You are on page 1of 4

The Performance Analysis of Non-primitive Turbo codes in MIMO System over Flat Fading Channel

Mingxin Tan .the Department of Information Technology, Central Chinese Normal University, Wuhan City, 430079, China tmingxin@yahoo.com.cn

Abstract
So far almost all the research work on turbo codes focus on primitive turbo code (PTC). In multiple input multiple output (MIMO) system this paper concentrates on the research of non-primitive turbo code (NOPTC) over flat fading channel. It shows that NOPTC outperforms PTC at any size SIR. Thus NOPTC promotes both capacity and reliability of the communications in MIMO system. This paper concludes as well there is no direct proportion between the state-number and the BER performance of turbo code.

2 Channel Model
Turbo and spacetime coding [3] are two of the most explored concepts in modern-day communication theory and wireless research (in this paper MIMO and spacetime coding are interchangeably used). In less than five years since the MIMO became widely known a great variety of techniques have evolved. In radio communications system antenna arrays are used to combat various types of channel impairments [4]. An antenna array with sufficient antenna spacing can provide spatial diversity to mitigate multi-path fading. Transmitting and reception diversity can be employed to combat the effect of delay spread and cochannel interference. Spacetime processing is a way to increase the possible capacity by exploiting the multi-path nature of fading wireless environments [5], while turbo coding/processing is a way to approach the Shannon limit on channel capacity. In this paper, Turbo code is considered as a channel code, and multiple antennas at both transmitter and receiver are used to overcome the interference over Rayleigh flat fading channel. Figure 1 shows the system block diagram. We consider a wireless communication system that employs n antennas both at transmitter and receiver. Normalise the total power transmitted from all transmitter antennas to 1. Let the signal transmitted from antenna i be represented by s i . Assume that all the antennas transmit simultaneously and have same transmission period. So equation (1) below:

1 Introduction
The research on turbo codes [1] has clearly shown its advantages over the conventional convolutional codes, and it is being widely used in wireless communications. Undoubtedly, the invention of Turbo codes is a milestone in the development of channel coding / decoding theory. Compared to other coding system, the improvement in performance obtained with turbo coding is so big that, for many applications, the gap between practical systems and Shannons theoretical limit essentially closed, a situation which was probably not predicted by anyone before the advent of turbo codes. However, So far almost all the research work[2] on turbo code only focus on PTC whose encoders/decoders consist of both primitive feedback polynomials and primitive generator polynomials, and when at a lower signal to interference (SIR) turbo code has poor bit error rate (BER) performance which influence the establishment and maintenance of the communications. Therefore, turbo code will be more widely applied in data communications once its Low SIR Performance is improved over severe wireless channel. This paper analyses in MIMO system NOPTCs performance over Rayleigh flat fading channel, and discusses their merits. Here NOPTCs consist of nonprimitive component codes, including both non-primitive feedback polynomials and non-primitive generator polynomials. This paper is organized into several sections. It is arranged as follows: Section 2 describes the system diagram concerning this papers research, Section 3 gives simulation results and analysis. Section 4 concludes the paper.

(1) S = [ s1 s 2 sj sn ] ' can be used to indicate the signal vector transmitted from all antennas (where X denotes the transpose conjugate of the vector or matrix X).The information bits are encoded by channel encoder and mapped to n signal streams corresponding to the n transmit antennas. Information bits are not encoded by n encoders corresponding to n transmit antennas but encoded by single block Turbo encoder and then mapped to each antenna to form transmission frames. The signal at each receive antenna is a noisy superposition of the transmitted signals corrupted by Rayleigh fading. The coefficient h ij is the path gain from
transmit antenna i to receive antenna j. We assume that the

1-4244-0517-3/06/$20.00 2006 IEEE

system works on Rayleigh fading channel, so hij are modeled as samples of zero mean complex Gaussian random

power N 0 / 2 replaces r.m.s. amplitude a. Then R is the signal set received from all antennas: R = [r1 r2 rj rn ] ' (5) Where

N = [n1

n2

nj

nn ] '

(6)

is the noise set on receive antenna array. Hence, the wireless link of the system can be describe as below: R = HS + N (7) Usually in coding/decoding theories and applications encoders/decoders consisting of primitive polynomials secure better BER/FER (frame error rate) performance than non-primitive polynomials, including linear block code (such as cyclic code), convolutional code and turbo code. Here several kinds of forward error correction (FEC) examples are provided to show PTCs wide application. For example, Cyclic Hamming codes [7, 4, 3] was earliest explored block code several decades ago, of which 7, 4, 3 stand for code length, information length and minimum Hamming distance, respectively, has the generator g ( x ) = x 3 + x 2 + 1 , It is seen g (x) is a primitive polynomial. So is the BCH code, such as BCH code [31, 11, 5], of which 31, 11, 5 denote code length, information length and maximum number of rectifiable error bits, respectively. Here the generator [31, 11, 5] is a primitive polynomial as well [6]. As to PTC, some probable reasons and analysis of poorer BER performance have been given in reference [7]. However, it does not provide practical measures to promote BER performance at lower SIR. It is known the BER performance of communications system is poorer when at a lower SIR [8], and in everyday life we often face severe wireless communications environment of lower and moderate SIR. In a real channel, the received signal consists of a combination of attenuated, reflected, refracted, and diffracted replicas of the transmitted signal. In addition to all these factors, the channel adds noise to the signal and can cause a shift in the carrier frequency if the transmitter/receiver is moving (Doppler effect). Understanding of these effects on the signal is of importance because the performance of a radio system depends on the radio channel characteristics. The block fading channel is a widely used model for dispersive wireless channels. In this model, the channel state remains xed over a block (or frame) of a given size, There are two ways to describe the block fading channel, first: to assume that the path gains do not change during a whole frame; or second: to assume that the path gains remain constant over a fixed transmission period T which is the number of the symbols transmitted from each antenna within the period. In this paper, the simulation program uses the second approach. It will be the same as the first when T is set to the length of the frame. The performance of BER vs SIR in SISO system is

Figure 1 System Block Diagram variables with variance 0.5 per dimension, as the equation below: (2) hij = a (0.7071 x + j 0.7071 y ) Where x, y are zero mean real Gaussian variables with variance 1, r.m.s. amplitude a is a constant: in the following parts of the report we assume that a is 1. The path gains along different path fade independently of one another. So we can use: hi1 hn1 h11 (3) H = h1 j hij hnj h1n hin hnn to annotate the path gains of the whole system. Extremely, for the fast Rayleigh fading channel, we assume that h changes from one symbol period to next, but remains same during each symbol period and for adjacent symbol periods hij h is uncorrelated. We here describe the block fading channel used in this paper: to assume that the path gains remain constant over a fixed transmission period T which is the number of the symbols transmitted from each antenna within the period. The signal received by antenna j at time t can be denoted by r j :

r j = hij s i + n j
i =1

(4)

n j are noise samples on receive antenna j, modelled as


independent samples of a zero mean complex Gaussian random variable with variance N 0 / 2 per dimension.Its mathematical expression is similar to hij except noise

1-4244-0517-3/06/$20.00 2006 IEEE

discussed in reference [2], and it is only limit to AWGN channel. This paper provides an applicable model, Rayleigh flat fading model.

10

10

-1

PTC-(23,35)-permuter=400 PTC-(37,21)-permuter=400 NOPTC-(33,25) -permuter=400

3 Performance Analysis

Simulation

Results

and
BER

10

-2

10

-3

In this paper, NOPTC is applied to MIMO system. Figure 3 compares the BER curves. As stated above, the simulation circumstance is block fading. In our simulation, (17,11) 8 and (11,17) 8 , (33,25)8 , (71,47 ) 8 and

10

-4

10

-5

10

-6

(55,63) 8 , (137,101) 8 and (167,101) 8 are investigated that


are 8-state 16-state 32-state and 64-state NOPTC, respectively.
10

-7

(13,15) 8 , (37,21) 8

4 Eb/N0(dB)

and

( 23,35) 8 ,
10
0

(b)
PTC-(73,45)-permuter=400 NOPTC-(71,47)-permuter=400 NOPTC-(55,63) -permuter=400

(73,45) 8 (177,103) 8 and (167,101) 8 are selected that are 8-state, 16-state, 32-state and 64-state PTC, respectively. Obviously PTC (13,15) 8 is what we often read in publications that is chosen as 3GPP standard [9] [10] for its excellent performance. The goal of performance comparison here is to promote turbo codes BER performance and ensure NOPTCs BER performance surpasses the existing application specification of international standard at least when working at lower and moderate SIR. The simulation conditions also include the following: interleaver depth 400 bits; code rate 1 / 3 ; the number of iteration 5; the number of transmit antenna and the number of receive antenna are both 4. Synthesize (a), (b), (c) and (d) of figure 3, we can see no matter at a lower, moderate, or higher SIR, PTC (13,15) 8 has the worst BER performance among the 8state turbo code; PTC (37,21) 8 the worst BER performance among the 16-state turbo code; PTC (73,45) 8 the worst
10
0

10

-1

10

-2

BER

10

-3

10

-4

10

-5

10

-6

4 Eb/N0(dB)

(c)
10
0

10

-1

PTC-(177,103)-permuter=400 NOPTC-(137,101)-permuter=400 NOPTC-(167,101)-permuter=400

10

-2

BER

10

-1

PTC-(13,15)-permuter=400 NOPTC-(11,17)-permuter=40 NOPTC-(17,11)-permuter=400

10

-3

10

-2

10

-4

BER

10

-3

10

-5

10

-6

10

-4

4 Eb/N0(dB)

10

-5

10

-6

3 4 Eb/N0(dB)

(a)

(d) Figure 3 BER comparison betweenNOPTCand PTC over flat fading channel (a) BER performance of 8-state turbo code; (b) BER performance of 16-state turbo code; (c) BER performance of 32-state turbo code; (d) BER performance of 64-state turbo code.

1-4244-0517-3/06/$20.00 2006 IEEE

BER performance among the 32-state turbo code; and PTC (177,103) 8 the worst BER performance among the 64-state turbo code. In summary, among all the investigated PTC and NOPTC, NOPTC secures better BER performance when compared under the equal both state-number and other simulation conditions. Among the investigated 64-state NOPTCs, (167,101) 8 has best performance, and the transfer function for (167,101) 8 is:

differences between NOPTC and PTC over Rayleigh flat fading channel. As a result, For the explored codes, NOPTC outperforms PTC at lower, moderate and high SIR. Better performance will be expected in the future.

References
[1] C. Berrou and A. Giavieux. Near optimum error correcting coding and decoding: Turbo-codes. IEEE Trans. Commun.,

vol. 44, no. 10. Oct., 1996:1261-1271P (8)


[2] Mingxin Tan, Alister G. Burr, Non-primitive Turbo codes, Proceedings of First International Symposium on Broadband Communications, Harrogate, England, 12-15, Dec., 2004, page 61 [3] Branka Vucetic, Jinhong Yuan, Space-Time Coding, Wiley Editorial Office 2003 [4] Da-shan Shiu, Wireless Communication Using Dual Antenna Arrays, Kluwer Academic Publishers, 2000 [5] Sirikiat Lek Ariyavisitakul, Turbo Space Time Processing to Improve Wireless Channel Capacity, IEEE Transactions on communications, vol. 48, no. 8, August 2000 [6] Shu Lin. An Introduction to Error Correcting Codes. PrenticeHall, Englewood Cliffs, NJ, 1970 [7] L. C. Perez, J. Seghers, and D. J. Costello. A distance spectrum interpretation of turbo codes. IEEE Trans. Inform. Theory, vol.42., Nov, 1996: 1698-1709P [8] Oscar Y. Takeshita, Oliver M. Collins, Peter C. Massey, and Daniel J. Costello, Jr. On the Frame Error Rate of Turbo Codes, IEEE Transctions on Communications, Vol. 49, No. 4, April 2001. [9]Third Generation Partnership Project Multiplexing and channel coding, 3G TS 25.222 V3.2.0, [10]Third Generation Partnership Project Multiplexing and channel coding, 3G TS 25.212 (DRAFT) [11]Third Generation Partnership Project Quality of Service concept and architect- ture (Release 6), 3GPP TS 23.107 V6.2.0 (2004-12) [12] Third Generation Partnership Project QoS Concept and

g ( D) G ( D) = 1, 1 g 0 ( D)
Where

g 0 ( D) = 1 + D + D 2 + D 4 + D 5 + D 6

(9)

(10) From the simulation results of figure 3, comparing with other 8-state, 16-state and 32-state it is seen as well that 64state turbo code doesnt have best BER performance when at lower and moderate SIR , nor at higher SIR. It is NOPTC (17,11) 8 has best BER performance at any SIR. Obviously, the assertion the more state numbers turbo code has, the better performance turbo code achieves doesnt meet in SISO system, neither in MIMO system here. In a word, there is no direct proportion between the state-number and the BER performance of turbo code. Here we analyze the structure of feedback polynomial. From the simulation results, it is seen those feedback polynomials that contain more component 1 (not component 0) in the feedback polynomials have better BER performance when working at the equal SIR and other simulation conditions. The number of 1 doesnt include both the Least Significant bit (LSM) and the Most Significant bit (MSB). In convolutional code and turbo code theory, LSB and MSB corresponding to a polynomial item stand for output and input, respectively. In feedback polynomial structure, apart from LSB and MSB, other 1s mean the information feeds back to the input, 0s mean the information doesnt feed back to the input. Therefore, in feedback polynomial structure, more 1 denotes more external information feeds back to the input, which will contribute to recovering the original transmission information reliably. In practical application standards, for the 3G speech codec and the MPEG-4 video codec [11] [12], specification of PTC can implement both visible artefacts and speech communications. According to the results above, NOPTC will run better in both visible artefacts and speech communications. 4 Conclusions This paper investigates the BER performance

g1 ( D ) = 1 + D 6

Architecture (Release 1999), 3GPP TS 23.107 V6.2.0 (2004-12)

1-4244-0517-3/06/$20.00 2006 IEEE

You might also like