Reduced complexity detection and decoding for a receiver in a communication system

(54) Title: REDUCED COMPLEXITY DETECTION AND DECODING FOR A RECEIVER IN A COMMUNICATION SYSTEM (30) Unionist Priority: 01/02/2006 us 11 / 345,976,

11/18/2005 US 60 / 738,159 (73) Holder (s): Qualcomm Incorporated (72) Inventor (s): Bjorn Bjerke, Irina Medvedev, Jay Rodney Walton, John W. Ketchum, Mark S. Wallace (74) Attorney (s): Montaury Pimenta, Machado & Lioce (86) International Order: pct US2006045031 de

11/20/2006 (87) International Publication: wo 2007 / 06202ide

05/31/2007 (57) Abstract: detection of reduced complexity and DECODING FOR A RECEIVER IN A COMMUNICATION SYSTEM Techniques for performing detection and decoding in a receiver are described. In a scheme, the receiver obtains R received symbol streams for M data streams transmitted by a transmitter, performs spatial processing of the receiver on received symbols to obtain detected symbols, performs LLR computation independently for each of the best D data streams , and performs the LLR computation together for the remaining MD data streams, where M> D> 1 and M> 1. The best D data streams can be selected based on SNR and / or other criteria. In another scheme, the receiver performs LLR computation independently for each of the D best data streams, performs joint LLR computation for the remaining MD data streams, and reduces the number of hypotheses to be considered for joint LLR computation by performing a search for candidate hypotheses using list sphere detection, MCMC technique or some other search technique.

<img file="BRPI0618716A2_D0001.tif" />

<img file="BRPI0618716A2_D0002.tif" />

DETECTION OF REDUCED COMPLEXITY AND DECODING FOR A RECEIVER IN A COMMUNICATION SYSTEM

The present application claims priority of provisional US patent application No. 60 / 738,159, entitled REDUCED COMPLEXITY INTERATIVE DETECTION AND DECODING FOR MIMO-OFDM SYSTEMS, filed on November 18, 2005, designated for the assignee thereof and incorporated herein by reference.

FUNDAMENTALS

Field

The present description refers to communication, and more specifically to techniques for performing detection and decoding on a receiver in a communication system. Foundations

A multiple input and multiple output (MIMO) communication system employs multiple transmitting antennas (T) on a transmitter and multiple receiving antennas (R) on a receiver for data transmission. A MIMO channel formed by the T transmitting antennas and R receiving antennas can be decomposed into M spatial channels, where M &lt;min (T, R). Spatial M channels can be used to transmit data in order to achieve the highest overall throughput and / or greater reliability.

transmitter can encode and transmit M data streams in parallel through T transmission antennas. The receiver obtains R symbol streams received through R receiving antennas, performs MIMO detection to separate the M data streams, and decodes the detected symbol streams to retrieve the transmitted data streams. To achieve optimal performance, the receiver needs to evaluate many hypotheses for all possible data bit streams that may have been transmitted based on all the information available at the receiver. Such

<img file="BRPI0618716A2_D0003.tif" />

2/4 exhaustive and intense search in terms of computing and prohibitive for many applications.

There is, therefore, a need in the art to create techniques for performing detection and decoding with reduced complexity while achieving good performance.

SUMMARY

Techniques for performing detection and decoding with reduced complexity while achieving good performance are described here. These techniques are embodied in several low complexity detection schemes as described below.

In a low complexity detection scheme, a receiver obtains R symbol streams received for M data streams transmitted by a transmitter, performs spatial processing on the receiver (or combined spatial filtering) on ​​received symbols to obtain detected symbols, performs the computation of log-likelihood ratio (LLR) independently for each of the D best data streams, and performs LLR computation together for MD remaining data streams, where in general M&gt; D&gt; leM&gt; l. The best D data streams can be selected based on signal-to-noise plus interference (SNR) and / or other criteria. In another low complexity detection scheme, the receiver performs LLR computation independently for each of the best D data streams, performs LLR computation together for the remaining MD data streams, and reduces the number of assumptions to consider for the computing LLR together by conducting a search for candidate hypotheses using list sphere detection. Markov chain Monte Carlo technique or some other search technique.

3/43

For both detection schemes, scaling is reduced from M to M - D by performing LLR computation by flow for the best D data flows. The reduction in scaling can substantially reduce the number of assumptions to be considered for joint LLR computing for the remaining MD data streams. The number of hypotheses can be further reduced by conducting a search for candidate hypotheses. These detection schemes can be used for (1) a single-pass receiver that performs detection and decoding only once and (2) an iterative receiver that performs detection and decoding iteratively. These and other detection schemes are described in detail below.

Various aspects and modalities of the invention are also described in greater detail below.

BRIEF DESCRIPTION OF THE DRAWINGS

The characteristics and nature of the present invention will become more apparent from the detailed description presented below when taken in conjunction with the drawings in which similar reference characters identify corresponding parts from all views.

Figure 1 illustrates a block diagram of a transmitter and a receiver;

Figure 2 illustrates a block diagram of a transmission data processor (TX) and a space processor TX in the transmitter;

Figure 3 illustrates a block diagram of a space receiving processor (RX) and an RX data processor for a single pass receiver;

4/43

Figure 4 illustrates a block diagram of an RX space processor and an RX data processor for an iterative receiver;

Figure 5 illustrates a flowchart for a reduced scaling detection scheme;

Figure 6 illustrates an equipment for the reduced dimension detection scheme;

Figure 7 illustrates an illustrative search tree for the detection of list sphere;

Figure 8 illustrates a flow chart for a reduced order detection scheme;

Figure 9 illustrates an equipment for the reduced order detection scheme.

DETAILED DESCRIPTION

The term illustrative is used here to mean serving as an example, case or illustration. Any modality or design described here as illustrative should not necessarily be considered as preferred or advantageous over other modalities or designs.

detection and decoding techniques described

At

<td>here they can</td><td>to be</td><td>used</td><td>for</td><td>several</td><td>systems</td><td>in</td>

<td>Communication</td><td>we</td><td colspan="2">which multiples</td><td>streams</td><td>of data</td><td>are</td>

<td>transmitted</td><td>in</td><td>parallel</td><td colspan="2">across</td><td>a channel</td><td>in</td>

<td>Communication.</td><td>Per</td><td>example,</td><td>these</td><td colspan="2">techniques can</td><td>to be</td>

<td>used</td><td>for</td><td>a system</td><td>MIMO</td><td>with one</td><td>sub-band</td><td>in</td>

system frequency

MIMO with multiple sub-bands, single, an Access system

Code Division Multiple (CDMA), a Frequency Division Multiple Access (FDMA) system, a Time Division Multiple Access (TDMA) system, and so on. Multiple sub-bands can be obtained with orthogonal frequency division multiplexing (OFDM), single carrier frequency division multiple access (SC-FDMA), or some other

5/43 modulation. OFDM and SC-FDMA divide the system bandwidth as a whole into multiple orthogonal (L) sub-bands, which are also called subcarriers, tones, compartments, and so on. Each subband is associated with a subcarrier that can be modulated independently with the data. In general, the modulation symbols are sent in the frequency domain with OFDM and in the time domain with SC-FDMA. For the sake of clarity, much of the description below is for a MIMO system that uses OFDM.

Figure 1 illustrates a block diagram of a modality of a transmitter 110 and a receiver 150 in a MIMO 100 system.

multiple antennas

Transmitter 110 is equipped with (T), and receiver 150 is equipped with multiple antennas (R). For downlink (or direct link) transmission, transmitter 110 may be part of, and may contain all or some of the functionality of a base station, an access point, a Node B, and so on. The receiver 150 can be part of and can contain all or part of the functionality of a mobile station, a user terminal, a user equipment, and so on. For uplink (or reverse link) transmission, transmitter 110 can be part of a mobile station, a user terminal, user equipment, and so on, and receiver 150 can be part of a base station, a point access, a Node B and so on.

In the transmitter 110, a TX data processor 120 receives traffic data from a data source 112 and processes (for example, formats, encodes, merges and maps symbols) traffic data to generate data symbols, which are symbols modulation for traffic data. A TX 130 space processor multiplexes the data symbols with pilot symbols, which are

6/43 modulation for the pilot. A pilot is a transmission that is known a priori by both the transmitter and the receiver and can also be referred to as a sequencing signal, a reference, a preamble and so on. The TX 130 space processor performs spatial processing of the transmitter and provides T transmission symbol streams for T transmitting units (TMTR) 132a through 132t. Each transmitting unit 132 processes (for example, modulates into OFDM, converts to analog, filters, amplifies and converts upwardly) its transmission symbol stream and generates a modulated signal. The T modulated signals from transmitter units 132a to 132t are transmitted from antennas 134a to 134t, respectively.

At receiver 150, R antennas

152a to 152r receive the

T modulated signals, and each antenna

152 provides a received signal to a respective receiving unit (RCVR) 154.

Each receiving unit 154 processes its received signal in a complementary way to the processing performed by the transmitting units 132 to obtain received symbols, provides received symbols for traffic data to an RX 160 space processor and provides received symbols to the pilot for a 194 channel processor . Channel processor 194 estimates the MIMO channel response from transmitter 110 to receiver 150 based on the symbols received for the pilot (and possibly the symbols received for traffic data) and provides channel estimates for the space processor RX 160 The space processor

RX 160 detects the received symbols for traffic data with channel estimates and provides smooth decisions, which can be represented by

LLRs as described below.

A data processor

RX 170 additionally processes (for example, deinterleaves and decodes) smooth and

7/43

Λprovides decoded data for a 172 data store. Detection and decoding can be performed with a single pass through processors 160 and 170 or iteratively between processors 160 and 170.

receiver 150 can send feedback information to assist transmitter 110 in controlling data transmission to receiver 150. The feedback information may indicate a particular transmission mode for use for transmission, a particular rate or packet format for use for each data stream, acknowledgments (ACKs) and / or negative acknowledgments (NAKs) for packets decoded by receiver 150, information about the status of the channel, and so on, or any combination of them. Feedback information is processed (for example, encoded, merged, and mapped in symbols) by a TX 180 signaling processor, multiplexed with pilot symbols and spatially processed by a TX 182 space processor and further processed by transmitting units 154a to 154r for generate R modulated signals, which are transmitted through antennas 152a to 152r.

At transmitter 110, the R modulated signals are received by antennas 134a to 134t, processed by receiver units 132a to 132t, spatially processed by an RX 136 space processor and further processed (for example, deinterleaved and decoded) by an RX 138 signaling processor to retrieve feedback information. A controller / processor 140 controls the data transmission to the receiver 150 based on the feedback information received. A channel processor 144 can estimate the response to the MIMO channel from the receiver 150 to the transmitter

8/43

110 and can derive the spatial mapping matrices used by the TX 130 space processor.

Controllers / processors 140 and 190 control operation on transmitter 110 and receiver 150, 5 respectively. Memories 142 and 192 store data and program codes for transmitter 110 and receiver 150, respectively.

Figure 2 shows a block diagram of a TX 120 data processor and TX 130 spatial processor modality in transmitter 110. For this modality, a common coding scheme is used for all data streams, and a code rate A separate modulation scheme can be used for each data stream. For the sake of clarity, the description below considers 15 that M data streams are sent in M ​​spatial channels.

Within the TX 120 data processor, an encoder 220 encodes traffic data according to an encoding scheme and generates code bits. The coding scheme can include a convolutional code, a Turbo code, a low density parity check code (LDPC), a cyclic redundancy check code (CRC), a block code, and so on, or a combination of them. A demultiplexer (Demux) 222 demultiplexes (or analyzes) the code bits in M ​​streams and 25 provides the M code bit streams for M sets of processing units. Each set includes a punching unit 224, a channel interleaver 226, and a symbol mapper 228. Each punching unit 224 punctures (or eliminates) code bits, as needed, to obtain a selected code rate for its flow and supplies the retained code bits to an associated channel interleaver 226. Each channel interleaver 226 interleaves ( or reorder) your code bits based on a

9/43 interleaving scheme and provides interleaving bits for an associated symbol mapper 228. Interleaving can be performed separately for each data stream (as shown in figure 2) or through all or some of the data streams (not shown in figure 2).

Each 228 symbol mapper maps its interleaved bits according to a modulation scheme selected for its flow and provides a flow of data symbols {s_m}. The symbol mapping for flow m can be achieved by (1) grouping sets of Q_m bits to form Q values_m-bit, where Q_m &gt; 1, and (2) mapping each Qm-bit value to one of the points 2 ⁰⁷¹ in a signal constellation for the selected modulation scheme. Each signal point mapped is a complex value and corresponds to a data symbol. The symbol mapping can be based on Gray mapping or non-Gray mapping. With Gray mapping, neighboring points in the signal constellation (in both horizontal and vertical directions) differ by only one of the Q_m bit positions. Gray mapping reduces the number of bit errors for most likely error events, which corresponds to a received symbol being mapped to a location close to the correct location, in which case only one coded bit would be detected with an error. With non-Degree mapping, neighboring points may differ by honey from a bit position. Non-Gray mapping can result in greater independence between encoded bits and can improve performance for iterative detection and decoding.

Within the TX 130 space processor, a multiplexer (Mux) 230 receives the M data symbol streams from symbol mappers 228a to 228m and maps the data symbols and pilot symbols to the appropriate sub-bands in each symbol period. A matrix multiplier 232 multiplies the data and / or the pilot symbols to

10/43 each subband 1 with a spatial mapping matrix P (l) and provides transmission symbols for that subband. Different spatial mapping matrices can be used for different transmission modes, and different spatial mapping matrices can be used for different sub-bands for some transmission modes, as described below.

Figure 2 illustrates an embodiment in which a common coding scheme and separate code rates and modulation schemes can be used for M data streams. Different code rates can be achieved for M data streams by using different punching patterns for these streams. In another embodiment, a common coding scheme and a common code rate are used for all data streams, and separate modulation schemes can be used for M data streams. In another modality, a scheme of

<td>coding</td><td>common,</td><td>an</td><td>rate of</td><td>code</td><td>common</td><td colspan="2">, and a scheme</td>

<td>modulation</td><td>common</td><td>are</td><td colspan="2">used for</td><td>all</td><td>the M flows</td><td>in</td>

<td>Dice. In</td><td>another</td><td colspan="2">modality,</td><td>each</td><td>flow</td><td>of data</td><td>is</td>

<td>processed</td><td colspan="2">based</td><td>on a</td><td>scheme</td><td>in</td><td>coding</td><td>and</td>

modulation selected for that data stream. In general, the same coding schemes or different schemes, the same code rates or different code rates, and the same modulation schemes or different modulation schemes can be used for the M data streams. In addition, the same coding schemes or different coding schemes, the same code rates or different code rates, and the same modulation schemes or different modulation schemes can be used across the subbands.

transmitter 110 typically encodes each packet separately. In one modality, the M flows of

11/43 data is encoded together so that a single packet can be sent across multiple spatial channels (for example, all M). In another modality, the M data streams are coded independently so that each packet is sent on a spatial channel. In another Danish modality, some data streams are encoded together while the remaining data streams are coded independently.

For the sake of clarity, the following description considers that a data stream is sent on each spatial channel. The terms data flows and space channel are thus interchangeable for much of the description below. The number of data streams can be configurable and

<td colspan="2">can be selected</td><td>based</td><td>in</td><td>conditions</td><td colspan="2">channel and / or</td>

<td>others</td><td>factors. Per</td><td>reasons</td><td>in</td><td>clarity, the</td><td>description</td><td>The</td>

<td>follow</td><td>considers that</td><td>M flows</td><td>in</td><td>data are</td><td>sent in</td><td>M</td>

<td>channels</td><td>space.</td><td></td><td></td><td></td><td></td><td></td>

<td></td><td>Figure 3</td><td>illustrates</td><td colspan="2">a diagram of</td><td>blocks of</td><td>one</td>

<td colspan="2">space processor</td><td>RX 160a.</td><td>it is a</td><td colspan="2">data processor</td><td>RX</td>

170a for a single pass receiver. Processors 160a and 170a are a modality of processors 160 and 170, respectively, at receiver 150 in figure 1. For this modality, processors 160 a and 170a perform detection and decoding with a single pass through each of processors 160a and 170a .

Within the RX 160a space processor, a spatial filter matrix computing unit 308 receives channel estimates from channel processor 194 and derives the spatial filter matrices based on the channel estimates and spatial mapping matrices used by transmitter 110 , as described below. A MIMO 310 detector obtains the symbols received from the R receiver units 154a to 154r, the channel estimates of the

12/43 channel processor 194, and the spatial filter arrays of unit 308. The MIMO 310 detector performs detection as described below and provides K smooth decisions for K code bits of M data symbols sent in each subband in each symbol period used for data transmission. A soft decision is a multi-bit value that is an estimate of a bit of transmitted code. Soft decisions can be represented as LLRs and can be referred to as extrinsic LLRs. If M data symbols are sent in a subband in a symbol period, κ = Σ &lt;λ.

then K can be computed as, where Q_m is the number of code bits used to form a data symbol for flow m. If the same modulation scheme is used for all data streams M, then K can be computed as K = M · Q, where Q is the number of code bits for each data symbol.

Within the RX 170a data processor, channel deinterleavers 316a to 316m receive extrinsic LLRs for M data streams. Each channel deinterleaver 316 deinterleaves the extrinsic LLRs for its flow in a complementary way to the interleaving performed by the channel interleaver 226 for that flow. A multiplexer 318 multiplexes (or serializes) the deinterleaved LLRs of channel deinterleaver 316a to 316m. A decoder 320 decodes the interleaved LLRs and provides decoded data. Detection and decoding are described in detail below.

Figure 4 illustrates a block diagram of an RX 160b space processor and an RX 170b data processor for an iterative receiver. Processors 160b and 170b are another modality of processors 160 and 170, respectively, on receiver 150. For this modality,

13/43 processors 160b and 170b perform iterative detection and decoding.

Within the RX 160b space processor, a unit 408 derives the spatial filter matrices based on the channel estimates and the spatial mapping matrices used by the transmitter 110. A MIMO 410 detector obtains the symbols received from the R 154a to 154r receiver units, the channel estimates from channel processor 194, the spatial filter arrays of unit 408, and a priori LLRs from a decoder 420. A priori LLRs are denoted as L_The(B_k) and represent a priori information from the decoder 420. The MIMO 410 detector performs the detection as described below and provides K detector LLRs for K code bits of M data symbols sent in each subband in each symbol period used for data transmission. Detector LLRs are denoted as L (bk). K adders 412a to 412k subtract LLRs a priori from the detector LLRs and provide extrinsic LLRs, which are denoted as L_and(B_k). The extrinsic LLRs represent extrinsic or new information from the MIMO 410 detector to the decoder 420.

Within the RX 170b data processor, an M 416 flow channel deinterleaver deinterleaves the extrinsic LLRs for each flow in a complementary way to the interleaving performed by the channel interlayer 226 for that flow. Channel deinterleaver 416 can include M channel deinterleaver 316a to 316m illustrated in figure 3. A multiplexer 418 serializes the deinterleaved LLRs, which are denoted as L^D_and(B_k). A decoder 420 decodes the interleaved LLRs and provides decoder LLRs. An adder 422 subtracts the interleaved LLRs from the decoder LLRs and provides LLRs a priori, which represent extrinsic information from the

14/43 decoder 420 for the MIMO 410 detector for the next iteration. The a priori LLRs for the MIMO 410 detector are demultiplexed in M ​​flows by a demultiplexer 424. An M 426 flow channel interleaver interleaves the a priori LLRs for each channel in the same way as that performed by channel interlayer 226 for that flow. The channel interleaver 426 can include M channel interleaver 226a to 226m illustrated in figure 2. 0 channel interleaver 426 provides the interleaved a priori LLRs for the next iteration for the MIMO 410 detector.

space processor RX 160b and data processor RX 170b can perform any number of iterations. In one embodiment, processors 160b and 170b perform a predetermined number of iterations (for example, 4, 68 or more iterations). In another modality, processors 160b and 170b perform an iteration, then check if a packet was decoded correctly and / or if it was coded in error or if the decoder reliability metric is low. Error detection can be achieved with a CRC and / or some other error detection code. Processors 160b and 170b can therefore perform a fixed number of iterations or a variable number of iterations up to a maximum number of iterations. Iterative detection and decoding are described in detail below.

The symbols received at the receiver 150 can be expressed as:

y (O = fi_cb(Z) -P (Z) -s (Z) + n (Z) = H (Z) -s (Z) + n (Z), with 1 = 1, .., L, Eq (l) where s (l) is a vector Μ x 1 with M data symbols sent in subband 1;

P (l) is a spatial mapping matrix T x M used by transmitter 110 for subband 1;

15/43

H_ch(l) is a MIMO RxT channel response matrix for subband 1;

0 (Ο = Η_λ(Ι) · Ρ (Ζ) a channel response matrix

Effective MIMO RxM for subband 1;

y (l) is an Rxl vector with. R symbols received for subband 1; and n (l) is a noise vector Rxl for subband 1.

The noise can be considered as additional white Gaussian noise (AWGN) with an average vector equal to zero and a -2 r 2 covariance matrix equal to where is the noise variation and I_ is the identity matrix. The effective MIMO channel response H (l) includes the actual MIMO channel response H_Çh (l) θ the spatial mapping matrix P (l) used by transmitter 110.

In one embodiment, a MIMO detector (for example, MIMO 310 or 410 detector) performs detection separately for each subband based on the received symbols and the channel estimate for that subband and, if available, the prior LLRs for the data symbols sent in that subband. In another mode, the MIMO detector performs detection together for multiple sub-bands. A decoder (for example, decoder 320 or 420) performs decoding in a sequence of LLRs for a packet, which can be transmitted in one or more multiple sub-bands. For the sake of clarity, subband index 1 is omitted in the description below.

A packet can be divided into multiple blocks, with each block containing K bits of code. The K code bits for each block can be mapped to M data symbols, as follows:

s = map (b),

Eq (2)

16/43 where s = [si, S2 ... s_M) is a vector with M data symbols;

- ~ [—i ^ 2 · = (¾ ·· b₂₁... ... 6^; ...6^^] = ^ 6, ...6_K]

6 · 62 èm is a vector with K bits of code for a block;

<td>B_m is a vector with Q_m bits</td><td>in</td><td>code</td><td>used</td>

<td>to form data symbol s_m to 0</td><td colspan="2">flow m;</td><td></td>

<td>B_m,_q for m = l, ..., M eq =</td><td> 1, ·</td><td>• · rQm θ</td><td>0 bit</td>

<td>q code in vector b_m; and</td><td></td><td></td><td></td>

<td>B_k, with k = 1, ..., K, is 0 bit</td><td>in</td><td>code</td><td>k in vector</td>

B.

There is a one-to-one mapping between a given bit vector b and a corresponding data vector s. In general, Q can be the same or different for M data symbols sent in a given sub-band, and K can be the same or different for L sub-bands.

An optional receiver for the transmission scheme illustrated in equation (1) is a maximum probability sequence (ML) receiver that performs detection and decoding together for the entire package. This ideal receiver will make joint decisions on all data bits in the packet using knowledge of the correlation introduced by the coding scheme across the blocks, sub-bands, and OFDM symbols for the packet. The ideal receiver would perform an exhaustive search through all possible sequences of data bits that may have been transmitted to the packet to find the sequence that is most likely to have been transmitted. This ideal receiver would be prohibitively complex.

A receiver that performs iterative detection and decoding, for example, as illustrated in figure 4, can achieve almost ideal performance with less complexity. 0 detector and decoder

17/43 compute smooth decisions in the code bits and exchange this information iteratively, which increases the reliability of smooth decisions with the number of iterations performed. The MIMO detector and decoder can each be implemented in several ways.

In one embodiment, the MIMO detector is a maximum posterior probability (MAP) detector that minimizes the probability of error for each code bit and provides a smooth decision for each code bit. The MAP detector provides smooth decisions in the form of a posteriori probability (APPs) which is often expressed as LLRs. The detector LLR for code bit b_k, L (b_k), can be expressed as:

W ' ^with k = 1 * - ·. K, Eq (3) where^P{^-⁺¹ly) £ _The probability that the code bit b_k be +1 according to the vector received y; and

Úy} _aa probability that the code bit b_k be -1 according to a received vector y,

The detector LLR can be separated into two parts, as follows:

L (b_k) = L_The(B_k) + L_and(B_k) Eq (4) where L_The(B_k) is the a priori LLR for the code bit b_kprovided by the decoder or possibly other sources for the MIMO and L detector_and(B_k) is the extrinsic LLR for the code bit b_k provided by the MIMO detector to the decoder. The a priori LLR for the code bit b_k can be expressed as:

m = -i}

Eq (5) where P {b_k= + 1} is the probability that the code bit b_k be +1; and

18/43

P (b_k= -1} is the probability that the code bit b_k be -1.

The MAP detector can be a log-MAP detector, a max-log-MAP detector, or some other type of MAP detector.

Extrinsic LLR of a log-MAP detector, which is called the LLR log-MAP, can be computed as:

Eq (6) where transmission hypotheses are;

ÍUi is a vector with all code bits in the vector - except for code bit b_k;

—Is a vector with LLRs a priori for all code bits in;

represents a Euclidean distance cost function; and^T denotes a transpose.

Equation (6) illustrates an expression for

Extrinsic LLR from the log-MAP detector. Extrinsic LLR can also be expressed in other ways. The receiver typically derives S, which is an estimate of the effective MIMO channel response matrix H, and uses S in LLR computation. For the sake of simplicity, the description here does not assume any channel estimation error, so S = H.

Equation (6) is evaluated for each code bit, in the transmitted bit vector b. For each bit of code b_k, 2^K creates a hypothesis of bit vectors for all possible sequences of the code bits {bi ... b_K} (or all possible combinations of code bit values) that

19/43

ι.

»May have been transmitted to vector b are considered. 2^K_1 creates hypotheses of bit vector vectors £ having bk = + l, and the other 2^k_1 creates hypotheses of bit vectors having bk = -l. Each hypothesized bit vector has a corresponding hypothesized data vector -. Ά expression

<td>within the sum</td><td>is</td><td>computed for</td><td>each</td><td>vector of</td><td>bit</td>

<td>hypothesized</td><td>for</td><td>obtaining a</td><td colspan="3">result for that</td>

<td>bit vector. The</td><td colspan="3">results for 2^K1 To the</td><td>vectors of</td><td>bit</td>

<td>hypothesized 6</td><td>with</td><td>B_k= + l are added</td><td>for</td><td>obtaining</td><td>one</td>

<td>total result</td><td>for</td><td>the numerator. The</td><td colspan="2">results for</td><td>2K-i</td>

bit vectors hypothesized with b_k= -l are added together to obtain a total result for the denominator. LLR logMAP for code bit b_k is equal to the natural logarithm (ln) of the total result for the numerator divided by the total result for the denominator.

A max-log-MAP detector approximates LLR log-MAP in equation (6) and provides an LLR max-log-MAP, as follows:

W '| g «{- ^ - 111 - Η ί II¹ + Ê&gt;

The max-log-MAP approximation in equation (7) replaces the sums in equation (6) with operations max {}. Only a small performance degradation typically results from using the max-log-MAP approach. Other approximations of the LLR logMAP in equation (6) can also be used.

The log-MAP detector in equation (6) and the max-log-MAP detector in equation (7) make joint decisions on the symbols received in the vector y and compute the extrinsic LLRs for the code bits associated with these received symbols. In order to optimally compute extrinsic LLRs, each MAP detector performs a

20/43 exhaustive search through all possible combinations of data symbols that may have been transmitted to the vector s. This exhaustive search is intense in terms of computing and can be prohibitive for many applications. Ά LLR computation complexity is exponential in the number of bits (K) in the transmitted bit vector b for both log-MAP and max-log-MAP detectors. In particular, hypotheses 2^K are considered by both MAP detectors for each bit of code bk- Several detection schemes with reduced complexity are described below.

To reduce computational complexity, the receiver can perform spatial processing of the receiver (or combined spatial filtering) on ​​the received symbols to obtain detected symbols and can then perform LLR computation independently for each detected symbol. The detected symbols are estimates of the data symbols transmitted by the transmitter. The receiver can perform spatial processing of the receiver based on a technique of imposing minimum mean squared error with maximum technical ratio combination. A filter matrix based on the ZF, MMSE technique or • Η + ^ -ΐΓ-Η, zero (ZF), a (MMSE) technique, a (MRC) technique, or some other spatial one can be derived from MRC, as follows:

Eq (8)

Eq (9)

<img file="BRPI0618716A2_D0004.tif" />

where .Η} '¹ ;

= diag [H^H H] ¹ ;

- &lt;/&gt; and M,_nn. _s^_{the race}M x R spatial filter strips for ZF, MMSE and MRC techniques; and

21/43 ^H denotes a conjugated transpose.

The spatial processing of the receiver can be expressed as:

§ = My, Eq (ll) where M can be equal to M_zf, M ^^ p or M_mrc; e is a vector Μ x 1 of detected symbols and is an estimate of data vector s.

LLR computation can be performed independently for each detected symbol. Extrinsic LLRs can be computed for Q code bits_mof each symbol detected with the max-log-MAP detector, as follows:

Á ~ 'I ~ I llajTí | ~ 2nd⁷ '^Sm ~ ^Sm I where is the element m of 5;

is a data symbol

Eq (12) hypothesized for transmitted data symbol s_m;

- »(«] is a vector with all code bits for the data symbol s_m except for the code bit b_{m / q};

- «μη, Μ _{a ve}tor with LLRs a priori for all five code bits in; and

L_and(B_m,_q) is extrinsic LLR for the code bit bm, q ·

Equation (12) is evaluated for each code bit in each transmitted bit vector b_m, with m = l, ..., M. For each bit of code b_m,_q in bit vector b_m, the hypothetical bit vectors 2 °¹¹¹ 6 «for all possible code bit strings {b_m, i. . .B_m,} that may have been passed to vector b_m are considered. Each vector

The hypothesized bit 22/43 has a corresponding hypothesized data symbol. The expression within the

<td>max operation {}</td><td>is computed</td><td>Stop</td><td>every vector</td><td>in</td><td>bit</td>

<td>hypothesized</td><td>to obtain</td><td>in</td><td colspan="2">a result for</td><td>that</td>

<td colspan="2">vector. The results for</td><td>the</td><td> 2°^{111 1} vectors</td><td>in</td><td>bit</td>

<td>hypothesized 1</td><td>with b_m, _q=+1</td><td>are</td><td>used in</td><td colspan="2">first</td>

<td>max operation {}.</td><td>The results</td><td>for</td><td>the 2nd^111-1 vector</td><td>s of</td><td>bit</td>

<td>hypothesized 1</td><td>with b_m,_q= -l</td><td>are</td><td>used in</td><td colspan="2">Monday</td>

max operation {}. ,

The complexity of the receiver's spatial processing is linear in number of data flows (M) and does not depend on the size of the signal constellation. Computing extrinsic LLRs by flow reduces the number of hypotheses to be evaluated from 2^M'^Q for M-2^Q, considering that the same modulation scheme is used for all M data streams. Flow LLR computing can substantially reduce computing complexity, but it can result in greater performance degradation than desired.

In one aspect, the receiver performs the spatial processing of the receiver on the received symbols to obtain detected symbols, performs the LLR computation independently for each of the best detected D symbols, and performs the LLR computation together for the remaining MD detected symbols, where M&gt; D&gt; 1. The best D symbols detected can be from the D data streams with the highest SNRs, the D data streams with the smallest SNR variation, the D data streams with the most robust coding, and so on. This detection scheme is referred to as a scaled-down detection scheme and can be used for the pass-through receiver

23/43 «, the only one shown in figure 3 and the iterative receiver shown in figure 4.

The receiver can perform spatial processing, from the receiver on the R symbols received to obtain D best detected symbols (instead of M detected symbols). A reduced spatial filter matrix M_D response dimension of best receiver RxD

D x R can be derived based on a channel matrix and includes complex symbols for the reduced H_D.

D detected columns.

D best of matrix H_D has dimension

H corresponding to the D spatial processing of the detected symbols is less in terms of computation.

The receiver can perform LLR computation independently for each detected, for example, illustrated in equation (12) receiver can perform with or one of the base in

D best max-log-MAP detector symbols some other type of detector.

joint LLR computing for the remaining detected MD symbols

In one mode, the remaining computations detected.

LLR by

For the best D for a symbol modulation that is among all the signal used extrinsic detected max-log-MAP in various receiver ways.

performs the joint for the consideration of the detected symbols

Better symbols give this modality, decisions detected symbols. A determined determined closest to hard symbols are made hard decisions_m in terms of modulation for the rest, illustrated

THE

Sm is a distance symbol in the receiver constellation then code bits for example, in equation (7) detector. For this LLR computation, of computes LLRs the symbols based on the detector or some other type of the best detected D symbols are restricted or fixed to hard decisions

24/43 determined for these detected symbols. In this way, the M number of hypotheses for evaluation is reduced²”'For MD ο ^Σ?^λ' ’²”, Considering that the best D symbols detected have indexes of m = M-D + l, ..., M.

In another mode, the receiver performs joint LLR computation for the remaining detected symbols without considering the best detected D symbols. For this modality, the receiver forms vectors reduced to - '-^TheW and — in addition to a matrix reduced to H. The reduced vectors and matrix include only elements corresponding to the remaining detected symbols, for example, based on the max-log-MAP detector illustrated in equation (7) or some other type of detector. The reduced vectors and matrix are used for LLR computing. Thus, the number of hypotheses for evaluation and computation for each hypothesis are both reduced.

In one embodiment, D is a fixed value that can be selected based on an exchange between complexity, performance, and / or other considerations. In another embodiment, D is a configurable value that can be selected based on the selected transmission mode, channel conditions (for example, SNRs for data flows) and / or other factors. D can be adaptive and can be selected by package, frame, subband or some other way.

reduced sizing detection scheme can provide good performance with less complexity than the exhaustive detection scheme illustrated in equations (6) and (7). The scaling detection scheme may be well suited for data transmission in which some data streams observe high SNRs and / or low frequency selectivity and the remaining data streams

25/43 observe lower SNRs and / or higher frequency selectivity. Frequency selectivity refers to the variation in channel gains across the frequency, which results in a variation of SNR across the frequency. The reduced scaling detection scheme takes advantage of high SNRs and / or low frequency selectivity for the best D data streams to reduce computing complexity. Nearly optimal detection is performed on the remaining data streams with lower SNRs and / or higher frequency selectivity to improve performance through a detection scheme that performs LLR computation by flow for all M data streams.

Figure 5 illustrates a process 500 for performing detection and decoding based on the reduced scaling detection scheme. The receiver's spatial processing performed on the received symbols for multiple data streams (for example, based on the zero imposition or MMSE technique) to obtain detected symbols (block 512). Detection is performed independently for each of the at least one data stream to obtain smooth decisions for the flow (block

514) .

Detection is performed together for the remaining data streams to obtain smooth decisions for those streams (block 516). Smooth decisions (for example,

Extrinsic LLRs) for each of the at least one data stream can be derived independently based on the detected symbols and a priori information (for example, a priori LLRs) from the decoder, if any, for that stream. The smooth decisions for the remaining data streams can be derived together based on the received symbols and the prior information, if any. Smooth decisions for multiple data streams can be decoded to

26/43 obtaining a priori information for the detector (block 518).

A determination is then made as to whether another iteration of detection and decoding is performed (block 520). The answer for block 520 is No after an iteration for a single-pass receiver and it is also No for an iterative receiver if a shutdown condition is found. If the answer is Yes for block 520, then the a priori information is used for detection in the next iteration (block 522), and the process returns to block 514. Otherwise, the detector's smooth decisions are further processed to obtain decoded data (block 524). Soft decisions can also be decoded before block 520, and the result of decoding can be used in block 520 to determine whether or not to perform another iteration.

A packet can be encoded and demultiplexed into multiple sub-packets that can be sent in multiple data streams. A CRC can be used for each subpackage or each data stream. In that case, the CRC for each detected data stream can be verified after block 514, and subsequent processing can be terminated if CRC fails for any of the data streams detected in block 514.

Figure 6 illustrates equipment 600 for performing detection and decoding based on the reduced scaling detection scheme. Equipment 600 includes means for performing spatial processing of the receiver on symbols received for multiple data streams to obtain detected symbols (block 612), means for performing detection independently for each of at least one data stream to obtain smooth decisions for the flow (block 614), means for the

27/43 conducting joint detection for the remaining data flows to obtain 'smooth decisions for those flows (block 616), means for decoding the smooth decisions for multiple data flows to obtain a priori information (block 618) , performing an additional iteration means for detecting decoding, if applicable, using prior information (block 620), and means for processing smooth decisions to obtain decoded data (block 622).

The number of hypotheses to be considered in joint LLR computing can be reduced by performing a search for candidate hypotheses using list sphere detection (LSD), which is also referred to as sphere decoding, spherical decoding, and so on. Sphere detection can be used to reduce the complexity for the log-MAP detector in equation (6), the max-log-MAP detector in the equation. (7), and other types of detector. List sphere detection seeks to reduce the detector's search space by eliminating the least likely assumptions based on a cost function. As an example, the log-MAP and max-log-MAP detectors can only consider 'assumptions that satisfy the following condition:

|| yHs ||²&lt;r² , Eq (13)

II y — Hs II² where - is the cost function er² it is a sphere or limit radius used to retain or eliminate the hypotheses of the consideration.

The cost function in equation (13) can be expanded, as follows:

28/43

J (s) = || yHs ||²= || H-s + nHs ||²= || H- (ss) + n ||² = + - [H-Cs-D + n] „Eq (14) = (ss)^ff-H^fl-H- (s-í) + C = (ss)^H R -R- (ss) + C where R is an upper triangular matrix obtained from the QR decomposition of H; and

C is a constant, which can be omitted since it is not a function of £.

QR decomposition can be performed on the effective MIMO channel response matrix H to obtain an orthonormal matrix Q and an upper triangular matrix R, or H = QR where Q^H-Q = I. The upper triangular matrix R contains zeros below the main diagonal. The fourth equality in equation (14) can be obtained based on the following: H^H - H = j Η ^—· _RH

For the cost function in equation (14), the detected symbols can be used as the central point for the search. The cost function can then be expressed as:

/ (ã) = || R- (8-8) 11² . Eq (15) where s can be obtained from equation (11) based on any technique of spatial receiver processing (for example, force zero or MMSE). The cost function in equation (15) can be computed for 2^K hypothetical data vectors? that can be passed to vectors s to get 2^K cost values ​​for these 2^K hypotheses. The hypothetical data vectors with cost values ​​that are less than or equal to ar² can be considered for LLR computation, for example, in equation (6) or (7).

The structure of the upper triangular matrix R can be explored to reduce the number of hypotheses for the

29/43 which to compute the cost function. Equation (15) can be expanded as follows:

<td>5, i 5.2 '' '5jn</td><td></td><td>(Â - Ã) '</td>

<td>0 5.2 · 5, m</td><td></td><td>(S₂ - s₂)</td>

<td>^{0 0} - · 5th.</td><td></td><td> .(¾ ~^?m).</td>

Eq (16)

For M = 4, equation (16) can be expressed as:

/ = η + τ₂+ τ₃+ τ₄,

Eq (17) where ^T\ = 15, i · (Â - $ 1) + 5.2 · (¾ - s₂) +5.3 · (¾ “Sj) +5.4 · (« 4 - s₄) I² .

^Tz = I ^r2.2 · (Â ~ Ã) + r₂'₃ · (¾ - s₃) + r_2A (THE - ?₄) |² , ^3 ^— 15.3 '(^ 3 5.4' (Â “^ 4) I» q

T₄= | r_4j4-(At₄)|² .

Equation (17) can be generalized to any value of M, as follows:

where j =&gt;

í = M

<img file="BRPI0618716A2_D0005.tif" />

Eq (18)

Eq (19)

As illustrated in equations (18) and (19), the cost function can be decomposed into a sum of M terms Ti to T_M. The term T_M only depends on the detected symbol for flow Μ, the term T_M-i depends on detected symbols ⁵μ-ι e for flows Ml and M, and so on, and the term ΊΤ depends on the detected symbols ^swent to all M flows.

The cost function can be computed incrementally over M levels, a term Ti for each level, starting with the last term T_M to the first level. It is indicated by the inverse sum for J in equation (18), which starts with i = M. For each level, Ti is computed for

30/43 all the hypotheses applicable for that level, and the cost function is updated.

The cost function can be computed by considering an additional symbol at a time, starting with and ending with. To improve search performance, streams can be stored in a way that^{5 m} correspond to the best flow (for example, with the highest SNR) and correspond to the worst flow (for example, with the lowest SNR). The received vector y and the H channel response matrix can be reordered to achieve the desired ordering for the detected vector

The LSD search can be performed in several ways. In a first modality, all hypotheses with cost values ​​equal to or less than air² are retained, and all other hypotheses are discarded. In a second modality, only the best B hypotheses are retained at each level, and all other hypotheses are eliminated. For both modalities, the total number of hypotheses to be considered is reduced by eliminating hypotheses with high cost values.

The LSD search can be performed as follows. For the first level, a list P_M is formed with 2 hypothetical data symbols ^IMM that may have been transmitted to the data symbol s_M. T_M is computed for the 2⁰”¹ data symbols hypothesized in list P as illustrated in equation (19) to obtain the 2nd cost values. For the first modality, all hypothesized data symbols with cost values ​​less than or equal to ar² are retained and stored in a candidate list C_M. For the second modality, up to B data symbols hypothesized with the lowest values ​​of

31/43 costs that are less than or equal to air² are retained and stored in candidate list C_M. For the second modality, the hypothetical data symbols with cost values ​​less than or equal to ar² can be eliminated if these cost values ​​are not among the lowest B cost values. For both modalities, all other hypothesized data symbols are eliminated, which has the effect of pruning all hypothesized data vectors - containing the symbols deleted.

For the second level, a list P_M_i is formed with pQM-i hypothesized data symbols ^514-1 that may have been transmitted to the data symbol SM_i. TM-i is computed for all valid hypothesized symbol pairs (Sm-u ^ m) and added with TM to obtain the cost values ​​for these hypothesized symbol pairs. Valid hypothesized symbol pairs include all possible combinations of each symbol in the CM candidate list with each symbol in the PM-i list · The updated cost values ​​can be composed as: JM-i = ΤΜ-χ + TM. For the first modality, all symbol pairs hypothesized with cost values ​​less than or equal to ar² are retained and stored in a CM-i candidate list. For the second modality, up to B hypothesized symbol pairs with lower cost values ​​that are less than or equal to ar² are retained and stored in candidate list C_M-i. For both modalities, all other pairs of hypothesized symbols are eliminated.

Each of the remaining levels can be assessed in a similar way. A Pi list is formed with 2^Q1 hypothetical data symbols ^Si that may have been transmitted to the Si data symbol. Ti is computed for all

32/43 valid hypothesized symbol sets plus Ti + ia T_M to obtain cost values ​​for these hypothesized symbol sets. Valid hypothesized symbol sets include all possible combinations of each hypothesis in the candidate list Ci + i with each symbol in the list Pj. The updated cost values ​​can be computed as: Ji = Tt +. . . + T_M. For the first modality, all hypothesized symbol sets with cost values ​​less than or equal to ar² are retained and stored in a candidate list Ct. For the second modality, up to B hypothetical symbol sets with the lowest cost values ​​that are less than or equal to air² they are retained and stored in the candidate list Ci. For both modalities, all other hypothesized symbol sets are eliminated.

After all M levels have been assessed, extrinsic LLRs can be computed for all hypotheses stored in candidate list C_íf for example, based on the max-log-MAP detector, as if

5: c;

-Λ-lly-a-sF.

Z SX * u

L n J follows:

Eq (20) where c⁺_k is a subset of candidate list Ci and contains hypotheses for which b_k= +1;

a subset of candidate list Ci and contains hypotheses for which bk = -l.

Extrinsic LLRs can also be computed based on the log-MAP detector or some other type of detector.

The cost values ​​already computed for the candidate hypotheses can be used to compute the

Extrinsic LLRs for these hypotheses. For clarity, the description above uses different lists

33/43 candidates for different levels. A united candidate list C can be used for all M levels and can be updated at each level.

Figure 7 illustrates an illustrative search tree for the second modality, which retains the best B hypotheses at each level. For this example, M = 4, four terms T_x to T₄ are computed, B = 2 and candidate list C contains up to the two best hypotheses for 2^Q4possible data symbols that may have been passed to the data symbol s₄, which are denoted as the

<td>at</td><td>figure</td><td> 7.</td><td>Both</td><td>best</td><td>hypotheses are</td>

<td>illustrated</td><td>with us</td><td colspan="2">filled</td><td>in black.</td><td>For 0 seconds</td>

<td>level with</td><td>i = 3, T₃</td><td>is</td><td>computed</td><td>for B - 2^Q3</td><td>hypotheses for</td>

B · 2^q3 possible symbol pairs that may have been passed to the data symbols₃ es₄. The two best hypotheses are illustrated with nodes filled with black. For the third level with i = 2, T₂ is computed for B • 2^q2 hypotheses for B · 2^Q2 possible symbol sets that may have been passed to the data symbols₂, s₃ es₄. The two best hypotheses are illustrated with nodes filled with black. For the last level with i = l, Τχ is computed for B · 2^Q1 hypotheses for B · 2^Q1 possible symbol sets that may have been transmitted for data symbols Si, s₂, s₃ es₄. The two best hypotheses are illustrated with nodes filled with black. The set of symbols that results in the lowest cost value is illustrated by heavy lines.

In one embodiment, the number of candidate hypotheses to store in list C is a fixed value that 30 can be selected based on an exchange between detection performance, complexity and / or other considerations. This fixed value (B) can be for each

34/43 level, as described above. This fixed value can also be for each node at a given level, in which case B^M_1-2^Q1assumptions are considered by level. In another modality, the number of candidate hypotheses is configurable and can be selected based on the size of the signal constellation, number of iterations, detection performance, complexity, and / or other considerations. For example, B can be a function of the size of the signal constellation so that more candidate hypotheses are stored for larger signal constellations to ensure sufficient accuracy in LLR computation. B can also be restricted so that it is equal to or greater than some minimum value (for example, Bmin = 2), which ensures that at least B_min candidate hypotheses are stored at each level.

Two modalities for cost computation for list sphere detection have been described above. Cost computation can also be performed in other ways.

List sphere detection is a search technique for reducing the number of hypotheses to be considered for LLR computing. Other techniques can also be used to reduce the number of hypotheses for LLR computing. In another modality, an MCMC technique is used to generate a list of candidate hypotheses. The MCMC technique considers the M elements in the data vector s in a sequential way, evaluates a hypothesis for each element, and recycles through these M elements for multiple iterations. List and MCMC detection techniques are known in the art and described in several articles.

In another aspect, the receiver performs LLR computation independently for each of the best detected D symbols, and reduces the number of hypotheses to be

35/43 considered for joint LLR computing by performing a search using LSD, MCMC or some other search technique. This detection scheme is referred to as a reduced order detection scheme and can be used for the single-pass receiver shown in figure 3 and the iterative receiver shown in figure 4. For the reduced order detection scheme, scaling is reduced from M to MD by performing LLR computation by flow for the D best flows, and the number of hypotheses to be considered for the remaining MD flows can be reduced by MD from a limit higher than ²'“' By conducting a search. D can be selected adaptively based on channel conditions (for example, SNRs) and / or some other factors. The SNR information can be readily obtained from the spatial processing of the receiver that provides the detected symbols used for the central point of an LSD search.

Figure 8 illustrates a process 800 for performing detection and decoding based on the reduced order detection scheme. The receiver's spatial processing is performed on a set of received symbols (for example, vector γ) to obtain a set of detected symbols (for example, vector I) for a set of transmitted data symbols (for example, vector s) through a MIMO channel (block 812). Extrinsic LLRs are computed independently for each of the at least one data symbols based on a corresponding detected symbol and

A priori LLRs, if any, of the decoder (block 814). A list of candidate hypotheses determined for the remaining data symbols, for example, by conducting a search using

LSD,

MCMC or some other technique (block

816) .

36/43

Each candidate hypothesis corresponds to a different combination of symbols hypothesized as having been transmitted from the remaining data symbols. Extrinsic LLRs are then computed together for the remaining data symbols with the candidate hypothesis list (block 818). Extrinsic LLRs for all data symbols are decoded to obtain a priori LLRs for the detector (block 820).

A determination is then made as to whether or not to perform another detection and decoding iteration (block 822). If the answer is yes then LLRs a priori are computed to be used for computing LLR on the detector in the next iteration (block 824), and the process returns to block 814. Otherwise, extrinsic LLRs from the detector are processed to obtain decoded data (block 826).

Figure 9 illustrates an equipment 900 for performing detection and decoding based on the reduced order detection scheme. Equipment 900 'includes means for performing spatial processing of the receiver on a set of received symbols to obtain a set of detected symbols for a set of data symbols transmitted over a MIMO channel (block 912), means for computing LLRs extrinsic independently for each of the at least one data symbol based on a corresponding detected symbol and LLRs a priori, if any, from the decoder (block 914), means for determining a candidate hypothesis list for the remaining data symbols by performing a search (block 916), means for computing extrinsic LLRs together with the remaining data symbols with the candidate hypothesis list (block 918), means for decoding extrinsic LLRs for all symbols

<img file="BRPI0618716A2_D0006.tif" />

♦ · i

of data for obtaining LLRs a priori (block 920), means for performing an additional iteration of LLR computation and decoding, if applicable, using LLRs a priori (block 922), and means for processing extrinsic LLRs from detector for obtaining decoded data (block 924).

For the iterative receiver, the search for candidate hypotheses can be performed in several ways. In an environment, the search is performed for only the first iteration to obtain a list of candidate hypotheses, and this candidate list is used for all subsequent iterations. For this modality, extrinsic LLRs for each subsequent iteration are computed with LLRs a priori from the decoder and for the candidate hypotheses determined in the first iteration. In another mode, the search is performed for each iteration. In yet another modality, the search is performed for each iteration until a closing condition is found. This closing condition can be triggered, for example, after a predetermined number of iterations has been completed, if the search provides the same list of candidate hypotheses for two consecutive iterations, and so on.

For the modalities in which the search is performed for multiple iterations, limit (for example, sphere radius) can be a fixed value for all iterations or it can be a configurable value that can be determined for each iteration based, for example, in the decoder's prior information.

For the reduced order detection scheme, the reduction in complexity with respect to the ideal log-MAP or max-log-MAP detector depends on the complexity of the search. For an LSD search, the complexity is determined by the number of nodes visited in the search tree, which, in turn, depends

38/43 of several factors such as number of hypotheses to be stored at each level (for example, the value of B), the radius of sphere r², the specific H channel response matrix, the SNR, and so on.

In another low complexity detection scheme, the receiver performs (1) LLR computation per flow for all M flows for the first iteration and (2) joint LLR computation for the worst MDs or all M flows for an iteration subsequent, if necessary, for example, if the package is decoded in error.

For the iterative receiver, channel estimates can be updated during the iterative detection and decoding process to obtain improved channel estimates. For example, if pilot symbols are received and monitored several times during the iterative process, then the effective SNR of channel estimates can be improved over time by performing media processing and / or other processing. The improved channel estimates can be used for spatial processing of the receiver, for example, as illustrated in equations (8) to (10), for LLR computation, for example, as illustrated in equations (6), (7) and (12), for cost computation for a search for candidate hypotheses, for example, as illustrated in equations (18) and (19).

Several low complexity detection schemes have been described above. These schemes reduce the number of hypotheses to be considered for LLR computing while achieving good error rate performance that is close to the performance of the ideal detector.

The decoder 320 in figure 3 and the decoder 420 in figure 4 can be implemented in several ways depending on the encoding schemes used in the transmitter 10. The decoder 420 receives records of

39/43 smooth decision and generates smooth decision outputs for the MIMO 410 detector and can be a smooth output Viterbi algorithm decoder (SOVA) if a convolution code is used on transmitter 110, a Turbo decoder if a serial concatenated code tube or parallel is used, and so on. The decoder 320 can be a Viterbi decoder or a SOVA decoder for a convolution code, a Turbo decoder for a Turbo code, and so on. A Turbo decoder can be a MAP decoder that can implement a BCJR smooth input and smooth output MAP algorithm and a derivative of lower complexity. These various types of decoding are known in the art and described in various literature. For example, the SOVA decoder is described by J. Hagenauer et al., And an article entitled A Viterbi Algorithm with Soft-Decision Outputs and its Applications, IEEE Globecom, 1989, PP. 47.1.1-47.1.7. The MAP decoder is described by LR Bahl et al. in an article entitled Optimal Decoding of Linear Codes for Minimizing Symbol Error Rate, IEEE Transaction on Information Theory, March 1974, vol. 20, PP 284-287.

The detection and decoding techniques described here can be used for various MIMO transmission schemes, which can also be referred to as transmission modes, spatial modes, and so on. Table 1 lists some illustrative modes of transmission and their short descriptions. Targeted mode can also be called a beamforming mode, a self-targeting mode, a beamformed MIMO mode, and so on. 0 undirected mode can also be called basic MIMO mode. The spatial scattering mode can also be called the targeting mode

40/43 pseudo-random transmission, spatial expansion mode, and so on.

Table 1

<td>Way of Streaming</td><td>description</td>

<td>Directed</td><td>Multiple data streams are transmitted on multiple orthogonal spatial channels (or automodes) of a MIMO channel.</td>

<td>Not Directed</td><td>Multiple data streams are transmitted over multiple spatial channels (for example, from multiple transmitting antennas)</td>

<td>Scattering Space</td><td>Multiple data streams are spatially spread across spatial channels to achieve similar performance for all data streams.</td>

Each transmission mode has different capacities and requirements. Targeted mode typically provides better performance, but requires spatial processing on transmitter 110 and receiver 150. Undirected mode does not require any spatial processing on transmitter 110. For example, transmitter 110 10 can transmit a data stream from each transmitting antenna. 0 Spatial spreading mode transmits M data streams with different spatial mapping matrices so that these data streams observe a group of effective channels and achieve similar performance. A suitable transmission mode can be selected based on the availability of channel status information, transmitter and receiver capacities, and so on.

For the directed mode, data is transmitted in up to M MIMO channel automodes, which can be obtained by diagonalizing the response matrix of

41/43

<td>MIMO channel</td><td>H through</td><td>an</td><td>decomposition</td><td colspan="2">of singular value</td>

<td>of H or</td><td>decomposition</td><td>in</td><td colspan="2">eigenvalue of a matrix</td><td>in</td>

<td>correlation</td><td>of H, which</td><td>is</td><td>A = H * H. _THE</td><td>decomposition</td><td>in</td>

<td>eigenvalue i</td><td>of A can be</td><td colspan="2">expressed as:</td><td></td><td></td>

<td>A = H</td><td>• H = EAE^W ,</td><td></td><td></td><td>Eq (21)</td><td></td>

where E is a TxT unit array of eigenvectors from A; and

Λ is a diagonal TxT matrix of eigenvalues ​​of A.

Transmitter 110 can perform spatial processing with E-vectors to transmit data in

M odos automodes. Diagonal matrix A contains real non-zero values ​​possible along the diagonal and zeros in all other locations. These diagonal records are referred to as eigenvalues ​​of A and represent the power gains for the M automodes.

Table 2 illustrates the spatial processing of the transmitter for the three transmission modes and the effective MIMO channel response matrix for each transmission mode. In Table 2, the subscript denotes the directed (or self-directed) mode, us denotes the non-directed mode, and SS denotes the spatial spreading mode.

Table 2

<td></td><td>Mode Directed</td><td>No mode Directed</td><td>Scattering Space</td>

<td>Transmitter</td><td>x_es= Es</td><td>Xus S</td><td>Xss = VS</td>

<td>Effective channel</td><td>Hes⁼Hch'E</td><td>Hus,^—H_çH</td><td>H_ss= H_ÇhV</td>

For a MIMO system with multiple sub-bands, the spatial processing of the transmitter illustrated in Table 2

<td>can</td><td>be carried out</td><td>for</td><td>each</td><td>subband 1.</td><td>At</td><td>equation (1),</td>

<td>P (l)</td><td>= E (l) for the</td><td>mode</td><td colspan="2">directed, P (l)</td><td>= I</td><td>for mode</td>

<td>not</td><td>directed and</td><td>P (l) =</td><td>= V (1)</td><td>for mode</td><td>in</td><td>scattering</td>

42/43 space. V (l) is a TxT spatial mapping matrix used for spatial spreading and can be generated based on a Hadamard matrix, a Fourier matrix, and so on.

For a MIMO system with multiple sub-bands, the

M eigenvalues ​​for each subband 1 can be ordered from largest to smallest, so that the H (l) automodes are classified from the highest SNR to the lowest SNR. The broadband motor m can be formed with a motor for each L sub-bands. The main broadband auto (with m = 1) is associated with the largest eigenvalues ​​for all L sub-bands, the second broadband auto (with m = 2) is associated with the second largest eigenvalue for all L sub-bands , and so on. M data streams can be sent on M broadband devices.

The main broadband automobile has the highest average SNR and typically also has the least variation in SNR across time and frequency. Conversely, the weakest broadband auto has the lowest average SNR and typically has the most SNR variation.

The techniques described here can be used for M-band data transmission modes, the receiver can flow to one or more of the forts and the remaining broadband LLR computation. The GA. In one, the LLR computation by broadband automobiles more joint to the number of hypotheses to be considered for the joint LLR computation can be reduced by performing a search using LSD, MCMC, or some other type of search technique.

The detection and decoding techniques described here can be implemented by various means. For example, these techniques can be implemented in hardware, firmware, software or a combination of them. For

43/43 hardware implementation, the processing units used to perform detection and decoding can be implemented within one or more of the application-specific integrated circuits (ASICs), digital signal processors (DSPs), signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable port sets (FPGAs), processors, controllers, microcontrollers, microprocessors, electronic devices, other electronic units designed to perform the functions described here, or a combination of them.

For a firmware and / or software implementation, the techniques can be implemented with modules (for example, procedures, functions and so on) that perform the functions described here. Software codes can be stored in memory (for example, memory 192 in figure 1) and executed by a processor (for example, processor 190). The memory can be implemented inside the processor or outside the processor.

The foregoing description of the described modalities is provided to allow anyone skilled in the art to create or make use of the present invention. Various modifications to these modalities will be readily apparent to those skilled in the art, and the generic principles defined here can be applied to other modalities without departing from the spirit or scope of the invention. Accordingly, the present invention should not be limited to the modalities illustrated here, but the broader scope consistent with the principles and novelty features described here must be agreed.