Design of a MIMO OFDM baseband transceiver for cognitive radio system

Design of

A

MIMO OFDM Baseband

Transceiver

for Cognitive Radio System

Jui-Ping Lien, Po-An Chen and Tzi-Dar Chiueh Graduate Institute ofElectronics Engineering and

Department ofElectrical Engineering, National Taiwan University, Taipei, Taiwan

Email: pa@analog.ee.ntu.edu.tw

Abstract- As wireless communication services become more use of spectrum, the CR system needs to adapt its band-prevalent, bandwidth requirement increases dramatically. As width, carrier frequency and spectrum allocation to avoid such, the concept ofCognitive Radio (CR) has recently received systemmalfunction. Inthe

proposed

CR system, the

operating

much attention. In this paper, we present aMIMO-OFDM based

stem

ion. In

the

system, the opEri

CRsystem capable ofdynamically adjusting its system parame- frequency

ils

5GHz

UNII

band, the same as that of IEEE ters according to spectrum allocation. From system simulation, 802.1la wireless system. The targeted 5GHz UNJI band can theproposed transceiver isshownto be apromising solutionfor be divided into lower band, 5.15 - 5.35GHz, and upper band, CRapplications. 5.725-5.825GHz, and our system will use either one of them according to the overall system requirements. Based on the

I. INTRODUCTION above system requirements and OFDM design considerations, Recently, wireless communication is becoming pervasive. we define the overall system specifications.

The available spectrum, however, is getting scarce and the

conventional fixed spectrum allocation and usage has now A. System Specification become inefficient [1]. Cognitive Radio (CR) is proposed

and widely discussed to solve this dilemma [2]. The basic Summary of the major specification is shown in Table I,

idea of CR is to provide a system with the ability to sense manifesting the flexibility of the proposed system. The basic

available spectrum slots not occupied by existing users and parameter setting is from that of IEEE

802.11a,

bearing the

detect whether any primary useris demanding the bands that OFDM symbol duration of 4-,us, resulting from the sum of

the CR system currently uses. In 2003, a regulation enacted FFT and guard interval (GI) period. Since one subband is

byFCC formally initialized the commercialization of CR [3]. 20MHz and the possible number of subband allocation is

Currently, CR plays a major role in many applications to 1, 2, 4 or 8, we can get the corresponding bandwidths. As

enhance communicationperformance and/or provideadvanced sampling rate is the sameas bandwidth, differentFFT lengths

services [4]. Since CR only uses non-contiguous bands in can be obtained. The number of subcarrier used is less than

the spectrum, Orthogonal Frequency Division Multiplexing the corresponding FFT lengths since there are guard bands

(OFDM) is considered a suitable transmission technique. In to avoid out-of-band interference. Finally, to provide multiple brief, the primary target of CR is to find out available bands data rate, V-BLAST 2 x 2 MIMO scheme is implemented [5].

and change, if necessary, system parameters such as carrier

frequency, transmission bandwidth, power consumption and TABLEI

SYSTEM SPECIFICATIONS

modulationtype to achieveefficientuse of spectrum resource.

Inthis paper, a MIMO-OFDM CR system for 5GHz UNII Lower band Upper band

band to coexist with IEEE 802.11a Primary Users (PU) is Frequency band(GHz) 5.15-5.35 5.725-5.825

proposed.

The

proposed

system

can sensethechannel charac-

Subband

Number 8 4

teristics to actively modify its physical layer settings through FFTLength

64/128/256/5812

64/128/256 parameterized designsothat existing PU are notinterfered by Subcarrier Used 52-500 52-244

theCR signals. SubcarrierSpacing 312.5KHz

The rest of thepaper is organized as follows. In Section 2, FFTPeriod

302us

thesystemspecification and channel model will be introduced.

The design of the transceiver is discussed in Section 3. In

Section4, simulation of the CR system will be given. Finally, The packet format of the system is devised as shown in

Section 5 concludes this paper. Figure 1. There are 14 symbols in preamble, including 10 short and 4 long preamble. In order to support adaptivemodulation,

II. SYSTEM SPECIFICATION AND CHANNEL MODEL we design a

1-bit

EXTEND bit; when set to 1, itinforms that

There are two basic considerations in the overall physical there is an extra extended field containing more parameters

layer specification design: dynamic spectrum allocation and about the packet. Also, when the system is operating as an

channel characteristic. As the primary users change their MIMO transceiver, Alamouti code [6] is adopted in the long

preamble pattern transmitted by two antennas to estimate B. Receiver

channel frequency response. The receiver can be partitioned into three parts: initial

syn-chronization, CFO and SCO tracking loops, channelestimation

RATE EXTEND LENGTH Parity Tail ISERVICE PSDU Tail

I

Pad as well as data recovery, which also represents the operation

4bits II bit 122bits 1Ibit 16bits 6 bits I6bitsl bits Ircvr,rpeet

_Coded/IOFDM Coded/OFDM flow of the receiver. The proposed receiver architecture and

,.

VQ-BPSK,

r=112) (RATE is indicated inSIGNAL) initial synchronization flow are depicted in Figures 3 and 4,

PLCP Preamble SIGNAL SIGNAL DATA .t . .

114Symbols 1OFDM symbol Extend Several OFDM Symbols respectvely. Since OFDM system iS very sensitve to tming

____I__PLCP ____Preamble______SIGNAL_____DATA__

andfrequency deviation, thesystembegins synchronization by

oPLCPPreamble SIGNAL DATAshrprabe dlycreaon stme

14Symbols 1OFDMsymbolI Several OFDMSymbols

computing

short

preamble (SP) delay

correlation to estimate

EXTEND coarse symbol timing and coarse CFO. From Figure 4, with

bit

the autocorrelation property of SP, the coarse CFO can be

Fig. 1. Packet Format. estimated from the peak position of delay correlation in step 2 while the coarse symbol timing can be predicted from the position of the half peak value found in step 3. Similarly, the

B. Channel Model

long preamble

(LP)

,

passed

into matched

filter,

offers the

mechanism to find fine symbol timing by firstly getting the The channel model used for simulation is based on IEEE peak position of matched filter output as in step 5 and trace 802.11 TGn standard, which stipulates 6 different MIMO back in time until seeing a strong enough multipath component

channelmodels, representing various indoor environments [7]. to determine the window containing maximum energy as in

Based on these MIMO channels, we construct the equivalent step 6. Lastly, fine CFO estimation can be performed by

baseband channel modelby adding multipath Rayleigh fading, observing delay correlation result from LP, depicted in step

additive white gaussian noise (AWGN) , carrier frequency 7.

offset (CFO) and sampling clock offset (SCO) . To speed up Time Domain Frequency Domain

thesimulationprocess, we propose anapproach that considers [d 'L>

the

interpolation

needed in

modeling multipath

and SCO I CF FFTBlock sco Channel

1> * 1> 1> 1 * 1 1> 1 DerotatorI1TltTC1~~~~~~~~~~D (64/128/ Compensation «lfiEstimation

effects at the same time, which also gives a closer channel

InDot

(26/512

t) Etmio emulation to the real environments. CoarseSymbol

|EtCFO

Estimation

|s|t||mato

|0 Processing |

III. PARAMETERIZED TRANSCEIVER DESIGN Delay Integrator LowPass

E||

stimaton Tiino | Fl Modification ||

A. Transmitter

The 2 x 2 MIMO-OFDM transmitter block

diagram

is

X

---showninFigure 2. The binary data streamfirstpassesthrough Synchronization_{Synchronization} Tracking Loop

_~~~~Channel

Data Recove_Estimation 2

constellationmapperaccordingtothe modulationtypes.

After-wards, the data stream is allocated to the corresponding data Fig. 3. Blockdiagramof the baseband CRReceiver. subcarrier positions and then sentinto the IFFT block, where

preamble is inserted. Finally, the GI is added to complete Used for signal Usable

the generation of the time-domain transmitted signal. The detection, AGC 1 Short Preamble Long Preamble 1

baseband

signal

is then

up-converted and sent out through the

N/4|N/4|N/4|N/4|N/4|N/4|N/4|N/4|N/4|N/4 Gl2

N

N

transmitting antennas.

1.SPmovingaveraged 5. PeakPositionof 7.FineestimatedCFO

delay correlation Match Filter found beLPdelaycorrelation

spectrumallocation

informatIn-X

I ±1" @ oX _ 3. Halfpeakpeak

| I ~~~~~Insert_'Insert ' X i* __{H4 @ ,' value found}

Preamble

input bit

stream1 @ lConstellation > Symbol 2.IFFT ;>OInsertGI 2* Peak positionfound. 4.Derotatesignal with 6.Find the window g|Mapping | | Shaping | x (64112812561612pt) Insert GIloc1u CoarseCFO estimated coarse estimated CFO containing

* W --W--- .... ;;;;-;;F;-E;;;;--;;;;;EE;oup- maximum energy

ConstellationiiSym1bol IFFT iInsertG

Mapping

Shaping

)-3

Ij(64112812512pt)

5

setG

Fig.

4.

Operation

flow of the

proposed

receiver.

i;uput

modu~~~tion Insert When the system completes initial synchronization, it then

type

LPeablpl

Inner

Transmitter

enters into tracking mode. As shown in Figure 3, SCO

compensation, JWLS estimator [8], low pass filter, NCO form Fig. 2. Block diagram of the baseband CR transmitter. the feedback path to control CFO derotator. And the scaling block, integrator, FFT window try to set the proper FFT timing 4099

boundary so as to keep track of the SCO. In the mean time, TABLEII

ADPIESYSTEMPARAMETERS.

the channel estimation,

phase

modification

and

slicer

start to

recoverthe received data. Fromsimulation result, theaccuracy Coarse Symbol Timing Synchronization

of estimated channelfrequencyresponsedegrades moreunder Matched Filter lower SNR condition. As channel estimation is pivotal to the FFT Block_JWLS entire data recovery flow, we then smooth acquired channel SCO Compensation

frequency response through raised cosine filter, of which the Channel Estimation

time domain response is: _{Phase Modification}Equalization

if 0 < ltl <

1-OT

Slicer

(1 +COS

(|t-1

T)) if 1

2T

<

Itl

<

I0T

t0 elsewhere. correlation result to estimate the SNR value [9]. Meanwhile,

Thus, the frequency domain filtering function for smoothing matched filter is used tocomputethe channelimpulseresponse

is: forestimatingchanneldelay spread.Hardwaresharing concept

X(f) =sinc(fT)

cos(7/3f

T) is realized

by reusing

the

delay

correlation and matched filter

1

-4f

T circuits,

which

are also needed

during

initial

synchronization.

where Q is the roll-offfactor, and T is the width of the time

IV.

SIMULATION RESULT

domain window, the length of which is chosen to be that

ofguard interval because the maximum interval of statistical A System Verifcation

delay spread profile is less than the duration of GI. To

verify

the correctness of the

proposed

system, webuild Lastly, V-BLAST processing for equalization and detection a functional simulation model for our system to compare can further enhance data rate and lower bit error rate when our

design

with theoretical

performance.

Hence, we set the MIMO is applied. The linear combinational nulling concept receiver under SISO mode with

perfect

channel

estimation,

is adopted for the detection of V-BLAST scheme, which is timing synchronizationand AWGNonlychannelenvironment, outlined as follows: which enables us to

plot packet

error rate

(PER)

to see if

Step 1: Determine the optimal detection order of received thesystemperformancemeetstherequirements from 802.1 la.

signal.

Figure

5 shows that the PER curves of all four modulations

Step 2: Suppose the determined detection order is coincide with those obtained from theoretical

analysis,

indi-ki, k2,

... cating competitiveness of the proposed receiver.

Wethenpick the firststream astarget signalwhile treat other PakcetError Rate(PER) - AnalyticBPSK

streams as interference, which are later canceled

by

using

_using Analyti

~~~~~~~~~~~~~~~~Analytic

_16QAM

proper weights

and vector

multiplication

so as to

get

the

0.9

Analytic

---640M-

---

t---

---

AnatedBPSI<M

0 Simulated BPSK

detected first stream. 0.8- - --- --- SimulatedQPSK

Step

3: Subtract all the

remaining

streams

by

the detected 0 Simulated

64QAM

value we get in Step 2, then return to Step 1, continuing _0.6

-iteration of the above stepsuntil we detect all streams.

C.

Parameterized Design

QPSK

16

--QA-M

:64-QAM

Torealize aCR system, theproposed transceiver is capable BPS of adjusting the parameters of certain functional blocks as

the channel changes its characteristics. Table II lists those [[2 , ---_

receiver functional blocks in which some circuit parameters o.

1-will be

adjusted according

tothe channelconditions;the

major

o T111l

criterion is the FFT size, which can be obtained through the

SNR

persymb 25

informationprovided by the sensing block. The capability of dynamically adjustingtheseparametersindicates theprofound

reconfigurability of our system, meeting the demand for CR Fig.5. Packet error rate(PER) comparison. system.

D. Sensing block B. Simulation with/without CFO and SCO

Another major CR system requirement is the ability to as- The bit error rate (BER) comparisons between the two cases sess the characteristics or quality of the channel. In our system, of no CFO/SCO and CFO/SCO with compensation are shown we realize two functions: SNR estimation and delay spread in Figure 6. Under different types of channel, we can see the estimation, as shown in Figure 3. We use short preamble delay BER performance in the case with SCO = 13.7ppm, CFO

=0.254 Af and

compensation

for both is very close to 1 UncodedBERComparison(QPSK)

the no CFOISCO case. The simulation result

using

64

QAM

Channel ModelB

under channel model B, D, and B demonstrates the

system's

Channel.._ChannelModel....D_{Model E} effectiveness in

compensating

SCO and CFO effects.

Uncoded BERComparison (64-QAM):_{_________________________________________}with &w/oCEO/SCOC.... _0~~...

.1 ..D. ..-. _..10.

m -2 RU/fl_0. @... ...J.. .w... ..U... ..Band.. _{.7 -C O- ..7 r..}

7 .2x2 ..I.O.V..LASTsc-

--eme-.ub.. .s in

wit.C..SC(Model ...B).. .1.10 15 20 25

30---...w/o .. C...C...(Model....B)...b..No-per-

antenna--withCFO/SCO(Mode...

w/o CEC/SCO (ModelD)

LU 'M~~~~~EbN per...antennaR(dB)weUIIBad

Fi.6 opriosbtennoCOSOcs

and....CFOISCO...

wihdsgSndcanlqaitCseset.Tepooe

compensation2271case.memetricn

0~ ~~

~ ~ ~ ~ ~~~~~~~~oae

MIMO...technique...

toIM

enabe

BL

fexibe

ajsmnfdt

C.OealSse Performance.rate.according.to.the.estimated.channel.characteristics. The ToevaluatewitheCOveralModlsyte pefrane1e3s i prpoe reeie inerae sycroiato,rckn,0 n

ForU= 6, the moultondie6CF/SC QAM;hence,- th maximu

transissio typewithCFC Modulatio

of....QPSK..in...Figure....7.

From

iterorraeFperEomnceSo

h rpseeevr

the4imultio

resltweOSC sMoee thtucddBR esta)

5 1 0 1 5 20 25 30~~~~~~Ode, ov 202

more seer

Cmutipanbthwchnnels i.e.,C

model

and

B,OSC

some

[2]t.il

A.weis

F.hK.Jondal, "Specrumeig PoolsdeaingoAnnovative

Strateg

BERmdegraatini

sobeve,hwigaaERos1e.wt

fotri

thesiEnhandcmentnofSpecitru

Efficency,"eVo. 42, pp.o814,Mar

about SNR of 20 and 30 dB

underccordng

channeltimatmodelnelDcandctB,st2004.h

To evauateovrallhe

ystemperfomance

weuesi

FCC.poeT

DcketiNo. 03-322,te

"ynotchfropiatosed

trulacking, and

differnt trnsmision

tpes,

hree o whic show in [4]e DanijelaCsiabric, "ResachaOpportunitiesrdincognitiv Radio,"AN

11,to

imuate ur n TABLeeivrBeasIwIsIowrUI BWRCWinkterRedtrenatl,Jn 2004.

igan

aa rnmiso

Tnransmissionain Type Taial

dntst6ecpblt

T4 fnthRich-lscmuatteiong Wonirel Chne,nentonahffcieess

Symprposiudo

ForU= BanmduAlloationi 14 1-4 1-8ce

Signas,aSytems,AnElCtrOnis,ppE25-00M

Sp.198

DataRatei(ur 1,2,4,6)s 234U 161404 234U [6F.MoAaot,"ASmlrTasi ivriyTehiuUorWrls

2,or

4 BWemHz)ain 20mcresodst

80S160TCommunctions,

IEEEdJik

tof

SelectMediArasTin

Comm. pp. 145po1-148

transmission

type

with modulatioEsimationoffCarier requenygand

imingFrequecyFOffetCfo

ischivedwit dB

nCLUSIOanneyeB

lone

OFDM.E

Sysktem

Nover155

MuStpacth

umFading

Channl,kIE

Torans

Reont,

V. V~~~~~~~~~~~rehclr, Tehnlgyvol.4,N.12p.21-2,0auay205

Inr thsepaper wetiatpresnentsabasebanmdetransceivEr system [9] T.M.

Wchidl,D.. C.Joxda, "Rbctust

Frequency And

Invtimin

Synchro-aERchietredafori

CR

sapplication.ingorde

not tofdistur wthe niaton for OFDncmen" IEESetrans

onficomm.y,"

Vol.45, pp.161-162,Mr

respectively.

Order,"

~~~~~~~~~Dec.

1997.

prmr

Tsrasminspeaion,yp

The setru alctio of

ourcteig

iees hne,"Itratoaymoimo

sysemneds

tlotobednamcal

adused Such arrangementad lctoics p.25-0,e. 98