Académique Documents
Professionnel Documents
Culture Documents
Abstract - Multiple Input Multiple Output (MIMO) of coding and transmissions schemes. Some diversity options
communication systems are used to increase the channel capacity are combined in Space Time Codes (STC), Space Frequency
or the data rate for a particular connection through multiplexing Codes (SFC), or Space Time Frequency Codes (STFC), where
or to improve performance through diversity. In this paper, we the orthogonal property is a key factor for the system
present and study the performances of an Orthogonal Frequency performances.
Division Multiplexing (OFDM)-MIMO system using different Alamouti introduced in [2], a simple antenna diversity
Forward Error Correcting (FEC) codes. Several constituent codes
have been tested: block codes, convolutional codes and
scheme for two transmit antennas, which provides maximum
concatenated codes. The implemented system considers only QAM diversity in flat fading MIMO channels. STC codes, introduced
constellations, Rayleigh flat fading channels, no Channel State by Tarokh et all in [3], combines traditional channel code
Information (CSI) at the transmitter and with CSI at the receiver. design (for temporal diversity) and multi-antenna signal design
(for spatial diversity). Two performance criteria are provided:
I. INTRODUCTION firstly, is the rank criterion for diversity advantage, which gives
In the Information Theory (IT) community, the notion of the degree of antenna diversity achieved and describes the
space can be broadly defined as the simultaneous use of asymptotic slope of the decoding error probability versus
multiple, possibly coupled, channels. The notions of space-time Signal to Noise Ratio (SNR). Secondly, is the determinant
and Multiple Input Multiple Output (MIMO) channels are criterion for coding advantage, which quantifies the coding
therefore often used interchangeably. The connection between advantage and results in a shift of the performance curve.
space and MIMO is most transparent when we view the This paper is organized as follows. In Section II, we present
multiple channels as created by two or more spatially separated the signal model and the capacity of the MIMO channel.
antennas at a wireless transmitter or receiver. Section III presents a general MIMO OFDM system without
In the last decade, the academic and industrial communities any error correcting codes. In Section IV are presented general
started a flood of research to redesign the traditional codes used aspects and characteristics of the block codes, convolutional
for one-dimensional streams in time. These researches are codes and concatenated codes, with the emphasis on the
focused on the design of modulation and coding techniques parameters of the implemented codes. Section V presents the
that are two-dimensional, covering both space and time overall system performances (BER) of the MIMO OFDM
domains. system for different error correcting codes: block codes,
To achieve high data rates in wireless communications convolutional codes and concatenated codes.
systems, usually more bandwidth is required. Due to
II. THE MIMO CHANNEL MODEL
limitations in the spectrum, it is often impractical and
expensive to increase the bandwidth. In this case, multiple We consider a single user communication system with NT
transmit and receive antennas are used to increase the spectral transmit and N R receive antennas at the input and output of a
efficiency, offering a spatial diversity at both sides of the link
block Rayleigh flat fading channel with coherence time T. The
and obtaining the well-known MIMO channel. In addition, a
MIMO channel can be represented by the following complex
capacity gain and improvement of robustness and reliability are
discrete time signal model [4]:
obtained. In most scattering environments, spatial diversity is
used to reduce the detrimental effects of multi-path fading [1].
y = Hx + n , (1)
In fact, MIMO concept is much more general and embraces
many other scenarios such as wired networks, multi-user
systems, single antenna frequency selective channels. where x is a NT × 1 matrix of complex transmitted symbols
The different diversity techniques (time, space- antenna, (one for each transmitting antenna) with covariance matrix
frequency, polarization) should be properly exploited by means
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 21, 2009 at 07:57 from IEEE Xplore. Restrictions apply.
C x , y is a N R × 1 matrix of complex received symbols (one for In this paper, we consider an unknown channel at the
each receive antenna), n is a N R × 1 matrix of independent and transmitter side, saying that the transmitter has no CSI. In this
case, each antenna transmits an equal power σ 2x so the
identical distributed (i.i.d.) complex Gaussian noise samples
with zero-mean and σ 2n variance. The channel matrix H, covariance matrix of the transmitted signal is C x = σ2x I NT and
defined by: the total transmitted power is P = tr(C x ) = σ2x NT . The capacity
is [5,6]:
⎡ h11 h12 L h1NT ⎤
⎢ ⎥
⎢ h21
H=⎢
h22 L h2 NT ⎥
, (2)
(
C = log 2 det I N R + SNR ⋅ HH
H
) , [bit/s/Hz], (5)
M M O M ⎥
⎢ ⎥
⎢⎣ hN R 1 hN R 2 L hN R NT ⎥⎦ where SNR = σ2x σn2 is the average signal to noise ratio per
receiving antenna. The upper limit for the capacity can be
is a random matrix with complex elements hij describing the stated as:
gain of the radio channel between the jth transmitting antenna Cmax = min{NT , N R } ⋅ log 2 (1 + SNR ⋅ max{NT , N R }) , (6)
and the ith receiving antenna. For the block flat fading channel,
we assume that the fading coefficients hij remains constants
so the maximum limit of the information capacity depends on
within a frame of length T symbol periods and changes into the number of the transmitting and receiving antennas, the
new independent values from frame to frame. larger ones determines a logarithmic increase of it.
For uncorrelated Rayleigh fading channels (e.g. a wireless The MIMO channel, due to multipath propagation, causes a
transmission from a PC to a router, where the antennas can be time dispersion of the transmitted signals resulting an overlap
spaced sufficiently far apart from each other) the elements hij of the various transmitted symbols at the receiver. This is
are i.i.d. Gaussian random variables with zero-mean and σ2n called Inter Symbol Interference (ISI), which can causes high
variance. error rates if it is left uncompensated.
When the correlation between the receiving antennas exists, III. THE UNCODED MIMO-OFDM SYSTEM
e.g. a downlink transmission from a base station – BS to a
mobile station – MS, where the antennas of BS can be spaced Orthogonal Frequency Division Multiplexing (OFDM) is
sufficiently far enough to eliminate the correlation between standardized as a modulation scheme for a variety of
them. For MS this is more difficult due to physical size applications, such as Digital Audio Broadcasting (DAB),
constrains, the columns of H are independent random vectors Terrestrial Digital Video Broadcasting (DVB-T), Wireless
with the same mean and covariance matrix, but the elements of Local Area Networks (WLAN). Combining OFDM technique
each column are correlated. with multiple antenna architectures provides high-rate high
The mutual information and the information capacity of a performance transmissions.
general MIMO channel are given by [4,5]: In OFDM, the basic principle is to split a high rate data
stream into a number of K lower rate substreams that are
modulated and transmitted simultaneously over a K orthogonal
⎛ 1 ⎞ subcarriers [7]. Due to this division the ISI is reduced and the
I = log 2 det ⎜ I N +
H
HC X H ⎟ , (3)
⎝
R
σn
2
⎠ bandwidth of each substream will be less compared to the total
bandwidth of the data stream, converting the frequency
⎛ 1 ⎞ selective fading channel into flat fading channel.
C = max I = max log 2 det ⎜ I N +
H
HC X H ⎟ , (4)
σn Having an orthogonal transmission, the information signal
⎝ ⎠
2
Cx Cx R
296
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 21, 2009 at 07:57 from IEEE Xplore. Restrictions apply.
Likelihood (ML) or a Minimum Mean Square Error (MMSE)
Cyclic
detection algorithm can be used to produce hard estimates of
QAM IDFT
Prefix the bits transmitted on the respective subcarriers [10]. Using a
data
QAM demodulator, the complex symbols are demodulated and
then multiplexed.
QAM IDFT Cyclic
prefix
IV. THE FEC CODED MIMO-OFDM SYSTEM
In this section, we present the coded MIMO-OFDM system
Remove
(Fig. 2). Forward Error Correcting (FEC) codes may be applied
DFT QAM-1
C.P.
MIMO
Detection
to MIMO systems to improve the performances. Although
data
for there exists the STC codes [3], which have been developed
subcarriers
(ML, specifically for MIMO systems, any convolutional code or
Remove DFT QAM-1
C.P.
MMSE)
block code may be applied to improve the performance
[11,12].
Figure 1. The uncoded MIMO-OFDM system (transmitter and receiver).
Cyclic
The simulated system considers only Quadrature Amplitude
FEC
Π QAM IDFT
Prefix
data
Modulation (QAM) constellations that are suboptimal in the
sense that they do not maximally space the constellation points
for a given energy, but they have the considerable advantage
FEC
Π QAM IDFT Cyclic
prefix
that they may be easily transmitted as two Pulse Amplitude
Modulation (PAM) signals on two quadrature sinusoidal
carriers. Because the carriers have the same frequency band
and differ by a π / 2 phase, they can be modulated
Remove
C.P.
DFT MIMO
QAM-1
Π −1 FEC
Dec.
detection data
for
independently, transmitted over the same frequency band, and subcarrier
separated by demodulation at the receiver. (ML,
Remove
C.P.
DFT MMSE) QAM-1
Π −1 FEC
Dec.
The most used M-QAM modulations are 2-QAM (equivalent
with BPSK), 4-QAM (equivalent with QPSK), 16-QAM, 64-
QAM, 128-QAM and 256-QAM. M is the number of points Figure 2. The FEC coded MIMO-OFDM system (transmitter and receiver).
(signals) of the constellation, log 2 M is the number of bits for
the binary representation of the signals. If the amplitude of the In the MIMO-OFDM system we can use following codes:
farthest constellation point is required to be unitary, the whole 1) Block codes (n,k,d) divides the data stream in message
constellation should be divided by the coefficient words i of length k and associates to each a codeword v of
length n, adding a number of m = n − k parity check bits. For
k = 2 ( 2 p − 1) , with p = log 4 M .
instance, the classical Hamming (7,4,3) code, with minimum
To minimize the bit error rate of an uncoded system we distance d=3 is used. With hard decoding, the Hamming code
combine classical QAM modulation with Gray coding, so two is a SEC/DED (Single Error Correction / Double Error
horizontally or vertically adjacent symbols differs by precisely Detection) code. Another used block code is Hamming
one bit, while two diagonally adjacent symbols differ by (15,11,3).
exactly two bits. Other block codes which we have used in the coded MIMO-
The pilot symbols have not been included in the OFDM OFDM system are the binary Golay (23,12,7) code denoted as
system, as their only purpose is to estimate the channel matrix G23 and Golay (11,6,5) = G11. The Golay code G23 is able to
in the case of an unknown channel at the receiver. They would correct 3 errors and detect 6 errors in the case of hard
have been, for instance, Constant Amplitude Zero decoding.
AutoCorrelation (CAZAC) sequences, also called polyphase For block codes the decoding is based on the trellis
sequences. In the implemented system, we consider a perfect representation of the block codes, using the Viterbi algorithm
CSI at the receiver and before the MIMO detection; the with hard decision (1-bit quantization on the received channel
channel equalization is made using an estimated channel matrix values and Hamming distance metric) or soft decisions (multi-
based on the inverse of the channel matrix written in Toeplitz bit quantization and Euclidean distance metric). For the ideal
form. soft-decision decoding (infinite-bit quantization), the received
At the receiver, the signal at each receive antenna will have channel values are directly used in the decoder.
components from the all transmitted signals coming from 2) Convolutional codes, used for instance in GSM mobile
different paths of the MIMO channel. Assuming that the phones, as well as satellite and military communication
receiver is perfect time and frequency synchronized, for each devices, make every codeword symbol be the weighted sum of
receive antenna the cyclic prefix is removed and the FFT is
the various input message symbols. Fundamentally,
applied. After retrieving the subcarrier signals, MIMO
convolutional codes do not offer more protection against noise
detection is applied on a subcarrier basis. A spatial Maximum
than an equivalent block code, but they are generally easier to
297
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 21, 2009 at 07:57 from IEEE Xplore. Restrictions apply.
implement compared to the block code of equal correction V. SIMULATION RESULTS
power [13,14].
In the coded MIMO-OFDM system we use the nonrecursive A. Uncoded system
nonsystematic [7,5]8 convolutional code, with generator The first results presented in Fig. 4 are meant to show the
polynomials 1 + D + D 2 and 1 + D 2 . The decoding strategy for performances of the uncoded MIMO OFDM system (Fig. 1),
convolutional codes is based on the Viterbi algorithm using which not include any error correcting codes, as a starting point
hard decision or soft decision. and comparison reference for future simulations.
The source information bits {0,1} are generated randomly
3) Concatenated codes. A code adapted to the QAM
with uniform distribution. The transmitter includes QAM
mapping must certainly have more than one level of protection,
mapping with Gray coding. For the MIMO system we consider
as well as different correction capabilities within the same
NT = 2 transmit antennas and N R = 2 receive antennas. The
level. The general idea can be traced back to Elias product
codes construction, developed later by Forney, [15], into a number of OFDM subcarriers is K = 2 . The transmitted
simple and powerful concatenated code construction (Fig. 3), OFDM symbols are grouped in frames of T = 200 symbols
and the number of trials is 1000 frames or more to ensure a
Outer
Encoding
Π Inner
Encoding high accuracy up to BER = 10−3 . The program automatically
Interleaver generates the channel matrix H, in a randomly manner. The
Channel
channel fading is flat, channel matrix varying from frame to
frame. Fig. 4 shows the simulation results for MIMO OFDM
Inner
systems with different QAM constellations, from 4-QAM to
Outer
Decoding Π −1 Decoding 256-QAM.
Deinterleaver 0
10
Figure 3. Concatenated codes.
298
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 21, 2009 at 07:57 from IEEE Xplore. Restrictions apply.
• In the simulated system we considered that the and more than 8 dB (for BER<10-4) compared to the uncoded
transmitted signals have unitary variance σ 2x = 1 . But transmission.
After testing separately the proposed error correcting codes on
for 16-QAM or denser constellations, not all the points
a SISO system and the MIMO OFDM black box, the next
are equally distanced from the origin and the signal
logical step was to try and combine them. First, only
don’t have a unitary energy, so a coefficient
convolutional code [7,5]8 has been applied to the MIMO
k = −10 log10 σ2x should rectify the OX axis of the plots,
OFDM system, in order to have intermediary results. The
the equation (7) is modified:
MIMO OFDM system parameters are the same as those used in
the uncoded system ( NT = 2 , N R = 2 , K = 2 , T = 200 , trials
SNR[dB]= E b N 0 [dB] + 10 log10 ( log 2 M ) + k . (8)
1000 frames).
0
10
4 ⋅ 1 + 4 ⋅ (1 / 3) + 8 ⋅ (5 / 9)
2 2
5
For 16-QAM results σ2x = = so
16 9
another k = 2.55 dB must be decreased to get the real
-1
10
value. Therefore, a point of 10 dB for SNR corresponds
to a point of 1.45 dB for Eb/N0.
-2
10
B. Coded system
BER
For the presented error correcting codes in Section IV, the
BER performance results from Fig. 5 are useful to test the 10
-3
algorithm (hard decoding with Hamming distance) on the Figure 6. Simulation results for convolutional coded MIMO OFDM system
codes trellises. Using only the outer code or only the inner code (4-QAM, 16-QAM).
the gain is around 4 dB as compared to the uncoded 10
0
-2
10
-2
-3 10
10
BER
299
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 21, 2009 at 07:57 from IEEE Xplore. Restrictions apply.
Following results regards the concatenated coded MIMO Block Codes (STBC) and Low Density Parity Check
OFDM system, with the Golay (23,12,7) as inner code and (LDPC), in conjunction with MIMO-OFDM systems.
convolutional [7,5]8 as outer code. The Viterbi algorithm is
REFERENCES
used for both codes, with hard or 8 levels soft decoding. Fig. 8
[1] E. Telatar, “Capacity of multi-antenna Gaussian channels”, AT&T Bell
presents the results for the uncoded MIMO OFDM system, the Labs., Internal Tech. Memo, Jun. 1995.
convolutional coded one and the concatenated coded one. For [2] S. M. Alamouti, “A simple transmit diversity technique for wireless
concatenated codes a gain of 8 - 10 dB must be noticed. communications”, IEEE Journal Selected Areas Comm., Vol. 16, No. 8,
10
0 pp. 1451-1458, Aug. 1998.
[3] V. Tarokh, N. Seshadri, A. R. Calderbank, “Space-Time codes for high
data rate wireless communication: performance criterion and code
construction”, IEEE Trans. Inform. Theory, vol. 44, pp. 744-765, Mar.
-1
10 1998.
[4] G. Foschini, M. Gans, “On limits of wireless communications in a fading
environment when using multiple antennas,” Wireless Personal Comm.,
-2 Vol. 6, No. 3, pp. 311–335, 1998.
10
[5] Andrea Goldsmith, S. A. Jafar, N. Jindal, S. Vishwanath, “Capacity limits
of MIMO channels”, IEEE Journal Selected Areas Comm., Vol. 21, No.
BER
300
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 21, 2009 at 07:57 from IEEE Xplore. Restrictions apply.