2012 IEEE Wireless Communications and Networking Conference

Coverage analysis ofSkype VoIP services

over 3G cellular networks

Renny E. Badra and Hayat AbdullaDepartamento de Electrónica y Circuitos

Universidad Simón BolívarCaracas, Venezuela

renny@usb.ve , habdulla@usb.ve

This presentation

MotivationSkype TrafficSimulation Set-upResultsConclusions

Motivation

Skype ® is the most popular commercial platformfor peer-to-peer IP telephonyOver 700 million registered users worldwideOften used over data-enabled cellular devices

Skype can be significantly bit rate-hungryCellular coverage is compromised for high rate servicesAir interface parameters may affect Skype traffic

This work seeks to perform a side-by-sidecoverage comparison between conventional CSvoice services and Skype over 3G networks

Skype Traffic

CODEC G.729 iSAC SVOPC

Constant / Variable Bit Rate

(CBR/VBR)

CBR VBR VBR

Speech Bandwidth 4 kHz 8 kHz 8 kHz

Codec frame interval [ms] 10 30 or 60 20 - 60

Skype Frame interval Variable

Codec output bit rate [kbps] 8 Variable

10 – 32

Variable

20 – 50

Approx. max. sending rate

(no redundancy) [kbps]

12

34

54

Application Scenario Skype-

out

Skype-

to-Skype

Skype-to-

Skype

(3.1 or

later)

Main Skype Voice Codecs

Skype Traffic

Three factors influence Skype data rate:Choice of voice codecChoice of compression ratio (only for VBR codecs)Skype redundancy ratio

Under random packet loss conditions, Skypedata rate can be modeled by:

rs(t) is Skype’s sending rater(t) is the redundancy ratioLI(t) is the bit rate of the codec’s i-th operating mode

!

rs( t) = (1+ r( t))•L

i( t)

Skype Traffic

Skype’s redundancy ratio reflects doubletransmission of some voice packetsRedundant packets are piggybacked to fresh packets.For random packet loss rate under 2%, r(t) approaches

zero.For random packet loss rate above 5%, r(t) approaches

its maximum value (unity).

This work assumes that VBR codecs operateat their maximum rateThis ensures MOS values comparable to CS voice

(about 3.4 for AMR at 12.2 kbps).

Skype Traffic

CODEC AMR G.729 iSAC SVOPC

B L E R 1% 1% 10% 1% 10% 1% 10%

Max codec rate [kbps]

12.2 8

32 50

Max SKYPE sending rate [kbps]

N/A 12 24 34 68 54 108

Max MAC rate [kbps]

15.6 17 29 39 71 59 113

Required Eb/No [dB]

4.5 4.5 2.7 3.7 2.4 3.4 2.5

Traffic simulation scenarios

Simulation set-up

W-CDMA (release 99) system assumed Skype packets carried by the uplink DPDCH RTP/UDP/IP header compression (down to

about 1.6 kbps) Signaling overhead of 3.4 kbps (average) Small MAC and RLC overhead Two BLER set-point values considered for

Skype service: 1% and 10%.

Simulation set-up

Urban hexagonal macrocells Random system parameters:

Large scale fading (Shadowing)Small-scale fading (Rice/Rayleigh)Line-of-sight condition for outdoors locationsSoft-Handoff gain

520K uniformly spaced Monte Carlo trials percellCoverage probability defined as the fraction of Monte

Carlo trials where minimum required power is less thanmaximum UE transmit power

Simulation set-up

Parameter Value

UE max transmit power 24 dBm (1900 MHz)

UE ant. gain + cable loss 0 dB

Receiver Noise Figure 5.0 dB

Receiver Sensitivity Service-dependent

BS Antenna Type Sectorized, 90° beamwidth

BS Antenna Max Gain 16.1 dBi

BS Cable losses 2 dB

Uplink Load Factor 3.0 dB (50% cell load)

BS ant. diversity gain Included in (Eb/No)req

Link Budget Parameters

Simulation Set-up

Location Outdoors Indoors

Environment Urban

Frequency 1900 MHz (UMTS/PCS uplink bands)

UE location Street-level pedestrian In-building

Linearly decreasing LOS

probability as a function of

distance to cell

LOS/NLOS

LOS NLOS

NLOS at

all

locations

Distance-

dependent

Propag. Loss

Walfisch-Ikegami

LOS model

COST 231

Shadowing

Standard

Deviation

4 dB 8 dB 11 dB

Small-Scale

Fading

Rician with

distance-dependent

KR

Rayleigh

Building

Penetration

Loss

0 dB 0dB 12 dB

Body Loss 3 dB

Propagation Models

Simulation Set-up

Both indoors and outdoors locationsconsidered separately

Two cell-area coverage confidencescenarios: 90% and 95% (outdoors).

Simulation Scenario A B

Cell radius 1480 m 1130 m

Street-level coverage probability

for AMR speech

90 % 95 %

In-building coverage probability

for AMR speech

58 % 69 %

Results: G.729

Scenario: A Scenario: B

Cell-area coverage expressed as a fraction of CS-voice coverage

Results: iSAC

Scenario: A Scenario: B

Cell-area coverage expressed as a fraction of CS-voice coverage

Results: SVOPC

Scenario: A Scenario: B

Cell-area coverage expressed as a fraction of CS-voice coverage

Conclusion

Skype coverage is always a fraction of CS-voicecoverage because of higher peak bit rates:Such fraction is close to 100% for G.729 codecCoverage drops to 80-90% for iSAC codec, and even

less for SVOPC (down to a minimum of 67%)

Smaller cells (and thus, looser link budgets)improve Skype coverage relative to CS-speech

Setting BLER to 1% (as opposed to 10%)improves Skype coverage relative to CS-speechSkype redundant packet transmission is avoided by

lowering the air interface BLER

Conclusion

Coverage is confirmed to be a potential obstacle inmassive deployment of proprietary VoIP over cellularnetworks.

This problem can be significantly mitigated through:Adapting traffic management rules to cellular networks at the

VoIP application (i.e., redundancy)Appropriate setting of air interface parameters

Similar problems are expected in HSPA because ofcoverage sensitivity to bit rates and Eb/NorequirementsHowever, layer-2 retransmissions in HSPA should improve

coverage at the expense of increase packet jitter.

2012 IEEE Wireless Communications and Networking Conference

Coverage analysis ofSkype VoIP services

over 3G cellular networks

Renny E. Badra and Hayat AbdullaDepartamento de Electrónica y Circuitos

Universidad Simón BolívarCaracas, Venezuela

renny@usb.ve , habdulla@usb.ve