P3 score improvement through Parameter Optimization

By:

Shehraz Shaikh/s00363887

Keywords: P3, Benchmark, Mobile, LTE, CSFB, Optimization, Parameter, Feature

Abstract:

P3 CommsDay Mobile Benchmark is considered as the crème de la crème

of all benchmarks performed in Country A and has kept all Country A
Operators on their feet, fighting out for the top spot since 2014.

Operator V has been in the third position after Operator T and Operator O in
the 2014 and 2015 benchmarks and stepped up the game in 2016 with
Huawei Country A as a partner in working towards improving OPERATOR
V’s User Experience and thereby translating into a higher score in 2016 P3
CommsDay Mobile Benchmark.

Huawei as an experienced Optimization solution provider with extensive

knowledge in working on such projects world over worked from grass root
level, built a solid foundation by optimizing the coverage/quality for the LTE
and UMTS network for Operator V using specialized tools and techniques,
further adding on top features/parameters and guiding Operator V to have a
holistic E2E review of the network to bring them on top in the 2016
benchmark results.

By: Shehraz Shaikh P3 OPERATOR V Parameters Optimization ID: S00363887

Contents
Introduction/Situation....................................................................................................................3
1. Description of Problems/Challenges..................................................................................4
2. Analysis, Strategies and Actions...............................................................................................6
2.1 CS Parameter tuning................................................................................................................6
2.1.1 CSFB setup time delay.........................................................................................................6
2.1.2 Cross procedure optimization for CS/PS Qualifiers.........................................................7
2.1.3 U2L/FR Redirection Optimization for CS and PS calls....................................................8
2.1.4 LAC/TAC Optimization.........................................................................................................9
2.1.5 Other Issues (Neighbor missing, baseline alignment, equipment issues, etc)...........10
2.1.6 MOS improvement..............................................................................................................12
2.2 PS Parameter tuning..............................................................................................................13
2.2.1 LTE Throughput Optimization (During Handover/CQI Steps/Interference avoidance)
13
2.2.2 LTE CA performance Optimization...................................................................................14
2.3 Transmission issues...............................................................................................................18
2.4 Other notable observations and planned trials...................................................................20
3. Result..............................................................................................................................................24
4. Suggestion.....................................................................................................................................25

Huawei Confidential 2|Page

Introduction/Situation

OPERATOR V’s strong point has been UL throughput, however, it always lost points on DL
throughput, CS call setup delay, qualifiers resulting in a snowball effect and coming out 3 rd in
most of these categories and overall scores.

Huawei as an experienced Optimization solution provider with extensive knowledge in working

on such projects world over worked from grass root level, built a solid foundation by optimizing
the coverage/quality for the LTE and UMTS network for OPERATOR V using specialized tools
and techniques, further adding on top features/parameters and guiding OPERATOR V to bring
then on top in the 2016 benchmark results.

It was 9 months joint project with Operator V team to cover:

a. Entire network re-planning using Huawei ACP and MR data with A-GSP data.
b. Parameters optimization specifically designed based on opportunity area from P3 score
gap.
c. Feature activation to utilize all available spectrum as well as improving signal quality.
d. Transmission link quality optimisation to improve packet drop reported from P3.
e. Adjust TCP optimizer setting to improve video resolution and throughout rate.
f. Fast Deployment of missing technology.

Team Structure was as below,

This report describes the approach of using Optimization parameters and techniques in
improving KPIs such as Service access delay, DL/UL Throughput, etc.

Huawei Confidential 3|Page

 1. Description of Problems/Challenges

Huawei performed P3 like benchmark tests (in two major cities) all through the year of 2016 in
each of the Quarters. Q1 results were deeply analyzed and certain focus areas were selected
for optimization.

CS Benchmark Analysis:

CSFB KPIs and breakdown analysis showed OPERATOR V’s short comings in many areas, few
are listed below:

• Large call setup duration for CSFB

• Overall lower MOS than competitors.
• Call setup failures/Drops due to coverage/quality (NCCI/ACP project was setup for
overall RF tuning)
• U2L cross border issues causing calls to be initiated in UMTS rather than LTE.
• Reselection from U2L causing missed pages
• UL issues such as incorrect RACH settings.
• Missing neighbors and inconsistent baseline settings.
• No RRC Release received by UE causing RRC state mismatch

Huawei Confidential 4|Page

PS Benchmark Analysis:

PS KPIs and breakdown analysis showed OPERATOR V’s short comings in many areas, few
are listed below:

• Call setup failures/Drops due to coverage/quality (NCCI/ACP project was setup for
overall RF tuning)
• Poor CA performance with low DL grants.
• Poor throughput during handover (large number of handovers in network due to poor RF
tuning).
• Poor 10%ile and average throughput performance.
• Transmission bandwidth limitation/Packet loss/DNS Failures.

Huawei Confidential 5|Page

 2. Analysis, Strategies and Actions

2.1 CS Parameter tuning

2.1.1 CSFB setup time delay

A breakdown analysis was performed and certain actions were performed to improve call setup
delay. Good improvement in Call setup delay was noted (see chart below).

Parameter/switches tuning performed:

Adding onto the above actions, CN team implemented LAC/TAC mapping. Also, Flash CSFB
with RIM was implemented to save delay between LTE RRC Release and UMTS RRC Request.
These actions and good improvement in RF conditions due to RF tuning using ACP
compounded into good improvement in CSFB delay for both 4G-4G and 4G-3G scenarios.

Huawei Confidential 6|Page

Good improvement in setup delay allowed for OPERATOR V to have top score for call delay in few of
the cities and towns.

2.1.2 Cross procedure optimization for CS/PS Qualifiers

Cross procedure cases were identified from the P3 log analysis which were affecting CS call setup and
causing unqualified call setup. Few of the cross procedure actions were performed and KPIs were
observed for validation as reproducing failures may not be possible in normal testing scenarios.

Good improvement in Paging and Cell Update SR noted along with improved CS/PS Drop rate.
Reduction in RRC and RAB attempts due to Paging and CU overlap optimization as users stay in PCH.
Overall traffic/User number/throughput is maintained.

Huawei Confidential 7|Page

 2.1.3 U2L/FR Redirection Optimization for CS and PS calls
Several CS calls were noted to have initiated from UMTS rather than LTE from the Q1 drives. Also,
several call setup failures were noted due to page being missed while the UE is performing reselection
from UMTS to LTE.

The main cause for UE to remain in UMTS after the CS call was due to the UE crossing RNC border and
then remaining in UMTS until idle mode and then reselecting back to LTE. When the UE crosses the
RNC border the information that the UE is a CSFB UE is not transferred to the DRNC. Related switches
were enabled and cross border U2L/FR issue was eliminated. Few areas still had UEs being released in
UMTS due to the U2L/FR strategy (non LTE co-located UMTS cells not enabled with U2L/FR).

Further, the U2L Redirection parameters were tuned for quicker return to LTE for PS users that handed
over to UMTS (punishment timer & periodic timer changed from 20s to 10s, Hysteresis 3C=0) and Low
activity trigger for users performing D2F or D2P.

CSFB IE over IuR:

As seen from the above DT results, border area U2L issue has been eliminated. Samples for Calls
starting from UMTS still exist due to no Fast Return enabled in U900 cells and sites with no co-located

Huawei Confidential 8|Page

high band LTE cells.

U2L Timers and Low activity switch tuning:

Large increase in attempts for U2L redirection noted with other KPIs being stable. Parameters were
rolled out and overall LTE %age increased.

2.1.4 LAC/TAC Optimization

Some of the LAC areas were optimized due to unreasonable plan causing LAC/TAC updates to occur in
busy areas. Due to the unnecessary LAC/TAC updates, paging messages would be missed resulting in
missed call setup and therefore impacting CS call qualifiers KPI.

Paging capacity was noted to be available and there was no congestion due to layered paging being
previously rolled out. However, unreasonable LAC borders needed optimization.

Area was studied and LAC/TAC border was moved away from busy areas. Board capacity and RNC
load needs to be reviewed before planning out the areas.

Below DT picture shows the pre and post LAC areas.

Huawei Confidential 9|Page

 2.1.5 Other Issues (Neighbor missing, baseline alignment, equipment issues, etc)
From the P3 logs analysis, missing neighbors, incorrect RACH settings and other baseline misalignment
issues were identified. A network wide audit was done and baselines were corrected for all
eNBs/NBs/RNCs in the network by OPERATOR V.

Sample missing neighbor issue:

Several call drops reasons were noted to be due to missing neighbors from P3 logs analysis. A sample
analysis is below. Suggest to perform neighbor audit and deploy SON related features if available.

Sample for RACH issue:

Few samples of call setup failures were detected due to incorrect RACH settings. Incorrect baseline
setting can lead to unnecessary failures. On further indoor testing, more setup failures were found due to

Huawei Confidential 10 | P a g e
low UL Tx Power for the test UEs. 1. PreambleRetransMax was modified to 32 to improve the possibility
to receive RRC CONN REQ message and the issues were resolved.

Sample for test equipment issue:

During one of the rounds of testing, large numbers of drops were noted for 2 handsets in a test car and
the handsets in other car were noted to have low drops as expected. Typical scenario for the drop was
as follows; when NODEB detected an uplink synchronization failure (uplink quality bad), NODEB sent a
RL FAIL IND to RNC and triggered call re-establishment with RL ACT CMD message, but there was no
cell update report from UE with new radio link for the call re-establishment in 5 seconds and the call was
released. Call drops occurred in good DL RF conditions at reasonable TP distances. It was suspected
that the antenna system in the problem car had incorrect settings causing poor sensitivity in UL for the 2
handsets. Issue was raised to P3.

Huawei Confidential 11 | P a g e
 2.1.6 MOS improvement
Overall MOS for all operators in Country A is good, however, OPERATOR V MOS values were noted to
have slightly abnormal behavior where the trend suddenly changes at a certain point. Following steps
were taken to narrow down the issue

a. Drive test analysis was analyzed and the issue was noted to be across the entire drive area.
b. Top worst VQI cells were analyzed for any issues.
c. Detailed DT was performed with different locations and scenarios. Testing was performed using
different handsets.
d. From tests performed between different handsets, operator V test handset was noted to have the
issue while handset from other operators had good performance.
e. Issue was narrowed down to the software version of Operator V.
f. Handset vendor provided an updated version and the issue was resolved.

Huawei Confidential 12 | P a g e
 2.2 PS Parameter tuning
PS parameters tuning was mainly based on LTE related parameters as UE remains most times in LTE
rather than UMTS. CA performance, MIMO and throughput during handover were the main focus for
optimization.

2.2.1 LTE Throughput Optimization (During Handover/CQI Steps/Interference avoidance)

The First batch of parameters and switches tuning performed for LTE DL throughput improvement were
related to MIMO, CQI steps and RB allocation as the OPERATOR V network was noted to have several
handovers to overshooting cells as the RF tuning needed much improvement to have dominant servers
rather than ping pong handovers. These large number of handovers caused unnecessary step up and
down of MIMO configuration, CQI step, etc.

Below mentioned parameters were implemented and throughput during and after handover were
improved.

Implementation was done in an area with several handovers with mixed coverage scenarios (good
coverage and poor coverage). Comparison table below shows different KPIs and the improvement they
brought.

Results were evident from KPIs as well, below charts show some of the KPIs and the impact of the

Huawei Confidential 13 | P a g e
parameters on them.

• DT and KPIs such as MIMO Rank2 %age, MCS and Throughput show improvement with the
trials.
• Overall, no KPI degradation noted.

2.2.2 LTE CA performance Optimization

LTE Scell activation optimization:

LTE CA performance in the network was noted be poor with large number of Scell configurations while
low activation or delayed attempts.

Huawei Confidential 14 | P a g e
Active Buffer length and delay settings were default (9KB/50ms) making it difficult to activate Scell and
PCell would handle all the data traffic. Few recommendations were trialed and implemented.

With the above setting SCell activation rate increases and more users can perform DL based on CA.

After implementing CA Instant Activation Strategy, more small packets will be transmitted with SCC
activated, which will cause the “CA User DL Throughput” KPI decrease, the reason is: Throughput =
Traffic Volume / Traffic Time, more small packets are taken into account will cause the Traffic Time
Huawei Confidential 15 | P a g e
increased more significantly compared to the Traffic Volume. However, the user experience will be
improved due to the user has more chance to be scheduled in the MAC layer.

LTE 3CC Activation:

At the beginning of 2016, OPERATOR V network was ready to be trialed and deployed with 3CC as it
had 3 carriers of LTE with up to 40MHz of total bandwidth in most of the cities. Band combinations of
B3/B3/B5 and B1/B3/B5 were available and compatible handsets such as Samsung Note 5 and Galaxy
S7 were recently launched, also handsets from other vendors were now being available and increasing
the requirement of 3CC to be provided to users.

Above PCC/SCC strategy with CA Group based configuration was trialed and deployed in early 2016
bringing in a good boost to the DL throughput KPI.

LTE CA DL Grant improvement:

In OPERATOR V network, SCell DL Grants were noted to always be low with almost all RBs being
available from the carrier. There were no limitations on the server side of the transmission bandwidth as
the PCell DL Grants were at healthy levels. DRX receive DTX Process Optimization was trialed and
improvements in DL grants for SCell were very evident.

Huawei Confidential 16 | P a g e
From DT, Obvious increase in DL throughput was noted in the SINR 0 to 15 ranges (most test samples)
with similar RB counts as Pre data and much higher DL RB grants. This change was double confirmed
with a rollback of the switch and the grant/throughput levels returned to previous levels. These changes
were verified from network KPIs as well, as follows.

Huawei Confidential 17 | P a g e
KPIs show clear increase in SCell traffic and Throughput with increase rBler due to increased
retransmissions for UEs that were actually on DRx and not moved to DRx due to wait timer introduced
by the switch. Overall good results were observed and the optimization was rolled out over the network
resulting in better utilizing the Scell in a CA activated scenario.

2.3 Transmission issues

Transmission bandwidth plays an important role in the throughput performance and issues in
transmission side can lead to packet loss, low grants, drops, etc. One of the major issues with
OPERATOR V network is noted to be transmission bandwidth limitation of several sites and/or other
issues. A quick Tx utilization analysis (based on database available), showed several sites (4G+3G) to
have very high utilization during busy hours. Over the course of the year 2016, transmission related
issues were fixed to a great extent by OPERATOR V allowing to greatly improved user experience.

DNS Failures case:

PS data qualifier, especially for webpage download was noted to have poor performance. On initial
analysis, large number of DNS failures was suspected to be causing the failures.

Most of the qualifier failures were due to timeout (max download duration is 10s). 10.143.147.147 is the
main DNS server and if DNS fails or has no reply, secondary server 10.143.147.148 is sent a DNS
request. The trend for almost all failures was that the main server did not send a response and then the
secondary server was requested after 5s, effectively halving the download time available for webpage

Huawei Confidential 18 | P a g e
download. Sample failure session shown below,

RF conditions were noted to be normal.

P3 logs were analyzed and few areas where repeated failures occurred were selected for testing and
tests were performed at off-peak and peak hours. No failures were observed for off-peak, however peak
hours tests showed large number of DNS failures.

PCAP traces from PGW/eNB/UE were compared and end to end check for DNS messages was done,
identifying that all DNS requests were received by the PGW in the UL from UE via eNB to PGW and the
PGW responds to the request but is not received by either the eNB or the UE. This observation was
shared with OPERATOR V CN and Transmission teams and it was found that transmission capacity
from the Hub site towards a very large number of sites in the metro area were reaching full capacity
during peak hours and required expansion.

TCP window full case:

Several qualifier failures were found for data stream DL and file DL scenarios with a wireshark message
called TCP window full. It occurs when Wireshark has identified that the current packet has the exact
payload size that will fill up the receive buffer on the receiving end (based on the latest seen "Window
Size" on a packet from that receiver in the current TCP session). These cases were escalated to
OPERATOR V transmission team and they found bandwidth limitation being the cause.

Huawei Confidential 19 | P a g e
Comparing the Q1 and Final Benchmark results for Data stream DL throughput, good improvement in
throughput was observed.

2.4 Other notable observations and planned trials

DSCP QOS settings:

It was identified that the DSCP settings for UMTS had a higher priority compared to LTE in the current
network scenario where incase high utilization of transmission bandwidth occurs, LTE packets will be
dropped in favor of UMTS. This is still under and study and is planned to be corrected by OPERATOR V.

Huawei Confidential 20 | P a g e
Other cases:

From DT logs, eNB X was identified to have call drops for LTE. Further checking the KPIs found large
number of drops due to MME reasons (eUTRAN generated drops).

Further analyzing using FMA it was found that UE is reporting other PCIs in MR but no handover occurs.
On checking the neighbor list, no neighbors except for own eNB cells were added and the ANR switches
were OFF. ANR was switched ON and the issue was fixed.

Other Activities:

• RF optimization/tuning was a key action in providing a solid base of parameter/switches/features

tuning and was performed over 2 quarters of year 2016.
• UMTS low band re-farming to LTE 10MHz in metro and 5MHz in outer metro/rural areas, L21
activation in certain regions and bandwidth expansion greatly expanded the LTE capabilities and
improved performance.
• Power increase in outer metro and rural areas provided a slightly larger coverage and allowed
calls to sustain for a longer period.
• UMTS RL power increase for CS, further improved edge user performance.
• MOS, Drops rate, handover rate kpis were improved with changing SHO RL number from 3 to 4
with slight impact on capacity.
• Baseline alignment for the entire network, corrected any inconsistencies and completed the
required health check.

Future planned parameter tuning activities:

CS related parameters/switches/features:

Huawei Confidential 21 | P a g e
PS related parameters/switches/features:

Huawei Confidential 22 | P a g e
 3. Result

Huawei Confidential 23 | P a g e
A great partnership between Huawei and OPERATOR V along with relentless hard work by both parties
resulted in outstanding results in the final benchmarks. OPERATOR V was termed #1 in most cities with
best performance in both Voice and Data with the best year on year improvement compared all other
competitors.

The approach of building a strong base and adding on top enhancements translated to a consistent
performance in all test scenarios in every clutter area of the network, proving that the tried and tested
bottom top method is bound to provide results.

P3 Results for 2016 testing in County A is shown here. Based on scoring, Operator V network scoring
has improved 57% in 2 years and close the gap with other 2 operators.

The main target for Operator V was to achieve best score in major where it has most of its customers.
This target was well achieved with Operator V score being highest for almost all major cities.

Huawei Confidential 24 | P a g e
 4. Suggestion

The above details of parameter optimization activities on top of detailed re-planning, layering
strategies, transmission expansions, capacity expansions accelerated the improvements in
Operator V network in order to achieve the results. However, each network is unique and it has
its own challenges and difficulties. In order to optimize network for P3 purposes, the best
starting point is to understand area where it is possible to gain points based on gap analysis and
improve rather than blindly start optimizing everything. The approach has to be holistic and not
narrow.

P3 tests user experience and thereby P3 optimization is different than improving network KPI
and not necessary that activities above improve network KPI. Customer expectations have to
be set and managed to avoid unnecessary rework.

Finally, it is team work that pays off and both internal and external parties need to work in sync
towards the target.

Huawei Confidential 25 | P a g e