VoIP over Wireless. IMA Summer Program on Wireless Communications. IMA Summer Program on Wireless Communications

1 2 2 2 3 3 1 1 2 ₃ 1 2 2 2 3 3 1 1 2 ₃

IMA Summer Program on

Wireless Communications

VoIP over Wireless

Phil Fleming

Network Advanced Technology

Group

Network Business

Motorola, Inc.

Acknowledgements:

Material on

PTT over GPRS

provided by

John Harris

and

Pranav Joshi

in the Network Advanced

Technology Group at Motorola, Inc

Material on

VoIP over Broadband Wireless

provided by

Amitava Ghosh

and members of his team in the

Network Advanced Technology Group at Motorola, Inc

•

Fan Wang

Outline

• Push to talk over GPRS

–

Description

–

GPRS simulator

–

Performance results

• VoIP over Broadband Wireless

–

History and evolution of broadband wireless

–

Voice path delay and user satisfaction

–

VoIP over HRPD-A (EV-DO-A)

•

Why it is going to happen

•

Voice path delay results from simulation

–

VoIP over HSDPA/HSUPA

1 2 2 2 3 3 1 1 2 ₃ 1 2 2 2 3 3 1 1 2 ₃

IMA Summer Program on

Wireless Communications

PTT over GPRS

- Simulation Modeling and Analysis

Pranav Joshi

John Harris

Motorola Inc., Networks

Business

Overview: Push to Talk (PTT)

• PTT over GPRS

–

service has been available for over a year

–

it is the first true commercial VoIP over cellular service.

• Walkie-talkie-type service

–

focus on person-to-person rather than group call

• Simulator – GENeSyS

• Performance Results

Push-to-Talk between Alice and Bob

•

Alice (Originator)

–

Pushes the PTT button to talk with Bob

–

Waits for TPT (Talk-Permit-Tone)

–

Continues to hold button while speaking,

–

Releases when she is done

•

Bob (Target)

–

Audio from Alice plays out

–

TPT “beep” indicates Alice has released her PTT button

–

Pushes the button and waits for TPT

–

Continues to hold button until he is done speaking

•

Single user can transmit at given time (half-duplex channel)

•

Mobile plays “Bonk” if

–

User is not available

–

Any other user is speaking

PTT over GPRS, initial call setup

UL TBF

Setup

Alice

Presses PTT

Alice

Bob

DL TBF

_Setup

DL Trans-

_mission

UL

Trans-UL TBF

Setup

UL

Trans-mission

DL TBF

Setup

DL

Trans-mission

: MS processing

UL TBF

Setup

Alice

Presses PTT

Alice

Bob

DL TBF

_Setup

DL Trans-

_mission

UL

Trans-mission

UL TBF

Setup

UL

Trans-mission

DL TBF

Setup

DL

Trans-mission

: MS processing & packet Assembly

Incl paging delay

Talk

Proceed

Tone

Infrastructure

PTT over GPRS, subsequent PTT

UL TBF

Setup

Bob Presses

PTT

Bob

Alice

Talk

Proceed

Tone

Infrastructure

UL

Trans-mission

DL TBF

Setup

DL

Trans-mission

Server

processing

Server keeps state information,

i.e is aware of Alice’s situation

Subsequent Push timeline

: MS processing &

packet assembly

EtoE Packet

delay

Response time

EtoE Packet

delay

Performance and Capacity

•

User experience metrics

–

initial call setup delay

–

subsequent PTT delay

–

mouth-to-ear audio delay

–

audio turn around time – time from when Alice stops talking

until she hears Bobs response playing out.

•

Service Provider Business Metrics

–

erlangs - average number of subscribers supported at sufficient

user experience per

•

RF carrier

•

network element

•

system

Features of the GENeSyS System Simulator

• C, C++

• Discrete time simulator

–

Basic time unit of 20ms (one block period)

–

8 timeslots per carrier

• Simulator has:

–

Detailed modeling of RLC/MAC

–

Air Interface (RF)

–

Various traffic profiles

–

Simplified version of Core Network

GPRS Adaptive Coding Schemes

Throughput per slot vs C/I

0

2

4

6

8

10

12

14

16

18

20

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Throughput (kbps)

CS1

CS2

CS3

CS4

System planning

for GSM voice

Simulation Configuration

•

4.75 kbps vocoder with 20 msec. framing

–

95 bits every 20 msec. when the user is speaking

•

6 frames per IP packet

•

43 bytes of overhead per IP packet

–

uncompressed IP header

•

Overall audio bit stream of 7.6 kbps

–

6*95 + 43*8 = 570 + 344 = 914 bits per IP packet

–

914 * 50 / 6 = 7616.7 bits per second

•

One PTT talk spurt ~ 5 sec of Audio

•

RLC Acknowledge Mode

Impact of Increasing Number of Dedicated

Slots

1 Uplink timeslot mode, coding scheme 1-2, dedicated timeslot 1 to 6

•

Trunking benefit similar to that

of Erlang-B formula observed:

–

Doubling the number of

dedicated slots more than

doubles the PTT capacity.

–

For example, increasing

number of dedicated slots

from three, up to six,

increases the capacity by

3.1x

•

Capacities quoted correspond

to knee of delay versus load

curve

95th % tile Delay Vs. Load

1 UL TS Mode, CS 1-2, No switchable TS

0

1

2

3

4

5

6

0

0.5

1

1.5

2

# of simultaneous PTT talk Bursts per sector

95th %-tile audio packet delay

(seconds)

1 TS

2 TS

3 TS

4 TS

5 TS

6 TS

Impact of mobile station capability

1 Uplink timeslot mode Vs. 2 Uplink timeslot mode

•

Increasing the number of uplink

timeslots the mobile is capable of

from 1

à

2 significantly improves

performance:

–

Audio delay

•

1 UL timeslot: Light load

~2.0 sec and typical load

~2.4 sec

•

2 UL timeslot: Light load

~1.6 sec and typical load

~2.1 sec

•

This benefit results from better

fractional timeslot utilization with

the improved mobile station

capability – see next slide for

visualization

Mobile's Uplink mode Comparison

4 Dedicated timeslots, CS 1-2, No Switchable timeslots

0 1 2 3 4 5 6 7 8 9 10 0 0.5 1 1.5 2

# of simultaneous PTT talk bursts per sector

95th %-tile audio packet delay

(seconds)

1 UL timeslot Mode 2 UL timeslot Mode

Impact of mobile station capability

1 Uplink timeslot mode Vs. 2 Uplink timeslot mode

•

This benefit results from better fractional timeslot utilization with

the improved mobile station capability

•

Consider a system with 2 dedicated non–hopped slots, & three

CS2 mobiles

•

If 1 slot uplink capable

à

–

Capacity = 2 PTTs or 1/TS

•

If 2 slot uplink capable

à

–

Capacity = 3 PTTs

~12 Kbps

Slot 1

~12 Kbps

Slot 2

MS1

MS2

~12 Kbps

Slot 1

~12 Kbps

Slot 2

MS1

MS3

MS2

Impact of CS3/4

•

CS 3-4

–

Higher Throughput

–

Higher BLER or FER

•

High load

à

Lower delay

•

Light load

à

Higher delay

due to additional re-tx

95th %-tile PTT audio packet delay comparison with

different coding schemes

0

1

2

3

4

5

6

7

8

0

0.5

1

1.5

2

# of simultaneous PTT talk bursts per sector

95th %-tile audio packet delay

(seconds)

4 TS, CS 1-2, 2 UL TS 4 TS, CS 1-4, 2 UL TS

Changing Vocoder and Packetization

•

Vocoder & Packetization

(scenarios)

1. 4.75 kbps, 6 frames/IP

packet, service bit rate

7.6 kbps

2. 5.15 kbps, 10 frames/IP

packet, service bit rate

6.9 kbps

•

Scenarios 2 has lower

service bit rate over

scenario 1

•

Scenario 2 accumulated

lower delay especially at

high load

95th %-tile Delay vs. Load

4 Dedicated PTT TS, 1 UL TS, CS 1-2

# of simultaneous PTT talk bursts per sector

95th %-tile audio packet delay (seconds)

4.75 Kbps Vocoder/ 6 Frame per IP packet

5.15 Kbps Vocoder/10 Frame per IP packet

Performance Impact Study

•

Capacity improvement as numbers of dedicated slots increase

•

Delay and capacity benefit with 2 uplink timeslot mode

•

Effect on delay and capacity as we add coding scheme 3 & 4

capable system

•

Trunking efficiency improvement with switchable timeslots

•

Impact of vocoded frames per IP packets with different vocoder

rate

1 2 2 2 3 3 1 1 2 ₃ 1 2 2 2 3 3 1 1 2 ₃

IMA Summer Program on

Wireless Communications

VoIP over Broadband

Wireless

Phil Fleming

Network Advanced

Technology Group

Motorola, Inc.

Evolution of Broadband Access Technology : Air-Interface

2000

1999

2001

2002

2003

2004

2005

2006

2007

2008

2009

3GPP2

3GPP

802.16

???

???

Voice Delay vs User Satisfaction

100

200

300

400

500

600

Mouth-to-Ear Voice Delay (msec.)

Very

Satisfied

Satisfied

Some

Dissatisfied

Many

Dissatisfied

Almost All

Dissatisfied

Source: ITU-T G114

• mix: Ped A/B, Veh A 3 km/h

*DO-A results assume mobile diversity; Additional capacity in the FL w/ MAC mux

•2-frame bundling = Encapsulation of 2 EVRC frames into 1 RTP/UDP/IP packets

Performance Analysis Summary

9-11

8-10

Set-up Time M -M (sec)

40*

(mix)

18-23

(mix)

Voice Erlangs

(Voice only)

3%

3%

Vocoder FER

(1% RL + 2% FL-delay)

248

250

Voice Delay M -M (msec)

CDMA2000 1xEV-DO-A

(VoIP with 2 -frame)+MAC Mux

CDMA2000 1X

(Circuit Voice)

Internet

Internet

Access

Node

Base Site

(Access Point)

Packet

Core

Packet

Core

Interface to the

Public Network

Base Site

Controller

Packet

Core

Packet

Core

Client

VoIP Delay Components (2-Frame Bundling)

10 ms

Voc Decode

Voc Accum

40 ms @ 2-frm

bundling

20 ms

Voc De-jitter

Voc Encode

15 ms

40 ms @ 40

Erlangs

Air (HARQ)

Air (HARQ)

50 ms @ 40

Erlangs

35 ms

BSC/PDSN

Network/Cor

e

Mob-Mob =

248

ms

BSC/PDSN

Network/Cor

e

38 ms

15 ms

Voc Encode

Voc De-jitter

20ms

40 ms @ 2-frm

bundling

Voc Accum

Voc Decode

10 ms

Forward Link = 160 ms

Reverse Link =173 ms

HRPD Channel Structure

Forward Channel Structure

•

HRPD-0 Forward Channel Fundamentals

–

Time multiplexed channels

•

Pilot

–

initial acquisition, phase & timing recovery, channel estimation & combining,

–

means for predicting received C/I for setting DRC (data rate control)

,

•

MAC – three sub-channels (all users’ RPC+DRCLock CDMed along with RA)

–

Reverse Power Control (users RPC CDMed using MACIndex – size 64 Walsh)

»

RPC data rate = 600*(1- 1/DRCLockPeriod) bps

–

Reverse Activity (RA)

à

one reverse link bit (RAB) per slot (MACIndex=4)

»

combined busy bit (CBB) is set to 1 if any sector sets RAB=1 else CBB=0

»

CBB & Reverse link persistence value used by AT to determine max uplink rate allowed

–

DRCLock

à

AN admission control – when asserted AT to stop selecting sector

•

Traffic (Forward Traffic Channel or FTC)

–

PHY Packet based variable rate traffic channel, data rates 38.4 Kbps to 2.4576 Mbps

–

QPSK, 8-PSK, and 16-QAM modulation ---- R=1/5, 1/3 Turbo Codes

–

PHY Packet sizes: QPSK: 1024 & 2048 bits, 8PSK: 3072 bits, 16QAM: 4096 bits

–

Frame length from 1 to 16 slots, slot length = 1.67ms

•

Control

–

Combines functions of IS -95 sync & paging channels w. rates of 38.4 & 76.8 Kbps

–

Transmitted 8 out of every 256 slots.

Forward Channel Structure

•

No power control: full cell power when transmitting.

–

Time division nature of the burst versus code division for

cdma2000-1x

•

Scheduling done at access point (base station)

–

Multi-user diversity benefit

Forward Channel Structure

•

Transmit slots use a 4-slot interlacing technique. Subsequent transmissions

occur until decoding successful or maximum allowed re-transmissions.

–

Data sent @153.6 kbps is sent in four slots and repeated the following four

slots

•

Maximum # of re-transmissions is dependent on the data rate selected.

•

Users can be scheduled for consecutive slots

.

Users can be

scheduled for

consecutive slots

Slot Structure of Forward Channel

•

No data transmissions occur at the same time as pilot (TDM)

–

High SNR for pilot signal resulting in accurate channel estimates

Active Slot

Idle Slot

Data

400

Chips

Data

400

Chips

Data

400

Chips

Data

400

Chips

1/2 Slot

1,024 Chips

1/2 Slot

1,024 Chips

Pilot

96

Chips

MAC

64

Chips

MAC

64

Chips

Pilot

96

Chips

MAC

64

Chips

MAC

64

Chips

Pilot

96

Chips

MAC

64

Chips

MAC

64

Chips

Pilot

96

Chips

MAC

64

Chips

MAC

64

Chips

HRPD-0 Reverse Link

•

Similar to IS-2000-1x with the addition of the following channels

–

Reverse Rate Indicator (RRI) Channel

--MAC--•

Indicate whether data channel is being transmitted or not & its corresponding rate

–

Data Rate Control (DRC) Channel

--MAC--•

Indicate supportable forward traffic channel data rate

•

Best serving sector on the forward channel

–

Acknowledgement (ACK) Channel

•

The data packet transmitted on the forward traffic received successfully? (600Hz rate)

•

Parameters for Reverse Link

–

Traffic Channel Data Rate Support -- 9.6, 19.2, 38.4, 76.8 and 153.6 Kbps

–

HPSK (variation of BPSK modulation with better peak to average)

–

Packet duration of 26.67 msec (fixed frame length), 16 slots of 1.67msec

–

R=1/2 and R=1/4 Turbo encoder

–

CDM & SHO supported for reverse link traffic channels (not fast cell sel.)

–

Fast reverse power control supported (600*(1- 1/DRCLockPeriod) bps)

Physical Layer Enhancements in HRPD-A

•

Reverse Link Enhancements

–

Higher data rates and finer quantization

•

Support of data rates ranging from 4.8 kbps to 1.8 Mbps with 48 payload sizes

–

4 slot sub-packets (6.66 ms)

–

Hybrid ARQ using fast re-transmission (re-tx) and early termination

–

Support of QPSK and 8-PSK modulation

–

Flexible rate allocation at each AT via autonomous as well as scheduled

mode

–

3-channel synchronous stop-and-wait protocol

•

Forward Link Enhancements

–

Peak rates increased from 2.4 Mbps to 3.1 Mbps

–

Additional small payload sizes (128, 256, 512 bits)

•

Improves frame fill efficiency

–

Data Source Control (DSC) Channel introduced (on RL) to indicate the

desired forward-link serving cell

•

Minimize service interruption due to server switching on FL

System Simulation Assumption: Forward Link

•

VoIP-only and mixture of VoIP and web browsing are modeled.

•

Voice traffic modeled by 4 state Markov chain.

•

1/8

th

rate frames are blanked and not blanked

•

Vocoder frame bundling (multiple voice frames per IP packet)

–

2-frame bundling and no-bundling

•

Overhead used in the simulation

–

3 bytes RTP/UDP/IP, 5 bytes for PPP, and 3 bytes for RLP (11 bytes)

–

6 bytes overhead (PPP header elimination)

•

Hybrid-ARQ included

•

Channel model:

–

34/33/33% Ped-A/Ped-B/Veh-A

•

Scheduler

–

Proportional fair scheduler with delay multiplier.

–

Delay multiplier is a function of delay constraint.

–

Multi-user MAC multiplexing included ( up to 8 users)

•

AT Receive diversity

–

2 way

•

Advanced Receiver

System Simulation Assumption : Reverse Link

•

VoIP-only

•

Voice traffic modeled by 4 state Markov chain.

•

1/8

th

rate frames are blanked and not blanked

•

Vocoder frame bundling

–

2-frame bundling and no-frame bundling

•

Overhead

–

3 bytes RTP/UDP/IP, 5 bytes for PPP, and 3 bytes for RLP (11 bytes)

–

Also modeled a total of 6 bytes overhead

•

HARQ included

•

Channel model:

–

34/33/33% Ped-A/Ped-B/Veh-A

•

Scheduler

–

Rate control scheduling

•

Reverse link overhead

–

DRC and Pilot channel

(DRC to Pilot power ratio is set to –6 dB for repetition 8)

–

Ack/Nack channel

(Ack/Nack to Pilot power ratio is set to 4 dB)

•

BTS Rx Diversity

–

2-way

•

Interference Canceller at the BTS

System Simulation Parameters

Parameter

Explanation/Assumption

Comments

Cellular layout

Hexagonal grid, 3-sector sites

19 sites

Site to Site distance

2500 m

Antenna pattern

As proposed in 1XEV-DV Evaluation Method

Only horizontal pattern specified

Propagation model

L

= 28.6 + 35

Log

(R)

R in meters

Power allocated to 1XEV-DO

Total cell power

Slow fading

Similar to UMTS 30.03, B 1.4.1.4

Std. deviation of slow fading

8.0 dB

Correlation between sectors

1.0

Correlation between sites

0.5

Correlation distance of slow fading

50 m

See D,4 in UMTS 30.03.

Carrier frequency

2000 MHz

BS antenna gain

14 dB

UE antenna gain

0 dBi

UE noise figure

9 dB

Max. # of retransmissions

Variable (dependent on selected data rate)

Retransmissions by fast HARQ

Fast HARQ scheme

IR and Chase as per specification

4 channel stop-and-wait

BS total Tx power

43 dBm

Active set size

3

Maximum size

Reverse Link Delay:

2 Frame Bundling, Ped-A+Ped-B+Veh-A, RF Delay only (Maximum allowable pathloss = 130 dB)

• 40 users can be

supported with RF

delay of 50 ms with 11

byte overhead

• Users out of range are

assumed dropped.

Reverse Link Delay:

2 Frame Bundling, Ped-A+Ped-B+Veh-A, RF Delay only (Max allowable pathloss = 160dB)

• 40 users can be

supported with RF

delay of 50 ms with

11 byte overhead

• No restriction on

location of users

• Capacity drops by 5

Erlangs when the

Reverse Link Delay:

No Frame Bundling, Ped-A+Ped-B+Veh-A, RF Delay only

(Maximum allowable pathloss = 160 dB)

•40 users can be supported

with RF delay of 50 ms

with 11 byte overhead and

no-frame bundling

Reverse Link Delay:

6 bytes overhead, Ped-A+Ped-B+Veh-A, RF Delay only

(Maximum allowable pathloss = 130 dB)

• With a 40 ms delay bound the 2-frame bundling performs better than

no-frame bundling

• If the delay bound is increased to 50 ms the performance with no frame and

2-frame bundling are identical

(need some explanation)

Reverse Link Delay:

6 bytes overhead, Ped A+Ped-B+Veh-A, 1/8 rate frames included

• Capacity drops from 45 to 40

Erlangs if 1/8 rate frame is

included with a RF delay of

50 ms

• Complete blanking of 1/8

th

rate frames causes poor voice

quality at the beginning of a

talk spurt

• Loss in performance not as

significant as in forward link

Forward Link Delay:

2 Frame Bundling, Ped A+Ped-B+Veh-A, RF Delay only,

w/ MAC Mux

• MAC multiplexing

of up to 8 users

• 40 Erlangs can be

supported with RF

delay of 40 ms

• Significant increase

in capacity with

MAC multiplexing

Forward Link Delay:

6 bytes overhead, Ped A+Ped-B+Veh-A, 1/8 rate frames suppression,

w/ MAC Mux

Forward Link Delay:

6 bytes overhead, Ped A+Ped-B+Veh-A, 1/8 rate frames included,

w/ MAC Mux

•

Capacity drops from 45 to 30 Erlangs if 1/8 rate frame is included with a RF

delay of 50 ms

•

Blanking of 1/8

th

_{rate frame results in a drop in voice quality at the}

VoIP over HRPD-A Capacity: Conclusions

•

VoIP only capacity is balanced between forward and reverse links

•

40-45 VoIP (Mobile to Mobile) Erlangs

with a mixture of channels and 2 frame

bundling with Mobile to Mobile delay of

less than 250 ms

•

Two frame bundling has slightly greater capacity to no-frame bundling

•

Inclusion of 1/8

th

rate frames significantly degrades VoIP Capacity

–

Need to study the effect on MOS with various 1/8

_{rate frame blanking schemes}

–

Trade-off between capacity and quality

•

Two frame or no frame bundling is the desired mode of operation since gaps in

speech will lead to inferior voice quality with loss of packets with frame bundling

greater than 2.

•

Mixture of VoIP and Web services can be supported.

–

Graceful degradation in data capacity as VoIP users are increased

•

~250 kbps of FL data traffic

with 25 Erlangs in both directions

⇒

Mobile diversity brings extra FL data throughput

•

Advance receiver at the MS will further increase FL throughput

•

Work in progress includes

–

Performance with both advanced receiver and RX diversity enabled in FL

–

Performance with 4-way RX diversity at BTS

•

High Speed Downlink Packet Access

•

3G (WCDMA Rel-99 & CDMA2000-1x Rel-C ) spectral efficiency ~ 2.5G

(GSM/GPRS) for wireless packet data

•

SE significantly increased with

technology enablers

:

–

Fast distributed scheduling

(at Node-B)

–

Fast AMC

: H-ARQ + IR, higher-order modulation (

ala EDGE

),

multi-code tx, smaller frame sizes, fast channel quality

feedback.

•

HSDPA (Rel-5 WCDMA)

is the evolution of WCDMA Rel-99/4 using

technology enablers

–

2ms sub-frames (vs 10ms frames), Peak Rate 14Mb/s

–

QPSK, 16QAM w. Stop&Wait H-ARQ and IR

–

Fast scheduling, Fast Ack/Nack & ch. quality feedback

User Equipment

Node B

HS-DPCCH

conveys channel

quality report and

ACK/NACK

UL-DPCCH

UL-DPDCH

HS-DSCH

high-speed downlink

shared channel

HS-PDSCH

carries HS-DSCH

transport channel

HS-SCCH

signalling info

for the HS-DSCH

(max. 4 channels)

DL-DPDCH

DL-DPCCH

Associated dedicated physical channels (DPDCH and DPCCH) must be

present : can simultaneously transport R’99/R4 DCH, e.g. voice

call in parallel with HS-DSCH data (subject to UE capability)

•

HS-PDSCH Physical Channel Fundamentals

–

Partitioned into static 2ms (3 timeslot) periods or “sub-frames” (2560 chips)

•

i.e. HSDPA Transmission Time Interval (TTI) is 2ms sub-frame

•

5 sub-frames per 10ms frame

–

QPSK and 16-QAM modulation

–

Fixed length-16 symbol spreading

–

Code division multiplexing (CDM) of users per TTI

TTI

2ms

UE 3

UE 1

UE 2

UE 3

UE 1

HS-PDSCH Code

Space

(Not necessarily

contiguous)

UE 2

UE 2

UE 1

UE 3

CPICH

Time Ref.

TTI 0

TTI 1

TTI 2

TTI 3

TTI 4

System Overhead + HSDPA Control

R'99 Services

(e.g. voice)

Power &

Code

Time

HS-PDSCH Physical Channel

HSUPA Concept

à

lower delay & higher capacity

•

High Speed Uplink Packet Access

•

E-DCH – enhanced uplink dedicated channel

–

Fast Node-B scheduling with HARQ and IR

•

control & minimize Rise over Thermal (RoT) variation

•

avoid RLC re-transmission delay and benefit from previous tx energy

–

#UEs per TTI depends on assignment of available RoT margin (left over R99/4/5)

–

Higher Rates

•

BPSK or QPSK modulation

•

Variable length (64 to 2) symbol spreading

TTI

2ms / 10ms

UE 3

UE 1

UE 2

UE 3

UE 3

RoT Margin

Used

UE 2

UE 2

UE 1

UE 3

CPICH

Time Ref.

TTI 0

TTI 1

TTI 2

TTI 3

TTI 4

R'99/4/5 Services

(e.g. voice) and

signalling

RoT

Margin

Available

Time

UE 1

UE 2

E-DCH Physical Channel Structure

User Equipment

Node B

E-DPDCH

Carries E-DCH

Transport channel

E-DPCCH

Signaling info for

E-DPDCH (TFRI,RSN)

UL-DPCCH

UL-DPDCH

ACK Ch (E-HICH)

Conveys ACK/NACK

E-RGCH

Conveys ternary

up/down/hold

relative grant from

serving cell and

binary down/hold

command from

non-serving cell

E-AGCH

Conveys absolute

grant information

DL-DPDCH

DL-DPCCH

Physical Layer Features for HSDPA/EUL: VoIP

•

Duplexing: FDD

•

Multiple access: CDMA

•

TTI: 2 ms for both UL and DL

– For EUL both 2ms and 10ms TTI is supported

•

Wide range of payload size supported

– 137 bits-28000 bits for HSDPA

•

Bandwidth: 5 MHz

•

Modulation levels

– Downlink: QPSK, 16QAM

– Uplink: BPSK, QPSK

•

AMC support

•

H-ARQ at uplink and downlink

– Key feature for VoIP

– 6 and 8-channel stop-and-wait protocol on DL and UL respectively

•

Fast CQI feedback and Ack/Nack

•

Node-B based scheduling for both DL and UL

VoIP Delay Components

10 ms

Voc Decode

Voc Accum

40 ms @ 2-frm

bundling

20 ms

Voc De-jitter

Voc Encode

15 ms

70 ms @ 70

UEs, 98%-tile

Air (HARQ)

Air (HARQ)

20 ms @ 70

UEs, 99%-tile

40 ms

Network

RNC/GSN

Mob-Mob =

255 ms

Network

RNC/GSN

40 ms

15 ms

Voc Encode

Voc De-jitter

20 ms

40 ms @ 4-frm

bundling

Voc Accum

Voc Decode

10 ms

Forward Link =

40+55+80=195

Reverse Link =

40+35+70=145

Vocoder Modeling

•

12.2 kbps vocoder

•

Number of information bits in 20 msec

–

12.2 kbps : 244 information + 16 CRC = 260 bits

•

Two state Markov Model

–

Voice activity factor should be set to 0.32 by randomly choosing on and off

periods of appropriate duration.

•

No and two frame vocoder frame bundling (20 ms and 40 ms)

•

Results shown without SID frame modeling (no “comfort noise”)

Speech Activity Time Series

0 0.2 0.4 0.6 0.8 1 0 200 400 600 800 1000 Activity

DL/UL VoIP Simulation Parameters

Parameter

Assumption

Cellular layout

Hexagonal grid, 19 sites, 3sectors

Macro-cell propagation model

L=128.1+37.6(R)

Shadowing Model

Log Normal stdev 10.0dB

Fading Model

50% PB, 50% VA

UE receive diversity

With and Without considered

UE antenna gain

0 dBi

Penetration loss

10 dB

Speed Assignment

3 km/h

Carrier Frequency

1.9 GHz

Node B configuration

3 sectors, 1 carrier

Site to site distance

1000 m

Node B/UE HSDPA capability

15/5 codes

Max CDM

4

HS-SCCH

Not explicitly modeled (10% fixed

power assignment)

HSDPA scheduler

Proportional Fair (

α

=1,

β

=0.75)

HSDPA Resource Allocation

Greedy Algorithm

Common channel power overhead

20%

HARQ with IR, AMC

Enabled – HARQ/IR

Vocoder

12.2 Kbps

Number of bundled speech frames

2

l

7 Bytes Packet Overhead

l

RTP 3 bytes

l

RLC 3 bytes for

unacknowledged mode

VoIP RF Capacity for DL (HSDPA): 2 Frame Bundling

•

A UE is in outage if it has more than 2% of voice frames either lost or arrive later than the delay bound

•

RF capacity with 2-frame bundling

•

Without RxDiv: 60-70 Erlangs/sector; With RxDiv: 140-160 Erlangs/sector.

•

Need F-DPCH to support Rx-Diversity

50

55

60

65

70

75

80

85

90

95

100

0

10

20

30

40

50

60

70

80

90

100

Delay - ms

Percentage of Users with FER < 2%

120 Erlangs/Sector 140 Erlangs/Sector 160 Erlangs/Sector 180 Erlangs/Sector

50

55

60

65

70

75

80

85

90

95

100

0

10

20

30

40

50

60

70

80

90

100

RF Delay - ms

Percentage of Users with FER<2%

50 Erlangs/Sector

60 Erlangs/Sector

70 Erlangs/Sector

80 Erlangs/Sector

90 Erlangs/Sector

VoIP RF Capacity for DL (HSDPA): No Bundling

•

Without RxDiv: 70 Erlangs/sector with a delay bound of 70 ms

•

No difference in performance between 2 frame bundling and no bundling case

Without RxDiv

VoIP Packet Delay - No Speech Bundling

Retry count: 6

50

55

60

65

70

75

80

85

90

95

100

0

10

20

30

40

50

60

70

80

90

100

RF Delay - ms

Percentage of Users with FER<2%

40 Erl/Sector

50 Erl/Sector

60 Erl/Sector

70 Erl/Sector

80 Erl/Sector

VoIP RF Capacity for UL (EUL) : 2 Frame Bundling

VoIP RF Capacity for UL (EUL) : No Bundling

3GPP VoIP RF Capacity: Conclusions

•

VoIP capacity with one UE Rx antenna is 70-80 Erlangs/sector with a

mobile-to-mobile delay bound of approx 255 ms

(NO SID)

•

Circuit voice from EUL SI was ~70 erlangs/sector

•

Significant improvement in FL capacity with 2 UE Rx antenna

–

Requires implementation of F-DPCH

–

Capacity bounded by uplink

–

Increase in Uplink RF delay bound numbers

Emerging

802.16 Standards

•

Specifications: completed and

in-progress

–

802.16d – was due May 2004, but significant technology insertion Jan.-May 2004

•

Examples: CTC coding + H-ARQ, Tx Diversity, MIMO, Adaptive Antenna System enhancements

•

Will be approved in July 2004, but additional ‘corrigendum’ document created to accept changes

•

Corrigendum will be proposed as PAR to 802.16#32

–

802.16e – L1 changes (‘Scalable OFDMA’, LDPC codes etc.), plus detailed mobility

support

MAC changes: handoff, sleep modes. PHY changes: scalable

FFT length for PHY (variable BW channelisation)

Primary support: nomadic,

mobile

operation.

MAC & PHY

(<11GHz)

Q3-2005

802.16e

802.16e

Extended 802.16 LOS operation to non-LOS

PHY (2-11GHz)

Apr. 2003

802.16a

802.16a

Enhanced fixed operation – but also radical changes to PHY –

H-ARQ, MIMO, Tx Diversity etc.

Primary support: fixed, nomadic operation.

PHY (<11GHz)

July 2004 (?)

(+ Corrigendum)

802.16d

802.16d

(802.16

(802.16

-

-2004)

2004)

802.16 ‘reduction to practice’ specification – analogous to RAN4

(25.101/104) & UE capability specifications

Profiles

Jan. 2003

802.16c

802.16c

mm-wave (LOS, line of sight) operation

MAC & PHY

(10-66GHz)

Apr. 2002

802.16

802.16

Comments

Scope

Pub.

Spec.

Scope of 802.16d&e Specifications

•

802.16d Specification (published as 802.16-2004)

–

Physical layer specification

•

3 physical layers defined: Single Carrier, OFDM -256, OFDMA-2048

•

Signalling, FEC (H-ARQ, RV), multi-antenna operation

–

MAC layer specification

•

MAC management

–

Addressing, MAC PDU/SDU definitions, fragmentation/concatenation support etc.

–

MAC-layer ARQ, window management etc.

•

QoS management

•

Security sub-layer

•

802.16e Specification (Enhancements)

–

Mobility management

•

Inter-BS (AP) HO, scanning (adjacent cell measurement), sleep modes (paging)

–

Now vehicle for Scalable OFDMA (SOFDMA)

•

Elements Not

Specified by IEEE 802.16

–

Network management (OMC functionality etc.)

–

Layer 3 & Core Network (CN) interfaces

802.16e OFDMA Key Attributes

•

Band AMC mode:

Frequency-selective

adaptive modulation and coding

and

scheduling

•

Scalable bandwidth

mode support

•

Low-complexity Subscriber Station (SS)

receiver design

•

Sustainable Base Station transmitter

efficiency

–

Peak-average ratio similar to HSDPA

in DL

•

Improved broadcast mode

performance

–

Using synchronous or

quasi-synchronous network operation

OFDM DL Waveform Requirements

•

Basic OFDM DL Waveform Requirements

–

Simple

OFDM waveform construction

•

Classical approach to OFDM waveform (inc. cyclic prefix) construction

–

Simplifies UE receiver and BS PA out-of-band emission control

•

Low complexity baseline receiver

, including simple extension to MIMO

•

Design goal:

–

Occupied bandwidth ~90%

–

Inter sub-carrier separation ~11.6 kHz

•

Consistent with target Doppler frequency range and BS/UE impairments

–

Regular scaling in time and frequency domains

•

Integer sample cyclic prefix and guard sub-carrier allocations desirable but not essential

when required to support flexible bandwidth modes

802.16e Scalable Channel Bandwidths and Duplexing

•

802.16e nominally limited to ‘operation in <11GHz

licensed

spectrum’

–

But, could be modified to include unlicensed operation

•

802.16e ‘Scalable OFDMA’ nominal system bandwidths:

•

Constant symbol duration = 100.8us

•

Constant sub-carrier frequency = 11.6kHz

•

12.5% guard period = 12.6us; 6.25% guard period = 6.3us

–

Other 802.16e bandwidths also specifiable

•

e.g. 3.5MHz (European fixed wireless allocations), 7MHz

•

Scalable OFDM approach to realizing these bandwidths under study

–

Options: OFDM sample rate scaling or sub-carrier suppression

•

802.16e Duplexing

–

Observes same duplexing rules as 802.16d

•

Nominally, FDD, TDD and Half-Duplex FDD (HD-FDD) supported

•

Most popular option is TDD

20

10

5

2.5

1.25

System BW (MHz)

11.2

11.2

11.2

11.2

11.2

Sub-carrier Separation (kHz)

100.8

100.8

100.8

100.8

100.8

Symbol Duration (Tb) (us)

2048

1024

512

256

128

FFT Length

802.16 OFDMA Frequency Re-Use

•

Frequency Re-Use Modes

–

Full Usage of Sub-channels (FUSC) (mandatory mode)

•

All frequencies re-used in every sector of every cell (1/1, or full re-use)

•

2004/D5 Section 8.4.6.1.2.2

–

Partial Usage of Sub-Channels (PUSC) (mandatory mode)

•

Nominal

1/3 re-use pattern – used for FCH & DL Map signalling

•

2004/D5 Section 8.4.6.1.2.1

–

Full Usage of Sub-Channels (FUSC) (optional mode)

•

2004/D5 Section 8.4.6.1.2.3

–

Also, ‘adjacent’ or ‘AMC’ DL sub-channelisation mode

•

2004/D5 Section 8.4.6.3

Frequency

(

Logical

Sub-channel

Number)

Segment 1

(Sector A)

Segment 2

(Sector B)

Segment 3

(Sector C)

Note – Adjacent logical

sub-channels are not

necessarily adjacent in

physical frequency

domain.

Typical 802.16d/e TDD Frame Structure

•

Key Elements

–

DL and UL maps indicate per burst data regions, modulation, coding etc.

–

DL bursts are arbitrary congruent blocks – uplink follow in sequence

Peak Data Rate Summary : Example

•

Peak Data Rate Summary

–

Assumes 5MHz channel bandwidth – length-512 FFT

–

Assumes 802.16e/D3 mandatory DL FUSC sub-channelisation (Table 272c)

–

Assumes 70/30 DL/UL split

•

DL Peak Data Rate Limitation

–

Is MSS restricted to 1 data region per frame, peak data rate is limited by maximum

block size that can be signaled.

–

For H-ARQ mode, maximum block size is 24000 bits

–

Resulting user peak data rate = 4.8Mbps (5ms frame)

0.25

0.5

0.75

0.25

0.5

0.75

0.595

1.190

1.786

0.230

0.461

0.691

1.190

2.381

3.571

0.461

0.922

1.382

2.381

4.762

7.142

0.922

1.843

2.765

3.571

7.142

10.714

1.382

2.765

4.147

(Code Rate)

DL Data Rate Mb/s

UL Data Rate Mb/s

(Code Rate)

QPSK

16-QAM

64-QAM

BPSK

VoIP characteristics and requirements

•

VoIP traffic

–

One packet every 20 ms (full, half, quarter, null)

–

Strict delay latency requirement (RF delay 70 ms in simulation)

•

Outdated packets are dropped

•

Outage should be less than 1%

•

Scheduling and resource allocation for VoIP

–

Channel adaptive only (e.g. proportional fair) scheduler does not perform good

–

Multi user diversity gain is less for VoIP

–

Delay constraint must be included in the scheduler

–

Similar scheduler for NRTSV has been applied for VoIP and performs good

•

VoIP packet is usually small, but requires high transmission reliability

–

Small Packet size should be supported

–

High R ratio (spreading gain * (1/coding rate)) should be supported

–

Channel diversity (e.g. multiple antenna) can significantly improve the coverage and

capacity

•

Ped-B/Veh-A has higher capacity than flat fading

–

Multiple user multiplexing is critical

•

OFDMA: 32 users in every 5 ms per sector

•

HSDPA: 4 users in every 2 ms per sector

System Simulation Parameters ( 802.16e – DL)

Parameter

Value

Cell layout

17 W ERP, 3 sector, hex grid, 19 sites, 2.8km site-site dist.

Frequency Reuse

1

Propagation/shadowing/ antenna models

Lognormal std dev. = 8.0 dB; L=128.1+37.6log10(R) , R in km

Fading model

50% Ped-B, 50% Veh-A

Cyclic prefix overhead

10%

Pilot allocation

Approximately 10%

Tx diversity

Off

Rx diversity

Off

Re-transmission

IR

SS Receiver type

1-tap frequency domain equalization

Scheduler/resource allocation

Prop. fair – Delay constraints for VoIP, Non-frequency selective

scheduling

Traffic Model

VoIP

Number of data sub carriers

384 (5 MHz)

Number of downlink data symbols per 5ms frame

22

Modeling of control channels

1 symbol preamble and 2 symbols for MAP

VoIP Frame Bundling

1

System Simulation Parameters ( 802.16e – UL)

Parameter

Value

Cell layout

200 mW ERP, 3 sector, hex grid, 19 sites, 2.8km site-site dist.

Propagation/shadowing/ antenna models

Lognormal std dev. = 8.0 dB; L=128.1+37.6log10(R) , R in km

Fading model

50% Ped-B, 50% Veh-A

Cyclic prefix overhead

10%

Pilot allocation

Approximately 10%

Tx diversity

Off

Rx diversity

On

Soft/Softer handoff

On

Re-transmission

IR

BS Receiver type

1-tap frequency domain equalization

Scheduler/resource allocation

Prop. fair – Delay constraints for VoIP

Traffic Model

VoIP

Number of data sub carriers (including control)

384 (5 MHz)

Number of uplink data symbols per 5ms frame

24

Modeling of control channels

48 subcarriers for control, including Ranging, Bandwidth request

VoIP Frame Bundling

1

Scheduling and resource allocation

• Scheduling

–

Two major classes

•

Frequency non-selective

– PUSC, FUSC (Random/Interleaved)

•

Frequency Selective

– Band AMC (Contiguous)

• Resource allocation

–

Resource requests are satisfied according to user

priority

–

Allocated resources are calculated based on CQI

feedback and scheduled packet size

• Retransmission

–

Retransmission resources are determined according to

VoIP Performance (5 MHz, 50/50 DL/UL Split, FER<1%)

Downlink

_Uplink

200 SS per sector

220 SS per sector

80 SS per sector

100 SS per sector

120 SS per sector

Delay in VoIP over 802.16e Forward Link

1% frame outage

0.01

0.10

1.00

0

10

20

30

40

50

60

70

Delay (msec.)

Prob[Delay > x]

200 user/sector

220 user/sector

VoIP Performance (5 MHz, 50/50 DL/UL Split, FER<2%)

802.16e VoIP Delay Components

10 ms

Voc Accum

20 ms

20 ms

Voc De-jitter

Voc Encode

15 ms

30 ms @ >100

Erlangs

Air (HARQ)

Air (HARQ)

70 ms @ 100

Erlangs

35 ms

Packet core

delay

Mob-Mob =

238

ms

Packet core

delay

38 ms

15 ms

Voc Encode

Voc De-jitter

20 ms

Voc Accum

Voc Decode

10 ms

Forward Link = 130 ms

Reverse Link =173 ms

10 ms

Voc Decode

Voc Accum

20 ms

Voc De-jitter

Voc Encode

15 ms

Erlangs

Air (HARQ)

Air (HARQ)

Erlangs

35 ms

Mob-Mob =

ms

38 ms

15 ms

Voc Encode

Voc De-jitter

20 ms

Voc Accum

Voc Decode

10 ms

802.16e VoIP RF Capacity: Conclusions

•

VoIP capacity with one UE Rx antenna is ~200 Erlangs/sector for DL

and ~ 80 Erlangs /sector for UL with a mobile-to-mobile delay bound

of approx 240 ms (no SID)

•

Capacity limited by RL

•

Analysis very preliminary

•

Performance will improve with frequency selective scheduling

•

Effect of SID frames on VoIP capacity currently being studied

Summary of VoIP Performance over

Broadband Wireless

Features HRPD-A WCDMA- Rel 5/Rel6 WCDMA- Rel-99 802.16e

Specturm Occupancy 1.25 MHz (FDD) 5 MHz (FDD) 5 MHz (FDD) 5/10/20 MHz (TDD/FDD) Data shown for TDD mode only

Chip rate or #sub-carriers 1.2288 Mcps 3.84 Mcps 3.84 Mcps 512/1024/2048

F/L 3.1 Mbps 13.97 Mbps 1.92 Mbps 11/22/44 Mbps (TDD, 70% DL with 64-QAM and R=3/4 code)

R/L 1.8 Mbps 4.0 Mbps (Tentative) 384 kbps 2.7/5.5/11 Mbps (TDD, 30% UL with 16-QAM _{and R=3/4)}

F/L 1.6666 2 10 5 (TDD, 70% DL)

R/L 6.66 2 and 10 10 5 (TDD, 30% UL)

F/L QPSK/8-PSK/16-QAM QPSK/16-QAM QPSK BPSK/QPSK/16-QAM/64-QAM

R/L BPSK/QPSK/8-PSK BPSK/QPSK BPSK BPSK/QPSK/16-QAM/64-QAM (?)

HARQ, IR, Chase, AMC 4-channel HARQ stop-and-Fast wait protocol on FL

Fast

6-channel HARQ stop-and -wait protocol on FL

No _{HARQ stop-and -wait protocol on FL}Fast

Tx Diversity at BTS Yes (open loop only) Yes (Open loop and Closed loop) Yes (Open loop and _{Closed loop)} Yes (STBC and MIMO)

Advanced Receiver Yes Yes No Yes

Adaptive Antenna Support No Yes using Dedicated Pilots Yes using Dedicated

Pilots Yes

Multiple Access (MAC multiplexing used TD-CDMA on the FL)

CDMA

(Up to 4 users can be CDM'ed per

TTI) CDMA Multiple users scheduled per TTI

Enhanced broadcast

Present: Plain broadcast using CDMA Proposal for OFDM based

MBMS support is being

MBMS

(Upto 256 kbps can be supported

using selection/soft combining) NA

MBMS

(3 Mbps in 5 MHz using SFN) Max Data Rate per User w/o

MIMO

TTI Size (msec)

Broadband Access Technology VoIP Capacity Comparison

Features

HRPD-A (EV-DO Rev A)

W-CDMA, Rel 5/6

802.16e

Spectrum Occupancy

1.25 MHz (FDD)

5 MHz (FDD)

5/10 MHz (TDD)

Vocoder

8 kbps EVRC

12.2 kbps AMR

8 kbps AMR

Voice Capacity (F/L)

(Erlangs)

45 (VoIP)

70-80 (CS)

80-90 (VoIP)

150-200

(5 MHz, 50% DL)

Voice Capacity (R/L)

(Erlangs)

35-40 (VoIP)

~80-90 (VoIP)

80-100

(5 MHz, 50% UL)

Voice Spectral Efficiency

(Erlangs/MHz)

24.5-28 (normalized to 10 MHz)

21-24 (normalized to 5 MHz)

14-16 (CS)