10.1.1.26.5946.pdf

Embed Size (px)

Citation preview

  • 8/14/2019 10.1.1.26.5946.pdf

    1/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 1

    Space-Time Block Codes versus Space-Time

    Trellis Codes

    Sumeet Sandhu, Robert W. Heath Jr., Arogyaswami Paulraj

    The authors are with Information Systems Laboratory, Stanford University, Stanford, CA 94305. R. Heath was

    supported by Ericsson Inc. through the Networking Research Center at Stanford University.

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    2/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 2

    Abstract

    Transmit diversity schemes for the coherent multiple-antenna flat-fading channel range from space-

    time block codes (STBC) to space-time trellis codes (STTC). In this letter we compare the performance

    of STBC and STTC by means of frame error rate. We discover that a simple concatenation of STBCwith traditional AWGN (additive white Gaussian noise) trellis codes outperforms some of the best known

    STTCs at SNRs (signal to noise ratios) of interest. Our result holds for small numbers of receive antennas

    and trellis states, and may extend to greater numbers of antennas and states with improved block codes.

    Keywords

    Block codes, Diversity methods, MIMO systems, Modulation coding, Fading channels.

    I. Introduction

    The challenge of transmit diversity design for the multiple-antenna fading channel has been

    met with several novel modulation and error correction techniques in the recent past. Prominent

    among these space-time block codes ([1], [2]) and space-time trellis codes ([3], [4], [5], [6]).

    Space-time block codes operate on a block of input symbols producing a matrix output over

    antennas and time. Unlike traditional single-antenna AWGN block codes, full rate space-time

    block codes do not provide coding gain. Their key feature is the provision of full diversity with

    extremely low encoder/decoder complexity. The best known orthogonal block codes are provided

    in [2].

    Space-time trellis codes operate on one input symbol at a time producing a sequence of spatial

    vector outputs. Like traditional TCM (trellis coded modulation) for the single-antenna channel,

    space-time trellis codes provide coding gain. Since they also provide full diversity gain, their

    key advantage over space-time block codes is the provision of coding gain. However, they are

    extremely difficult to design and require an expensive encoder and decoder.

    This paper proposes a compromise between STBC (known designs, simple ML (maximum

    likelihood) decoding, no coding gain) and STTC (difficult to design, expensive ML decoding,

    coding gain). We concatenate AWGN trellis codes with STBC in order to obtain coding gain.

    Then we compare the performance of concatenated STBC with STTC by keeping transmit power

    and spectral efficiency fixed. The results are somewhat surprising - with the same number of

    trellis states (4 and 8), concatenated STBC outperforms STTC for 1 and 2 receive antennas.

    Concatenation of TCM with STBC was also proposed in [7] but its performance was not com-

    pared with STTC. A performance comparison of STTC and concatenated STBC was provided

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    3/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 3

    in [8], but it was not fair in terms of date rate and number of trellis states. A fair comparison

    of STTC and concatenated STBC was provided in [9] for the fast fading channel but not for the

    quasi-static channel.

    II. Data and Channel Model

    Consider a system with Mr receive antennas and Mt > 1 transmit antennas. The channel is

    flat-fading and quasi-static. It is unknown at the transmitter but is known at the receiver. At

    time nT, the channel output corresponding to the nth input block spanningTsymbol times is

    YnT = HXnT+ VnT (1)

    where the received signal YnT is MrT, the fading channel H is MrMt, the encoded codewordXnT isMtT, and receiver noise VnT isMrT. The entries ofH are i.i.d.(independent, identicallydistributed) circular complex Gaussian random variables with variance 0.5 in each dimension,

    i.e. hi,j c(0, 1). The entries ofVnT are i.i.d., vnTi,j c(0,N0), and independent over n.The average power transmitted on Mt antennas is Es. The codeword XnT is encoded using

    concatenated STBC or STTC, both of which are described in detail in the following sections.

    III. Space-Time Block Codes

    The input to the encoder is a stream of real or complex modulated symbols. The encoder

    operates on a block ofKsymbols producing an MtT codeword XnTwhose rows correspondto transmit antennas and columns correspond to symbol times. At the receiver, ML decoding is

    simplified by the orthogonal structure of the codewords.

    The effective channel induced by space-time block coding of input symbols (before ML de-

    tection) is an AWGN channel with receive SNR equal to ||H||2FEs

    MtN0[10], [11]. This motivates the

    concatenation of traditional single-antenna TCM with STBC. The system used here consists

    of an outer TCM encoder/decoder concatenated with the STBC encoder/decoder. The overall

    coding gain is only due to the outer TCM encoder since we consider full rate STBCs.Recently, new block codes were designed in [12] by maximizing average capacity, some of which

    outperform the orthogonal codes in [2]. Here we will only consider the codes in [2], which are the

    best orthogonal linear codes with respect to the union bound on symbol error probability [13].

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    4/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 4

    IV. Space-Time Trellis Codes

    Space-time trellis codes encode the input scalar symbol stream into an output vector symbol

    stream. Unlike space-time block codes, space-time trellis codes map one input symbol at a time

    to an Mt1 vector output. Decoding is performed via ML sequence estimation.Code performance is quantified by the diversity advantageand the coding advantage [4]. For

    a given number of transmit antennas, the design objective is to construct the largest possible

    codebook with full diversity advantage (= Mt) and the maximum possible coding advantage.

    A number of hand-crafted codes with full diversity advantage were provided in [4]. Codes with

    greater coding advantage than those in [4] were reported in [5] and [14] after exhaustive computer

    searches over a feedforward convolutional coding (FFC) generator. A structured method of code

    construction that ensures full diversity was provided in [15] along with new designs. Recently

    new codes with better distance spectrum properties and performance were reported in [6].

    V. Performance Analysis

    Note that any space-time code can be analyzed in terms of the criteria presented for space-

    time trellis codes, namely diversity advantage and coding advantage. These criteria affect the

    performance curve in different ways. Diversity advantage affects the asymptotic slope of the

    FER (frame error rate) versus SNR graph - greater the diversity, the more negative the slope.

    Coding advantage affects the horizontal shift of the graph - greater the coding advantage, thegreater the shift to the left.

    Since all codes considered here are full diversity, the asymptotic slopes of their FER graphs

    will be the same. The difference in coding advantage can be quantified as follows. At high SNRs

    the logarithm of the minimum PEP for the kth code isPk MrMtskMrMtck whereMrMtisthe diversity advantage, sk =log(

    Es4N0

    ) is the SNR term, and ck is the coding advantage. Defining

    P =PkPl, c = ckcl, and s = sksl for codes k and l we have

    P MrMts MrMtc (2)

    At a given SNR, s = 0, and if c > 0 then the PEP for k is lower than that for l by

    P MrMtc. Therefore the effect of improved coding advantage increases linearly with thenumber of receive antennas, as also observed in [14].

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    5/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 5

    A. Coding Advantage

    We will only consider full diversity codes for 2 transmit antennas at a bit rate of 2 bits per

    symbol time. We will focus on a small number of trellis states, equal to 4 and 8, both for STTCs

    and concatenated STBCs. In order to predict the performance of different codes, we computed

    their coding advantages by means of a computer search and listed them in Table I. These values

    agree with those found in published literature [16], [5], [6]. Note that the values in Table I are

    normalized by Mt = 2.

    The block code labeled STBC stands for the complex Mt= 2 Alamouti code [1] and has a

    4-PSK constellation. The codes labeled STBC-TCM represent the Alamouti code concatenated

    with outer AWGN trellis codes and have 8-PSK constellations. The outer codes are both rate 2/3

    and are the best 4-state (parallel transition) and 8-state trellis codes in [16]. The codes labeled

    STTC denote space-time trellis codes, of which the 4 and 8 state codes in [4] (STTC-Tarokh),

    [5] (STTC-Grimm), and [6] (STTC-Yan) are shown, all of which have 4-PSK constellations. The

    4-state Grimm and Yan codes have the best possible coding advantage in the class of FFC codes

    [5].

    As demonstrated in [6], although the 4 and 8 state Yan codes have the same coding advantage

    as the Grimm codes, they outperform the Grimm codes because of better distance spectrum

    properties. Since the Grimm codes were shown to outperform the Tarokh codes [14], [6], we will

    only consider the Grimm and Yan codes in the sequel.

    VI. Simulations

    In this section Monte Carlo simulations of the codes in Table I are presented. Performance

    is measured in terms of FER for a burst of 130 symbols. The Rayleigh channel described

    in (1) is used with two transmit antennas. Channel realizations are i.i.d.from burst to burst.

    Performance is plotted for one, two and three receive antennas in Figures 1, 2 and 3 respectively.

    The performance of STBC is plotted in all the figures for reference.

    In Figure 1 (a) note that STBC by itself performs as well or better than all the STTCs, even

    though it provides no coding gain. This is explained by the multidimensional structure of STBC

    that spans two time symbols.

    With one receive antenna, concatenated STBC performs dramatically better than all the 4-

    state STTCs. The performance gap reduces with increasing receive antennas in Figures 2 and 3.

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    6/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 6

    The performance gap is also less noticeable with 8-state codes for all receive antennas.

    With three receive antennas and 8 trellis states, in fact, STTC-Yan outperforms STBC-TCM.

    The loss in performance of STBC-TCM with increasing receive antennas is explained by the

    capacity loss incurred by STBC [11]. Considering the capacity-optimized block codes in [12], we

    expect that those block codes will outperform STTC-Yan for even greater numbers of receive

    antennas and trellis states.

    Note that in all these comparisons, the encoding/decoding complexity of STTCs is much

    higher than that of concatenated STBCs for large numbers of trellis states and multiple receive

    antennas. A comparison that is fair with respect to complexity would entail concatenation of

    more sophisticated codes with improved block codes, potentially extending their performance

    advantage to multiple receive antennas.

    VII. Conclusions

    We have shown that for a small number of receive antennas and trellis states, a simple concate-

    nation of space-time block codes with traditional AWGN trellis codes can significantly outperform

    space-time trellis codes. The concatenated scheme separates spatial modulation from temporal

    error correction and is therefore attractive from the aspect of computational complexity as well.

    References

    [1] S. M. Alamouti, A simple transmitter diversity scheme for wireless communications, IEEE J. Select. Areas

    in Comm., vol. 16, pp. 14511458, October 1998.

    [2] V. Tarokh, H. Jafarkhani, and A. R. Calderbank, Space-time block codes from orthogonal designs, IEEE

    Trans. Information Theory, vol. 45, pp. 14561467, July 1999.

    [3] J. C. Guey, M. P. Fitz, M. R. Bell, and W. Y. Kuo, Signal design for transmitter diversity wireless commu-

    nication systems over Rayleigh fading channels, in Proc. VTC, vol. I, pp. 136140, 1996.

    [4] V. Tarokh, N. Seshadri, and A. R. Calderbank, Space-time codes for high data rate wireless communication

    I: performance criterion and code construction, IEEE Trans. Information Theory, vol. 44, pp. 74465, March

    1998.

    [5] J. Grimm, Transmitter diversity code design for achieving full diversity on Rayleigh fading channels. PhDthesis, Purdue University, 1998.

    [6] Q. Yan and R. S. Blum, Optimum space-time convolutional codes, in Proc. WCNC, 2000.

    [7] S. M. Alamouti, V. Tarokh, and P. Poon, Trellis-coded modulation and transmit diversity : design criteria

    and performance evaluation, in Proc. ICUPC, 1998.

    [8] J.-C. Guey, Concatenated coding for transmit diversity systems, inProc. VTC, vol. 5, pp. 25002504, 1999.

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    7/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 7

    [9] A. Yongacoglu and M. Siala, Space-time codes for fading channels, in Proc. VTC, vol. 5, pp. 24952499,

    1999.

    [10] G. Ganesan and P. Stoica, Space-time diversity using orthogonal and amicable orthogonal designs, in

    Proc. ICASSP, 2000.

    [11] S. Sandhu and A. Paulraj, Space-time block codes : a capacity perspective. Accepted IEEE Comm. Letters,

    June 2000.

    [12] B. Hassibi and B. M. Hochwald, High-rate codes that are linear in space and time, in Proc. 38th Annual

    Allerton Conference on Communication, Control and Computing, 2000.

    [13] S. Sandhu, R. Heath, and A. Paulraj, Union bound for linear space-time codes, in Proc. 38th Annual

    Allerton Conference on Communication, Control and Computing, 2000.

    [14] S. Baro, G. Bauch, and A. Hansmann, Improved codes for space-time trellis-coded modulation, IEEE

    Comm. Lett., vol. 4, pp. 2022, January 2000.

    [15] A. R. Hammons and H. E. Gamal, On the theory of space-time codes for PSK modulation,IEEE Trans.

    Information Theory, vol. 46, pp. 524542, March 2000.

    [16] E. Biglieri, D. Divsalar, P. J. McLane, and M. K. Simon, Introduction to Trellis-Coded Modulation with

    Applications. Macmillan, 1991.

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    8/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 8

    Table I : Coding Advantage : 2 transmit antennas, 2 bits/symbol time

    Figure 1 : One receive antenna

    Figure 2 : Two receive antennas

    Figure 3 : Three receive antennas

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    9/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 9

    TABLE I

    Code States Coding Advantage in dB

    STBC 1 1 0

    STBC-TCM 4 2 3.01

    STTC-Tarokh 4 1 0

    STTC-Grimm 4

    2 1.51

    STTC-Yan 4

    2 1.51

    STBC-TCM 8

    4.5858 3.31

    STTC-Tarokh 8

    3 2.39

    STTC-Grimm 8 2 3.01

    STTC-Yan 8 2 3.01

    Fig. 1.

    10 12 14 16 1810

    3

    102

    101

    100

    (a) 4 state codes

    SNR in dB

    FER

    STBCSTBC + 4state TCMGrimm 4stateYan 4state

    10 12 14 16 1810

    3

    102

    101

    100

    (b) 8 state codes

    SNR in dB

    FER

    STBCSTBC + 8state TCMGrimm 8stateYan 8state

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    10/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 10

    Fig. 2. Two receive antennas

    4 6 8 10 1210

    3

    102

    101

    100

    (a) 4 state codes

    SNR in dB

    FER

    STBCSTBC + 4state TCMGrimm 4stateYan 4state

    4 6 8 10 1210

    3

    102

    101

    100

    (b) 8 state codes

    SNR in dB

    FER

    STBCSTBC + 8state TCMGrimm 8stateYan 8state

    November 9, 2000 DRAFT

  • 8/14/2019 10.1.1.26.5946.pdf

    11/11

    IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. Y, MONTH 2000 11

    Fig. 3. Three receive antennas

    0 2 4 6 810

    3

    102

    101

    100

    (a) 4 state codes

    SNR in dB

    FER

    STBCSTBC + 4state TCMGrimm 4stateYan 4state

    0 2 4 6 810

    3

    102

    101

    100

    (b) 8 state codes

    SNR in dB

    FER

    STBCSTBC + 8state TCMGrimm 8stateYan 8state

    November 9, 2000 DRAFT