Professional Documents
Culture Documents
SD
HSDPA Optimization Guideline for CELCOM
version 0.1
2/21
25/01/2008
CHANGE HISTORY
Version
0.1
Date
25.01.2008
Created by
Khoo Nee Sern, Endy
Comments
First draft version
3/21
Table of Contents
1.
4
Objective
2.
4
Cluster Tuning
2.1
4
Cluster Preparation
2.4
6
2.5
6
RF Optimization
3.
7
Parameter Optimization
3.1
7
HSDPA Power
3.4
9
4.
10
4.1
10
25/01/2008
4/21
4.1.1
10
4.1.2
11
4.1.3
11
4.1.4
11
25/01/2008
1. Objective
This document describes the optimization process for the HSDPA. The optimization process
is divided into the following categories that are separately explained:
Cluster tuning where the purpose is to identify any poor HSDPA performance which
is caused by RF issues or neighbor planning problem. This mainly focuses on RF
optimization, especially on the hardware changes, i.e. antenna tilt and orientation,
or changes in the neighbor lists.
Parameter optimization where the purpose is to optimize the HSDPA performance
via parameter tuning, e.g. HSDPA power, Iub parameter, etc.
KPI counters monitoring can be further used in daily optimization to identify any
worst performance cells and troubleshoot what is the root cause of performance
degradation.
2. Cluster Tuning
Figure below shows the cluster tuning process flowchart. Cluster tuning is mainly to identify
any RF issues or missing neighbour problem. If the problem is not related to RF issues, then
it can be further optimised from parameter tuning or KPI counter troubleshooting.
5/21
25/01/2008
2.1
Cluster Preparation
Before starting any drive test campaign, the cluster area should be well prepared to ensure
minimal disruption to the data collection and subsequent troubleshooting activities. The drive
test route should cover a good percentage of main roads, motorways and different clutter
types. HSDPA is not selected during SRNC relocation when UE coming from CELL_FACH, so
we need to avoid RNC borders for HSDPA mobility.
2.2
Drive test KPIs are calculated after the measurement collection campaign. Among the KPIs to
be measured are:
HSDPA Throughput
The other possible reasons that HSDPA throughput may be bad because the BTS is not
scheduling sufficient data every TTI due to lack of incoming data from the RNC/core network.
This is not a radio problem, and it should be solved if more than 1 parallel FTP session is run
at the same time.
6/21
25/01/2008
RTT
This refers to the number of successful HS-DSCH allocation and the number of drop calls
happen during the throughput measurement.
2.3
7/21
25/01/2008
Before drive test starts, we have to ensure that the sites in the cluster are free of any major
alarms. Alarm Check can be done with NetAct Reporting Suite Fault Management reports or
using standard NetAct GUI tools. The HSDPA related alarms are:
7772 PHYSICAL SHARED CHANNEL CONFIGURATION FAILED
This alarm indicates that there has been an error when RNC has requested
configuration or reconfiguration of HSDPA channels to WBTS.
7776 HSDPA FAILURE IN WCEL
Possible reason: HSDPA enabled in too many cells
2.4
It is important to make sure the parameter is consistent, the following parameter consistency
checks should be done regularly:
2.5
RF Optimization
HSDPA throughput depends directly on the radio channel conditions. These conditions are
changing rapidly all the time due to fast fading of the radio channel. BTS is able to change
the LA for each 2ms TTI based on the channel measurements. Average throughput in a
certain location can be estimated if the average SINR (signal to interference + noise ratio) is
known. SINR is dependent on the EcNo, RSCP and power allocation for HS-DSCH and total
WCDMA power, and orthogonality.
Benefits that can be gained with RF Optimisation to HSDPA performance:
Improving average SINR for HS-DSCH, i.e. improving average HSDPA cell
throughput
Well defined dominance areas reduces HS-DSCH to FACH transition for mobile users
if HSDPA mobility is disabled (RAS51)
Well defined dominance areas improves probability to get HS-DSCH at the call setup
Therefore, it is extremely important to perform RF optimisation in areas having poor
dominance, e.g. physical change of antenna tilt, (azimuth, type and height) based on scanner
data. In general, if the optimization is done already for R99, then there is no need for separate
RF optimization for HSDPA.
3. Parameter Optimization
Many parameters for HSDPA can be optimized to improve the performance. The parameter
optimization can be categorized into few subjects:
8/21
3.1
25/01/2008
HSDPA throughput
HSDPA power setting
Iub configuration and parameter setting
HSDPA selection and mobility
HSDPA FMCS/HOPS
HSDPA impact to R99 users
HSDPA priority
SHA/SHFCA parameter setting
HSDPA Power
Increasing HSDPA max power increases average HSDPA cell throughput. However, impact
on R99 users should always considered, when increasing HSDPA power. The following is the
effect of different settings of HSDPA power to the throughput.
9/21
HsdpaPriority =1
HsdpaPriority = 2
25/01/2008
10/21
25/01/2008
3.2
Iub Parameter
MaxBitRateNRTMACDFlow
This parameter defines the maximum bit rate of the MAC-d flow. For 16QAM, it should be set
as 3456kbps, and for QPSK it should be 1664kbps.
SHA
SharedHsdpaAllocation is ATM protocol level parameter (includes the AAL2 overhead and
ATM overhead). It defines the amount of Iub bandwidth on a VCC allocated for HSDPA traffic
only. The application throughput can be multiplied by ~1.30 to get the ATM level SHA value.
Table below shows an example of the effect of different SHA setting on 1E1 site. As the SHA
setting is higher, the achievable HSDPA throughput is higher as the AAL2 resources are
reserved in the Iub for HSDPA. Below shows some test results from Kepong test bed:
11/21
SHA (kbps)
0
0.3
0.5
Table 2 SHA setting
1 HSDPA
1000
1000
1000
25/01/2008
Another impact of SHA setting, especially to Iub capacity-limited sites with HSDPA enabled, is
the increase in AAL2 CAC Rejection rate, which affects to the RRC and RAB establishment
success rate, soft handover success rate, and call drop rate. Care must be taken to tune the
SHA parameter to match the available E1 configuration of the site.
SHFCA
3.3
Release Margin Average Ec/No , Release Margin Peak Ec/No in HSDPA HOPS can be
changed to increase HSDPA selection probability in SHO areas. Tuning AdditionTime in
HSDPA FMCS and Enable RRC Connection Release, Release Margin Average Ec/No ,
Release Margin Peak Ec/No = 3.5dB in HSDPA HOPS can reduce the amount of HS-DSCH
to FACH transitions. This would lead better average mobile HSDPA throughput.
3.4
12/21
25/01/2008
Figure below shows an example of the degradation of EcNo if PtxMaxHsdpa is set to 8W. In
low RF area, degraded EcNo will lead to degraded R99 throughput performance.
4.1
It is important to ensure that there is sufficient capacity in the Node B hardware (CE), air
interface (codes, power), and the transmission (Iub) after HSDPA activation.
13/21
25/01/2008
Enabling HSDPA, with or without, active HSDPA users will reserve 32 CE if HSDPA is enabled
in BTS. If HSDPA is Enabled in 3 Cells it will reserve 3*32=96 CEs (RAS5.1).
In RAS5.1, there is dedicated measurement in BTS for the channel element usage.
M5001C0, MAX_AVAIL_CE
M5001C2, AVE_AVAIL_CE
M5001C3, MAX_USED_CE_DL
M5001C4, MAX_USED_CE_UL
M5001C7, AVG_AVAIL_CE_DL
M5001C8, AVG_AVAIL_CE_UL
We can verify the average utilised CE against the available CE to identify any possibility of
CE blocking.
M5001C7 AVG_USED_C E_DL
x100%
M5001C2 AVE_AVAIL_ CE
M5001C8 AVG_USED_C E_UL
RNC_731A Average Ratio of utilised CE for UL in BTS
x100%
M5001C2 AVE_AVAIL_ CE
CE_Uti_DL
250
45.00%
40.00%
35.00%
30.00%
25.00%
20.00%
15.00%
10.00%
5.00%
0.00%
200
150
100
50
Figure 7 DL CE Utilization
If there is problem due to CE capacity, it is worth to cross check RRC connection setup failure
or RAB setup failure due to BTS.
DL spreading codes
Enabled HSDPA will reserved together with Common Channels 45 codes (SF 128), so when
HSDPA is enabled in the cell, there are free codes left for 4 x 384 users:
Code tree occupancy:
RNC_113a
CODE_CAPAC ITY
DENOM_CODE_CAPACITY
100%
Code blocking can be calculated from counters which are triggered when there is no code
for SFx (x=4,8256) are available:
1/27/2008
1/26/2008
1/25/2008
1/24/2008
1/23/2008
1/22/2008
1/21/2008
1/20/2008
1/19/2008
1/18/2008
1/17/2008
1/16/2008
1/15/2008
1/14/2008
1/13/2008
1/12/2008
1/11/2008
1/10/2008
1/9/2008
1/8/2008
1/7/2008
1/6/2008
1/5/2008
1/4/2008
1/3/2008
1/2/2008
1/1/2008
CE Utilisation, %
MAX_AVAIL_CE
14/21
25/01/2008
256
100%
256
Code tree occupancy is heavily increased due to HSDPA activation, roughly this means 36%
(together with codes for CCCH) code tree occupancy without any load. So introduction of
HSDPA increase possibility of code blcoking.
WBTS Power
Iub Capacity
Iub reserved load and bandwidth can be monitored to verify the reservation of the Iub by R99
and specifically by HSDPA traffic.
15/21
25/01/2008
Iub reservation success rate can be monitored to check for any rejection on the AAL2 CAC
Reservation algorithm. HSDPA reservation success rate is calculated by below formula. If the
reservation fails then one of M800C7, M800C8 or M800C9 is incremented. M800C7 indicates
a lack of external resources whereas M800C8 indicates a lack of internal resources. M800C9
is incremented if the failure is for any other reason.
sum(AAL2_SUCCEEDED_HSPDA)
x100%
sum(AAL2_SUCCEEDED_HSPDA TRANSPORT_ REJECTED_EXT_HSDPA
TRANSPORT_ REJECTED_INT_HSDPA OTHER_REJECTED_HSDPA )
The number of HSDPA AAL2 reservation rejected because of too many users can be verified
from counter REJECTED_HSDPA_TOO_MANY_USERS. This counter is incremented in
failure cases where Shared HSDPA Allocation has failed and the number of MAC-d flows
(HSDPA users) is limited to the number given by parameter NbrOfOverbookedHSDPAUsers
AAL2 CAC rejection also useful to see what is the impact to R99 users for different SHA
setting. It can be further verified from Service Level counters whether it affect to the RRC or
RAB performance.
sum(AAL2_CAC_REJECTED)
x100%
sum(AAL2_RM_SUCCEEDED AAL2_CAC_REJECTED)
Below figure shows the improvement of AAL2 CAC rejection after fine tuning the SHA
parameter.
AAL2 CAC Reservation vs Rejection
Ave_Reserved_Bandm idth
Peak_CAC_Reservation
Iub_CAC_Reservation_R99
CAC_Rejection
4.2
HSDPA Usage
These measurements cover the data volume transferred and the Mac-hs throughput
HSDPA received data (Mbit) in RAN access points (=WCELLs). It is based on received MACd PDUs in HS-DSCH data frames at BTS.
11/27/2007-18
11/27/2007-12
11/27/2007-06
11/27/2007-00
11/26/2007-18
11/26/2007-12
11/26/2007-06
11/26/2007-00
11/25/2007-18
11/25/2007-12
11/25/2007-06
11/25/2007-00
11/24/2007-18
11/24/2007-12
11/24/2007-06
11/24/2007-00
11/23/2007-18
11/23/2007-12
11/23/2007-06
11/23/2007-00
11/22/2007-18
11/22/2007-11
11/22/2007-05
11/21/2007-22
11/21/2007-16
11/21/2007-10
11/21/2007-04
11/20/2007-08
11/20/2007-01
11/19/2007-19
11/19/2007-12
11/19/2007-06
11/19/2007-00
11/18/2007-18
11/18/2007-12
11/18/2007-06
11/18/2007-00
11/17/2007-18
11/17/2007-12
11/17/2007-06
11/17/2007-00
11/16/2007-18
11/16/2007-12
11/16/2007-06
11/16/2007-00
11/15/2007-18
11/15/2007-12
11/15/2007-06
11/15/2007-00
100.00%
90.00%
80.00%
70.00%
60.00%
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
16/21
25/01/2008
The average active HS-DSCH MAC-d throughput from network perspective calculated as the
HSDPA MAC-d throughput at BTS, divided by the active HS-DSCH time (time when there is
scheduled TTIs) from the network perspective. This formula does not include retransmissions
from the WBTS to the UE.
HSDPA Users
M550C11 SUM_AAL2_CONNECTIONS_HSDPA
M550C7 NBR_SAMPLES
And the maximum HSDPA simultaneous users, MAX_AAL2_CONNECTIONS_HSDPA can be
used as a reference to consider BTS to upgrade to 15 codes multiplexing in RAS06.
17/21
4.3
25/01/2008
HSDPA Performance
HSDPA Accessibility
The accessibility of all started allocations for HS-DSCH for NRT Traffic from user point of view
calculates the number of times when HS-DSCH channel has been established divided by the
number of times when HS-DSCH channel has been selected by cell specific PS.
RNC _ 605a
HSDPA MAC-d flow setup failures typically indicates lack of the resources (Radio, BTS, Iub)
for HSDPA or for UL return channel. HSDPA Setup can also fail if maximum amount of
HSDPA users in the cell exceeds. Failure counters itself are not enough to indicate
congestion but together with other counters e.g. BTS CE usage it is possible to identify cells
having capacity problems.
18/21
25/01/2008
This KPI is based on Traffic Measurement. The normal transition from HS-DSCH to
FACH/DCH is considered as a normal HS-DSCH release (including transitions due to mobility
and pre-emption). And the failure releases are mainly RL Failures releases or release other
than RL failure. RL failure release refer to the release of HS-DSCH allocation due the radio
link failure indication from BTS, RLC protocol reset internally in the RNC or UL RCL
unrecoverable error (Cell Update sent by UE).
RNC _ 609a
sum( REL _ ALLO _ NORM _ HS _ DSCH _ INT REL _ ALLO _ NORM _ HS _ DSCH _ BGR)
100%
sum(REL _ ALLO _ NORM _ HS _ DSCH _ INT REL _ ALLO _ NORM _ HS _ DSCH _ BGR
REL _ ALLO _ OTH _ FAIL _ HS _ DSCH _ INT REL _ ALLO _ OTHER _ FAIL _ HS _ DSCH _ BGR
REL _ ALLO _ RL _ FAIL _ HS _ DSCH _ INT REL _ ALLO _ RL _ FAIL _ HS _ DSCH _ BGR)
MAC-hs efficiency quantifies HSDPA retransmission ratio between BTS and HSDPA capable
UE done by MAC-hs. This is number of all successful sent MAC-hs PDUs divided by total
number of all transmitted MAC-hs PDUs including retransmissions
19/21
25/01/2008
MAC _ HS _ PDU _ RETR _ DIST _ CL _ O MAC _ HS _ PDU _ RETR _ DIST _ CL _1 MAC _ HS _ PDU _ R
MAC _ HS _ PDU _ RETR _ DIST _ CL _ 3 MAC _ HS _ PDU _ RETR _ DIST _ CL _ 4 MAC _ HS _ PDU _ RE
RNC _ 607b
ORIG _ TRANS _1_ CODE _ QPSK ORIG _ TRANS _ 2 _ CODE _ QPSK ORIG _ TRANS _ 3_ CODE
ORIG _ TRANS _ 4 _ CODE _ QPSK ORIG _TRANS _ 5 _ CODE _ QPSK ORIG _ TRANS _1_ CODE
ORIG _ TRANS _ 2 _ CODE _16QAM ORIG _TRANS _ 3 _ CODE _16QAM ORIG _ TRANS _ 4 _ CO
ORIG _ TRANS _ 5 _ CODE _16QAM RETRANS _1_ CODE _ QPSK RETRANS _ 2 _ CODE _ QPSK
RETRANS _ 3_ CODE _ QPSK RETRANS _ 4 _ CODE _ QPSK RETRANS _ 5 _ CODE _ QPSK
RETRANS _1_ CODE _16QAM RETRANS _ 2 _ CODE _16QAM RETRANS _ 3 _ CODE _16QAM
RETRANS _ 4 _ CODE _16QAM RETRANS _ 5 _ CODE _16QAM
Distribution of MAC-hs retransmissions indicates the number of retransmissions needed to
correctly deliver the MAC-hs PDU. MAC_HS_PDU_RETR_DIST_CL_0 below indicates no retransmission percentage is quite high, more than 80%.
M AC-hs Re-Transmission Distribution
MAC_HS_PDU_RETR_DIST_CL_0
MAC_HS_PDU_RETR_DIST_CL_1
MAC_HS_PDU_RETR_DIST_CL_4
MAC_HS_PDU_RETR_DIST_CL_5
MAC_HS_PDU_RETR_DIST_CL_2
MAC_HS_PDU_RETR_DIST_CL_3
100.00%
80.00%
60.00%
40.00%
20.00%
1/27/2008
1/26/2008
1/25/2008
1/24/2008
1/23/2008
1/22/2008
1/21/2008
1/20/2008
1/19/2008
1/18/2008
1/17/2008
1/16/2008
1/15/2008
1/14/2008
1/13/2008
1/12/2008
1/11/2008
1/10/2008
1/9/2008
1/8/2008
1/7/2008
1/6/2008
1/5/2008
1/4/2008
1/3/2008
1/2/2008
1/1/2008
0.00%
20/21
25/01/2008
CQI
HS-DSCH Link Adaptation and HSSCCH power control are based on the instantaneous
signal quality information received from the UE, i.e. Channel Quality Indicator, CQI.
Theoretically CQI can be used as a reference of potential throughput. Example of distribution
can be seen in figure below.
31
21/21
25/01/2008