The present invention relates to a system and method for simplified Maximum Likelihood (ML) decoding method for Multiple-In-Multiple-Out (MIMO) wireless communication systems
Multiple in Multiple out (MIMO) is gaining increasing attention in wireless network applications because MIMO has the potential to provide better frequency efficiency than a single antenna system, although the implementation complexity is higher than for a single antenna communication system.
The implementation cost is higher not only because of the hardware increase resulting from the increased number of antennas, but also because of the costs associated with the decoding higher complexity. When a system has multiple inputs, joint detection of the multiple inputs is optimal. However, joint detection is subject to very high computational complexity. Thus most of the MIMO decoding focuses on non-optimal but computationally affordable methods, such as Zero Forcing (ZF) and Ordered Successive Interference Cancellation (OSIC), see Wolniansky, P. W. et al., “V-BLAST: an architecture for realizing very high data rates over the rich-scattering wireless channel”, ISSSE 98, pp 295-300, which is hereby incorporated by reference on its entirety. Speth, et al, “Low complexity spaced frequency MLSE for multi-user COFDM”, GLOBECOM '99, Vol. 5, 1999, pp 2395-2399, which is hereby incorporated by reference in its entirety, provides a simplified decoding based on optimal ML decoding. But the main idea is still ZF. The difference is that the method proposed in Speth et al. uses the channel state information to weight the bit metrics, whereby it came to be called Weighted Zero Forcing (WZF) decoding.
Although WZF currently is the choice for MIMO decoding because of its simplicity, there is still a several dB performance gap between the optimal ML decoding and WZF. Research is on going to find a decoding method that eliminates the performance gap with reasonable implementation cost.
Thus, there is a need for a system and method for improving on existing WZF MIMO decoding.
The present invention provides a totally different decoding method than ZF for MIMO wireless systems.
The system and method of the present invention is based on the optimal ML decoding that, instead of performing a norm 2 calculation of the joint detection, performs a more simple calculation. The simpler calculation reduces the number of multiplications required to a more affordable level for implementation.
According to a simulation, the performance of a first embodiment of the simplified ML decoding method provides a 2 dB gain over WZF for a packet error rate (PER) level of 10−.
According to a simulation, the performance of a second embodiment of the simplified ML decoding method provides a 4 dB gain over WZF for PER level of 10−2 for 16/64 Quadrature Amplitude Modulation (QAM) modulation and 8 dB gain for Binary Phase Shift Keying (BPSK) modulation.
It is to be understood by persons of ordinary skill in the art that the following descriptions are provided for purposes of illustration and not for limitation. An artisan understands that there are many variations that lie within the spirit of the invention and the scope of the appended claims. Unnecessary detail of known functions and operations may be omitted from the current description so as not to obscure the present invention.
Referring now to
The present invention provides a system and method for a distance calculation according to the observation. In optimal ML decoding, the following equation is used to calculate the bit metrics for a Viterbi decoder:
thereby reducing the computation cost by replacing the multiplication with comparison while providing a 2 dB gain over WZF for PER level of 10−2.
In a second embodiment of the invention, the metrics calculation is simplified to
that reduces the computation cost by replacing the multiplication with addition while providing a 4 dB gain over WZF for PER level of 10−2 for 16/64QAM modulation and 8 dB gain for BPSK modulation.
Optimal Single Input Single Output SISO Coded Orthogonal Frequency Division Multiplexing (COFDM) System Decoding
The basic principle of Orthogonal Frequency Division Multiplexing (OFDM) is to split a high-rate data stream into a number of lower rate streams that are transmitted simultaneously over a number of sub-carriers. Because the symbol duration increases for the lower rate parallel sub-carriers, the relative amount of dispersion in time caused by multi-path delay spread decreases. Intersymbol interference eliminated almost completely by introducing a guard interval in every OFDM symbol.
In a single carrier system, since a time-dispersive channel (frequency selective fading channel) brings the channel memory into the system, joint maximum likelihood equalization and decoding is not realistic because of the high computation cost. The general practice is to first use minimum mean-square error (MMSE) as the criteria to equalize the channel. Then the equalized signal is sent to a maximum likelihood detector for further decoding. This is a sub-optimal system. In a COFDM system, since the system is designed to let each sub-carrier experience flat fading channel, the real maximum likelihood equalization and decoding can be implemented with affordable computation cost.
The architecture of a typical transceiver of an IEEE 802.11a based COFDM system can be found in Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: High-speed Physical Layer in the 5 GHz Band, IEEE Std 802.11a, 199, that is hereby incorporated by reference in its entirety. In this transceiver, bitwise interleaving combined with Gray mapping is used. This technique, referred to as Bit Interleaved Coded Modulation (BICM), provides significantly better performance over Rayleigh fading channel than TCM with comparable complexity. The received signal in the frequency domain can be expressed as
r
k
=h
k
s
k
+n
k (1)
where
P(rk|bki),
where
and where
to be sent to the Viterbi decoder. For the 64 QAM case, to obtain a bit metric for a bit to be ‘1’ or ‘0’, one needs to calculate 32 distances
∥rk−hka∥2
with different constellation points a, in which bit i is ‘1’ or ‘0’, to find the minimum value of these distances.
A diagram of a 2 by MIMO system based on IEEE 802.11a 54 Mbps mode OFDM system is illustrated in
where
hij represents the channel from transmitter antenna i to receiver antenna j on subcarrier k. Since the operation of each subcarrier is the same, to save space, the subscript k is omitted in the following discussion. Without losing generality assume the four channels are independent Rayleigh fading channels. The received signal can be expressed as
Analogous to the ML detection for the SISO system, for each received signal pair, r1 and r2, in order to find out the bit metrics for a bit in symbol S1 and S2, the following equation should be satisfied
where m1ip and m2ip represent bit metrics of bit i in transmitted symbol s1 and s2 to be ‘p’, pε{0,1}; C represents the whole constellation point set and Cip the subset of the constellation point such that bit i is equal to p. Then the bit metrics pairs (m1i0,m1i1)(m2i0,m2i1) are sent to corresponding deinterleavers and Viterbi decoders for decoding both the bit streams.
For simplicity, only 2 by 2 MIMO analysis is given in detail. The same principle can be readily used by n-by-n MIMO system. Rewriting the received signal in (4) in matrix format results in
r=Hs+n (6)
With ZF filter, the estimation of the transmitted signal is
ŝ=(HHH)−1HHr (7)
With MMSE filter, the estimation of the transmitted signal is
ŝ=((HHH)−1HH+σ2)r (8)
The estimation error (for ZF only, the MMSE analysis will follow the same principle with additional noise variance item) is
ε=ŝ−s=(HHH)−1HHn (9)
With variance
R
ε
=E[ε·ε
H]=σN2(HHH)−1 (10)
Thus the multi-dimensional complex joint probability distribution is
Since Rε is not a diagonal matrix, the bit metrics calculation cannot be separated for these two symbol detections. To simplify the calculation, an uncorrelated normal distribution is assumed for the components of (11). Then the bit metrics can be calculated as
where m1ip and m2ip represents bit metrics of bit i in transmitted symbol s1 and s2 to be ‘p’, pε{0,1}; Cip the subset of the constellation point such that bit i is equal to p. Then the bit metrics pairs (m1i0,m1i1)(m2i0,m2i1) are sent to corresponding deinterleavers and Viterbi decoders for decoding both the bit streams.
ZF decoding has the problem of boosting noise if the channel matrix is ill conditioned. Weighted ZF partly solves this problem by giving low weight to the bit metrics calculated from the ill-conditioned channel. Using this approach a Viterbi decoder is not disturbed by some very erroneous bit metrics. This is similar to puncturing a sequence to set those erroneous bit metrics almost to zero. The following decoder attempts to recover those bits using some reliable bits information around those bits.
In the present invention, a decoding method is used having a basic principle that is the same as ML decoding. The only difference is replacing the distance calculation of the ML criteria by an approximate distance calculation that is less computationally intensive. The bit metrics is then simplified as follows.
(1) In a first embodiment
reducing the computation cost by replacing the multiplication with comparison while providing a 2 dB gain over WZF for PER level of 10−2.
The number of multiplications needed for the simplified ML decoding to find the bit metrics for all the bits transmitted in one packet is 4*M, where M is the constellation order, assuming the channel remains quasi-static for one OFDM packet. For WZF method, the multiplications needed for one packet is 8*N*K, where N is the number of data sub-carriers in one OFDM symbol and K is the number of OFDM symbols in one packet.
As an example only, the number of multiplications for SML, WZF and ML decoders are calculated as follows:
For the whole packet:—
For a single bit:—
From
(2) In a second embodiment
The simplified ML decoding requires 4*M complex multiplications to find the bit metrics for all the bits transmitted in one sub-carrier, where M is the constellation order. For a WZF method, the multiplications needed for one sub-carrier is 8 no matter what modulation scheme is used. As an example only, the number of multiplications for SML, WZF and ML decoders are compared. Assuming one packet contains 1000 information bytes, 96 data sub-carriers in one OFDM symbol, 64QAM is used to transmit the data and compared to a 2 by 2 MIMO system according to the present invention. A packet contains about 10 OFDM symbols for the given OFDM and modulation parameters.
To decode one packet:
SML requires 4*64*96=24576 complex multiplications,
WZF requires about 6*96+8*96*10=8256 complex multiplications,
optimal ML decoding requires about 4*64*96+64*64*96*10*2=7888896.
For a single bit:
SML requires about 24576/(1000*8)˜=3 complex multiplications,
WZF requires 8256/(1000*8)˜=1 multiplications, and
optimal ML requires 7888896/(1000*8)˜=986 multiplications.
To find the minimum distance:
SML requires 64*64=4096 additions,
WZF requires 64*2=128 additions, and
optimal ML requires 64*64=4096 additions.
The second embodiment has been simulated to determine the performance for the simplified ML decoding method as compared to SML and WZR. Results are presented as the PER vs. SNR curves for different modulation schemes illustrated in
Referring now to
A MIMO system and method according to the present invention can be used in systems and methods according to the IEEE 802.11n standard since the decoding method in this invention can be used in any 11n decoder. However, the application of the present invention is not limited to only 11n devices since a general decoding system and method are provided. That is, the system and method of the present invention can be used for any convolutional coded BICM MIMO system, and especially good a wireless system having a small number of antennas, such as 2 by 2 MIMO. Since the system and method of the present invention comprises a simplified ML decoding method, any communication system using ML decoding can take advantage of the present invention to reduce the computation cost for system implementation.
A decoder according to the system and method of the present invention provides a level of error performance that approximates a level of error performance provided by a maximum likelihood decoder ML.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that the superframe as described herein is illustrative and various changes and modifications may be made and equivalents may be substituted for elements thereof without departing from the true scope of the present invention. In addition, many modifications may be made to adapt the teachings of the present invention to a particular situation without departing from its central scope. Therefore, it is intended that the present invention not be limited to the particular embodiments disclosed as the best mode contemplated for carrying out the present invention, but that the present invention include all embodiments falling with the scope of the appended claims.
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/IB05/52128 | 6/27/2005 | WO | 00 | 12/28/2006 |
Number | Date | Country | |
---|---|---|---|
60584203 | Jun 2004 | US |