GSM WIRELESS NETWORKS

Nortel's Fractional Reuse Technical Aspects

November 98

"NORTHERN TELECOM AND NORTEL MATRA CELLULAR CONFIDENTIAL: Information contained in this document is the property of Northern Telecom and/or Nortel Matra Cellular. Except as specifically authorized in writing by Northern Telecom and Nortel Matra Cellular, the holder of this document shall keep the information contained herein confidential and shall protect same in whole or in part from disclosure and dissemination to third parties and use for evaluation, operation and maintenance purposes only". "You may not reproduce, represent, or download through any means, the information contained herein in any way or in any form without prior written consent of Northern Telecom and Nortel Matra Cellular ".

1. Introduction
Fractional re-use is a Nortel technique for increasing the maximum subscriber capacity that can be supported by a given spectrum without increasing the number of cell sites. It achieves this whilst at the same time offering improvements in call quality and dropped call rates compared with non-hopping 4x12 re-use systems. The technique has been assessed by and remains in use with; France Telecom, Bouygues Telecom, Sunday and various operators in USA and China. These networks were designed with conventional patterns and fractional reuse was introduced at a later date. Nortel has led the way in fractional re-use techniques and their implementation and its approach offers the highest capacity gain available in the market. To achieve this increase in BSS network capacity we increase the number of transceivers that can be installed in a sector beyond that achievable with simple frequency re-use patterns. Our reference is the commonly used 4x12 re-use pattern which for many operators represents the basic limit achievable with acceptable interference levels.

Quality

Simplicity

Fractional re-use brings into play many of the standard GSM features designed to reduce interference under Capacity various conditions; Frequency hopping, DTX, DRX and up and downlink power control. In addition Nortel BTSs offer an interference suppression algorithm which can reduce cochannel interferers by 4dB. Since 1992 Nortel has had Synthesised frequency hopping available, in this technique it results in no impact on the number of transceivers required. Where only base-band hopping is available to the operator, additional expensive transceivers are required (total of one per hopping frequency) to produce the same results. We offer two versions of fractional re-use, 1*3 and 1*1. They each give a similar increase in capacity but are in use by different operators. 1*1 is applicable in low bandwidth networks to maximise the number of hopping frequencies available, in large bandwidth networks this becomes less significant. 1*1 has a particular advantage in frequency planning for the traffic carriers, none is required! Clearly this reduces the workload, however where large bandwidths are available 1*3 can offer a better C/I ratio since it is not necessary to hop on all frequencies in every cell to realise the interference reducing benefits of hopping. Implementing fractional re-use on networks designed for conventional re-use patterns has given each of our customers significant quality of service benefits immediately (typically a 20% reduction in dropped call rates for example) and the opportunity of increasing capacity without adding cells. For a network designed from the start to use these techniques, the benefits can be expected to be even greater.

2. Technique
Two re-use patterns are used in this technique. The first is used for one transceiver in each sector and typically carries the BCCH on a 4*12 pattern. This transceiver is non-hopping to meet the fixed frequency requirement of the Broadcast Control Channel. All subsequent transceivers in a sector frequency hop.
1 A 2 3 4 B 7 C 8 9
11

4*12

1 A 2 3 1 B

1*3

1 A 1 1 1 B

1*1

5
10

1 C 2 3

2 1 D 2

1 C 1 1

1 1 D 1

D
12

These frequency hopping transceivers use some of the total number of RF channels available to the operator, after removing the 12 required for the BCCH carriers. To reduce interference in the system we install only a fraction of the number of transceivers per sector compared with the number of frequencies allocated. We have two fractional re-use alternatives available 1*3 and 1*1. In 1*3 we use 1/3 of the radio channels as hopping frequencies in each sector and 50% load these frequencies with transceivers. In 1*1 we take this to the ultimate and use all hoping frequencies in all sectors and 20% load these frequencies with transceivers. 2.1 2.1 WHY IS FRACTIONAL RE-USE ESSENTIAL ?

Most GSM operators are now facing increased market competition and booming subscriber demand. Therefore they must not only find ways to expand cost effectively their network capacity, but also continue to improve and provide the highest quality of service. Spectrum being a limited and expensive resource, the essential initial stage is to optimise the use of this primary resource. Several solution, called spectrum optimisation techniques, among which Fractional Re-use, exist on the market. In addition, networks are expanding continuously, with thousands of BTSs, and operators are keen on finding easy but reliable solutions to manage their growth and their daily optimisation tasks. These three major issues are answered with Fractional Re-use : Unmatched capacity increase : over 100% increase without having to acquire new sites Perfect end-user quality : after having deployed Fractional Re-use, Nortel customers have noticed major improvement of their quality matrix Unique engineering flexibility : transceivers are added la carte according to subscribers demand, with no single modification of the frequency plan, therefore saving major operational costs.

Example : 7.2 MHz Spectrum 4x12 12 2 3 14 1x3@50% 12 4 5 27 1x1@20% 12 5 6 35

BCCH channel TCH TRX / cell Total TRX / cell Erlang / cell Capacity Gain

S666 105 Erl / site

150% 100%
S333 42 Erl / site

S555 81 Erl / site

4x12

1x3@50%

1x1@20%

Nortel has pioneered Fractional Re-use technology and has commercial service references for three years. Operators that, a few years before, were reluctant to deploy Fractional Re-use, are now eager to understand it and recognised it as today's leading spectrum optimisation solution. Our leadership is mainly due to : Leading BTS performances : Synthetised Frequency Hopping, essential to Fractional Re-use, is standard in our BTS portfolio since 92 All the powerful GSM radio features : VAD (Voice Activity Detection) and DTX (Discontinuous transmission), dynamic uplink and downlink power control, also being standard in our BTS portfolio since 92 Voice quality thrust : features such as Interference Cancellation Algorithm or a leading diversity technique, despite not being essential, improves the performance of the solution Engineering expertise : Nortel has three years engineering expertise with Fractional Re-use, which is, today, an invaluable differentiation factor. 2.2 2.2 WHAT ABOUT 1*1 AND 1*3 PATTERNS ?

Nortel has launched Fractional Re-use technology with 1*3 patterns, and some customers are now using 1*1 patterns. Both have to be advertised, 1*1 having obviously more marketing impact. 1*1 and 1*3 Fractional Re-use are two alternatives of the same technique and there is no real difference as far as performance (capacity gain, ...) is concerned between the two. Whereas, 1*1 should be easier to implement in most cases. In fact, these two alternatives allow Nortel to generalise the well-known benefits of Fractional Re-use to almost all networks as they give further flexibility to adapt the technique to the concerned environment. The deployment of one or the other should be made according to operators specific constraints :

 For a small bandwidth networks (up to 9 MHz according to the environment), 1*1 Fractional Re-use allow to match the 100% capacity gain, that was not achievable with 1*3 pattern (getting then full benefits of the Synthesised Frequency Hopping). For an area where the design constraint have lead to an heterogeneous network (mixture of omnidirectional sites and tridirectional sites,,...), a 1*1 design will be easier to implement For large bandwidth network with regular design (from 11 to 13 MHz according to the environment), 1*3 patterns are in general better adapted Up to now, Nortel states that we can reach 50% fractional loading with 1*3 Fractional Reuse. With 1*1 Fractional Re-use, we can get 20% fractional loading , which correspond roughly to the same capacity gain as the one we get with 1*3 patterns. (Note : For 1*1 , the fractional loading corresponds to the ratio of traffic frequencies within the site over the number of non BCCH transceivers within the site). Interference cancellation algorithm The ICU (available from V10) is an attractive feature for Fractional Re-use networks. It combats uplink interference, and when introduced over a Fractional Re-use network, it enhances the end user quality, and in some case may even allow to increase on some cells (on case by case basis, with network performance monitoring) the fractional load and therefore the capacity. Whereas It do not guarantee a value for the C/I improvement if the network is asynchronous (as it is today) and offers benefits only in the uplink side. That s why it is only one of the building block of our future solution.

3. Blossoming Fractional Re-use networks

Nortel has Fractional Re-use in commercial service network world-wide for 3 years. Since the last 6 months, this solution has raised much interest by our customers and many of them have tested and deployed it successfully. Omnipoint, Salt lake City, USA, GSM1900 Due to the bowl shape valley of the area, the quality of the signalling (BCCH) needed to be reinforced to ensure a high quality of service. The solution proposed by Nortel to the operator was to implement Fractional Reuse, with a mix of 1*3 and 1*1 patterns, in order to free some frequencies for the signalling (BCCH) pattern. Bellsouth, Charlotte, USA, GSM1900 In order to anticipate their subscriber growth and to reduce their operating costs, Bellsouth decided to deploy Fractional Re-use technique with 1*3 pattern over their Charlotte network. In particular, this allowed to smoothly introduce, according to subscriber demand, the third and fourth transceivers in the concerned cells. USA - GSM 1900

SOUTH EAST ASIA- GSM 1800 HONK KONG, SINGAPORE

Sunday, Hong Kong, GSM1800

In a particularly challenging environment (11 cellular operators, coverage issues), Sunday deployed a very dense network - 500 BTS over 10km in city centre - to ensure a perfect coverage quality. In order to improve the service quality and to be enhance their network capacity, they decided to test Fractional re-use patterns and MRP. The result of the tests was that Fractional Reuse provided the best compromise between performance, capacity and flexibility, and they decided to generalised it over their network. Today the deployment of 1*1 Fractional Re-use has been performed over Kowloon (part of Hong Kong territories) and is still on going for the rest of the area. Singtel, Singapore, GSM1800 Singtel has deployed a Fractional Re-use solution with 2*6 pattern over their whole network in order to enhance the voice quality they provide to their subscribers.
EUROPE - GSM 900 FTM, ESAT DIGIFONE ...

France Telecom Mobile, France, GSM900

After intensive quality tests started in June 97, France Telecom decided to generalised over the whole south region the Fractional Re-use technology with 1*1 pattern. The 1*1 pattern was more suitable than the 1*3 because of the network heterogeneity, due to the non regular topology of Marseille. The generalisation was initialised during the 98 GSM World Congress where in particular the Palais des festivals BTS supported a 1*1 reuse with 20 % fractional loading. More than 300 sites over Cannes and Nice have been put in service in February 98. FTM was particularly attracted by the engineering easiness provided by Fractional Re-use patterns. Bouygues Telecom, France, GSM1800 Bouygues Telecom has recently conducted Fractional Re-use tests over Paris in order to be able to increase their network capacity. Results have been very satisfactory, with in particular quality enhancement, and Bouygues Telecom decided to generalise Fractional Re-use technology over part of their network. ESAT Digifone, Ireland, GSM900 ESAT Digifone, the second operator in Ireland, captured 45% market share in 9 months. The operator chose 1*1 Fractional Re-use, as the very best technical solution , to keep their performance promises while continuing to welcome new subscribers. 1*1 Fractional Re-use was deployed in march 98 over the capital city of Dublin and the surrounding region. Switching the network from its previous configuration to the new solution was accomplished

overnight, minimising the impact on customer service. The overall network capacity was improved by 55 % and instantaneous result was an improvement of the quality metrix and voice quality. ESAT Digifone has now the flexibility to further improve their network capacity la carte . The technique will be generalised over other cities in Ireland (Cork,...). CHINA - GSM 900 PTA China (4 network), GSM900 Fractional Re-use, with 1*3 pattern, has been deployed from launch in 4 PTA networks, among which Shanxhi and Hebei that both comprise hundreds of BTS. In Xian city wall capital city of Shanxhi-, where the subscriber demand is very high and where the engineering is a real challenge (taking into account the Chinese PTA constraints...), S666 BTS with 50% fractional loading are in operation and present high quality metrix. A recent quality audit (including all standard metrix) performed by PTA over their 30 networks even ranked Shanxhi Fractional Re-use network 3rd out of the 30 audited networks !

4. References
The following table gives the GSM networks where Nortel Fractional Reuse (1*1 or 1*3) is implemented. Operator China Hebei PTA GSM900 CUTC Daqing GSM900 Singapore GSM1800 network China Shaanxi PTA GSM900 EAE Saudi Arabia FTM Marseilles GSM900 China Shaanxi PTA GSM900 BS Charlotte GSM1900 WW Salt lake City GSM1900 FTM Nice/Cannes GSM900 Sunday in HK Esat Digifone Dublin GSM900 Bytel Paris, Lyon and Nice GSM1800 China Tianjin PTA GSM1800(Dual band trial) China Shaanxi PTA GSM900 Taiwan CHT-LDM GSM1800 Ph.2,3 Esat Digifone GSM900 BCCH/TCH Implementation Reuse Pattern date 4*12/1*3 09/1995 4*12/1*3 09/1995 4*12/2*6 11/1995 4*12/1*3 12/1995 4*12/1*3 11/1996 18/1*1 07/1997 4*12/1*3 09/1997 7*21/ 1*1 09/1997 7*21/1*1 12/1997 7*21/1*1 01/1998 19/1*1 02/1998 16//1*1 03/1998 7*21/2*6/1*3/1*1 03/1998 7*21/1*1 03/1998 4*12/1*3 04/1998 7*21/1*3 08/1998 4*12/1*1 09/1998

5. TECHNICAL OBJECTIVES

The frequency allocation technique used to design a cellular network is a critical factor determining the maximum capacity. Traditional regular models cannot offer enough flexibility to fully benefit of the GSM technology and maximise the spectrum efficiency to cope with the growing cellular demand. In urban areas where interference are dominating for fixing the Quality of Service (QoS), an appropriate application of frequency hopping in conjunction with other standard GSM features can allow to control their distribution with a greater spectrum efficiency. This is based on a statistical optimisation of these interference over the whole frequency band and a strict control of the load of each RF channel. By comparison with the most aggressive regular frequency reuse pattern applicable to GSM, the fractional technique may allow to multiply the offered capacity by 2 for the same QoS. Bandwidth consuming services such as data, which rely on channels aggregation to increase the end user rate , are the first ones to leverage this resources increase. In addition, it offers an extreme flexibility to simplify the TRX count evolution while traffic grow without having to reconsider the frequency allocation plan. This improvement on the spectrum efficiency and the network management is independent of other current investigations on minimisation of interference based on traffic management and self optimisation of network parameters and keep the door open to further capacity increase. This paper introduces the theoretical principles which sustain the fractional reuse technique as well as the operational and engineering aspects. Results achieved within networks where Nortel has deployed fractional reuse are then presented. In Annex are given the results of the tests made on the frequency hopping in a real network.

6. PRINCIPLES
GSM specifications impose the necessity to ensure full power and full time transmission of the BCCH beacon frequency across the cells , in order not to compromise the measurement process performed by mobiles connected in neighboring cells or currently in idle or selection mode inside the cell. The BCCH layer, onto which no frequency hopping nor other GSM features are thus activated, is protected from interference by applying a large conventional pattern such as 4/12 or 7/21. These patterns provide sufficient robustness and call quality. The TCH TRXs, on the contrary can fully benefit from the frequency hopping , power control and VAD / DTX GSM features. The fractional reuse technique relies on a split of the available bandwidth in two separate segments, one dedicated to the BCCH frame, the other to the remaining TCH traffic frames. BCCH sub-band : TCH sub-band : Implementation of a conventional pattern Implementation of the Nortel 1/3 or 1/1 reuse pattern

Conventional pattern A1 B1 B3 D1 D3 A2 C1 C2 T3 T3 T1 T1

Nortel 1/3 Fractional pattern

T1 T2 T1 T3 T2 T2

A3 B2 C3 D2

T3 T2

In a 1/x reuse pattern , the total TCH bandwidth is split between x cells and the cluster made by those x cells is repeated periodically throughout the network. In a 1x1 network, it means that all the TRX within any cell use the whole TCH sub band. Introduction of such a tight reuse pattern would obviously degrade the C/I that the TCH TRXs experience in comparison with the conventional BCCH pattern if several GSM features were not activated to compensate this C/I degradation.

7. GSM features
1. Frequency Hopping with a fractional load. When combined with channel coding and interleaving, frequency hopping in itself enables to improve the system's robustness against C/I interference. Rayleigh fading is smoothed , especially for slow mobiles , and hopping on several frequencies creates a situation of
9

jammers diversity : co-channel communications become statistically spread over close and far distances from the communication. The important factor to keep in mind is that the Frame Erasure Rate becomes more meaningful than the C/I as soon as Frequency Hopping is introduced. The C/I becomes variable from one burst to the other and its average value or its worst case value no more represents the voice quality that will eventually be achieved in the network. The reason is that the interleaving and code correction capabilities that the GSM offer allow to recover from errors instantaneously encountered on some bursts :

Samples from RX

Equalisation

De-interleaving

Channel decoding

Bad Frame Filter

bad burst

BER

RxQual

FER

Same average nb of errors -> same RxQual Different distribution of errors -> different FER
104 bursts

The gain in FER can be accounted for as much as 2 dB, in case frequency hopping is performed over 8 frequencies , and 3 dB when performed over 12 frequencies (cf. Annex). Whereas it may not be possible to activate frequency hopping with a conventional reuse pattern, fractional reuse make it possible as more frequencies are allocated to each cell.
FER versus SFH at -104 dBm
14,00

12,00

1 km/h 3 km/h 5 km/h 10 km/h 50 km/h

10,00

8,00

6,00

4,00

2,00

0,00 1 2 3 4 5 6 7 8

NUMBER OF FREQUENCIES

Frequency hopping can also be applied with some frequencies being unused during any given time slot.. This ratio TRX/number of TCH frequencies is often referred to as "use rate", or "fractional" load, since it depicts the utilization rate of each frequency in time, within the cell. The number of TRXs provisioned in each cell is smaller than the number of TCH frequencies for a fractional reuse pattern. Each hopping frequency is then

10

used a fraction of the time in each BTS. Moreover, random frequency hopping laws (MAIO,HSN) are chosen to maximize interferer diversity. 2. VAD/DTX Feature Activation of the Voice Activity Detection and Discontinuous Transmission means that transmission doe not occur when there is no speech. This reduces the transmission requirements by over 50%. This feature brings a 3dB gain if a voice activity detection ratio of 0.5 is considered. 3. Power control Power control activation brings an estimated C/I gain of 1.5 to 2 dB. Intensive simulations have been conducted within Nortel to come to this figure after demonstration that the partial path loss compensation algorithm was more efficient that the full path loss approach
proposed by the GSM 05 recommendations.
Optimum POWER CONTROL NO POWER 14 CONTROL

1,5 to 2 dB gain

% FOR FER AND COM, DB FOR C/I

12 10 8 6 4 2 0 0 0, 1 0, 2 0,3 0, 4 0, 5 0, 6 0, 7 0,8 0, 9 1 FER R1 2 % COM FER>5 % 1 0 % C/ I

PATHLOSS COMPENSATION

4. Nominal Traffic Effect Last gain comes from the fact that all the TCH channels can not be fully utilized simultaneously at nominal traffic load. In fact, Erlang B law application for 3 TRX per cell configurations shows a traffic load of about 70%, which brings in 1.5 dB gain. This 70% value represents the trunking efficiency of the network. 5. Optional Feature : Interference Cancellation Voice quality is highly affected by the interference level on both the downlink and the uplink. This interference is mainly self-generated by the GSM system, i.e., by the BTSs on the downlink and by the mobile handsets on the uplink. Uplink interference is more difficult to control as it is not constant in time, and depends on mobile handset location, allocated frequencies and other terminals. Nortel proprietary interference cancellation algorithms provide a solution against uplink interference.

11

This feature combines the use of receive diversity and Nortel signal processing technology. By eliminating the dominant source of uplink interference from the expected signal after processing, a much higher voice quality is produced.

f1 WANTED

Xf

1 interferer 1 WANTED

f1 wanted INPUT f1 interferer DSP

f1 wanted OUTPUT

Thus, the BTS estimates the channel response associated with the expected signal. Then, the system calculates the characteristics of the interfering signal so as to allow the processor to focus the digital beam forming (i.e., the diversity processing) and to cancel the interfering signal. Adding up these gains gives an indication of the circumstances under which the transition from one conventional BCCH/TCH combination to a conventional BCCH / fractional TCH pattern is "possible", i.e. without loss of call quality. These "acceptable" transitions are obtained when gains out balance losses. Some examples are given in the table below, along with the capacity gains they permit, while the same quality is maintained :
BCCH pattern 12 12 21 TCH pattern (use rate) 3 1 6 (@ 50%) (@ 15%) (@ 50%) Capacity gain (with spectrum) +75% (with 6 MHz) + 70% (with 5 MHz) +75% (with 12 MHz)

For small spectrums , such as 5 MHz , Nortel recommends to go directly to 1x1 to maximize the frequency hopping benefits and the statistical gain obtained from interferer diversity.

8. IMPLEMENTATION
Fractional reuse has been successfully implemented by Nortel over 15 networks throughout the world. A strong engineering expertise has been acquired which now allows Nortel to propose clear guidelines or optimisation activities to guarantee success when fractional reuse is introduced :

12

 Optional Automatic Frequency planning on BCCH : Nortel CT3100 Radio optimisation tool may be used to build a C/I matrix from radio planning tool predictions or field measurements and optimize automatically the BCCH plan. Such plan is normally prepared in advance and activated during the night to minimize loss of service. Modification of the BSIC neighbouring plan Optimized Handover settings : Introduction of frequency hopping over a large set of frequencies modify the RxQual distribution, which no more represents accurately the voice quality. In consequence, RxQual handovers criteria must reflect this evolution ,and handovers on Bad RxQual condition will be minimized. Those parameters are modified at OMC level during normal operation. Radio access optimisation : due to the above mentioned RxQual versus RxLev profile modification, the Minimum Access Field required to allow a mobile to camp on a new cell can be decreased. As an example, those Network Optimisation activities conducted in parallel with other activities of monitoring, troubleshooting, network datafill have been achieved within Esat commercial network in 6 weeks for 300 cells. Drive Tests are also carried out at different stages to monitor and qualify the Quality of Service. A big advantage offered by 1x1 or 1x3 is that it drastically simplifies the network densification. Provided that appropriate coupling devices have been installed in the BTS since the beginning, TRXs will be added in sites without any modification on the radio planning. For instance, an operator with 5MHz could start with S222 (H2D) and upgrade to S334 (H2D) when traffic requires it, increasing simply the 1x1 fractional load from 8% to 18% . Compared to other densification techniques like cell splitting or micro cellular , which require new sites acquisition , and new radio planning activities , fractional reuse offers considerable operational savings to the operator.

9. RESULTS
The fractional techniques has been deployed over many Nortel networks and intensively qualified in China (1*3 in Xian with Shaanxi PTA) and in Ireland (1*1 in Dublin with Digifone). In the same time, France Telecom has conducted its own qualification campaign in the south of France and widely deployed the 1*1 in Cannes for the last GSM world congress. Shaanxi PTA network initial situation was an existing 1*3 fractional reuse deployed in a 6 MHz spectrum with no optimisation and traffic saturation. Additional 2MHz spectrum could be obtained and up to 6 TRX/Sector were put in service over more than 10 sites. After optimisation the drop call rate was reduced from 5.5 % to 1.8% and the successful hand over rate increased from 80% to 95%. This network is still using very high towers for antennas and some more improvements are expected when the antennas can move down to more reasonable heights.

13

Digifone network was highly congested with a traditional "4*12 like" reuse pattern in a 7.2 MHz spectrum. Migration to 1*1 was realised in one night over 300 cells. Today , 1x1 has allowed the addition of more than 250 TRX in Dublin area with some sites having up to 5 TRX/sector. After a few weeks of optimisation, a reduction of congestion was achieved so that the initial rate of sites having less than 2% blocking was increased from 68% to 97%. The dropped call rate was also reduced from 2.5% to 1.5% and the Handover failure rate decreased relatively by 17%. In addition some voice quality tests were performed at busy hours and the most significant impact has been noticed on the downlink where the excellent/good notes increase from 89% to 94%. Another test is presently ongoing in Australia with a fractional load of up to 20%. Results are expected to be available within the coming weeks. Thanks to the flexibility that fractional reuse offers, all those operators have been able to tremendously reduce their frequency planning activities. New TRX may be added without any modification of the frequency plan as long as the fractional load limit is not exceeded. Moreover, TRXs may easily be withdrawn from some sites to be added in others according to the observed -non homogeneous- traffic distribution.

10. CONCLUSION
The fractional reuse technique is now a well known solution combining the benefits of the advanced GSM radio management features with high engineering experience. Pioneered by Nortel in 95, widely deployed since three years, intensively qualified and enhanced during several fields campaign , this technique now leverages a high operational experience and optimised radio algorithms to guarantee that the predicted capacity and QoS performances are achieved. The impact of fractional reuse on the deployment and maintenance cost of GSM networks is very significant. Several measurements have confirmed the predictions and have even sometimes overpassed them particularly for what regards the impact of frequency hopping over a wide amount of frequencies. This is why research is continuing in Nortel to keep on improving this technique by adapted traffic load management techniques like the cell tiering , which keeps the door open to further spectrum efficiency improvement by significant factors. Nortel has been always concerned by offering the best solution to minimise the site counts, either in rural areas where the S8000 BTS features the best sensitivity , or in high density areas with the fractional reuse. For narrow spectrum allocation, fractional 1*1 is definitively considered within Nortel to be the most appropriate solution.

14

11. Annex : SFH Trial

In mid 1997, VIAG Interkom has requested from Nortel to perform a Frequency Hopping Trial in order to assess the practical benefits which can be expected from the activation of this BSS feature. The measurement campaign, from preparation phase to completion of postprocessing, has lasted from August 12th to September 19th, 1997. 1. Objective The main objectives of this trial was to evaluate the impact of frequency hopping over network quality and assess how this improvement could be converted into a potential cell range increase. Measurements have been conducted in a range-limited environment, without any traffic load. Samples have been recorded in buildings at low-speed, in a cell-edge situation, so as to reproduce worst-case configurations of a network in commercial service, designed for slowmoving mobiles and high degree of indoor penetration. 2. Data Collection The variety of the test equipment used throughout the trial has enabled the collection of multiple parameters : GSM-specific parameters (RxLev, RxQual, FER) and Voice Quality, in both Uplink and Downlink paths for most of them. Data collection has been performed under three operating modes (no hopping, random hopping and cyclical hopping), each mode being activated consecutively, on the same indoor measurement paths. Post-processing tools have produced distribution curves as well as correlation curves, which allow to understand the immediate impact of frequency hopping upon quality, but also how the benefits of frequency hopping can be integrated in the network design. In addition, some specific tests have been conducted in order to assess whether the necessity for any given GSM call to use a non-hopping BCCH carrier in the early stages of call establishment could prove to be a bottleneck.

15

3. Results

FER cumulative distribution

Uplink FER cdf - Test Location 3
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0% 4% 8% 12% 16% 20% 24% 28% 32% 36% 40% 44% 48% No FH Random FH Cyclical FH

Uplink FER

Clear-cut effect of frequency hopping ->reduction of mean FER No inflexion point -> improvement all across the FER range Enhancement is most noticeable in the critical 0-15% FER range for design
Overall Voice Quality Results (Mean values over 6 indoor test locations)
4

Poor

3,5

Fair
3 Voice Quality

2,5

Good
2

1,5

Excellent
1

Uplink
No FH

Downlink
Random FH Cyclical FH

Total

Downlink effect striking : improvement from Fair/Poor to Good ! Overall effect : improvement from Fair/Poor to Good/Fair
16

FER vs. RxLev correlation

Mean Uplink FER vs RxLev - 7 indoor test locations
20% 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% -99 -98 -97 -96 -95 -94 -93 -92 -91 -110 -109 -108 -107 -106 -105 -104 -103 -102 -101 -100 -90 No FH Random FH Cyclical FH

Uplink RxLev (dBm)

Warning : low nb of samples with Rx > -98 dBm

Gain in dB : Rx reduced from -104 to -107 dBm @ FER = 8%

Gain in quality : FER reduced from 11% to 6% @ Rx = -106 dBm

12. Conclusion
Measurements have focused on the FER (Frame Erasure Rate) parameter, which provides a more meaningful insight on the speech frames quality. In all test locations, activation of frequency hopping has resulted in improvements of the mean FER value : the value of this shift depends on the original FER range. A typical test location shows reduction of the mean FER from 7% to less than 2% ; another test performed in somewhat harsher conditions shows a reduction from 16% to 8%. In parallel with the audit of these GSM parameters, assessment of voice quality has been performed using a dedicated call-generation and MOS-analysis device ; significant improvement on the Downlink path have been observed, with the mean communication quality being increased from Fair to Good thanks to frequency hopping. Moreover, the Qvoice vs. FER correlation indicates that FER trends are consistent with Qvoice trends, which is a confirmation that the FER parameter is the best representative GSM-parameter for voice quality.

17