A programmable transceiver structure of multi-rate OFDM-CDMA for wireless multimedia communications

A Programmable Transceiver Structure of Multi-rate OFDM-CDMA for

Wireless Multimedia Communications

Po-Wei Fu

Department

of

Electronical Engineering and

Graduate Institute

of

Communication Engineering

National Taiwan University, Taipei, Taiwan, R.O.C.

powei@santos.ee.ntu.edu.tw

Abstract

TIVO multi-rate transmission schemes, multi-code (MC) and

variuhle-spreading-length (VSL) code, for realizing niultiniediu communications on three @pes of OFDM-CDMA systems are proposed. We integrate them into a programmable structure such that the operation can be controlled and adjusted by system parameters and thus the transceiver can be used in dif'lirrent systems without chan&g the fundamental hardware and sofmare architecture, which serves the trend of sojtware-radio for ,fiture application.

1. Introduction

Multimedia communication will be the main stream in the future communication services and it introduces challenges in effective transmission. As code-division multiple-access (CDMA) being utilized for the third generation and future communication systems, it is an important subject to realize multimedia services

based on CDMA. Based on previous research, Multi-Code (MC)

and Various-Spreading-Length (VSL) access schemes are the two

most fundamental and widely applied multi-rate schemes [l]. On

the other hand, Orthogonal-Frequency-Division-Multiplexing (OFDM) has been widely used in high-speed digital

communications, which can be implemented efficiently by the

digital technique of Fast-Fourier-Transform (FFT)[2] and known

as a solution to combat the problem of the highly hostile mobile chaunels in high-speed transmissions. Furthermore, combining CDMA and OFDM results in finer partition of radio resource, which makes the resource allocation more effective. Therefore,

OFDM-CDMA for multimedia applications is an attractive

candidate for the 4th generation wireless communication systems

and its realization by a flexible sofnn-are-defined architecture is of

significant interests. Recent years, multiple access schemes based on the combination of CDMA and OFDM have been proposed,

and they can be generally divided into three types, namely MC-

CDMA, MC-DS-CDMA, and MT-CDMA respectively [3]. To

realize multi-data-rate transmission, we separately proposed two

transmission methods, based on MC and VSL strategies, for the

three OFDM-CDMA scenarios. Besides, we design a transceiver architecture accommodating these six multi-rate OFDM-CDMA R.O.C. and by the Integrated Progmmable Communications, Inc.

*This research is supported in part by Ministry of Education, Taiwan,

Kwang-Cheng Chen

Department of Electrical Engineering and

Graduate Institute

of Communication Engineering

National Taiwan University, Taipei, Taiwan, R.O.C.

chenkc@cc.ee.ntu.edu.tw

I942

scenarios and show its programmability such that system operation can be controlled and reconfigured by adjusting parameters in software. It supports the accommodation in different systems with only one fundamental hardware and software architecture.

2.

Multi-Rate OFDM-CDMA Systems

Assume there is a basic data rate supported in systems and the data rate o f each user is an integer multiple m of the basic data

rate, said a user with rate m. The detailed structures of the multi- rate (MR)-OFDM-CDMA systems are described in the following.

In transmitting aspects,

(1) Multi-rate M C - C D U

a.MC access

The data stream of a user with rate /n is first multiplexed into m different streams with basic rate and then

each is

treated as an individual (effective) user with individual spreading codes. After

multiplexing, each stream is serial-to-parallel (UP) converted to P

outputs, where the number of sub-carriers in the system depends

on P. For frequency-domain spreading, symbols on each output is

copied iuto F branches, where F is the constant spreading factor

o f the spread spectrum operation in the system and the signal at

each sub-stream is then multiplied by the corresponding bit of the

spreading codes. Thus, there are

PF

parallel outputs of each

effective user after the frequency domain spreading. After

combining all the PF parallel outputs from other users, they are

transmitted by PF orthogonal carriers correspoudiigly, where

Inverse Discrete Fourier Transform (LDFT) can perform this

modulation on orthogonal carriers equivalently. Label the

effective users as user 1,2,.

.

.,

and the transmitted BPSK signal of a system containing IM data rates in baseband is

x ( f ) =

-&kbbAck,f?J2? k=l PI (I

0

I t

I

PT, ,

(1) U

where K =

mK,

is the number of total effective users, b b is

the pth symbol of the kth effective user, and AL is the transmitted

amplitude of the kth effective user. Af

fp,,+,

-

,f,,f

= ,&9,

where

T,

is the symbol duration of the data stream with basic

m=l

then..spxac..by. the..same..spreading ..codes . and transmitted .via..

different sub-carriers. Thus, the number of sub-carriers in transmission is generally PL. The transmitted signal is,

x(t)= ~ ~ ~ , b , , ~ c ~ q ( t

- JT,

&ej2@+', 0 < t <

PT.,

(3)

MK, is the number of effective users, and p(t

)

is

. . .

K P F

k=l pi J = I /=I

M

where

K

=

the unit-rectangular function with duration T,, and

Af

=

F/

m=l

P T s '

rate before S/P conversion.-ew E fk1)denotes t h e m bit ofthe

spreading codes used by the kth effective user.

u.ku* -,I

R g Ila) Mulh-Code Access ofMuln-Xa~c UC-CDM.4

b. VSL access

The data stream from a user with rate in is directly SIP

converted into Pnr sub-streams, where the number of sub-carriers

in the system is still PF as in MC access. Symbols on each sub-

stream are copied into F/m branches and then the symbols on each

branch are multiplied by the corresponding bit of spreading codes. Note that F/m should be chosen as an integer in our system design.

Regardless any dare rate, there are totally PF parallel signal

outputs after such fiequency domain spreading. Combining all the

PF parallel signals From other users, they are transmitted by

orthogonal carriers. The bandwidth of each sub-carrier and the

overall occupied bandwidth are the same as in MC access. The

transmitted signal in a VSL system containing M data rates is

: ?

rfknr.uaj+r P

Yiy 2(.) Mdh-Cndr Ai-m*l U T M r - I I I C W A

b. VSL access

Regardless any date rate, the data stream of each user is

directly S/P converted into P sub-streams. If the user is with rate

mi, each sub-stream is spread with factor F/nz by cyclically

multiplying the same spreading codes. Combining all the

P

parallel signals form other users, they are trausmitted via

orthogonal carriers. Also note that if the transmission diversity is expended with a factor L, each sub-stream before spreading stage

should be copied into L identical branches and these data-streams

fkom the same user are then spread by the same spreading codes and transmitted by different sub-carriers The number of sub-

carriers in transmission is still generally

PL.

The trausmitted

signal is

m=1 klp 4 I

where

K,,,

denotes the number of users with rare m, h,, is thepth

symbol of the kth user with rate m, Ad is the transmitted

amplitude, and Af

=

.

c , , ~ E {k 1) denotes the@ bit of

the spreading codes used by the kth user with rate m.

( r s n n w x u m n ) 0

s t <

PTs

7 (4)

where K,,, is the number of users with rate m and Af =

ypT,.

I@) \inahle-SpmAvylengdi Axes o f M bRate MC CDMA (2) Mirlti-rate MC-DS-CDMA

a. MC access

The data stream of rate ni is first multiplexed into m different

streams with basic rate and each is treated as an individual

(effective) user with individual spreading codes. Each stream is then S P converted into

P

parallel sub-streams, where

P

is the

number of sub-carriers in transmission. At the spectrum-spreadiug

stage, the sub-streams from the same effective user are spread by

the same spreading codes with factor F via cyclically multiplying

the codes. Combining all the

P

parallel spread signals fiom other

effective users, they are transmitted by orthogonal carriers correspondingly. Note that if the system adopts the strategy in [5] to expend the transmission diversity with a factor L, each sub-

stream before spreading stage should be copied into L identical

branches and these data-streams from the same effective user are

U

P

P

Fig a%) WUDleSprcadlng-Lengtl~ A-* nf MC-DS-CDUA

(3) Multi-rate MT-CDMA a.MCaccess

The data stream of rate m is first multiplexed into m different

streams with basic rate and each is treated as an individual

(effective) user with individual spreading codes. Each streams is

then SP converted to P outputs, where P is the number of sub-

carriers in the system. The

P

parallel outputs i?om the same

effective user are spread with factor F by identical spreading

codes via cyclical multiplying the codes. Unlike in MC-DS-

CDMA, &er combining all the

P

spread signals from other users,

they are transmitted by carriers whose implicit orthogouality

exists corresponding to the signals before spreading. NOC-IDFT

(see Appeudix) performs this modulation equivalently instead of regular IDFT. The transmitted signal is

M

where K = CtnK,,, is the number of total effective users and

,H=l

rffccli\c wr8

. .

E U ~ l t v r u u r j

--=

_{ELTall". "=/+/} Fig. 3 ~ ) Multi-Codc AceersofMTCDMA

b. VSL access

The data stream from a user regardless any rate is SP

converted into

P

parallel sub-streams, where

P

is the number of

sub-carriers in the systems. Signal at each sub-stream is spread with factor Fhn by cyclical multiplying the spreading codes. Afier combining all the

P

spread signals &om other users, they are transmitted by carriers, where the implicit orthogouality exists between signals before spreading. The transmitted signal is

m=1 k=l g=I p=l

0 5 t 5

PT,

,

(6)

where K,, is the number of users with rate n1 aud

Af

=

)/pT,

. c.,.c.

Low-Raw w i

7

P

Rg. 3(b) V m a b l e - S ~ m g - L m @ b Accers of MT-CDMA

Tn receiving aspect, a general structure is illustrated as Fig. 4.

In general, the receiver perfom inverse functions corresponding

to those in the transmitter.

Fig 4 TI= geeneirl receiving )tmcwc of MR-OFDM-CDM4

The detailed architectures and functions of different scenarios are

described in the following sections for both transmitter and receiver in programmable design.

3. Programmable Architecture

Programmable OFDM-CDMA transmitter

Two multi-rate access strategies and three OFDM-CDMA

methods result in totally six multi-rate OFDM-CDMA scenarios. The proposed architecture of programmable transmitter is depicted in Fig. 5 , which also shows the control functions, F I-F6, that should be executed in operation. Each function can be adjusted by some parameters, which includes: choice of access scheme, choice of adopted OFDM-CDMA scheme, data rates, and the spreading factors of each data stream. These fuuctions are defined as:

F1: It determines the number of enabled Main Branches

(MB).

It

is executed on a multiplexer after receiving the data stream from a

user. If the system is operated on multi-code access mode, it multiplexes the data stream into m sub-streams by enabling m

branches. If it operates on VSL mode, just forward the data

stream on one branch.

F2: It determines how many Parallel Output Branches (POB)

correspouding to the SE' ratio will be enabled. Only under the

scenario of VSL-MC-CDMA, POB is set to be mP, otherwise,

POB is P .

F3: Function 3 determines the number of enabled Sub-Branches

(SB), which is a copy of the former stream, of each POB and the type of spreading. If the scenario is MC-CDMA, the enabled SB

is F for multi-code access and Flm for VSL access. If the scenario

is MC-DS-CDMA and MT-CDMA, SB is L and 1 respectively for

both access methods. Cyclical multiplying by corresponding

codes on each SB is set for MC-DS-CDMA and MT-CDMA. In

MC-CDMA mode, signal on each SB is spread by multiplying

one corresponding bit of spreading codes.

F4: To satisfy the required number of parallel inputs for JFFT

operation, fuuction 4 pads zeros accordiug to the selected scenario,

which results in regular padding for regular IFFT and circular-

shifi padding for NOC-TFFT. The circular period is F in MC

mode and F/m in VSL mode. In example of radix-2 algorithm, the

number of Padded Zeros (PZ) should be 2r'0g1.vb1 -Nh in regular

IFFT, where Nb denotes the number of parallel inputs. For NOC-

E F T operation in MT-CDMA modes, PZ should be

F(2r1082 N h l

-

N ,

)

and allocated in circular-shift type.

F5: Function 5 decides the number of operation points (OP) in

IFFT for each scenario. OP equals the number of outputs in F4.

F6: Function 6 determines the enabled input branches of the PIS converter and the conversion ratio at different scenarios.

- J

Table 1 summarizes the parameters of Fl-F5 in different

scenarios, the operation period in F6, and the resultant bandwidth.

The involved notations are defined as:

MC: multi-code, VSL Variable-Spreading-Length

m: data rate in unit of basic rate

F: spreading factor of a basic data-rate user

L : the diversity expansion of MC-DS-CDMA

CNS: constant bit multiplying, CYC: cyclical multiplying

A: MC-CDMA, B: MC-DS-CDMA, C: MT-CDMA

'The oolauw 7 is delinrd in srcuon 4 for hwoducing -pard lime ai each "io.

According to the scenario selection, appropriate parameters and

subroutines are chosen and adjusted to perform these functions by

micro-controller or by DSP. Briefly speaking, these parameters

determine the number of enabled hardware branches, the diversity

in transmission, zero padding, the time-domain or frequency-

domain spreading, and the spectrum profile.

Pi-ogrankable OFDM-CDMA receiver

The same as transmitting, there are six receiving variations

should be accommodated for six system scenarios. A software- based Rake receiver is illustrated in Fig. 6. Due to the principle of OFDM transmission, the number of sub-carrier is generally

selected that signals on each sub-band suffer fiom frequency-

nonselective fading. Therefore, only one finger is sufficient for

most cases. However, there may be some situations that the fading in sub-bands is hard to be maintained Frequency-

nonselective. Thus, the number of tumed-on fingers could be a

pre-defined value

or

is controlled by the result of channel

estimation to combat multi-path effect. The delays between fingers are also adjusted according to the estimation result. Fig.7 depicts the programmable structure of the fingers, where eight

functions, FI-F8, are defined. The decision of symbol detection

could be used for next step synchronization if the precision is high

enough. These hnctions are defined as:

F1: It controls the sample rate of the received signals according to

the selected scenario, including the required number of input in

the following FFT.

F2: Function 2 determines the ratio of S/P conversion and the

number of parallel outputs (PO) according to the selected scenario.

F3: Function 3 decides which type of FFT will be executed. It

consists of two sub-functions where the fmt performs circular-

shift zero padding for MT-CDMA modes and the second

determines the operation points of FFT device, General FFT

operates in NOC form for MT-CDMA modes and in regular form

for MC-CDMA and MC-DS-CDMA modes by disabling the

fxst

sub-function. (See Appendix)

F4: Function 4 controls the filtering of samples after FFT to

discard the signals out of the sub-bands and determines the

enabled tapped-delay-lines. The pass window (PSW) of zero

forcing equals the number of sub-carriers used in transmission.

F5: Function 5 is executed in the dispreading stage such that each

output of F4 is despread according to the spreading type (constant

or cyclic type) of the selected scenario by multiplying the

corresponding codes. The despreading structure is a tapped-delay-

line (TDL) form and the number of enabled TDL equals ZFW in

F4. F5 determines the number of enabled tap (ET) and the time

spacing (TS) of each tap. For MC-CDMA modes, only one tap is

needed and each line corresponds to one bit of the spreading

codes. For MC-DS-CDMA and MT-CDMA modes, the number

of taps in each line equals the spreading factor.

F6: Functions 6 determines the sampling rate (SR) for taking

samples on each line at the dispreading stage. The sampling rate

equals the OFDM symbol duration.

F7: Based on the collection of samples from each line of all

fingers, the detector makes symbol decision by combining these diversities. Different combining methods could be applicable in this stage by DSP.

FE: It controls the de-multiplexing in multi-code mode to reconstruct the original data sequence fkom the effective streams.

Clunnel E~umarton **IC

4.

Implementation Issues

The software architecture of the proposed programmable multi- rate OFDM-CDMA transceiver can be realized by the general

hardware structure in [4] as an extension. Different scenarios

have different requirement of bandwidth and sampling rate. Due

to the accommodation of the six multi-rate OFDM-CDMA

scenarios, it forces the specification of the D/A converter and the

low-pass filter should satisfy all the requirements. Thus, the D/A

converter should support input rate higher than any possible sampling rate and the bandwidth of the low-pass filter should accommodate all scenarios. Another practical issue is the peak-to-

average power (PAP) ratio problem, which causes the inefficiency

of power amplifiers i11 RF. The PAP problem exists inherently iu

OFDM systems and can be reduced by techniques such as signal distortion, error correcting, and scramblkg [6]. However, the introducing of multi-rate traffjc challenges the power AMP severer. Among these scenarios, multi-code access may cause

larger instant power due to its concept of parallel transmission.

Therefore, we should select a power AMP whose linear range

could accommodate the largest requirement o f multi-rate

transmission under the aid of PAP reduction mechanisms.

To eliminate the der-symbol interference and inter-carrier

interference, a guard time

T,

which is larger than the multi-path

spread of channels, should be added on each OFDM symbol after

P/S conversion in the transmitter and the corresponding removing

should exist before F1 in the receiver. ARer cyclic guard time

extension, the OFDM symbol duration Twill be PTs+Tg in MC-

CDMA, and PTs/F+Tg in MC-DS-CDMA and MT-CDMA. In addition, to improve the performance, some techniques such as

forward error control (FEC) and interleaving could be easily

added to this structure without any difficulty.

Although our design focuses on multi-rate applications, this architecture is backward conipatible to single rate OFDM-CDMA

systems by setting m=l, to DWCDMA systems by setting

P=L=l

in MC-DS-CDMA mode, and to conventional OFDM systems by

setting F=L=I in single user case. Due to the occupied bandwidth

of each OFDM-CDMA systems is kept fixed for users of any rate

in both MC access and VSL access modes, multi-rate applications

do not increase the requirement on some hardware devices, such

as the sampling rate of A/D and the bandwidth of low-pass filter.

Processing delay is a major challenge in OFDM systems,

especially sensitive in real-time applications. In fact, this

programmable transceiver supports the possibility that the number

of subcarriers could be adjusted dynamically according to the

channel conditions such that the least sub-carriers attain frequency-nonselective fading at each sub-channel. In the same

way, delay-sensitive data transmission is realizable by reducing

the number of sub-carriers with the aid of more fingers in Rake receiving.

5.

Conclusion

We developed a transceiver architecture o f multi-rate OFDM-

CDMA systems and showed its programmability such that the general system can operate under different scenarios with a common hardware structure and reconfigure by software implementation. The hture work shall be analyzing the performance of each scenario and figuring out its relationship with these adjustable parameters such that optimal resource

programming is realizable for future wireless multimedia

applications.

References

[I] T. Ottosson and A. Svensson, “Multi-rate schemes in DS/CDMA

systems”, Proc. IEEE Vehic. Tech. Cor$, pp. 1006-1010, 1995.

[2] K. F a d , S. Kaiser and M. Schiiell, “A flexible and high-performance cellular mobile communication system based on orthogonal multi-carrier

SSM.4,” Wireless Personal Communications, vol. 2, No. 1, pp. 121-144,

1995.

[3] Shinsuke Ham, Ranijee Prasad, “Oveniew of multicarrier CDMA,” IEEE Conirniinication Magazine, pp. 126-133, Dec. 1997.

[4] K. C. Clen and S.T. Wu, “A Programmable architecture for OFDM-

CDMA,” IEEE Ctm”icntion Magm’ne, pp. 76-82, Nov. 1999.

[ 5 ] Sourour, EA.; Nukuguwu, bf “Pcrfomiancc of Orthogonal Multicanicr CDMA in a Multi-path Fading Channcl”, IEEE Trunsactions on Communications, pp. 356 -361, March 1996.

[6] R. Van. Nee, and R Prasad, OFDM Jbr wireless nrultimediu contwzmicutiorrs, Artech House, 2000.

Appendix

NOC-IFFT

For the kth user, the transmitted signals after modulation is

P F I.‘

x k ( t ) = ~ b b ~ c l i f ~ ( t - j ~ . ) e J 2 “ ‘ = c x k , ( t ) , O 5 t 5 T,(A.l)

p = l /=I f = l

where Af

= f,+,

- f,

=

I/.Ts

and the discrete equivalent form

of xu (t)is: for n=l

-N,

where d,

b ] = b , c w .

Dejinifion:

N point NOC-IDFT of a sequence s [ i ] , i=l-N, where

N I

N ,

=

F E N

is generated bv

( I ) , At lth sub-period, for I = I . 2

,...,

F. 2

- ( I - I ) N , ] ,

(I

- lws

+

1

I

i

I

0

,

otheiwise

where N , = N / F , n=I-N, and I-I-F.

Therefore, let N=FP and it can be easily shown that

where F{;N) denotes N-point regular DFT. Therefore, circular- shift zero padding aud regular FFT can implement NOC-FFT. Disabling of F4 goes back to regular EFT.

For receiving in MT-CDMA systems as [ 5 ] , we define the NOC-

FFT corresponding to NOC-TFFT. Let r ’ [nl n = 1

-

N, be the circular-shift zero padding for r[n],

Dqfinition:

Fpoints NOC-DFT oftfn],

n=l-N,

,is

s,

[i3=

F,,,.{r[I].

.

. r [ X J F , f }

=

2

r s

[n]..p(-

j 2 X

i%)

, n=(j.-l),V,+I

i=I-N, , where

iV,

= Ar I F E

N

.

By the same principle, it can be shown that circular-shift zero padding and regular FFT can also implement NOC-FFT.

Therefore, sampling

4t)

with rate N/FTc and taking F-point

NOC-FFT with period Twill get the desired result. I