Rough Guide to 3G and HSPA 2009

2

Contents

1. INTRODUCTION 5

2. GENERAL CONCEPTS OF 3G/UMTS/WCDMA 5

2.1 3G : General Information 5

2.2 UMTS Network 5

2.3 Why do we need 3G? Is 2G not enough? 6

2.4 What is the main difference between 3G and 2G? 7

2.5 Why does 3G have less coverage compared to GSM900? 7

2.6 Why is WCDMA called Wideband CDMA? 7

2.7 What are the frequency bands used in 3G? 7

2.8 QOS Classes in 3G 8

2.9 What are the main services available (and used) in 3G/HSPA as of October 2009? 8

2.10 Main difference in performance between R99 Packet and HSDPA 8

3. TECHNICAL CONCEPTS OF 3G/UMTS/WCDMA 9

3.1 Noise Floor 9

3.2 Pilot 9

3.3 Ec/No, RSCP 9

3.4 Codes 10

What is the difference between Scrambling, Spreading and Channelization Codes? 10

3.5 Scrambling Codes 10

3.6 Spreading Codes 11

3.7 Spreading Factor 12

3.8 Spreading and Processing Gain : What do they mean for us? 13

3.9 Soft & Softer Handover 14

3

3.10 Power Control 15

3.11 Achievable Speeds in 3G 17

3.12 What factors affect the data rates available to a user? 18

3.13 What are the main issues in a real 3G network? 18

3.14 What is the difference between RAB and RB? 20

4. HSDPA 20

4.1 HSDPA – Techniques 21

4.2 HSDPA Channel Structure 21

4.3 Advantages of HSDPA over R99 22

4.4 What is the maximum possible speed in HSDPA? 23

4.5 Why is CQI important? 24

4.6 Limiting factors of HSDPA 25

5. EUL 25

5.1 EUL – Techniques 25

5.2 EUL - Channels 26

5.3 Achievable Speeds in EUL 26

6. HSDPA & EUL 27

6.1 Resource Utilization in HSDPA and EUL 27

6.2 Difference between HSDPA and EUL 28

7. KPIS 28

8. CAPACITY MANAGEMENT 28

9. NETWORK ELEMENTS UTILIZATION 29

10. INTER-RAT & INTER-FREQUENCY HANDOVERS 31

4

10.1 Inter-RAT Handovers 31

10.2 Inter-Frequency Handovers 32

10.3 Compressed Mode 32

11. WHAT NEXT AFTER HSPA? 33

11.1 HSPA+ 33

11.2 MIMO 34

11.3 Dual Carrier HSPA 34

11.4 Continuous Packet Connectivity 35

12. APPENDIX 37

12.1 UE Categories 37

12.2 Modulation Schemes 38

12.3 SIB List 39

12.4 UTRAN Protocols 39

Acknowledgements & References 40

5

Rough Guide to 3G and HSPA

1. Introduction

This Rough Guide has been written with the objective of aiding those, who already have

some experience with 3G. Prior knowledge will be helpful for deeper understanding of

the material presented in this guide.

Please note that only WCDMA is considered in this guide and for 2G, only GSM is

considered. Most of the topics covered are Radio related. Core Network details are not

explained.

2. General Concepts of 3G/UMTS/WCDMA

2.1 3G : General Information

UMTS – Universal Mobile Telecommunications System

Provides mainly Speech, Video, R99 data and HS services

3GPP Releases

Rel 99 3G UMTS

Rel 5 HSDPA

Rel 6 EUL

Rel 7 HSPA +

Rel 8 LTE, All IP network (SAE)

Rel 9 SAES Enhancements, WiMax and LTE/UMTS Interoperability

Rel 10 LTE advanced

2.2 UMTS Network

UMTS can be considered as an evolution of GSM. While UMTS has its own radio access network

known as UTRAN (UMTS Terrestrial Radio Access Network ), usually UMTS and GSM/EDGE

have a shared Core Network.

Generally UMTS networks are built up on existing GSM networks and both networks co-exist.

UMTS networks in general have lesser coverage due to the fact that most of them operate at

higher frequency bands. This is not a big issue as UMTS-GSM handover is possible.

6

Please note that the network below has a common core network for both 3G and 2G.

Fig 1: UMTS/GSM Network

2.3 Why do we need 3G? Is 2G not enough?

3G gives much higher data rates compared to 2G. 2G was mainly designed keeping in

mind the requirements for Speech traffic. 3G has been developed mainly to cater to data

services, in addition to Speech traffic. Multiplexing of services with different QOS

requirements on a single connection is possible with 3G.

7

2.4 What is the main difference between 3G and 2G?

WCDMA GSM

Carrier Bandwidth 5MHz 200kHz

Frequency Re-use Factor 1 1-18

Frequency Diversity Multipath diversity with

rake receivers achieved

with 5MHz bandwidth

Frequency Hopping

Packet Data Load based Scheduling Time Slot based Scheduling

with GPRS

Power Control Frequency 1500Hz 2Hz or lower

2.5 Why does 3G have less coverage compared to GSM900?

GSM900 works at a lower frequency band than 3G, which usually works at the 2GHz

band. Lower frequency signals are attenuated less, which gives them greater propagation

capability.

2.6 Why is WCDMA called Wideband CDMA?

WCDMA has a higher bandwidth of 5 MHz compared to IS-95(cdmaOne), which has

only 1.25MHz.

2.7 What are the frequency bands used in 3G?

FDD – Frequency Division Duplexing is mainly used for UMTS. Hence, for uplink and

downlink, we have different frequency bands.

UL – Uplink (mobile to base station) 1920-1980 MHz

DL - Downlink (base station to mobile) 2110-2170 MHz

Point to remember: Generally, operators are given 5MHz Carriers and can have one or

more carriers depending on the operator requirements as well as frequency band

availability.

8

2.8 QOS Classes in 3G

CSIB – Conversational, Streaming, Interactive, Background

Traffic

Class

Conversational

(Real Time)

Streaming

(Real Time)

Interactive

(Best Effort)

Background

(Best Effort)

Basic

Features

- Preserve time

relation (variation)

between

information

entities of the

stream

- Preserve

time relation

(variation)

between

information

entities of the

stream

- Request

response pattern

-Destination is not

expecting the data

within a certain time

- Conversational

pattern (stringent

and low delay )

-Preserve payload

content

-Preserve payload

content

Example of

the

application

voice streaming

video

web browsing emails

2.9 What are the main services available (and used) in 3G/HSPA as of

October 2009?

Service

CS12 – Speech Service with 12.2 kbps dedicated channel

CS64 – Video Telephony with 64kbps dedicated channel

PS64 - Packet Switching with 64kbps dedicated channel

PS128 – Packet Switching with 128kbps dedicated channel

PS384 – Packet Switching with 384kbps dedicated channel

HSDPA - High Speed Downlink Packet Access – shared channel

EUL – Enhanced Uplink

2.10 Main difference in performance between R99 Packet and HSDPA

- R99 Packet service requires dedicated channels whereas HSDPA users have a

shared channel

- Speeds of HSDPA are much higher compared to 3G(R99). In real networks, an

average HS subscriber gets around 5-8 times throughput, compared to an R99

data user. We can easily say that an average HS user can get between 1100kbps

to 2000kbps..whereas an average R99 user can get around 250- 280kbps.

9

Ofcourse, all these values depend on the configuration of the network. For

example speeds of about 6Mbps was reported during random field tests in one of

the networks in Kuwait. Introduction of higher capacity UEs as well as higher

modulation schemes will further increase the HS throughputs.

3. Technical Concepts of 3G/UMTS/WCDMA

3.1 Noise Floor

Main idea of WCDMA is to spread the User signal over the whole band, pushing the signal under

the noise floor. Only a receiver with knowledge of the correct PN (pseudorandom noise)

sequence can detect the signal. Any other receiver will see only the noise. Hence the security is

high.

3.2 Pilot

Pilot coverage decides the coverage boundary for a particular site. Proper Pilot power planning

is very important. Too-weak pilot will lead to coverage holes, whereas too-strong pilots will lead

to overshooting and interference.

Point to remember : In real networks Pilot power normally varies between 27-33dBm. Usually in

urban areas Pilots have values between 27-30dBm.

3.3 Ec/No, RSCP

Ec/No signifies the level difference between received pilot signal and the overall noise floor.

No is the noise floor, which signifies all the signals (useful and interfering) present at the

receiver side.

For example: A value of Ec/No= -8dB tells us that the spread signal is 8 dB below the noise floor

Higher the Ec/No value, the better it is….

Please note that the existing receivers have rake receiver functionality which enables them to

decode multiple pilots and use them accordingly based on their strength.

For example:

If there are 3 pilots present….the mobile receiver will compare Ec1/No, Ec2/No and Ec3/No and

decide which pilot will be the best server. More details are provided in Handover and Pilot

Pollution Sections.

RSCP : Received Signal Code Power is the received power on one code after despreading,

defined on the pilot symbols.

Ec/No = RSCP/RSSI

3.4 Codes

What is the difference between Scrambling, Spreading and Channelization Codes?

Spreading Code = Channelization Code

Usage

Length

No: of Codes

Fig 2: Usage of Scrambling Codes and

3.5 Scrambling Codes

Downlink Scrambling Codes

3 types of scrambling codes are available in DL: primary, secondary and alternative.

Downlink primary scrambling codes are used for cell separation. One primary scrambling

code, is allocated for each cell. Secondary scrambling codes

scrambling codes can be used in compressed mode.

10

What is the difference between Scrambling, Spreading and Channelization Codes?

Spreading Code = Channelization Code

Channelization Code/

Spreading Code

Scrambling Code

DL – Separation of DL

dedicated user channels

UL – Separation of Data

and Control channels from

the same terminal

DL – Separation of Cells

(Sectors)

UL – Separation of UEs

Variable Fixed

Depends on SF DL – 512

UL – Unlimited

Fig 2: Usage of Scrambling Codes and Channelization codes

Scrambling Codes

odes



code, is allocated for each cell. Secondary scrambling codes are not used. Alternative

scrambling codes can be used in compressed mode.

Scrambling Code

Separation of Cells

Separation of UEs

512

Unlimited (Millions)



re not used. Alternative

11

How many Scrambling Codes are available in DL? – 512

Uplink Scrambling Codes

2 types of scrambling codes are available in UL : long and short. Only the long ones are

used. Uplink scrambling codes are used for separating the different UEs in the same cell.

RNC allocates the code.

3.6 Spreading Codes

Downlink Spreading Codes (Channelization Codes)

DL spreading codes differentiate the dedicated user connections/channels within one cell.

Ideally they are orthogonal to each other, though due to multipath propagation, some

orthogonality might be lost.

Channelization codes are managed with the help of a code-tree. Basic rule is that codes

are orthogonal, if they do not descend from an already used code. If a code is used, then

all the codes below and above on the same branch are unavailable for service. Resource

manager keeps track of the codes allocated so that orthogonality of the code tree is

preserved.

Fig 3: Code Tree for orthogonally spreading codes

Example : Code management with the help of the code tree

If code C2(0) in the Tree of orthogonal spreading codes (in the figure above) is allocated,

then:

12

All codes below it in the same branch become unavailable, starting with C3(0) and C3(1),

then, on the next level, C4(0), C4(1), C4(2) and C4(4), and so on.

All codes above it in the same branch to root become unavailable, that is, C1(0) and

C0(0) cannot be assigned to any user .

Spreading codes of some channels (mainly Pilot and P-CCPCH) are fixed. Spreading

codes for all other downlink physical channels are allocated by the resource manager.

3.7 Spreading Factor

Higher the bit rate of the data service, lesser the spreading factor.

Service Spreading Factor

Half Rate – AMR 256

Speech 128

CS64 32

PS64 32

PS128 16

PS384 8

HSDPA 16

Table giving DL spreading factors for different services

Points to remember :

Usually UL spreading factor for a service is half the value of that in the DL (when the

RAB bearer rates are the same in both UL and DL).

For example: DL SF for speech(AMR12.2) service is 128, where as in UL, it has a SF of

64.

Why should we avoid pulsed transmission in the UL?

During the silent periods, only information for link maintenance purposes are needed in

UL direction. A typical example is Power Control commands at 1.5KHz which can

interfere with the telephony voice frequency band.

To avoid audible interference to audio devices in UL, data and control channels are not

time multiplexed in WCDMA. Continuous transmission is achieved with I/Q code

multiplexing or by using parallel control and data channels.

13

3.8 Spreading and Processing Gain : What do they mean for us?

In WCDMA, the baseband signal is spread using a spreading code…

By spreading,

1) Baseband signal is spread over the entire spectrum (3.84MHz), with help of a

spreading code

2) Overall noise floor rises, but the baseband signal is hidden below the noise floor

and hence difficult to detect

3) Effect of Narrow-band interference is reduced, since only a small part of the

signal will be affected and data can be recovered with effective techniques

4) Effect of Multipath fading is also reduced

5) Higher the bit rate of the service, lower the SF (Speech SF = 128, PS384 SF= 8)

and lower the processing gain

Despreading is done at the RX side.

By despreading

1) We get the baseband signal back and gain from the processing gain.

Point to remember : Spreading and despreading can be considered as a process of

pushing the actual baseband signal below the noise floor and then retrieving it.

Processing Gain = 10 log (chiprate / bit rate)

To get a good service, the requirement is

Rx Sig Level + Processing Gain > Eb/No

Eg: PG for speech = 10 log ( 3.48Mcps / 12.2Mbps) = 25dB

Eb/No requirement for speech = 5dB (for good service)

Rx sig level = 5 – 25 = -20dB (which implies that even if the received signal is 20 dB

below the noise floor, the WCDMA receiver can detect the speech signal).

In GSM, the C/I requirement is about 9-12dB. This directly gives an advantage of about

20-25 dB for WCDMA.

14

3.9 Soft & Softer Handover

Soft handover is the condition in which the UE is connected to more than one NodeB at

the same time. While in connected mode, UE continuously measures the neighbouring

signals and compares the measurement results with specific handover thresholds set by

the operator. When the threshold is exceeded, UE sends a measurement report to the

RNC. RNC decides if the SHO should take place.

Soft Handover is also called MEHO – Mobile Evaluated Handover

Active Set : Set of cells which are in soft handover.

There are 3 types of Soft Handover

1) Handover between sectors in the same site (Softer Handover)

2) Intra-RNC SHO

3) Inter-RNC SHO

Majority of Soft handovers are usually Intra-RNC SHO.

Advantages of SHO:

1) Seamless handover without disconnection of RAB

2) Macro diversity gain..achieved in both UL and DL due to the combining of

signals from different cells

3) Better performance in areas where a single cell is not strong enough

Disadvantages of SHO

1) Increased consumption of radio resource as one UE in SHO, will use more than

one radio link at a time

Point to remember : SHO is kept in mind during the initial planning and ideally an

overhead of 30-40% is assumed.

Events

Mobile sends Measurement Report to RNC, when certain thresholds are crossed. For

SHO, it is important to know Event 1a, 1b, 1c and 1d.

Event 1a : addition of a new cell to the Active Set

Event 1b: deletion of a cell from the Active Set

15

Event 1c: replacement of weaker cell in Active Set by another stronger cell (not in the

Active Set)

Event 1d : replacement of best cell in Active Set by a stronger cell (from Active Set,

Monitored Set or Detected Set)

3.10 Power Control

Main purpose of Power control mechanism is to

1) maintain the quality of service

2) minimize the transmitted power in both UL and DL

In WCDMA, downlink transmitted power determines the interference and hence the air

interface capacity. So it is important to avoid excessive transmission in DL.

A single UE can create problems with excessive transmission in the UL. Power control

mechanism takes care of this.

Power control is done on both common and dedicated channels. Power control in

common channels ensure that sufficient coverage is available to setup UE-originating and

UE-terminating calls as well as data transfer on RACH and FACH. Power control in

dedicated channels ensure that connection quality is maintained in terms of BLER (Block

Error Rate)

There are mainly 3 types of power control.

1) Open loop power control

2) Closed loop power control (Fast Power Control)

3) Outer loop power control

Open Loop Power Control – When the UE accesses the system it first sends a preamble

and waits for a response from the NodeB. If this expected response, AI (Acquisition

Indicator), is not obtained, the UE transmits another preamble with slightly higher power.

The process of ramping up preamble power continues till either a response is obtained

from the NodeB or the allowed number of preamble steps are used. When the maximum

number of steps in a preamble cycle is used, another preamble cycle is started, which in

turn is limited by a maximum number of preamble cycles set by the operator.

16

Point to remember : Three parameters are controlled by the operator in the case of

Open loop power control ( preamble step, number of preamble steps in a preamble cycle

and the number of preamble cycles).

Closed Loop Power Control (Fast Power Control) – setting of TX power based on SIR

target (in NodeB). Done with a frequency of 1500Hz.

UE and BTS continuously compare the actual SIR of the received signal with a target

SIR. Based on the comparison, BTS/UE tells the UE/BTS to either increase or decrease

the transmission power.

Outer Loop Power Control – setting of SIR target based on Frame quality (in RNC).

Outer loop power control aims to provide the required quality in both UL and DL, by

monitoring the BLER of the received signal. Based on the BLER, the SIR target for the

Fast Power Control is increased or decreased.

For example: if the received BLER is not meeting the expected quality, then the SIR

target is increased and if the received BLER is higher than the expected quality, then the

SIR target is decreased.

Fig 4: Power Control Mechanism

17

3.11 Achievable Speeds in 3G

How do we get the speed of 2Mbps for R99 ?

Data Rate = Chip Rate / Spreading Factor

In R99, 3 codes with SF4 gives the max possible data rate.

Data Rate for one SF4 code = (3.84Mcps / 4 ) = 960ksps

ksps = kilo symbols per second

Since R99 uses only QPSK, 1 symbol = 2 bits

Hence, Data Rate = 480ksps = 960 * 2 bits = 1920kbps = 1.92Mbps

For 3 SF4 codes, data rate = 3 * 1.92Mbps = 5.76 Mbps

BUT, Data Rate = Net User Data + Channel Code Redundancy + Control Data

After taking out Channel Code Redundancy and Control data, Net User Data == 2Mbps

(the above value is for one sector with one carrier)

Point to remember : The code rate used in R99 is 1/3

Why do we have 384kbps as the max possible data for a single R99 Packet user in

3G?

Currently PS384 is the highest RAB available in DL for R99 Packet users.

SF for PS384 = 8

Data Rate for one SF4 code = (3.84Mcps / 8 ) = 480ksps

ksps = kilo symbols per second

Since R99 uses only QPSK, 1 symbol = 2 bits

Hence, Data Rate = 480ksps = 480 * 2 bits = 960kbps

BUT, Data Rate = Net User Data + Channel Code Redundancy + Control Data

After taking out Channel Code Redundancy and Control data,

Net User Data == 384kbps (max possible)

18

How do you calculate maximum possible speed in HSDPA?

Using the formula (data rate = chiprate/spreading factor),

1 QPSK code at SF16 = 480kbps

1 16-QAM code at SF16 = 960kbps

1 64-QAM code at SF16 = 1440kbps

For HSDPA after applying ¾ coding rate

1QPSK Code = 360kbps

1 16-QAM Code = 720kbps

1 64-QAM Code = 1080kbps

10 codes with 16QAM = 720 * 10 = 7200 kbps = 7.2Mbps

15 codes with 16QAM = 720 *15 = 10.8Mbps (max per cell or sector)

15 codes with 64QAM = 1080 *15 = 16.2Mbps (max per cell or sector)

Theoretical max of HSDPA with one carrier = 15 Codes * 1440kbps = 21.6Mbps for a

single carrier (assuming coding rate of 1, which is impossible in actual conditions)

3.12 What factors affect the data rates available to a user?

- User position in the cell

- Interference from other users and neighbouring cells

- Number of subscribers accessing the same cell

- Speed of the customer (if he is mobile)

3.13 What are the main issues in a real 3G network?

Pilot Pollution (Improper Pilot Power Planning)

Main objective of Pilot planning is to have a dominant signal at a given place. In

practice, this is difficult to achieve. 2 to 3 strong signals are still ok, since Soft handover

will manage the situation. But if you have more signals coming at the same place with

more-or-less equal strength, then the UE gets confused and cannot correctly decode the

19

signals due to low Useful Signal-to-Interference ratio (Ec/No) and hence the call gets

dropped.

Points to remember:

- Strive to have ONE dominant Pilot signal at a given place.

- In Ec/No, Ec is good (as long as there is no pilot pollution). No is interference.

Power, Tilt and Azimuths optimization mainly used to avoid pilot pollution.

Missing Neighbour Definitions

This can be observed on Field with Tems or any monitoring tool (as Detected Set).

When the UE is getting a strong signal which is not defined as a neighbour to the existing

cells in the Active Set, the new signal adds to the interference. Soft handover does not

take place and as a result Ec/No degrades. As a result the call drops when the new signal

is about 15dB higher than the cells in the Active set.

Improper UEs

Though not observed on a wide scale, this can be a problem. A malfunctioning UE can

cause many problems like

- Demanding too much power from the base station

- In-efficient channel switching

- Excessive transmission of power in UL

IRAT HO Parameter Definition

- Improper definitions can lead to un-necessary handover between 3G and 2g. This

can be a problem especially for indoor customers using HSDPA or data services.

Overall throughput of the data user will be affected due to unnecessary

handovers/cell changes.

Cell Breathing

With more and more users coming into a cell, the actual power available for services is

lesser than the power available in an empty cell. So the overall coverage of the cell

shrinks.

Cell breathing is more of a planning issue and has to be considered at the planning stage

itself. Proper handover regions should be planned, to avoid any coverage gaps.

20

Prioritizing Neighbours

More efficient handover can be achieved by proper prioritization of neighbours. It is

possible to give higher priority for some cells than to other cells, so as to make sure that

chances for a handover is higher between certain cells. This parameter can also be used to

avoid handover in certain locations between certain cells to some extent. Improper

allocation of priority can lead to bad handover decisions.

Low Sites

One major mistake RF planners did in the beginning was to install low sites for UMTS,

thinking mistakenly that since interference is to be avoided in UMTS, it is better to have

low sites with lower coverage areas.

In actual practice, low sites are generally problematic as they overshoot and contribute to

Pilot Pollution. Down-tilting of low sites can lead to coverage holes…(we should keep in

mind that down-tilting is an efficient way of reducing overshooting).

3.14 What is the difference between RAB and RB?

RAB – Radio Access Bearer – Link between UE and Core (Radio + Iub + Iu)

RB – Radio Bearer – Link between UE and RNC (Radio + Iub)

4. HSDPA

HSDPA has a fixed spreading factor of 16. Multiple codes can be reserved for HSDPA at

this SF level and depending on the number of codes available, the speed varies. Details

are given in the section What is the maximum possible speed in HSDPA?

Generally operators reserve 5 or 10 codes per carrier (out of the 15 available) for HSDPA

service, which implies that these codes are not available for other R99 services like

Speech, CS64 and PS. There are different ways of code allocation for HSDPA, and this

varies from vendor to vendor.

When there is a shortage of codes, due to higher traffic, the operators can go for a second

carrier. Operator can decide how to distribute HS and R99 traffic in different carriers. It

is also possible to have a carrier fully allocated to HS, which implies that 15 codes will be

available solely for HS and no other services will be possible in that carrier.

Point to remember: Greater the number of codes you reserve for HS, lesser the

resources available for R99 services.

21

4.1 HSDPA – Techniques

- Shared Channel Transmission (enabling one user to have more than one code)

- Shorter TTI (2ms)

- Higher Modulation Technique (16QAM )

- Hybrid ARQ Retransmission

- Faster Scheduling based on Radio conditions

- Better Scheduling Techniques(code rate, modulation technique)

Fig 5: HSDPA Techniques

4.2 HSDPA Channel Structure

In addition to the new downlink shared channel HS-DSCH, some control channels are

also required for HSDPA. Mainly they are HS-SCCH and HS-DPCCH.

22

Channel Direction Content

HS – DSCH DL User Data

HS-SCCH DL

Control information to address UEs and

information for decoding the transport block.

UEs can see upto 4 HS-SCCH

HS-DPCCH UL ACK/NAK, CQI

A-DCH UL and DL

SRB (Control signaling: RRC and NAS) in DL

SRB and User data in UL

Table giving HSDPA Channels and related R99 Channels

Fig 6: HSDPA Channels

4.3 Advantages of HSDPA over R99

- Faster Retransmission (due to control in NodeB), leading to much lower RTT

23

Fig 7: Retransmission methods in R99 and HSDPA

As seen in the picture above, in case of R99, retransmission decision is taken in the RNC

(RLC layer), whereas in HSDPA, the retransmission decision is taken in NodeB (MAC-

hs layer). This leads to a great reduction in overall RTT (Round Trip Time)

- More codes used by a single user, hence higher throughputs

- Shorter TTIs, hence better response time and RTT

- 16QAM is not used in R99

- Soft Combining of re-transmission

Point to remember : There are mainly 2 types of scheduling in HSDPA – Round Robin

and Proportional Fair. Round Robin scheduling, allocates resources to every user in a

round robin manner regardless of the radio conditions, the users are in.

Proportional fair scheduling takes into account, the radio conditions also and tries to

improve the overall cell throughput by giving slightly higher preference to users in better

radio conditions.

In actual testing conditions, not much difference in overall cell throughput was observed

between the two scheduling techniques and since Round Robin scheduling came free of

charge, with most vendors, it was the preferred scheduler.

4.4 What is the maximum possible speed in HSDPA?

Check section 3.11

24

4.5 Why is CQI important?

CQI is the feedback which the system receives from the UE and it mainly indicates the

radio condition of the UE. Depending on the CQI values, NodeB scheduler allocates

resources to the UE.

Higher the CQI, better the network. An average CQI value of about 22 and above,

indicates a reasonably good network. CQI values less than 17, indicates a low quality

network and optimization is required.

Fig 8: Overall picture of how radio conditions affect HS Throughput and Power

Requirement

The figure above summarizes the tests conducted for a HS user in both bad and good

radio conditions.

In all the 3 graphs above, the left side represents a user in bad radio condition and the

right side represents a user in a good radio condition.

25

A user in a very bad radio condition reports an average CQI of 14, whereas the same user

in excellent radio conditions reported an average CQI of 26. In bad radio conditions, the

user consumed much more power, though he got almost the same throughput as the user

in good radio condition.


- It is very important to have a HS user in good radio conditions, since higher

throughputs can be achieved with lesser transmitted power, leading to increased

capacity for the system.

- For higher order modulations to work, CQI values should be high.

4.6 Limiting factors of HSDPA

Channelization Codes, Modulation Scheme, Channel Elements, Power, Simultaneous

users, UE Category

5. EUL

Main idea of EUL is to effectively use the interference headroom available in the uplink.

Currently achievable peak individual user throughputs are around 1.4 to 2Mbps in EUL

where as it is 384 kbps with R99. Overall, EUL should give higher throughputs and

greater capacity than R99.

For example: Assume that 4 users want to upload big amounts of data…(let us say,

movies)

If EUL is used, we need to have upto 32 channel elements.

With R99, assuming that they are using 384 RAB in uplink, total CE requirement = 4 * 16

= 64 CEs, since each 384UL RAB requires 16 CEs. Hence in this case, 32 CEs are saved by

using EUL.

5.1 EUL – Techniques

- Hybrid ARQ with Soft Combining

- Fast Channel Dependent Scheduling

- Multi-code Transmission

26

- Power Control

- Soft Handover

5.2 EUL - Channels

Fig 9: EUL Channel Structure

5.3 Achievable Speeds in EUL

Case 1: Assuming that the UE category available can support only upto 2 SF4,

Data rate per channel = 3.84/4 = 0.96Msps

1symbol = 1bit since BPSK is used in EUL

So, Data rate per channel = 0.96Mbps

Since 2 channels (2 SF4) are possible, maximum rate = 0.96 * 2 = 1.92Mbps

27

After taking out all FEC, CRC, MAC-headers and L3 signaling, data rate at RLC level =

1.376Mbps

Data rate at L1 (transport block level) = 1.46Mbps

Point to remember: Above figure is the total bit rate achievable with EUL in one cell,

when the maximum possible configuration is 2 * SF4 channels (and only BPSK is

available). If we have SF2 available, we will be getting higher UL throughputs.

Case 2: Assuming that the maximum channel capacity of 2SF2 + 2SF4 is available,

Data rate per SF2 channel = 3.84/2 = 1.92Mbps

Data rate per SF4 channel = 3.84/4 = 0.96Mbps

Total data rate = (2*1.92) + (2*0.96) = 5.76Mbps

Realistically with ¾ coding Max EUL Data Rate = 5.76 * ¾ = 4.32Mbps

Why is it NOT beneficial to have 16-QAM in EUL ?

Since UL is interference limited:

- It is better not to have power-inefficient higher-order modulation schemes

- Cost effective design of UE power amplifier is possible with lower-order

modulation schemes, since they have lesser PAR (Peak to Average Ratio) which

in turn lead to lesser Electromagnetic Interference (EMI) generated by the UE.

6. HSDPA & EUL

6.1 Resource Utilization in HSDPA and EUL

In HSDPA, the shared resource is DL Transmission Power, Channelization Codes and

Channel Elements

In EUL, the shared resource is UL interference and Channel Elements

28

6.2 Difference between HSDPA and EUL

HSDPA EUL

Spreading Factor Fixed = 16 Variable from 256-2

Soft Handover No (only A-DCH in SHO) Yes

Power Control No (Check RPA ) Yes

Modulation Scheme 16QAM & QPSK BPSK

Link Adaptation Rate Control Rate & Power Control

7. KPIs

Accessibility – both RRC and RAB phases considered

Mobility – Soft/softer handover (30-40%), IRAT handover

Retainability – Mainly Voice and HS drops. Currently the practice is to monitor

Minutes/Drop

Traffic – Erlangs for Speech/CS64 services, Data Volume for PS/HS services

Integrity – CQI for HS, BLER for R99 (if needed)

8. Capacity Management

Main purpose of capacity management is to provide sufficient QOS and coverage for

users. Admission Control and Congestion Control are the two main mechanisms used

for capacity management.

Admission Control ensures that a new user will be connected only if there are enough

resources available for him.

Congestion Control tries to keep the usage of the system within reasonable limits. For

example, if there are 3 PS384 users in a cell and one of them moves into a bad signal

area and requires more power to maintain the data rate, the system checks the used DL

transmitted power. If it has crossed a threshold, the user is downgraded from PS384 to

PS128 or to PS64, depending on the available power. By doing this, channel element

29

utilization is also reduced from 16 to 8(PS128) or 4(PS64), which effectively means that

more speech users can be accommodated.

Congestion control is based on 3 parameters

- Downlink overload (when the downlink transmitted power is exceeding some

threshold for a set period of time)

- Uplink overload (when RTWP –received total wideband power exceeds a


- DL HSDPA Overload (when total power, which includes HS power exceeds a


Point to remember: Generally congestion control comes into play before admission

control. Speech and video call users have higher priority over HS and PS users.

Admission for speech and video calls have strict criteria. Speech/video call users are

connected only if dedicated resources are available for them. Data services have easier

admission policies. EUL, especially has a very lenient admission policy, as connected

users are allocated capacity based on availability and do not use other system resources.

Resources Monitored for Load Control:

Parameters monitored and used for capacity management are

- Downlink Transmitted Carrier Power

- Downlink Channelization Codes

- Uplink Received Total Wideband Power

- Interference

- No: of radio links in compressed mode

- No: of serving HS connections

- No: of serving connections

- No: of non-serving connections

- Node B Hardware Utilization (mainly Channel Elements)

9. Network Elements Utilization

This section gives a rough idea of the parameters to be monitored to calculate the

utilization of different network elements

30

- RNC: Total Traffic, Simultaneous number of HS users, ATM connectivity, total

number of NodeBs which can be connected to one RNC

- NodeB: Channel Elements, Code Tree, DL Transmit Power

Channel Elements are one of the major hardware resource in NodeB to be planned

and monitored carefully. Different services have different requirement of CEs. In

most of the vendors, there is a fixed allocation of CEs for HS services. R99

services use CE when required. The tables below give sample CE requirements

for different services. HS requirements are not included in these tables, as they

are different for different vendors.

Spreading

Factor Bearer Data Rate (kbps) Channel Element Requirement

128 AMR 12.2 1

32

32 64 2

16 128 4

8 384 8

Sample Table for DL Channel Element Requirement

Spreading Factor Bearer Data Rate (kbps) Channel Element Requirement

64 AMR 12.2 1

32 32 2

16 64 4

8 128 8

4 384 16

Sample Table for UL Channel Element Requirement

Channelization Codes : With the introduction of HSPA, channelization codes

have become a major limiting factor in terms of resource utilization. Since atleast

5 to 10 codes are reserved for HS, only the remaining codes are available for R99

services like Speech, CS64 and R99 Packet. Generally, vendors go for a second

carrier in case of code congestion.

31

DL Transmit Power : In WCDMA, downlink is power limited, assuming that we

have enough resources like CEs and channelization codes. Hence it is important

to monitor the DL power consumption. We can say that Power == Capacity. We

have to keep in mind that Packet users require more power compared to Speech

users.

Point to remember : Channel Element is a NodeB level resource. Channelization

code is a cell level resource.

- Iub: Proper planning should be done for VP/VC. Different methods are

available. One of the main limitations if you have AAL2 switching is the number

of CIDs available per VC.

For example: If you have one STM1 link with 155Mbps, you can divide it into

any number of VCs as you need.

Case 1: If you assign just one VC, you have a total of 248 CIDs available…

Case 2: If you assign 10 VCs, you have 248 * 10 = 2480 CIDs available….

Assuming only voice users in the network, since each Voice user needs 2 CIDs,

Total possible subscribers in case 1 = 248 / 2 = 124 speech users

Total possible subscribers in case 2 = 2480 / 2 = 1240 speech users

So in case1, even when there was more than enough capacity (155Mbps), we have

a limitation of 128 speech users due to the definition of VC.

In case2, with the same capacity available as in Case1, we have 10 times more

speech users.

Please keep in mind that the each HS user require 3 CIDs. Further, separate CIDs

are needed for Control purpose also.

10. Inter-RAT & Inter-Frequency Handovers

10.1 Inter-RAT Handovers (event 3a)

Required since 3G coverage is generally less compared to 2G.

It is important to have proper parameters defined for Inter-RAT handovers (mainly

UMTS-GSM).

Event 2d occurs when the 3G measured quality is below a certain threshold for a certain

period of time and this triggers measurement on IRAT or Inter-Frequency (depending on

32

vendor). Compressed mode measurements on 2G start after event 2d. Once event 2d is

triggered, if the measured quality of 2G is above a certain threshold for a certain period

of time, then event3a occurs. Actual 3G-2G handover is triggered by event 3a.

10.2 Inter-Frequency Handovers (event 2b)

Required when 2 or more frequencies are implemented in a network.

Event 2d occurs when the measured quality is below a certain threshold for a certain

period of time and this triggers measurement on IRAT or Inter-Frequency (depending on

vendor). Compressed mode measurements on the 2nd

frequency start after event 2d.Once

event 2d is triggered, if the measured quality of the 2nd frequency is above a certain

threshold for a certain period of time, then event2b occurs. Actual IF HO is triggered by

event 2b.

Point to remember :

- Event 3a : 3G-2G HO

- Event 2b : Inter-Frequency HO

Event 2f occurs when the measured quality is above a certain threshold for a certain

period of time and this triggers the stopping of IRAT/Inter-Frequency measurements.

Depending on the settings, when event 2d occurs, the system decides if IRAT or IF

handover should take place . In some vendors both are possible.

For example: In Ericsson you have to set either IRAT or IF HO, where as in Nokia it is

possible to have IRAT and IF handovers from the same carrier.

Event 6d occurs when the UL UE Tx power exceeds a certain threshold for a certain

period of time. Event 3a (IRAT HO) or Event 2b(IF HO) follows.

Event6b, occurs when the UL UE Tx power is below a certain threshold for a certain

period of time. All ongoing HO attempts are aborted if DL Quality for both Ec/No and

RSCP are good.

10.3 Compressed Mode

Compressed mode mechanism enables the UE to carry out measurements on another

frequency. Certain idle periods are created in radio frames during which the UE can

perform measurements on other frequencies. No user data is lost as it is compressed in

the time domain using one of the below 2 methods

33

- Halving the spreading factor so that the same amount of data can be sent in half

the time

- Higher layer scheduling in which layer2 restricts the high bit rate TFC (transport

format combinations) so that the user throughput is reduced temporarily


- Currently, compressed mode is not used for HS-DSCH or EUL. It would be

available soon.

- Compressed mode can be used for both UL and DL (depending on UE capability)

- The transmission/reception gap is always 7 slots (out of the total 15 slots in a

frame)

Fig 10: Transmission Gaps created with Compressed Mode

11. What next after HSPA?

11.1 HSPA+

HSPA+ is a natural evolution to HSPA and can be considered as an upgrade to the

existing HSPA system. Many techniques are specified in HSPA+ for improved

performance. They are

- MIMO

- Higher Order Modulation (64QAM)

- Multi-carrier HSPA

- Continuous Packet Connectivity

34

- Enhanced Cell_FACH

- Voice Over HSPA

Below sections will give you a brief idea of some of these features.

11.2 MIMO

Multiple Input Multiple Output involves using multiple antennas at both transmit and

receive side which leads to significant increase in achievable throughputs, without the

necessity for additional bandwidth or transmit power.

Point to remember:

HSPA+ Rel: 7 (MIMO) can theoretically support up to 28Mbps with a single 5MHz

Carrier

HSPA+ Rel: 8 (Higher Order Modulation + MIMO) can theoretically support up to

42Mbps with a single 5MHz carrier

11.3 Dual Carrier HSPA (also known as Dual Cell HSPA)

DC- HSPA aims to increase the available user data rates by merging 2 carriers of 5MHz

each, thus making available up to 10MHz carrier bandwidth for a user.

Higher Bandwidth available to a user = = Higher Throughput for the user

Basic idea of DC-HSPA is to achieve better resource utilization by means of joint

resource allocation and load balancing across the carriers.

Some of the features for DC-HSPA are

- New MAC entity, MAC-ehs which supports HS-DSCH transmission/reception in

more than one cell served by the same Node-B

- New UE categories required (Categories 21 to 24)

- Anchor Carrier : Carrier with all physical channels (as shown below)

- Supplementary Carrier: Carrier with just HS-PDSCH and HS-SCCH

35

Fig 11: DC-HSPA Channel Usage in the Multiplexed Carriers

Advantages of DC-HSPA are

- Higher data rates possible compared to the 5MHz single carrier, since a user can

get all the code and power resources of both carriers in a single TTI

- Improved load sharing due to dynamic statistical multiplexing of users at

connection management level

- Greater frequency selectivity and improved QOS due to joint scheduling. User

can be assigned resources dynamically either on the anchor or on the

supplementary carrier

-

Point to remember : Theoretical DL throughputs achievable with DC-HSPA without

MIMO is around 43.2 Mbps

11.4 Continuous Packet Connectivity

In future, data users are expected to stay connected for long times, even if they are not

doing anything for a majority of the time they are connected. So it will be good to avoid

unnecessary transmissions during these idle periods, so as to avoid interference and

reduce system resource utilization.

CPC consists of two main features UE DTX/DRX and HS-SCCH-less operation.

36

UE DTX (discontinuous transmission from UE) enables the UE to switch off continuous

transmission of DPCCH (Dedicated Physical Control Channel) when there is no

information to be transmitted in the uplink. This leads to

- Reduced battery consumption

- Reduced interference, resulting in increased uplink capacity

UE DRX (discontinuous reception at UE) enables the UE to switch off their receivers,

when there is no data to be received in downlink. This also leads to reduced battery

consumption.

Services like VoIP, require transmission of lots of small packets in DL. This leads to

significant overhead due to the HS-SCCH control channel. One solution to this problem

is to remove HS-SCCH transmission completely for the first HARQ transmission. This

involves blind decoding of up to 4 different formats of HS-DSCH, the DL data channel.

37

12. Appendix

12.1 UE Categories

Knowledge of different categories of UEs available is essential to understand the

achievable throughputs.

Category Max. number of

HS-DSCH codes Modulation

MIMO -

Dual

Carrier

Code rate

required to

achieve max.

data rate

Max. data rate

[Mbit/s]

1 5 QPSK and 16-QAM 0.76 1.2

2 5 QPSK and 16-QAM 0.76 1.2

3 5 QPSK and 16-QAM 0.76 1.8

4 5 QPSK and 16-QAM 0.76 1.8

5 5 QPSK and 16-QAM 0.76 3.6

6 5 QPSK and 16-QAM 0.76 3.6

7 10 QPSK and 16-QAM 0.75 7.2

8 10 QPSK and 16-QAM 0.76 7.2

9 15 QPSK and 16-QAM 0.7 10.1

10 15 QPSK and 16-QAM 0.97 14.4

11 5 QPSK only 0.76 0.9

12 5 QPSK only 0.76 1.8

13 15

QPSK, 16-QAM and 64-

QAM 0.82 17.6

14 15

QPSK, 16-QAM and 64-

QAM 0.98 21.1

15 15 QPSK, 16-QAM MIMO 23.4

16 15 QPSK, 16-QAM MIMO 27.9

19 15 QPSK, 16-QAM MIMO 35.3

20 15 QPSK, 16-QAM, 64-QAM MIMO 42.2

21 15 QPSK, 16-QAM DC 23.4

22 15 QPSK, 16-QAM DC 27.9

23 15 QPSK, 16-QAM, 64-QAM DC 35.3

24 15 QPSK, 16-QAM, 64-QAM DC 42.2

25 15 QPSK, 16-QAM DC + MIMO 46.8

26 15 QPSK, 16-QAM DC + MIMO 55.9

27 15 QPSK, 16-QAM, 64-QAM DC + MIMO 70.6

28 15 QPSK, 16-QAM, 64-QAM DC + MIMO 84.4

Table giving UE categories for HSDPA

38

Table giving UE categories for EUL

12.2 Modulation Schemes

Fig : Constellation diagrams of different modulation schemes

39

12.3 SIB List

System information is broadcast regularly to the UE on the BCCH. It contains

parameters related to Cell Selection, Reselection, Location and routing registration,

Handover, Power Control etc. Any parameter change in the system information is

notified to all UEs in the cell by a paging message or by a system information change

indication message. The table below list the different SIB messages available.

12.4 UTRAN Protocols

RRC : Radio Resource Control

- Handles control plane signaling of Layer3 signaling between UEs and RNC

NBAP : NodeB Application Protocol (Iub)

- Signaling protocol responsible for the control of NodeB by RNC

- NBAP has two parts: C-NBAP and D-NBAP

C-NBAP (Common NBAP) controls the overall functionality of the NodeB

System Information

Blocks Contents

MIB PLMN identity for serving cell, SIB Scheduling Information

SB1 SIB Scheduling Information

SIB1

Paging parameters, Timers and counters in Idle and Connected mode, LA and

RA updating

SIB2 URA identity list

SIB3 Cell selection and reselection parameters

SIB4 Cell selection and reselection parameters. Connected mode only

SIB5 and SIB5bis Paging parameters, Cell and common channel configuration

SIB7 Power control on common channel

SIB11 Measurement management, Cell selection and reselection parameters

SIB12 Measurement management

SIB18 PLMN identity for GSM neighbors listed in SIB11.

40

D-NBAP (Dedicated NBAP) controls radio links specific to UEs

RANAP : Radio Access Network Application Part (Iu)

- For signaling between Core Network( MSC or SGSN) and RNC

RNSAP : Radio Network System Application Part (Iur)

- Signaling protocol responsible for communication between RNCs

Acknowledgements & References

I would like to thank my colleagues at Wataniya Telecom, Kuwait as well as Mobitel,

Slovenia for the support extended to me. I would like to thank specially,

- Naveen Krishnapillai, Wataniya Telecom, Kuwait

- Amol Rajan Pradhan , Wataniya Telecom, Kuwait

- Santosh Tummala , Wataniya Telecom, Kuwait

- Amin Sudhir Vasanth , Wataniya Telecom, Kuwait

- Iztok Saje, Mobitel, Slovenia

Material for this guide has been compiled from

- Author’s experience in 3G from year 2002 with Mobitel, Slovenia and Wataniya

Telecom, Kuwait

- WCDMA for UMTS by Harri Holma and Antti Toskala

- Internet (especially Wikepedia)

- White Paper – Dual Cell HSDPA and its Future Evolution - Nomor Research

GmbH

- Articles from different vendors, especially Ericsson and NSN

Rough Guide to 3G and HSPA 2009

Documents

signaling

round robin

preserve time

power control

fast power

channel code

bearer data

event 2d occurs