Professional Documents
Culture Documents
Uppgjord (ven faktaansvarig om annan) - Prepared (also subject responsible if other) Nr - No.
1 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
Contents
1 2 3 4 5 5.1 5.2 5.3 5.4 6 6.1 6.2 6.3 6.4 6.5 7 7.1 7.2 7.3 7.4 8 INTRODUCTION ............................................................................2 FRACTIONAL LOADING NETWORKS.......................................2 TRIAL NETWORK CHARACTERISTICS....................................3 FEATURES ......................................................................................3 TEST CASES....................................................................................4 BASELINE............................................................................................4 NETWORK 1 ........................................................................................5 NETWORK 2 ........................................................................................5 MRP...................................................................................................5 TRIAL RESULTS ............................................................................5 STS PERFORMANCE ............................................................................5 TEMS PERFORMANCE .........................................................................6 EFFECTS OF UNEVEN TRX DISTRIBUTION ...........................................7 EFFECTS OF BCCH/TCH ADJACENCIES ................................................8 NONHOPPING BCCH VERSUS HOPPING TCH.........................................8 DISCUSSION ON THE RESULTS..................................................9 1/3 LOADING .......................................................................................9 TWO CHANNEL GROUPS .....................................................................11 BCCH/TCH ADJACENCIES ................................................................11 1/3 VERSUS MRP...............................................................................12 CONCLUSIONS.............................................................................12
C:\temp\1_3_telstra_B.doc
2 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
Introduction
This document summarises the 1/3 trials in a major CBD capital macrocell network within Telstras GSM network, which were performed in September and October of 1998. The objective with the trials was to investigate and obtain practical experience about 1/3 frequency reuse. Also, to learn about differences and potential advantageous to the MRP methodology. For example, find out when MRP should be used, and vice versa when 1/3 is preferred. In addition to trial results, the report includes thoughts and findings about 1/3 and MRP. The trials in short, the interference load of a network with 1/3 TCH frequency reuse was increased by reducing the number of hopping frequencies. In addition, an MRP frequency plan was also implemented and tested for comparison reasons.
FRQLoad =
The frequency load reflects how much time a frequency is being transmitted. Sometimes, the frequency load is referred to as the fractional load.
Note that 1/3 and 1/1 can only applied on the TCH frequencies. The BCCH frequencies must still use a sparse reuse for satisfactory control channel performance.
C:\temp\1_3_telstra_B.doc
3 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
Another commonly used measure is the hardware load defined as the average number of TRXs per cell divided by the number of hopping frequencies per cell. This measure is similar to the frequency load and the hardware load can be expressed as:
HWLoad =
However, the hardware load does not take the trunking efficiency into account. The same hardware load results in different served traffic loads depending on the number of frequencies and thereby different frequency loads. 1/3 frequency reuse can be compared to MRP networks, which usually use baseband-hopping resulting in that the number of transceivers is equal to the number of hopping frequencies per cell. Thus, increased frequency hopping gain (interference and frequency diversity) due to the large number of frequencies in the hopping sequence is a potential advantage with 1/3. However, this might be compensated by the sparser reuse of the TCH frequencies (e.g. 8/6/4) with MRP.
Features
The following features were applied in the test area: Random synthesizer hopping The key characteristic of a network with fractional loading.
C:\temp\1_3_telstra_B.doc
4 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
Power control (PC) Utilised in both up- and downlink. In the downlink a careful down regulation was applied, only a maximum of 8dB was allowed. Overlaid/underlaid (OL/UL) A special OL/UL configuration was used in two of the trial cases. The OL was made to have the same size as the UL. All TCH frequencies were put to the OL subcells, while the UL subcells contained the BCCH frequencies only. The main reason for this set up was to be able to direct all traffic to OL in order to increase the load on the TCH frequencies during the tests. In addition, it also made it possible to have different QLIM urgency handover settings in OL and UL. Further, all macrocells were set to co-sited (CS=1), allowing handover and assignment to macrocell OL directly. However, only a handover from a microcell to UL in macro could be performed since the microcell were not set to co-sited. Cell load sharing Used in all cells. RHYST=75.
Test Cases
The BCCH frequencies were planned in a 12-reuse pattern using a staggered allocation. This plan was kept unchanged throughout the entire trial. For the 1/3 TCH frequency planning, Telstra uses strict group planning, i.e. all cells pointing in the same direction use the same hopping frequencies. In addition, a large number of cells (all but 3) were configured with one or two TCH frequencies adjacent to own BCCH frequency. The CBD network was in a very good shape; the handover borders had been optimised for the 1/3 plan since July-98. During the trials, extra transceivers were added to the macrocell in the CBD area, this to make sure that all traffic could be carried on the TCH frequencies only, see below for more details. Thus, the average no of TCH TRX was 2.2 per cell.
5.1
Baseline
This case was used as the reference case and it was run from 14/9 to 27/9. 6 frequencies were used as hopping TCH frequencies in each cell (i.e. totally 18 TCH frequencies). Carrier-zero filling was applied for the BCCH frequencies, which means that the synthesizer hopping was performed over 7 frequencies. The BCCH frequency was included in the hopping set. QLIMUL/DL were set to 45.
C:\temp\1_3_telstra_B.doc
5 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
5.2
Network 1
The second case was tested from 28/9 to 7/10 and it was configured in a two-channel group structure using the special OL/UL. The 4/12 BCCH frequency plan was nonhopping and placed in the UL, while the 1/3 TCH was synthesizer hopping over 6 frequencies in OL, (i.e. still 18 TCH frequencies were utilised). This implies increased frequency load and reduced hopping gain compared to the baseline due to less hopping frequencies. The macrocells were configured to carry all traffic on the OL. QLIMUL/DL was 45 in UL and 55 in OL.
5.3
Network 2
The third case was run from 8/10 to 20/10 and it was also configured in a two-channel group structure using the OL/UL. As before, the 4/12 BCCH frequency plan was nonhopping and placed in the UL, while the 1/3 TCH was now synthesizer hopping over 5 frequencies in OL, (i.e. only 15 TCH frequencies were utilised). One frequency per sector was removed. This implies an even higher frequency load and reduced hopping gain compared to network 1. Still, QLIMUL/DL was 45 in UL and 55 in OL.
5.4
MRP
A 12/6/6/6 fractional load MRP configuration using one channel group was also tested during 21/10-5/11. The MRP used the 4/12 BCCH plan and 3 TCH layers, each with a reuse of 6. Carrier-zero filling was applied; i.e. synthesizer hopping was performed over totally 4 frequencies. Thus, the BCCH frequency was included in the hopping sequence. The TCH frequency planning was quite rough, done in 2 days. This resulted in that 15 cells had TCH frequencies adjacent to own BCCH frequencies. Further, QLIMUL/DL was set to 45 in all cells. The MRP was configured so it could be directly compared to the baseline case since both cases utilise the same number of frequencies (both BCCH and TCH) and they utilise carrier-zero filling. The method of distributing the TCH frequencies (i.e. the 1/3 or MRP methodology) was the only difference.
6
6.1
Trial Results
STS Performance
STS statistics were collected every hour from 0700 to 1900 every weekday during the trial. The shown STS figures represent average values over all CBD macrocells and all weekdays for each trial. Table 1 gives an overview of the STS performance. The STS statistics show a slight performance degradation for network 2. However, there is not a large drop in performance, just a gradual degradation. The other cases have similar performance. The network performance was acceptable for all cases according to Telstra.
C:\temp\1_3_telstra_B.doc
6 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
Baseline Traffic load SDCCH drop SDCCH drop bad qual TCH drop TCH drop - bad qual HO in drop HO out drop TCH Erlang min per drop Table 1 STS performance 4.6 Erl/cell 0.30% 18% 1.55% 22% 0.51% 0.47% 114
Network 1 4.4 Erl/cell 0.31% 19% 1.55% 18% 0.45% 0.43% 118
Network 2 4.9 Erl/cell 0.34% 19% 1.59% 22% 0.47% 0.48% 109
MRP 5.2 Erl/cell 0.38% 12% 1.52% 16% 0.50% 0.46% 114
Observe also that the traffic successively increased for network 2 and the MRP. This perhaps also had an influence on the performance.
6.2
TEMS Performance
A further investigation of the downlink network performance was also done using TEMS measurements. TEMS routes were defined for some selected streets and buildings. Drive tests were performed for the street routes and walk tests were performed in the buildings. The same routes at the same time of the day were investigated 1-3 times for each trial case. Samples from the microcells and the nonhopping BCCH frequencies were excluded in the analysis. Table 2 includes a summary of the TEMS results and the amount of data for the hopping TCH in terms of hours. For comparison reasons, it is here assumed that SQI>15dB reflects acceptable speech quality. The results indicate approximately the same performance for the street drives. However, in the high-rise buildings, network 2 and MRP have somewhat worse results. In the streets, users have most probably favourable shadowing to the surrounding base stations and the increase in interference will not be severe. On the contrary, high up in the buildings, users more likely have unfavourable shadowing (line-of-sight in the worst case) to the surrounding base stations which results in increased interference. This problem is difficult to avoid unless the high-rise buildings are covered by dedicated indoor systems. The performance in the highrise buildings had always been rough said the Telstra people.
C:\temp\1_3_telstra_B.doc
7 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
Street drives SQI>15 dB RXQUAL>5 Data length In building walks SQI>15 dB RXQUAL>5 Data length Table 2
Baseline 90% 4% 5h
Network 1 88% 4% 7h
MRP 88% 4% 6h
TEMS Performance for measurements in the streets and in the high-rise buildings
6.3
Figure 1
C:\temp\1_3_telstra_B.doc
8 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
Consequently, 1/3 reuse seems to be very robust; it can handle networks with uneven transceiver distributions without problems. It thus easy to add more transceivers to the network when needed. This can be done until the average load reaches the acceptable limit.
6.4
6.5
C:\temp\1_3_telstra_B.doc
9 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
7
7.1
7.1.1
Frequency loading The average traffic during busy hour was about 7.36 Erlang/cell for network 2. Using this traffic load, the frequency load was about 7.36/(5*8)0.18. A frequency was used maximum 18% of the time, but in average only 9% since DTX was applied (assuming 50% voice activity). This frequency load value represents a practical limit and it can be used for planning other networks. A network using 1/3 should be dimensioned so that the frequency load is maximum 18%. It is possible to add transceivers until the frequency load reaches 18%. The allowed average peak traffic load per cell can in general be calculated as:
C:\temp\1_3_telstra_B.doc
10 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
Figure 2
Average peak traffic load per cell versus total number of hopping TCH frequencies for a frequency load of 0.18 Hardware loading The hardware load was ~40% (2 TRX / 5 TCH freq.) on the TCH frequencies for network 2. This hardware load corresponds to 4 TRX/cell with 10 TCH frequencies per cell. This might not work because of too high fractional load (interference). The frequency load is not the same due to the trunking efficiency. Figure 3 illustrates the hardware load for different number of TCH frequencies assuming the same frequency load as given by a 40% hardware load with 5 hopping frequencies per cell and 2% GoS. Approximately 34% hardware load will be the corresponding value in a network hopping over 10 frequencies per cell. Thus, this must be considered when using the hardware load as a loading measure for 1/3 reuse networks. However, the decrease in hardware load will most likely be slightly compensated by the increased hopping gain.
7.1.2
Figure 3
The hardware load versus total number of hopping TCH frequencies for a constant frequency load
C:\temp\1_3_telstra_B.doc
11 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
7.2
Drawbacks: Requires probably a reuse sparser than 12 for the nonhopping BCCH frequencies otherwise the speech quality will be too bad. Somewhat reduced hopping gain since the BCCH frequency is not included in hopping sequence (depending on the number of hopping TCHs).
Thus, using two channel groups is a good solution in many cases. However, carrierzero filling might be the best option for operators with very little spectrum (when the hopping gain is critical). Typically, when only 4 TCH frequencies per cell are available.
7.3
BCCH/TCH adjacencies
The trials showed that it was possible to run the network even though a lot of BCCH/TCH adjacencies were present (see Section 6.4). This means probably that the performance would be even better, the 1/3 can be pushed even further, if these adjacencies are removed. In addition, a more aggressive downlink power control strategy can then be used. One must however keep in mind that this might lead to a trade off between performance and frequency planning complexity.
C:\temp\1_3_telstra_B.doc
12 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
Removing the BCCH/TCH adjacencies implies increased number of constraints for the frequency planner.
7.4
Conclusions
The trails indicate that a 3/3/4 transceiver (3.3 transceivers per cell in average) configuration is possible with 1/3 reuse and fractional loading using only 27 frequencies (including 12 BCCH frequencies, which were planned in 4/12 pattern during the trials). Thus, corresponding to an average hardware load slightly higher than 40%. Further, the effect from an uneven transceiver distribution appears to small in 1/3 networks according to results where hardware loads between 20 and 60% were achieved in the trial area cells. In addition, BCCH/TCH adjacencies appear to be a minor problem. The reason is that severe interference hits are few due to the use of DTX and synthesizer
C:\temp\1_3_telstra_B.doc
13 (13)
Rev File
LVR/D-98:0398
Datum - Date
LVR/DTC
1998-12-22
hopping. Thus, there is a potential for further increased load without these adjacencies. Moreover, the fractional load MRP showed somewhat worse performance in high-rise buildings compared to a corresponding 1/3 solution based on the same amount of spectrum. The reduced number of hopping frequencies with the MRP might be a reason for this result. However, since the MRP was not optimised, it was difficult to draw a conclusion about its performance compared to that of the 1/3. Nevertheless, MRP seems better suited for networks with many transceivers per cell and a lot of spectrum, more than about 6 MHz. 1/3 frequency reuse in combination with synthesizer hopping and fractional loading appears to be a very good solution for networks with narrow bandwidths (less than 6 MHz) and a small number of transceivers per cell. The 1/3 methodology offers: simplified frequency planning with a low grade of complexity cheap solutions larger frequency hopping gain compared to MRP, resulting in good quality high flexibility regarding adding and removing transceivers in the network
C:\temp\1_3_telstra_B.doc