7. The management entities in both BS and CS communicate trough ATM connections routed through a switch and based onto virtual paths. Specific or standard signalling virtual channels, such as Q.2931 can be used to carry the information [Itu97]. Other acceptable solutions are such based onto Internet based data networks as in GPRS [Pah00].

BSBS

ATMswitchATM

switch

CSCS

BSBS

UNI

UNI

Figure 2.7. ATM based BS to CS interconnection.

Page 25: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 25 of 84

3. System Models A review of existing system models is a key objective in order to obtain significant performance measures for the BAP. The models used for simulations are based on adaptations of well-known models to the OBANET scenario. The system parameters that must be modelled are radio-channel and antennas, traffic generation and users mobility. This models will be implemented as simulation libraries over and under the BAP. 3.1. Traffic Models Several traffic models have been proposed in order to model traffic patterns and evaluate the BAP for the simulation tool. Voice, video and data are intended with fixed-length packet (ATM cells) data generation, while Ethernet and TCP/IP models are related with variable-length (internet) data generation.

3.1.1. Voice Traffic A voice source generates a signal that follows a pattern of talk-

spurts separated by silent gaps. A speech activity detector can be used to detect this pattern. Voice packets are thus transmitted only during periods of speech activity to reduce the traffic and enable statistical multiplexing (with activity detection). Therefore, a voice source can be described by an On/Off model [Fri01]: the source alternates between the On state where the source generates packets at rate Rv (the encoder bit rate) and the OFF state where no packets are generated. The duration of On and Off states are modelled by exponential distributions with mean values ton and toff respectively. Table 3.1. summarises typical voice generation parameters used in the simulations.

Parameter Value

Source Bit Rate 64 Kbps Average On Time 1.00 s Average Off Time 1.35 s Activity Ratio 0.426 Packet Size 48 bytes Frame Time 0.5 ms

Table 3.1. Voice source model parameters. To model this behaviour, a two-states Markov Chain as shown in

figure 3.1 is used, where the “0” and “1” states correspond to the ON and OFF states, respectively.

Page 26: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 26 of 84

10

a01

a11a00

a10 Figure 3.1. A two-states Markov Chain.

The state transition probabilities parameters are related with the mean time on each state, as

01on a

1t = 0100 a1a −= 10

off a1t = 1011 a1a −= (3.1)

The simulation results can be summarised in figure 3.2, where the

voice packets per time generated are depicted. It can be observed that the pattern of talk-spurts are separated by silent gaps, as described above.

0 50 100 150 200 250 300 350 400 450 5000

1

Time (s)

Voice

Pac

kets

Figure 3.2. Trace of a voice packets source.

3.1.2. Video traffic

The video traffic may be modelled as a combination of On/Off sources [Fri01]. Each On/Off source alternates between the On and Off states independently, with states durations assumed to be exponentially distributed. Although this mathematical model is not representative of every type of variable bit-rate source, it provides an easy approach to evaluate the efficiency of the system protocols. The video source model parameters are defined in table 3.2.

Parameter Value Source Bit Rate 128/256/512 Kbps Number of On/Off 20 Average On Time 100 ms Average Off Time 200 ms Activity Ratio 0.33 Packet Size 48 bytes

Page 27: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 27 of 84

Frame Time 0.5 ms Table 3.2. Video source model parameters.

The average number of On sources and so the generated bit-rate

are related to the activity ratio of the sources, as

offon

onon tt

tNN

+= vonRNR = (3.2)

Some traces of this type of traffic, with different source bit-rate are

shown in figure 3.3. The low latency due to the short frame duration limits the number of video packets that must be stored before transmission.

0 5 10 150

0.5

1

1.5

2

2.5

3

3.5

Time (s)

Vid

eo P

acke

ts

0 5 10 15

0

1

2

3

4

5

6

7

8

9

10

Time (s)

Vid

eo P

acke

ts

Figure 3.3. Trace of video packets sources: 128 kbps (left) and 512 kbps

(right). 3.1.3. Data traffic

The model for data traffic generation represented by batches of packets that arrive at a certain rate [Fri01]. The inter-arrival time between groups of packets is assumed to be exponentially distributed with a mean value of td. The number of packets in each batch arrival is given by a truncated gamma distribution. The data source model parameters are summarised in table 3.3.

Parameter Value Source Bit Rate 19.2 Kbps Average Inter-arrival Time 100 ms Beta Parameter 3 Mean Burst Length 5 packets Packet Size 48 bytes Frame Time 0.5 ms

Table 3.3. Data source model parameters

Using these values, the traffic trace shown in figure 3.4 is obtained.

Page 28: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 28 of 84

0 50 100 150 200 250 300 350 400 450 5000

2

4

6

8

10

12

14

16

Time (s)

Data Pac

kets

Figure 3.4. Trace of a data packets source.

Data traffic implies no need of real-time transmission, which means

that in a real environment simulation (a packet may not be transmitted as it is generated) the generation peaks can be absorbed by the source without congestion. 3.1.4. Ethernet Traffic

The traffic produced by a large number of traffic sources has self-similar [Dra96] behaviour over large time scales. Self-similarity indicates that the behaviour of a process is very similar in time. This self-similarity indicates a form of invariance regarding changes of time-scales [Rob97].

Ethernet measurements have long term dependence, at least over 4 or 5 orders of magnitude. If we represent the number of Ethernet packets arriving in a time-interval, the statistics of the number of packets looks the same for a 10, 100 or 1000 times period and is distinctively different from a pure noise. However a stabilisation in the dispersion index has been observed, indicating a lack of self-similarity. So a short-term dependence process would be enough to model LAN traffic.

This kind of traffic can be modelled with pseudo long-range dependent processes, which are able to model aggregated traffic over several time-scales. We will approximate a hyperbolic decaying auto-covariance function as a sum of exponentials using a n-states Markov chain with transition probabilities aij as depicted in figure 3.5.

10

a02

a11a00

a10

2 N

a20

aN0

a0N

a01

a22 aNN

Page 29: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 29 of 84

Figure 3.5. A n-states Markov chain. The suggested Markov chain may be characterised by means of

the transition probabilities matrix, where the q parameter gives an estimation in the degree of data correlation (Hurst parameter) [Rob97], as given by.

−−

−−−−

=

−−

−−

1n1n

22

1n21n2

)a/q(1...00)a/q(...............0...)a/q(10)a/q(0...0a/q1a/qa/1...a/1a/1a/1...a/1a/11

A (3.3)

As in the voice or video models, an exponentially-distributed dwell time is equivalent to the Markov chain. The transition probabilities are related to the average state-stay time as follows.

∑≠

=

jiij

i a1t

(3.4)

For instance, considering the four-states Markov Chain, with the transition matrix given by 3.5, a good agreement for this equivalence is obtained, as shown in table 3.4.

a=1.99 q=0.6737

=

961199.000038801.00885389.00114611.000661457.0338543.0

126894.0252519.0502513.0118075.0

A (3.5)

Mean Dwell Time Simulated Theoretical

state 0 1.13467 1.133883

state 1 2.898625 2.953834

state 2 9.028729 8.725165

state 3 26.369877 25.772530 Table 3.4. Equivalence of Markov chain and exponential dwell time.

3.1.5. TCP/IP Traffic Several models have been proposed for emulating Internet traffic. Among them Interrupted Poisson Process (IPP) and Markov Modulated

Page 30: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 30 of 84

Poisson Process (MMPP) are the most typical [Bau01, Raa01]. IPP model is a contribution to the IEEE 802.16 standard and defines a generic traffic model) that provides self-similar traffic modelling. IPP may be considered as a particular case of MMPP.

The MMPP takes into account the arrival process of service requests and it is a doubly stochastic Poisson process. The arrival process is determined by the state of a continuous time Markov chain with k states. When the chain is in state k, the arrival process follows a Poisson distribution with rate λi (i=0,...,N). For instance, a two-state MMPP, may be characterised by four parameters: c1, c2, λ1 and λ2 where The mean sojourn times of the chain in states 1 and 2 are 1/c1 and 1/c2 respectively. The average number of arrivals is related to these parameters as,

[ ]21

12212211 cc

ccPPR

+λ+λ

=λ+λ= (3.6)

This model can be simplified assuming that the arrival process has

stationary increments. The stationary of increments implies that the statistical characteristics depend solely on the interval and not on the time. The MMPP takes the auto-correlation of the increments into account.

The IPP generates self-similar traffic found in Ethernet and Internet traffics. This model characterises the correlation (not just the variance) of aggregated traffic with a minimum number of parameters (i.e. only three parameters implicitly related to average, variance and correlation). It is also analytically solvable when fed into a queuing system by using the matrix geometric method.

IPP considers also a two-states model. During the On state λ packets per unit of time, with a Poisson distribution, are generated. During the Off state no packets are generated. The transition probability rates α are related with the state dwell or sojourn times. To model the self-similarity traffic found in Ethernet and internet traffic samples, four IPP are superimposed. Each process has different parameters in order to represent four different time scales. Fig. 3.6 show these different time scales.

Page 31: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 31 of 84

ON OFFλ1

time

ON ON ON ON

OFF OFF OFFIPP1 Covers short time scale

ON OFFλ1

time

ON ON

OFFIPP4 Covers long time scale

Figure 3.6. Short and Long term IPP components. The TCP/IP model will take into account aspects from both MMPP and IPP models. A four-states Markov, as depicted in figure 3.7, chain will be considered where the state 0 corresponds to the Off and the other states corresponds to the On, but with different Poisson-distributed packet arrival rates. The parameters chosen for our simulations are defined in table 3.5.

00a

33a

03a

0

1

01a

10a

11a

2

02a20a

22a

3

30a

λ1 λ2λ3

00a

33a

03a

0

1

01a

10a

11a

2

02a20a

22a

3

30a

λ1 λ2λ3

Figure 3.7. MPP philosophy of the TCP/IP traffic model.

State Packet Rate

0 0 1 2.679 2 1.698 3 1.234

Table 3.5. TCP/IP traffic model parameters. The generated data rate is related to the transition probabilities and the average packet arrival rate of each state, as given by

33003

032

2002

021

1001

01aa

aaa

aaa

aR λ

+

+

+

= (3.7)

Page 32: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 32 of 84

The results obtained using this model must be scaled to give the appropriate data rate for the test cases. From Telcordia and Lawrence Berkeley Labs data [Bau01], the average packet size is 192 bytes. Thus, the number of packets per second for a 100 Kbps data rate is 65.104. So all the parameters of the basic model must be scaled to ensure this is the mean value for the states addition, as shown in table 3.6.

State Packet Rate 0 0 1 46.9662 2 29.7680 3 21.6335

Table 3.6. 100 Kbps scaled TCP/IP parameters.

A trace of the obtained Internet packet pattern is shown in figure 3.8.

Figure 3.8. Trace of a 100 Kbps internet traffic source.

Finally, the size in bytes of each data bursts may be modelled using a Pareto distribution [Jab00]. A trace of the packet length can be seen in figure 3.9.

0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 00

5 0

1 0 0

1 5 0

Time (s)

Pack

et le

ngth

(byt

es)

Figure 3.9. Packet length with a=1 and b=0.9 Pareto distribution.

Taking into account both packets per time generation and variable packet length the instantaneous bit-rate for an internet connection can be represented, as shown in figure 3.10.

Page 33: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 33 of 84

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 00

1

2

3

4

5

6

7

8

9

1 0

x 1 04

T im e (s )

Bits

/s

T rá fi c o E t h e rn e t

Figure 3.10. Binary rate for an Internet traffic source.

3.2. Antenna Models

Realistic antenna models are essential to develop and study the BAP. Differentiated BS and SS antenna models are needed: BS antenna will be directional and its radiation pattern will change dynamically due to beamforming, while SS antenna will be simple and will differ from the mobile (omnidirectional) to the fixed (directional) scenarios. The antenna model is required in order to evaluate the propagation loss equation, given by

( ) ( )θ+−−θ+= RrainpropTtxrx GAAGPP (3.8)

where GT is the gain of the BS/SS antenna in the downlink/uplink and GR is the gain of the BS/SS antenna in the uplink/downlink. It should be noticed that the gain depends on the azimuth angle and therefore on the antenna model. Only azimuthal variation in the radiation pattern will be considered, whilst remains constant in elevation. 3.2.1. BS antenna

The BS antenna radiation pattern will change dynamically based on a Photonic Integrated Beamformer (PIB). The beamforming is achieved by using an integrated multi-wavelength laser in combination with a delay-line based on a combination of fast InP-based switch matrix and dispersive fibres [Oba01a].

Beam switching and beam shaping capabilities are possible using this PIB. Beam switching is achieved changing the possible combinations of the delay-line time. For instance, with a 3 switches delay line, 8 possible optical delays are selected, corresponding to 8 possible beam positions. On the other hand, beam shaping can be achieved controlling the bias current of each laser in the multi-wavelength laser, which varies each wavelength output power.

Page 34: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 34 of 84

The antenna radiation pattern may be modelled as a linear equidistance antenna array with N sinusoidal or omnidirectional elements. In antenna arrays the normalised antenna power pattern can be described by the following equation,

( ) ( )( ) ( ) 2

maxmax

2

)(θθ

θθθ

e

e

dAF

dAFt

⋅= (3.9)

where AF stands for the array factor, θmax is the direction where the

array radiation is maximum, and de is the radiation pattern of each individual element. Elements have a static radiation pattern pointing to the broadside direction. In order to attenuate secondary lobes of the antenna array, the antenna element will be directive covering a 90º sector. The simplified model for this element will follow a parabolic beam pattern with a secondary lobe floor level of 20 dB, whose radiation pattern is given by,

( )

θ−

≤θ≤

−θ⋅−=θ

of rest

20626-

2090

9012d

2

e (3.10)

The single-element radiation pattern is depicted in figure 3.11.

0.2

0.4

0.6

0.8

1

30

210

60

240

90

270

120

300

150

330

180 0

Figure 3.11. Parabolic radiation pattern of an element.

The array factor may be fully defined by the amplitudes (an) and phases (nα) of each element and the distances between them, as

( ) ( )∑−

=

α+θ⋅=θ1N

0n

coskdjnn eaAF

λπ

=2k (3.11

) So, the antenna directivity in dB for any direction, in terms of the maximum directivity, D, and the normalised array power pattern, can be expressed as follows,

Page 35: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 35 of 84

( ) ( )( )θθ tDD ⋅⋅= 10log20 (3.12)

Beam switching may be achieved by varying the phase of the elements while beam shaping is obtained by modifying the amplitudes of the elements. As the elements are in fixed positions, the inter-element spacing is not a control parameter. The pointing or scan angle of the array is given by the progressive phase of each element as follows,

( )θα cos⋅−= kd (3.13)

In the meantime, the beamwidth may be modified by acting over

the array amplitudes and therefore the array gain is modified. But these beam pattern changes also affect to other antenna parameters such as SLL or Front-Back ratio (F/B). The current amplitude distribution is characterised by the Ap parameter which stands for the ratio in dB between the current amplitude fed to the central element and that to the external element, and is given by,

( )0102/

010 log20log20 a

aa

AN

p ⋅=

⋅= (3.14

)

if the central element current amplitude is normalised to 1. The remaining current amplitudes may be obtained as follows,

( )( )

−=

−=−−

⋅−=

−− 1,...,2/

12/,...,0 12/

1

1

00

NNaa

NnNa

naa

nnN

n (3.15)

For instance, table 3.7 shows different current distributions for several values of Ap.

Ap a0 a1 a2 a3 -20 0.1 1 1 0.1 -10 0.3162 1 1 0.3162 -5 0.5623 1 1 0.5623 0 1 1 1 1 5 1.7783 1 1 1.7783

Table 3.7. Current distributions for a four-elements array. 3.2.2. SS antenna Simplified models will be considered for the SS antennas. In the fixed scenario a highly directive antenna will be taken into account. This antenna may be easily described by its pointing direction, the

Page 36: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 36 of 84

beamwidth and the SLL. The fixed SS antenna radiation pattern is given by Eqn. 3.16 and has been depicted in figure 3.12.

∆+≤≤

∆∆∆=

−−

−−

ofrest SLL

2

2

- 4)(

3max

3max

33

θ

θθθ

θθ

φθπ

θdBdB

dBdBd (3.16)

1

2

30

210

60

240

90

270

120

300

150

330

180 0

Figure 3.12. Fixed SS antenna radiation pattern.

On the other hand, for the mobile scenario, an omnidirectional antenna in the azimuthal range will be considered, as depicted in figure 3.13.

0.5

1

1.5

30

210

60

240

90

270

120

300

150

330

180 0

Figure 3.13. Mobile SS antenna radiation pattern.

3.3. Radio Channel Models Radio channel is the signal path between the transmitting and the receiving antennas. The simulation of system performances for the BAP requires channel models that accurately describe relevant features of the radio propagation at the given frequency [Cra99]. The radio

Page 37: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 37 of 84

channel model choice will have a great impairment in most of the beamforming strategies. 3.3.1. Radio Propagation at 40 GHz The link budget of a radio system is expressed in Eqn. 3.17, where the radio channel imposes the level of attenuation. In fact, the attenuation term is an statistical variable due to phenomena like rain, multipath fading or shadowing. However, a deterministic approximation can be done. The main attenuation mechanism at the 40 GHz band are the free space propagation losses at 40 GHz can be calculated as follows and it only depends on the transceivers distance, as given by

)log(204412.124)(log204.92)( 10 KmKmGHzPROP ddfdBA ⋅+=⋅⋅+= (3.17)

The other main effect for attenuation are the rain losses, provided that a LoS link exists. Rain loses have a large-scale component which can be considered to be deterministic but also a short-scale fading due to atmospheric inhomogenities or changes in the electromagnetic properties (i.e., reflectivity) of the receiver surrounding scatters, such as building surfaces [Xu99].

Large-scale attenuation, also called average rain attenuation, can be obtained using the following expression,

7.2dRaA Kmb

h/mmRAIN +⋅⋅= (3.18)

In this equation R are the rain intensity and d is the radio path length affected by the rain zone. The a and b coefficients depends on several conditions such as frequency, polarisation, geographical zone [Itu94].

The small-scale fading is a random variable which has a Rice

probability distribution [Xu99]. Rice distribution can be parameterised by the total received power and the K factor, the ratio of coherent to incoherent powers.

h/mmR04.088.16K ⋅−= (3.19)

3.3.2. Multipath and Shadowing Fading

Page 38: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 38 of 84

The propagation of a radio signal in a non free-space environment produces a random variability in the strength of the received signal power. This variability is commonly divided into a fast or multipath fading and a slow or shadowing fading. Fast fading is caused by the alternate constructive and destructive summation of multipath signal components as the terminal or the surrounding scatters in the radio path move. Shadowing and blockage occur mainly due to man-made degradations, such as users intercepting the beam or vehicles traffic into the streets, or also natural impairments, such as trees motion due to the wind.

A great number of distributions which describe the statistics of the signal under fading condition, are described [Yac00]. However the statistical behaviour of the signal propagation in the 40 GHz frequency band is not certainly clear at this moment [Cra99]. Different approaches can be found in the literature. Although K factors were a good measure of the variability of the signal [Pit00], the Rice distribution predicted deeper fades than those which were observed [Pap97]. Table 3.9 summarizes empirical K factors obtained in [Pit00]

Scenario K factor high dense urban 10.13 typical urban 10.38 suburban 14.28 rural 14.86

Table 3.8. Rice K factor in several environments. In general, log-normal or Nakagami-m distribution are recommended for high directive gain antennas providing LoS path and limiting non-LoS paths and the Rice distribution is recommended for high gain antennas providing LoS path and a large number of non-LoS paths [Zha00].

While multipath fading and rain attenuation may be overcome by using appropriate fade margin and advanced transmission techniques, blockage effects are more difficult to mitigate, resulting in high bit error rates and temporary service unavailability. In the fixed scenario the BS and SS antennas will be located high on rooftops and blockage effect will occur with lower probability than in the mobile scenario.

Different approaches may be considered in order to characterise the changes in the signal statistical behaviour when radio path is blocked. A log-normal shadowing statistical has been used under non-LoS conditions and a free space transmission with exponent of two is used under LoS [Gon01]. Also, a change from Rice to Rayleigh distributions have been considered in the same conditions [Fla00].

Page 39: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 39 of 84

3.3.3. Statistical Radio Channel Model

To characterise the statistical nature of the radio signal a two states Markov chain, as shown in figure 3.14, has been chosen. Similar models have been used in mobile satellite systems in order to evaluate the service quality and reliability [Xie00]. The chosen propagation model is based in the Lutz model [Lut91] for a Land Mobile Satellite (LMS) propagation channel which has commonalities with the OBANET scenario. This model has been adapted to take into account the special characteristics of the radiowave propagation in the millimetre-wave band and the performance measurements to be evaluated in the simulations, which require of the estimation of the variations of the signal strength level.

LOS NLOS

NLP

LNP

NLP−1LNP−1 LOS NLOS

NLP

LNP

NLP−1LNP−1

Figure 3.14. Two-states Markov chain radio channel model.

The LoS or Good state represents the normal behaviour of the radio link. In this state a deterministic value will be calculated based on Eqns. 3.17 and 3.18. In a second step, the received power will be randomly varied in a short range based on a statistic distribution. Several statistic distributions will be simulated depending on the scenario (fixed/mobile), the environment (outdoor/indoor) and weather conditions. The non-LoS or Bad state represents an anomalous behaviour of the channel such as blocking. An additional and severe power drop will be added to the link budget and also the statistical nature will change. In both states, BER can be obtained from the received power depending on the modulation scheme. For instance, the BER distribution of a Grey-encoded 64-QAM signal transmitted over a Rayleigh fading channel with coherent detection is given by equation 3.20, in which the parameter is the average SNR [Sam97],

γ+−⋅=

o/7111

247P (3.20

)

where γ0 is the average SNR. γ0 is equivalent to Eb/N0 when a root

Nyquist filter is employed as the transmitter and receiver filter. The difficulty of using stochastic modelling is the parameter estimation. In the

Page 40: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 40 of 84

proposed channel model, transition probabilities may be obtained from equilibrium state probability of being in the non-LoS state, PNLOS,

NLLN

LNNLOS PP

PP

+= (3.21

)

This parameter also represents the unavailability of service due to the blockage effect and it will be different depending on the scenario (mobile/fixed) and environment. In equation 3.21 both probabilities may be estimated with a certain degree of freedom. PNLOS is a key parameter which makes the difference between the mobile and the fixed scenario. The dwell time in each state will follow an exponentially distributed random variable with a mean value related to the transition probabilities. 3.4. Mobility Models

Mobility models become very important to properly simulate the performance of the BAP focusing in the tracking algorithm but also to study handover and radio resource management [Mit09]. Therefore, mobility models are used to obtain parameters such as the call blocking probability, the handover blocking probability or the call dropping used to evaluate the performance of the system protocols, or to evaluate signalling and teletraffic parameters [Lam97] as mean sojourn or dwell time, the handover rate and the channel holding time.

Next generation wireless networks are envisaged as small cells, which lead to smaller mean sojourn times. Also, since the transmitting power is reduced and LoS is required, obstacles such as buildings, trees, hills, etc. will have a greater effect on cell size and shape. The result is that cells tend to be less regular in shape and more variable throughout the coverage area. This large variation in cell size and shape, together with the shorter mean sojourn time, suggest the consideration of mobility models which consider both the heterogeneity of the environment and the subscriber units (pedestrians, cars) [Bra99]. Three possible models have been considered, depending on the scenario. A random or a street model for out-door environments and a model for in-door environments. The microscopic mobility behaviour of the users in these models is its most important characteristic, so they can properly simulate the performance of the beamforming strategies. 3.4.1. Out-door Mobility The random model can be used to model users in a suburban or rural scenario or, for example, to model the mobility of an HDTV camera covering a golf match. In this model the position of the user is continuously updated with its velocity and directional factor as follows,

Page 41: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 41 of 84

( )( )rnd0

rnd0costvyycostvxx

α⋅⋅+=α⋅⋅+=

(3.22)

This directional factor is a random variable with a given distribution in the 360º range. Mainly the uniform distribution will be used. Figure 3.15 schematises this model.

(x0,y0)

(x,y)

v

range of newpossible positions

Figure 3.15. Two dimensional random motion.

The street model can be employed to model pedestrian and

vehicles in a urban scenario, as depicted in figure 3.16. This model considers a grid of possible positions according to the streets in an urban environment [Mar97]. An important parameter is the turn probability in the cross of two directional ways.

Pt/2

Pt/2

1-Pt

Figure 3.16. Street grid movement and turn probability.

This model can be used for both mobile and pedestrian users as they only considering that they only differ in their speed [Bra99]. Some parameters for a pedestrian user are given in [Ets97], as summarised in table 3.9.

Parameter Value Mean speed 3 km/h

Page 42: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 42 of 84

Speed Standard Deviation 0.3 km/h

Probability to change speed at position update

0.2

Probability to turn at cross street 0.5

Table 3.9. Pedestrian user parameter. 3.4.2. In-door Mobility The in-door model considers several offices rooms communicated by a corridor [Ets97-Mbs95]. Those models are of interest to consider WLAN scenarios. Mobile users can be stationary or moving with a constant speed in a room or into the corridor to another room [Mbs95]. The motion along the corridor will be neglected as is short time and has no practical interest in the beamforming strategies performance, as schematised in figure 3.17.

Each mobile user can be described by three possible: the Sleep state in which it does not move and with a high probability, the Move in the Room state in which it moves randomly inside the room, and the Move to other Room state, in which follows a given path to another randomly chosen room. For each mobile the initial state is set to Sleep. This is represented in the states diagram depicted in figure 3.18.

Office Room

Corridor

Figure 3.17. In-door motions.

Sleep Move toother room

Move inroom

outinp ,

outslp ,

inslp ,

)(1 ., outinslin pp +−

)(1 ,, outslinsl pp +−

sloutp ,

slinp , inoutp , )(1 ,, inoutslout pp +−

Page 43: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 43 of 84

Figure 3.18. In-door motion states diagram. The time spent in the corridor (Tcorridor) is related to the time spent in the office room (Toffice) and the ratio of in-room users (r), assuming equal flows to and from the office rooms, and it is given by

( )

rr1TT officecorridor

−= (3.23

) The time that each mobile user spends in each state can be approximated to an exponential distribution. The mean time spent at each state can be expressed as follows,

sleepofficesleep PT ⋅=λ

)P1(T sleepofficein −⋅=λ

corridorout T=λ

(3.24)

where Psleep is the probability of being in the sleep state. On the

other hand, transition between states is a random process. The transition probabilities are related to the spent times as follows.

out,slin,slsleep pp

1+

out,insl,inin pp

1+

sl,outin,outout pp

1+

(3.25)

To determine the transition probabilities, it must be considered that the transition probability from any state to the Sleep state should be very high and the transition probability from any state to the Move to other Room state should be very low. For instance, with a moving time of 30 s., a rate of 85% in-room users and a sleeping probability of 0.9 the mean times and transition probabilities are summarised in the tables 3.10 and 3.11.

Parameter Value sleepλ 27 s.

inλ 3 s.

outλ 5.29 s.

Table 3.10. In-door motion mean times.

Page 44: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 44 of 84

in,slp 0.037 sl,inp 0.33 sl,outp 0.039

out,slp 0.001 out,inp 0.0034 in,outp 0.15

Table 3.11. In-door motion transition probabilities.

Page 45: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 45 of 84

4. Protocols Simulation and Specification In the OBANET project, the performance of beam adaptation protocols (BAP) is going to be tested for fixed and mobile broadband wireless access networks. A simulation tool will be developed in WP3. This tool will be used to evaluate the proposed beamforming strategies that will be carried out by the beamformed BS by means of the BAP. Next a description of the simulation, specification and programming techniques that will be used in the project are discussed. 4.1. Discrete Event Simulations Discrete-event simulation [Law91] concerns the modelling of a system as it evolves over time by a representation in which the state variables change instantaneously at separate time instants. Mathematically speaking, it might be said that the system state can change only at certain time instants. These instants are those at which an event occurs, where an event is defined as an instantaneous occurrence that may change the state of the system. One of the most important aspects in a discrete-event simulation is the time-advance mechanism (the mechanism to advance simulated time from one value to another). In a simulation model, the variable that gives the current value of a simulated time is the simulation clock. Two principal approaches have been suggested for advancing the simulation clock: next-event time advance and fixed-increment time advance. The first approach is used by all major simulation languages and by most people coding their model in a general-purpose language, and it is the approach used in our simulation. The second is a special case of the former. With the next-event time-advance approach, the simulation clock is initialised to zero and the times of occurrence of future events are determined. The simulation clock is then advanced to the time of occurrence of the most imminent (first) of these future events, at which point the state of the system is updated to account for the fact that an event has occurred. Then the simulation clock is advanced to the time of the new most imminent event, the state of the system is updated, and future events time are determined, and so on. This process of advancing the simulation clock from one event time to another is continued until eventually some pre-specified stop condition is satisfied. Since all state changes occur only at event times for a discrete-event simulation model, periods of inactivity are skipped over by jumping the clock from event time to event time.

Page 46: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 46 of 84

The following components will be found in most discrete simulation models using the next-time advance approach:

System state. The collection of state variables necessary to describe the system at a particular time.

Simulation clock. A variable giving the current value of

simulated time.

Event list. A list containing the next time when each type of event will occur.

Statistical counters. Variables used for storing statistical

information about system performance.

Initialisation routine. A subprogram to initialise the simulation model at time zero.

Timing routine. A subprogram that determines the next event

from the event list and then advances the simulation clock to the time when that event is to occur.

Event routine. A subprogram that updates the system state

when a particular type of events occurs (there is one event routine for each event type).

Library routines. A set of subprograms used to generate random

observations from probability distributions that were determined as part of the simulation model.

Report generator. A subprogram that computes estimates

(from the statistical counters) of the desired measures of performance and produces a report when the simulation ends.

Main program. A subprogram that invokes the timing routine to

determine the next event and then transfers control to the corresponding event routine to update the system state appropriately. The main program may also check for termination and invoke the report generator when the simulation is over

The flow-control process for a generic next-event time-advance

simulation process is depicted in figure 4.1. The simulation begins at time 0 with the main program invoking the initialisation routine, where the simulation clock is set to zero, the system state, the statistical counters and the event list are initialised. After

Page 47: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 47 of 84

control has been returned to the main program, it invokes the timing routine to determine which type of event is most imminent. If an event of type i is the next to occur, the simulation clock is advanced to the time that event type i will occur and control is returned to the main program. Then the main program invokes event routine i, where typically three types of activities occur: (1) the system state is updated to account for the fact that an event of type i has occurred; (2) information about system performance is gathered by updating the statistical counters; and (3) the times of occurrence of future events are generated and this information is added to the event list. Often, it is necessary to generate random observation from probability distributions in order to determine these future event times. After all processing has been completed, either in event routine i or in the main program, a check is typically made to determine if the simulation should be finished. If so, the report generator is invoked from the main program to compute estimates of the desired performance measurements and to produce a report. If it is not the case, the control is returned to the main program and the cycle is repeated until the stop condition is eventually satisfied.

Start

0. Invoke the initialization routine1. Invoke the timing routine

2. Invoke event routine i

1. Update system state2. Update statistical counters

3. Generate future events and add to event list

1. Set the simulation clock = 02. Initialize system state and statistical counters

3. Initialize event list

1. Determine the next event tpe, say i2. Advance the simulation clock

Generate randomvariates

Is simulation over?

1. Compute estimates of interest2. Write report

Stop

i

0

1

No

2

Yes

Main programInitialization routine

Event routine i

Timing routine

Library routines

Report generator

Figure 4.1. Flow control for the next-event time-advance approach.

4.2. Protocols Specification The Specification and Description Language (SDL) is an object-oriented, formal language defined by The International

Page 48: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 48 of 84

Telecommunications Union-Telecommunications Standardisation Sector (ITU-T) as recommendation Z.100 [Itu93]. The language is intended for the specification of complex, event-driven, real-time, and interactive applications involving many concurrent activities that communicate using discrete signals.

SDL is a graphical specification language that is both formal and object-oriented. The language is able to describe the structure, behaviour, and data of real-time and distributed communicating systems with a mathematical rigour that eliminates ambiguities and guarantees system integrity. It has a graphic syntax that is extremely intuitive. The most important characteristic of SDL is its formality. The semantics behind each symbol and concept are precisely defined. Above all, the great strength of SDL lies in describing large real-time systems.

Systems described in SDL consist of many processes running

simultaneously which communicate with each other via signals. Each process is described by an extended finite state machine. The state machines are labelled extended since variables and timers may also be defined in processes.

SDL has a number of advantages compared to other high-level languages. SDL has a rich grammar that describes behaviour and is unambiguous. Therefore, it is possible to build tools for the simulation of SDL systems and for the validation of formal characteristics. In short, this means that errors are detected at a very early stage.

The precision and formality of SDL provide the possibility for tool-supported code compilation into lower level languages, such as C/C++ or Java. This means that the SDL system may be translated into an executable application without manual coding, leading to shortened development time and increased quality. As a side effect, due to the readability of SDL diagrams, the SDL specification becomes the documentation in itself, ensuring simplified maintenance and post-development.

SDL is used worldwide for the development of all kinds of complex, communicating systems. In the telecommunications field, SDL is the language of choice for the development of a broad range of software and hardware. Examples are 3G products, cellular phones, switches, WAP stacks, Bluetooth devices, GPRS systems, DECT phones, radio systems, network management platforms and network services systems. Other strong examples are telecommunication standards like UMTS, GSM, ISDN, etc.

Page 49: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 49 of 84

4.3. Simulations Structure The programming structure of the system simulations is schematised in figure 4.2. This structure has been widely employed [Hon95, Ber95, Ber95a, Pie97].

The simulation tool will be written in an object oriented language such as C/C++ or Java, allowing parallel processing and direct SDL specification. It consist of several modules and components:

The simulation parameters (simulation time, frame length, number of active users and their location, etc.) will be introduced by the input interface. The simulation kernel establish the initial conditions of the current simulation scenario with these parameters and launch the simulation.

The operation of the simulation tool will be controlled by the

Discrete-event Simulation Scheduler Process. It is based in the use of event tables which rule the event-driven simulation process.

Traffic generation library. This module will allow to set which kind

of traffic generates each user. Therefore, the system performance could be analysed for different traffic conditions.

Kernel

Traffic ModelLibrary

Channel ModelLibrary

Mobility ModelLibrary

Antenna ModelLibrary

...

...

BSN

BS1

SSN

SS1 ...

...

Discrete-Event Simulation Scheduler Library

FilesGraphs

Animations

OutputInterface

InputInterface System

Parameters

Figure 4.2. Simulation Tool Structure.

Page 50: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 50 of 84

Radio Channel Library. This module will allow to choose the characteristics of an active radio link. So the effects of the channel impairments in the link quality may be evaluated.

Mobility Library. This module controls terminals mobility within

the cell (speed, type of movement, …) and also its geographical location and distribution.

Antenna Library. This module will calculate the antenna

parameters, mainly its radiation pattern and gain, as a function of the scan angle and network conditions for both BS and SS.

BS and SS behaviours. This is the main portion of the simulation

code and will implement the BWA protocols including the BAP.

The simulation kernel will integrate all the above mentioned modules, establishing the required control and communications among them and generating the outputs for a simulation post-processing and results obtaining.

Page 51: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 51 of 84

5. Beamforming Functionalities and Strategies

In this section, the smart functionalities and strategies of the OBANET proposed optically beamformed BS antennas that will be implemented and analysed by simulations are presented. Also a description of the proposed algorithm and its possible implementation is given. 5.1. Beam Switching

Beam-switching comprises the capability of the BS beamformed antenna of blanketing different directions in order to provide full coverage in the service area. A real-time control algorithm will provide dynamic beam positioning among the resource allocation schemes in such a way that a sectorial coverage is provided. This concept in combination with DCA techniques results in a dramatic reduction of the system interferences. 5.1.1. Switching and Steering

The proposed beamformer can switch the beam into a number of fixed positions that must be designed for each beamformed antenna. These positions are related to the phase difference of the signals that feed each elemental antenna. The number of positions is determined by the number of switches in the beamformer.

In the system scenario, the beam switches among N possible pointing directions in order to provide coverage to the whole 90º sector with a directive beam in a timeslot basis. The beam position at a given instant depends on the location of the user to whom the current timeslot is assigned. The different allowable positions of the beam, must be designed in a way that a certain degree of overlapping between sub-sectors is ensured, in such a way that network issues as handover or user tracking are simplified. Given a sectorial coverage of Θ degrees, the following design relationship must be accomplished for the antenna beamwidth (∆θ-3dB).

NdB3Θ

≥θ∆ − (5.1)

If the different positions are evenly allocated along the sector, the

angular direction (measured respect to the array axis) for each maximum gain position is given by,

N21i2i

maxΘ−

=θ (5.2)

Page 52: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 52 of 84

Equations 5.1 and 5.2 are represented in figure 5.1.

Switching among allowed beam positions is done by acting on the beamformer delay profiles and therefore over the antenna signal phases. There is also the possibility of modifying the pointing angle around the designed angular position by slightly steering the beam. This may be achieved by acting over the emission wavelengths of the beamformer multiwavelength source.

θmax

θmax

θmax

θmax

1

2

3

4

Θ

Figure 5.1. Beam switching angular positions.

Beam-steering may allow for instance to improve the quality of

service and to optimise the network throughput if a group of user sites are geographically close or if there are hot-spots with great amount of generated traffic. The beam may also be steered to avoid/reduce an interference, as depicted in figure 5.2.

Group of users

Traffic Hot-Spot

Figure 5.2. Beam-steering functionality.

Both switching and steering functionalities require the modification of several beamformer parameters, so the beam points to a different angular position. As seen in chapter 3, the ∆θ–3dB is assumed to keep constant independently of the pointing direction. However, the pointing angle can not be varied without changing simultaneously the radiation

Page 53: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 53 of 84

pattern and its main characteristics: beamwidth, directivity, Side-Lobe Level (SLL) and Front-Back ratio (F/B), as depicted in figure 5.3. It may noticed from figure 5.3 that when the angular position changes from θmax=90º to θmax=135º, the SLL, ∆θ–3dB and F/B increase.

But also a pointing error due to the elemental antenna radiation pattern occurs. A more sophisticated control than acting only on the signal phases at the antenna input becomes necessary. One of the main objectives of the beamformer control algorithm will be to remain beam conditions constant independently of the pointing direction, especially the directivity and therefore the bandwidth which are approximately inversely proportional.

0.2

0.4

0.6

0.8

30

210

60

240

120

300

150

330

Figure 5.3. Radiation pattern modification due to pointing.

As an important remark there should be noticed that when the

beam changes from an angular position to another it requires of a switching operation, which is made at a certain speed and in a certain time. The switching time, defined as the time spent as the beam changes its scan position, influences the protocol efficiency as introduces some necessary guard times. Therefore, fast switching capability has a great importance in the proposed scenario [Oba01]. 5.1.2. Beamwidth Control

Several practical considerations regarding the radiation pattern behaviour as the beam is steered must be taken into account in order to accurately model the proposed system functionalities. In the OBANET scenario, 90º sectors are proposed. However, as explained in section 3.4, occasionally it could be desirable to increase/reduce the sector size to adapt the system to different traffic conditions. Assuming that the direction of maximum gain is θmax=90º and 90º sectors, θmax will vary

Page 54: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 54 of 84

between 45º and 135º, but if we assume that the sector can be widened up to 120º, θ max will vary then between 30º and 150º. In addition, the radiation pattern is symmetrical, as the coefficients of the array are always real and positive. Therefore, the radiation pattern will vary identically and symmetrically if the beam points from 30º to 90º or from 90º to 150º, Therefore, in the forthcoming work in WP3, only the radiation pattern parameters modifications when the pointing direction varies from 90º to 150º degrees will be taken into account.

Figures 5.4, 5.5 and 5.6, show the variation of the ∆θ–3dB as a

function of the pointing angle for 4, 8 and 16 elements antenna arrays and for several current distributions. It can be seen that by increasing the number of elements a reduction in the ∆θ–3dB is obtained. For a given pointing angle, the ∆θ–3dB reduces as the Ap factor (as defined in section 3.2) grows, and for a given current distribution, the ∆θ–3dB increases as the scan angle moves away from the broadside direction. In addition, the ∆θ–3dB variations may be assumed to be approximately independent on the current distribution at the antenna input ports.

Also in figures 5.4, 5.5 and 5.6, it is represented the SLL variation as a function of the pointing angle for 4, 8 and 16 elements antenna arrays and for several current distributions. The SLL decreases as the pointing angle moves from the broadside direction and also as Ap increases. When the antenna array has few elements, the SLL decreases sharply and an inflection point appears. This is due to a change in the most significant side lobe which grows quickly. This inflection point appears for those pointing angles closer to the broadside direction as Ap decreases.

In any case, the F/B is always greater than the individual antenna element one.

90 100 110 120 130 140 15020

25

30

35

40

45

50

55

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Bea

m-w

idth

( º)

Scan angle (º)

90 100 110 120 130 140 15020

25

30

35

40

45

50

55

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Bea

m-w

idth

( º)

Scan angle (º) 95 100 105 110 115 120 125 130 135 140 145 150

0

5

10

15

20

25

30

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Scan angle (º)

SLL

(dB

)

95 100 105 110 115 120 125 130 135 140 145 1500

5

10

15

20

25

30

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Scan angle (º)

SLL

(dB

)

Figure 5.4. Beamwidth and SLL for an 4-elements antenna array.

Page 55: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 55 of 84

90 100 110 120 130 140 15010

12

14

16

18

20

22

24

26

28

30

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Bea

m-w

idth

(º)

Scan angle (º)90 100 110 120 130 140 150

10

12

14

16

18

20

22

24

26

28

30

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Bea

m-w

idth

(º)

Scan angle (º)

95 100 105 110 115 120 125 130 135 140 145 1500

5

10

15

20

25

30

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Scan angle (º)

SLL

(dB

)

95 100 105 110 115 120 125 130 135 140 145 1500

5

10

15

20

25

30

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Scan angle (º)

SLL

(dB

)

Figure 5.5. Beamwidth and SLL for an 8-elements antenna array.

90 100 110 120 130 140 1504

6

8

10

12

14

16

18

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Bea

m-w

i dth

(º)

Scan angle (º)90 100 110 120 130 140 150

4

6

8

10

12

14

16

18

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Bea

m-w

i dth

(º)

Scan angle (º) 95 100 105 110 115 120 125 130 135 140 145 150

0

5

10

15

20

25

30

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Scan angle (º)

SLL

(dB

)

95 100 105 110 115 120 125 130 135 140 145 1500

5

10

15

20

25

30

Ap=-20dBAp=-10dBAp=-5dBAp=0dBAp=5dB

Scan angle (º)

SLL

(dB

)

Figure 5.6. Beamwidth and SLL for an 16-elements antenna array.

As mentioned above, the antenna element affects to the overall array antenna radiation pattern as the scan angle varies. This results into a pointing error (the scan angle is slightly lower than the expected) and a reduction of the antenna gain as the scanning direction moves away from the broadside direction. The former degradation is greater when a fewer elements array antenna is used, and also when low values for Ap are chosen. Regarding the gain reduction, the opposite behaviour is observed. In this case, the reduction of the antenna gain value varies with the number of antenna elements due to the pointing error. Both the pointing angle error and the gain reduction effects are depicted in figures 5.7 and 5.8, respectively. In figure 5.7, it may be observed that the scan angle error is negligible for 8 and 16-elements arrays. However, the net antenna gain reduction becomes higher in these configurations.

Page 56: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 56 of 84

90 100 110 120 130 140 15090

100

110

120

130

140

150

N=4N=8N=16

Scan angle desired (º)

Scan

an g

le o

btai

n ed

(º)

90 100 110 120 130 140 15090

100

110

120

130

140

150

N=4N=8N=16

Scan angle desired (º)

Scan

an g

le o

btai

n ed

(º)

Figure 5.7. Worst case pointing angle error due to element antenna.

Once all these effects are characterised an algorithm to maintain

the ∆θ–3dB constant independently on the scan angle is developed. The mechanism is quite simple and consists of the control of the current distribution, by varying the Ap parameter.

90 100 110 120 130 140 150-6

-5

-4

-3

-2

-1

0

N=4N=8N=16

Scan angle (º)

Gai

n a t

tenu

atio

n (d

B)

90 100 110 120 130 140 150-6

-5

-4

-3

-2

-1

0

N=4N=8N=16

Scan angle (º)

Gai

n a t

tenu

atio

n (d

B)

Figure 5.8. Worst case gain reduction due to element antenna.

A polynomial approximation between beamwidth (in the

broadside direction) and Ap has been obtained from the simulations. A first order polynomial approximation is valid in the available Ap range, from –20 to 6 dB. These expressions are summarised in table 5.1.

N Approximation Range 4 90

dB3p 0423.353575.1A −θ∆⋅+−= 6.4146.22 90dB3 ≤θ∆≤ −

Page 57: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 57 of 84

8 90dB3p 5521.466275.3A −θ∆⋅+−= 48.1846.11 90

dB3 ≤θ∆≤ −

16 90dB3p 9043.534208.8A −θ∆⋅+−= 78.878.5 90

dB3 ≤θ∆≤ −

Table 5.1. Estimated approximations for Ap and ∆θ–3dB (@ 90º). However, the approximations shown in table 5.1 are valid only for

the broadside direction since they do not take into account the scan angle error. As the ∆θ–3dB variation with the scan angle is almost independent on the Ap value, an equation for the relationship between the ∆θ–3dB in the broadside direction from a given ∆θ–3dB in any other direction may be obtained

<θ<θ∆≥θ≥θ∆−θ∆

=θ∆−

−− º100

º100

maxdB3

maxerrdB390dB3 80º

80º (5.3)

The variation of the beam-width against the mispointing from the broadside to any other direction is given by the expressions summarised in table 5.2,

N ∆θerr 4 2

maxmaxerr 7842.72515.00017.0 θ⋅+θ⋅−=θ∆

8 2maxmaxerr 6555.346899.00034.0 θ⋅+θ⋅−=θ∆

16 2maxmaxerr 1622.275195.00025.0 θ⋅+θ⋅−=θ∆

Table 5.2. Approximations for beamwidth and scan angle. The beamwidth errors and the SLL obtained by using the

approximations of table 5.2 are depicted in figures 5.9, 5.10 and 5.11. The scan angle error is greater than 1º (which is inherent to the approximations). This is due to the limitations in the current distribution variation range (Ap) that has been limited to the –5 and 5 dB range, which is coincident with the experimentally available within the OBANET test platform. This constraint implies that if a certain ∆θ–3dB is required for a certain scan angle, it will be achieved only if the required Ap is within the available range. As can be seen from figures 5.9 to 511, the error increases if the scan angle moves away from the broadside direction.

90 100 110 120 130 140 1500

1

2

3

4

5

6

BW =30BW =28BW =26

Scan angle (º)

Bea

m-w

d ith

err

or (

º)

90 100 110 120 130 140 1500

5

10

15

20

25

BW =30BW =28BW =26

Scan angle (º)

SLL

(dB

)

90 100 110 120 130 140 1500

5

10

15

20

25

BW =30BW =28BW =26

Scan angle (º)

SLL

(dB

)

Page 58: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 58 of 84

Figure 5.9. Beamwidth estimation error and SLL for a 4-elements array.

90 100 110 120 130 140 1500

1

2

3

4

5

6

BW =17BW =15BW =14.5

Scan angle (º)

Bea

m-w

dith

err

or (

º)

90 100 110 120 130 140 150

0

5

10

15

20

25

BW =17BW =15BW =14.5

Scan angle (º)

SLL

(dB

)

90 100 110 120 130 140 1500

5

10

15

20

25

BW =17BW =15BW =14.5

Scan angle (º)

SLL

(dB

)

Figure 5.10. Beamwidth estimation error and SLL for a 8-elements array.

90 100 110 120 130 140 1500

1

2

3

4

5

6

BW =8.5BW =8BW =7

Scan angle (º)

Bea

m-w

dith

err

or (

º)

90 100 110 120 130 140 150

0

5

10

15

20

25

BW =8.5BW =8BW =7

Scan angle (º)

SLL

(dB

)

90 100 110 120 130 140 1500

5

10

15

20

25

BW =8.5BW =8BW =7

Scan angle (º)

SLL

(dB

)

Figure 5.11. Beamwidth estimation error and SLL for a 16-elements array.

If a greater Ap range was technologically available, the ∆θ–3dB error

would be only produced by the estimated approximations. For instance, figure 5.12 represents the error for a 4-elements array with a designed ∆θ–

3dB of 35º and a Ap range from -20 to 6 dB.

90 100 110 120 130 140 1500

1

2

3

4

5

6

Scan angle (º)

Bea

m-w

dith

err

or (

º)

Figure 5.12. Beamwidth error for an extended currents range.

Page 59: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 59 of 84

Optimal ∆θ–3dB taking into account the currents range constraint and the SLL behaviour have been chosen. The values are given in table 5.3.

N ∆θ-3dB 4 30º 8 15º 16 7º

Table 5.3. Optimal beamwidth values.

As explained above a beam fine-tuning or beam steering could be necessary to perform several system functionalities. This fine-tuning could be achieved, in the proposed beamformer, acting over the lasers wavelengths. This action will change the delay and therefore the progressive phase.

Two main assumptions can be done about the variation of the scan angle as a function of the progressive phase. The first one is that scan angle dependence with the progressive phase can be considered approximately linear in the 30º-150º margin. Secondly, a scan angle variation of one degree implies a variation around three degrees of the progressive phase. Table 5.4. shows several values of the relative delay variation needed for different scan angles variations at the 40 GHz frequency.

Scan Angle Variation

Progressive Phase Relative Delay

±1º ±3º ±0.208 ps ±3º ±9º ±0.625 ps ±6º ±18º ±1.250 ps

Table 5.4. Phases and delays for the scan angle fine-tuning functionality. 5.1.3. BWA Protocols Modification

Some modifications in the standard BWA protocols, mainly in the PHY and MAC layers (such as frame or system messages) must be carried out in order to introduce the beam-switching technique:

First, we must introduce mechanisms allowing for the broadcast of control messages such as frame mappings, which is clearly infeasible with a directed antenna. This implies that this broadcast information has to be retransmitted in each sub-sector. In order to minimise the protocol overhead the whole-sector common control information and the sub-sector control information should be transmitted in different control messages.

Page 60: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 60 of 84

On the other hand, contention access time intervals have to be provided for each beam position, allowing for the registration of new users. The mechanism that controls which slot/mini-slot is used in contention mode has to take into account that the main sector has been divided in several sub-sectors.

In [Oba01] two main approaches for modifying the standard MAC

protocols were identified: inter-frame or frame based switching and intra-frame or slot based switching. In the first approach a different and complete frame is transmitted per each beam position. This solution does not suppose any protocol modification except the fact that different control messages and information must me managed for each frame. However, this approach has a poor management of the bandwidth resources and has problems with the latency requirements.

In the second approach the switching is performed inside the frame, pointing the beam to the required direction. This solution requires the mentioned control information replication and contention periods for each sub-sector. It is more efficient regarding management of the bandwidth resources, but the high number of beam switchings of this solution requires of low switching times. Figure 5.13 schematises the slot-based beam-switching approach.

subsector 1

subsector 2

subsector N

pream bleM A Ps

CD s dow stream data ups tream dataREG s

BW s

frame Figure 5.13. Slot based beam-switching approach.

Assuming a uniform density of users in the cell so that terminals may

be found in each beam, the minimum number of position switches needed to provide coverage is 7N, where N is the number of possible beams (related to the number of switches in the beamformer), and 7 comes from the different frame portions (as can be seen in figure 5.13). Then the efficiency loss due only to the switching time can be expressed as 7*N*Ts/Tf , where Ts and Tf are the switching and frame times. As can be seen in figure 5.14, a switching time in the range of few tens of nanoseconds or less nanoseconds is needed to avoid an efficiency loss over 2%. This figure increases as the number of subsectors is higher.

Page 61: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 61 of 84

5µs

10

20

30

40

50

60

70

80

90

1µs 0.5µs 0.1µs 50ns 10ns 5ns

sw itching time

effi

cien

cy lo

se (%

)

16 sector

8 sector

Figure 5.14 Efficiency loss due to switching time.

The above is a simplified estimation and worse scenarios may be

expected in a realistic implementation, i.e. when using adaptive modulation, DCA techniques, etc., where the number of switches may be in the order of 10N or greater. However, the casuistry of these situations is too wide and should be more precisely estimated in the simulation results that will be reported in D3.2. 5.1.4. Beam Switching Algorithm

At a first glance, the beam-switching control algorithm is very simple. Only the angular position in which every registered SS is located is necessary. This information may be obtained with a location algorithm in the registering phase for each SS jointly with a tracking algorithm for the mobile SS. If the position of the SS and the resources allocation table used by the MAC protocol are given as parameters, a control signal for the phase and the amplitude of the beamformer may be easily generated. Two auxiliary tables should be used, one relating the different SS and its angular position, and another relating the different directions and the needed signal phase and amplitudes.

However, due to the directive feature of the beams, this algorithm may be refined in order to minimise the Co-Channel Interference (CCI). This improvement will be introduced by modifying the resources allocation mechanisms. The algorithm may be implemented as distributed between the different BS or as centralised in a CS. In both cases the algorithm is based onto the following premises.

Instantaneous power measurements for the non-assigned time intervals in the BS. This measurements are a prediction of the interference level in that time instant.

Page 62: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 62 of 84

Interference levels estimation in the assigned time intervals, by means of the corrected errors by the FEC and the packet loss rate.

If the interference level is higher than a given threshold, the

assigned time instants will be reallocated to others in which the measured interference level follows a given criteria [Che96] (least interfered instant, least interfered instant below a given threshold, highest interfered instant bellow a threshold, first interfered instant bellow a threshold, etc.)

If a centralised control is done, every BS will report the

interference and location information to the CS. The assignation and interference reduction will be done in such a way that the BS beams pointing configuration in the networks gets minimised.

5.2. Tracking and Locating

One of the most important features in next generation broadband mobile wireless access networks is to be able to determine the mobile user position and to track it with an antenna beam [Lib99]. With this feature, the location of terminals may be updated. But it also allows a better network performance using adaptive antenna capabilities by means of pointing the BS antenna in the desired direction, and the development of advanced traffic management strategies.

Location becomes necessary when new users are registered in the network and also when a registered user has been idle for a long time. In these cases, the BS must evaluate the angular position of the user in the sector. The locating process can be carried out as a scanning trace in the whole sector. As explained above, contention mode time intervals in the uplink direction must be provided to each sub-sector. Therefore, the angular position of the terminal is obtained at the registering time-interval, in which the terminal accesses to the system. There are several parameters/aspects to consider, that may affect to the definition of the location process, as for instance, the number of already registered users in this sub-sector and limitations for traffic due to hot-spot in certain zones, etc.

Several approaches have been used to implement users tracking. Among them, two different possibilities are being considered: mono-pulse tracking and signal processing.

The use of pilot tones for broadband wireless access system protocols has been previously proposed [Aca00]. The BS expects the SS

Page 63: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 63 of 84

to send a pilot tone at a predetermined instant, as shown schematically in figure 5.15, such as the preamble of a data burst. In response to this pilot the BS may adapt the beam pattern to the user motions. A similar technique has also been used in [Sta96, Mbs95], where RADAR monopulse tracking or additional beams were used. The algorithm was tested using several vehicle trajectories and speeds and it was proved that the algorithm was very robust and quite insensitive to changes in velocity or trajectory.

PREAMBLE

PILOT

DATA

Figure 5.15. Pilot tone in the data burst preamble.

This pilot tone may be used to monitor the power received from a

user terminal by the BS, and therefore to implement the user tracking. The tracking algorithm will be the following and is shown schematically in figure 5.16.

BSSS

Figure 5.16. Monopulse tracking algorithm.

During the preamble, the BS measures the power of the pilot tone in the current SS position Pi, but also in the adjacent sub-sectors Pi-1 and Pi+1. Using the measurement results the terminal position may be estimated as follows.

i

1i1iP

PPd −+ −= (5.4)

Ideally, three possible cases may be considered: (1) d=0 means

the user is within the current subsector; (2) d>0 means that the user moves to the next subsector; and (3) d<0 means that the user goes to the former subsector. To make the decision of changing the user dedicated beam a threshold value should be considered. The more adequate levels should be d=1 and d=-1 because these situations

Page 64: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 64 of 84

indicate that the user is in the half between two adjacent beams and that it will probably move to any adjacent subsector.

The value of d is independent on the BS to SS distance and also of the beams overlapping. It is also independent on the power fluctuations due to the channel impairments due to their geographical neighbourhood but a fast switching time becomes necessary. This feature is also necessary to minimise the protocol overhead due to the tracking algorithm. In addition, an analysis of the latest power measurements may be used as a prediction of the terminal motion in the coverage area. This prediction may be useful in network functionalities as for the handover process is concerned.

There is also the possibility of using conventional signal-processing algorithms to estimate the Direction of Arrival (DoA) as proposed as fall-back option [Oba01a] for OBANET project. It requires of additional hardware and processing equipment to implement traditional subspace based estimation methods, as MUSIC or ESPRIT.

Once the tracking algorithm is defined, there are several parameters/aspects to be taken into account that may affect the system efficiency and accuracy. It is necessary to fix a hysteresis margin in the beam switch to avoid the well-known “ping-pong” effect. This situation may happen when the terminals move in such a way that d values around –1 and 1 are obtained continuously in a short time.

The tracking algorithm is physically implemented in the uplink direction and must be iteratively repeated to get the right track. Therefore, there is a minimum time-interval between two consecutive iterations in the tracking algorithm It can be calculated as follows,

v2

tgr2T

dB3

scan

θ∆⋅⋅

=

(5.5)

In (5.5), r is the distance between BS and SS and v is the terminal speed. This minimum time is calculated in the worst case illustrated in figure 5.17. It does not take into account the switching and processing time

Page 65: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 65 of 84

v

t t+Tr

rr

v

t t+Tr

rr

Figure 5.17. Tracking scanning time worst case.

Figure 5.18 shows Tscan as a function of the user speed and the

number of beams. The distance value r is ten meters corresponding to the most critical case. It can be seen how Tmin becomes more restrictive when the user speed increases but also when the number of sub-sectors increases. Also some scan time values for several speeds are summarised in table 5.5.

0 20 40 60 80 100 120 1400

10

20

30

40

50

60

70

80

90

100

Speed (km/hr)

Tim

e (m

s)

Figure 5.18. Tracking scan time for 4, 8 and 16 beams (right to left).

Terminal Speed Beams 5 Km/h 60 Km/h 120 Km/h

4 286.4338 ms 70.7427 ms 11.9347 ms 8 141.8276 ms 11.8190 ms 5.9095 ms 16 70.7427 ms 5.8952 ms 2.9476 ms

Table 5.5. Tracking scan times values.

Page 66: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 66 of 84

Finally, the maximum range error in the tracking algorithm is given by equation (5.6), in which R is the cell radius. This scenario is illustrated in figure 5.19.

θ∆⋅⋅= −

2tgR2e dB3 (5.6)

23dB−θ

⋅⋅= −

22 3

maxdBtgRerr

θ

R: Cell radius

SSReal position

SS

Stored position

23dB−θ

⋅⋅= −

22 3

maxdBtgRerr

θ

R: Cell radius

SSReal position

SS

Stored position

Figure 5.19. Tracking estimation error worst case.

However, if beam overlapping is taken into account, as shown schematically in figure 5.20, the error might be reduced when a mobile is detected by two adjacent beams. In this case, the user position can be estimated as the mean-value angle between adjacent sectors, if a small overlapping exists. Using such estimation, the tracking error is given by,

e)S1('e ⋅−= s>0.5

eS'e ⋅= s<0.5 (5.7)

The minimum error is produced when S=0.5. In that case, the

maximum errors are reduced in a 50%, as shown in table 5.6.

Page 67: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 67 of 84

TdS ⋅

( ) TdS ⋅−1

Td

No overlapping

OverlappingTdS ⋅

( ) TdS ⋅−1

Td

No overlapping

OverlappingTdS ⋅

( ) TdS ⋅−1

Td

No overlapping

Overlapping

Figure 5.20. Tracking algorithm with overlapping estimation.

Beams w/o Overlapping with Overlapping 4 397.825 198.9125 8 196.983 98.4915 16 98.254 49.1270 Table 5.6. Maximum error in the tracking algorithm.

5.3. Beam Shaping

Under certain conditions the modification of the antenna radiation pattern shape in order to manage traffic hot-spots or to overcome link impairments or interfering sources could be useful. Shaping the beam means to vary the beamwidth and therefore the directivity, to modify the SLL, to implement nulling, etc. The beamwidth variations of the beamformed antenna are limited by the number of elements and the ranges for the signal amplitudes and phases. Next, some scenarios in which beam-shaping could be useful are described:

User dedicated beams. A user with high QoS requirements may require that a high directivity beam to be generated. This feature needs that a network level protocol ensures that the increase in directivity does not degrade the CIR of adjacent sectors/cells. Another application of this feasibility is to counter temporary link fadings due to man-made or natural obstacles obstructing the LoS link.

Cell-range extension. Under certain network conditions, it could

be of interest to extend (in order to temporary provide service to users of an adjacent sector) or reduce the array gain (if users

Page 68: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 68 of 84

are near the BS, the signal power may be reduced and CIR for adjacent sectors/cells increased)

User grouping. If users are geographically close, the beam may

be widened to provide service to all of them and therefore reduce the number of switches in the coverage area. This will reduce the protocol overhead as the control information is transmitted less times during a frame.

5.3.1. Gain Control The gain of a high directive antenna is related to the beamwidth in the azimutal and elevation planes as follows.

dBdB

G33

4

−− ∆∆≅

φθπ (5.8)

This allows increasing the antenna gain by narrowing the antenna beamwidth, for example, to compensate for any power drop in the link. Those power drops are assumed to be detected by the network monitoring capability. Figure 5.20 shows the relationship between antenna gain and the azimuthal beamwidth. There should be noticed that this relation has been obtained for the broadside direction. A reduction as the position moves away from this direction will occur as explained previously

5 10 15 20 25 30 3516

18

20

22

24

26

28

30

BW =10BW =15BW =20

Beam-width (º)

Gai

n(d

B)

5 10 15 20 25 30 3516

18

20

22

24

26

28

30

BW =10BW =15BW =20

Beam-width (º)

Gai

n(d

B)

Figure 5.21. Gain control acting over the beamwidth.

In order to obtain a given beamwidth at a given angular position, a current control technique was studied above. Figure 22 depicts the relation between the obtained beamwidth and the current Ap factor and also the obtained SLL. There can be seen that a limited beamwidth range is obtained due to the beamformer constrains in the current variations. Also, that the variation range is higher as less elements antenna arrays are used.

Page 69: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 69 of 84

-6 -4 -2 0 2 4 65

10

15

20

25

30

N =4N =8N =1 6

Bea

m-w

idth

(º)

Relative amplitude (dB)-6 -4 -2 0 2 4 6

5

10

15

20

25

30

N =4N =8N =1 6

Bea

m-w

idth

(º)

Relative amplitude (dB) -6 -4 -2 0 2 4 6

5

10

15

20

25

30

N=4N=8N= 16

SLL

(dB

)

Relative amplitude (dB) Figure 5.22. Beamwidth control acting over the current distribution.

A possible solution to the beamwidth variation range limitation is to

obtain an array with less elements by switching off some of the elements. In such a way, a rough beamwidth variation may be implemented whilst a beamwidth fine-tuning may be done by acting over the current distribution. However, a reduction for the antenna gain as several elements are switched-off is obtained, which imposes extra requirements in the power amplifiers dynamic range. A careful design should be done to ensure that the power budget is kept constant. As mentioned above, the beamwidth range depends on the beam position but also on the currents range. In tables 5.7 and 5.8 the beamwidth variation range is given for several positions and ranges. It can be seen that the range is not very wide for the 10 dB current range.

N Variation range at 90º Variation range at 150º 4 22.84º<∆θ-3dB<28.72º 31.90º<∆θ-3dB<36.83º 8 11.62º<∆θ-3dB<14.04º 20.36º<∆θ-3dB<23.55º 16 5.86º<∆θ-3dB<6.96º 11.30º<∆θ-3dB<13.21º

Table 5.7. Beamwidth ranges for the –5 to 5 dB current variation.

N Variation range at 90º Variation range at 150º 4 22.46º<∆θ-3dB<41.6º 31.56º<∆θ-3dB<46.59º 8 11.46º<∆θ-3dB<18.48º 20.12º<∆θ-3dB<28.60º 16 5.78º<∆θ-3dB<8.78º 11.50º<∆θ-3dB<16.28º

Table 5.8. Beamwidth ranges for the –20 to 6 dB current variation.

Additional possibilities in the beamwidth modification could be considered, for instance, the use of current distributions different than the raised-triangular or the variation in the element spacing, but these techniques are not envisaged for the project. 5.3.2. Beam Shaping Algorithm

Page 70: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 70 of 84

The dynamic modification of the beamwidth is another capability

that must be introduced in the beamformer control algorithms. This functionality may be exploited to overcome channel degradations. Traditionally this has been realised using power-flow control, but with the beamformed antenna it might be carried out using a more directive beam.

As antenna directivity is related to the beamwidth in the case of directed antennas. From that relationship, we may obtain an estimation of the improvement introduced in the link budget by modifying the beam shape. Also in this case, the beamformer control algorithm must have a table relating the different beam positions and the possible amplitudes of the signals for each beam-width.

On the other hand, the detected power must be monitored at the BS in real-time. This kind of degradations (i. e. an artificial obstacle in the LoS) is characterised by a rough, very constant and long term variation in the channel attenuation and in the received power. However, the range of variations in the signal levels is limited in the proposed beamformer, so only a low range of variations in the beamwidth may be obtained. And as seen in the previous section, this range will depend on the angular direction of the beam. 5.4. Cell Reconfiguration

Another strategy that may be implemented is to vary dynamically the sector shape. In this case the proposed mechanism may be described as follows:

Although 90º sectors are implemented, the beamformed

antenna is designed is such a way that each sector may be capable of covering a highest or smallest sector.

As depicted in figure 5.21 (left), in a normal mode of operation

each BS sector will serve a 90º coverage area but this range may be dynamically varied increasing or reducing the number of active switched beams per sector, as shown in figure 5.21 (right).

Page 71: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 71 of 84

Figure 5.23. Cell reconfiguration.

However, there is a limitation in the maximum angle that may be

obtained, due to the arising SLL when moving away from the broadside direction. Using the above referred antenna models, maximum angles of 110º are obtained. 5.5. System Monitoring

Network monitoring is required for implementation of most of the functionalities provided by the beamformed network. There are several ways of monitoring the channel state, as:

Pilot tones.

Training sequences.

Power measurements in control or data transmission intervals

Higher level layer error controls that report to the physical layer when the number of errors exceeds those allowed by the protocol.

Reports from users. The SS monitors the received power from

the BS and reports the BS of those measurements. In order to implement monitoring functionalities, it is necessary that the physical layer provides means for performance monitoring, as SNR, BER/FER, power levels, among others [Hyp01]. ITU-T G826/821/827 and M2100 series ITU-T-Recommendations will be applied for performance monitoring purposes.

In the simulation tool, power control and monitoring will be simulated employing the antenna and propagation models described in chapter 3 of this deliverable. The obtained values will be used to

Page 72: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 72 of 84

implement the required functionalities, as beam-switching, beam-shaping and cell reconfiguration.

Power measurements: Both direct received signal power measurements or pilot-tone power measurements will be analysed in the simulations.

CIR estimation: direct measurements of BER will allow to calculate the carrier-to-interference ratio (CIR) level depending of the modulation format used.

Page 73: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 73 of 84

6. Conclusions and Remarks

In this deliverable, the specifications of all the aspects related to the implementation of beam adaptation protocols (BAP) under the simulation tool to be developed in WP3 (D3.2) are provided. Those BAP must be compliant with emerging standardisation proposals.

The main conclusions of this deliverable are as follows,

1. Physical models of the SS and BS antennas and the radio propagation channel, as well as traffic generation and users mobility models have been proposed based on adaptation of existing and well-known models to the OBANET scenarios.

Traffic models: due to the multimedia nature of next generation fixed and mobile broadband wireless access networks, voice, data, video and Internet (TCP/IP) traffic generation models must be taken into account in the simulations. A mixed MMPP/IPP model has been proposed to model Internet traffic generation.

Antenna model: both SS and BS antenna models have been defined. The SS model takes into account variations on the beam pattern as the beam switches from one position to another. Algorithms to control beam pattern parameters have been proposed. Those algorithms make use of the possibility of varying the current amplitude distribution fed to the BS antenna. The limitations imposed by the PIB that will be tested at the OBANET platform have been taken into account.

Radio-channel model: a 2-states Markov model has been chosen to model the radio propagation channel. Using this model, it is possible to simulate the channel behaviour in LoS and non-Los conditions (rain fading, temporary obstacles, etc.) and to obtain received power estimations that will be used to implement beamforming strategies and network monitoring. Network monitoring will be implemented using outputs from the physical layer, as power levels, BER measurements, and so on.

Mobility model: Indoor and outdoor mobility models have been chosen to evaluate the impact of users mobility in the beamformed BS antennas performance and to test the proposed users tracking and location algorithms.

2. A description of the simulation tool that is going to be implemented to evaluate the proposed beamforming strategies has been provided. A discrete-event simulation tool is being developed using SDL and an object oriented programming language (C/C++ or

Page 74: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 74 of 84

Java). The simulation tool will make use of the models described in section 3 of this deliverable.

The proposed beamforming functionalities and strategies performance will be evaluated using the simulation tool in order to characterise the performance improvement in both fixed and mobile BWA networks, under different traffic and environment conditions.

3. Several beamforming functionalities and strategies (beam switching, beam-shaping, user tracking and location and cell reconfiguration) have been specified and their implementation issues discussed. Algorithms to control the beam parameters as the beam-switches have been proposed and studied, as well as a description of how those strategies are going to be implemented in the simulation tool through the beam adaptation protocols.

Page 75: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 75 of 84

Appendix. Statistical Distributions A.1. Probability Distributions Next follows several probability distributions used in the system models. A.1.1. Geometric The Geometric distribution gives the dwell time in each state for the two-states Markov chain. This distribution defines the number of successes for a given number of independent Bernouilli trials.

x)p1(p)x(p −= (A.1)

0 1 2 3 4 5 6 7 8 9 1 00

0 .0 5

0 .1

0 .1 5

0 .2

0 .2 5

p = 0 .2 5

Figure A.1. Geometric mass function.

A.1.2. Poisson The Poisson distribution gives the process arrival rate in each state for a Markov Modulated Poisson Process. It defines the number of events that occur in an interval of time when the events are occurring at a constant rate.

!x)x(p

xe λ=

λ− (A.2)

Page 76: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 76 of 84

0 1 2 3 4 5 6 7 8 9 1 00

0 .0 5

0 .1

0 .1 5

0 .2

0 .2 5

λ = 2

Figure A.2. Poisson mass function.

A.1.3. Exponential The Exponential distribution is the only continuous distribution with the memoryless property, and it is the equivalent for the geometric distribution.

β−−= /xe1)x(f (A.3)

0 1 2 3 4 5 6 7 8 9 1 00

0 .1

0 .2

0 .3

0 .4

0 .5

β = 2

Figure A.3. Exponential density distribution. A.1.4. Gamma The Gamma distribution is a generic case for the exponential distribution. For a positive integer, the gamma is called Erlang-m distribution.

)a(x)x(f

/x1aa eΓ

β=

β−−− (A.4)

Page 77: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 77 of 84

Figure A.4. Gamma density distribution.

A.1.5. Pareto

The Pareto distribution is a skewed, heavy-tailed distribution that is usually used to model the distribution of incomes.

1a)xa(a)x(f

++= (A.5)

Figure A.5. Pareto density distribution.

A.2. Fading Distributions Next follows several distributions used for the radio-channel fading estimations. A.2.1. Lognormal distribution

In a real propagation ambient environment, a transmitted signal traverses different environments, therefore the received envelope can be written as follows.

( )γ+⋅σ= xexpr2 (A.6)

Page 78: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 78 of 84

In this equation x has a unit normal distribution and σ and ϒ are the process standard deviation and mean value. After a standard transformation of variables there can be found the following expressions.

( )

σ−⋅

⋅σ⋅π=

w2rln2exp

r22rp 2

2log

σ−⋅=

2expww

2 ( )w2ln=γ

(A.7)

The correspondent power density distribution can be expressed as

follows.

( )

σ−⋅

⋅σ⋅π=

wwln

21exp

w21wp 2

2log (A.8)

0 5 10 150

0. 05

0 .1

0. 15

0 .2

0. 25

0 .3

0. 35

power

3.1=σ&

1=σ&

5.0=σ&

Figure A.6. Lognormal density distribution.

A.2.2. Rayleigh distribution

The Rayleigh distribution characterizes an environment dominated by multipath propagation, where the partial waves are assumed to have a homogeneous phase distribution. A signal envelope with a Rayleigh distribution is written as follows.

222 yxr += (A.9)

In this equation x and y are the in-phase and quadrature components of a narrowband Gaussian processes. Therefore, the signal envelope distribution can be expressed as follows.

( )

⋅−

⋅=w2

rexpwrrp

2rayleigh

2σ=w

(A.10)

Page 79: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 79 of 84

The correspondent power density distribution can be expressed as

follows.

( )

−⋅=

wwexp

w1wprayleigh (A.11

)

0 5 10 150

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

Power Figure A.7. Rayleight density distribution.

A.2.3. Rice distribution

The Rice distribution characterises an environment where a direct component is superimposed onto the scattered waves, therefore the signal envelope distribution can be expressed as follows. A signal envelope with a Rice distribution is written as follows.

( ) 222 yaxr ++= (A.12)

Therefore the signal envelope distribution can be expressed as

follows.

( ) ( ) ( ) ( )

+⋅

+

−−⋅+

=w

k1k2rIrw2k1kexp

wk1rrp o

2rice

( )22 2/ak σ= ( )k1w 2 +σ=

(A.13)

In this equation k is the Rice factor, corresponding to the ratio of

the power of the direct wave and the scattered waves and I0 is the zeroth-order modified Bessel function. The correspondent power density distribution is given by the following expression.

( ) ( ) ( ) ( )

+⋅

+

−−⋅+

= ww

k1k2Iww

k1kexpw

k1wp orice (A.14)

Page 80: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 80 of 84

0 5 10 150

0.05

0.1

0.15

0.2

0.25

0.3

0.35

power

K=10dB

K=15dB

K=12.5dB

Figure A.8. Rice density distribution.

Page 81: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 81 of 84

References [Aca91] Acampora A. S., Chu T., Dragone C. and Gans M. J., “A metropolitan area radio system suing scanning pencil beams”, IEEE Transactions on Communications, pp. 141-151, vol. 39, no. 1, January 1991.

[Aca00] Acampora A. S. and Krishnamurthy S. K., “A New Adaptive MAC Layer Protocol for Broadband Packet Wireless Networks in Harsh Fading and Interference Environments”, IEEE/ACM Transactions on Networking, vol. 8, no. 3, pp. 328-336, June 2000.

[And96] Anderson B. L., Collins S. A., Klein C. A., Beecher E. A. and Brown S. B., “Highly parallel optical true time delay device for phased array antennas”, in Proceedings of the Sixth Annual ARPA Symposium on Photonic Systems fo Antenna Aplications, Monterey, USA, March 1996.

[Bau01] Baugh C. R., “Traffic Model for 802.16 TG3 MAC/PHY Simulations”, IEEE 802.16 Broadband Wireless Access Working Group contribution, 2001.

[Ber95] Bergholm P., Honkanen M. and Häggman S. G., “Simulation of a Microcellular DS-CDMA Radio Network”, IEEE Vehicular Technologies, pp. 838 –842, 1995.

[Ber95a] Bergholm P., “Development of a DS-CDMA Radio Network Simulator”, Technical Report, Technology Development Centre, Finland, December 1995.

[Blo01] Blogh J. S., Cherriman P. J. and Hanzo L., “Adaptive antenna array assisted dynamic channel allocation techniques”, IEEE Journal on Selected Areas in Communications, pp. 305-311, vol. 19, no. 2, February 2001.

[Bra99] Bratanov P. I., “User Mobility Modeling in Cellular Communications Networks”, Ph.D. Thesis, Technical University of Vienna, Austria, 1999.

[Cha99] Chavero M., Polo V., Ramos F. and Marti J., “Impact of vegetation on the performance of 28 GHz LMDS transmission", in Proceedings IMS 1999 Symposium, paper WE2D-5, vol. 3, pp. 1063-1066, Anaheim, USA, 1999.

[Che96] Cheng M. L. and Chuang C. I., "Performance evaluation of distributed measurement-based dynamic channel assignment in local wireless communications", IEEE Journal on Selected Areas in Communications, pp. 698-710, vol. 14, no. 4, May 1996.

[Cra99] CRABS Project, “Specification of next-generation of LMDS architecture”, CEC Deliverable AC215/TEL/RD/DR/P/D2P1B/b1, February 1999.

[Dro96] Droz P. and Boudec J., “A high speed self similar ATM VBR Traffic Generator”, GLOBECOM 1996, London, November 1996.

[Emb00] "System Requirements", Deliverable 2.1, EMBRACE IST-1999-11571 Project, November 2000.

[Ets97] ETSI, “Universal mobile telecommunications syste (UMTS), Selection procedures for the choice of radio transmission technologies of the UMTS”, TR 101 112, UMTS 30.03 version 3.1.0, November 1997.

[Fer93] Fernandes L., “Overview of the MBS project R2067”, RACE Mobile Workshop, Metz, France, pp. 75-79, June, 1993.

Page 82: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 82 of 84

[Fla00] Flament M. and Unbehaun M., “Impact of shadow fading in a mm-wave band wireless network”, WPMC'2000, Bangkok, Thailand, November 2000.

[Fri01] Frigon J. F., Leung V. and Chan H., “Dynamic Reservation TDMA Protocol for Wireless ATM Networks”, IEEE Journal on Selected Areas in Communications, vol.19, no.2, February 2001.

[Gan01] Gong S. Q., “Cochannel Interference in Cellular Fixed Broadband Access Systems with Directional Antennas”, Wireless Personal Communications, 2001.

[God97] Godara L. C., “Applications of antenna arrays to mobile communications: part I”, Proceedings of the IEEE, pp. 1031-1060, vol. 85, no. 7, July 1997.

[Han95] Honkanen M., “The Ray Tracing Program”, Technical Report, Technology Development Center, Finland, December 1995.

[Hel97] Hellebrandt M., Mathar R. and Scheibenbogen M., “Estimating position and velocity of mobiles in a cellular radio network”, IEEE Transactions on Vehicular Technologies, pp. 65-71, vol. 46, no. 1, February 1997.

[Hor95] Horneffer M. and Plassmann D., “Directed Antennas in the Mobile Broadband System”, Proceedings in RACE Mobile Telecommunication Summit, Cascais, Portugal, November 1995.

[Hyp01] HYPERACCESS Functional Specifications, Part 1 - Physical (PHY) Layer, 1st draft, ETSI BRAN, June 2001

[Iee01] IEEE, “Standard Air Interface for Fixed Broadband Wireless Access Systems”, IEEE P802.16.1/D3-2001, May 2001.

[Itu93] ITU-T, “CCITT Specification and Description Language (SDL)”, Recommendation Z.100, Helsinki, 1993.

[Itu94] ITU-R, “Characteristics of precipitation for propagation modeling”, Recommendation PN.837-1, Geneva, 1994.

[ITU97] ITU-T, “Digital Subscriber Signalling System No.2 Layer 3 Specification for Basic Call/Connection Control”, Recommendation Q.2931, 1997.

[Jab00] Jabbari B., “Internet Traffic Modeling”, Advanced Internet Lab, 2000.

[Khu99] Khurram S., Gesbert D., Gore D. and Paulraj A., “Smart Antennas for Broadband Wireless Access Networks” IEEE Communications Magazine, pp.100-105, November 1999.

[Lam97] Lam D., Cox D. C. and Widom J., “Teletraffic modeling for personal communications services”, IEEE Communications Magazine, pp. 79-87, February 1997.

[Law91] Law A. M. and Kelton W. D., “Simulation Modeling and Analysis. Second Edition.”, Ed. McGraw-Hill, 1991.

[Leh99] Lehne P. H. and Pettersen M., “An overview of smart antenna technology for mobile communications systems”, IEEE Communications Survey, pp. 2-13, vol. 2, no. 4, October 1999.

[Lib99] J. C. Liberti, Jr., T.S. Rappaport, “Smart Antennas for wireless communications: IS-95 and Third Generation CDMA applications, Prentice Hall PTR, NJ, USA, 1999.

Page 83: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 83 of 84

[Lut91] E. Lutz, et al., “The land mobile satellite communication channel – recording, statistics and channel model,” IEEE Trans. Veh. Technol., vol. 40, no. 2, pp. 375-386, 1991.

[Mar97] Markoulidakis J. G., Lyberopoulos G. L., Tsirkas D. F. and Sykas E. D., “Mobility Modeling in Third-Generation Mobile Telecommunications Systems“, IEEE Personal Communications, pp. 14-46, August 1997.

[Mbs95] MBS Project, “Requirements for suitable channel management and efficient handover procedures”, CEC Deliverable RS2067/UA/WP224/DS/P/0.33.b1, RACE Project 2067, September 1995.

[Mbs96] MBS Project, “Report on MBS applications and services”, CEC Deliverable MBS/CT1/T1.2/BBC016.1, July 1996.

[Mit01] Mitchell K. and Sohraby K., “An Analysis of the Effects of Mobility on Bandwidth Allocation Strategies in Multi-Class Cellular Wireless Networks”, INFOCOM 2001, Proceedings IEEE, vol. 2, pp. 1005-1011, 2001.

[Oba01] OBANET Project, “System conception and specification: network and beamformer aspects”, Deliverable D2.1, April 2001.

[Oba01a] OBANET Project, “Beamformer choice", Deliverable D2.2, April 2001.

[Pah00] Pahlavan K. et al., “Handoff in Hybrid Mobile Data Networks”, IEEE Personal Communications, pp. 34-46, April 2000.

[Pap97] Papazian P., Hufford G. A., Achatz R. J., and Hoffman R., “Study of the local multipoint distribution service radio channel”, IEEE Transactions on Broadcasting, no. 43, June 1997.

[Pie97] Pietilä A., “Development of the Netsim program”, Proceedings of the IRC Workshop’97, pp. 90-91, 1997.

[Pit00] Pitchaiah S., “Recommendation on Time Varying Radio Propagation Channel Models and Study of System Performance for LMDS”, IEEE 802.16.1, 2000.

[Pol96] Pollini G., “Trends in Handover Design”, IEEE Communications Magazine, pp 82-90, March 1996.

[Raa94] Raatikainen K., “Symptoms of Self-Similarity in Measured Arrival Process of Ethernet Packets to a File Server”, University of Helsinky, Department of Computer Science, 1994.

[Rap95] Rapeli J., “UMTS: targets, system concept, and standardisation in a global framework”, IEEE Personal Communications Magazine, pp. 20-28, February 1995.

[Rob97] Robert S. and Boudec J., “New Models for Pseudo Self-Similar Traffic”, Performance Evaluation, vol. 30, no. 2, pp. 57-68, July 1997.

[Sam97] Sampei S., “Applications of digital wireless technologies to global wireless communitcations”, Ed. Feher/Prentice Hall Wireless and Digital Communications Series, 1997.

[Sta96] Stapleton S. P. et alt, “Tracking and Diversity for a Mobile Communications Base Station Array Antenna”, IEEE Vehicular Technology Conference, Atlanta, GA, April 1996.

Page 84: “Coverage area management specifications”morse.uml.edu/Activities.d/Summer-05/PAPERS/KC/D3_1_PUB.pdf · “Coverage area management specifications” Contractual Date of Delivery

OBANET IST-2000-25390 Deliverable D3.1

Page 84 of 84

[Xie00] Y. Xie, Y. Fang, “A general Statistical Model for Mobile Satellite Systems”, IEEE Trans. Veh. Technol., vol. 49, no. 3, pp. 744-751, 2000.

[Xu00] Xu H. and Rappaport T. S., “Measurement and Models for 38 GHz Point-to Multipoint Radiowave Propagation”, IEEE Journal on Selected Areas in Communications, vol. 18, no. 3, March 2000.

[Xu00a] Xu H., Rappaport T. S., Boyle R. J. and Schaffner J. H., ”38 GHz wide-band point-to-multipoint measurements under different weather conditions”, IEEE Comm. Lett., vol. 4, no. 1, pp. 7-8, 2000.

[Yac00] Yacoub M. D. et alt., “Fading Distributions and Co-Channel Interference in Wireless Systems”, IEEE Antennas and Propagation Magazine, vol. 42, no. 1, February 2000.

[Zha00] Zhang W., “Classification of Statistical Channel Models for Local Multipoint Distribution Service Using Antenna Height and Directivity”, 2000.