Active Set Size Trial

NE region UMTS Performance

T-Mobile - New York Market

Leandro Megale
RF Engineer
 Objective

 To improve the performance of the New York network by increasing the maximum
allowed cells in the Active Set

 The expectation is that this would decrease the number of RF drops in the network
and in turn lead to an increase in MoU/Drop (Minutes of Use per Drop)

 It is also expected that this improvement would NOT be at the cost of significant
degradation in other major KPIs

maxActiveSet 2 Leandro Megale 2008-12-07

 Methodology
 Change the parameter “Active Set Size” on the RNC level from 3 to 4

OSS:
- Open WCDMA RAN EXPLORER
- Right click in the RNC – Select Element Manager
- Select Radio network
- Right click in “RNC Function” – Select Properties
- Go to HANDOVER tab.
- Change “maxActiveSet” to 4
- Save CV
OR
Moshell:
- Connect to RNC
- lset Handover maxActiveSet 4
- Save CV

maxActiveSet 3 Leandro Megale 2008-12-07

 Pre-requisites

 Prioritize the 3G-3G neighbor list according to handover statistics. This becomes
imperative as the AS neighbor list gets further truncated when the AS size is
increased from 3 to 4

 Prioritize the IRAT neighbor list. Again, as in the case of 3G-3G handovers, the IRAT
neighbor list will get truncated when the AS size is increased from 3 to 4

 Ensure that no other significant changes are made in RNC01 area during the trial

 A stable network for 5 working days prior to the start of the trial is essential to monitor
the effects before and after implementation of the parameter change

maxActiveSet 4 Leandro Megale 2008-12-07

 The Neighbor List

 A UE is constantly monitoring the connection

quality of adjacent cells and informs the
RNC when measured quantities fulfill the
criteria for specified events. Based on such
events, the RNC keeps the active set
updated with cells providing the best
connection quality. For the UE to know which
cells to measure, a list of scrambling codes
is provided by the RNC. This is the IAF (intra
frequency) listed subset, created based on
the configured neighbors of the cells
currently in active set.
 The number of cells in the IAF listed subset
(active set cells + IAF monitored subset
cells) is by 3GPP limited to 32. If the number
of active set cells plus the union of their intra
frequency neighbors (IAF neighbor subset)
exceeds 32, the total list is truncated to form
a list of 32 scrambling codes.

maxActiveSet 5 Leandro Megale 2008-12-07

 Neighbor List Prioritization

 Prioritization of neighbor relations is a method of limiting the negative effects of

truncation.

 It gives the opportunity to limit or even remove the risk for frequently used
neighbors to become unmonitored, without having to remove other neighbors from
the neighbor list.

 In this way all defined neighbors are made available for detected set capable UEs.

maxActiveSet 6 Leandro Megale 2008-12-07

 Expectations

 Increase of dropped calls due to missing neighbor and congestion. (-)

 We will probably see a reduction in capacity of the network and will have to get
visibility to this metric to properly evaluate the pros and cons of this trial. (-)

 The SHO overhead target has been limited to a value for the cluster and market
acceptance drives so far. This number could potentially increase (-)

 Potentially see a degradation in user experienced call quality, again due to an

increase in the number of calls held in poor coverage areas. (-)

 Since the calls will stay on the UMTS network longer, we will probably see a
reduction in the number of IRAT HOs. (+)

 According to previous comment, the number of IRAT HOs attempts will reduce,
reducing consequently the number of HOs failures.

maxActiveSet 7 Leandro Megale 2008-12-07

 KPI’s to be monitored

 MoU/Drop
 %Secondary (Voice) Traffic
 Number of times compressed mode is triggered/cancelled
 IRAT Attempts, Failures and ratio
 UE TX power (if visible in counters)
 Overall load in the network (DL power)
 SHO overhead
 T1 utilization
 CE utilization

 PrimaryVoiceTraffic= (pmSumBestCs12Establish + pmSumBestAmr12200RabEstablish +

pmSumBestAmr7950RabEstablish + pmSumBestAmr5900RabEstablish +
pmSumBestAmr4750RabEstablish)/720

 SecondaryVoiceTraffic= (pmSumCs12RabEstablish + pmSumAmr12200RabEstablish +

pmSumAmr7950RabEstablish + pmSumAmr5900RabEstablish + pmSumAmr4750RabEstablish)/720
-
(pmSumBestCs12Establish + pmSumBestAmr12200RabEstablish + pmSumBestAmr7950RabEstablish +
pmSumBestAmr5900RabEstablish + pmSumBestAmr4750RabEstablish)/720

maxActiveSet 8 Leandro Megale 2008-12-07

 Area and duration of trial

 RNC 1 in Manhattan was chosen for it’s challenging RF conditions

 A duration of 5 working days should give us concrete details regarding the affect of
the trial on the UMTS network.

 Target will be to finalize NB prioritization on time and have next week as base line

maxActiveSet 9 Leandro Megale 2008-12-07

 Following slides show statistical results
of the trial

maxActiveSet 10 Leandro Megale 2008-12-07

 MoU/drop and traffic

NYRNC001 Voice Traffic & MoU/Drop NY RNC Network Availability (%)

6000 250
100.6
100.2
5000
200 99.8
99.4
4000 99
98.6

MoU /Drop
150
Erlang

98.2
3000

%
97.8
100 97.4
97 RNC 1
2000
96.6 RNC 2
50 96.2
1000 95.8 RNC 3
95.4
0 0 95 RNC 4

6/2/2008
6/4/2008
6/6/2008

6/8/2008
5/27/2008
5/29/2008
5/31/2008

6/10/2008
6/12/2008
6/14/2008
6/16/2008
6/18/2008
6/20/2008
6/22/2008
6/24/2008
05/27/08
05/29/08
05/31/08
06/02/08
06/04/08
06/06/08
06/08/08
06/10/08
06/12/08
06/14/08
06/16/08

06/18/08
06/20/08
06/22/08
06/24/08
Primary Traffic Secondary Traffic Voice MoU/Drop

•Traffic is increased in what seems a ‘step’ increase compared with a more gradual
increase in the other RNC’s
•Weekly average Primary traffic increased by ~10% while secondary traffic
increased by ~14%
•Ratio primary traffic to secondary traffic increased from 1.44 to 1.47
•Cell Availability had a big impact on MoU/drop.

maxActiveSet 11 Leandro Megale 2008-12-07

maxActiveSet
1.45
1.55
1.65

1.4
1.5
1.6
05/25/08
05/26/08
05/27/08
SHO overhead

05/28/08
05/29/08
05/30/08
05/31/08
06/01/08
06/02/08
06/03/08
06/04/08
06/05/08

12
06/06/08
06/07/08
06/08/08
06/09/08
06/10/08
06/11/08
06/12/08
06/13/08
SHO overhead

06/14/08
06/15/08
06/16/08
06/17/08
06/18/08
06/19/08
06/20/08

Leandro Megale
06/21/08
06/22/08
06/23/08
06/24/08
SHO overhead

2008-12-07
 05

0
200
400
600
800
1000
1200
1400
1600
05/27
/0

maxActiveSet
05/28 8
/ /0
05 29 8
/ /0
05 30 8
/ /0
06 31 8
/0
06/01 8
/0
06/02 8
/0
06/03 8
/0
06/04 8
/0
06/05 8
/0
RAB releases

06/06 8
/0
06/07 8
/0
06/08 8
/0
06/09 8
/ /0
06 10 8
/ /0
06 11 8
/ /0
06 12 8
/ /0
06 13 8
/0
06/14 8
/0
06/15 8
/0

13
06/16 8
/0
06/17 8
/0
06/18 8
/0
06/19 8
/0
06/20 8
/0
06/21 8
/0
06/22 8
/ /0
06 23 8
/ /0
RAB releases

06 24 8
/ /0
06 25 8
/2 /08
6/
08

0
20000
40000
60000
80000

Leandro Megale
100000
120000

2008-12-07
pmNoSysRelSpeechSoHo (Cell)
pmNoSysRelSpeechNeighbr (Cell)

pmNoSysRelSpeechULSynch (Cell)

pmNoNormalRabReleaseSpeech (Cell)
pmNoSystemRabReleaseSpeech (Cell)
 SHO Signaling
NYRNC001 SHO Signaling

1400000
Total RL add/del/rep & RL Rep 2500000
1200000

RL Add & RL Del

2000000 1000000

1500000 800000

600000
1000000
400000
500000
200000

0 0
5/27/2008

5/29/2008

5/31/2008

6/2/2008

6/4/2008

6/6/2008

6/8/2008

6/10/2008

6/12/2008

6/14/2008

6/16/2008

6/18/2008

6/20/2008

6/22/2008

6/24/2008
SUM_RL Add/Del/Rep RL Addition RL Deletion RL Replaced

•SHO signaling has decreased significantly with increase of AS size (~16%) in

the first week of trial.
•In the first 3 days of the second week however, the reduction is only ~2%
•Improvement is in the # AS replacements, this had a huge reduction of 285%
in the first week and still 237% in second week.

maxActiveSet 14 Leandro Megale 2008-12-07

maxActiveSet
0
20000
40000
60000
80000
100000
120000
140000
160000
180000
05/27/08
05/28/08
05/29/08
05/30/08
05/31/08
06/01/08
06/02/08
06/03/08
06/04/08
06/05/08
06/06/08
06/07/08
Compressed mode

06/08/08
06/09/08
06/10/08
06/11/08
06/12/08

15
06/13/08
06/14/08
06/15/08
06/16/08
06/17/08
06/18/08
06/19/08
06/20/08
06/21/08
06/22/08
06/23/08
06/24/08
06/25/08
06/26/08

Leandro Megale
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
0.70%

pmCmStop (Cell)

CM activation failure
pmCmAttDlSf2 (Cell)

pmCmSuccDlSf2 (Cell)

2008-12-07
 IRAT HO attempts and failure rate
NYRNC001 CS IRAT Failure Rate
16000 20.00

14000 18.00
16.00

CS IRAT Failure Rate (%)

12000
# IRAT Attempts 14.00
10000 12.00
8000 10.00

6000 8.00
6.00
4000
4.00
2000 2.00
0 0.00
5/27/08
5/29/08
5/31/08
6/2/08
6/4/08
6/6/08
6/8/08
6/10/08
6/12/08
6/14/08
6/16/08
6/18/08
6/20/08
6/22/08
6/24/08
IRAT Attempts Voice IRAT Failure Rate

•Irat attempts seems to be more stable throughout the

week and are slightly reduced ~2% in the first week
(~1% in the second week)
•Take into consideration the increase in overall traffic!

maxActiveSet 16 Leandro Megale 2008-12-07

 DL load

maxActiveSet 17 Leandro Megale 2008-12-07

 Results: AS size in numbers

maxActiveSet 18 Leandro Megale 2008-12-07

 Results: AS size distribution

maxActiveSet 19 Leandro Megale 2008-12-07

 Failures due to resources

 From the counters before and after implementation of max AS size=4; NO

occurrences yet of RAB setup failures due to:

DL power
DL channelization codes
UL/DL CE capacity
UL/DL HW resources

 T1 and CE utilization counters are currently not available in the XPM backend;
Region is working on it to get those counters populated in the database.

maxActiveSet 20 Leandro Megale 2008-12-07

 Conclusion

Max AS size has been increased from 3 to 4

 From the trial results we can conclude that the max AS size increase from 3 to 4 did
not have major negative results.
 MoU/drop shows more stable after increasing AS size (+)
 Stable number of IRAT attempts (+)
 Less HO signaling (+)
 All other PI’s are behaving as expected with this setting (+)

 Cell Availability heavily influenced the MoU/drop over the last few weeks (-)
 Drops due to missing NB has drastically increased with this setting, although this KPI
is influenced by Cell Availability NB-list optimization is required (-)
 SHO overhead increases (-)
 Increase traffic (-)

maxActiveSet 21 Leandro Megale 2008-12-07

 End

maxActiveSet 22 Leandro Megale 2008-12-07