Near-optimal low-complexity decoding of space-time codes for wireless applications

Description

REFERENCE TO RELATED APPLICATIONS

This application is a continuation of U.S. patent application Ser. No. 10/234,407, filed Sep. 3, 2002, now U.S. Pat. No. 6,741,635 which is a continuation of U.S. patent application Ser. No. 09/690,542, filed Oct. 17, 2000 (now U.S. Pat. No. 6,470,043), which is a continuation of U.S. patent application Ser. No. 09/063,765, filed Apr. 21, 1998 (now U.S. Pat. No. 6,188,736), which claims the benefit of U.S. Provisional Application No. 60/068,613, filed Dec. 23, 1997.

BACKGROUND OF THE INVENTION

This invention relates to wireless systems and, more particularly, to systems having more than one antenna at the receiver and at the transmitter.

Physical constraints as well as narrow bandwidth, co-channel interference, adjacent channel interference, propagation loss and multi-path fading limit the capacity of cellular systems. These are severe impairments, which liken the wireless channel to a narrow pipe that impedes the flow of data. Nevertheless, interest in providing high speed wireless data services is rapidly increasing. Current cellular standards such as IS-136 can only provide data rates up to 9.6 kbps, using 30 kHz narrowband channels. In order to provide wideband services, such as multimedia, video conferencing, simultaneous voice and data, etc., it is desirable to have data rates in the range of 64–144 kbps.

Transmission schemes for multiple antenna systems may be part of a solution to the problem of the currently available low data rates. Such schemes were first proposed in papers by Wittneben, and by Seshadri and Winters, where the problem was addressed in the context of signal processing.

One prior art arrangement having a single transmitter antenna and multiple receiver antennas is shown in FIG. 1. Each of the receiver antennas receives the transmitted signal via a slightly different channel, where each channel i is characterized by transfer function α_i. Using an approach known as “Maximum Ratio Combining”, the prior art approach to detection contemplates multiplying each received signal that had been influenced by α_i by the complex conjugate signal, α_i*, summed, and then processed.

In a co-pending application titled “Method and Apparatus for Data Transmission Using Space-Time Codes and Multiple Transmit Antennas”, filed on May 6, 1997, bearing the Ser. No. 08/847,635, and assigned to the assignee of this invention, a coding perspective was adopted to propose space-time coding using multiple transmit and receive antennas. Space-time coding integrates channel coding, modulation, and multiple transmit antennas to achieve higher data rates, while simultaneously providing diversity that combats fading. It may be demonstrated that adding channel coding provides significant gains over the schemes of Wittneben and Seshadri and Winters. In said co-pending application, space-time codes were designed for transmission using 2–4 transmit antennas. These codes perform extremely well in slowly varying fading environments (such as indoor transmission media). The codes have user bandwidth efficiencies of up to 4 bits/sec/Hz which are about 3–4 times the efficiency of current systems. Indeed, it can be shown that the designed codes are optimal in terms of the trade-off between diversity advantage, transmission rate, decoding complexity and constellation size.

It can also be shown that as the number of antennas is increased, the gain increases in a manner that is not unlike a multi-element antenna that is tuned to, say, a particular direction. Unfortunately, however, when maximum likelihood detection is employed at the receiver, the decoding complexity increases when the number of transmit and receive antennas is increased. It would be obviously advantageous to allow a slightly sub-optimal detection approach that substantially reduces the receiver's computation burden.

SUMMARY

Such an approach is achieved with a receiver arrangement where signals received at a plurality of antennas are each multiplied by a respective constant and then summed prior to being applied to a maximum likelihood detector. The respective constants, λ_j where j is an index designating a particular receiver antenna, are derived from a processor that determines the largest eigenvector of the matrix A, where Λ is a vector containing the values λ_j, and A is a matrix containing elements α_ij, which is the transfer function between the i^th transmitter antenna to the j^th receiver antenna. The α_ij terms are determined in the receiver in conventional ways.

BRIEF DESCRIPTION OF THE DRAWING

FIG. 1 presents a block diagram of prior Maximal Ratio Combining detection; and

FIG. 2 presents a block diagram of an embodiment including a transmitter having a plurality of antennas, and a receiver having a plurality of antennas coupled to an efficient detection structure.

DETAILED DESCRIPTION

FIG. 2 presents a block diagram of a receiver in accord with an embodiment of the invention. It includes a transmitter 10 that has an n plurality of transmitting antenna 1, 2, 3, 4, and a receiver 20 that has an m plurality of receiver antennas 21, 22, 23, 24. The signals received by the receiver's antennas are multiplied in elements 25, 26, 27, and 28, and summed in adder 30. More specifically, the received signal of antenna j is multiplied by a value, λ_j, and summed. The collection of factors λ_j can be viewed as a vector Λ. The outputs of the receiver antennas are also applied to processor 40 which, employing conventional techniques, determines the transfer functions α_ij for i=1, 2, 3, . . . , n and j=1, 2, 3, . . . , m. These transfer functions can be evaluated, for example, through the use of training sequences that are sent by the different transmitter antennas, one antenna at a time.

The evaluated α_ij signals of processor 40 are applied to processor 45 in FIG. 2 where the multiplier signals λ_j, j=1, 2, 3, . . . , m are computed. Processor 45 also evaluates a set of combined transfer function values y_i, i=1, 2, 3, . . . , n (which are described in more detail below). Signals y_i of processor 45 and the output signal of adder 30 are applied to detector 50 which detects the transmitted symbols in accordance with calculations disclosed below.

It is assumed that the symbols transmitted by the antennas of transmitter 10 have been encoded in blocks of L time frames, and that fading is constant within a frame. A codeword comprises all of the symbols transmitted within a frame, and it corresponds, therefore, to
c₁¹c₁²c₁³ . . . c₁⁴c₂¹c₂²c₂³ . . . c₂⁴c₃¹c₃²c₃³ . . . c₃⁴ . . . c_m¹c_m²c_m³ . . . c_m⁴,  (1)
where the superscript designates the transmitter's antennas and the subscript designates the time of transmission (or position within a frame).

From the standpoint of a single transmitting antenna, e.g., antenna 1, the signal that is received from antenna 1 in response to a transmitted symbol c_t¹ at time interval t is:

  R t = c t 1 ⁡ ( α 11 ⁢ λ 1 + α 12 ⁢ λ 2 + α 13 ⁢ λ 3 + … + α 1 ⁢ m ⁢ λ m ) = c t 1 ⁢ ∑ j = 1 m ⁢ λ j ⁢ α 1 ⁢ j = c t 1 ⁢ γ 1 ( 2 )
(when noise is ignored). If each λ_j value is set to α*_1j, (where α*_1j is the complex conjugate of α_1j) then the received signal would simply be

R t = c t 1 ⁢ ∑ j = 1 m ⁢  α 1 ⁢ j  2 ( 3 )
yielding a constructive addition.

Of course, the values of λ_j cannot be set to match α*_1j and concurrently to match the values of α*_ij where i≠1; and therein lies the difficulty.

When all n of the transmitting antennas are considered, then the received signal is

  R t = ∑ i = 1 n ⁢ ( c t i ⁢ ∑ j = 1 m ⁢ λ j ⁢ α ij ) = ∑ i = 1 n ⁢ c t i ⁢ γ i ( 4 )

In accordance with the present disclosure, the objective is to maximize

∑ i = 1 n ⁢  γ i  2
because by doing so, signal R_t contains as much information about c_tⁱ, i=1, 2, 3, . . . n as is possible. However, it can be easily shown that if a matrix A is constructed such that

A = ∑ i = 1 n ⁢ ( Ω i * ) T ⁢ Ω i , ( 5 )
where Ω_i=(α_i1, α_i2, α_i3 . . . α_im), then

∑ i = 1 n ⁢  γ i  2 = Λ ⁢ ⁢ A ⁡ ( Λ * ) T . ( 6 )

The receiver, thus, has to maximize ΛA(Λ*)^T, subject to the constraint ∥Λ∥²=1. The solution to this problem is to choose Λ to be the eigenvector of A which corresponds to the maximum eigenvalue of A. Accordingly, processor 45 develops the matrix A from the values of α_ij, finds the eigenvalues of A in a conventional manner, selects the maximum eigenvalue of A, and creates the vector A. Once Λ is known, processor 45 develops signals γ_i for 1=1, 2, 3, . . . , n, (where

γ i = ∑ j = 1 m ⁢ λ j ⁢ α ij ⁢   ) ,
and applies them to detector 50. Finally, detector 50 minimizes the metric

∑ t = 1 L ⁢  R t - ∑ i = 1 n ⁢ γ i ⁢ c t i  2
from amongst all possible codewords in a conventional manner. As can be seen, this approach reduces the complexity of decoding by almost a factor of m.

FIG. 2 depicts separate multipliers to multiply received signals by multiplication factors λ_i, and it depicts separate blocks for elements 30,40,45, and 50. It should be understood, however, that different embodiments are also possible. For example, it is quite conventional to incorporate all of the above-mentioned elements in a single special purpose processor, or in a single stored program controlled processor (or a small number of processors). Other modifications and improvements may also be incorporated, without departing from the spirit and scope of the invention, which is defined in the following claims.