HSDPA

Introduction and Troubleshooting

for Low Throughput

Sadok Ben Ali

28 Nov 2008

Ericsson Internal 1 2008-11-28

 Key characteristics of HSDPA
Scheduled
user
high data
rate

User 1

User 2 low data rate

Time
#1 #2 #1#2 #1 #2 #1

Scheduling

Power
High data rate

Total cell
HS-DSCH (rate controlled)

power
Dedicated channels
Low data rate (power controlled)
Common channels
t
Fast Link Adaptation HS-DSCH with dynamic power allocation

Ericsson Internal 2 2008-11-28

Shared Channel Transmission (1)

 A set of radio resources dynamically shared among multiple users,

primarily in the time domain
SF4

SF8

SF16

SF32 HS-PDSCH Cell CCH(s)

SF64 s-CCP CH
(PS32Kbps on FACH)

SF128
HS-SCCH
SF256
p-CCPCH PICH
p-CPICH AICH

 Up to 15 codes (SF16) can be allocated and shared between the

users. It also depends on what the UE can support.

Ericsson Internal 3 2008-11-28

 Shared Channel Transmission (2)
HSDPA channel structure
 DL: High-Speed Downlink Shared Channel – HS-DSCH
 DL: High-Speed Shared Control Channel(s) – HS-SCCH (SCHEDULING)
 Associated Dedicated Channel – A-DCH bidirectional DCH for SRB3.4 signaling (
DPCCH+DPDCH)
 UL :HS Dedicated Physical Common Control Channel - HS-DPCCH (CQI, HARQ
ACK/NACK)
Iu Iu
RNC Iur RNC

Iub Iub
Associated
Dedicated
Channels

Ericsson Internal 4 2008-11-28

 Fast Link Adaptation (1)
CQI – Channel Quality Indicator
 UE sends CQI info in the UL to aid rate adaptation and scheduling
 CQI (1-30) provides the Node B with a measure of the UE's perceived channel quality and the UE receiver
performance
 The CQI report estimates the number of bits that can be transmitted to the UE using a certain assumed
power with a block error rate of (blerQualityTargetDl =1%)
 UE assumes a HS- PDSCH power to calculate CQI : PHS_assumed = RSCPCPICH +
hsMeasurementPowerOffset +  (3GPP 25214)
 CQI = 0 if SNR <= -16, CQI=SNR/1.02 + 16.62 if -16<SNR<14, CQI = 30 if SNR=>14

Ericsson Internal 5 2008-11-28

Fast Link Adaptation (2)
 Rate control
– Adjusts data rate based on the Radio conditions (CQI)
– Fast Adaptation : 2 ms TTI basis
– Adaptive Modulation (QPSK and 16 QAM) and Coding
– Use “available power”
Power

High data rate

Total cell power

HS-DSCH (rate controlled)

Dedicated channels
(power controlled)
Low data rate
Common channels

t
HS-DSCH with dynamic power allocation

Ericsson Internal 6 2008-11-28

Fast Hybrid ARQ with Soft Combining
 Rapid retransmissions of erroneous data
– Hybrid ARQ protocol terminated in Node B
 short RTT (typical example: 12 ms)
– Soft combining in UE of multiple transmission attempts
 reduced error rates for retransmissions

Ericsson Internal 7 2008-11-28

Fast Channel-dependent Scheduling
(MAC-hs)

 Scheduling => which UE to transmit to at a given time

instant and at what rate
– MAC-hs (a new MAC sub-layer in RBS)

 Basic idea: transmit at fading peaks

– May lead to large variations in data rate between users

Scheduled
user high data rate

User
1
low data rate
User Time
2
#1 #2 #1 #2 #1 #2 #1

Ericsson Internal 8 2008-11-28

MAC-hs Functions
Resource Estimation
• Estimation for available HS-PDSCH power and codes
• PHS = Pmax - hsPowerMargin - Pnon-HS
• PHSPDSCH = PHS - PHSSCCH
• HS-SCCH power is set with reference to CPICH
• Code information available from RNC

Queue Validation
• Data to transmit is available
• ADCH in uplink in synch for the UE
• UE’s minimum TTI capability is ok
• Adjusted CQI exists for the UE
• Suitable HARQ process exists
• MAC-hs transmission window is not full

Queue Selection
• f(CQI)
• f(C/I )
• f(average rate)

Ericsson Internal 9 2008-11-28

Queue Selection Algorithms
 Ericsson supports the following Scheduling Algorithms
and is configurable per Node B (queueSelectAlgorithm)

– Round Robin (RR)

 Cyclically assign the channel to users without taking channel
conditions into account
 Simple but poor performance

– Proportional Fair (PF)

 Assign the channel to the user with the best relative channel quality
 High throughput, fair

– Max C/I Ratio

 Assign the channel to the user with the best channel quality
 High system throughput but not fair

Ericsson Internal 10 2008-11-28

Channel Switching (New in P5)
P4 P5

HS

64/HS
hsdschinactivityTimer

IDLE

Switching between HS-DSCH and DCH happens to support

Iur, IRAT and IFHO Mobility…
Ericsson Internal 11 2008-11-28
 HSDPA Throughput Transport Number of Reference power NIR XRV
CQI value Modulation adjustment 
Block Size HS-PDSCH
0 N/A Out of range
1 137 1 QPSK 0 9600 0
Maximum Minimum Modulation Coder
2 173 1 QPSK 0
HS-DSCH category number of HS- inter-TTI Throughput 3 233 1 QPSK 0
DSCH codes interval Mbit/s 4 317 1 QPSK 0
received 5 377 1 QPSK 0
6 461 1 QPSK 0
Category 1 5 3 QPSK&16QAM 1.2 7 650 2 QPSK 0
8 792 2 QPSK 0
Category 2 5 3 QPSK&16QAM 1.2 9 931 2 QPSK 0
10 1262 3 QPSK 0
Category 3 5 2 QPSK&16QAM 1.8 11 1483 3 QPSK 0
12 1742 3 QPSK 0
Category 4 5 2 QPSK&16QAM 1.8 13 2279 4 QPSK 0
14 2583 4 QPSK 0
Category 5 5 1 QPSK&16QAM 3.6 15 3319 5 QPSK 0
16 3565 5 16-QAM 0
Category 6 5 1 QPSK&16QAM 3.6 17 4189 5 16-QAM 0
18 4664 5 16-QAM 0
Category 7 10 1 QPSK&16QAM 7.3 19 5287 5 16-QAM 0
20 5887 5 16-QAM 0
Category 8 10 1 QPSK&16QAM 7.3 21 6554 5 16-QAM 0
22 7168 5 16-QAM 0
Category 9 15 1 QPSK&16QAM 10.2 23 7168 5 16-QAM -1
24 7168 5 16-QAM
Category 10 15 1 QPSK&16QAM 14.4 25 7168 5 16-QAM -3
26 7168 5 16-QAM -4
Category 11 5 2 QPSK only 0.9 27 7168 5 16-QAM -5
28 7168 5 16-QAM -6
Category 12 5 1 QPSK only 1.8 29 7168 5 16-QAM -7
30 7168 5 16-QAM -8

• HS L1 Throughput = W * Number of codes * Modulation scheme /SF=(3.84 M cps 5 * 4 )/16 = 4.8 Mbps

• Max HS MAC-d user data = ( Tbsize (7168) / 2ms) = 3.584 Mbps

The deference between MAC-d and L1 throughput due to Rate matching & turbo code

•The real Max HS MAC-d user data = ( Tbsize / TTI)*(1 - BLER)

Ericsson Internal 12 2008-11-28

 HSDPA Throughput Analysis

• Radio Environment
• Low CQI or Low Modulation rate : it could be due to no enough power
(Much power allocated for R99), bad coverage....
• High DTX : UE is not being scheduled. It could be that are many users in
the cell
• ACK/NACK rate : UE is not receiving well the data and is sending NACKs,
it can be due to improper parameters setting

• USIM
• HLR need to be verified for the USIM profile (DL throughput /UL
throughput)

• UE Capability
• ueCapability-Info : UE category shall be more than 6

Ericsson Internal 13 2008-11-28

HSDPA Throughput Analysis
 Laptop
 TCP Window Size should be 102400，MTU=1460

 RBS
 Check RBS parameters
 IMA configuration(number of E1s)
 AtmTrafficDescriptor on AAL2 (aal2QoSAvailableProfiles) in RBS and RNC side
 AAL2 Path parameters such as PCR, SCR...
 Check Hardware (some times replacing the HSDPA board can resolve the
problem)

• Iub Flow control

• Proper Setting for maxHsRate
• The HSDPA Iub flow control algorithm adjusts to the Iub congestion by adapting a reference value
(targetHsRate) every second. The reference value is initially set to maxHsRate, which also determines
the maximum value for targetHsRate. The more congestion detections, the lower the available bit rate
over Iub
• Iub flow capacity is affected by quality of the transport network

Ericsson Internal 14 2008-11-28

 HSDPA Throughput Analysis
• CN
In R4 GGSN by default there is limitation 2M bps in the downlink. To allow more
bandwidth, the maximum-bandwidth-downlink has to be set to a bigger value with
the policing in the PDP context definition.
RAB-Assignment Request:
{ trafficClass interactive, rAB-AsymmetryIndicator asymmetric-bidirectional,
maxBitrate
{ 2048000, (>>> DL)
384000 }, (>>> UL)}
 In SGSN, check if the correct QoS profile has been linked to the IMSI series that
were used in the test;
 In GGSN, verify the policing with the PDP context to see if there is any limitation on
the downlink.
 The link between the GGSN and the IP router in the IP backbone has to have
enabled the auto-negotiation and fixed 100 Mbit/s speed in full-duplex mod

Ericsson Internal 15 2008-11-28

 HSDPA Throughput Analysis
Counters :
Aal2PathVccTp pmBwErrBlocks VplTp pmReceivedAtmCells
pmBwLostCells pmTransmittedAtmCells
pmBwMissinsCells pmSecondsWithUnexp
pmFwErrBlocks
pmFwLostCells
pmFwMissinsCells VclTp pmReceivedAtmCells
pmLostBrCells pmTransmittedAtmCells
pmLostFpmCells

HsDschResources pmAckReceived
RncFunction pmSentPacketDataHs1 (is stepped only if < 1kB)
pmAverageUserRate pmSentPacketDataHs2 (is stepped only if < 10kB but > 1kB)
pmIubMacdPduCellReceivedBits pmSentPacketDataHs3 (is stepped only if < 100kB but > 10kB)
pmNackReceived pmSentPacketDataHs4 (is stepped only if > 100kB).
pmNoActiveSubFrames pmTotalPacketDurationHs1 - duration for data < 1kB.
pmTotalPacketDurationHs2 - duration for data < 10kB but > 1kB
pmNoInactiveRequiredSubFrames
pmTotalPacketDurationHs3 - duration for data < 100kB but > 10kB
pmReportedCqi pmTotalPacketDurationHs4 - duration for data > 100kB
pmSumAckedBits
pmSumNonEmptyUserBuffers
pmSumTransmittedBits
pmTransmittedCarrierPowerNonHs
pmUsedCqi

IubDataStreams pmCapAllocIubHsLimitingRatio
pmHsDataFramesLost
pmHsDataFramesReceived
pmIubMacdPduRbsReceivedBits
pmTargetHsRate

Ericsson Internal 16 2008-11-28

 HSDPA Throughput Analysis
KPIs:
RNC :
 HsTputAverg = 8 * 1000 * ( pmSentPacketDataHs1 / 1000 + pmSentPacketDataHs2 / 1000 +
pmSentPacketDataHs3 + pmSentPacketDataHs4 ) / ( pmTotalPacketDurationHs1 + pmTotalPacketDurationHs2 +
pmTotalPacketDurationHs3 + pmTotalPacketDurationHs4 )

 HsTputGt100KbAverg = 8 * 1000 * (pmSentPacketDataHs4) / (pmTotalPacketDurationHs4)

HsTputAverg is looking at HS throughput over all data transmission sizes, even if a small amount of data is sent.
HsTputGt100KbAverg is for data transmissions > 100kB, the HsTputGt100KbAverg should give a better indication
of the performance achieved with HS because you'd be most interested in the speed when performing large
downloads rather than when using smaller transmission sizes where you may not be utilizing the full 2ms air
interface frame.

RBS :
 The average MAC-hs throughput per PS Interactive HSDPA user: If MAC-hs Data rate mush
pmSumAckedBits / (TTI * pmSumNonEmptyUserBuffers) bigger than MAC-hs throughput
 The average MAC-hs Data rate per PS Interactive HSDPA user: per user there is a high NACK
pmSumTransmittedBits / (TTI * pmSumNonEmptyUserBuffers) subframe, we need to check
 The MAC-hs throughput on cell level:
BLER, R99 used power…
pmSumAckedBits / (TTI * pmNoActiveSubFrame)
 The MAC-hs data rate on cell level:
pmSumTransmittedBits / (TTI * pmNoActiveSubFrame)

Ericsson Internal 17 2008-11-28

HSDPA Throughput Analysis
IubDataStreams
Is to check capacity limitation in the Iub Interface :
 pmCapAllocIubHsLimitingRatio : is the relative number of occurrences when the calculated Capacity Allocation
figure is limited by the Iub HS bandwidth during a 100 ms period (compared to the total number of 100 ms periods in
the PM interval. This is related to Physical throughput, so it will include all the overheads, retransmissions, and the
HSDSCH throughput is not mapped to this counter directly. If the counter is pegging, that means the Iub bandwidth
limit as been met sometimes, so the configuration of the Iub, i.e., the AtmTrafficDescriptor on AAL2 VC's should be
verified.
 pmtargetHsRate : is real capacity of the Iub interface is calculated by the Iub Flow Control according to
retransmission ratio and MaxHsRate

Transport Network
 VplTp and VclTp counters are used to verify that everything is sent from RNC is received in RBS.
 Is to check the quality of transport network.

Ericsson Internal 18 2008-11-28

 HSDPA PARAMETRS (1)

MO Class Parameter Value According to…

ChannelSwitching downswitchThreshold 0 E/// Default
ChannelSwitching downswitchTimerThreshold 1 BaseLine 1.6
ChannelSwitching hsdschInactivityTimer 8 BaseLine 1.6
Rcs hsDschRcLostT 100 E/// Default
RncFunction hsCellChangeAllowed 1 BaseLine 1.6
RncFunction hsOnlyBestCell 1 BaseLine 1.6
RncFunction hsToDchTrigger. servHsChangeInterRnc ON BaseLine 1.6
RncFunction hsToDchTrigger. servHsChangeIntraRnc ON BaseLine 1.6
RncFunction hsToDchTrigger. changeOfBestCellIntraRnc ON BaseLine 1.6
RncFunction hsToDchTrigger. poorQualityDetected ON BaseLine 1.6
UeMeasControl hsQualityEstimate 1 BaseLine 1.6
UeMeasControl hsTimeToTrigger1d 1280 BaseLine 1.6

MO Class Parameter Value According to…

IubDataStream maxHsRate 29 BaseLine 1.6, 2 E1s links
NodeBFunction flexibleSchedulerOn TRUE BaseLine 1.6
NodeBFunction supportOf16qam TRUE BaseLine 1.6

Ericsson Internal 19 2008-11-28

HSDPA PARAMETRS (2)
MO Class Parameter Value According to…
Carrier cqiAdjustmentOn 1 BaseLine 1.6
Carrier hsScchMaxCodePower -20 E/// Default
Carrier queueSelectAlgorithm 1 BaseLine 1.6
HsDsch codeThresholdPdu656 0 BaseLine 1.6
HsDsch deltaAck1 5 BaseLine 1.6
HsDsch deltaAck2 7 BaseLine 1.6
HsDsch deltaCqi2 6 BaseLine 1.6
HsDsch deltaNack1 5 BaseLine 1.6
HsDsch deltaNack2 7 BaseLine 1.6
HsDsch hsMeasurementPowerOffset 80 E/// Recommended
HsDsch numHsPdschCodes 5 BaseLine 1.6
HsDsch numHsScchCodes 3 BaseLine 1.6
RbsLocalCell dynamicHsPdschCodeAdditionOn 1 BaseLine 1.6
RbsLocalCell maxNumHsdpaUsers 16 BaseLine 1.6
RbsLocalCell maxNumHsPdschCodes 15 BaseLine 1.6
UtranCell beMarginDlCode 1 BaseLine 1.6
UtranCell dlCodeAdm 85 BaseLine 1.6
UtranCell hsdpaUsersAdm 12 BaseLine 1.6

Ericsson Internal 20 2008-11-28

Ericsson Internal 21 2008-11-28