WiPPETsignal

Multi-cell WCDMA Signal Processing Simulation

Vikram KaulRoy D. Yates

Wireless Propagation and Protocol Evaluation Testbed

Outline

ØMotivation and Background

ØWiPPETsignal Design

ØWiPPETsignal Processing Rates, Complexity,Parallelization and Speedups

ØWiPPETsignal Studies

ØConclusions and Future

Motivation & Background

Motivation

ØSimulation studies at physical layers o Link level (e.g. interference cancellation)o System Level (e.g. power control and handoff)

ØHow do system level algorithms respond to link level processing ??

ØNeed for fast, reusable and configurable simulation setups

Motivation & Background

Background Motivation & Background

Ø1994 - 96 MADRASo Uniprocessor simulation in C

Ø1997 WiPPET Sessiono Calls - Blocking, Dropping, Handoff

Ø1998 WiPPET Packeto Packets - Dropping, Retransmission, Throughput

Ø1999 WiPPET Signalo Bits, Symbols, Chips, Waveforms, BER, SIR

WiPPETsignal Design

WiPPETsignal Features

ØMulti-cell waveform simulation at chip level precisiono Asynchronous Bases and Mobileso Uplink dedicated physical channels o Downlink common pilot channel

ØPropagation and Interferenceo Distance loss, shadow fading, fast fading, multipath

ØProtocols and algorithmso Transceiver signal processing, Power control, Soft

handoff

WiPPETsignal Design

WiPPETsignal CSSF Implementation

ØScalable Simulation Framework

Entity1 Entity2

Entity3

Entity4

SSF Channel

SSF Channel

Event

WiPPETsignal Design

ØSolaris and Linux

ØParallel discrete event simulation API

ØEntities interact with other over SSF channels

Block DiagramWiPPETsignal Design

MS 1

MS 2

MS N

BS j

BS 1

BS 12

MS x

RAYLEIGH (short time-scale losses)

MOBILITY/GEO (long time-scale losses)

SWITCH

FwdRCH

RevRCH

Message Sequence Chart: TPC Loop

STEADY STATE

MS 1 Rev RCH BSSTEADY STATE

Channel Processing

Set Next TxPow

Channel Processing Accumulate

Accumulate

Generate TPC

Rx_DATA_SYMBOL

Rx_DATA_SYMBOL

CHANGE_POWER

MS x

Set Next TxPow Channel Processing AccumulateRx_DATA_SYMBOL

WiPPETsignal Designtim

e

Tx_DATA_SYMBOL

Tx_DATA_SYMBOL

Idealized feedback

channel

Radio Channel ProcessingWiPPETsignal Design

ØPropagation and interference from every Tx to every Rx

ØRayleigh Fadingo Impulse response has an approximate closed formo Polyphase subfilter implementation, WSSUS Model

ØDistance loss and shadow fadingo Manhattan geography, Streets and avenues, Wrap-

aroundo Distance resolution, correlation distance, Shadow

variation

Transceiver Signal ProcessingWiPPETsignal Design

Ø3GPP and NTT DoCoMo

ØWCDMA physical layer suite

ØCoherent multi-stage serial interference canceller (COMSIC)

ØTransceiver processing inherited by MOBILE, BASE and SWITCH

ØTransceiver software implementation decoupled from radio channel implementation

DATUM WiPPETsignal Sample studies

ØWINLAB DATUM o Socket interfaceo Generic/Independento SIR, BER

comparisonso Adding more

parameters to track

ValidationWiPPETsignal Design

ØUnit and System level validationo Rayleigh fadingo Shadow fadingo Single user BPSK

validation

ØValidation against NTT DoCoMo o 10 user single cell

simulationso NTT DoCoMo

measurements Eb/N0

BE

RB

ER

Eb/N0

WiPPETsignalProcessing rates,ComplexityParallelization & Speedups

Processing RatesWiPPETsignal Processing rates & Complexity

Users/cell

ØAs load (K) increaseso RR = O(1)o RMF = O(1)o RIC = O(1/K)

ComplexityWiPPETsignal Processing rates & Complexity

ØAnalysis of relative complexity for each moduleo Theoretical gradientso Software profiles

ØBit Detection (Multi-stage Interference Cancellation)

ØRadio Channel Processing

Users/cell

Per

cen

tage

ParallelizationWiPPETsignal Parallelization & Speedup

MS 1

MS 2

MS N

BS j

BS 1

BS 12

MS x

RAYLEIGH

MOBILITY/GEO

SWITCH

FwdRCH

RevRCH

………

………

Radio Channel Parallelization

Processor 3 Processor X

Processor 1 Processor 2

results

BS j

BS 1

BS 12PARALLEL

Bit detection job

Bit detection job

Bit detection job

PARALLEL

Transceiver Parallelization

Ø Classically parallel problem for multi-cell simulations

SpeedupsWiPPETsignal Parallelization & Speedup

Ø Reverse link and Forward-&-Reverse link simulationsØ Load balancing on all threads helpfulØ Exploit heavy transceiver processing on reverse link

Only RCH parallelized Both RCH and Trx. parallelized

WiPPETsignal Studies

BER Plots ØHow “cool” is the interference cancellation after all?

1.8 2.4

Ø Antenna diversityØ 10 users/cellØ Perfect power

controlØ No codingØ 64 SFØ Doppler 80Hz

WiPPETsignal Studies

Feasibility Study of SIR targetØ 10 and 20 users (doppler 80Hz) in a single cell (1200 m dia.) ØMax Tx.power 30dBm (1 Watt)

Average Transmitted power Average SIR

Higher targets cause mobiles to reach Tx.power limits

WiPPETsignal Studies

WiPPETsignal Studies

ØUsers with low SIRso Farther away from BSo Have very bad BERo Cause bad average BER

4dB feasible target SIR

10dB infeasible target SIR

Target SIR

BE

R

Feasibility Study of SIR target

Soft HandoffWiPPETsignal Studies

Average Transmitted powerFraction in soft handoff

ØAs soft handoff range (H) increaseso Lower Tx. and Rx. Powers, higher fraction in

soft handoffo More effective power control !!

Soft Handoff

ØAs H increaseso Average SIR closer to target SIRo Average BER decreases although avg SIR also decreases !!o “Bad” users are being handled by “more capable” BSo Macro diversity

Average SIR Average BER gains

WiPPETsignal Studies

Conclusions and Future

ConclusionsConclusions & Future

ØSimulation model validated

ØBER Plots, Feasibility studies, Soft handoff gains

ØRobust setup for WCDMA simulations

ØFully scalable, object oriented (easy to replace transceiver)

ØEasy to monitor during runtime (integrated with DATUM)

ØFast and parallel

ØSolaris and Linux implementations over SSF

FutureConclusions & Future

ØOuter loop and Joint IC and power control

ØExtensions to cdma2000, MC-CDMA

ØCall admission and load activity factors, Downlink data

ØMore realistic models of geography and propagation

ØHigher rate systems, Turbo codes

ØAdaptive antenna processing, Pulse shaping