HSDPA/HSUPA

Performance Assessment
Ericsson Cluster

Airtel-Nigeria

Roshan Karunakaran – Staff Engineer

Qualcomm Technologies, Inc.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 1
 Confidential and Proprietary – Qualcomm Technologies, Inc.
© 2015 Qualcomm Technologies, Inc. All rights reserved.

NO PUBLIC DISCLOSURE PERMITTED: Please report postings of this document on public servers or websites to: [email protected].
Not to be used, copied, reproduced, or modified in whole or in part, nor its contents revealed in any manner to others without the express written permission of Qualcomm Technologies, Inc.
Qualcomm Snapdragon is a product of Qualcomm Technologies, Inc. Other Qualcomm products referenced herein are products of Qualcomm Technologies, Inc. or its subsidiaries.
Restricted Distribution: Not to be distributed to anyone who is not an employee of either Qualcomm Technologies, Inc. or its affiliated companies without the express approval of Qualcomm
Configuration Management.
Qualcomm and Snapdragon are trademarks of Qualcomm Incorporated, registered in the United States and other countries. All Qualcomm Incorporated trademarks are used with permission.
Other product and brand names may be trademarks or registered trademarks of their respective owners.
This technical data may be subject to U.S. and international export, re-export, or transfer (“export”) laws. Diversion contrary to U.S. and international law is strictly prohibited.
Qualcomm Technologies, Inc.
5775 Morehouse Drive
San Diego, CA 92121
U.S.A.

HSDPA/HSUPA Performance Assessment Ericsson Cluster – Airtel Nigeria

80-W4325-1 Rev September 2015

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 2
Outline
Project Information

Executive Summary

HSDPA Performance Assessment

• Single User Stationary Tests Analysis
• Mobility Tests Analysis

HSUPA Performance Assessment

• HSUPA Introduction
• HSUPA Network Configuration
• Single User Stationary tests
• Mobility Test Analysis

Web Browsing Performance

Idle Mode Performance Assessment

AMR Speech Performance Assessment

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 3
Project Information
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 4
Project Objective
Overview
Focus Areas
The Project scope was to assess the
End User perception for :
− HSDPA/HSUPA Data Services Data
Performance
CS on FACH
Performance
Idle Mode
Performance
Performance
− DL Throughput, UL Throughput
− HSDPA/HSUPA Mobility HSDPA HSUPA
− HSDPA/HSUPA Retainability Performance Performance
− Ping latency
− Assessment of features as observed during
initial drive testing
− Idle Mode Performance
− AMR (CS on FACH) performance
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 5
Test Details FTP Test Setup
Test Methodology & Project Timelines

PS test executed using: Stationary/Mobile UE

− Huawei Data Dongle (Qualcomm MDM9200

chipset). Multiple FTP DL/UL
… sessions in parallel
CS test executed using:
− Samsung Galaxy S4 Mini (Qualcomm
MSM8974 chipset). QXDM logs

UE is kept in Mobility scenario under Normal

Network Load conditions. Aug 3rd , 2015 Aug. 28th , 2015

Qualcomm’s QXDM used for UE tracing Analysis and

Reporting
Analysis performed using Qualcomm Post
processing tools (Qcat), post processing
scripts and Actix Software.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 6
Executive Summary
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 7
Executive Summary
Key Observations and Concerns
RRC and RAB Setup Success Rate

Accessibility for the Voice and Data Services are significantly below the RABESTABLISHSUCCESRATE 98.30%
Industry Average, mainly due to RRC Connection Rejections because of RRCCONNSETUPSUCCESSR
86.0%
Admission Control, and RRC Setup failures after Admission ATE
75% 80% 85% 90% 95% 100%

25
21.1

DL THROUGHPUT
20
User Experience for Data Services is below the Industry Average 15

[MBPS]
10 8.4
due to low HSPA Throughput, High Latency in DNS Server 5
Responses and PDP Context Activation Delays 0
Max Possible Measured
Throughput Throughput
Frequent Cell
Changes Ping-
High Non-HS Power Utilization due to Excessive Signaling and non- Pongs generates
optimal Target BLER Dl/Ul SIR settings Signaling Tsunami
and affects HSPA
Throughput

CPICH and common channel powers across the RNC is not

harmonized, resulting in RF Coverage Holes, Low HSPA
Throughput and a Call Drops

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 8
HSDPA performance
HSDPA Throughput Results for Stationary and Mobility Test Cases
~60% of the time # HS-DSCH
Mobility L1 Throughput (Mbps) codes was less or equal to 5
40% 100%

30% 80%
60%

CDF
20%
40%

PDF
10% 20%
0% 0%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
# HS-DSCH Codes
25
21.1 Stationary L1 Throughput (Mbps)
PDF CDF
DL THROUGHPUT [MBPS]

20

15
Resources (code and/or power) was the
10 8.4 main limiting factor to obtain better
5
throughput.
Site by Site Capacity and Resources
0
Utilization Review should be performed to
Max Possible Throughput Measured Throughput
remove the capacity bottleneck.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 9
HSDPA Throughput Test
Factors impacting HSDPA Throughput All factors shown on the
diagram have significant
impact on HSDPA
High Serving Cell throughput
Changes and data Low Scheduling
interruption
Due to low throughput
obtained during the
Sub-Optimal
stationary and, especially,
Limitation of
Radio Conditions available DL mobility test, the parameter
CQI Codes for HS changes and resources
(power and code)
evaluation should be
carefully done site by site.

In the areas with weak

Limited DC-
HSDPA HSDPA Limitation of
available DL HS coverage the sites
configuration in
the cluster Throughput Power inspection, neighbor lists and
over-shooters audit should
be done to investigate
coverage issues
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 10
HSDPA Throughput Test
Recommendations

High Serving Cell Unavailability of DL Unavailability of DL

Limited DC-HSDPA Sub-optimal Radio Suspected Iub
Changes and resulting Codes for HSDPA Power for HSDPA
Configuration Conditions/CQI Limitation
in data interruption Traffic Traffic

Implement
parameter Optimize
Perform
Implement changes to Baseline
detailed counter
parameter Reduce Code parameters as
study to
Increase DC changes to tree Utilization. recommended
Optimize SCC troubleshoot
and 64-QAM Reduce Code Disable higher to improve
Parameters outliers.
licenses tree Utilization. RABS. EcNo and
Qualcomm to
Trial RSCP
Reduce DPCCH share the
SRBoHSDPA. conditions
Power/ BLER counter details.
across the NW.
target.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 11
HSDPA Throughput Test
Summary & Recommendations
Excessive HSPA Serving Cell Changes (SCC) impact RNC Signaling Load and
reduce the HSPA Throughput. The frequent SCC ping-pongs are due to non-
optimized event 1d parameters
− Recommended to tune event_1h-HS parameters to reduce Ping-Pong mobility
Most of the cells don't have DC capability across the test cluster
− It is recommended to enable DC across all cells for better throughput and trucking gain
Throughput restriction shows sub-optimal allocation of 64-QAM and TBS size.
Points to DL power limitations.
− Optimize parameters to reduce R99 DL power usage and Code tree.
− Reduce BLER Target for R99 PS RABS
− Disable higher PS R99 RABS in UL and DL.
− Trial SRBoHSDPA+F-DPCH for Code tree gains
− Optimize DPCCH power offsets
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 12
HSUPA Throughput Test
Factors impacting HSUPA Throughput

Low Grant
ETFCI restriction
in 10ms TTI. UE
Low RTWP
gets 1.4 Mbps
headroom
instead of 2
Mbps.

ETFCI restriction
in 2ms TTI. UE Misdetection
gets 5.4 Mbps
instead of 5.7 of HICH
Mbps.

Restricted 2ms HSUPA Missing features

for RTWP
TTI usage
Throughput reduction

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 13
HSUPA Throughput Test
Recommendations

ETFCI restriction in
ETFCI restriction in Instances of Low Grant
2ms TTI usage 2ms TTI. UE gets 5.4
10ms TTI. UE gets 1.4 allocation across the Low RTWP headroom Misdetection of HICH
restricted Mbps instead of 5.7
Mbps instead of 2 Mbps route
Mbps

Implement Implement Optimize Enable

Optimize POC
parameter parameter eulMaxOwnUu required
Increase 2ms parameters to
changes to changes to Load and licenses for E//
TTI License. reduce
remove ETFCI remove ETFCI eulMaxRotCov features to
misdetection
restriction restriction erage reduce RTWP

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 14
HSUPA Throughput Test
Summary & Recommendations – Cat 5
Current network configuration limits the 10ms TTI peak UL physical layer
throughput to 1.45Mbps
− It is recommended to set maxChannelizationCodes = 2xSF2 to increase peak UL L1
throughput to 2Mbps

MAC Layer throughput is slightly impacted by high HARQ NACK rate (~15%)
− It is recommended to optimize HARQ NACK rate to 10% to enhance end user UL throughput
− Tradeoff between cell throughput & end-user throughput

7% cells in the 2 RNCs have eulMaxOwnUuLoad = 3dB

− EUL capacity in these cells is limited to 50%
− It is recommended to set this parameter to 8dB to increase EUL cell capacity.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 15
HSUPA Throughput Test
Summary & Recommendations – Cat 6
Current network configuration limits the 2ms TTI peak UL physical layer
throughput to 4.2Mbps and in certain sites it is even less, 3.5Mbps.
− At 78% sites, eulMaxAllowedSchRate = 4480, limiting highest SG = 31 (E-TFCI = 119 & TBS
= 8596)
− At 22% sites, eulMaxAllowedSchRate = 3968, limiting highest SG = 30 (E-TFCI = 114 & TBS
= 7173)
− It is recommended to set eulMaxAllowedSchRate ≥ 5398 with GRAKE & interference
suppression feature active to enhance the 2ms TTI peak UL data rate.
Though transmissionTargetTti2 is set to 5%, the HARQ NACK rate is in the
range of 17%.
− It is recommended to optimize E-HICH power control setting to improve the robustness of
HICH channel and avoid potential HICH misdetections.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 16
Vendor specific Features to improve HSUPA Throughput
Recommendations
Parameter Current Setting Recommendations GRAKE is a receiver technique that is
used to suppress the self-interference,
1 (Enabled)  95% of the sites Enable feature across all allowing a higher SNR and higher bit
featureStateEul2msTti
2 (Disabled)  5% of the sites cells rates

≥ 5398 for all 2ms TTI

3968  22% of the sites enabled sites with The GRAKE is applied to EUL users with
eulMaxAllowedSchRate 2ms TTI for the spreading factor
4480  78% of the sites interference suppression
feature active combinations 2xSF2 + 2xSF4,
approximately corresponding to bit rates
1 (Enabled)  95% of the sites ON for all 2ms TTI enabled above 1.28 Mbps (on the RLC layer)
featureStateGrake
2 (Disabled)  5% of the sites sites

IF GRAKE feature is ON, and, feature

1 (Enabled)  76% of the sites StateInterferenceSuppression is ON then
featureStateInterferenceSup ON for all 2ms TTI enabled
0 (Disabled)  5% of the sites
pression
- (Not set)  19% of the sites
sites set eulMaxAllowedSchRate to 5398 Kbps

transmissionTargetErrorTti2 50 (5%) 50 (5%) Increasing value of sirMaxTti2 parameter

to 10dB -12dB should improve the EUL
sirMaxTti2 80 (8dB) 100-120 (10dB -12dB)
data channel

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 17
Observation on Accessibility KPIs
Low Accessibility Rate (Data from Counters Measurements)
RRC and RAB Setup Success Rate RRC/RAB Connection failures due to Admission Control
accounts for 41% of all RRC Connection Request
RABESTABLISHSUCCESRATE 98.30% Failures. The distribution of failures due to AC is shown
at the diagram
RRCCONNSETUPSUCCESSRATE 86.0%
The other 59% of RRC/RAB Connection Request
Failures happened after the AC.
75.0% 80.0% 85.0% 90.0% 95.0% 100.0%

DeniedAdmDlChnlCode
1%
1%

25% DeniedAdmDlHw

~49% of DeniedAdmDlPwr
RRC/RAB
41% Conn
pmNoReqDeniedAdm Request DeniedAdmUlHw
4% failures are 49%
pmNoFailedAfterAdm due to lack of
59% DL Power pmNoRrcConnReqBlockTn

18%
pmNoFailedRrcConnectReqHw
2%
pmNoRejRrcConnRrcLc

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 18
DNS Server Responses Statistics
Large delays in DNS Responses Observed

Large delays in DNS responses observed,

while browsing popular websites ‘Facebook’
and ‘google’. In these cases the DNS server
was a home server, when min delay in
response time is expected.

Reasons for large DNS response time

could be DNS server(s) CPU load and high
memory utilization
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 19
PDP Context Activation Delay
PDP Context Activation plus DNS server response time are more than 2 sec in
average
PDP Context Activation Delay
Day/Time Lat Long PDP Context Activation Delay is high, negatively
(ms)
impacting user experience
Jun/30/15 07:24:07 3.46184 6.45478 1556
Jul/01/15 07:36:53 3.43013 6.45502 1847
Jun/30/15 06:37:55 3.40809 6.4441 1806
Jul/03/15 05:56:59 1767
average 1744

DNS resolution latency was in a range of ~250ms and above, reference values are much lower
− DNS server CPU load and Memory usage should be checked

Average PDP context activation delay for APN “internet.ng.Zain.com” was 1744 ms, which is
higher than reference networks
− network topology and elements load have to be revised

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 20
Idle Mode Performance – Summary of
Power settings for Primary and
Observations Secondary synchronization channels
Synchronize might impact performance in cell Edge/
Poor RF
Read
BCCH
BCCH power offsets are not set correctly
for some sites, which could impact SIB Cell reselection Parameters, are
decoding in Cell Edge / Poor RF Cell Reselection exhaustive foe UE battery and in some
cases restrict the UE from camping on
the best available cell.
Receive
Paging
PICH and PCH power offset are not set
correctly to many sites impacting the
reachability of the users, and DRX cycles RACH
RACH Parameters are not
could be extended to enhance battery life. homogeneous in the network with many
templates available, most of them could
L.A.
Update induce high UL interference.
LAC Borders need to be redesigned in
order to improve signaling load, Idle
Operations
accessibility and mobility for UEs

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 21
HSDPA Performance Assessment
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 22
Introduction

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 23
What Single User Throughput Shell be Expected ?
Per layer Throughput Overview
1. Max Throughput 27.95
42.19

21.10

3GPP Layer 1 7.21

10.13
13.98

3.65
1.82

2. Max Throughput 21.10

27.95
42.19

13.98
UTRAN Layer 1 7.08
10.13

3.58
1.72

41.64

3. Max RLC 13.79

20.81
27.57 Cat 24 TP (Mbps)
Cat 16 TP (Mbps)
Throughput 3.36
6.72
9.60
Cat 14 TP (Mbps)
1.60
Cat 10 TP (Mbps)
37.85 Cat 9 TP (Mbps)
25.07
4. Consider average 12.53
18.92
Cat 8 TP (Mbps)
8.73
Cat 6 TP (Mbps)
10% S-BLER 1.45
3.05
6.11
Cat 12 TP (Mbps)

36.69
24.29

5. Discount 8.46
12.15
18.34

5.92
TCP/IP headers 1.41
2.96

0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 45.00

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 24
Summary of HSDPA Test Cases
Test Case Description Comments
Stationary test with FTP Download for
Single User Stationary - Good RF. No MC support
UE Categories (CAT 24, CAT 14)
58% of the cells do not have MC
support. This sub-optimal setting
Mobility Test Mobility test with FTP Download
impacts the DL throughput. NW
serves Cat 24UE as Cat 14 device

42%
No MC Support
MC Support Enabled
58%

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 25
Near Cell Stationary Test
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 26
Static Test – CAT 14 – Near Cell
Single Carrier Throughput Performance
RF conditions : the test was done in good RF in front of the site and KPI – Average Primary
with CQI average maintained ~ 29.6 Sched Phy Rate [kbps] 8748
Average served rate was 8351 Kbps, which is considered low if
Served Phy Rate [kbps] 8351
compared to UE capability
Scheduling is showing good values (>80%), however, number of MAC Rate [kbps] 7470
served codes was low. SBLER on 1st Tx% 11

Scheduling Rate% 95.5

Average MAC Layer CQI 29.6

throughput ~ 7.5 Mbps
64 QAM utilization 98.4

16 QAM Utilization 1.3

QPSK 0.3

Served #Codes 6.5

Residual BLER 0

Transport Block Size 17497

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 27
Static Test – CAT 14 – Near Cell Single Carrier
Stationary Test Time-Line

MAC Layer Throughput

MAC Layer throughput
occasionally reaches 18000
Kbps when UE was scheduled
with 13 HS-DSCH codes and,
HS-DSCH codes
correspondently, with larger
TBS size

Most of the time (75%) UE

TBS scheduled
was scheduled only with 5-6
codes and TBS less or equal
to 16648 bits

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 28
Static Test – CAT 14 – Near Cell
TBS and HS-PDSCH Codes Distribution 11 HS-PDSCH 12 HS-PDSCH
10 HS-PDSCH
0% 2%
0% 13 HS-PDSCH
TBS Utilization 9 HS-PDSCH 1% 5 HS-PDSCH
35 31.25 8 HS-PDSCH 17% 47%
0%
30
25 23.42 7 HS-PDSCH
5%
Percentage

20
15 11.94
10 7.22
3.99
5 1.432.26
0

6 HS-PDSCH
TBS 28%

Less than 10 codes were used in 97% of the time. Code and/or power starvation could limit the DL
throughput significantly

In 60% of the time scheduled TBSs were equal or less than 16648 bits  corresponding to 8.3
Mbps, which is only 40% of the maximum data rate of 21.1 Mbps for CAT 14 UE
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 29
HS-DSCH category
UE CAT 24 (CAT 14 for Single-Cell) Specification

25
21.1
DL THROUGHPUT [MBPS]

20

15
It seems that resources (code and/or Tx
10 8.4 power) was the main limiting factor to obtain
5
better throughput

Max Possible Throughput Measured Throughput

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 30
Static Test – CAT 14 – Near Cell
Proposed Changes: Increase available codes for HSDPA
Parameters Current QC Recomm Minimizing channel switching and limiting/de-activating high
R99 bit rate RC is an efficient way to increase code
RabCombination016 featureState 1 0 availability for HS and minimize Code congestion

RabCombination018 featureState 1 0
While R99 only users will be impacted, HS users are
1 0 supposed to have more resources, which will be translated to
RabCombination028 featureState
more throughput  1 User with R99 384 Kbps uses 1SF8, if
RabCombination029 featureState 1 0 spared could provide 2 SF16 codes to HS codes pool.

RabCombination030 featureState 1 0

RabCombination031 featureState 1 0
Guaranteed HS admission requests of Spreading Factor 4 in the uplink
1 0 (384kb) are denied when the use of Spreading Factor 4 in the uplink exceeds
RabCombination032 featureState
sf4AdmUl
RabCombination052 featureState 1 0
Guaranteed HS or Non-Guaranteed admission requests of Spreading Factor 8
sf4AdmUl 4 0 in the uplink (128kb) are denied when the use of Spreading Factor 8 in the
uplink exceeds sf8AdmUl
sf8AdmUl 8 4 Non-Guaranteed admission requests of Spreading Factor 8 in the downlink
are denied when the use of Spreading Factor 8 exceeds sf8Adm. The 384kb
sf8Adm 8 0 Radio Access Bearer uses Spreading Factor 8 on the downlink.
sf16Adm 16 0 Non-Guaranteed admission requests of Spreading Factor 16 in the downlink
are denied when the use of Spreading Factor 16 exceeds sf16Adm . The
,RncFeature=IfIratMobilityHsdpaEul 1 0 128kb Radio Access Bearer uses Spreading Factor 16 on the downlink.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 31
Mobility Test
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 32
 Mobility test was done with CAT 24 UE, but

Mobility Test Radio Conditions Overview since Multi-carrier was not activated on all
cells in the test cluster, UE was served by
single carrier during entire route and was

Observed RF Conditions were below the Industry Average considered as CAT 14. Correspondently,
HSDPA throughput was altered according to
UE category

The actual targets for

Ec/No and RSCP
depend on the NW
design objectives and
can differ from Operator
to Operator

RSCP target of 95%

better then-95 dBm
assumes that the NW
was designed to
provide in-building
coverage

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 33
Mobility Test Summary
Single Carrier Throughput Performance (CAT 14 UE)
Average MAC Layer KPI - Average Primary
throughput 1640 Kbps Sched Phy Rate [kbps] 2681
Physic L1 Rate [kbps] 1660
MAC Rate [kbps] 1640
SBLER on 1st Tx% 11
Scheduling Rate% 46.2
CQI (average) 20.6
64 QAM utilization % 16
16 QAM Utilization % 45.6
QPSK % 38.4
Served #Codes 6.2
Residual BLER 0.3
Transport Block Size
6315
bits (average)

Observed MAC Layer throughput

is significantly lower than expected
CAT 14 values

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 34
Mobility Test Summary
Mobility Test Time-Line Sample. Code Limitations Observed
The entire test was done with CAT 24
UE, but since Multi-carrier was not
MAC Layer Throughput activated on all cells, the UE was
considered as CAT 14 along the entire
route

At times, when UE was scheduled with

14 codes, the throughput has reached
the peak value of 14000 Kbps

It seems that number of available codes,

and, possible, NodeB Tx power, were
Average # of scheduled the limiting factor for HSDPA throughput
HS-DSCH codes

Low utilization of 64QAM due to low

Ec/No and CQI values has also
contributed to the degraded HSDPA
performances

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 35
Mobility Test Summary
Good Scheduling Rate and good CQI – still Low Throughput
UE was considered as CAT 14 device
along the entire route

Maximum L1 throughput one can expect

from CAT 14 UE is 21.1 Mbps

Even under good reported CQI (25-26),

64 QAM modulation and 95%-100%
scheduling rate, UE was served only
with 8-9 HS-DSCH codes and relatively
small TBSs, resulting in MAC layer
throughput of only ~4000 kbps

The observed throughput in good

coverage area was only about 20% from
the maximum possible for CAT 14!

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 36
Mobility Test Summary
Good Correlation between Weak Coverage and Low Throughput
There is an obvious correlation
between low RF coverage (RSCP
map shifted by -200 m south), low
Ec/No (map shifted by 200 m North),
and, low throughput (MAC layer
throughput map, no shift).

Three PS Call Drops happened

in the area of weak coverage.
ASET Best Pilot E/No
The Central area of the test route
MAC Layer Throughput should be inspected to find the
cause of weak coverage.
ASET Best Pilot RSCP Neighbor list discrepancies
should be also checked

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 37
HSDPA Mobility Tests
HS Serving Cell Pin-Pongs: Event 1d Parameters need Adjustments
Many areas suffer
from HS Serving cell
ping-pong

With no SHO for

HSDPA, the data flow
is interrupted each
time a new serving cell
is selected
Too many Cell
changes and Ping-
Pong areas impact the
throughput negatively

Intensive serving cell

changes also create
significant RNC
signaling load
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 38
Mobility Test Summary
Excessive HSPA Cell Changes Observed
Excessive HSPA Serving Cell Changes (SCC) impact RNC
Signaling Load and reduce the HSPA Throughput

The frequent SCC ping-pongs are due to non-optimized

event 1d parameters

Current Recommen
Parameter Comments
Setting dations
Increasing
hsTimeToTrigger1d from
640ms to 1280ms has
hsTimeToTrigger1d 640 ms 1280 ms shown that the number of
ping pong cell changes
reduces, which has lead to
an increase in throughput.

Testing on
hsHysteresis1d has
shown that an increase
hsHysteresis1d 1 dB 2 dB
from 1dB to 2dB results in
a reduction of mobility In the above sample, 6 SCCs
events at whole RNC level
happened during the 14 sec
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 39
HSDPA Mobility Tests
HS-DSCH Codes Distribution and Cell Changes Time Statistics
40% 100% Only ~15% of the
80% time # of
30%
assigned HS-
60%
DSCH Codes

PDF
20% ~60% of the time # HS-DSCH
codes was less or equal to 5 40% was more than
CDF

10% 10
20%
0% 0%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Limitation on
# HS-DSCH Codes
codes
assignment limits
PDF CDF the scheduled
TBSs and leads
to low throughput
values
Average cell change time = 319 msec which
is higher than on the reference networks

Data Interruption time was < 100 mses,

which considered to be a satisfactory
number
Parameter Current Value QC Recom.
hsHysteresis1d 10 (1dB) 20 (2 dB)
hsTimeToTrigger1d 640 ms 1280 ms

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 40
HSDPA Throughput Test
Factors impacting HSDPA Throughput All factors shown on the
diagram have significant
impact on HSDPA
High Serving Cell throughput
Changes and data Low Scheduling
interruption
Due to low throughput
obtained during the
Sub-Optimal
stationary and, especially,
Limitation of
Radio Conditions available DL mobility test, the parameter
CQI Codes for HS changes and resources
(power and code)
evaluation should be
carefully done site by site.

In the areas with weak

Limited DC-
HSDPA HSDPA Limitation of
available DL HS coverage the sites
configuration in
the cluster Throughput Power inspection, neighbor lists and
over-shooters audit should
be done to investigate
coverage issues
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 41
HSDPA Optimization Focus Areas
HSDPA Throughput Improvement Steps

L1, L2, L3 Timers Alignment

CPICH Power Optimization

HSDPA
Performance Increase Dual-Cell Capability
Optimization
Dl Code/Power Optimization to reduce
Congestion

Minimize Serving Cells Ping-Pongs through

e1D settings optimization

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 42
 Ways to Reduce high R99 DL Power & Code Usage
Based on current Parameter setting in the test RNC

Power

PWR Available: HS-PDSCH + HS-SCCH

 Reduce DPCCH Power usage

Total cell Power

 Disable PS 384 & 128 in DL

Dedicated channels (power  Introduce SRBoHSDPA+FDPCH
controlled)  Increase BLER Target for PS RABs
AMR + PSR99 + SdAlone_SRB +
HSUPA control + HS_A-DCH

Common channels

To enhance HSDPA performance, Downlink Power usage and Code tree Utilization should be optimized.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 43
Summary & Recommendations
HSDPA Throughput Performance
Excessive HSPA Serving Cell Changes (SCC) impact RNC Signaling Load and
reduce the HSPA Throughput. The frequent SCC ping-pongs are due to non-
optimized event 1d parameters
− Recommended to tune event_1h-HS parameters to reduce Ping-Pong mobility
Most of the cells don't have DC capability across the test cluster
− It is recommended to enable DC across all cells for better throughput and trucking gain
Throughput restriction shows sub-optimal allocation of 64-QAM and TBS size.
Points to DL power limitations.
− Optimize parameters to reduce R99 DL power usage and Code tree.
− Reduce BLER Target for R99 PS RABS
− Disable higher PS R99 RABS in UL and DL.
− Trial SRBoHSDPA+F-DPCH for Code tree gains
− Optimize DPCCH power offsets
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 44
Changes for Interactive RABs BLER Targets
New values for blerQualityTargetUl and blerQualityTargetDl (1of 2)
Current settings of blerQualityTargetUl/Dl are not
uniform and are not optimal.

The proposed changes will increase BLER target

for interactive RABs.
The changes should reduce the transmitted power
on uplink and downlink. This has resulted in
improved RSSI in uplink and lower power
consumption in downlink.
The parameter change also improves capacity in
uplink and downlink.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 45
Changes for Interactive RABs Bler Targets
New values for blerQualityTargetUl and blerQualityTargetDl (2of 2)

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 46
HSUPA Performance Assessment
Ericsson Cluster
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 47
Introduction

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 48
Summary of HSUPA Test Cases

Test Case Description Comments

Stationary test with FTP upload for
Single User Stationary Near cell
10ms and 2ms TTI

Mobility test with FTP upload in

Mobility Test
selected route

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 49
Key HSUPA Performance Indicators
General UE Side KPI Definition

Average Physical Layer Rate Physical Layer rate over all received blocks (new and Re-transmitted blocks)

Average Scheduling Serving Grant Average assigned grant by Network by means of E-AGCH or E-RGCH

Sub-block Error Rate on all UL transmissions. This is calculated from UE side based on
% UL BLER
received E-HICH ACK/NAK/DTX

% of re-transmissions that don’t recover at HARQ level after reaching the maximum
% UL Residual BLER
number of UL re-transmissions

% Throughput limited by SG % of time a certain E-TFCI is selected based on the assigned SG

% of time a certain E-TFCI is selected based on the available buffer in the UL (e.g. grant
% Throughput limited by Buffer Occupancy
may be available, but UE is limited by buffer occupancy)

% of time a certain E-TFCI is selected based on the available Power in the UL (e.g.
% Throughput limited by Maximum UE Tx Power
grant/buffer may be available, but UE is limited by UL Tx power)

% Happiness % of time UE reports Happy in Happy Bit transmitted on E-DPCCH

RF Related KPIs EcNo, RSCP, UE Tx Power

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 50
HSUPA Operation Overview

HSUPA Operation
1. The UE sends a Transmission Request (Happy
Bit or SI) to the Node B for getting resources.

2. The Node B responds to the UE with a Grant

Assignment (initially on E-AGCH from Serving
3dTower.emf

EUL Cell), thus allocating uplink band to the

Node B UE.

3. The UE uses the grant to select appropriate

transport format for the Data Transmission (E-
TFCI) to the Node B.

4. The Node B attempts to decode the received

data and send ACK/NAK (on E-HICH) to the
UE. In case of NAK, data may be retransmitted.

5. Network can adjust the grant using E-AGCH

(from serving cell) and E-RGCH (from serving
or non-serving EUL cells)
UE

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 51
HSUPA UE Capabilities
Max number of E- Peak rate for TTI = Peak rate for TTI =
E-DCH Category Minimum SF Supported TTI
DPDCH channels 10 ms* 2 ms

Category 1 1 SF 4 10 ms 711 kbps --

Category 2 2 SF 4 2 & 10 ms 1448 kbps 1448 kbps

Category 3 2 SF 4 10 ms 1448 kbps --

Category 4 2 SF 2 2 & 10 ms 2000 kbps 2886 kbps

Category 5 2 SF2 10 ms 2000 kbps --

Category 6 (used
4 SF2 + SF 4 2 & 10 ms 2000 kbps 5742 kbps
for test)

• Maximum Peak data rate for 10 ms E-DCH TTI operation is 2 Mbps

• For 2 ms TTI, 4 codes are required to achieve 5.76 Mbps. 4 codes are sent as 2xSF2 + 2xSF4

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 52
Peak User Throughput for different EUL Categories
3gpp Max 5.74
2
phy layer 1.45

UTRAN Max 5.74

2
phy layer 1.45

RLC 5.44
1.89
throughput 1.38

Remove 5.29
1.84
TCP/IP headers 1.34

IOT Max FTP 5.2

1.77
Throughput 1.27

0 1 2 3 4 5 6
Cat. 6 TP [Mbps] Cat. 5 TP [Mbps] Cat. 3 TP [Mbps]

Typical Breakdown of HSUPA UL Throughput at different protocol layers

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 53
Serving and Non-Serving E-RGCH & E-HICH
Network Configuration There are three different types of Radio Links in the UE
Active Set. Current Network configuration is as follows:
Example with an Active
Set of 4 cells Serving E-DCH Cell – The cell from which UE receives
E-AGCH from scheduler
− In Ericsson, E-RGCH is not configured from the serving cell
3dTower.emf

Node B
Serving (E-DCH) RLS – Set of cells that contain at
Serving
least the serving cell and from which the UE can
E-DCH cell
(cells from Non-Serving Radio receive and combine the serving RGCH
same NodeB) Links (RL) − Because is NO E-RGCH is configured by Ericsson on Serving E-
DCH Cell, then the Serving RL Set does not exist in this
configuration
− However, different E-HICH can be received/combined from Serving
RLS

Non-Serving RL – Cell that belongs to the E-DCH

Active Set but does not belong to the serving RLS and
from which the UE can receive a E-RGCH
−
3dTower.emf

Any cell added to E-DCH ASET will be considered as Non-Serving

Node B
3dTower.emf
RL from E-RGCH perspective
Node B Serving E-DCH Radio
Link Set (RLS)
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 54
 Ericsson does not use E-RGCH
UL Scheduling in Ericsson from Serving RLS
How things are done?
E-RGCH Grant from Non- E-RGCH Grant from
E-AGCH Absolute Grant
Serving Cell Serving RLS
from Serving Cell

Only Down/Hold can be Down, UP or Hold can

Apply T2P Power
sent be sent
translated into # of bits

UE reduces Grant by one Grant UE UP Grant depending on

Maximum E-TFCI selected based threeIndexStepThreshold &
depending on the Last Used Power Ratio,
on E-AGCH Grant Sent twoIndexStepThreshold. DOWN grant
LUPR (i.e. last HARQ Payload Bits)
by one Grant, with reference to LUPR
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 55
Non-Scheduled MAC-d Flow
Control Plane

With this configuration, UE can send UL NAS

SRB messages at any time even when
RRC
Scheduling Grant is not assigned
144 bits RLC
− Pre-assigned non-scheduling grant at call SDU
SRB #1 SRB #2 SRB #3 SRB #4

setup, guarantees SRB to always be RLC-AM RLC-AM RLC-AM RLC-AM

RLC
transmitted on UL  162 bits reserved for
SRB
DCCH #1 DCCH #2 DCCH #3 DCCH #4
MAC-d
This configuration is the recommended one Non-Scheduling Grant
to: MAC-d Flow #1

− Increase Control Plane reliability MAC-es/e

Scheduled
MAC-es/e
− Reduce latency on UL SRB MAC-e Payload from Flow #1
− Mitigate the risk of dropping calls Header

− Improve Handover performance, especially E-DCH

when Measurement Report Messages are PHY
sent with less latencies E-DPCCH E-HICH
E-DPDCHs

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 56
Airtel Nigeria – HSUPA Network
Configurations

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 57
Airtel Nigeria HSUPA Network Configuration
Parameter 2ms TTI Configuration 10ms TTI Configuration
e-TFCI-TableIndex Table 0 Table 1
maxChannelizationCodes sf4x2-and-sf2x2 Sf4x2
e-DCH-MinimumSet-E-TFCI 31 4
E-DPCCH Power Offset (Beta_EC) 5 (0 dB) 4 (-2 dB)
RLC PDU Size = 144 bits RLC PDU Size = 144 bits
Non-Scheduled MAC-d Flow Parameters
HARQ Power offset = 0 dB HARQ Power offset = 0 dB
(SRB)
Max # of Re-tx = 7 Max # of Re-tx = 3
RLC PDU Size = 336 bits RLC PDU Size = 336 bits
HARQ Power offset = 0 dB HARQ Power offset = 0 dB
Scheduled MAC-d Flow Parameters
(PS RAB) Max # of Re-tx = 7 Max # of Re-tx = 3
E_DCH_reference_ETFCI = [[31, E_DCH_reference_ETFCI = [[3, 11], [53,
17], [124, 24]] 24], [74, 23], [86, 24]]

Multiplexing of E-DCH MAC-d flows FALSE FALSE

Maximum Possible UL Physical Layer

Throughput achieved with these 5.7 Mbps 1.4 Mbps
parameters

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 58
Scheduling Serving Grant to E-TFCI Mapping – 10ms TTI
100 97 1600

ULL Physical Layer Throughput [Kbps]

90 Maximum UL Throughput 85 1400
80 supported by Airtel Nigeria
73
configuration is 1.45 Mbps 1200
70 63
60 1000
53
E-FTCI

50 43 800
40 37
31 600
30 27
23
19 400
20 15
11
10 7 7 200
3 3 3
0 0
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Serving Grant

Selected E-TFCI UL Physical Layer Throughput [Kbps]

Based on the Network configuration:

− Minimum Scheduled E-TFCI UE can select = 3 (354 bits) with Serving Grant of 11
− Maximum Scheduled E-TFCI UE can select = 97 (14484 bits) Serving Grant of 28
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 59
Serving Grant to Channelization Codes Mapping – 10ms TTI
Scheduling Grant E-TFCI UL Spreading Factor
11 3 sf32
12 3 sf32 UE UL Category 3 allows up to 2 codes
13 3 sf32 usage (2 codes of SF4) – 1.45 Mbps
14 7 sf16
15 7 sf16
UE UL Category 5 allows up to 2 codes
16 11 sf8
usage (2 codes of SF2) – 2.0 Mbps
17 15 sf8
18 19 sf4
19 23 sf4 Maximum Channelization Code of 2xSF4 is
20 27 sf4 possible with the current Airtel10ms TTI
21 31 sf4 configured parameters – limiting the UL
22 37 sf4 throughput to maximum of 1.45 Mbps even
23 43 sf4 for Cat. 5 devices
24 53 sf4
25 63 sf4x2
26 73 sf4x2
27 85 sf4x2
28 97 sf4x2
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 60
Scheduling Serving Grant to E-TFCI Mapping – 2ms TTI
140 Maximum UL Throughput 7000

UL Physical Layer Throughput [Kbps]

125 127
supported by Airtel Nigeria 119
120 114 6000
configuration is 5.74 Mbps 107
101
100 94 5000
88
80
75
E-TFCI

80 69 4000
61
60 50 50 3000

40 31 31 31 2000

20 1000

0 0
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Serving Grant

Selected E-TFCI UL Physical Layer Throughput [Kbps]

Based on the Network configuration:

− Minimum Scheduled E-TFCI UE can select = 31 (355 bits) with Serving Grant of 17
− Maximum Scheduled E-TFCI UE can select = 127 (11484 bits) Serving Grant of 33
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 61
Serving Grant to Channelization Codes Mapping – 2ms TTI
Scheduling Grant E-TFCI UL Spreading Factor
UE UL Category 6 allows up to 4 codes
17 31 sf4
usage (2 codes of SF 2 + 2 Codes of SF
18 31 sf4 4)
19 31 sf4
20 50 sf4
Maximum Channelization Code of
21 50 sf4
2xSF2 + 2xSF4 is possible with Airtel
22 61 sf4
configured parameters
23 69 sf4
24 75 sf4x2
25 80 sf4x2
Usage of 4 codes start with Serving
26 88 sf4x2
Grant = 30 (E-TFCI = 114)
27 94 sf2x2
28 101 sf2x2 These E-TFCIs are possible to be used by the
29 107 sf2x2 Network Configuration, but not currently utilized by
30 114 sf4x2-and-sf2x2 Network scheduler
(refer to next section for explanation)
31 119 sf4x2-and-sf2x2
32 125 sf4x2-and-sf2x2
33 127 sf4x2-and-sf2x2
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 62
10ms TTI Near-cell Stationary
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 63
Peak Throughput Test Results
Outdoor, Near Cell – UL Throughput Statistics

KPI [Average Values] EUL Cat. 5

E-DCH ASET Size 1.0
Avrg Ec/No [dB] -8.8
Avrg RSCP [dBm] -50.0
Avrg UE Transmit Power [dBm] -17.5
TTI Usage [%] 10ms TTI = 100%
Average Serving Grant 26.4
Average Physical Layer UL Throughput [Kbps] 1176.6
Average MAC Layer Throughput [Kbps] 1000.1
Average UL Scheduling Rate [%] 100%
Average UL SBLER [%] 15.0%
Average UL Residual BLER [%] 0%
Average Happiness reported by UE [%] 0%

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 64
UL Throughput Impact in 10ms TTI
Factors Impacting UL Physical Layer Throughput

UL T HRO UG HPUT U L T H R O U G H P U T L IMIT IN G FAC T O R S

1600 60%
1448 19% loss in peak
average throughput 51.8%
1400 50% 48.2%
1176.6
UL Throughput [Kbps]

1200
40% Current network configuration
1000 doesn’t allow using 2xSF2
30%
codes, hence the reason why
800 max. spreading factor is seen
600
as a limiting factor.
20%

400
10%
200
0.0% 0.0%
0%
0 Serving Grant UL Buffer UL Max Tx Power Max Spreading
Max. Possible Throughput L1 Throughput Occupancy Factor

Current network configuration allows using 2xSF4 codes, thus limiting the peak UL throughput to 1.45Mbps.
− It is recommended to set “maxchannalizationcodes” = 2xSF2 to enhance the peak 10ms TTI Uplink throughput to 2Mbps.
The L1 throughput in the example above was primarily limited due to “Serving Grant” (refer next slide)
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 65
UL Throughput Impact in 10ms TTI
Serving Grant Distribution and BLER
Serving Grant Distribution Mostly Used
Selected L1 Throughput
60 Serving UL BLER
52.5
E-TFCI [Kbps]
Grant
50
23 43 14.65% 566.4
40
PDF [%]

25 63 13.29% 882.9
30

26 73 11.11% 1052.4
20
11.5 11.1 11.4
10 7.9 27 85 16.56% 1252.1
0.4 1.3 0.4 0.9 1.8 0.8
0 28 97 16.16% 1448.4
12 18 20 21 22 23 24 25 26 27 28
Serving Grant Average L1 Throughput with these SG 1040 Kbps

EUL Scheduler assigned highest Serving Grant (28) more than 50% of the time, L1 throughput = 1.4Mbps
− Rest of the time, scheduler assigned SG ≤ 27, slightly impacting the average L1 throughput.
The average HARQ NACK rate (UL BLER) >10% (impacts user perceived Throughput – refer next slide)
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 66
UL Throughput Impact in 10ms TTI
Factors Impacting UL MAC Layer Throughput i.e. end-user Perceived Throughput

UL T HRO UG HPUT For 10ms TTI, HARQ retransmissions in Ericsson is

1200 controlled by a parameter: transmissionTargetError%
1176.6

1150
Parameter Current Setting
UL Throughput [Kbps]

Average transmissionTargetError 100 (10%)

1100
15% BLER
1050
Per E///, the actual HARQ NACK rate is generally higher
than the set value of transmissionTargetError. Table below
1000
1000.1 shows the measured approximate HARQ NACK rate for
multiple users performing simultaneous uploads for
different settings of the parameter transmissionTargetError
950

900 transmissionTargetError % ~HARQ NACK rate %

L1 Throughput MAC Throughput
1 1.3
10 15.7
Higher target number of HARQ transmissions has a
tradeoff between cell throughput and end-user throughput 20 27.4

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 67
UL Uu Load Estimator
eulMaxOwnUuLoad & eulMaxRotCoverage Definitions
The UL Uu load estimator function provides the EUL scheduler with an estimate of the available
air interface load.
− The purpose of the UL Uu load estimator is to give the EUL scheduler a running update of the interference headroom
that is available to distribute among the EUL users both in CELL_DCH and CELL_FACH (the latter if Enhanced
Uplink for FACH is active).
Two types of headroom are estimated as follows:
− “Own cell interference” estimate defined by “eulMaxOwnUuLoad”, primarily to prevent power rushes
− “Coverage” estimate defined by “eulMaxRotCoverage”, primarily to adjust the EUL cell coverage
EUL scheduler assign UL grants based on E-DCH headroom
− E-DCH headroom = Minimum(“eulMaxOwnUuLoad” – measured OwnUuload, “eulMaxRoTCoverage” – measured
RoT)

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 68
Theoretical Aspects of EUL
Uplink Cell Throughput vs. Rise Over Thermal Simulations

The normalized throughput is the uplink

load in the system. Higher RoT target
provides ~90% of the total expected UL
Cell Throughput

The UL cell throughput is normalized with respect to the

maximum throughput at the pole capacity
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 69
UL Uu Load Estimator
eulMaxOwnUuLoad & eulMaxRotCoverage Definitions
Current Airtel settings from the site database

Maximum EUL UL Cell

eulMaxOwnUuLoad eulMaxRotCoverage % Cells
Capacity = 1 – 10-Noise Rise/10
100 (10dB) 100 (10dB) 90% - 90% 78%
80 (8dB) 100 (10dB) 84% - 90% 15%
30 (3dB) 100 (10dB) 50% - 90% 7%

100% of the cells in ELGRNC07 & ELGRNC08 have “eulMaxRotCoverage” set to 10dB targeting
90% UL load. Ideal settings to control the balance between UL cell capacity and coverage.
− Low value – low capacity but large coverage, high value – high capacity but less coverage.

In 7% cells with eulMaxOwnUuLoad = 3dB, the EUL capacity in these cells will be limited to 50%
− It is recommended to set eulMaxOwnLoad = 8dB across all these cells to enhance EUL cell capacity, unless these
cells have very high AMR R99 traffic.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 70
Impact of EUL on AMR Service
Example Slide

Initial Preamble Tx Power

1

0.9
~ 3 dB Increase in Initial
0.8 Preamble Power when
AMR Only
One EUL user present
0.7 AMR + 1 EUL
AMR + 2 EUL
0.6
CDF

0.5

0.4
~ 10 dB Increase in
0.3 Initial Preamble Power
need when Two EUL
0.2
user present
0.1

0
-20 -15 -10 -5 0 5 10 15 20
Power [dBm]

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 71
Summary & Recommendations
HSUPA with 10ms TTI
Current network configuration limits the 10ms TTI peak UL physical layer
throughput to 1.45Mbps
− It is recommended to set maxChannelizationCodes = 2xSF2 to increase peak UL L1
throughput to 2Mbps

MAC Layer throughput is slightly impacted by high HARQ NACK rate (~15%)
− It is recommended to optimize HARQ NACK rate to 10% to enhance end user UL throughput
− Tradeoff between cell throughput & end-user throughput

7% cells in the 2 RNCs have eulMaxOwnUuLoad = 3dB

− EUL capacity in these cells is limited to 50%
− It is recommended to set this parameter to 8dB to increase EUL cell capacity.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 72
2ms TTI Near-cell Stationary
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 73
Stationary EUL Throughput Test Results
Outdoor, Near Cell – UL Throughput Statistics

KPI EUL Cat. 6

E-DCH ASET Size 1.0
Avrg Ec/No [dB] -4.4
Avrg RSCP [dBm] -54.6
Avrg UE Transmit Power [dBm] -37.3
TTI Usage [%] 2ms TTI = 100%
Average Serving Grant 22.6
Average Physical Layer UL Throughput [Kbps] 4195
Average MAC Layer Throughput [Kbps] 3458
Average UL Scheduling Rate [%] 100%
Average UL SBLER [%] 17.5%
Average UL Residual BLER [%] 0%
Average Happiness reported by UE [%] 1.8%

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 74
Stationary EUL Throughput Test Results
Stationary EUL Throughput Test Results
Parameter EulMaxAllowedSchRate limits the
L1 throughput to 4480Kbps
Average of 40% less
3456
Uu_HSUPA_Throughput_MAC Serving Grant limits the L1 throughput down
to 4195 Kbps

MAC Re-transmissions limit throughput further

Average of down to 3458 Kbps
4195
Uu_HSUPA_Throughput_L1

High UL BLER Rate (17.5%) results in large

number of retransmissions and is much higher
than the target BLER for Tti2ms set to 5%
Max CAT 6 HSUPA Throughput
5742
L1 Airtel should contact Ericsson, if, after
implementation of proposed changes the
measured UL BLER will not converge with
0 2000 4000 6000 8000 target number of 5%

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 75
Stationary EUL Throughput. Test Time-Line
Strong RF Environment – Large Fluctuations in MAC Throughput

Parameter EulMaxAllowedSchRate limits the L1

throughput to 4480Kbps

Serving Grant limits the L1 throughput down to

4195 Kbps

MAC Re-transmissions limit throughput further

down to 3458 Kbps

After the removing MAC layer overhead, the

observed RLC layer throughput was 3206 Kbps, or
44% from MaxPossibleSchRate

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 76
UL Throughput Impact in 2ms TTI
Investigations
High level of re-transmissions for ETFCI =
119 leads to SBLER of 20%, in spite of
parameter transmissionTargetErrorTti2 is
set to 5%

To increase the reliability of EUL data

channel, parameter sirMaxTti2, which is set
currently to 80 (8dB) might be increased to
100-120 (10 dB-12dB)

SIR Target of 8 dB seems low and could be

recommended only for high loaded NW, with
large number of 2ms TTI active users

In current NW settings, max number of

active 2ms users is set to 2, which makes
Total number of HARQ re-transmissions due to E-HICH NACKs % 17.16
the sirMaxTti2 of 120 (12dB) more
Total number of HARQ re-transmissions due to E-HICH DTXs % 0 appropriate

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 77
UL Throughput Impact in 2ms TTI
Factors Impacting UL MAC Layer Throughput i.e. end-user Perceived Throughput

UL T HRO UG HPUT For 2ms TTI, HARQ retransmissions in Ericsson is

4500 controlled by a parameter: transmissionTargetErrorTti2%
4000
Average 17.5% BLER Parameter Current Setting
3500
transmissionTargetErrorTti2 50 (5%)
3000

2500 Per E///, the actual HARQ NACK rate is generally higher than
2000 4195 the set value of transmissionTargetErrorTti2. Table below
1500
3458 shows the measured approximate HARQ NACK rate for
multiple users performing simultaneous uploads for different
1000 settings of the parameter transmissionTargetErrorTti2
500

0 transmissionTargetErrorTti2 % ~HARQ NACK rate %

L1 Throughput MAC Throughput
2 5.8
Field logs reveal HARQ NACK rate much higher than the
5 10.5
expected HARQ NACK rate. Review E-HICH power
control settings – refer next slide 10 24.1
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 78
Potential Case of HICH Misdetection
Potential case of HICH misdetection where UE
decodes NACK on HICH channel, but EUL scheduler
doesn’t decrease the grant on AGCH channel.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 79
UL Throughput Impact in 2ms TTI
E-HICH Power Control
Feature State:
RNC DlPowerControlEul eHichMinCodePower maxUserEHichERgchPowerDl maxUserEHichPowerDlTti2
(Activated)

Current / Recommended Current Setting: -183 (-18.3dB) Current Setting: -123 (-12.3dB)
ELGRN07 75% Sites
Setting: -220 (-22dB) Recommended Setting: -140 (-14dB) Recommended Setting: -80 (-8dB)

Current / Recommended Current Setting: -183 (-18.3dB) Current Setting: -123 (-12.3dB)
ELGRN08 72% Sites
Setting: -220 (-22dB) Recommended Setting: -140 (-14dB) Recommended Setting: -80 (-8dB)

It is recommended to activate FAJ 121 1345 (DL Power Control for EUL) feature across all Node-Bs.
When the feature is activated, the E-HICH power is regulated between the values of the maxUserEhichErgchPowerDl and
eHichMinCodePower parameters for 10ms TTI and between the values of the maxUserEhichPowerDlTti2 and
eHichMinCodePower parameters + 6 dB for 2ms TTI
Increasing value of maxUserEHichERgchPowerDl & maxUserEHichPowerDlTti2 parameter should improve the robustness
of E-HICH channel and prevent ACK/NACK mis-detection, decreasing unwanted retransmissions and improve end-user
UL throughput experience.
Note: If DL power Control for EUL feature is disabled, network uses fixed E-HICH power settings.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 80
UL Throughput Impact in 2ms TTI
Other Recommendations
Parameter Current Setting Recommendations GRAKE is a receiver technique that is
used to suppress the self-interference,
1 (Enabled)  95% of the sites Enable feature across all allowing a higher SNR and higher bit
featureStateEul2msTti
2 (Disabled)  5% of the sites cells rates

≥ 5398 for all 2ms TTI

3968  22% of the sites enabled sites with The GRAKE is applied to EUL users with
eulMaxAllowedSchRate 2ms TTI for the spreading factor
4480  78% of the sites interference suppression
feature active combinations 2xSF2 + 2xSF4,
approximately corresponding to bit rates
1 (Enabled)  95% of the sites ON for all 2ms TTI enabled above 1.28 Mbps (on the RLC layer)
featureStateGrake
2 (Disabled)  5% of the sites sites

IF GRAKE feature is ON, and, feature

1 (Enabled)  76% of the sites StateInterferenceSuppression is ON then
featureStateInterferenceSup ON for all 2ms TTI enabled
0 (Disabled)  5% of the sites
pression
- (Not set)  19% of the sites
sites set eulMaxAllowedSchRate to 5398 Kbps

transmissionTargetErrorTti2 50 (5%) 50 (5%) Increasing value of sirMaxTti2 parameter

to 10dB -12dB should improve the EUL
sirMaxTti2 80 (8dB) 100-120 (10dB -12dB)
data channel

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 81
Summary & Recommendations
HSUPA with 2ms TTI
Current network configuration limits the 2ms TTI peak UL physical layer
throughput to 4.2Mbps and in certain sites it is even less, 3.5Mbps.
− At 78% sites, eulMaxAllowedSchRate = 4480, limiting highest SG = 31 (E-TFCI = 119 & TBS
= 8596)
− At 22% sites, eulMaxAllowedSchRate = 3968, limiting highest SG = 30 (E-TFCI = 114 & TBS
= 7173)
− It is recommended to set eulMaxAllowedSchRate ≥ 5398 with GRAKE & interference
suppression feature active to enhance the 2ms TTI peak UL data rate.
Though transmissionTargetTti2 is set to 5%, the HARQ NACK rate is in the
range of 17%.
− It is recommended to optimize E-HICH power control setting to improve the robustness of
HICH channel and avoid potential HICH misdetections.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 82
Mobility Test
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 83
 HSUPA Mobility Test
HSUPA Throughput Observations
MAC Layer
Good correlation between Throughput
low throughput, bad RF
and low scheduling rate
observed

HSUPA_Throughput
Statistic
MAC (Kbps)
Mean 862.0
Median 942.6
Maximum 4298
Minimum 0

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 84
Mobility Throughput Test Results
EUL MAC Throughput was at 15% from CAT 6 Device Maximum Throughput
Factors impacted the HSUPA Throughput
85% less
 Low SG
Average EUL MAC
862
Layer Throughput [Kbps]  Frequent Serving Cell Changes

 Lower 2ms TTI usage under supported

Average Physical Layer sites
1032
EUL Throughput [Kbps]
 High UL BLER Rate (~16%) results in
large number of retransmissions and is
much higher than the target BLER for
Max EUL CAT 6 Tti2ms (5%)
5742
Throughput
Airtel should contact Ericsson, if, after
implementation of proposed changes the
0 2000 4000 6000 8000
measured UL BLER will not converge to
5% target
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 85
 Airtel HSUPA Mobility Test –General Observations
Snapshot of the RF conditions during the Mobility Test
CPICH RSCP distribution CPICH EcN0 distribution
RF Average
100.00% 7.00%
90.00% 100.00% 12.00%
UARFCN RSCP EcN0 RSSI Tx Power RSCP PDF 6.00%
80.00% 10.00%
80.00%
All -80.5 -9.6 -71.3 0.0 70.00% 5.00% EcN0 PDF
EcN0 CDF 8.00%
10712 -77.6 -10.7 -67.3 -2.5 60.00% 4.00% 60.00%
6.00%
50.00%
10737 -82.5 -8.8 -74.1 1.8 40.00% 3.00% 40.00%
4.00%
30.00% 2.00% 20.00% 2.00%
Tx power distribution 20.00%
1.00% 0.00% 0.00%
100.00% 8.00% 10.00%
-30 -27 -24 -21 -18 -15 -12 -9 -6 -3 0
Tx Power PDF 0.00% 0.00%
80.00% 6.00%

-96
-90
-84
-78
-72
-66
-60
-54
-48
-42
-120
-114
-108
-102
Tx Power CDF
60.00% SGI distribution
4.00% 100.0% 40.0%
40.00% UARFCN distribution SGI PDF
20.00% 2.00% 10712 10737 80.0%
30.0%
0.00% 0.00% 60.0%
-60 -54 -48 -42 -36 -30 -24 -18 -12 -6 0 6 12 18 24 20.0%
40.0%
41%
59% 10.0%
20.0%

0.0% 0.0%

1.1
4.1
7.1
9.9

22.1

27.8
13
16
19

25
-9.5
-5.5
-1.9
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 86
Airtel HSUPA Mobility Test –General Observations (cont)
Non-optimized event 1d parameters lead to frequent CC ping-pongs

Zoomed ping-
pong area MAC Throughput

Serving Cell Changes:

12 cell changes for 60
sec

MAC TBS Retransmissions

Excessive HSPA Serving Cell Changes (SCC) impact
RNC Signaling Load and reduce the HSPA throughput

The observed frequent SCC ping-pongs are due to non-

optimized event 1d parameters
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 87
Mobility Route Essential KPIs
2ms and
KPI
10ms TTI
E-DCH ASET Size 1.5
Avrg Ec/No [dB] -9.1
10ms vs 2ms: 83% of
Avrg RSCP [dBm] -80.5 10ms TTI

Avrg UE Transmit Power [dBm] 0

10ms TTI = 83%
TTI Usage [%]
2ms TTI = 17%
Average Serving Grant 24.1
Average Physical Layer UL Throughput [Kbps] 1032
Average MAC Layer Throughput [Kbps] 862
Average UL Scheduling Rate [%] 84.0%
Average UL SBLER [%] 16.4%
Average UL Residual BLER [%] 0.3%
Average Happiness reported by UE [%] 22.4%

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 88
Serving Grant Utilization
Investigations

Serving Grant Distribution Throughput MAC

40% 37.2%
35% Mostly used SG=20 2357
2516
corresponds to MAC
30% 2012
throughput of 1412 kbps
PDF

25%
1412
1306
20%
1068
931
15% 774
565
8.6% 384 463
10%
6.3% 6.3% 6.0% 194 258 237
5.2%
5%
0

11
12
13
14
15
16
17
18
19
20
21

23
1.1
2.1
2.9
4.1
5.1

7.1
8.1
8.9
9.9

22.1
10 11 12 13 14 15 16 17 18 19 20 21 22.1 23
Serving Grant Serving Grant

No SG allocation above 23

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 89
UL Throughput Impact Breakdown
From Mobility Route Observation

UL Throughput Impacted By
Lower 2ms TTI
Low SG Frequent Serving Cell
usage under High BLER
Parameters and features Changes
supported sites Implementation
review Parameters review
Parameters review

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 90
Throughput Impact in Mobility
Mobility Test
ETFCI Limitations
0 Throughput Throughput Throughput is
7% limited by limited by limited by
13% UL Power in Buffer in Serving Grant
7% ~13% in 80%

Power Buffer Grant

80%

Uu_HSUPA_ETFCI_LtdMP_Rate Uu_HSUPA_ETFCI_LtdSG_Rate
Uu_HSUPA_ETFCI_LtdBO_Rate Uu_HSUPA_ETFCI_LtdHARQ_Rate

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 91
Serving Grant Allocation
Features to Improve Uu Overload, UL Interference and CE  All reflect on SG
Feature Name Feature ID Description Status
Enabled on 95% of sites  other
increases the achievable bit rates for EUL in a multi-path radio
featureStateGrake FAJ 121 1333 dependency parameters need to be
environment, compared to a standard RAKE receiver
reviewed
This Interference suppression feature replaces the existing G- Enabled on 95% of all sites  other
featureStateInterferenceSuppression FAJ 121 1714 Rake receiver and improves equalization, SIR estimation and dependency parameters need to be
interference mitigation capabilities reviewed
This feature is an advanced receiver enhancement that brings
Disabled on all sites  other
significant gains for EUL which in turn lowers interference
featureStateIntSuppEul10ms FAJ 121 2671 dependency parameters need to be
levels also for other radio bearers. Beneficial in U900 and
reviewed
GSM 900 deployment scenarios
Disabled on all sites  other
this feature avoids EUL starvation by releasing non
featureStateDchEulBalancing FAJ 121 1584 dependency parameters need to be
guaranteed UL DCH connection (UL PS R99)
reviewed
This feature provides a fast and accurate UL interference
featureStateUlFcc FAJ 121 2444 control mechanism for loaded conditions. Results in better Disabled on all sites
uplink throughput
This feature increases CE’s available for EUL scheduling by Enabled on 57% of the sites  other
featureStateCeExtForEul FAJ 121 2598 50% resulting in better throughput and significant decrease of dependency parameters need to be
number of EUL grant reductions reviewed
This feature improves CE capacity for EUL Smartphones with Enabled on 66% of the sites  other
featureStateCeCapEul FAJ 121 1883 2ms TTI, with lower interference levels (improved system dependency parameters need to be
capacity) and coverage for services such as voice reviewed

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 92
Uplink Throughput in Mobility Conditions
License Review Needed

Parameter Current Settings (# of users) Recommendations Comments

15, 16 (22% of sites), 31, 32 (78% of

eulServingCellUsersAdm Review Licensing Model
sites)
4 Enhanced Uplink users per cell  FAJ 121 0450
16 Enhanced Uplink users per cell  FAJ 121 0451
2 and 4 (98% on cells with 2ms TTI 32 Enhanced Uplink users per cell  FAJ 121 0452
eulServingCellUsersAdmTti2
enabled)
Review Licensing Model 64 Enhanced Uplink users per cell  FAJ 121 1676
96 Enhanced Uplink users per cell  FAJ 121 1677

maxNumEulUsers 32 (99% of sites), 4,16 Review Licensing Model

Admit as many as non-serving users to improve UL

Macro diversity and to avoid situation of mapping SRB
eulNonServingCellUsersAdm 100 (99% of sites), 50 100
on DCH when 2ms TTI is enabled causing reduced UL
throughput

 For better HSUPA performance, in the future it is recommended to increase # of 2ms TTI users
which would improve UL Throughput

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 93
Web Browsing Experience
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 94
SERVICE AVAILABITY
Web Browsing using popular sites ‘Facebook’, ‘google’

Typical Web browsing session includes

the following stages:
• Channel setup
• TCP transmission
• Content rendering by browser
Performance of each of the stages
affect end-user experience

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 95
DNS Server Responses Statistics
Large delays in DNS Responses Observed

Large delays in DNS responses observed,

while browsing popular websites ‘Facebook’
and ‘google’.
In these cases the DNS server was a home
server, when min delay in response time is
expected. The DNS response scenario 1 is
illustrated on the next slide

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 96
DNS Response Scenario I
1st Scenario of DNS response – domain information is in home DNS server
DNS response time for this scenario should be less than 50 ms
Observed IP address resolution for popular website facebook.com was 1.1 sec!

DNS Response Scenario One

UE RNC DNS Response 1.103 s
Gi DNS Movistar
DNS Request T1 If domain information is in
home DNS server the
UE IP: 10.159.29.140 T2 response time should be
DNS Response T3
DNS IP: 192.168.175.145 less than 50 ms. Delay in
DNS responses introduce a
negative experience for the
Total Time= T1+T2+T3 users browsing the WEB

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 97
DNS Response Scenario II
2nd Scenario of DNS response – domain information is in home DNS server
If domain information is not in home DNS server, home DNS server requests the data
from external (root) DNS server. In this case the response time will be significantly
higher. During the DNS tests conducted by ESG, the some amount of popular IP
addresses (yahoo, google) were resolved by external DNS server, rather than Airtel
home server. This could be explained by expiration of the cache memory in home DNS
servers.
The setup of the servers should be checked to minimize number of external
requests

UE RNC DNS (Movistar)

Airtel External DNS (Telefonica)
DNS Request T1 DNS Internal Table Resolution and Cashe verification
T2 DNS Intractive Request
T3
DBS Intractive Response
T4
DNS Intractive Request
T5
DBS Intractive Response

T6 Cache Storage in process

DNS Response T7

Total Time= T1+T2+T3+T4+T5+T6+T7

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 98
PDP Context Activation Delay
PDP Context Activation plus DNS server response time are more than 2 sec in
average
PDP Context Activation Delay
Day/Time Lat Long PDP Context Activation Delay is high, negatively
(ms)
impacting user experience
Jun/30/15 07:24:07 3.46184 6.45478 1556
Jul/01/15 07:36:53 3.43013 6.45502 1847
Jun/30/15 06:37:55 3.40809 6.4441 1806
Jul/03/15 05:56:59 1767
average 1744

DNS resolution latency was in a range of ~250ms and above, reference values are much lower
− DNS server CPU load and Memory usage should be checked

Average PDP context activation delay for APN “internet.ng.Zain.com” was 1744 ms, which is
higher than reference networks
− network topology and elements load have to be revised

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 99
Evolution of UTRAN Features - Motivation
CPC and Enhanced Cell FACH

+ HS-DCH
- Inefficient for + No dedicated - HS resources
can be used bursty traffic resources
not usable
required

High data UE Battery Power Low data

rates consumption efficient rates

Short delays Ideal for low State

data volume switching

• Improve efficiency in DCH • Make HS resources available

state through introduction of in FACH state
DTX / DRX • Reduce the need for state
switching

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 100
Idle Mode Performance Assessment
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 101
Introduction
Overview
For Idle performance assessment, the main Intra-
parameters for Idle activities will be audited, Freq.
which are:
− Cell Reselection for Mobility Inter-
− Intra Frequency CCCH
Freq.
− Inter Frequency
− Inter Rat
− Paging parameters for Reachability Idle
− Rach Parameters for Accessibility
− Common Control Channels (CCCH) for RACH
Inter-
proper performance RAT

Parameters for RNC ELGRN08 will be audited

and recommendations could be applied to Paging
other RNCs.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 102
W2W Idle Camping Parameters
3G Cell Re-selection
Below are the main Parameters for cell
reselection.
Parameter
Description
Name
FDD-quality-measure RSCP or Ec/No , used for ranking
• Squal ≤ Sintrasearch (Measure Intra-frequency neighbor cells)
Measure • Squal ≤ Sintersearch (Measure Inter-Frequency neighbor cells)
Sintersearch The Ec/No threshold below which the inter-frequency cell search starts
Neighbor
Sintrasearch The Ec/No threshold below which the intra-frequency cell search starts Cells?
Ssearch,RAT The Ec/No threshold below which the inter-RAT cell search starts
• Used for Cell Selection during Acquisition state
Ssearch,HCS The RSCP threshold below which the inter-frequency cell search starts • Suitable if Srxlev>0 & Squal>0
(valid for R99 with HCS, and from Rel5 onward even if HCS is not
used) • Srxlev = Qrxlevmeas – Qrxlevmin – max (UE_TX_PWR_MAX_RACH – P_MAX, 0)
Suitability • Squal = Qqualmeas – Qqualmin
SHCS,RAT The RSCP threshold below which the inter-RAT cell search starts
(valid for Rel5 onward, even if HCS is not used)

Qhyst1 The hysteresis used for ranking the serving cell (RSCP)

Qhyst2 The hysteresis used for ranking the serving cell (Ec/No) • Serving Cell Rank (Rs) = Qmeas,s + Qhyst,s
Treselection The cell ranking evaluation period before the UE changes serving cell. • Neighbor Cell Rank (Rn) = Qmeas,n - Qoffset,n
(A scaling factor for Inter-frequency reselection is allowed from Rel5
Ranking
onward)

Qoffsets1,s,n The offset used for ranking the non-serving cells (RSCP)

Qoffsets2,s,n The offset used for ranking the non-serving cells (Ec/No)
• When Rn > Rs for a time duration of Treselection
SIB11 Neighbor List The potential candidates specified by the network for cell reselection

Qrxlevmin Minimum CPICH RSCP for the cell to be suitable Cell Re-
selection
Qqualmin Minimum CPICH Ec/No for the cell to be suitable

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 103
W2W Idle Camping Parameters
qQualmin & qRxLevmin
Different values are seen among the cells: qQualmin and qRxLevmin
− 2 values for qQualmin (-17, -18) dB 800 730

− 7 values for qRxLevmin. (from -91 dBm to -115dBm) 700

600
− Most common combination is (-18, -115), which is the 500
nominal recommendation. 400
300
200
For a considerable number of cells, qRxLevmin 100
66 25 21
24 2 29 4
6 6 3 5
is set to very high values (-91 dBm, -103 dBm 0
-115 -113 -111 -109 -107 -103 -91
and -107 dBm), this will have a severe impact
qQualmin = -18 qQualmin = -17
on cell coverage, specially for indoor users.
Even if these cells are overshooting, decreasing qRxLevmin, will not decrease the DL interference or pilot
pollution, it will only restrict the UEs to camp on these cells while the DL power will be still interfering on the
surrounding cells  in case these cells are overshooting recommendation is first tilt the sectors and second
– less preferred – decrease CPICH power.

Parameter Current QC Recom.

qQualmin -17, -18 dB -18 dB
qRxLevmin -91, -103, -107, -109, -111, -113, -115 dBm -115 dBm
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 104
W2W Idle Camping Parameters
Sintra-Search & Sinter-Search Sintra-Search
700
613
Looking to the settings from the RNC Dump, different 600
configurations are identified. 500
400
Initiating WCDMA cell measurements is decided

Cells
300
based on Squal 184
− Squal = Qqualmeas – Qqualmin 200 116
− If Squal > Sintrasearch, Sintersearch  no measurement is initiated. 100
8
− If Squal ≤ Sintrasearch  Measure Intra-frequency neighbor cells. 0
− If Squal ≤ Sintersearch  Measure Inter-Frequency neighbor cells. 0 - Always Search 8 dB 12 dB 14 dB
dB
Setting Sintra/inter-Search to “0”, means it will not be
broadcasted in SIB 3, and per standard, when the
parameter is not broadcasted the UE will ALWAYS Sinter-Search
550
search for intra/inter-frequency neighbors, which will 500 463
exhaust UE battery quickly. 450
400
342
350
With qQualmin set to -18dB in most cases, with 300
Sintra-Search set to 12 means the UE is allowed to 250
200
start intra-frequency neighbors search once its EcN0 150 116
is equal or below = -6dB and from EcN0 distribution of 100
50
the drive test, only 3.3% of the samples showed EcN0 0
>= -6 dB, meaning the UE will be searching for 0- Always Search 8 dB 10 dB

neighbors most of the time!

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 105
W2W Idle Camping Parameters
Sintra-Search & Sinter-Search Neighbor cell
Intra-frequency neighbors
Serving cell

Signal Quality
Also, different set of combinations were seen:
− Sintra = Sinter [Both set to either “0” or “8dB”]
− 0 is not advised as it will drain the UE battery
− Both equal 8dB not advised as it will lead to many inter-freq reselection ping
Time
pongs.
Inter-frequency vertical neighbors

Signal Quality
− Sinter = “0”, while Sintra = any value other than “0”
− UE will start searching for Inter-frequency neighbors before Intra-freq, which Neighbor cell

will be pushing traffic out to inter-freq neighbors specially in good RF!

Serving cell

− Sintra = “0”, while Sinter = any value other than “0” Time

− Sintra = 0 will drain the UE battery Different RF characteristics – Example

• Intra-Freq: Ec/No & RSCP vary with
location/geometry
Recommendation is to have 2dB difference between Sintra and • Inter-Freq (co-site intra-band): at a given location,
Sinter, in order to minimize unnecessary interfreq reselections RSCP is similar while Ec/No can differ.
and achieve Idle load balancing at the same time.
Parameter Current QC Recom.
Sintra 0, 8, 12, 14 dB 10 dB
Sinter 0, 8, 10 dB 8 dB
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 106
W2W Idle Camping Parameters
Qoffset 2SN and Inter-freq Neighboring strategy
For 2 frequencies sites, different configurations are
applied, with no set strategy, the following is a way
to try to summarize the majority of the cases but still
many cells don’t follow these rules.
Load sharing is activated only from F1 to F2, but not
for all F1 cells with co-located F2.
− F1:
− Only Co-site F2 cells are defined [exceptions with few
extra F2 cells defined].
− If F1  F2 load sharing is active, then Qoffset 2Sn = 50 
Never Select F2. [exceptions with Qoffset 2Sn = 0 exist]
− If not, Qoffset 2Sn = 0 [exceptions with Qoffset 2Sn = 50/-
50 exist]
− F2:
− Co-site F1 + Surrounding Cells on F1 are defined as neighbors
− Qoffset 2Sn = -50  Always Select F1 or Qoffset 2Sn = 0  random camping

A complete audit for the mobility strategy and neighboring definition needs to be carried on accompanied with
counter analysis. Multi-carriers Strategy Assessment could be proposed.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 107
W2G Idle Camping Parameters
S-IRAT
UE measures GSM cells if
− Squal ≤ Qqualmin – SsearchRAT and TmeasureGSM expires
− From Rel5 onward, ShcsRAT specifies the limit for Srxlev in the serving cell
below which the UE starts measuring inter-RAT neighboring cells
− Or Measurement rules are skipped if suitability fails for Nserv consecutive
DRX cycles

When Treselection > 0, multiple GSM measurements are required to trigger

a W2G reselection
TmeasureGSM controls the frequency of GSM measurements
− For DRX cycle = 1.28 sec  TmeasureGSM = 6.4 sec
− For DRX cycle = 0.64 sec  TmeasureGSM = 5.12 sec

UE only measures and reselects to GSM cells provided in the SIB11 of

the serving WCDMA cell * Recommended Settings

Parameter Current QC Recommendation

SRatSearch 2, 4 dB 2 dB
ShcsRAT 3 dB 2-3 dB
Qoffset1sn (GSM neighbor relation) 7 dB 7dB
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 108
Idle Camping Parameters
G2W Reselection
WCDMA measurement trigger condition: FDD_Qmin – Cells Dist. FDD_RSCPMin - Cells Dist.
− RLA_C of serving GSM cell is below 2%
Qsearch_I

G2W reselection is triggered if all the 10 dB

28%
-104 dBm
conditions below are satisfied for 5 seconds: 42% 14 dB
-102 dBm
− CPICH Ec/No ≥ FDD_Qmin 56%
12 dB

− CPICH RSCP > BCCH RSSI + FDD_Qoffset 72%

− CPICH RSCP > FDD_RSCP_Threshold
− where FDD_RSCP_Threshold =
FDD_RSCPmin – min(P_MAX – 21 dBm, 3 dB)

Parameter Current (from EOGBS02 Dump) QC rec. Comment

QSearch_I 7 , (14 cells set to 15  Never select 3G) 7 (always) Default UE value is 15 (never) until receiving SI 2quater.
With Qqualmin = -18 dB and SsearchRAT= 2dB, UE would likely
FDD_Qmin -10, -12,-14 dB -14 dB trigger GSM reselection at CPICH Ec/No < -16 dB and reselect
back to WCDMA at -14 dB, this will facilitate 3G reselection.
0 , (28 cells set to 8  0 dB difference between Setting this value to 0 would imply that always select a cell if the
FDD_Qoffset 0 (-infinite)
GSM RSSI and CPICH RSCP to reselect 3G) quality is acceptable to maximize time on 3G.

Extending the range to -104 could increase the probability of

FDD_RSCP_min -102, -104 dBm -104 dBm returning to 3G.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 109
Paging
DRX Cycles
The UE listens to the PICH only at certain predefined
times, reducing power consumption. The periodicity of
these searches is set by the system and the time
interval is called Discontinuous Reception (DRX)
cycle. Different DRX cycles are used for circuit
switched and packet switched services in Idle Mode.

DRX cycle length = 2^k×10 (ms)

− Where:
− k Integer defined by the cnDrxCycleLengthCs parameter for
circuit-switched services and by the cnDrxCycleLengthPs for
packet-switched services. For paging of Connected Mode UEs
in URA_PCH state, k is an integer defined by the parameter
utranDrxCycleLength
Parameter Current QC Recomm.
The number of times the WCDMA RAN transmits the
paging information to a UE is determined by the cnDrxCycleLengthCs 6 7
configurable parameter noOfPagingRecordTransm.
cnDrxCycleLengthPs 7 7
Increasing DRX cycle could extend UE battery life, but
noOfPagingRecordTransm 2 2
at the same time could increase Call Setup time
slightly since it might take longer for the paging to utranDrxCycleLength 5 6
reach the Mobile terminating.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 110
Paging
PICH and PCH
The PICH is used to indicate to the UE when it Paging ch. Power
has to read the S-CCPCH. PCH is used to carry 700
575
the RRC message "Paging type 1", which 600
500
contains the actual paging information. 400

Cells
300
Decreasing these channels power might impact 200
182

the paging success rate in the network, while 100 21 13

65
14
51

assigning higher power would waste DL Power. 0

-4 -6 -8 -10 -12 -14 -16

dB
sccpchOffset is the PCH timing offset inside
the radio frame and it is recommended to be set Pich Power Offset from CPICH
to 20, while many cells in the network have this 800
offset set to 0, which may affect the ongoing 700
676

traffic. 600
500
400

Cells
Parameter QC Recommendation 300
200 149
pichPower -7 dB 100 29 47
20

pchPower -4  -0.4 dB 0
-5 -6 -7 -8 -9
dB
sccpchOffset 20
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 111
RACH Parameters
Preamble Initial Power
Initial PRACH transmit power depends on the open loop estimate:

Preamble_Initial_Power = Primary CPICH Tx power – CPICH_RSCP + UL_Interference +

Constant Value

Where:
− [Primary CPICH Tx power – CPICH_RSCP]
− Downlink path loss from the NodeB to the UE
− Primary CPICH Tx power is sent in SIB5
− UL Interference
− Interference seen on the UL. Higher the interference, the more transmit power needed to overcome it.
− Sent in SIB7
− Constant Value
− Can account for any difference between Uplink and Downlink propagation and/or transmit-receive chain
− Sent in SIB 5

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 112
RACH Parameters
SIB 5 RACH Parameters
Preamble Retransmission Max
− Setting Tradeoff: If set large, the uplink interference increases especially in areas requiring more RACH
re-transmissions (edge of coverage)

Mmax
− Setting Tradeoff: If set large, the uplink capacity consumption may unnecessarily increase

Constant Value
− Setting Tradeoff: If set too large, the preamble initial transmit power may be larger than needed for
reliable detection at the Node B. This causes unnecessary uplink interference

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 113
RACH Parameters
Power Offset P p-m
Power Offset: Power offset of control part of RACH message relative to last transmitted preamble
P p-m

RACH Message RACH Message

Preamble control part data part

The Preamble is used to synchronize the physical layers of the UE (Tx) and the NodeB (Rx).
Physical layer synchronization means that the NodeB has detected the location (time and
frequency) of the UE’s signal and demodulated it

After synchronization, the NodeB sends an acknowledgment on the AICH channel and the UE
then transmits the message part of the RACH message. This can be the RRC Connection Request
and is decoded by the RRC layer

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 114
RACH Preamble Count and Power
RACH Stats
UE RACH TxPreamble Count Stats extracted from short
70.00% 100.00%
90.00%
CS Over FACH test done
60.00%
50.00%
80.00%
70.00%
in the same cluster.
40.00% 60.00%
50.00%
30.00% 40.00% For Preamble Count:
−
20.00% 30.00%
20.00% 62.4% of the samples
10.00%
0.00%
10.00%
0.00%
required only 1 preambles
1 2 3 4 5 6 7 8 until an acknowledgement is
PDF CDF received  considered high
and could indicate that the
RACH Last Preamble Tx Power UE is using high power for
70.00% 100.00% its initial preamble power.
90.00%
60.00%
80.00%
50.00%
40.00%
70.00%
60.00%
For last Preamble power:
30.00%
50.00%
40.00%
− Around 10.5% of the
20.00% 30.00% samples showed last
20.00%
10.00%
10.00% preamble power higher or
0.00%
< -25.0 -25.0 to 0.0 0.0 to 10.0 10.0 to 15.0 15.0 to 20.0 > 20.0
0.00% equal to 15 dBm.
PDF CDF

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 115
RACH Performance
RACH Parameters
Too many different configurations are seen
in field for RACH, the plot aside is showing
7 different values for the constant value
only!

In general RACH performance is

correlated with HSUPA performance and
UL Capacity in general, since it uses
OLPC, suboptimal settings could lead to
high UL interference impacting the HSUPA
performance.

Comparing the areas with high constant

values with the UE Tx Power, a correlation
could be deduced, though other
parameters like step size should also be
considered.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 116
RACH Simulation – Current Settings
Cell Edge Conditions

Considering average busy hour UL interference of -97dBm, as observed in the test logs and under cell edge conditions
(e.g. Ec/No = -14dB and RSCP = -104 dBm)
− Preamble Initial Power starts at 21dBm and flats out at 24dBm after 1st RACH retransmissions, and for the next 7 retransmissions.
This may increase the UL interference.
Lowering Constant Value and Power Ramp Step may reduce UL interference
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 117
RACH Simulation – Proposed Settings
Cell Edge Conditions

Keeping UL interference value of -98dBm, at the edge of coverage (RSCP = -104dBm, Ec/No = -14 dB):
− Recommended RACH parameters yield to Initial RACH power of 15dBm, and then flats out to max UE Tx Power after 5th RACH re-
transmissions

The recommended parameters should reduce the overall UL interference, however monitoring is required to determine
any negative impact. Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 118
RACH Parameters
Parameters Review
Parameter Current Qualcomm Recom. Comments
POWERRAMPSTEP 1 - 3 dB 2 dB If set too low might slow CST and if set too high might increase UL interference

PREAMBLERETRANSMAX 8 - 32 6 to 10 If parameter is too large, the uplink interference increases

If parameter is too large, the uplink capacity consumption may unnecessarily
Mmax 4 - 32 3 to 6 increase
The shortest period of time after which a UE can reinitiate a random access
NB01min 0 0 - 10 ms attempt incase a negative AICH Ack. is received
This IE shall be set larger or equal to NB01min – too large value might delay
NB01max 0 10 ms retry in case of RACH collision.
11 11
Gain Factor ßc (PRACH)
Gain Factor ßd (PRACH)
15 15
PowerOffsetPpm If parameter is too small, decoding of the RACH message part may be
-5  2 dB -2 to 0 dB unreliable
With large value, the preamble initial transmit power may be larger than needed
CONSTANTVALUE -18  -27 -27 to -24 dB for reliable detection at the Node-B, causing unnecessary uplink interference.
Time between RRC Connection request repetitions, setting it too long increases
T300 2000 ms 1200 ms call setup time.
Number of allowed repetitions for RRC connection request, increasing it could
N300 5 5 increase the probability of reception and hence the overall RRC Setup Success
rate.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 119
RACH
AICH Channel
To access the network the UE is using the
RACH cycles increasing its preamble power in
steps in order to be able to reach the NodeB.
If a preamble reaches the NodeB, and was to
decoded properly, the NodeB will use the AICH
channel to replay to the UE.
The AICH carries AI (Acquisition Indicator -
response to PRACH preambles). Its power, the
aichPower parameter, is expressed as an offset
relative to the PCPICH power. Aich Power
If the parameter is set to low, it might be missed 800
676
700
by the UE which will continue its RACH cycles 600
and then goes to Idle and consider this a failed 500

Cells
400
Access attempts, if set to high it might consume 300
DL power unnecessarily. 200 119
100 12 17 20 40 37
Parameter Current QC Recomm. 0
-3 -4 -5 -6 -7 -8 -9

aichPower -3  -9 dB -6 dB dB

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 120
Trial Proposal
FAJ 121 1350 - Improved RACH Coverage

The purpose of the feature is to increase the coverage of RACH so that it more
closely matches the coverage of the Cell_DCH state.

This feature is a potential enabler of site reduction, and it improves the

accessibility and reduces the response time.

The features increases the detection sensitivity for the RACH preamble in the
RBS, If the license is in place and the increasedRachCoverageEnabled
parameter equals "TRUE", the sensitivity of the RBS is increased as the
preamble threshold will be lowered by 3 dB relative the current value.

With the increased sensitivity of NodeB, RACH Constant Value could be set to
lower values decreasing UL interference.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 121
Common Control Channels
CPICH Power
Different CPICH values were seen within the
drive test route.
From RNC dump, different values ranging from
28 dBm to 36 dBm were seen with 33 and 30
dBm being the most widely used values in the
network.
It is recommended to have a homogeneous
setting of CPICH power along the network to CPICH Ratio to Cell Power
minimize link imbalance. 450
400
394

350
Generally it is recommended to have the 300
250
CPICH power at 8 – 10 dB from the total cell 200
193

150
power, distribution is showing most cells have 100 56

CPICH ratio less than 8% of the total cell power 50 8 1 1 1 1

33
9 3 1 1 1
23
10 11 6 9 16
2 2
21
2 4 1
22 15
6 7 6 7
32
1 7 8
0
highlighting an optimization room to increase

10.00%
10.71%
11.76%
13.33%
1.67%
1.72%
1.82%
2.08%
2.17%
2.50%
2.63%
2.78%
2.86%
3.03%
3.13%
3.33%
3.45%
3.57%
3.85%
4.00%
4.17%
4.35%
4.55%
4.76%
5.00%
5.26%
5.56%
5.88%
6.06%
6.67%
6.90%
7.14%
7.69%
8.00%
8.33%
8.70%
3G coverage, maybe specially for indoor.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 122
Common Control Channels
CPICH Power Sub-Optimal Settings
Distribution of Cells with CPICH Power as % of
It is important to keep CPICH power in 7% - 10% range from cell MaxTxPower
MaxTxPower. If allocated CPICH power accounts for more than 10% of 50% 42%

Percentile of Cells
MaxTxPower, the DL Tx Power might become a limiting factor in loaded 40%
network.
30% 24%
If allocated CPICH power accounts for less than 7% from MaxTxPower, the 19%
CPICH Ec/No might become too low in loaded conditions, and could drop 20%

below Qqualmin at Cell Edge. 10% 3% 4% 2% 3%

0% 1% 1%
One should remember that power of the Common Channels (BCH, PCH, 0%
FACH, etc) is set as fixed offset from CPICH power. Selecting CPICH power 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
lower than 7% might affect other CCH operations, reducing accessibility, CPICH Power as % of MaxTxPower
retainability and limiting the HSPA throughput.
In tested cluster, 50% of the cells have CPICH power set to 5%
or less CPICH Power
59%
Ec/No Percentile of Cells with
Ec/No Cell Percentile
CPICH % NearCell Ec/No < Qqualmin at Cell

CDF
Edge Loaded of Cells
Loaded Edge

PDF
28%
10% -10 -15 3%
7% -12 -17 42%
5% -13 -18 24%
4% -14 -19 4%
50% 250 260 270 280 290 300 310 320 330 340 350 360
3% -15 -20 19%
2% -18 -23 3% CPICH Power [dBm]
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 123
Common Control Channels
Synchronization Channels Power Settings
The SCH consists of a Primary SCH (P-SCH) Deviating cells - Primary Sch Power
and a Secondary SCH (S-SCH), used in the cell 12 11

search procedure in the UE. Their power is set 10

independently and they are both expressed as 8 7 7

Cells
offsets relative to the PCPICH power. 6 5 5
4 4 4 4 4
4
2 2 2 2 2
2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

A total of 84 cells SCH Power - Cells Dist. 0

-19
-20
-21
-22
-23
-25
-26
-27
-28
-29
-30
-31
-32
-33
-35
-36
-37
-38
-39
-40
-42
-43
-44
-45
-47
-48
-50
-51
-53
-55
-60
-63
-67
-69
were found with
0.1 dB
deviating settings 9%

from the default /

recommended Recommended Deviating Cells - Secondary Sch Power
Deviating 12
settings for SCH 11

10
channels power 91%
8 7 7
offset.

Cells
6
6 5 5
4 4 4 4 4
4
Parameter QC Recommendation 2 2 2 2 2 2
2 1 1 1 1 1 1 1 1 1 1 1

primarySchPower -18  -1.8 dB 0

-36 -37 -38 -39 -40 -42 -43 -44 -45 -46 -47 -48 -49 -50 -52 -53 -54 -55 -56 -57 -59 -60 -61 -64 -65 -67 -68 -70
secondarySchPower -35  -3.5 dB 0.1 dB

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 124
Common Control Channels
BCH Power Settings
The PCCPCH carries the Broadcast Channel Deviating Cells - bch Power
(BCH), which is broadcast time multiplexed with 120 112

the SCH. The last 2,304 chips of every slot 100

transmit the PCCPCH. The bchPower parameter 80

Cells
determines its power and is expressed as an 60 49

offset relative to PCPICH power. 40

25
20 6 7
12 11
2 2 1 2 3 2 2 1 3 3 1 2 3 3 3 2 3 1 3
A total of 264 cells BCH Power - Cell Dist. 0
-30 -32 -33 -34 -35 -36 -37 -38 -39 -40 -41 -44 -45 -46 -47 -48 -49 -50 -51 -53 -54 -55 -56 -57 -58 -60
were found with
deviating settings 0.1 dB
29%
from the default / Recommended
recommended Deviating

settings for SCH 71%

channels power
offset.

Parameter QC Recom.
bchPower -31  -3.1dB
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 125
Common Control Channels
FACH
FACH channel plays a major role in accessibility as Deviating Cells - maxFach2Power
RRC connection Setup message is sent using FACH 120
to establish the initial SRB 100
102

− maxFach1Power: If the FACH frame carries a logical control

80
channel, the power is set through the maxFach1Power parameter.

Cells
− maxFach2Power: If the FACH frame carries a logical traffic
60 46

channel, the power is set through the maxFach2Power parameter. 40

21 21
16 16
20 8 6 9 6 9
4 4 2 5 4 4 3 2 5 3 3 3 3 4 5 3 5 3
While the default for maxFach1Power - Cell Dist.
1
0
maxFach1Power is 18, 0% -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 16 17 18 19 20 21 22 23 24 25

539 cells have it set to 0.1 dB

50 and 1 cell to 28.
41% 18 --> 1.8 dB Increasing maxFach1Power could have a positive impact
50 --> 5 dB RRC success rate, but is it’s set too high it might consume
For maxFach2Power a 59% 28 --> 2.8 dB DL power unnecessarily, a median value of 3.8 dB is
wide range of settings is proposed
seen.
− Monitoring required for both types of setting to assess gains
and losses
Parameter QC Recommendation
For maxFach2Power, setting the parameter too low could
maxFach1Power 38  3.8 dB impact data performance on FACH at cell edge increasing
the BLER on FACH, while setting it too high could consume
maxFach2Power 15  1.5 dB DL power unnecessarily.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 126
Other Observations
LAC Borders
Based on the provided cell data
base, the LAC borders needs to
be optimized in many areas in
the network.

Un-optimized borders will

increase the signaling load on
the border cells impacting cell
capacity and performance.
Increase IuR signaling which
might impact the overall
Retainability due to any system
limitation. (example call re-
establishment Over IuR)
Also ping pong between the LACs will cause a lot of missed pagings and impact users reachability

Many other areas from the cell data based were showing un-optimized borders, Airtel team to check and
confirm the configuration and redesign borders accordingly, rehoming of some sites might be required.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 127
Other Observations
RRC Connection Reject
The following case is to
highlight the consequences
of having a bad LAC
borders design.

In the highlighted area we

have many ping pongs
between LACs (156,178
and 302) as shown.

The UE being in FACH, was listening to DL CCCH, and at this same area it received 72 RRC Connection
Reject with cause Congestion directed to other users !

As stated before, this is the expected output with all the extra signaling that will be generated by all users
passing these border to perform LAC/RAC:
− If the UE was sending RRC for LAC/RAC  Rejection will delay the LAC Update which will lead to the UE missing its paging
− If the UE was sending RRC for service request  Rejection will impact user perception and might be considered as call setup
failure.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 128
Other Observations
RRC Connection Reject
Whenever RRC connection is rejected, the RCN will send to the UE a waitTime IE indicating the
time to wait in seconds for the UE before it retries the RRC connection, and optionally it can
redirect the UE to another cell either inter-frequency or IRAT.
If the timer is set too high, user perception will be
impacted delays in registration will occur and call
failures might happen (ex: Mobile terminating a
call and has to wait 15 seconds before retrying !).
If the timer is set too low it might lead to a snow
ball effect in case of congestion leading to more
connections being rejected.
2 different values were seen from the logs, 1 sec and 15 sec, both are exaggerated and require
fine tuning

Parameter Current Value QC Recomm.

Wait Time 1 - 15 2-4
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 129
Summary of issues Power settings for Primary and
Secondary synchronization channels
Synchronize might impact performance in cell Edge/
Poor RF
Read
BCCH power offsets are not set correctly BCCH
for some sites, which could impact SIB Cell reselection Parameters, are
decoding in Cell Edge / Poor RF exhaustive foe UE battery and in some
Cell Reselection
cases restrict the UE from camping on
the best available cell.
Receive
Paging
PICH and PCH power offset are not set
correctly to many sites impacting the
reachability of the users, and DRX cycles RACH
RACH Parameters are not
could be extended to enhance battery life. homogeneous in the network with many
templates available, most of them could
L.A.
Update induce high UL interference.

LAC Borders need to be redesigned in

Idle
order to improve signaling load, Operations
accessibility and mobility for UEs

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 130
AMR Performance Assessment
CS on FACH
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 131
Multi-RAB Call Definitions
AMR RAB + PS RAB Multiple PS RABs

UE NW UE NW

AMR Call established First PS RAB established

Cell_DCH Cell_DCH

PS Call established Second PS RAB established

Multi-RAB in this project is refers to “AMR RAB+PS RAB” call scenario

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 132
Scenarios of Multi-RAB
AMR RAB First then PS PS RAB first then AMR
RAB added added
UE NW UE NW

Signaling at call Setup Signaling at call Setup

(SRB) (SRB)

Cell_DCH PS Call established

AMR Call established Cell_DCH / FACH / PCH

PS Call established AMR Call established

SRB is common between the multiple RABs

AMR can be established when PS Bearer is active from DCH/FACH/PCH states

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 133
PS RAB Rates in Multi-RAB
PS RAB re-configured to UL of PS RAB re- PS RAB remains on HSPA
R99 during Multi-RAB configured to R99 during during Multi-RAB
Multi-RAB
UE NW UE NW UE NW
Signaling at call Setup Signaling at call Setup Signaling at call Setup
(SRB) (SRB) (SRB)

Cell_DCH Cell_DCH Cell_DCH

PS Call established HSPA PS Call established HSPA PS Call established HSPA

AMR Call established AMR Call established AMR Call established

PS Call re-configured R99 UL/DL PS Call re-configured R99 UL PS Call re-configured

Choice of PS RAB on R99 or HSPA is part of the optimization process

and vendor support  No restrictions in 3GPP
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 134
mRAB Implementation

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 135
CS Over FACH
Test Summary
Aside is the drive test route in Lagos.

Calls 147 calls done

− 135 starting from FACH.
− Average CST 2.35 Seconds
− 12 calls starting from DCH.
− Average CST 4.22 Seconds

1 call setup failure discussed in case

study#1.

Call setup time is showing shorter values for calls starting from FACH compared to the ones
starting from DCH which is an expected outcome.
− UE is Cell_FACH uses SRB 13.6 kbps for setting up the call with the CS core.
− UE in Cell_DCH uses SRB 3.4 kbps (4 time slower than 13.6 kbps) for setting up the call with CS core.
− State transition from FACH to DCH while setting up the call will increase call setup time  Discussed in
details in case Study #3.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 136
 mRAB Combination in Airtel
Observed mRAB Call Flow Currently, there is no fixed mRAB rate configured in Airtel

UE NW UeRC mRAB Combinations (CS 12.2 + One Interactive PS UL/DL)

UeRc=10 Conv. CS speech 12.2 + Interact. PS (64/64)

UE in Cell_FACH
UeRc=113 Conv. CS Speech (12.2, 7.95, 5.9, 4.75) + Interact. PS (16/HS)
CS Call (12.2Kbps) added
UeRc=19 Conv. CS speech 12.2 + Interact. PS (64/HS)

Frequent Switching Radio Bearer Reconfig UeRc=20 Conv. CS speech 12.2 + Interact. PS (384/HS)
of PS rates using
the configured rate
in the table in this UeRc=36 Conv. CS speech 12.2 + Interact. PS (64/128)
side RB Reconfig Complete

UeRc=37 Conv. CS speech 12.2 + Interact. PS (128/64)

Multiple UeRC for mRAB leads to two issues UeRc=38 Conv. CS speech 12.2 + Interact. PS (64/384)

High Signaling Load Link Budget Issue UeRc=60 Conv. CS Speech 12.2 + Interact. PS (128/HS)
Signaling procedures specific to Uplink or downlink coverage difference
combined services can lead to higher between the CS service in combined
drop rate services and a single CS service UeRc=9 Conv. CS speech 12.2 + Interact. PS (0/0)

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 137
mRAB Combination in Airtel
mRAB Combinations in Drive Route
The test done focused on call starting from
FACH which usually accounts for low
throughput requirements and low data in the
buffer, that’s why most of the cases the
mRAB combination used was with PS 0/0 or
HSDPA/16 kbps, and a small percentage for
HSDPA/64 kbps.
In case the voice call were starting from DCH
state or in case of high data volume in the UE
buffer, higher rates on Uplink since their
UeRc are already enabled in the network.
CS 12.2 Kbps + SRB 3.4 Kbps + the following PS rates
Higher R99 rates on the UL would have
caused power limitations from UE sides and UL PS Data Rate DL PS Data Rate Usage
could cause higher call drops cases, details R99 64 Kbps HSDPA 3.5 %
in next slides
R99 16 Kbps HSDPA 24.9 %
A complete study for the Multi-RAB
R99 0 Kbps R99 0 Kbps 71.6 %
performance is advised.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 138
 Impact of Current Configuration
Impact on Link Budget
Example of Link Budget impact with
different rates on UL Net effect of 4 to 10 dB link budget difference
LB Component AMR12.2+16/HS AMR12.2+64/HS AMR12.2+EUL/HS expected for MRAB compared to voice calls on UL
SF Impact 3dB 6dB 0dB
Ahs impact 1.5-2.5dB 1.5-2.5dB 1.5-2.5dB
(which is typically considered as the limiting link in
MPR impact 1-1.5dB 1-1.5dB 2.5dB terms of coverage).
TOTAL 5.5-7dB 8.5-10dB 4-5dB
The result would be increased drop call rate for
MRAB compared to CS only calls.
Service Type Uplink SF Downlink SF
SRB3.4 Kbps+AMR12.2 Kbps+PS DCH
Call Drop rate may significantly increase
SF64 SF128
0 Kbps especially with a usage of UeRc= 20, 37, 60
SRB3.4 Kbps+AMR12.2 Kbps+PS DCH
SF32 SF64
16 Kbps Using higher rate on UL in mRAB calls will add
SRB3.4 Kbps+AMR12.2 Kbps+PS DCH
64 Kbps
SF16 SF32 disadvantage to UL power and link budget due to
SRB3.4 Kbps+AMR12.2 Kbps+PS DCH higher Spreading Factor.
SF8 SF16
128 Kbps
SRB3.4 Kbps+AMR12.2 Kbps+PS DCH
SF4 SF8  Once PS call drops, the CS call would also drop during
384 Kbps
mRAB
SRB3.4 Kbps+AMR12.2 Kbps+PS The SF depends on the
SF128
HSUPA/HSDPA rate of HSUPA services
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 139
Coverage Difference between mRAB and Single CS RAB
Coverage Impact with mRAB
The choice for PS transport
channel in a mRAB call will

UL Coverage Impact with mRAB

have its impact on the If the PS BE services are carried If the PS BE services are carried
on (DCH) in the uplink: on the DCH in the downlink:
combined service coverage. The higher the PS BE data rate, the
Coverage lower to CS only due to Tx
Power required for PS more the code resources are
For UL, the combined service consumed by combined services and
the less spreading gains, hence

DL Coverage Impact with mRAB

coverage will be impacted in all If the PS BE services are carried coverage is lower than CS only.
on the (E-DCH) in the uplink:
cases, even more with the
Enhanced Dedicated Physical Control If the PS BE services are carried
utilization of HSDPA on the DL. Channel (E-DPCCH) consumes on the HS-DSCH in the downlink:
additional UE transmit power.
1- When the UE is not receiving PS
For DL, utilization of HSDPA BE data in the downlink, combined
services have the same downlink
If the PS BE services are carried
where the corresponding on the (HS-DSCH) in the downlink: coverage as a single CS service
physical channel uses the The uplink physical channel High
Speed Dedicated Physical Control 2-When the UE is receiving PS BE
shared SF16, which is Channel (HS-DPCCH) that carries data in the downlink there would be
orthogonal to SF128 used by the channel quality indicator (CQI) some impact on EcNo with HSDPA if
consumes additional UE transmit download is in progress.
AMR voice service seems to power
provide similar coverage as CS
only, specially when UE is not
receiving any data on DL.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 140
Recommended mRAB Configuration

Recommended Option #1 Recommended Option #2

mRAB with UL R99 & DL HSDPA mRAB with UL HSUPA & DL HSDPA
CS 12.2 + PS UL: 16 kbps ; DL: HSDPA CS 12.2 + PS UL: HSUPA; DL: HSDPA

Recommended UL/DL Rate Recommended UL/DL Rate

Use Fixed Rate of 16 Kbps on UL Use HSUPA 10ms on UL
Use HSDPA on DL Use HSDPA on DL
Always enable CS+Int. PS (0/0) Rate Always enable CS+PS (0/0) Rate

Other Recommendations Other Recommendations

Disable PS rate switching function based Keep enabled the UeRc Rate that
on throughput, to reduce the amount of supports UL 16 Kbps to be able to support
signaling during mRAB devices without HSUPA capability

Monitoring DCR is recommended for the strategy chosen to be implemented and further
fine tuning can be done based on KPIs
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 141
Expectations with Recommended Implementations
Coverage Difference Between Combined Services and a Single CS Service
Type of Combined Services
Uplink Downlink
(UL/DL)
Low Rate High Rate Low Rate High Rate
CS+PS 0k/0k Negligible difference Negligible difference

Coverage of a single CS service Coverage of a mRAB service is worse due to less

CS+PS (DCH > 0k) / (DCH > 0k) Negligible difference
is better spreading gains

Coverage of a single CS service is better because of additional power

CS+PS DCH/HSDPA Negligible difference with optimized HSDPA
need of CQI on HS-DPCCH  requires better CQI feedback cycle

Coverage of a mRAB service is worse because of additional power

CS+PS HSUPA/HSDPA Negligible difference with optimized HSDPA
needed for E-DPCCH & E-DPDCH & HS-DPCCH

The targets of Qualcomm recommendations are to use CS + a mix of UL/DL that provide good coverage compared to CS
only calls and reduce Call Drop Rate and improves PS throughput for better end-user experience

− Using DCH on UL with low rate (i.e. 16Kbps) is expected to provide negligible impact on UL coverage compared to CS only
− Using HSDPA on DL is expected to provide negligible impact on DL coverage compared to CS only, after optimizing HSDPA.
− Using 0/0 rate when there is no active data is expected to provide negligible impact on UL/DL coverage compared to CS only
− HSUPA can be a future option but after ensuring proper optimizations are done for HSUPA/HSDPA performance.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 142
Other Implementation Considerations
Inactivity Time Feature (FAJ 121 2134)
In general, the CELL_DCH to CELL_DCH Downswitch/Upswitch evaluation algorithm determines whether a
switch to a lower rate channel is made
− The same algorithm applies both to Single RABs and Multi-RABs
− It is important to optimize Downswitch/Upswitch  if the current configuration of UeRC remains the same
− If UeRC for Multi-RAB is chosen to be restricted to two rates (i.e. a transition between CS+DCH/HS & CS+0/0 as
discussed before)  Therefore, the next step below can be done

When Inactivity Time Feature (FAJ 121 2134) is activated, downswitchTimerSp is used for Multi-RAB
combinations of Speech and PS interactive on HSDPA, instead of hsdschInactivityTimer
− Currently, UtranCell::inactivityTimer (cell level) is disabled, hence the Multi-RAB down-switching will follow
hsdschInactivityTimer which is makes it harder to distinguish channel switching for PS-only calls from Multi-RAB
− It is recommended to enable UtranCell::inactivityTimer
− With Inactivity Timers on Cell Level enabled, downswitchTimerSp can be used after which the Multi-RAB of
Speech and single PS Interactive on HSDPA is downswitched to "Speech + Packet 0/0 kbps“ which would allow
to separate PS-only from Multi-RAB implementation
− Then setting of downswitchTimerSp can be optimized to control when to perform the switching to SP0 rate

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 143
Background on “Inactivity Timers on Cell Level”

Inactivity Timers on Cell Level (FAJ 121 2134) makes it possible to define inactivity timers on cell
level for RAB combinations with n×PsInteractive RABs
− When the feature is activated, inactivity timers are assigned values defined in parameters of the best
cell of the UE connection when the applicable channel switching algorithms are initialized
− Inactivity Timers on Cell Level feature enables to tune channel switching behavior for parts of the
network and balance the channel element load in RAN against the user experience
− Besides the cell level function, the Speech + PsInt (x/HS) combination during the downswitch evaluation
uses downswitchTimerSp instead of hsdschInactivityTimer
− Better distinction between Single RAB and Multi-RAB becomes possible in this case
− This feature becomes important when the UeRC rate chosen are CS+DCH/HS (or CS+EUL/HS) &
CS+0/0 rate is chosen as Multi-RAB strategy

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 144
Counters to monitor after applying mRAB Rate Recommendations
pmNormalReleaseSource pmNoRabEstSuccessMultiSpeech
pmReconfAttSource pmNormalRelMultiAmrNb
pmReconfAttTarget pmNormalRelMultiAmrWb
pmReconfSuccSource pmNormalRelMultiPsInt
pmReconfSysRelSource pmNormalRelMultiPsIntToAmrNb
pmReconfSysRelTarget pmNormalRelMultiPsIntToAmrWb
pmSystemReleaseSource pmNormalRelMultiPsIntToSpeech
pmReconfSuccTarget pmNormalRelMultiSpeech
pmNoCellUpdateAttMultiCs pmNormalRelMultiSpeechToPsInt
pmNoCellUpdateSuccMultiCs pmNoSystemNasSignReleaseMultiCs
pmNoNormalNasSignReleaseMultiCs pmNoSystemNasSignReleaseMultiPs
pmNoNormalNasSignReleaseMultiPs pmSystemRelMultiAmrNb
pmNoRabEstAttemptMultiAmrNb pmSystemRelMultiAmrWb
pmNoRabEstAttemptMultiAmrWb pmSystemRelMultiPsInt
pmNoRabEstAttemptMultiPsInt pmSystemRelMultiPsIntToAmrNb
pmNoRabEstAttemptMultiSpeech pmSystemRelMultiPsIntToAmrWb
pmNoRabEstSuccessMultiAmrNb pmSystemRelMultiPsIntToSpeech
pmNoRabEstSuccessMultiAmrWb pmSystemRelMultiSpeech
pmNoRabEstSuccessMultiPsInt pmSystemRelMultiSpeechToPsInt

mRAB Counters to Monitor mRAB Counters to Monitor

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 145
Case Study #1 – CC Disconnect: No Circuit

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 146
Case Study #1: Congestion
Failure Scenario
the call is setup normally, and then directly
after CC Call Proceeding, the UE receives CC
Disconnect with cause “resource unavailable;
no circuit/channel available”
This failure cause is attributed to the following
scenario in most cases:
− after the UE sends the CM Service Request to the
core, the core sends RAB Assignment to the RNC to
allocate resources for the Radio bearer to be setup.

− in case the RNC replies with RAB assignment

failure/reject due to congestion, the Core sends the
CC disconnect to release the Cs call with cause No
circuit/channel available.

The UE releases the call and the core sends

SCR to terminate the signaling connection.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 147
Case Study #2 – Out of Service in FACH

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 148
Case Study #2 : Out of Service
Overview
In CELL_FACH state, the
fachMeasOccaCycLenCoeff and
interFreqFddMeasIndicator parameters
are used to control the UE measurement
activities for inter-frequency and inter-RAT
neighbors.
If fachMeasOccaCycLenCoeff is set to a
value > 0, the UE performs inter-frequency
and inter-RAT measurements during FACH
measurement occasions.
During the FACH measurement occasions,
the RNC does not schedule any data for the
UE on FACH.

Few samples were seen in DT with

interFreqFddMeasIndicator set to False,
and from RNC Dump 114 cells have this
configuration.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 149
Reason for Out of Service in Cell_FACH
Draw Back
Many reasons could lead to the UE declaring Reason for
OOS in cell FACH which could categorized as OoS in
following: FACH

− Bad Coverage Problems: since the UE in Cell_FACH is

doing cell reselection, bins with quality below qQualmin Bad EcNo
Inter-Freq
and/or qRxlevmin will lead to the UE declaring OOS as and
neighbor
the cells will not be eligible for reselection. Acceptable
List
RSCP
− Inter-freq FACH MO is disabled: specially in hot spot
deployment scenarios (like Airtel), where the UE could be
restricted to go to a better carrier. Only collocated Inter-Freq cell is
Inter-Freq FACH
UE can’t find any available for UE measurements in
Measurement
eligible cell for SIB-11 neighbor list  does not
− Neighboring List: for the list the UE is restricted to camping.
Occasion is
allow UE to find another better
disabled
reselect better cell because they are not defined in its NL cell on the other carriers
(ex: Co-site cells only defined )

Parameter Current Value QC Recomm

interFreqFddMeasIndicator 0 (disabled) - 1(Enabled) 1 Enabled for all carriers

fachMeasOccaCycLenCoeff 4 4
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 150
 Scenario

service area.
the UE in FACH.
FACH were detected.

hence it declares OOS.

will present one of them in details.
Out of Service Cases

From declaring OOS to CU ~ 2.4 sec

Cell Update with cause : Re-entering
UE was in a Call with AMR + P0/0, the

UE then starts full band cell selection, and

Network sent Radio bearer release putting

The UE tries to camp on a cell but he EcN0

The 2 cases have almost similar flow so we

degrades quickly below qQualmin level, and

finds and eligible cell, camps on it and sends

Within the drive test 2 occurrences of OOS in
-8
-6
-4

-28
-26
-24
-22
-20
-18
-16
-14
-12
-10

Time
Jul/01/15 13:18:04
Jul/01/15 13:18:04
Jul/01/15 13:18:05
Jul/01/15 13:18:05
Jul/01/15 13:18:06
Jul/01/15 13:18:06
Jul/01/15 13:18:07
Jul/01/15 13:18:07
Jul/01/15 13:18:08
Jul/01/15 13:18:08
Jul/01/15 13:18:09
Jul/01/15 13:18:09
Jul/01/15 13:18:10
Jul/01/15 13:18:10
Jul/01/15 13:18:11
Jul/01/15 13:18:11
Jul/01/15 13:18:12
Jul/01/15 13:18:12
Jul/01/15 13:18:13
Jul/01/15 13:18:13
Jul/01/15 13:18:14
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION

Jul/01/15 13:18:14
Jul/01/15 13:18:15
Jul/01/15 13:18:15
Jul/01/15 13:18:16
Jul/01/15 13:18:16
Jul/01/15 13:18:17
Jul/01/15 13:18:17
Jul/01/15 13:18:18
Jul/01/15 13:18:18
Uu_ActiveSet_EcNo_0

Jul/01/15 13:18:19
Jul/01/15 13:18:19
Jul/01/15 13:18:20
Jul/01/15 13:18:20
Jul/01/15 13:18:21
Jul/01/15 13:18:21
80-W4325-1 Rev A

Jul/01/15 13:18:22
Jul/01/15 13:18:22
Jul/01/15 13:18:23
Jul/01/15 13:18:23
Jul/01/15 13:18:24
UE Declares OOS

Jul/01/15 13:18:24
Jul/01/15 13:18:25
Jul/01/15 13:18:25
Jul/01/15 13:18:26
Jul/01/15 13:18:26
Jul/01/15 13:18:27
151
Case Study #3 – Call Setup time

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 152
Case Study #3 : Call Setup Time
Call Setup time Distribution
Call Setup from FACH Time
134 successful calls made to 60.00% 100.00%
90.00%
call server starting from 50.00%
80.00%
70.00%
Cell_FACH. 40.00%
Analysis 60.00%
30.00% 50.00%
40.00%
Average call setup time 2.35 20.00%
30.00%
20.00%
seconds 10.00%
10.00%

− minimum 1.81 Sec

0.00% 0.00%
1.6 2 2.4 2.8 3.2 3.6 4 4.4 4.8 5.2 5.6

− maximum 5.27 Sec

Seconds
PDF CDF

For all the calls with extended call setup time, state transition was detected in
the middle of call setup with 2 main scenarios identified detailed in next slides.

Optimizing State transition thresholds and timers could minimize the occurrence
of such cases  will be discussed in the following section.
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 153
Case Study #3 : Call Setup Time
Extended Call Setup Time

State transition occurs before mRAB Bearer Setup  SRB used State transition occurs after mRAB bearer setup with PS 0/0 (UE
starting this point will be 3.4 Kbps increasing delays. was not receiving data in FACH), Radio bearer reconfiguration
Another Radio bearer Setup and radio bearer reconfiguration added to up switch PS Bearer of the UE from PS 0/0 to
process are added to the normal call flow ! HSDPA/32 Kbps

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 154
State Transition
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 155
 State transition based on Traffic Volume Measurements
Channel Switching Parameters
From CELL_FACH to DCH From DCH/HS to CELL_FACH
Parameter Current QC Recommended Parameter Current QC Recommended

ulRlcBufUpswitch 256 bytes 256 bytes downswitchThreshold 8 kbps 8 kbps

dlRlcBufUpswitch 500 bytes 300 bytes (Resolution 100) downswitchTimerThreshold 32 kbps 16 kbps

upswitchTimerUl 500 ms 100 - 300 ms downswitchTimer 10  1 sec 5 sec

upswitchTimer 500 ms 100 - 300 ms hsdschInactivityTimer 2 sec 2 - 5 sec

From CELL_FACH to URA_PCH

Parameter Current QC Recommended
Increase DCH
resources
inactivityTimer 2 sec 5 sec

inactivityTimerPch 10 min 30 min Reduce

FACH
Setting Improve
Tradeoff Latencies
Fast Dormancy Congestion

Parameter Current QC Recommended

Reduce
Signaling
T323 120 sec 120 sec
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 156
mRAB Channel Switching
Multi-RAB Upswitch
The Multi-RAB Upswitch evaluation algorithm is specific to the Speech + Interactive Multi-RAB. It
monitors the RLC buffer load on both the UL and the DL of the interactive part of the Speech +
0/0 state (SP0).
If the dlRlcBufUpswitchMrab parameter is set to 0, the Multi-RAB Upswitch evaluation
algorithm is turned off for the Downlink, meaning that no upswitches from SP0 occur, irrespective
of the RLC buffer load  Current setting at Airtel, recommended to increase the threshold in
order to enhance user perception.
For the Uplink the up-switching is controlled by ulRlcBufUpswitchMrab, currently it is set to 8
bytes which might trigger the up switch quickly and impact call setup time as seen in the previous
cases  recommended to increase this threshold.

Monitoring DCR for mRAB is recommended.

Parameter Current Proposed Trial
ulRlcBufUpswitchMrab 8 bytes 128 bytes
dlRlcBufUpswitchMrab 0 bytes 500 bytes
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 157
mRAB Channel Switching
Multi-RAB Downswitch
The Multi-RAB Downswitch evaluation algorithm monitors the user data throughput both in the
UL and the DL. If no data is transmitted for the duration specified by a downswitch timer, a
request is issued to switch the Interactive RAB to state Speech + 0/0 (SP0) in case of Speech +
Interactive DCH/DCH Multi-RAB or Speech + DCH/HSDPA Interactive Multi-RAB.

If the downswitchTimerSp parameter is set to 0, the corresponding Multi-RAB Downswitch

evaluation is turned off and no downswitches occur irrespective of how long the throughput has
been zero.

As a general rule, it is preferred to decrease the number of Radio Bearer Reconfigurations as

much as possible with a mRAB call, hence a long value should be assigned to this parameter,
but also speeding the downswitch to SP0 might help the call retainability, hence a trial is
proposed to set the timer to 5 sec and study the impact.

Parameter Current Proposed trial

downswitchTimerSp 60  30 Sec 10  5 Sec
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 158
Additional Observations
Based on PM Counter Data
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 159
General Observations
Low Accessibility Rate (Data from Counters Measurements)
RRC and RAB Setup Success Rate RRC/RAB Connection failures due to Admission Control
accounts for 41% of all RRC Connection Request
RABESTABLISHSUCCESRATE 98.30% Failures. The distribution of failures due to AC is shown
at the diagram
RRCCONNSETUPSUCCESSRATE 86.0%
The other 59% of RRC/RAB Connection Request
Failures happened after the AC.
75.0% 80.0% 85.0% 90.0% 95.0% 100.0%

DeniedAdmDlChnlCode
1%
1%

25% DeniedAdmDlHw

~49% of
RRC/RAB DeniedAdmDlPwr
41% Conn
pmNoReqDeniedAdm Request DeniedAdmUlHw
4% failures are 49%
pmNoFailedAfterAdm due to lack
59% pmNoRrcConnReqBlockTn
of DL Power

18%
pmNoFailedRrcConnectReqHw
2%
pmNoRejRrcConnRrcLc

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 160
Executive Summary
Understanding Accessibility Issue at Airtel Network (1 of 2)

1 If step 1 failed – RRC Conn Rejected due to AC

2 If step 2 failed – RRC Conn Rejected due to unsuccessful RL setup

If AC accept RRC Conn Request, RNC sends RRC Conn Setup

3 Message over the FACH

4 If no RRC Conn Setup Complete Message received

– RRC Connection Access Failure: counter pmNoFailedAfterAdm
has incremented
pmTotNoRrcConnectSetup 236802174
pmNoFailedAfterAdm 8906720
Precentile of RRCConnSetup Failure
3.8%
Possible reason: not sufficient FACH power
after Admission

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 161
Executive Summary
Understanding Accessibility Issue at Airtel Network (2 of 2)
Parameter maxFACH1Power adjustment is needed in order to improve CS and PS
Accessibility KPIs. With FACH control channel power increase, we expect less number
of failures in RRC Connection Setup messages, leading to better Accessibility KPI. That
way number of RRC failures after admission (detected by performance counter
pmNoFailedAfterAdm) will be significant reduced

Change maxFach1Power from 18 to

maxFACH1Power 38 (1.8 dB → 3.8 dB)
In 18% cells maxFach1Power has
been set to 50 (5.0 dB) relative
18% CPICH power which might deliver
even better accessibility results.
However, this setting is not
1.8 dB
recommended for high capacity
cells where downlink power may be
5.0 dB a limited resource

82% If FACH Power is too

low UE will not ‘hear’
the RRC Conn Setup
Message

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 162
Multi-Layers Timers Harmonization
PHY, RLC and RRC Layers parameters should be harmonized to prevent premature call drops due
to timers timeout (SRB specifically )

The RRC layer has specific timers for some

•Active Set
Updates, procedures, such as, RRC Connection
Layer 3 (RRC) •RB Reconfiguration
•RRC handshakes
Extra Time Establishment, Radio Bearer Reconfiguration,
•etc.. Active Set Update, etc. However, these timers
are not available in the regular RNC dumps.
•MaxDat
•TimerPoll Changing these values is done directly by
Layer 2 (RLC) •MaxRst
•TimerRst
Ericsson

•N313 In order to correctly harmonize the timers,

•T313
Layer 1 (PHY) •T314 Airtel and Ericsson need work together in the
•T315 optimization
•etc..

The common practice is to align the layers timeouts to 8-10 sec

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 163
Conclusion and Next Steps
Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 164
Conclusion & Next Steps

Airtel to identify test RNC and trial Qualcomm recommendations.

Harmonize baseline parameters for “Common channel power” as a first step
before implementing other changes.
Detailed Pre-Post analysis should be performed after every change to quantify
the impact.
Vendor consultation is recommended before performing the changes.

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 165
Thank you
Follow us on:
For more information on Qualcomm, visit us at:
www.qualcomm.com and www.qualcomm.com/blog

Confidential and Proprietary – Qualcomm Technologies, Inc. | MAY CONTAIN U.S. AND INTERNATIONAL EXPORT CONTROLLED INFORMATION 80-W4325-1 Rev A 166