Performance analysis of Enhanced Uplink in UMTS network

1
Performance analysis of Enhanced Uplink in UMTS
network

• Jukka Pihonen
• Supervisor Prof. Riku Jäntti
• Instructor Laura Koskela, M.Sc.
• 2008-05-30

2
Outline

• Background
• Objectives and Methodology
• Introduction to Enhanced Uplink
• Introduction to Measurements
• Measurement Cases and Results
• Conclusions

3
Background

• Enhanced Uplink (EUL) High Speed Uplink Packet
  Access (HSUPA) Enhanced Dedicated Channel
  (E-DCH)
• Currently in UMTS network, the uplink
  transmission rate is restricted to 384 kbps
• Downlink transmission rate was improved over a
  year ago with HSDPA, which enables up to 14.4
  Mbps throughput for downlink, today 3.6 Mbps is
  used in commercial network
• EUL balances uplink and downlink performances by
  enabling 1.45 Mbps uplink throughput
• EUL also shortens response times and uses
  frequency band more effectively

4
Objectives and Methodology

• Objectives
• Introduce Enhanced Uplink
• Carry out measurements using Enhanced uplink both
  in commercial network and in a closed laboratory
  network
• Find out and analyze the performance of Enhanced
  Uplink
• Compare the performance to previous R99 uplink
• Methodology
• Introduction to Enhanced Uplink is based on 3GPP
  specifications, IEEE papers and books.
• Measurements were made both in laboratory and in
  commercial network

5
Introduction to Enhanced Uplink (1/4)

• Enhanced Uplink is defined in 3GPP Release 6
• Combines HSDPA by enabling a fast uplink
  connection
• EUL enhances uplink throughput first to 1.4 Mbps,
  later to 5.76 Mbps
• Main features are
• Fast Node B scheduling for uplink
• Fast Hybrid Automatic Repeat Requests (HARQs)
• Short Transmission Time Interval (TTI)
• Multicode transmission
• Introduces 5 new physical channels and 2 new MAC
  layer protocols

6
Introduction to Enhanced Uplink (2/4)

• Provides following new features
• Fast Node-B scheduling
• Scheduling is moved from SRNC to Node B, enables
  faster response times to constantly changing
  radio environments
• Node B based scheduling keeps noise raise as high
  as possible -gt each user gets best possible
  uplink throughputs
• Fast HARQ
• Retransmission control moved from SRNC to Node B,
  enables faster retransmission
• Short TTI
• New 2 ms TTI option to combine mandatory 10 ms
  TTI
• Enables faster retransmissions -gt reduced round
  trip times
• 2 ms TTI was not tested
• Multicode transmission
• Up to 4 parallel E-DPDCHs (2 x SF2 2 x SF4
  5.76 Mbps uplink throughput in Layer 1)

7
Introduction to Enhanced Uplink (3/4)

• Introduces 5 new physical channels, 2 for uplink
  and 3 for downlink
• Uplink channels
• E-DPDCH (E-DCH Dedicated Physical Data Channel)
• Carries uplink user data E-DCH traffic channel
• SF 2-256, power controlled
• Number of parallel E-DPDCHs is 1-4
• E-DPCCH (E-DCH Dedicated Physical Control
  Channel)
• Carries uplink control information
• SF 256, power controlled
• Carries E-DCH Transport Format Combination
  Identifier (E-TFCI), Retransmission Sequence
  Number (RSN) and a single bit called happy bit
• Downlink channels
• E-AGCH (E-DCH Absolute Grant Channel)
• Carries absolute scheduling grants, SF 256
• E-RGCH (E-DCH Relative Grant Channel)
• Carries relative scheduling grants, SF 128
• E-HICH (E-DCH HARQ Indicator Channel)
• Carries ACKs/NACKs, SF 256

8
Introduction to Enhanced Uplink (4/4)

• New MAC-layer protocols
• MAC-e
• Between UE and Node B
• Controls HARQs and scheduling
• MAC-es
• Between UE and SRNC
• Reorders MAC-es Protocol Data Units (PDUs) in
  case of soft handover
• Disassembles dedicated channels in RNC

9
Measurements

• Measurements were done in commercial network and
  in a closed laboratory network
• Totally four measurement cases were carried out
• General performance of EUL
• Performance of EUL in different environments
• Uplink performance comparison between EUL and R99
• Connection setup times and round trip times
• Measurements in commercial network were made as a
  drive tests i.e. the UE was placed in a car and a
  certain route was driven through
• Measurements in laboratory were made in a stabile
  place, nearby Node B
• Laboratory has its own isolated network -gt
  minimize external interferences

• Measurement unit

10
Measurement routes (1/2)

• Long route
• Measurements General performance of EUL and EUL
  performance in different environments

• Lengths
• Urban 3.5 km
• Suburban 18.1 km
• Motorway 6.7 km

Purple line Urban
Green line Suburban
Red line Motorway
11
Measurement routes (2/2)

• Short route
• Measurement EUL to R99 Uplink comparison

• Urban area is rounded with a red circle
• Total length 11.1 km

12
Case 1 (1/7)

• First test case was about general EUL performance
• Idea was to figure out how the EUL works in
  practice
• Test was carried out in commercial network
• Results in this case covers following things
• Throughputs in different layers
• Happy bit status handling
• MAC-e retransmission rates
• E-DCH channel configurations
• Tx power in different RSCP values

13
Case 1 (2/7)

• Distribution of measured samples
• Best reliability between -55 dBm to -100 dBm

14
Case 1 (3/7)

• Uplink throughputs in different layers
• Application layer throughput was 89 of total
  carried bits in physical layer

15
Case 1 (4/7)

• Uplink application throughput and happy bit
  handling
• EUL performed even when RSCP was -110 dBm
• Happy bit started to be happy when RSCP reached
  -90 dBm
• To inform unhappy, following 3 conditions must be
  fulfilled
• UE transmit as much scheduled data as allowed by
  scheduling grant
• UE has enough power to transmit at higher data
  rate
• Total buffer status in UE requires more
  transmission

16
Case 1 (5/7)

• Retransmission rates in MAC-e
• Describes the retransmission rates, adjusted by
  HARQ process
• Retransmission were immune to variation of RSCP
  value -gt link adaptation worked properly

17
Case 1 (6/7)

• E-DPDCH channel configurations
• The highest throughputs (up to 1.38 in Radio Link
  Control Layer) can be reached with 2 x SF 4

18
Case 1 (7/7)

• Transmit power
• Increases linearly, reached maximum value when
  RSCP -103 dBm

19
Case 2 (1/3)

• Second test case analyzed the performance in
  different environments
• Idea was to figure out how the EUL works in
  different environments (urban, suburban
  motorway)
• Results in this case covers following things
• Uplink throughputs
• Transmit powers
• Retransmission rates
• Happy bit status
• Founds out that the performance was depended on
  the environment type
• The best field strength was noticed in urban
  part, the weakest field strength was in the
  suburban part
• Following figure shows the distribution of
  samples
• Following table summarizes the result of this case

20
Case 2 (2/3)

• Statistics and RSCP value distribution

  Urban Suburban Motorway
Average speed (km/h) 21.3 37.5 80.8
Measurement duration (min) 945 285700 500
Number of measured samples 585527 1737413 299745
Number of averaged samples 1133 3354 580
21
Case 2 (3/3)

• This table summarizes the results which were
  gathered from different environments
• Retransmission rates were immune to different
  environments
• In all environment, the same Max UL throughput
  was reached
• In suburban, the happy bit was most as happy

  Urban Suburban Motorway
Uplink application throughput (kbps) 919 719 618
Max UL application throughput (kbps) 1383 1382 1383
Tx Power (dBm) 1.36 7.8 7.62
 
MAC-e 1st Retransmission rate () 2.26 2.37 1.87
MAC-e 2nd Retransmission rate () 0.98 1.09 0.64
MAC-e 3rd Retransmission rate () 0.95 1.03 0.47
 
Happy bit status ( of happy) 1.34 7.58 1.82
22
Case 3 (1/2)

• Third test case compared R99 uplink and EUL in
  performance manner
• Idea was to find out how much better performance
  can be expected with EUL compared to R99 uplink
• EUL provided 2.3 times better average uplink
  throughput compared to R99
• EUL used 4.5 dB more transmit power
• Circumstances were not stabile between routes,
  when EUL was measured, the average RSCP value was
  4 dB higher due to better weather conditions.
• increases more the power difference between EUL
  and R99 uplink

23
Case 3 (2/2)

• Uplink application throughputs in EUL and R99
  networks
• Continuous file uploading
• R99 uplink was immune to variation of RSCP value
• In all RSCP values, EUL provides better
  throughput

24
Case 4

• The last test case was performed in laboratory
  environment
• Free of traffic and low interference levels.
• UE was in fixed place, used one certain Node-B
  located nearby the UE
• Purpose was to find out the connection setup
  times and round trip times
• Results are summarized in the following table
• 2 ms TTI was not tested

 PDP-context activation time (s)   PDP-context activation time (s)   PDP-context activation time (s)   PDP-context activation time (s) 
     R99/R99 R99/HSDPA EUL/HSDPA
    1.41 1.38 1.36
   
Round Trip Times (ms) Round Trip Times (ms) Round Trip Times (ms) Round Trip Times (ms)
Packet size R99/R99 R99/HSDPA EUL/HSDPA
32 bytes 129 79 68
128 bytes 131 90 84
512 bytes 260 172 160
25
Conclusions

• Max UL throughput was 1.38 Mbps in application
  layer
• Scheduling worked properly keeping retransmission
  rates constant through the RSCP value scale
• Happy bit status stayed unhappy until RSCP
  reached -85 dBm. In lower RSCP values, happy bit
  was more often happy
• Best throughputs measured in urban environment
• EUL provided better throughput in all RSCP values
  compared to R99 uplink
• EUL required more transmit power compared to R99
  uplink
• Round trip times decreased by 9 compared to
  R99/HSDPA network configuration and 40 compared
  to R99/R99 network configuration
• Future aspects in EUL
• Mobility enhances
• Uplink throughput raises up to 5.76 Mbps