Linear space-time block code with block STTD structure

[0001] This application claims priority from provisional patent application Serial No. 06/331,721, filed Nov. 21, 2001. Copending patent applications Ser. No. 10/144,114, filed May 13, 2002 disclose related subject matter and have a common assignee with the present application.

[0002] The present invention relates to wireless digital communications, and more particularly to space-time diversity transmission systems and methods.

[0003] Spread spectrum wireless communications utilize a radio frequency bandwidth greater than the minimum bandwidth required for the transmitted data rate, but many users may simultaneously occupy the bandwidth. Each of the users has a pseudo-random code for spreading information to encode it and for despreading (by correlation) received spread spectrum signals and recovery of information. Such multiple access typically appears under the name of code division multiple access (CDMA). The pseudo-random code may be an orthogonal (Walsh) code, a pseudo-noise (PN) code, a Gold code, or combinations (modulo-2 additions) of such codes. After despreading the received signal at the correct time instant, the user recovers the corresponding information while other users' interfering signals appear noise-like. For example, the interim standard IS-95 for such CDMA communications employs channels of 1.25 MHz bandwidth and a pseudo-random code pulse (chip) interval T_C of 0.8138 microsecond with a transmitted symbol (bit) lasting 64 chips. The recent 3GPP wideband CDMA (WCDMA) proposal employs a 3.84 MHz bandwidth and the CDMA code length applied to each information symbol may vary from 4 chips to 256 chips. Indeed, UMTS (universal mobile telecommunications system) approach UTRA (UMTS terrestrial radio access) provides a spread spectrum cellular air interface with both FDD (frequency division duplex) and TDD (time division duplex) modes of operation. UTRA currently employs 10 ms duration frames partitioned into 15 time slots with each time slot consisting of 2560 chips (T_C0.26 microsecond).

[0004] The air interface leads to multipath reception, so a RAKE receiver has individual demodulators (fingers) tracking separate paths and combines the finger results to improve signal-to-noise ratio (SNR). The combining may use a method such as the maximal ratio combining (MRC) in which the individual detected signals in the fingers are synchronized and weighted according to their signal strengths or SNRs and summed to provide the decoding. That is, a RAKE receiver typically has a number of DLL or TDL code tracking loops together with control circuitry for assigning tracking units to the strongest received paths. Also, an antenna array could be used for directionality by phasing the combined signals from the antennas.

[0005] Further, UTRA allows for space-time block-coding-based transmit antenna diversity (STTD) in which, generically, channel bits b₀, b₁, b₂, b₃ (values 1) are transmitted as the sequence b₀, b₁, b₂, b₃ by antenna 1 and simultaneously transmitted as the sequence b₂, b₃, b₀, b₁ by antenna 2. Note that interpreting (b₀, b₁) and (b₂, b₃) as two complex numbers (e.g., QPSK or QAM symbols) implies the sequence transmitted by antenna 2 differs from the sequence from antenna 1 by a rotation of /2 in a two-complex-dimensional space followed by complex conjugation. See Alamouti, A Simple Transmit Diversity Technique for Wireless Communications, IEEE J. Selec. Areas on Comm. 1451 (1998).

[0006] STTD is a diversity technique in the sense that it provides redundancy for the transmitted symbols. In particular, (1) it provides full transmit diversity gain without bandwidth expansion, (2) there is no loss in code rate, and (3) it is an orthogonal code and thus allow simple maximum likelihood detection by a maximum ratio combiner followed by a demodulator. When used with a modulation scheme of order M (e.g., QPSK has M4 and 16QAM has M16) the complexity of the detector is proportional to 2M.

[0007] Recently, attempts to generalize orthogonal codes with full diversity and are noted in Naguib et al, Increasing Data Rate over Wireless Channels, IEEE Signal Proc. Mag. 76 (May 2000). However, it was proven that orthogonal space-time block codes with full diversity for more than 2 antennas (with complex symbols) suffer from rate loss. That is, for more than 2 transmit antennas, space-time block codes with all three of the foregoing properties do not exist. Thus there are various possibilities of partial fulfillment of the three properties.

[0008] More specifically, space-time block codes with symbol constellation rotation appear in Xin et al, Space-Time Diversity Systems Based on Unitary Constellation-Rotating Precoders, Proc. ICASSP 2429 (2001). The space-time rotation codes can achieve full diversity gain and good coding gain, which can be measured in terms of distance from the full transmit diversity curves. However, this class of codes do not inherit any orthogonal structure and the complexity of the maximum likelihood detector for space-time rotation code is proportional to M^P where P is the number of antennas and M the symbol constellation size. Sub-optimal detectors such as linear or iterative minimum mean square error can also be used to reduce the complexity to be proportional to MP³.

[0009] Papadias et al, A Space-Time Coding Approach for Systems Employing Four Transmit Antennas, Proc. ICASSP 2481 (2001) and Tirkkonen et al, Minimal Non-Orthogonality Rate 1 Space-Time Block Code for 3 Tx Antennas, IEEE 6th Int. Symp. on Spread-Spectrum Tech. & Appl. 429 (2000) show for systems with 4 antennas that instead of achieving full orthogonality, one can achieve three-level orthogonality which leads to maximum likelihood detector with complexity proportional to (P/2)M^P/2, and for the suboptimal detectors complexity reduces to 16M.

[0010] However, these block codes have limited performance.

[0011] The present invention provides space-time block codes with a block STTD code matrix structure plus transmission and/or reception with such codes. Preferred embodiments with 22 blocks (i.e., four antennas and blocks of four symbols) allow simple partial orthogonality for low complexity detection plus provide large diversity gain and coding gain.

[0012] This has the advantages including increased performance (in terms of error rate and system throughput) for wireless communications.

[0013] The drawings are heuristic for clarity.

[0014] FIG. 1 is a flow diagram.

[0015] FIGS. 2a-2b illustrate preferred embodiment transmitter.

[0016] FIG. 3 shows preferred embodiment receiver.

[0017] FIGS. 4-5 present simulation results.

[0018] 1. Overview

[0019] FIG. 1 is a flow diagram for preferred embodiment code design methods which can be used for preferred embodiment space-time block encoding with a block STTD structure. This structure allows for partial orthogonality and consequent low-complexity detection. Preferred embodiments include 22 blocks which apply to four transmission antennas with four-symbol code blocks. The preferred embodiment 22 block STTD code matrices possess a channel-code 3-level orthogonality plus jointly maximized diversity gain and coding gain.

[0020] Preferred embodiment communications systems use preferred embodiment encoding and decoding methods, FIGS. 2a-2b illustrate preferred embodiment transmitter and receiver functional blocks.

[0021] In preferred embodiment communications systems base stations and mobile users could each include one or more digital signal processors (DSPs) and/or other programmable devices with stored programs for performance of the signal processing of the preferred embodiment methods. Alternatively, specialized circuitry could be used. The base stations and mobile users may also contain analog integrated circuits for amplification of inputs to or outputs from antennas and conversion between analog and digital; and these analog and processor circuits may be integrated as a system on a chip (SoC). The stored programs may, for example, be in ROM or flash EEPROM integrated with the processor or external. The antennas at the receiver may be parts of receivers with multiple finger RAKE detectors for each user's signals. Exemplary DSP cores could be in the TMS320C6xxx or TMS320C5xxx families from Texas Instruments.

[0022] 2. Four-Antenna, Four-Symbol, 3-Level Orthogonality

[0023] First preferred embodiment space-time transmit diversity block codes use blocks of four symbols and transmit with four antennas, have 3-level orthogonality (described below) of column vectors of the channel-code matrix for low complexity detection, plus jointly maximized diversity gain and coding gain as described below.

[0024] FIGS. 2a-2b illustrate functional blocks of a first preferred embodiment transmitter which could be a base station transmitter in a cellular wireless system. FIG. 3 illustrates a preferred embodiment receiver with Q antennas for reception, such as for a mobile user in a cellular wireless system. A base station may include the channel estimation and block coding transformation computation as in FIG. 2b.

[0025] In general a block code for blocks of N symbols and for P antennas can be represented by PN code matrices C. For example, the 22 STTD described in the background codes a block of two complex symbols (s₁, s₂) for transmission modulation as s₁, s₂ on a first antenna and s₂*, s₁* on a second antenna. This 22 STTD thus has code matrices $C\left(s\right)=\left[\begin{array}{cc}{s}_1& s_2\\ -s_2^{*}& s_1^{*}\end{array}\right]$

[0026] where s denotes the block of symbols s₁, s₂. The two elements of the first row of C are the two transmission modulations for the first antenna during the two symbol periods, and the elements of the second row are the two transmission modulations for the second antenna during the same two symbol periods. Thus over two symbol periods and using two antennas, the 22 STTD transmits two symbols. That is, this 22 STTD has a block code rate of where the block code rate is defined as the ratio of the number of symbols transmitted in a block divided by the product of the number of transmit antennas and the number of symbol periods in the block:

[0027] For systems with P transmit antennas, a space-time block code with full transmit diversity can have a block code rate of at most 1/P. The preferred embodiment space-time block codes use four antennas (P4), blocks of four symbols (N4), and have a block code rate of . In particular, with a block s of symbols s₁, s₂, s₃, s₄, the first preferred embodiment space-time block codes have 44 code matrices of the form: $C\left(s\right)=\left[\begin{array}{cc}{C}_{}& C_{}\\ -C_{}^{*}& C_{}^{*}\end{array}\right]$

[0028] with 22 submatrices:

[0029] where each of A_k and B_k is a 22 complex matrix. (The space-time code is defined by the set of A_ks and B_ks.) During the first and second symbol periods antennas 1 and 2 transmit the C combinations of s₁ and s₂ while antennas 3 and 4 transmit the C* combinations of s₃ and s₄. Then during the third and fourth symbol periods antennas 1 and 2 transmit the C combinations of s₃ and s₄ while antennas 3 and 4 transmit the C* combinations of s₁ and s₂. Thus preferred embodiment space-time block codes have the form of a higher dimensional analog of the 22 STTD; that is, they are block STTD. Indeed, for the special case of $A_1=A_3=\left[\begin{array}{cc}1& 0\\ 0& 0\end{array}\right],A_2=A_4=\left[\begin{array}{cc}0& 0\\ 0& 1\end{array}\right],\mathrm{and}$ $B_1=B_2=B_3=B_4=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]$

[0030] the space-time block code is essentially an interleaved version of two 22 STTD codes: during the first and third symbol periods the first and third symbols would be transmitted on the first and third antennas, and during the second and fourth symbol periods the second and fourth symbols would be transmitted on the second and fourth antennas.

[0031] Other possible variations of the block STTD structure that achieve the same purpose can also be used. For example: $C\left(s\right)=\left[\begin{array}{cc}{C}_{}& -C_{}^{*}\\ C_{}& C_{}^{*}\end{array}\right]\mathrm{or}\left[\begin{array}{cc}{C}_{}& -C_{}\\ C_{}^{*}& C_{}^{*}\end{array}\right]\mathrm{or}\left[\begin{array}{cc}{C}_{}& C_{}\\ C_{}^{*}& -C_{}^{*}\end{array}\right]$

[0032] and similar variations.

[0033] Initially, consider the cases with $B_1=B_2=B_3=B_4=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right].$

[0034] To describe the 3-level orthogonality and consequent low-complexity detection, first consider the received signal for the case of one receiver antenna. Denote the received signal during the four symbol periods as r₁, r₂, r₃, r₄; then with the four channels (one from each transmit antenna to the single receive antenna) defined by impulse responses h₁, h₂, h₃, h₄, the received signal in vector notation is (ignoring noise and presuming maximal ratio combining of multipath signals):

[0035] Thus the column index of C corresponds to time and the row index to space (antenna number). In the - notation of C as an analog of 22 STTD, this can be rewritten as

[0036] where r^Tr₁, r₂, r^Tr₃, r₄, and h^Th₁, h₂, h^Th₃, h₄.

[0037] And in terms of a 4-vector r of received signals r₁, r₂, r₃*, r₄* depending on a 4-vector s of transmitted symbols s₁, s₂, s₃*, s₄* through a channel-code matrix H, this translates to rHs, which written in terms of elements is: $\left[\begin{array}{c}{r}_1\\ r_2\\ r_3^{*}\\ r_4^{*}\end{array}\right]=\left[\begin{array}{cccc}{h}_{}^T{a}_{11}& h_{}^T{a}_{21}& -h_{}^T{a}_{31}^{*}& -h_{}^T{a}_{41}^{*}\\ h_{}^T{a}_{12}& h_{}^T{a}_{22}& -h_{}^T{a}_{32}^{*}& -h_{}^T{a}_{42}^{*}\\ h_{}^H{a}_{11}& h_{}^H{a}_{21}& h_{}^H{a}_{31}^{*}& h_{}^H{a}_{41}^{*}\\ h_{}^H{a}_{12}& h_{}^H{a}_{22}& h_{}^H{a}_{32}^{*}& h_{}^H{a}_{42}^{*}\end{array}\right]\left[\begin{array}{c}{s}_1\\ s_2\\ s_3^{*}\\ s_4^{*}\end{array}\right]$

[0038] where A_ma_m1, a_m2 for m1,2,3,4. That is, a_mj is a 2-vector column of A_m.

[0039] The first preferred embodiment 44 space-time codes, in addition to having code matrices of the form C, has 3-level orthogonality in H in the sense that each column vector of H is orthogonal to 2 or all 3 of the other three column vectors. Note that the jth component of the kth column vector represents the antenna weightings multiplied by channel impulse responses of the kth symbol transmitted during the jth symbol period. Thus the 3-level orthogonality is an orthogonality between symbol combinations and relates to low-complexity detection as illustrated in section 6 below. In particular, the inner products of the various column vectors of H are the elements of the 44 matrix H^HH used in maximum likelihood detection. That is, the (m,n)th element of H^HH is the inner product of the mth and nth column vectors, and thus 3-level orthogonality can be expressed visually as requiring at least four of the six upper triangle (and by symmetry at least four of the six lower triangle) entries of H^HH vanish. Of course, by re-labeling, the 3-level orthogonality could be taken as column 1 and column 2 each orthogonal to both column 3 and column 4, and conversely. In this case the 22 upper right and 22 lower left submatrices of H^HH are both 0.

[0040] Due to the separation of symbols s₁, s₂ from symbols s₃, s₄ by the structure of C, the inner product of H column j with H column k (with j3 or 4 and k1 or 2) equals:

[0041] where the components of the 2-vectors were used: h^Hh₃*, h₄*, h^Th₁, h₂, a_m1a_m1,1, a_m1,2^T, and a_m2a_m2,1, a_m2,2^T. Note the multipliers of h₁h₃* and h₂h₄* both vanish, which reflects the structure of C as a block STTD code matrix.

[0042] Conditions on the A_m (i.e., the a_mn) which make the sum of determinants deta_j1 a_k1deta_j2 a_k2 vanish for each of the four inner products will thus define codes with 3-level orthogonality, and conversely. This leads to space-time code design in the following sections. Of course, the simplest condition would be to have all four a_m1 be parallel and to have all four a_m2 be parallel to make all eight determinants deta_jn a_kn vanish. In fact, section 4 considers codes with all a_m1 as multiples of $\left[\begin{array}{c}1\\ 0\end{array}\right]$

[0043] and all a_m2 as multiples of $\left[\begin{array}{c}0\\ 1\end{array}\right]$

[0044] and imposes further constraints for practical code design methods.

[0045] Further, a 22 orthogonal transformation applied to all of the A_m preserves the sum of determinants and thus preserves 3-level orthogonality. In particular, let $U=\left[\begin{array}{cc}{u}_{11}& u_{12}\\ u_{21}& u_{22}\end{array}\right]$

[0046] and consider the transformation A_mA_mU for m1,2,3,4; then

[0047] Using the multilinearity of the determinant yields:

[0048] Thus U orthogonal preserves the sum of determinants, and more generally, U a multiple of an orthogonal matrix preserves 3-level orthogonality. Section 4 uses an orthogonal U to reduce the peak-to-average power ratio by transformation applied to C. Of course, an orthogonal U is just rotating the antenna weightings of a each symbol over time for a pair of antennas (1-2 or 3-4) and a pair of symbol periods (1-2 or 3-4).

[0049] In contrast, the inner product of H column j with H column k for j and k both 1 or 2 or both 3 or 4 do not have the channel coefficients, h₁, h₂, h₃, h₄, factoring out as in the previous inner products, so the code alone cannot achieve orthogonality. In particular, the inner product of column j with column k with j and k both 1 or 2 equals

[0050] where A_kA_j^H(m,n) is the row m, column n element of A_kA_j^H and equals the inner product of the nth row of A_j with the mth row of A_k, and $h_1=\left[\begin{array}{c}{h}_1\\ h_3^{*}\end{array}\right],h_2=\left[\begin{array}{c}{h}_2\\ h_4^{*}\end{array}\right].$

[0051] The inner product for j and k both 3 or 4 is the complex conjugate:

[0052] Thus conditions on code defining matrices A_m can make the inner products of H columns 1 and 2 appear similar to inner products of H columns 3 and 4 because the channel coefficient factors are the same.

[0053] FIG. 3 illustrates a receiver for transmissions using a block STTD space-time code possessing 3-level orthogonality. The 3-level orthogonality permits lower complexity detection as described in section 6 below.

[0054] Further, for sufficiently large signal-to-noise ratio (SNR), a space-time block code has performance determined by the diversity gain, G_D, and the coding gain, G_C, with the following definitions:

[0055] where the two code matrices have differing symbol blocks. Thus the preferred embodiment block STTD space-time codes also jointly maximize gains in addition to having 3-level orthogonality. This maximization requires consideration of all constellation symbols, and thus is more than just the 3-level orthogonality which only depends upon the A_m. These conditions on the block STTD space-time codes to have 3-level orthogonality provide limitations for searching to find optimal coding diversity and coding gain.

[0056] Note that the foregoing extends to systems with more than four antennas. Explicitly, for P antennas (P an even integer) and symbol blocks of size P, use the same block STTD structure for the PP matrices C(s) where the (P/2)(P/2) submatrices are:

[0057] where each of A_k and B_k is a (P/2)(P/2) complex matrix. Again, a partial orthogonality of the columns of the PP channel-code matrix H for the case of all B_k0 can be expressed in terms of sums of 22 determinants but where the sums include all combinations of from the P/2-vectors which make up the columns of the corresponding A_k. Again, the section 4 case of all A_k diagonal insures the partial orthogonality.

[0058] 3. 5-Condition Preferred Embodiments

[0059] As noted in the preceding section, 3-level orthogonality leads to low-complexity detection, but to find space-time block codes which also maximize diversity gain and coding gain requires searching over classes of 3-level orthogonal codes combined with symbol constellations. Practicality demands some restriction and/or parameterization of the class of space-time blocks codes to be searched. Thus consider a class of codes possessing 3-level orthogonality in the form: $H^{H}H=\left[\begin{array}{cc}R& 0\\ 0& R\end{array}\right]$

[0060] where 0 is the 22 matrix of 0s and $R=\left[\begin{array}{cc}{}_{11}& {}_{12}\\ {}_{12}^{*}& {}_{22}\end{array}\right].$

[0061] Thus the column 1 and column 2 vectors of H are orthogonal to the column 3 and column 4 vectors, the inner product of the column 1 vector with the column 2 vector equals the inner product of the column 3 vector with the column 4 vector (₁₂), the squared magnitude of the column 1 vector equals the squared magnitude of the column 3 vector (₁₁), and the squared magnitude of the column 2 vector equals the squared magnitude of the column 4 vector (₂₂).

[0062] The foregoing form of H^HH is equivalent leads to the following five conditions for the code defining matrices A_k:

[0063] 1) there exist a pair of phases , so that either a₁₁ a₁₂e^j a₃₁ e^j a₃₂ or a₁₁ a₁₂e^j a₃₂ e^j a₃₁

[0064] 2) there exist a pair of phases , so that either a₂₁ a₂₂e^j a₄₁ e^j a₄₂ or a₂₁ a₂₂e^j a₄₂ e^j a₄₁

[0065] 3) a₁₁a₄₁^Ta₁₂a₄₂^Ta₄₁a₁₁^Ta₄₂a₁₂^T (i.e., A₁A₄^T is symmetric)

[0066] 4) a₂₁a₃₁^Ta₂₂a₃₂^Ta₃₁a₂₁^Ta₃₂a₂₂^T (i.e., A₂A₃^T is symmetric)

[0067] 5) a₁₁a₂₁^Ha₁₂a₂₂^Ha₄₁a₃₁^Ha₄₂a₃₂^H (i.e., A₁A₂^HA₄A₃^H)

[0068] These conditions provide further limitations on the block STTD structure and thereby simplify searching to generate codes with 3-level orthogonality plus maximal coding gain and diversity gain.

[0069] 4. STTD-Based 3-Level Orthogonal Using Rotation Preferred Embodiments

[0070] Third preferred embodiment space-time block codes have 44 code matrices C(s) with diagonal code-defining matrices A_k: $A_1=\left[\begin{array}{cc}{}_{11}& 0\\ 0& {}_{21}\end{array}\right]$ $A_2=\left[\begin{array}{cc}{}_{12}& 0\\ 0& {}_{22}\end{array}\right]$ $A_3=A_1^{*}$ $A_4=A_2^{*}$

[0071] where the matrix $=\left[\begin{array}{cc}{}_{11}& {}_{12}\\ {}_{21}& {}_{22}\end{array}\right]$

[0072] is a unitary matrix, although it can be any arbitrary invertible matrix. ( invertible is equivalent to A₁ not a multiple of A₂.) The use of a unitary matrix to mix the pair of symbols s₁,s₂ and to mix the pair of symbols s₃,s₄ preserves energy and distance within the symbol constellation because unitary operations are isometric isomorphisms. Note that with I symbol s₁ is transmitted at time 1 on antenna 1 and symbol s₂ is transmitted at time 2 on antenna 2; whereas, with differing from the identity, symbols s₁ and s₂ are mixed to be transmitted at time 1 on antenna 1 and at time 2 on antenna 2. analogously mixes symbols s₃ and s₄ to be transmitted at time 3 on antenna 1 and at time 4 on antenna 2. Antennas 3 and 4 have the complex conjugates in reverse order and with a minus sign for s₃ and s₄.

[0073] The foregoing A_k define the following C and C: $C_{}=\left[\begin{array}{cc}{}_{11}{s}_1+{}_{12}{s}_2& 0\\ 0& {}_{21}{s}_1+{}_{22}{s}_2\end{array}\right]$ $C_{}=\left[\begin{array}{cc}{}_{11}*s_3+{}_{12}*s_4& 0\\ 0& {}_{21}*s_3+{}_{22}*s_4\end{array}\right]$

[0074] and as noted in section 2, these codes have 3-level orthogonality. Further, being unitary allows for a simple parameterization and optimization searching. Indeed, for in SU(2) there are angle parameters , , so that $=\left[\begin{array}{cc}{}^j\cos & {}^j\sin \\ -{}^{-j}\sin & {}^{-j}\cos \end{array}\right]$

[0075] Sufficient flexibility occurs with the angle parameters constrained to 0, 0/2, and <. Also, including a reflection effectively just relabels the symbols. Hence, take $\left(,\right)=\left[\begin{array}{cc}\cos & {}^j\sin \\ -{}^{-j}\sin & \cos \end{array}\right]$

[0076] That is, for each pair of parameter values (,) in the ranges (expressed in radians) 0/2 and <, the third preferred embodiment generates a block STTD space-time block code having the code matrix C(s) for the block of symbols s(s₁,s₂,s₃,s₄) as $C\left(,;s\right)=\left[\begin{array}{cc}{C}_{}\left(,;s\right)& C_{}\left(,;s\right)\\ -{C_{}\left(,;s\right)}^{*}& {C_{}\left({}^{}s\right)}^{*}\end{array}\right]$ $\mathrm{where}$ $C_{}\left(,;s\right)=\left[\begin{array}{cc}\left(\cos \right)s_1+\left({}^j\sin \right)s_2& 0\\ 0& \left(-{}^{-j}\sin \right)s_1+\left(\cos \right)s_2\end{array}\right]$ $C_{}\left(,;s\right)=\left[\begin{array}{cc}\left(\cos \right)s_3+\left({}^{-j}\sin \right)s_4& 0\\ 0& \left(-{}^{-j}\sin \right)s_3+\left(\cos \right)s_4\end{array}\right]$

[0077] And the diversity gain, G_D, for this space-time block code then can be determined by computation over pairs of symbol blocks as

[0078] and the coding gain, G_c, can be analogously determined as

[0079] As illustrated in FIG. 1, the preferred embodiment block STTD space-time block code design proceeds as follows.

[0080] 1) Given a symbol constellation and a minimum acceptable diversity gain G_D, determine a region in the parameter space 0/2, and < in which each space-time block code with code matrices C(,;s) has a diversity gain at least as large as G_D. This may be done by searching over the parameter space by evaluations on a grid or with a variational method.

[0081] 2) For the space-time block codes defined by the region from step 1), maximize with respect to coding gain, G_C. That is, find a subset (which may be a single point) of the region of step 1) which defines a space-time block code with maximum coding gain. Again, this may be done by searching over the parameter region by evaluations on a grid or with a variational method.

[0082] Alternatively, a minimum coding gain may be used, and other criteria imposed, such as the following preferred embodiment design method:

[0083] 1) Given a symbol constellation and a minimum acceptable diversity gain G_D, determine a region in the parameter space 0/2, and < in which each space-time block code with code matrices C(,;s) has a diversity gain at least as large as G_D.

[0084] 2) For the space-time block codes defined by the region from step 1), maximize with respect to coding gain, G_C. Then find a subset of the region of step 1) which defines space-time block codes with coding gain at least as large as 90% (or some other threshold fraction) of this maximum coding gain.

[0085] 3) For the space-time codes of step (2) with at least 90% of the maximum coding gain, minimize with respect to multiplicity.

[0086] Table 1 shows examples of such coding design for the QPSK constellation.

TABLE 1
Type	(deg)	(deg)	GD	GC
Real rotation	58.32	0	4	10.21
Complex rotation	45.54	128.52	4	16.00

[0087] Thus a preferred embodiment block STTD code for a QPSK constellation has the code matrices A_m defined by the unitary (,) with 0.795 and 2.243.

[0088] Again, the foregoing extends to PP for P an even integer greater than 4. In particular, with a (P/2)(P/2) unitary matrix, for k1, 2, . . . , P/2 take A_k to be a diagonal matrix with elements equal to the elements of the kth column of and take A_kP/2A_k*. Thus provides a mixing of symbols s₁, s₂, . . . , s_P/2 and a corresponding mixing of symbols s_P/21*, s_P/22*, . . . , s_P*.

[0089] 5. Transmitter

[0090] FIG. 2a illustrates a transmitter with a 44 space-time encoder which could be a preferred embodiment encoder and which may operate as follows. Presume a block of data bits (with FEC) is channel encoded, block interleaved, mapped to symbols, and then partitioned into blocks of length 4 with the nth block denoted s₁(n), s₂(n), s₃(n), s₄(n), where each symbol s_j(n) is interpreted as a discrete complex number (e.g., an element of a constellation such as QPSK, QAM, . . . ). The serial-to-parallel converter sends the block of four symbols to the 44 space-time encoder which outputs symbol streams to the four modulators and antenna drivers. Systems such as TDMA, CDMA, OFDM, and so forth could be used. Additionally (shown in broken lines in FIG. 2a), for the case of a CDMA system each of the four outputs is spread (multiplied by a common spreading code), and for the case of a multitone system the inverse discrete Fourier transform is applied. Then the (spread) transformed-symbol streams drive modulators (pulse generators (at the chip rate) with essentially cosine and sine pulses for the real (in phase) and imaginary (quadrature) components which modulate the carrier wave radiated by the corresponding antennas. For preferred embodiments not using CDMA, omit the spreading and operate the pulse generators at the symbol rate. For an OFDM (orthogonal frequency division multiplexing) system each sub-carrier (tone) would carry a portion of the transmitted symbols.

[0091] FIG. 2b illustrates the space-time encoder for the STTD-based 3-level orthogonality with rotation preferred embodiment of section 4 in terms of the unitary matrix . FIG. 2b uses 0 . . . T . . . 2T . . . 3T . . . 4T to indicate the four time intervals of the input four symbols and of the output four streams to the modulators. Note that (without the optional orthogonal transformation) during the first and third transmission intervals the first and third antennas are active and the second and fourth antennas are not, and conversely for the second and fourth transmission intervals.

[0092] Additionally, FIG. 2b shows an optional orthogonal transformation of the output space-time blocks with broken lines. This orthogonal transformation can solve a potential problem of large peak-to-average power ratio (PAR). In general, an orthogonal transform of the code matrix C for a block of symbols can reduce the PAR but not affect the performance of the space-time code. In particular, the orthogonal transformation could be a simple rotation of /4 plus reflection for antennas 1-2 and for antennas 3-4: $C->CW\mathrm{with}W=\left(1/2\right)\left[\begin{array}{cccc}1& 1& 0& 0\\ 1& -1& 0& 0\\ 0& 0& 1& 1\\ 0& 0& 1& -1\end{array}\right]$

[0093] Note that without W applied, the first antenna would successively transmit x₁, 0, x₃, 0, and the second antenna would successively transmit 0, x₂, 0, x₄; whereas, with W applied the first antenna would successively transmit x₁/{square root}2, x₁/{square root}2, x₃/{square root}2, x₃/{square root}2, and the second antenna would successively transmit x₂/{square root}2, x₂/{square root}2, x₄/{square root}2, x₄/{square root}2. Thus a large PAR due to transmitting 0s would be reduced. The orthogonal matrix W can be written as $W=\left[\begin{array}{cc}U& 0\\ 0& U\end{array}\right]$

[0094] where U is a 22 orthogonal matrix as in section 2, and thus this transformation preserves the 3-level orthogonality. Also, the reflection could be omitted to have a pure rotation which just reverses signs on antennas 2 and 4: $U=\left(1/2\right)\left[\begin{array}{cc}1& 1\\ -1& 1\end{array}\right].$

[0095] The forgoing orthogonal transformation also applies in the PP case to reduce PAR.

[0096] 6. Receiver

[0097] FIG. 3 illustrates a first preferred embodiment receiver with Q antennas In particular, each antenna connects to a down-converter which drives a RAKE detector having one or more tracking fingers (including for CDMA systems a sub-chip-rate sampler for A/D conversion, early-late chip synchronizer with despreading correlator or for multitone systems a DFT), path channel estimation to track resolvable received multipath signals, and maximal ratio combiner to combine the multipath detected signals using SNR weights. For the qth antenna, the received signal over four symbol periods corresponding to the block of four transmitted symbols s₁, s₂, s₃, s₄ is

[0098] where as in section 2, r_qr_q,1, r_q,2, r_q,3*, r_q,4*^T, ss₁, s₂, s₃*, s₄*^T, and H_q is the 44 channel-code matrix from the transmitter to the qth receive antenna. Detection with maximal ratio combining of multipaths for each antenna uses the detection statistic

[0099] where y_qy_q,1, y_q,2, y_q,3, y_q,4^T.

[0100] Next, combine the results of the Q antenna detectors: y_1qQ y_q; and write yy₁, y₂, y₃, y₄^T. Then split the four components of y into a pair of 2-vectors: yy_a y_b^T where y_ay₁, y₂^T and y_by₃, y₄^T. This split uses the 3-level orthogonality in that each H_q^H H_q has the section 2 form of vanishing upper right 22 and lower left 22 submatrices so y_a depends upon s and y_b depends upon s. Then for each 2-vector use a 22 detector as illustrated in FIG. 3. The detectors could be maximum likelihood detectors, linear detectors such as linear zero-forcing, iterative zero-forcing, linear MMSE, or iterative MMSE. The operation of each detector is described in the following paragraphs.

[0101] Also, y_a and y_b can be be written as follows

[0102] where the noise n_a and n_b are zero-mean, independent, and identical complex Gaussian 2-vectors with covariance matrix ²R. The maximum likelihood (ML) detector for the transmitted symbols s₁, s₂, s₃, s₄ is $\begin{array}{c}{\hat{s}}_a=\arg \underset{s{S}_1{S}_2}{\min }{\left(y_a-\mathrm{Rs}\right)}^H{R}^{-1}\left(y_a-\mathrm{Rs}\right)\\ {\hat{s}}_b=\arg \underset{s{S}_3^{*}{S}_4^{*}}{\min }{\left(y_b-\mathrm{Rs}\right)}^H{R}^{-1}\left(y_b-\mathrm{Rs}\right)\end{array}$ ₁s_a1, ₂s_a2, ₃s_b1*, ₄s_b2*

[0103] where x_p denotes the p-th element of vector x. Here, S_p is the signal constellation (set) corresponding to the symbol s_p and _p indicates the hard-decision estimate of s_p.

[0104] The linear detectors (linear ZF and linear MMSE) can be written as:

[0105] where D. is a single-input single-output maximum likelihood symbol decision device which generates a hard-decision symbol estimate.

[0106] A 22 iterative detector (iterative ZF or MMSE) with detection ordering is composed of two linear detection stages with decision feedback interference cancellation in between. Detection ordering is used to reduce the effect of decision feedback error. The idea is to first detect the most reliable symbol, and then subtract its contribution from the received signal. Let M_i,j denote the (i,j)-th element of matrix M and M_l,: denote the i-th row of matrix M. The complete iterative MMSE detector with (MSE-based) detection ordering can be described as follows: $e_1={\left[{\left(R+{}^2{}_a^{-1}\right)}^{-1}\right]}_{1,1},e_2={\left[{\left(R+{}^2{}_a^{-1}\right)}^{-1}\right]}_{2,2}$ $\mathrm{if}{e}_1{e}_2:z_1={\left[{\left(R+{}^2{}_a^{-1}\right)}^{-1}\right]}_1,:y_a->{\hat{s}}_1=\left\{z_1\right\}$ $z_2=\frac{{\left[y_a\right]}_2-{\left[R\right]}_{2,1}{\hat{s}}_1}{{\left[R\right]}_{2,2}+{{}^2\left[{}_a\right]}_{2,2}^{-1}}->{\hat{s}}_2=\left\{z_2\right\}$ $\mathrm{else}$ $z_2={\left[{\left(R+{}^2{}_a^{-1}\right)}^{-1}\right]}_2,:y_a->{\hat{s}}_2=\left\{z_2\right\}$ $z_1=\frac{{\left[y_a\right]}_1-{\left[R\right]}_{1,2}{\hat{s}}_2}{{\left[R\right]}_{1,1}+{{}^2\left[{}_a\right]}_{1,1}^{-1}}->{\hat{s}}_1=\left\{z_1\right\}$ $\mathrm{end}$ $e_3={\left[{\left(R+{}^2{}_b^{-1}\right)}^{-1}\right]}_{1,1},e_4={\left[{\left(R+{}^2{}_b^{-1}\right)}^{-1}\right]}_{2,2}$ $\mathrm{if}{e}_3{e}_4:z_3={\left[{\left(R+{}^2{}_b^{-1}\right)}^{-1}\right]}_1,:y_b->{\hat{s}}_3={\left\{z_3\right\}}^{*}$ $z_4=\frac{{\left[y_b\right]}_2-{\left[R\right]}_{2,1}{\hat{s}}_3^{*}}{{\left[R\right]}_{2,2}+{{}^2\left[{}_b\right]}_{2,2}^{-1}}->{\hat{s}}_4={\left\{z_4\right\}}^{*}$ $\mathrm{else}$ $z_4={\left[{\left(R+{}^2{}_b^{-1}\right)}^{-1}\right]}_2,:y_b->{\hat{s}}_4={\left\{z_4\right\}}^{*}$ $z_{31}=\frac{{\left[y_b\right]}_1-{\left[R\right]}_{1,2}{\hat{s}}_4^{*}}{{\left[R\right]}_{1,1}+{{}^2\left[{}_b\right]}_{1,1}^{-1}}->{\hat{s}}_3={\left\{z_3\right\}}^{*}$ $\mathrm{end}$

[0107] The complete iterative ZF detector with (MSE-based) detection ordering can be described as follows: $e_1={\left[R^{-1}\right]}_{1,1},e_2={\left[R^{-1}\right]}_{2,2}$ $\mathrm{if}{e}_1{e}_2:z_1={\left[R^{-1}\right]}_1,.y_a->{\hat{s}}_1=\left\{z_1\right\}$ $z_2=\frac{{\left[y_a\right]}_2-{\left[R\right]}_{2,1}{\hat{s}}_1}{{\left[R\right]}_{2,2}}->{\hat{s}}_2=\left\{z_2\right\}$ $z_3={\left[R^{-1}\right]}_1,y_b->{\hat{s}}_3={\left\{z_3\right\}}^{*}$ $z_4=\frac{{\left[y_b\right]}_2-{\left[R\right]}_{2,1}{\hat{s}}_3^{*}}{{\left[R\right]}_{2,2}}->{\hat{s}}_4={\left\{z_4\right\}}^{*}$ $\mathrm{else}$ $z_2={\left[R^{-1}\right]}_2,y_a->{\hat{s}}_2=\left\{z_2\right\}$ $z_1=\frac{{\left[y_a\right]}_1-{\left[R\right]}_{1,2}{\hat{s}}_2}{{\left[R\right]}_{1,1}}->{\hat{s}}_1=\left\{z_1\right\}$ $z_4={\left[R^{-1}\right]}_2,y_b->{\hat{s}}_4={\left\{z_4\right\}}^{*}$ $z_3=\frac{{\left[y_b\right]}_1-{\left[R\right]}_{1,2}{\hat{s}}_4^{*}}{{\left[R\right]}_{1,1}}->{\hat{s}}_3={\left\{z_3\right\}}^{*}$ $\mathrm{end}$

[0108] 8. Simulation Results

[0109] FIGS. 4 and 5 show simulation results for the case of Q1 receive antenna with QPSK and 16QAM modulation. Two code designs in Table 1 (real and complex rotation) that are optimized for QPSK are used.

[0110] 9. Modifications

[0111] The preferred embodiments can be modified in various ways while retaining one or more of the features of block STTD code matrices, partial (3-level) orthogonality, and diversity gain and coding gain optimization.

[0112] For example, the block size in the STTD could be larger than 22 with the orthogonality corresponding to the block size.