A Study On Channel Estimation Methods For Mc-cdma Systems

A Study on Channel Estimation Methods for MC-CDMA Systems Atsushi Nagate, Hiroyoshi Masui, and Temya Fujii Information Communication Lab., R&D, Product & Services Supplier Unit, Japan Telecom Co., Ltd E-mail: [email protected] In this paper, we propose a novel channel estimation method for time-domain spreading multi-carrier code division multiple access (MC-CDMA) systems. In time-domain spreading MC-CDMA systems, code-multiplexed pilot symbols are transmitted consecutively in both time and frequency domains. I n our proposed method, weighted pilot symbols in both domains are coherently added to improve signal-to-noise ratio (SNR) of pilot symbols in channel estimation. However, the channel estimation quality degrades when the weights ofpilot symbols are not selected appropriately according to the changes of propagation environments. We propose a method which the optimum weighting factors for pilot symbols are adaptively selected according to the changes of propagation environments, by measuring maximum Doppler frequency for the time domain and delay spread for the frequency domain at MS (Mobile Station). We evaluate our proposed method by computer simulation, and clarify improvement of channel estimation accuracy. Abstract-

Index

Terms-MC-CDMA,

channel estimation, weight

I. INTRODUCTION

M

ULTI-CARRIERcode division multiple access (MC-CDMA), based on a combination of CDMA and orthogonal frequency division multiplexing (OFDM), is a major candidate of transmission technology for new generation mobile communication systems. In MC-CDMA systems, data symbols can be spread in frequency or time domain by using a given spreading code [ 1-61, In time-domain spreading MC-CDMA system, data and pilot symbols are spread in the time domain and code-multiplexed [5,6]. Therefore, we can perform accurate channel estimation even in fast fading channels by utilizing successive pilot symbols in both time and frequency domains. In DS-CDMA systems, it is general that some pilot symbols in the time domain are added coherently to improve signal-to-noise ratio (SNR) of pilot symbols in channel estimation. In time-domain spreading MC-CDMA systems, code-multiplexed pilot symbols are transmitted consecutively in both time and frequency domains. Therefore, we can expeci an improvement of SNR by adding pilot symbols of both domains coherently. However, the correlation among pilot symbols in both domains varies according to propagation environments such as

maximum Doppler frequency and delay spread. Therefore, channel estimation quality degrades inversely, when pilot symbols for the coherent addition are not selected appropriately. In this paper, we propose a novel channel estimation method for time-domain spreading MC-CDMA systems which adaptively change the weights for pilot symbols used for the coherent addition according to propagation environments, by measuring changes in the time domain (maximum Doppler frequency) and the frequcncy domain (delay spread) at Mobile Station (MS).

11. TIME-DOMAIN SPREADING MC-CDMA SYSTEMS

The frame format of time-domain spreading MC-CDMA systems is shown in Figure 1. On each sub-carrier, data and pilot symbols are spread in the time domain, and code-multiplexed. A guard interval, which is a copy of the tail part of the following OFDM symbol, is inserted between OFDM symbols to avoid inter symbol interference (ISI) caused by multipath fading. The transmitter structure is shown in Figure 2. The complex equivalent low-pass transmitted signal of mth sample in xth chip’s OFDM symbol is written as iN-1 N,-I

%,,” = i = o ==n

d,,< . wc,x ’e x P ( m

(1)

where N, is the number of sub-carriers, N, is the number of multiplexed codes, d,., is the data symbol of sub-camer k for cth user, Wc,xis the xth chip of the spreading code for cth user. Here, the spreading code for 0th user is used for pilot symbols. The receiver structure is shown in Figure 3. Here, the compensation of each received data symbol is conducted after the received data symbol is despread. For the compensation, the channel estimate is obtained by coherent addition of despread pilot symbols.

111. CONVENTIONAL CHANNEL ESTiMATlON METHODS

In OFDM and frequency-domain spreading MC-CDMA systems, pilot symbols are time-multiplexed on all sub-carriers. Here, channel estimation is conducted by coherent addition of pilot symbols on appropriate sub-caniers [7].This means the

d

S

c_c

Prore$$ingGain

I

OFDM symbol

Fig. I Frame format Fig. 3 Receiver structure

‘1 P E

I-

Fig. 2 Transmitter structure

selection of pilot symbols is very important for accurate channel estimation. In OFDM systems, fading characteristics among snb-caniers and among consecutive time-domain symbols are highly correlated. Therefore, in the systems, pilot symbols on adjacent sub-camers and adjacent time slots can be coherently added for channel estimation. On the other hand, in MC-CDMA systems, the correlation of fading characteristics among pilot symbols becomes lower in proportion to the length of the spreading code Pc. Therefore, when we select the fixed number of pilot symbols, the accuracy of channel estimation may degrade in case the correlation among pilot symbols is low. Considering this degradation, the number of pilot symbols for channel estimation has been strictly limited for the worst propagation environment. However, in this case, accuracy of channel estimation degrades by the limit in static environments.

1v. PROPOSED CHANNEL ESTIMATION METHOD We propose a channel estimation method that keeps accuracy in different propagation environments for time-domain spreading MC-CDMA systems. In this method, pilot symbols in both frequency and time domains are selected for coherent addition. Here, pilot symbols in the fixed area, which mean the fixed number of pilot symbols, are used. To keep accuracy of channel estimation, different weight is given to each pilot symbol in the area according to propagation environments. This method is equivalent to changing the number of pilot symbols adaptively. And, its process is easier because pilot symbols used

v

2Ni+I

frequency Fig. 4 Weighted pilot symbols for coherent addition

for coherent addition are fixed. As for the area of pilot symbols, we investigate two types of channel estimation methods. One is “stored method” which selects pilot symbols that have already been received and will be received as shown in Figure 4. Here, each small block means a despread pilot symbol. Each variable in the blocks means the weight of the pilot symbol. The central block surrounded with a thick line (hereafter target symbol) is where the channel estimate is applied to compensate data symbols. In this method, we need to memorize the received data symbols until all pilot symbols for the channel estimation are received. Then, we also propose “real-time method which selects only pilot symbols that have been received. In this method, the memory for received data symbols is not necessary, and we can demodulate data symbols in real time. In the stored method, accuracy of the channel estimation is better than that of the real-time method, because it can use more pilot symbols that have high correlation. A. Stored Method When we use N , and N,pilot symbols in one side of each time and frequency domain, (2N,+l)(ZNfiI) pilot symbols are used for the channel estimation. Considering the facts that the correlation becomes lower in proportion to the distance between symbols, weights are given according to the distance. Here, we denote 1,and 1, as the weight factor (hereafter

2102

5 path

t,-

A

time

/path Separation A r

1

Fig 5 Path model

forgetting factor) in each time and frequency domain as shown in Figure 4. The channel estimate of t,th symbol on 5 t h sub-camer is written as

L Y

m

where r,,, is the complex symbol of tth received pilot symbol on /th sub-camer, andp is the complex symbol of the transmitted pilot symbol.

B. R e d t i m e Method In the real-time method, only pilot symbols that have already been received are used for the channel estimation. Here, we use N,+l pilot symbols in the time domain, ZN,+l pilot symbols in the frequency domain, and (N,+1)(2N,+l) pilot symbols in total. The channel estimate obtained bv this method is written as

v.

SIMULATIONS AND DISCUSSIONS

We evaluate our proposed method with computer simulation. In this paper, we show results of only the stored method. The parameters used in the simulation are shown in Table 1. The processing gain, Pc, is 16 for both pilot and data symbols, and 15 codes for users and one code for pilot TABLEI SIMULATION PARAMETERS I40.961MHzI 1024

Guard mtcwal length Pmccfsing gain: i',;

114 symbol lcnglh

Spreadingcode

OVSF

lrtodulatm

QPSK

16

2103

o,ol

I~

-10

~

..

.......i . 0

~

i

.........

IO

~

~

. .

~

20

Eb!N o for pilot signalldB1

(b) N,=N,=3,fu=ZOOHz. q=I.Ops, €JNu=lOdB Fig. 6 Effect ofpilol symbol's power VE. BER of user's data

symbol are used., In Figure 5, we show the path model that means delay profile as the propagation model. In this model, a 5-path Rayleigh fading channel has an exponential decay of the average received power with an equal interval of AT between adjacent paths. Its decay factor is 3dB per path, and each path is subjected to independent Rayleigh fading with maximum Doppler frequencyfD. In this model, we set the delay spread by changing the interval between paths. The longest delay path is not longer than the guard interval. We also assume that frequency synchronization and timing synchronization are ideal. And interference from other cells is not considered. In our simulation, received power of pilot symbols is not taken into account for the calculation of received EdNOof data symbols. In the following simulation, we assume that the number of pilot symbols used in one side of each time and frequency domain is N,=N,=3, that is 49 pilot symbols in total are used for coherent addition. And received EdNOequals 10dB. First, we show BER performance o f several forgetting factors, A, and Afias a function of received EdN0 of pilot symbols. Fig. 6(a) shows the result in case of maximum

I

6a;

0.1

. .. 0.011

0.2

’ 0.4

. ..

.

.

.

.

L

..

..

...

..

2

.

~

06

0.8

I

1.2

1.4

1.6

1.8

o)iN

(a) N, =3, A,= I.O,/b=

I OHz. EdNcrI OdB

i

..+. .

_--0

,

--e-

;~~~

0.01 0.2

0.4

j

:

0.6

....!...... ~

~: ~ ............~

0.8

~

I

1.2

:

1.4

:.~~~-.1 1.6

1.8

5 00

aJi4

(b) N, =3, A+.4,/o=ZWHz.

EhlN,,=lOdB Fig. 7 Relation between forgetting factors and delay spread

Doppler frequency,fu=lOHz and delay spread a,=0.25ps. Fig. 6(h) shows the result in case offu =200Hz and 4=1.0 ps. In figures, it is shown that optimal values of forgetting factors change according to propagation environments. For examplc, when of pilot symbols is IOdB, BER of user’s data becomes the best at A,= A ~ l . 0in case of Fig. 6(a), and at A= Ar0.2 in case of Fig. 6(h). Then, we clarify the relation between delay spread q and forgetting factor for the frequency domain Af HereaRer, receivedE&” ofpilot symbols is also IOdB. Fig. 7(a) shows UER performance versus UTas a parameter of 2, Here,fu is IOHz. The forgetting factor for the time domain A i s fixed at 1.0 because fu of IOHz is small. Fig. 7(h) shows BER performance when fu is 200Hz. Here, $, is fixed at 0.4 hecausefu of 200Hz is large. In these figures, it is shown that BER performance becomes better as A, becomes bigger in case of small delay spread, and UER performance becomes worse as 3, becomes bigger in case of large delay spread. Moreover, it is shown that the optimal +changes according tofu. Then, we calculate the relation betweenfu and.2,. Fig. S(a)

(b) N I = 3 , 2 ~ 0 . 4~=l.Op$, . EdN,rlOdB Fig. 8 Relation between forgelling factors and Doppler frequency

shows UER performance versus fD as a parameter of A,. Here, 0; is 0 . 2 5 ~A,~ is . fixed at 1 .O because rq of 0 . 2 5 is ~~ small. ~. 2 , Fig. 8(b) shows BER performance when 0;is 1 . 0 ~Here, is fixed at 0.4 by considering large osof 1.0ps. In these figures, it is shown that BER performance becomes better as becomes bigger in case of low maximum Doppler frequency, and BER performance becomes worse as A, becomes bigger in case of high maximum Doppler frequency. From these results, it is clear that acucracy of channel estimation can he improved by optimizing forgetting factors according to fD and 4. In Figure 9, we show a flow, which realizes our proposed methods. First, we generate a table that has optimal forgetting factors according tofu and U,. Then, MS selects the optimal forgetting factors according tofu and oswhich are measured by itself. With our proposed methods, accuracy of channel estimation can he improved, that is, BER can be improved. Nevertheless, our proposed methods are very simple because MS needs only a table with forgetting factors, and selects optimal forgetting factors according tofo >L,

2104

and q.

VI. CONCLUSION In this paper, we proposed a novel channel estimation method for time-domain spreading MC-CDMA systems. In this method, pilot symbols in both time and frequency domains are used for coherent addition. Each pilot symbol is multiplied by a weight, which is adaptively determined according to propagation environments such as maximum Doppler frequency and delay spread. Then, we clarified the optimal weighting factor for each propagation environment. Moreover, we showed a method that selects the optimal weighting factor according to maximum Doppler frequency and delay spread, which realize our proposed method. From our results, following things were clarified. -The optimal weighting factor changes drastically according to maximum Doppler frequency and delay spread -By optimizing weighting factors, BER performance is improved, in other words, transmission power of pilot symbols for the same BER performance is saved.

ACKNOWLEDGEMENT

The authors would like to thank Mr. Y. Nakano, a vice president of R&D Product & Services Supplier Unit, Japan Telecom Co., Ltd. for his helpful comments.

REFERENCES [I] N. Yep, J-P. Linnalz and G.Feltweis, "Mullicurrier CDMA in Indoor Wireless Radio Networks," in Proc. IEEE PIMRC'93, Sept. 1993, pp. 109-113. 121 K. Fazcl, and L. Papke, "On the performance of convolutionally-coded CDMAIOFDM for mobile communication system." in hoc. ~ l E E E PIMRC'93. Sept. 1993. pp. 468472. [3] A. Chouly, A. Brajal, and S. Jourdan, "Orthogonal multicarrier techniques applied to direct sequence spread spectmm CDMA system," in Roc. IEEE GLOBECOM93, Nov. 1993, pp. 1723-1728. (41 S. Hara and R. h s a d , "Design and Performance of Multicarrier CDMA System in Frequency-Selective Rayleigh Fading Channels." IEEE Trans. Veh. Technol., Vo1.48,No.9, pp.1584-1595, Sept., 1999. [SI A. Matsumoto, K. Miyoshi, M. Uerugi. and 0. Kato, "A Study on Time Domain Spreading for OFCDM," in Roc. WPMC'OZ, pp. 725-728, Oct.. 2002. [6] A. Nagate, H.Masui, and T. Fujii, "A Shldy on Channel Estimation Methods for MC-CDMA System," Technical Repon of IEICE, RCS2001-310,Mar., 2002 (in Japanese). [7] T. Onizawva, M. Miroguchi. and M. Morikura, "A Novel Channel Estimation Scheme Employing Adaptive Selection of Frequency-Domain Filters for OFDM Systems," IElCE Trans. Commun., vol. EXZ-B, No. 12, pp. 1923-1931, 1999.

2105

I+ Fig. 9 Control Algorithm

parameter tab'e for Optima' selection