Optimization of EGPRS Link Adaptation
Thesis work presentation
2007-01-16
Author: Jussi Nervola
Supervisor: Professor Riku Jäntti
Instructor: M.Sc. Petri Grönberg
Contents
• Objectives of the thesis
• Introduction
• Basics of EDGE and EGPRS
• EGPRS link adaptation
• Starting point for the thesis - performance with original link adaptation
• Methods
• Link adaptation optimization work
• Results of the optimization
• Laboratory measurement results
• Live network measurement results
• Conclusions
Objectives of the Thesis
• Problem:
• Current EGPRS link adaptation (LA) algorithm is not behaving ideally in all radio
conditions. LA is not using the most suitable Modulation and Coding Scheme (MCS)
in every radio condition. That reduces the performance of the system.
• Objective:
• To adjust the operation of the EGPRS link adaptation so that its performance it at the
ideal level (i.e. LA chooses the best MCS in every radio condition)
• If ideal performance level can be achieved the quality of the service will be improved:
• Increased throughput
• Smaller latency
• Improved coverage for EGPRS service
EDGE and EGPRS
AuC – Authentication Centre
BSC – Base Station Controller
Air A-bis A-ter A BTS – Base Transceiver Station
PSTN
EIR – Equipment Identity Register
HLR GGSN – Gateway GPRS Support Node
MS BTS BSC+PCU TC MSC/VLR
AuC HLR – Home Location Register
Gs
Gr EIR
Gb MS – Mobile Station
MSC – Mobile Switching Center
Gn
Internet PCU – Packet Control Unit
BTS SGSN GGSN SGSN – Serving GPRS Support Node
TC – Transcoder
Base Station Subsystem Network Switching Subsystem VLR – Visitor Location Register
• EGDE (Enhanced Data Rates for GSM Evolution) offers both packet switched and
circuit switched data services for the users of GSM network:
• Packet switched EGPRS (Enhanced General Packet Radio Service)
• Circuit switched ECSD (Enhanced Circuit Switched Data)
• EGPRS has been more popular among the manufacturers than ECSD
• Link adaptation is working on the Air-interface between MS and BTS and is controlled
by the PCU in BSC
EGPRS Link Adaptation
• In wireless networks the quality of the connection can change
from very good to very poor in short time 8.8 kbps
MCS-1 RLC/MAC block
22 (radio block)
• EGPRS has nine different Modulation and Coding Schemes MCS-2
11.2 kbps
(MCS-1 – MCS-9) in order to optimize the performance in different 28
radio conditions. All coding schemes have different amount of 14.8 kbps
MCS-3
37
redundancy from
channel coding
robustness and error correction. MCS-4
RLC data block,
number of bytes
17.6 kbps 44
• When user bit rate is low, robustness is high (e.g. MCS-1)
MCS-5
• When user bit rate is high, robustness is low (e.g. MCS-9) 22.4 kbps 56
• Two modulations are used: 29.6 kbps
MCS-6
74
• GMSK (Gaussian Minimum Shift Keying) for MCS-1 – MCS-4 MCS-7
44.8 kbps 56 56
• 8-PSK (8-Phase Shift Keying) for MCS-5 – MCS-9
MCS-8
54.4 kbps 68 68
MCS-9
• MCS that offers the best performance in current radio 59.6 kbps 74 74
environment should be selected
• Automatic selection of most suitable MCS is called link
adaptation. The selection is based on link quality measurements
Starting Point for the Thesis
Performance with Original Link Adaptation
LA is una ble to chan ge LA use s MCS-9 although MCS-8
to GMSK m odu lation wou ld offer higher throu ghput
• Throughput performance of link adaptation is DL FTP Thro ug hpu t, va ria ble RX- Le ve l, 4 TSLs
compared to the performance of all MCSs
1.00
MCS=1
0.90
MCS=2
MCS=3 0.80
FTP Th ro u g h p u t ( re la t iv e )
• Throughput is measured with FTP application in MCS=4
MCS=5
MCS=6
0.70
0.60
variable RX-level environment MCS=7
MCS=8
0.50
0.40
MCS=9
• Ideal LA would follow the envelope of the
LA o n 0.30
0.20
throughput of different MCSs -110 -105 -100 -95 -90 -85 -80 -75
0.10
0.00
-70
RX- Le v e l ( d Bm)
• Problem: LA is un ab le to ch ange
to GMSK m odu lation
• Original link adaptation was not always choosing UL FTP Thro ug hp ut, variab le RX-Le ve l, 2 TSL
most suitable MCS 1.20
MCS=1
MCS=2
1.00
MCS=3
FTP Thro ug hp ut (re la tive )
MCS=4
MCS=5 0.80
• Solution: MCS=6
MCS=7
MCS=8
0.60
• Change to GMSK modulation has to be done
MCS=9
0.40
LA o n
earlier (when link quality decreases)
0.20
0.00
• MCS-8 should be used more -110 -105 -100 -95 -90
RX-Le ve l (dBm )
-85 -80 -75 -70
Operation of EGPRS Link Adaptation
• In EGPRS link quality is measured in terms of Bit Error Probability (BEP).
BEP values are received in measurement reports from MS and BTS
• Mean 8-PSK BEP and coefficient of variation of 8-PSK BEP
• Mean GMSK BEP and coefficient of variation of GMSK BEP
• MCS is selected according to BEP values by using look-up tables
1. Selection of modulation
2. Selection of MCS
• from GMSK MCS-selection table or
• from 8-PSK MCS-selection table (below, from 3GPP specification)
Optimization Work
• Methods:
• Measure the performance of the current link adaptation (LA)
• Specify new parameter / table values for LA
• Measure the performance of the optimized LA and adjust values further
• The measurements were executed in laboratory using real network elements:
MS BTS BSC SGSN GGSN Server
• Three different laboratory scenarios:
• Variable RX-level, Variable C/I and Fading with variable RX-level
• Measured quantities:
• Downlink and Uplink throughput (with FTP and HTTP)
• RTT (with ping)
Results of the Optimization – Summary
Laboratory Throughput Coverage Throughput Coverage
measurements improvement improvement improvement improvement
RX-level DL +7% +4 dB UL +17% +5 dB
C/I DL +15% +1 dB UL +3% +6 dB
Live Throughput Coverage Throughput Coverage
Measurements improvement improvement improvement improvement
drive <not <not
route DL +11% measured> UL +31% measured>
• After the best version of optimized link adaptation was chosen, the performance
improvement was verified with several MSs in laboratory and live network
• Good results achieved through the optimization work:
• System performance was improved in all cases
• Throughput performance was improved in whole cell area in average by +11%
• Coverage of EGPRS service was improved by 1 – 6 dB depending on the conditions
• On the following slides some of the measurement results are explained briefly
Results of the Optimization – Laboratory Measurements
Variable RX-level Scenario UL Thro ug hput, va ria ble RX-Le ve l, 2 TSL
100%
Origina l LA
• Throughput performance was compared: Optim ize d LA11
80%
Optim a l LA
• Original LA
Thro ug hput (re lative )
60%
• Optimized LA
40%
• Optimal LA
20%
0%
• Uplink throughput performance of optimized -110 -105 -100 -95 -90 -85 -80 -75 -70
RX-Le ve l (dBm )
LA is at the ideal level with all MSs
DL Thro ug hput, va ria ble RX-Le ve l, 4 TSLs
• Downlink throughput performance of Origina l LA
100%
Optim ize d LA11 6280
optimized LA is at the ideal level with new Optim ize d LA11 7270
80%
Th ro u g h p ut (re lative )
Optim a l LA
MSs 60%
• With some older MSs there is 1 – 2 dB wide 40%
area where throughput is not optimal. This 20%
is not a problem in fading environment. 0%
-110 -105 -100 -95 -90 -85 -80 -75 -70
RX-Le ve l (d Bm )
Results of the Optimization – Live Network Measurements
• Drive tests executed in
Ruoholahti with Nokia UltraSite
• FTP throughput was measured
on uplink and downlink
separately
• Improvement in average
Downlink FTP throughput
+11%
• Improvement in average Uplink
FTP throughput +31%
Uplink Throughput Performance
• RLC uplink throughput figures along
different parts of the route:
RX-level during this part of the route: -85 – -105 dBm RX-level during this part of the route: -70 – -88 dBm
Original link adaptation Original link adaptation
RX-level
RX-level
Optimized link adaptation Optimized link adaptation
RX-level
RX-level
-> Reason for improvement: -> Reason for improvement:
•Earlier change to GMSK modulation •Using more MCS-8 instead of MCS-9
Conclusions
• The results achieved in the thesis are well in accordance with the original
objectives
• Performance improvement of optimized link adaptation was clearly seen in
laboratory and live network
• The optimized link adaptation version created in this thesis is available for the
customers in newest BSC software update in 1/2007
• Further Study Items
• Measurements with various MSs from different vendors
• Effect of certain GSM features (such as frequency hopping) on Link Adaptation
• Analysis in commercial operators’ networks by using BSC KPI (Key Performance
Indicator) data
• Link adaptation algorithm enhancement by introducing new algorithm inputs
• This would optimize modulation change with all MSs also on static conditions
Thank You!
Questions / Comments ?
