Frequency domain pre-equalization for MIMO broadband CDMA communication systems

26th NATIONAL RADIO SCIENCE CONFERENCE (NRSC2009) March 17-19,2009, Faculty of Engineering, Future Univ., Egypt

Frequency Domain Pre-equalization for MIMO Broadband CDMA Communication Systems Najib A.Odhah, Kamal H . Awadallah, Moawad I. Dessoukyand Fathi E. Abd El-Samie Department of Electronics and Electrical Communications, Faculty of Electronic Engineering Menoufia University, Menouf, Egypt E-mails:[email protected][email protected] , [email protected] and [email protected]

Abstract In this paper, frequency domain pre-equalization at the base station is studied. Some different schemes are proposed and compared for single input single output (SISO) and multiple input multiple output (MIMO) broadband code division multiple access (CDMA) systems. The proposed MIMO zero forcing pre-equalization scheme cancels completely the effect of the intersymbol interference (lSI) and provides a better performance than the other traditional schemes. Our simulation results show an appreciable performance improvement by using the proposed scheme with a very low complexity at the mobile unit.

1. Introduction A very high-speed wireless access is required for the fourth-generation mobile communication systems. However, for such high-speed data transmission, the channel is severely frequency-selective due to the presence of many interfering paths with different time delays leading to intersymbol interference (lSI). A promising wireless access technique that can overcome the channel frequency selectivity and even take advantage of this selectivity to improve the transmission performance is the CDMA. Cyclic prefix CDMA (CP-CDMA) that combines the use of cyclic prefixes with frequency-domain equalization was proposed as an improved transmission method for broadband CDMA cellular systems [1, 2]. There are many techniques for supporting high data rate transmission such as the MIMO technique. Under rich multipath environments with independent multipath fading between each transmit and receive antenna pair, MIMO wireless communication systems achieve significant capacity gains over conventional single antenna systems. MIMO techniques are commonly divided into two classes; space-time codes and spatial multiplexing [3, 4]. The former includes space-time block codes and space-time trellis codes, while the best-known example of the latter is the Bell labs Layered Space Time architecture (BLAST) [5, 6]. These two approaches have different motivations. The former is derived from the earlier transmit diversity schemes. Hence, its main motivation is to increase diversity, and thus improve the robustness of the communications link. The latter's main objective is to increase the capacity and spectral efficiency of the communication link [4]. Broadband transmission suffers from lSI and channel equalization must be applied to overcome this problem. However, channel equalization in broadband MIMO systems is very complex due to the superposition of all of the transmitted streams at each receive antenna. The complexity of the equalization process can be mitigated by performing equalization in the frequency-domain at the receiver. The main objective of this paper is to increase the capacity and spectral efficiency of the downlink CP-CDMA system. Another objective is to reduce the complexity of the MIMO receiver by using pre-equalization techniques. In this paper, an efficient low complexity pre-equalization MIMO transceiver scheme is studied and compared with the other traditional schemes. These schemes employ pre-equalization techniques to suppress the interference caused by the multipath fading channel. The efficiency of the proposed scheme lies in the frequency domain implementation of all filters in the transmitter and the low complexity of the receiver due to moving the equalization into the transmitter. The rest of this paper is organized as follows. Section 2 describes the frequency domain equalization for downlink MIMO CP-CDMA system. In Section 3, frequency domain Pre-equalization for SISO CP-CDMA and MIMO CP-CDMA systems is discussed and the two systems are compared. The simulation results and concluding remarks are presented in sections 4 and 5, respectively. Notations: Throughout the paper, ( ) H, ( ) T, and ( l are used to denote complex conjugate transpose of a matrix, transpose of a matrix, and inverse of a matrix, respectively. Vectors and matrices are represented in

r

26th NATIONAL RADIO SCIENCE CONFERENCE, NRSC'2oo9 Future University, 5th Compound, New Cairo, Egypt, March17- 19,2009

26th NATIONAL RADIO SCIENCE CONFERENCE (NRSC2009) March 17-19,2009, Faculty of Engineering, Future Univ., Egypt boldface.

2. MIMO Frequency Domain Equalization The architecture of the MIMO CP-CDMA system employing VBLAST with K users is depicted in Figure (1) [7, 8]. In this system, three types of interference should be suppressed at the mobile station; the multiple access interference (MAl), the interantenna interference (IAI) which exists between different base station antennas and the intersymbol interference (lSI) resulting from the multipath distortion. At the transmitter, each user's data are demultiplexed to N, substreams by the V-BLAST demultiplexer. Then the substreams of each user are spread by the same spreading code. After spreading, the resulting signal is scrambled using a complex scrambling sequence and a cyclic prefix of N cp chips is added at the beginning of each block to form a transmit block. The resulting substreams are transmitted through N, antennas. The scrambling code is used to prevent intercell interference and the CP is used to prevent interblock interference. The mth (m = O... M- 1) sample of the received signal at thejth receive antenna (1 ~j ~ N r) can be expressed as [6]: Nt

L

rj(m) == LLdi(m-l)hj,i(I)+nj(m)

(1)

i=l 1=0

where di(m) represents the chips transmitted by the ith antenna, hj,ll) is the channel impulse response coefficient between the ith transmit antenna and the jth receive antenna, and nj(m) is the noise contributed by jth receive antenna at mth sample. At the receiver, the cyclic prefix is removed to prevent the interblock interference (IBI). Then, the received signals are transformed into frequency domain. The Pth frequency tone of the received signal can be written as:

R(P) == H(P)D(P) + N(P)

(2)

R(P) == [RI (P), R2 (P), ..., RNr (p)]T

(3)

D(P) == [D I (P),D2 (P), ..., DNt(P)]T

(4)

where:

and

HI,1

H I,2

HI,Nt

H 2,1

H 2 ,2

H 2,Nt

H(P)==

(5)

H Nr,2

HNr,1

HNr,Nt

Rj(P) is the Pth frequency tone at the jth receive antenna. Di(P) is the Pth frequency tone at the ith transmit R(P) is the Pth frequency tone MIMO channel matrix. Then, a series of MIMO-FDE, FFT, antenna. descrambling, despreading and hard decision is carried out to obtain N r hard decisions. The output of the MIMO FDE can be written as: "

D==WR

(6)

where

W ==

HH

MRC

(H)-l

ZF

HH (HHH + NoI)-l

LMMSE

(7)

26th NATIONALRADIOSCIENCECONFERENCE, NRSC'2009 FutureUniversity, 5th Compound, New Cairo,Egypt, March17- 19,2009

26 th NATIONAL RADIO SCIENCE CONFERENCE (NRSC2009) March 17-19, 2009, Faculty of Engineering, Future Univ., Egypt

MRC denotes maximum ratio combining or rake combining, ZF denotes zero forcing FDE, and LMMSE denotes linear minimum mean square error FDE. Finally, N, data streams are multiplexed by the V-BLAST to provide a better estimate of the desired user's data [7-10] Sprea ding co de of user 1

Scrambling code

Tx I

Insertio n of C P

Use r 1

Insertion of C P

Sp readi ng co de of user K

User K

(a) Rx I Descrambli ng an d Dispreading Desired Dat a

Rx Nr Descramb ling an d Disp reading

(b)

Figure 1. Block diagram ofMIMO CP CDMA (a) Transmitter (b) Receiver.

3. Frequency Domain Pre-equalization This Section introduces the proposed frequency domain pre-equalization schemes for downlink CP-CDMA systems. In these schemes, we assume that the channel information is perfectly known at the base station.

3.1 SISO Frequency Domain Pre-equalization We consider the downlink CP-CDMA block transmission system with a single cell and K active users over a frequency selective channel. A schematic diagram of the proposed SISO pre-equalization scheme is depicted in Figure (2) . Each user transmits BPSK information symbols. Those symbols are spread using a certain spreading code of length N. After spreading, the resulting signal is scrambled and the pre-equalization operation is performed. Finally, a cyclic prefix of N cp chips is added at the beginning of each block to form the transmit block. At the receiver, the received signal after the removal of the cyclic prefix can be formulated as:

r = H SISO W pred + n

(8)

where

26th NATIONAL RADIO SCIENCE CONFERENCE, NRSC2009 Future University, 5th Compound, New Cairo, Egypt, March 17-19, 2009

26th NATIONAL RADIO SCIENCE CONFERENCE (NRSC2009) March 17-19,2009, Faculty of Engineering, Future Univ., Egypt

W pre

HZso

Pr e - MRC

(H SISO )-1

Pre-ZF

H ZSO (H SISO H ZSO +

YsNR

0

(9)

Pre - LMMSE

I)-I

H SISO is the SISO ciculant matrix. The received data are descrambled and despread after the removal of the cyclic prefix. Finally, a hard decision is performed to produce a better estimate of the desired data. User 1

Spreading

Tx User 2 Spreading Scrambling

~

'"'0

Insert CP

'"1

CD

~

tJ

t'11

UserK Spreading

(a) Desired Data

...

Rx

Hard Decision

Descrambling and Despreading

Remove CP

(b)

Figure 2. Downlink SISO pre-equalization for CP-CDMA. (a) Transmitter. (b) Receiver. 3.2 MIMO Frequency Domain Pre-equalization: In this subsection, the proposed SISO pre-equalization scheme in section (3.1) is developed for broadband MIMO CP-CDMA systems. Figure (3) depicts the developed MIMO Pre-equalization scheme for downlink CPCDMA system. At the transmitter each user's data stream is demultiplexed into N, data streams that are spread with their own codes and summed, then scrambled by a complex scrambling code to avoid the so called intercell interference. The Nt streams are equalized in frequency domain and then the cyclic prefix is added. Finally, the resulting signals are transmitted through the MIMO channels. At the receiver, the received data are descrambled and despread after removing the cyclic prefix. Then, hard decisions are made and the data are multiplexed via a V-BLAST multiplexer to produce the useful desired data. The complexity of the proposed MIMO Pre-equalization receiver shown in Figure (3) is greatly reduced as compared to the MIMO FDE receiver shown in Figure (1).In MIMO Pre-equalization CP-CDMA, the equalization is implemented at the transmitter as follows: lL(~)

= ~pre(~)I>(~)

(10)

where:

lL(~) = [~(~), T2(~), ..., TNt (~)]T

(11)

Ti(P) is the Pth frequency tone at the ith transmit antenna

W pre =

HH

Pre-MRC

(H)-l

Pre-ZF

HH (HHH + NoI)-l

Pre-LMMSE

(12)

26th NATIONAL RADIO SCIENCE CONFERENCE, NRSC'2009 Future University,5th Compound, New Cairo, Egypt, March 17- 19,2009

26t h NATIONAL RADIO SCIENCE CONFERENCE (NRSC2009) March 17-19, 2009, Faculty of Engineering, Future Unlv., Egypt

where W pre is the pre-equalization matrix . After removal of the cyclic prefix , the received signal can be written as follows : (13) R(P) = H(P)T(P) + N(P) R (P) is the Pth frequency tone vector at the jth receive antenna. After that, descrambling, despreading and a hard decision process are made . VBLAST multiplexing is performed to provide the desired user's data as shown in Figure 3.

Spreadi ng code of user

Scramb ling code

Tx I

OIl

Use r I

"

Insertio n ofC P

o'"

Insertio n ofC P

I- .~

(/)

E

Use r K Cha nne l estima tio n

(a) Rx 1 Descram bling and Disn readin a Desired Data

Rx Nr Descramb ling and Disnreadin a

(b) Figure 3. Block diagram ofMIMO Pre-equalization CP CDMA. (a) Transmitter (b) Receiver. The main advantages of the proposed pre-equalization schemes are the high spectral efficiency, the low receiver's complexity, and the high capacity. Its low complexity comes from the frequency domain implementation of the pre-equalization at the base station and the very low complexity of the receiver due to shifting the equalization into the base station . This indicates that the proposed scheme is very suitable for downlink broadband CP-CDMA systems .

4. Simulation Results In this section, computer simulations are carried out to evaluate the performance of the proposed preequalization schemes . For the comparison purpose, the conventional equalization at the receiver is also simulated. More details of the simulation parameters are given in Table (1). All users are assigned the same power. The SUI-3 broadband wireless channel model is assumed in the simulation. The SUI-3 is one of six channel models adopted by IEEE 802.16a for evaluating broadband wireless systems in the 2-11 GHz bands [11]. It has three Rayleigh fading taps at delays of 0, 0.5 and 1 us, with relative powers of OdB, -5 dB, and -10 26th NATIONAL RADIO SCIENCE CONFERENCE, NRSC2009 Future University, 5th Compound, New Cairo, Egypt, March 17- 19,2009

26th NATIONAL RADIO SCIENCE CONFERENCE (NRSC2009) March 17-19,2009, Faculty of Engineering, Future Univ., Egypt

dB, respectively. The fading is modeled as a quasi-static fading (unchanging during an IFFT block). A. SISO System In this case, the performance of the proposed SISO pre-equalization with different equalization Table 1 Simulation parameters. Modulation

BPSK, and QPSK

Spreading codes

Walsh-Hadamard codes with processing gain 16

Multipath channel

SUI3

FFT points

P=256

Equalization

LMMSE, ZF, and MRC equalizers.

Channel estimation

Ideal channel estimation

techniques is compared to that of the SISO FDE-MMSE, SISO FDE-ZF, and the RAKE receivers. Each user transmits BPSK information symbols. Figures (4) and (5) show the BER versus the SNR for K=4, and K=16, respectively. As shown in these figures, the proposed pre-equalization with the ZF equalizer outperforms all other schemes, with low complexity. It improves the performance significantly, especially at high SNRs. It is also seen that the proposed scheme with MMSE equalization provides a better BER only at high SNR values. This is because the MMSE operation is similar to that of the ZF equalization. At low SNR, the MMSE operation is similar to that of the RAKE receiver. Figures (4) and (5) show also that the performance of the proposed scheme is independent on the number of users. On the other hand, the BER performance of all other systems degrades as the number of users increases. B. MIMO System: In this case, the performance of the proposed MIMO pre-equalization scheme is studied and compared with the other traditional schemes. The simulation environment is based on the downlink synchronous MIMO CPCDMA system, in which each user transmits QPSK information symbols. Perfect channel knowledge with N f=2, and N r= 2 is assumed. Figures (6) and (7) depict the BER as a function of the SNR at K=4 and K=16, respectively. The figures show that the proposed scheme with ZF pre-equalization achieves a remarkable gain compared to the other schemes, even with a large number of interfering users. With a BER of 10-3, the required SNR for the MIMO FDE receiver has to be 18 dB, whereas the required SNR for the proposed MIMO ZF pre-equalization is about 10 dB, which demonstrates a 8 dB improvement. As in the SISO case, the proposed MIMO pre-equalization scheme with MMSE equalization provides a better BER only at high SNR values. These figures show that the performance of the proposed MIMO ZF pre-equalization scheme does not change with number of users. This makes the proposed pre-equalization scheme suitable for high-speed wireless communications. Figures (8) and (9) give a comparison between the proposed MIMO ZF pre-equalization receiver, the MIMO LMMSE FDE receiver, and the MIMO LMMSE FDE with parallel interference cancellation (MIMO LMMSE-PIC FDE) receiver. It is clear that MIMO ZF pre-FDE scheme provides a lower BER than that of the other schemes, even with a large number of interfering users with a low complexity.

5. Conclusions In this paper, an efficient low complexity transceiver scheme for SISO and MIMO downlink CP-CDMA transmissions has been proposed and studied. This scheme employs a pre-equalization at the transmitter to suppress the interference caused by the multipath fading channel. The efficiency of the proposed scheme lies in the frequency domain implementation of all filters in the transmitter part and the low complexity of the receiver 26th NATIONALRADIOSCIENCECONFERENCE, NRSC'2009 Future University, 5th Compound, New Cairo, Egypt, March17- 19,2009

26t h NATIONAL RADIO SCIENCE CONFERENCE (NRSC2009) March 17-19, 2009, Faculty of Engineering, Future Univ., Egypt due to moving the equalization part into the transmitter. The comparison studies show that the proposed transceiver scheme offers an appreciable performance improvement with a very low receiver complexity relative to the frequency domain equalization receivers . This makes the proposed MIMO ZF pre-equalization scheme suitable for high-speed wireless communications. SF=16 ,K=4

SF=16 ,K=16

:::::::::::r:::::::::::J: :::::::::::

~----------+-----------~-----------­

------ ----- T - - - - - - - - - - - i - - - - - - - - - - - -----------T-----------i-----------J

______ l

I

_

I

0:::

0:::

W III

w

CD

10

5

Figure 5. HER vs. the SNR for different SISO schemes. K=I6.

Pre-equalization MIMO ,OPSK,V-BLAST, SUI3, K=16

Pre-equalization MIMO ,QPSK,V-BLAST, SUI3, K=4

10° r - - - - - - - - , - - - - - - , - - - - - - - - - - - - - , - - - - - - - - ,

0:::

0:::

co

co

ui

15

SNR (dB)

SNR (dB) Figure 4. HER vs. the SNR for different SISO schemes. K=4.

10

5

15

W

--e-- MRC Pre-FOE -S- MMSE Pre-FOE

-+- ZF Pre-FOE

--e-- MRC FOE

--EJ-- MMSE FOE

10-4 l'== = = = = = ='-------L.-_ ,,_ _-----L_------.J o 5 10 15 SNR (dB) Figure 6. HER vs. the SNR for different MIMO schemes. K=4.

--+-- ZF FOE

5

10

15

SNR (dB) Figure 7. HER vs. the SNR for different MIMO scheme. K=I6.

261b NATIONAL RADIO SCIENCE CONFERENCE,NRSC2009 FutureUniversity, Sib Compound, NewCairo, Egypt, March17- 19,2009

26t h NATIONAL RADIO SCIENCE CONFERENCE (NRSC2009) March 17-19, 2009, Faculty of Engineering, Future Univ., Egypt

Pre-equalization MIMO ,QPS K,V-BLAST, SUI3 ,K=16

Pre-equalization MIMO ,QPSK,V-BLAST, SUI3,K=4

========= ::!= --------4-

---------,---------,--------- ,-

--_.1I I

--+-- MIMO LMMSE equalization

::::::::: _ _ _ _ _ _ _ _ _ 1: -1 _ ---------+-_ _ _ _ _ _ _ _ _ .1

--e-- MIMO LMMSE+PIC --e-- MIMO ZF Pre-equalization

~ MIMO LMMSE equalization

---e- MIMO

~~ - - ~ - - B - MIMO

LMMSE+PIC ZF Pre-equalization

-----_ 1_---

1

- -;===1= ======== = :t ===== -:1:::: ::::::1:::::

-

0:::

0:::

llJ

llJ

co

m

10-4 '--

o

-'--5

--"--_ ,, 10

--"-15

_

SNR (dB)

Figure 8. BER vs. the SNR for different receiver schemes. K= 4.

10-4 '--

o

-'--5

--"--_-'--_ _---"--_ _----' 10 15 SNR (dB)

Figure 9. BER vs. the SNR for different receiver schemes. K= 4.

References [1] K. L. Baum, T. A. Thomas, F. W. Vook, V. Nangia, "Cyclic-Prefix CDMA : An Improved Transmission Method for Broadband DS-CDMA Cellular Systems," IEEE WCNC-2002, Orlando, FL, March 2002 . [2] F. Adach i, D. Garge, S. Takaoka, and K. Takeda, "Broadband CDMA techniques," IEEE Wireless Communs., Vol. 12, Issue 2, pp . 8-18, April 2005 . [3] V. Kuhn , Wireless Communications over MIMO Channel Application to CDMA and Mult iple Antenna Systems, John Wiley &Sons, Ltd, 2006 . [4] GJ. Foschini, "Layered space-time architecture for wireless communication in a fading environment when using multi-element antennas," Bell Labs Technical Journal, 1996. [5] S. M. Alamouti,"A simple transmit diversity technique for wireless communications," IEEE Journal Selected Areas on Communications, vol.16 , pp.1451-1458, Oct. 1998. [6] C. Z. W . Hassell, J. S. Thompson, B. Mulgrew, and P. M. Grant, "A comparison of detection algorithms including BLAST for wireless communication using multiple antennas," IEEE Personal, Indoor and Mobile Radio Commun.'OO, vol. 1, pp . 698-703 , Sept. 2000 . interference [7] F. S. Al-kamali, M. 1. Dessouky, B. M. Sallam, and F. E. El-Samie, "Frequency Domain electromagnetics Research B, cancellation for single carrier cyclic prefix CDMA systems," Progress In PIERB 3, pp . 255-269, 2008 . [8] F. S. Al-kamali, M. 1. Dessouky, B. M. Sallam, and F. E. El-Samie,"Performance Evaluation of Cyclic Prefix CDMA Systems with Frequency Domain Interference cancellation," DSP Journal, Elsevier Inc, 2008 . [9] D. Falconer et aI., "Frequency domain equalization for single-carrier broadband wireless systems," IEEE Commun. Mag ., vol. 40, pp . 58-66, April 2002 . [10] X. Zhu, and R. D. Murch, "Novel frequency doma in equalization architectures for a single carrier wireless MIMO system," IEEE VTC, vol. 2, pp. 874-878, Vancouver, Canada, Sept 2002. [11] V. Erceg et aI., "Channel Models for Fixed Wireless Applications," IEEE 802 .16a cont. IEEE 802.16.3c0l/29rl

26th NATIONAL RADIO SCIENCE CONFERENCE,NRSC2009 FutureUniversity, 5th Compound, New Cairo, Egypt, March17- 19,2009