VoIP over Wireless

12

2

2

33

1

1

2 3

12

2

2

33

1

1

2 3

IMA Summer Program on Wireless Communications


For the personal use of the participants in the IMA summer program on wireless communications. May not be reproduced or distributed in whole or

in part without written consent of the author

VoIP over Wireless

Phil FlemingNetwork Advanced Technology GroupNetwork BusinessMotorola, 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• Wiemin Xiao




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– VoIP over 802.16e

12

2

2

33

1

1

2 3

12

2

2

33

1

1

2 3





PTT over GPRS- Simulation Modeling and Analysis

Pranav Joshi John HarrisMotorola 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

• Group Call: three or more users on single PTT call




PTT over GPRS, initial call setup

UL TBFSetup

Alice Presses PTT

Alice

BobDL TBF

SetupDL Trans-

mission

UL Trans-

UL TBFSetup

UL Trans-

mission

DL TBF Setup

DL Trans- mission

: MS processing

UL TBFSetup

Alice Presses PTT

Alice

BobDL TBF

SetupDL Trans-

mission

UL Trans- mission

UL TBFSetup

UL Trans-

mission

DL TBF Setup

DL Trans- mission

: MS processing & packet Assembly

Incl paging delay

TalkProceed

Tone

Infrastructure

1st PTT setup delay




PTT over GPRS, subsequent PTT

UL TBFSetup

Bob Presses PTT

Bob

Alice

TalkProceed

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

– system resources used per call




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– Multi-carrier, multi-cell, various reuse patterns etc




GENeSyS Schematic Layout




GPRS Adaptive Coding SchemesThroughput 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 25C/I

Th

rou

gh

pu

t (k

bp

s)

CS1

CS2

CS3

CS4

System planningfor 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

– radio frame (20 msec.) re-transmission protocol




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. Load1 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

%-ti

le a

udio

pac

ket d

elay

(s

econ

ds) 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 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 Comparison4 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

%-ti

le a

udio

pac

ket d

elay

(s

econ

ds)

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 – Capacity = 2 PTTs or 1/TS

• If 2 slot uplink capable – Capacity = 3 PTTs

~12 Kbps

Slot 1

~12 Kbps

Slot 2MS1 MS2

~12 Kbps

Slot 1

~12 Kbps

Slot 2MS1 MS2MS3




Impact of CS3/4

• CS 3-4 – Higher Throughput

– Higher BLER or FER

• High load Lower delay • Light load 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


95th

%-ti

le a

udio

pac

ket d

elay

(s

econ

ds)

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. Load4 Dedicated PTT TS, 1 UL TS, CS 1-2

0

2

4

6

8

10

12

14

16

0 0.5 1 1.5 2


95th

%-t

ile a

ud

io p

acke

t del

ay (s

eco

nd

s) 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

12

2

2

33

1

1

2 3

12

2

2

33

1

1

2 3





VoIP over Broadband Wireless

Phil FlemingNetwork Advanced Technology GroupMotorola, Inc.




Evolution of Broadband Access Technology : Air-Interface

20001999 2010-20022001 20042003 20062005 20082007 2009

cdma-1XRev-A

cdma-1XRev-B

1X EV-DVRev-C

1X EV-DORev-0

1X EV-DORev-A

1X EV-DORev-B

BCMCS

3GPP2

3GPP

802.16

1X EV-DORev-C

MC

3GPP2Evol ??

IS-95

cdma-1XRev-0

EDGEGPRSGERAN

(Adv Recvetc.)

GERAN(VoIP,

Multicastetc.)

???

UMTSRel-99

UMTSRel-4

UMTSRel-5

(HSDPA)

UMTSRel-6

(HSUPA)

3GPP EvolRel-7/8

(EUTRA)

802.16eMobility

???802.16dAdv Fixed

802.16aFixed802.16

1X EV-DVRev-D




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




3GPP2 HRPD-A VoIP Performance




• 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

Air Interface CDMA2000 1X

(Circuit Voice)

CDMA2000 1xEV-DO-A

(VoIP with 2-frame)+MAC Mux

Voice Delay M-M (msec) 250 248

Vocoder FER(1% RL + 2% FL-delay)

3% 3%

Voice Erlangs (Voice only)

18-23

(mix)

40*

(mix)

Set-up Time M-M (sec) 8-10 9-11




InternetInternet

MGX 8800 MGX 8800

Access Node

Base Site(Access Point)

PacketCore

PacketCore

Interface to the Public Network

Base Site Controller

PacketCore

PacketCore

Client




VoIP Delay Components (2-Frame Bundling)

Reverse Link =173 ms Forward Link = 160 ms

10 ms Voc Decode Voc Accum 40 ms @ 2-frm bundling

20ms Voc De-jitter Voc Encode 15 ms

38 ms BSC/PDSN Network/Cor

e

Mob-Mob = 248 ms

BSC/PDSN Network/Cor

e

35 ms

50 ms @ 40 Erlangs

Air (HARQ) Air (HARQ) 40 ms @ 40 Erlangs

15 ms Voc Encode Voc De-jitter 20 ms

40 ms @ 2-frm bundling

Voc Accum Voc Decode 10 ms




HRPD Channel Structure

DRC.Lock




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) 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 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• Provision for receive diversity (two antennas) at Access Terminal




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

Data400

Chips

Data400

Chips

Data400

Chips

Data400

Chips

1/ 2 Slot1,024 Chips

1/ 2 Slot1,024 Chips

Pilot96

Chips

MAC64

Chips

MAC64

Chips

Pilot96

Chips

MAC64

Chips

MAC64

Chips

Pilot96

Chips

MAC64

Chips

MAC64

Chips

Pilot96

Chips

MAC64

Chips

MAC64

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

– Multi-user packet support




System Simulation Assumption: Forward Link

• VoIP-only and mixture of VoIP and web browsing are modeled.• Voice traffic modeled by 4 state Markov chain.• 1/8th 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– Have capability, not currently used




System Simulation Assumption : Reverse Link• VoIP-only• Voice traffic modeled by 4 state Markov chain.• 1/8th 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– Not included




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 Log10(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

Specify Fast Fading model Jakes spectrum Generated by Filter approach




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.

40 Erl




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 restriction is removed




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 1/8th 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

• 40-45 Erlangs can be supported with RF delay of approximately 50 ms




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 1/8th rate frame results in a drop in voice quality at the beginning of a talk spurt




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 1/8th rate frames significantly degrades VoIP Capacity

– Need to study the effect on MOS with various 1/8th 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– Performance of RL with Interference Canceller




3GPP HSDPA/HSUPA VoIP Performance




• 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

HSDPA Background




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 Channels




• 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

TTI2ms

UE 3

UE 1

UE 2UE 3

UE 1

HS-PDSCH CodeSpace

(Not necessarilycontiguous) UE 2

UE 2

UE 1

UE 3

CPICHTime 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 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

TTI2ms / 10ms

UE 3UE 1

UE 2

UE 3

UE 3

RoT MarginUsed

UE 2UE 2

UE 1

UE 3

CPICHTime Ref.

TTI 0 TTI 1 TTI 2 TTI 3 TTI 4

R'99/4/5 Services(e.g. voice) and

signallingRoT

MarginAvailable

Time

UE 1

UE 2




E-DCH Physical Channel Structure




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• Vocoders: AMR at 12.2 kbps




VoIP Delay Components

Reverse Link = 40+35+70=145

Forward Link = 40+55+80=195

10 ms Voc Decode Voc Accum 40 ms @ 4-frm bundling

20 ms Voc De-jitter Voc Encode 15 ms

40 ms Network

RNC/GSN

Mob-Mob = 255 ms

Network

RNC/GSN

40 ms

20 ms @ 70 UEs, 99%-tile

Air (HARQ) Air (HARQ) 70 ms @ 70 UEs, 98%-tile

15 ms Voc Encode Voc De-jitter 20 ms

40 ms @ 2-frm bundling





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

20ms Frame Intervals

Acti

vit

y




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 (=1, =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

7 Bytes Packet Overhead

RTP 3 bytes

RLC 3 bytes for unacknowledged mode

PDCP 1 byte




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

Per

cen

tage

of

Use

rs w

ith

FE

R <

2%

120 Erlangs/Sector140 Erlangs/Sector160 Erlangs/Sector180 Erlangs/Sector

50

55

60

65

70

75

80

85

90

95

100

0 10 20 30 40 50 60 70 80 90 100RF Delay - ms

Per

cen

tage

of

Use

rs w

ith

FE

R<

2%

50 Erlangs/Sector

60 Erlangs/Sector

70 Erlangs/Sector

80 Erlangs/Sector

90 Erlangs/Sector

Without RxDiv With RxDiv




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 BundlingRetry 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

Per

cen

tage

of

Use

rs w

ith

FE

R<

2%

40 Erl/Sector50 Erl/Sector

60 Erl/Sector70 Erl/Sector80 Erl/Sector




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

• Approx 80 Erlangs/sector can be supported with a delay bound of 30 ms




VoIP RF Capacity for UL (EUL) : No Bundling

• Approx 80 Erlangs/sector can be supported with a delay bound of 30 ms




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

• Effect of SID frames on VoIP capacity currently being studied




802.16e VoIP Performance




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

Spec. Pub. Scope Comments

802.16802.16 Apr. 2002 MAC & PHY (10-66GHz)

mm-wave (LOS, line of sight) operation

802.16c802.16c Jan. 2003 Profiles 802.16 ‘reduction to practice’ specification – analogous to RAN4 (25.101/104) & UE capability specifications

802.16a802.16a Apr. 2003 PHY (2-11GHz) Extended 802.16 LOS operation to non-LOS

802.16d802.16d

(802.16-(802.16-2004)2004)

July 2004 (?)(+ Corrigendum)

PHY (<11GHz) Enhanced fixed operation – but also radical changes to PHY – H-ARQ, MIMO, Tx Diversity etc.Primary support: fixed, nomadic operation.

802.16e802.16e Q3-2005 MAC & PHY (<11GHz)

MAC changes: handoff, sleep modes. PHY changes: scalable FFT length for PHY (variable BW channelisation)Primary support: nomadic, mobile operation.




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

• Inter-BS (AP) interface, BS-CN interface, BS-RNC interface etc.




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

SINR

Frequency

Inst. SINR

Avg. SINR

MCS 0

MCS 2

MCS 1

Freq. Selective Scheduling and AMC




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

Cyclic PrefixUseful Symbol

t

Tu (Nu Samples)Tg (Ng Samples)

Ts (Ns Samples)Total Symbol t

Ng GuardBand

Ng GuardBand

Df Sub-carrierSeparation

Nu Used Sub-Carriers




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

FFT Length 128 256 512 1024 2048

System BW (MHz) 1.25 2.5 5 10 20

Sub-carrier Separation (kHz) 11.2 11.2 11.2 11.2 11.2

Symbol Duration (Tb) (us) 100.8 100.8 100.8 100.8 100.8




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-channelNumber)

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

– Allocated regions for UL for random access, channel quality (CQI), ACK’s

Sub

-cha

nnel

Log

ical

Num

ber

ss-1

s+1

1

Ns

0 1 3 5 7 9 N-1

FCH

DLMAP

ULMAP

DL Burst #2

DL Burst #1DL Burst #4

DL Burst #6

DL Burst #7

DL Burst #5

ULMAP(cont)

Pre

ambl

e

Coded symbol write order

ACK

0 M-1

CQI

DL Burst #3

UL Burst #1

UL Burst #2

UL Burst #3

UL Burst #4

UL Burst #5

UL Burst #6

OFDM Symbol Number

GuardDownlink Subframe Uplink Subframe

Ranging Sub-channel




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 0.750.595 1.190 1.786 0.230 0.461 0.6911.190 2.381 3.571 0.461 0.922 1.3822.381 4.762 7.142 0.922 1.843 2.7653.571 7.142 10.714 1.382 2.765 4.147

(Code Rate)DL Data Rate Mb/s UL Data Rate Mb/s

(Code Rate)

QPSK16-QAM64-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• DO-A: 8 users in every 1.667 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 Vocoder 8 kbps




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 Vocoder 8 kbps




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

the left-over information in a scheduled packet size




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

Downlink Uplink

0 10 20 30 40 50 60 70 800

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

RF-delay (ms)

Pro

b. o

uta

ge

< a

bsc

issa

Emperical CDF (outage: FER < 1.0 %)

200 SS per sector220 SS per sector

0 10 20 30 40 50 60 70 80 90 1000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

RF-delay (ms)

Pro

b. o

uta

ge

< a

bsc

issa

Emperical CDF (outage: FER < 1.0%)

80 SS per sector100 SS per sector120 SS per sector




Delay in VoIP over 802.16e Forward Link1% frame outage

0.01

0.10

1.00

0 10 20 30 40 50 60 70

Delay (msec.)

Pro

b[D

elay

> x

]

200 user/sector 220 user/sector




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

Downlink Uplink




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





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 Mbps11/22/44 Mbps (TDD, 70% DL with 64-QAM

and R=3/4 code)

R/L 1.8 Mbps 4.0 Mbps (Tentative) 384 kbps2.7/5.5/11 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, AMCFast

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

Fast6-channel HARQ stop-and -wait

protocol on FLNo

Fast HARQ stop-and -wait protocol on FL

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 PilotsYes using Dedicated

PilotsYes

Multiple AccessTD-CDMA

(MAC multiplexing used 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

discussed in 3GPP2)

MBMS(Upto 256 kbps can be supported using selection/soft combining)

NAMBMS

(3 Mbps in 5 MHz using SFN)

Max Data Rate per User w/o MIMO

TTI Size (msec)

Modulation

VoIP over Wireless

Documents

Transcript of VoIP over Wireless