inacon hsdpa networak optimization troubleshooting

HSDPA Network Optimization & Troubleshooting INACON GmbH Kriegsstrasse 154 76133 Karlsruhe Germany www.inacon.com e-mail: inacon@inacon.de... 9 -HSDPA Network Optimization & Troublesho

Trang 3

HSDPA Network Optimization

&

Troubleshooting

INACON GmbH Kriegsstrasse 154

76133 Karlsruhe

Germany www.inacon.com e-mail: inacon@inacon.de

Trang 4

Cover design by Stefan Kohler

Trang 5

HSDPA Network Optimization

&

Troubleshooting

Stefan Blomeier

Trang 6

All INACON publications use the same color codes to distinguish mandatory from optional or conditional parts in frame formats or optional from mandatory data blocks or signaling messages in scenarios The different color codes are explained underneath:

• Color Codes in Frame Formats:

• Color Codes in Scenarios:

Trang 7

Foreword of the Publisher:

Dear Reader:

Note that this book is primarily a training document because the primary business of INACON GmbH

is the training and consulting market for mobile communications As such, we are proud to providing high-end training courses to many clients worldwide, among them operators like AT&T Wireless, INMARSAT or T-MOBILE and equipment suppliers like ERICSSON, MOTOROLA, NOKIA or SIEMENS

INACON GmbH is not one of the old-fashioned publishers With respect to time-to-market, form-factor, homogenous quality over all books and most importantly with respect to after-sales support, INACON GmbH is moving into a new direction Therefore, INACON GmbH does not leave you alone with your issues and this book but we offer you to contact the author directly through e-mail (inacon@inacon.de), if you have any questions All our authors are employees of INACON GmbH and all of them are proven experts in their area with usually many years of practical experience

The most important assets and features of the book in front of you are:

• Extreme degree of detailed information about a certain technology.

• Extensive and detailed index to allow instant access to information about virtually every parameter, timer and detail of this technology.

• Incorporation of several practical exercises.

• If applicable, incorporation of examples from our practical field

experiences and real life recordings.

• References to the respective standards and recommendations on virtually every page.

Finally, we again like to congratulate you to the purchase of this book and we like to wish you success

in using it during your daily work

Sincerely,

Gunnar Heine / President & CEO of INACON GmbH

PS: Please check for our UMTS online encyclopaedia at www.inacon.com

Trang 9

9

-HSDPA Network Optimization &

Troubleshooting

Table of Contents

HSDPA in Practice 12

HSDPA Channel Overview 15

Example of HSPA Drive Testing with TEMS 21

(1) CQI during HSxPA Serving Cell Change 21

(2) HSDPA various Throughput Rates 24

(3) HS-DSCH Retransmission and BLER 26

(4)HS-DSCH Hybrid Automatic Repeat Request Processes 28

(5) HS-SCCH Decoding 30

(6) Iub-Bandwidth Limitation or bad Flow Control between NodeB and RNC 32

HS-PDSCH Symbol Rate 34

Code Rate R 34

HSDPA Throughput versus SIR – Simulation 38

QPSK with Code Rate R ¼ 38

QPSK with Code Rate R ½ 38

QPSK with Code Rate R ¾ 38

16-QAM with Code Rate R ½ 38

16-QAM with Code Rate R ¾ 38

HSDPA Throughput according to Antti Toskala 40

CPICH variation versus MPO variation 44

Session Throughput versus CPICH Power and MPO 46

Application Throughput vs CPICH Power and MPO 48

CQI versus Ec/No 50

Can 15 HS-PDSCH’s Codes be really allocated? 52

Solutions to the Problem of Code Shortage 54

Introduce 2nd UMTS Frequency 54

F-DPCH in Rel 6 54

Dynamic Code Tree Allocation between R99 and HS-PDSCHs 56

Semi Dynamic Code Tree Allocation 56

Dynamic Code Tree Allocation 56

Physical Shared Channel Reconfiguration 58

CQI mapping table for UE Categories 1-6, 7-8, 9 and 10 65

HSDPA during Compressed Mode Operation 67

(1) Inter Frequency Handover – Event 2D 69

(2) Inter Frequency Handover – Compressed Mode Parameter 71

(3) Inter Frequency Handover – Compressed Mode Parameter 73

Radio Bearer Setup: 75

Radio Bearer Release: 75

Radio Bearer Reconfiguration: 75

Cell Update Confirm 75

Transport Channel Reconfiguration: 75

Physical Channel Reconfiguration: 75

HS-DSCH Capacity Request procedure 89

HS-DSCH DATA FRAME 97

material is prohibited and will be prosecuted to the full extent of German and international laws Version Number: 1.2

Trang 10

HSDPA Network Optimization &

Troubleshooting

(1) Understanding Flow Control Flow Control over HSDPA 105

Flow Control of Cat 8 113

114

Iub Flow Control 115

Solutions for Practical Exercises 116

Trang 11

11

-HSDPA Network Optimization &

Troubleshooting

Intentionally left blank

material is prohibited and will be prosecuted to the full extent of German and international laws Version Number: 1.2

Trang 12

HSDPA Troubleshooting

Trang 13

13

-HSDPA Troubleshooting

Objectives

After this Lecture the Student will be able to:

throughput performance

via RRC

Trang 15

mac-d-PDU-Size : 656 mac-d-PDU-Index : 0 -deltaACK : 7

deltaNACK : 7 ack-NACK-repetition-factor : 1 -dl-HSPDSCH-Information

hs-scch-Info modeSpecificInfo : fdd hS-SCCHChannelisationCodeInfo : [0 ] : 4

[1 ] : 5 [2 ] : 6 [3 ] : 7 measurement-feedback-Info modeSpecificInfo : fdd measurementPowerOffset : 21 feedback-cycle : fc2

cqi-RepetitionFactor : 1 deltaCQI : 7

dl-InformationPerRL-List

DL-InformationPerRL-List-r5 : [0 ] :

modeSpecificInfo : fdd primaryCPICH-Info primaryScramblingCode : 390 servingHSDSCH-RL-indicator : True

15

HSDPA Channel Overview

Trang 16

Practical Exercise:

Trang 17

17

Please fill in the Physical Channel Names!

Answer:

Trang 18

Example of HSDPA Operation in ROMES

Trang 19

19

Example of HSDPA Operation in ROMES

[ROMES Drive test tool (Rohde & Schwarz)]

Trang 20

Trang 21

21

Example of HSPA Drive Testing with TEMS

(1) CQI during HSxPA Serving Cell Change

• CQI goes with Ec/No of the serving HSxPA cell, so late change of HSXPA serving cell leads to low HSDPA throughput in mobility

Trang 22

Trang 23

23

(2) HSDPA various Throughput Rates

• Ec/No of HS Serving Cell  CQI  HS Physically Requested Throughput  HS Physically Scheduled Thp  HS

Physically Served Thp

Trang 24

Trang 25

25

(3) HS-DSCH Retransmission and BLER

• HS-DSCH NACK Rate ~ HS-DSCH Retransmission Rate ~ HS-DSCH BLER 1st Transmission

Trang 26

Trang 27

27

(4)HS-DSCH Hybrid Automatic Repeat Request Processes

• HS-DSCH HARQ Processes: max 8; typical 6 for Cat 6 and Cat 8, 9 and 10

Trang 28

Trang 30

to be transmitted!!!

As there is only 1 hs-scch, codes are not shared with other user(s).

good Ec/No but HSDSCH decoding success rate is not 100% means that the UE was not scheduled all the time.

Trang 33

33

Throughput and Code Rate R

Here we would like to focus on the user plane throughput and physical amount of bit to be transferred

Trang 34

Achievable Throughput without Retransmissions

Trang 35

35

Achievable Throughput without Retransmissions

Max throughput = R * {2 | 4 [bits/symbol]} * 240 ksymbol/s * #HS-PDSCH’s

• (1) Example: QPSK with R = 3/4 and 10 HS-PDSCH’s

Max throughput = 3/4 * 2 bits/symbol * 240 ksymbol/s * 10 = 3.6 Mbit/s

• (2) Example: 16-QAM with R = 2/4 and 10 HS-PDSCH’s

Max throughput = 2/4 * 4 bits/symbol * 240 ksymbol/s * 10 = 4.8 Mbit/s

Trang 36

HSDPA Throughput versus SIR – Simulation

Trang 37

37

HSDPA Throughput versus SIR – Simulation

QPSK with Code Rate R ¼

240 Ksymbols x 2 Bits/Symbol x ¼ = 120 kbit/s per HS-PDSCH

QPSK with Code Rate R ½

240 Ksymbols x 2 Bits/Symbol x ½ = 240 kbit/s per HS-PDSCH

QPSK with Code Rate R ¾

240 Ksymbols x 2 Bits/Symbol x ¾ = 360 kbit/s per HS-PDSCH

16-QAM with Code Rate R ½

240 Ksymbols x 4 Bits/Symbol x ½ = 480 kbit/s per HS-PDSCH

16-QAM with Code Rate R ¾

240 Ksymbols x 4 Bits/Symbol x ¾ = 720 kbit/s per HS-PDSCH

Trang 38

HSDPA Throughput according to Antti Toskala

Trang 39

39

HSDPA Throughput according to Antti Toskala

Usually a proportional fair resource scheduler is used in HSDPA

The Network Efficiency or Effective Load for a UMTS System is best (load is lowest) for a BLER between 10% 20% up to 30%

Depending on UE speed and channel interference characteristic (e.g indoor, outdoor macro cell, micro cell), more or

less BLER gives better System Performance

The lower the Eb/No and amount of (re-)transmissions is/are for successful decoding, the less Noise/Interference is generated and the

higher System Capacity in UMTS gets Unfortunately, the lower Eb/No is, the more retransmission might be necessary for successful decoding

(depending on the interference channel characteristics) and the higher the BLER becomes With higher BLER however, time and power is

wasted so the system becomes inefficient again So somewhere between 1% BLER and 70% BLER there is an optimum where BLER and

Eb/No for successful decoding offer the most benefit for UMTS System Capacity And that BLER versus Load relation is best for about

20% BLER based on first transmission

I want to emphasize that the UE has to be delivered with a CQI fitting to 10 % BLER otherwise it will fail acceptance test It is up to the NodeB to allow the UE to cheat by adding an offset to the CQI or by requesting an offset to the measured CPICH power

Actually there is a potential harm: Since a higher BLER will result in more retransmission the latency of the system will be increased

This might lead to more latency than tolerated by the QoS

FYI: LTE (Long Term Evolution project of 3GPP) calculates with 30% BLER

Is there any harm caused to the network if more and more UE's cheat in their CQI reporting and report CQI for e.g 20% BLER (single retransmission)instead of 10%?

Generally speaking: not really However, if more and more UE's report CQI significantly higher than 20%, e.g 30% then the network efficiency and

overall network throughput suffers As long as the CQI reporting is for a BLER between 10% to 20% BLER I see not real issue Of course, reporting CQI for a BLER of 5% or only 1% BLER is not good either for UE's individual throughput and neither for the overall system throughput "The NodeB would be then too anxious to go for high Blocksizes and is not trying its luck over air"

You must be aware that the NodeB Scheduler may always attend the UE promising the highest throughput with its CQI So in the field such "cheating" UE's may get served more often despite their high BLER

This can be partially mitigated when the Scheduler uses the ACK/NACK ratio as an additional weighting when selecting the UE's for 2ms

HS-DSCH transmisson For example if the NodeB targets a 10%BLER of ACK/NACK based on first transmission, then a good performing UE will get the max Ack/Nack weight of 9/9 (=100%) <=> 90% success rate is the target of the NodeB if 10% BLER is set A "cheating UE" with a high BLER of e.g 30% only achievs a 70% success Rate on first transmission, so the scheduler weight for such an UE would be reduced by a factor of 7/9 However,from my experience, the most important scheduler input is always the indicated Blocksize from the CQI report All the big operators use a

so called Proportional Fair Scheduler in NodeB: The higher the Blocksize is (indicated by CQI), the higher the priority of such an UE to get served

by the NodeB The correction imposed by the Ack/Nack ratio is most of the time not so effective implemented so a cheating UEs might be still served more 2ms TTI Intervals than a correct performing/ CQI reporting UE

Trang 40

Trang 41

41

Cell Breathing due to HSDPA

• Scanner CPICH RSCP versus CPICH Ec/No with different CPICH power (27.5 dBm or 29.5 dBm) and different HSDPA MPO (6 or 7.5)

• MPO = Measurement Power Offset sent to UE via e.g Radio Bearer Setup in order to inform UE about the

“imaginary HS-DSCH Power” to be used for CQI reporting.

Trang 42

CQI distribution

0.2000.3000.4000.5000.6000.7000.8000.9001.000

CDF (27.5|15)CDF (27.5|12)CDF (29.5|12)CDF (29.5|15)

CPICH variation versus MPO variation

Trang 43

43

CPICH variation versus MPO variation

It is shown that increasing the CPICH power leads to better CQI when e.g comparing 27.5|12 to 29.5|12 showing lower CQI values

But Session Throughput shows an improvement

Trang 44

Session Throughput versus CPICH Power and MPO

Trang 45

45

Session Throughput versus CPICH Power and MPO

Trang 46

Application Throughput vs CPICH Power and MPO

Trang 47

47

Application Throughput vs CPICH Power and MPO

Trang 48

CQI versus Ec/No

Trang 49

49

CQI versus Ec/No

Trang 50

Can 15 HS-PDSCH’s be really allocated? Considering Downlink HS-SCCH and E-XXCH’s

Trang 51

51

Can 15 HS-PDSCH’s Codes be really allocated?

Please add the Channel Names to the marked / occupied spreading codes!

Trang 52

Solutions to the Problem of Code Shortage

o F-DPCH allows to support up to 10 users with 1 x SF256

Trang 53

53

Solutions to the Problem of Code Shortage

Please be advised the 2nd frequency does not solve the issue of code shortage for the A-DCH (associated DCH) and HS-SCCH, however code shortage between Rel 99 and downlink HS-XXXCH’s could be partially mitigated In order to optimize the OVSF codes in a flexible manner a flexible/dynamic code assignment strategy could be introduced on networks with only one frequency layer

Introduce 2nd UMTS Frequency

The commonly wide spread 1+1+1 (3 sectorized NodeB) configuration can be expanded to 2+2+2 where on both frequency layers 3 concentric sectors are deployed The handover among the concentric cells can be done blindly without the need for compressed mode (CM) Of course the addition of another frequency layer is only justified in hot spot areas and must go along also with the upgrade of Iub capacity

• F1 used for Idle Mode and Rel 99; F2 is HSDPA preferred Frequency Layer

To enforce camping of all UE’s being in RRC Idle state on F1, there are two possibilities First possibility is to work with negative Q-Offset-S-N for RSCP and CPICH_Ec/No on F2 In order not to affect the reselection between cells on F1, the parameter Q-Hyst-S should be not used for this The alternative

CPICH-is to work with hierarchical cell structure The important F1 layer gets a high priority value assigned and the least important layer to camp on (F2) gets a low priority value assigned Then it has to be ensured that the H-criterion on F1 is always fulfilled and never fulfilled on F2 so the UE is disregarding the cells on F2 from HCS reselection

• Possible Load Balancing between F1 and F2  IFHO

F-DPCH in Rel 6

• F-DPCH allows to support up to 10 users with 1 x SF256 but is only available in Rel 7 (as in Rel 6 it is quasi optional)

• Requires DCCH to be mapped on HS-DSCH

Định dạng
Số trang	116
Dung lượng	2,21 MB