WO2002049219A2 - Systeme et procede de decodage de codes en treillis - Google Patents

Systeme et procede de decodage de codes en treillis Download PDF

Info

Publication number
WO2002049219A2
WO2002049219A2 PCT/US2001/048238 US0148238W WO0249219A2 WO 2002049219 A2 WO2002049219 A2 WO 2002049219A2 US 0148238 W US0148238 W US 0148238W WO 0249219 A2 WO0249219 A2 WO 0249219A2
Authority
WO
WIPO (PCT)
Prior art keywords
group
state
states
branches
value
Prior art date
Application number
PCT/US2001/048238
Other languages
English (en)
Other versions
WO2002049219A3 (fr
Inventor
Eran Arad
Efraim Dalumi
Shachar Kons
Donald B. Eidson
Original Assignee
Conexant Systems, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US10/013,492 external-priority patent/US6865711B2/en
Application filed by Conexant Systems, Inc. filed Critical Conexant Systems, Inc.
Priority to AU2002232576A priority Critical patent/AU2002232576A1/en
Publication of WO2002049219A2 publication Critical patent/WO2002049219A2/fr
Publication of WO2002049219A3 publication Critical patent/WO2002049219A3/fr

Links

Classifications

    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/37—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/39—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
    • H03M13/3905—Maximum a posteriori probability [MAP] decoding or approximations thereof based on trellis or lattice decoding, e.g. forward-backward algorithm, log-MAP decoding, max-log-MAP decoding
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/25—Error detection or forward error correction by signal space coding, i.e. adding redundancy in the signal constellation, e.g. Trellis Coded Modulation [TCM]
    • H03M13/256—Error detection or forward error correction by signal space coding, i.e. adding redundancy in the signal constellation, e.g. Trellis Coded Modulation [TCM] with trellis coding, e.g. with convolutional codes and TCM
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/25—Error detection or forward error correction by signal space coding, i.e. adding redundancy in the signal constellation, e.g. Trellis Coded Modulation [TCM]
    • H03M13/258—Error detection or forward error correction by signal space coding, i.e. adding redundancy in the signal constellation, e.g. Trellis Coded Modulation [TCM] with turbo codes, e.g. Turbo Trellis Coded Modulation [TTCM]
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/29—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes combining two or more codes or code structures, e.g. product codes, generalised product codes, concatenated codes, inner and outer codes
    • H03M13/2957—Turbo codes and decoding
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/37—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/39—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
    • H03M13/3961—Arrangements of methods for branch or transition metric calculation
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/37—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/39—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
    • H03M13/3972—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes using sliding window techniques or parallel windows
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/37—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/39—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
    • H03M13/3988—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes for rate k/n convolutional codes, with k>1, obtained by convolutional encoders with k inputs and n outputs
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/37—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/39—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
    • H03M13/41—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes using the Viterbi algorithm or Viterbi processors
    • H03M13/4161—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes using the Viterbi algorithm or Viterbi processors implementing path management
    • H03M13/4169—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes using the Viterbi algorithm or Viterbi processors implementing path management using traceback
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/65—Purpose and implementation aspects
    • H03M13/6502—Reduction of hardware complexity or efficient processing
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/65—Purpose and implementation aspects
    • H03M13/6577—Representation or format of variables, register sizes or word-lengths and quantization
    • H03M13/6583—Normalization other than scaling, e.g. by subtraction
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
    • H04L1/0045—Arrangements at the receiver end
    • H04L1/0047—Decoding adapted to other signal detection operation
    • H04L1/005—Iterative decoding, including iteration between signal detection and decoding operation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
    • H04L1/0056—Systems characterized by the type of code used
    • H04L1/0059—Convolutional codes
    • H04L1/006—Trellis-coded modulation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
    • H04L1/0056—Systems characterized by the type of code used
    • H04L1/0064—Concatenated codes
    • H04L1/0065—Serial concatenated codes

Definitions

  • This invention generally relates to decoders, and, more specifically, to decoders of trellis codes.
  • Trellis codes such as convolutional codes, or parallel or series combinations or concatenations of convolutional codes, are codes which are decoded through use of a trellis.
  • Trellis coded modulation (TCM) codes are groupings of trellis coded bits which result from mapping the bits into symbols, such as MPSK symbols, prior to transmission. The symbols may then be used to modulate a carrier, and the modulated carrier transmitted over a wireline or wireless interface.
  • TCM codes serial concatenated trellis coded modulation codes (SCTCM) codes
  • Patent Number 6,023,783 "Turbo Codes: Analysis, Design, Iterative Decoding and Applications," Course 909, Part II, International Courses for Telecom and Semiconductor Professionals, S. Benedetto & D. Divsalar, October 25-29, 1999, Barcelona, Spain, pp. 324-339; "A Serial Concatenation Approach to Iterative Demodulation and Decoding," K. Narayanan et al., IEEE Transactions on Communications, Vol. 47, No. 7, July 1999; "Turbo DPSK': Iterative Differential PSK Demodulation and Channel Decoding," P. Hoeher et al., IEEE Transactions on Communications, Vol. 47, No.
  • Decoders of trellis codes may be configured for computing soft or hard estimates of the underlying source bits or the encoded symbols, or for computing extrinsic outputs for the underlying source bits or encoded symbols, i.e., soft or hard estimates with a priori information about the bits or symbols removed.
  • decoders include, for example, maximum a posteriori (MAP) decoders, log-MAP decoders, MAX-Log-MAP decoders, Viterbi decoders, Soft Output Viterbi (SONA) decoders, A Posteriori Probability (APP) decoders, Soft List Viterbi (Soft-LNA) decoders, etc.
  • Known decoders of TCM codes are configured to handle QPSK symbols, but are susceptible to performance limitations in applications involving MPSK or QAM symbols beyond QPSK.
  • the problem is particularly acute in applications involving log- MAP decoders, in which probabilities are expressed in the natural log domain. The computations needed to perform log-domain calculations places additional demands on the decoders.
  • the invention provides ba system for determining one or more state probabilities for one or more states in a trellis representation.
  • branch metric logic determines branch metrics for one or more of the branches between one or more states in a first portion of the trellis and one or more states in a second portion of the trellis
  • state probability logic determines state probabilities for one or more of the states.
  • the state probability logic in this embodiment groups one or more of the branches into groups. This state probability logic also performs, for one or more of the groups, a first group operation on one or more of the branch metrics for the branches in the group to determine a group value for the group; and then updates, for one or more of the groups, the group value for the group based on one or more previously determined state probabilities.
  • This first group Operation may be but is not limited to the MAX* operation.
  • This state probability logic also performs, for one or more of the states, a second group operation on one or more of the updated group values. This second group operation may be but is not limited to the MAX* operation.
  • the state probability logic comprises p state probability logic modules for determining in parallel state probabilities for each of p states, where p is an integer of two or more.
  • each of the p modules groups one or more of the branches into groups.
  • Each of the p modules performs, for one or more of the groups, a first group operation on one or more of the branch metrics for branches in the group to determine a group value for the group; and then updates, for one or more of the groups, the group value for the group based on one or more previously determined state probabilities.
  • Each of the p modules also performs a second group operation on one or more of the updated group values to determine a state probability for the state.
  • Normalization logic may also be included in this second embodiment for normalizing one or more of the p state probabilities as determined in parallel by the p modules. In one implementation, this logic functions by determining the maximum state probability of one or more of the p states, and subtracting the same from one or more of the p state probabilities.
  • the invention also provides a system for determining branch metrics.
  • First logic determines a first correlation value representing the level of correlation between one or more symbols and one or more symbols associated with a branch.
  • Second logic determines a second correlation value representing the level of correlation between one or more a priori probabilities and one or more bits associated with the branch.
  • Third logic derives a branch metric for the branch from the first and second correlation values.
  • the invention also provides a system for determining an estimate or extrinsic output of one or more bits using a trellis representation.
  • this trellis representation there are one or more branches between one or more states in a first portion of the trellis and one or more states in a second portion of the trellis.
  • One or more of the states in the first portion of the trellis have forward state probabilities, and one or more of the states in the second portion of the trellis have backward state probabilities.
  • one or more of the branches have branch metrics.
  • first logic in the system groups one or more of the branches into groups.
  • Second logic in this embodiment performs, for one or more of the groups, a group operation on one or more of the branch metrics for the branches in the group to determine a group value for the group.
  • This group operation may be but is not limited to the MAX* operation.
  • Third logic in the system updates, for one or more of the groups, the group value for the group based on one or more state probabilities and, possibly, one or more a priori probabilities for the input bits.
  • Fourth logic in the system derives a first value from one or more of the updated group values for groups which imply release of a logical "1" for the bit in question, derives a second value from one or more of the updated group values for groups which imply release of a logical "0" for the bit in question, and determines an estimate of or extrinsic output for the bit in question from the first and second values.
  • the invention further provides a system for computing the MAX* of operands A and B.
  • first logic in the system tests the difference A - B relative to zero, and outputs a signal indicative thereof.
  • Second logic in the system determines the maximum of the operands A and B,
  • MAX(A,B) by outputting a signal representative of the operand A if the signal from the first logic indicates that the difference A - B is greater than zero, and outputs a signal representative of the operand B otherwise.
  • Third logic in the system determines the absolute value of A - B by outputting a signal representative of the difference A - B if the signal from the first logic indicates that the difference A - B is greater than zero, and outputs a signal representative of the difference B - A otherwise.
  • Fourth logic outputs a value corresponding to ln(l + exp(- - R
  • Fifth logic determines a value corresponding to iA"( ⁇ (,R)+ ln(l + exp(-
  • the invention also provides a system for performing a MAX* 2 P ->1 operation, ' where p is an integer of two or more.
  • the system comprises a hierarchical arrangement of MAX* 2->l logic modules having p levels, ranging from
  • a level q, 1 ⁇ q ⁇ p, of the hierarchy comprises 2 (p"q) MAX* 2->l logic modules configured in parallel.
  • the invention further provides a system for computing one or more forward state probabilities or one or more backward state probabilities in a trellis representation.
  • This trellis representation has one or more branches between one or more states in a first portion of the trellis and one or more branches in a second portion of the trellis.
  • branch metric logic in the system computes one or more branch metrics for one or more of the branches, and indication logic in the system indicates whether the system is configured to compute forward state probabilities or backward state probabilities.
  • State probability logic in the system (1) computes one or more forward state probabilities for one or more states in the second portion of the trellis, provided the indication logic indicates the system is configured to compute forward state probabilities; and (2) computes one or more backward state probabilities of one or more states in the first portion of the trellis, provided the indication logic indicates the system is configured to compute backward state probabilities.
  • the system may further include estimation extrinsic output logic for computing an estimate of or extrinsic output for one or more bit responsive to one or more state probabilities, one or more branch metrics, and, possibly, one or more a priori probabilities for the bits.
  • the backward state probability for a state may be computed by grouping the branches exiting the state according to the state at which the branches terminate, determining for each group exiting the state a group value based on the branch metrics for the branches in the group, and then determining the backward state probability for the state responsive to the group values for one or more of the groups exiting the state and one or more of the backward state probabilities of the successor states of the groups.
  • the forward state probability for a state in this implementation may be computed by grouping the branches entering the state according to the state at which the branches originate, determining for each group entering the state a group value based on the branch metrics for the branches in the group, and then determining the forward state probability for the state responsive to the values of one or more of the groups entering the state and one or more of the forward state probabilities of the predecessor states of the groups.
  • the estimates or extrinsic outputs in this implementation may be computed one bit at a time.
  • the branches for the corresponding portion of the trellis may first be grouped according to the origination and termination states, and whether the branch implies release of a . logical "1" or a logical "0". Then, a group value may be determined for each group based on the branch metrics for the branches in the group. An estimate of or extrinsic output of the bit may then be determined responsive to the group values, the forward state probabilities of the predecessor states of the groups, the backward state probabilities of the successor states of the groups, and possibly the a priori probabilities of the bits. This process may then be repeated for each of the bits.
  • the system is part of a log-MAP decoder, and there are only two states in the trellis at a time.
  • the branches exiting a particular state can be divided into two groups: those which terminate at the first state in a successive time period, and those which terminate at the second state in a successive time period.
  • the backward state probability for a particular state may be computed by first determining for each group exiting the state the MAX* of the branch metrics for all the branches in the group.
  • the MAX* value for the first group is added to the backward state probability of the state at which the branches in the first group terminate
  • the MAX* value for the second group is added to the backward state probability of the state at which the branches in the second group terminate.
  • the backward state probability for the state in question may then be derived from the MAX* of these two values.
  • the branches entering a particular state can also be divided into two groups: those which originate at the first state in a previous time period, and those which originate at the second state in a previous time period.
  • the forward state probability for a particular state may be computed by first determining for each group entering the state a MAX* value of the branch metrics for all the branches in the group. Then, the MAX* value for the first group may be added to the forward state probability of the state at which the branches in the group originate, and the MAX* value for the second group may be added to the forward state probability of the state at which the branches in the group originate. The forward state probability for the state in question may then derived from the MAX* of these two values.
  • the extrinsic outputs may be computed one bit at time.
  • the branches in the section of the trellis may be divided into groups according to their origination and destination states, and whether the branch implies release of a logical " 1 " or "0". A maximum of eight groupings are possible.
  • the MAX* of each grouping is then computed and added to the forward state probability for the originating state for the group, and added to the backward state probability for the terminating state for the group. After subtracting out the a priori probability for the bit in question, the result is a group value assigned to the group.
  • a first value may then be derived from the MAX* of the group values for the four groups which imply a release of logical "1”, and a second value may be derived from the MAX* of the group values for the four groups which imply a release of logical "0".
  • the second value may be subtracted from the first to form an extrinsic output for the bit in question. This process may then repeated for each of the k input bits.
  • FIG. 1 is a block diagram of one example of a decoder of serial concatenated TCM codes.
  • FIG. 2 is a block diagram of the inner decoder module of the decoder of FIG. 1.
  • FIG. 3 is a block diagram of an encoder which corresponds to the decoder of FIG. 1.
  • FIG. 4A is a block diagram of one example of the inner encoder module of the encoder of FIG. 3.
  • FIG. 4B is a block diagram of one example implementation of the inner encoder portion of the module of FIG.4A.
  • FIG. 5 are constellations illustrating one example of a four dimensional bit to
  • FIG. 6 is a timing diagram illustrating concurrent operation of forward and backward engines according to one example application of a decoding engine according to the invention.
  • FIG. 7A is a block diagram of a first embodiment of a system for decoding
  • FIG. 7B is a block diagram of a second embodiment of a system for decoding TCM symbols according to the invention.
  • FIG. 8 A is a portion of trellis illustrating the computation of backward state probabilities according to embodiment of the invention.
  • FIG.8B is a portion of a trellis illustrating the computation of forward state probabilities according to one embodiment of the invention.
  • FIG. 8C is a portion of a trellis illustrating the computation of extrinsic outputs according to one embodiment of the invention.
  • FIG. 8D is a portion of a trellis illustrating the computation of forward and backward state probabilities according to a second embodiment of the invention.
  • FIG. 8E is a portion of a trellis illustrating the computation of extrinsic outputs according to a second embodiment of the invention.
  • FIG. 9 is a block diagram illustrating one implementation of branch metric logic and state probability logic according to the invention.
  • FIG. 10 is a block diagram illustrating one implementation of branch metric logic and extrinsic output logic according to the invention.
  • FIG. 11A is a block diagram illustrating one implementation of symbol correlation logic according to the invention.
  • Figure 11B illustrates an 8-PSK symbol constellation on the I-Q plane rotated counter-clockwise by ⁇ /8 radians (22.5°).
  • FIG. 12 is a block diagram illustrating one implementation of bit correlation logic according to the invention.
  • FIG. 13 is a block diagram illustrating one implementation of 16 to 1 MAX* computation logic according to the invention.
  • FIG. 14 is a block diagram illustrating one implementation of 8 to 1 MAX* computation logic according to the invention.
  • FIG. 15 is a block diagram illustrating one implementation of 4 to 1 MAX* computation logic according to the invention.
  • FIG. 16A is a block diagram illustrating one implementation of 2 to 1 MAX* computation logic according to the invention.
  • FIG. 16B illustrates a lookup table employed in the implementation of the MAX* 2->l computation logic illustrated in FIG. 16 A.
  • FIG. 17 is a block diagram illustrating one implementation of normalization logic according to the invention.
  • FIG. 18 is a flowchart of one embodiment of a method of determining state probabilities according to the invention.
  • FIG. 19 is a flowchart of one embodiment of a method of determining extrinsic outputs according to the invention.
  • Example Application An example application of a system for decoding TCM symbols according to the invention will first be described followed by a discussion of the system itself. This example application is included in order to provide context and aid in understanding the invention. However, many other applications of the invention are possible. Therefore, the inclusion of this application should not be construed as limiting.
  • SCTCM serially concatenated trellis coded modulation
  • m encoded symbols are input to inner decoder module 102 over one or more signal lines 100, where m is an integer of one or more.
  • the symbols may be represented in quadrature form, i.e., in the form of two dimensional in-phase (I) and quadrature (Q) components, although other forms of representation are possible.
  • the quadrature form of representation is convenient since the I and Q components of a symbol may be simultaneously represented as a single point or number on a complex I-Q plane. Therefore, in the ensuing discussion, this form of representation may be assumed unless otherwise noted.
  • the encoder comprises the series combination of outer encoder module. 302, parallel-to-serial converter 306, bit interleaver 308, serial-to-. parallel converter 310, inner encoder module 314, bit to symbol mapper 318, and symbol multiplexor 322.
  • outer encoder module 302 is assumed to be a rate r/s convolutional encoder, receiving r input bits over one or more signal lines 300, and producing therefrom s parallel output bits over one or more signal lines 304, where both r and s are integers of one or more.
  • Parallel-to-serial (P/S) converter 306 serializes the s parallel bits output by outer encoder module 302, bit interleaver 308 interleaves these bits, and serial-to-parallel converter 310 converts the serial bit stream from interleaver 308 into successive renditions of k bits each.
  • Inner encoder module 314 is assumed to be a rate k/n convolutional encoder, receiving in parallel over one or more signal lines 312 k input bits and producing therefrom in parallel n output bits on one or more signal lines 316, where both k and n are integers of one or more.
  • Bit to symbol mapper 318 converts each n bit rendering from inner encoder module 314 into m symbols, where m is also an integer of one or more.
  • the mapper 318 may output the m symbols in the form of 2m symbol components for each n bit rendering from module 314. The m symbols may then be serialized onto signal line 324 through multiplexor 322. If the symbols are represented in quadrature form, the 2m symbol components may be serialized onto signal line 324 two symbol components at a time.
  • Module 314 in this example comprises logic 320 and storage register 324.
  • the k inputs to module 314, identified with numeral 322, are input to logic 320.
  • Another input to logic 320 is formed from the output of storage register 324.
  • the n outputs of module 314, identified with numeral 316, are output from logic 320.
  • One or more of these n outputs may comprise one of the k inputs passed through unaltered.
  • the n outputs are of module 314 are input to mapper 318.
  • logic 320 comprises five two-input exclusive OR modules 326a, 326b, 326c, 326d, and 326e (or equivalently modulo 2 modules.
  • Four of the input bits, u 0 , u 1? u 3 , and U 4 are passed through unaltered to respectively form output bits Y ls Y 2 , X ls and X 2 .
  • One of the output bits, Y 0 is derived from the modulo two addition of all five input bits, Uo, ui, u 2 , u 3 , and U
  • Another of the output bits, Xo is derived from the modulo two addition of Y 0 and the output of storage register 324.
  • mapper 318 may map each 6 bit output of inner encoder module 314 into two 8-PSK symbols. As illustrated, the output bits X 0 , Xi, and X 2 are mapped into a first symbol, and the output bits Yo, Yi, and Y 2 are mapped into a second symbol.
  • Figure 5 One example of such a mapping is illustrated in Figure 5.
  • each 3 bit grouping within the 6 bit output is mapped into a separate 8-PSK symbol.
  • a Gray scale mapping is employed, whereby the 3 bit groupings for adjacent symbols differ by a single bit.
  • the I and Q components for the first symbol may be represented as I ⁇ and Qx, and the I and Q components for the second symbol may be represented as I ⁇ and Q Y .
  • this particular implementation is provided by way of illustration only, and that other examples are possible in which any number of bit groupings or subgroupings are mapped into one or more MPSK or QAM symbols, including BPSK, QPSK, 8-PSK, 16-QAM, 64-QAM, and beyond.
  • inner decoder module 102 receives as an input over one or more signal lines 116 a priori probabilities of each of the k input bits to the inner encoder module 314 of Figure 3. Responsive to this information, inner decoder module 102 outputs on one or more signal lines 104 the extrinsic probabilities of each of the k input bits to the inner encoder module 314. In one embodiment, these extrinsic probabilities may be log likelihood ratios that comprise soft estimates of these input bits.
  • De-interleaver 106 receives the k extrinsic probabilities and de-interleaves them to form a priori probabilities of each of the s coded bits output from the outer encoder module 302 in Figure 3. Responsive to this information, outer decoder module 110 outputs on one or more signal lines 112 extrinsic probabilities of the s coded bits output from the outer encoder module 302 in Figure 3. (The other output of outer decoder module 110, comprises estimates of the r input bits to outer encoder module 302.) Again, these s extrinsic probabilities may be derived from soft estimates of the coded bits output from outer encoder module 302.
  • interleaver 114 which interleaves these values to produce, on one or more signal lines 116, the a priori probabilities of the k input bits to the inner decoder module 314. These k a priori probabilities form the second input to inner decode module 102.
  • the decoder illustrated in Figure 1 may iterate one or more additional times so that, at the conclusion of these iterations, the estimates of the r input bits to the outer encoder module 302 may be considered reliable. Thus, at this point, the estimates of the r input bits may be taken from the outer encoder module 302.
  • the inner decoder module 102 of Figure 1 is shown as a standalone module in
  • this module receives as inputs the m symbols output by mapper
  • the inner decoder module 102 may comprises one or more forward decoding engines and one or more backward decoding engines, each embodying the system of the invention. These decoding engines may operate concurrently to process incoming blocks of symbols.
  • the inner decoder module 102 comprises two backward engines, identified with numerals 651a and 651b, and one forward engine 652, each of which are configured according to the invention.
  • the engines may operate concurrently on portions (sub-blocks) of a block of encoded symbols within a movable sliding window according to the timing diagram illustrated in Figure 6.
  • the locations within the block of symbols being processed are identified with numeral 650.
  • the sub-blocks are the areas which are demarcated by the locations L 0 , Li, etc.
  • the backward engines can begin operating at any portion of the block and will eventually begin producing reliable results, i.e., reliable backward state probabilities, after a certain amount of processing.
  • reliable backward state probabilities i.e., reliable backward state probabilities
  • a rule of thumb is that the backward engines require about 5-6 constraint lengths of processing before they begin producing reliable results. These 5-6 constraint lengths are referred to as the traceback length. Processing by the backward engines over the traceback length is referred to in the figure with the "acquire" label. The period over which the backward engines produce reliable results is referred to in the figure with the "store” label. That is because these results are typically stored in a volatile memory such as RAM.
  • backward engine 65 la performs acquire processing of sub-block L 1 -L 2 , and during time TrT 2 , it performs reliable processing of sub-block Lo-Li.
  • backward engine 651b during time T!-T 2 , performs acquire processing of sub-block L 2 -L 3 , and during time T 2 -T 3 , performs reliable processing of sub-block LrL 2 .
  • backward engine 651a during time T 2 -T 3 , performs acquire processing of sub-block L 3 -L 4 , followed, during time T 3 -T 4 , by reliable processing of sub-block L 2 -L 3 .
  • backward engine 651b during time T 3 -T 4 , performs acquire processing of sub-block L -L 5 , and during time T -T 5 , performs reliable processing of sub-block L 3 -L 4 .
  • the backward engines then continue to operate in tandem as illustrated in the figure until the entire block has been processed.
  • reliable results for the sub-blocks are sequentially produced in the same order as the physical sub-blocks. That is to say, during T T 2 , reliable results are produced (by engine 651a) for sub-block Lo-Li. Then, during time T 2 -T 3 , reliable results are produced (by engine 651b) for sub-block L ⁇ -L 2 . This is followed by time T 3 -T 4 , when reliable results are produced (by engine 651a) for sub-block L 2 -L 3 . This is followed by time T -T 5 , when reliable results are produced (by engine 65 lb) for sub-block L 3 -L 4 . The process then continues in this fashion as illustrated in the figure.
  • the forward engine 652 operates in tandem with the completion of processing of reliable backward state probabilities for the various sub-blocks. That is to say, after reliable backward state probabilities are determined for sub-block Lo-Li during time T!-T 2 , the forward engine 652 performs, during time T 2 -T 3 , the calculation of forward state probabilities for sub-block L 0 -L ⁇ . At the same time, or synchronized with the calculation of the forward state probabilities, the forward engine 652 may use the immediately available forward state probabilities and the stored backward state probabilities to calculate and release soft outputs for sub-block Lo-Lj.
  • the forward engine 652 performs, during time T 3 -T 4 , the calculation of forward state probabilities for sub-block L 1 -L 2 .
  • the forward engine 652 may calculate and release soft outputs for sub-block L 1 -L 2 . The process then continues in the same manner until soft outputs for the entire block have been released. Note that the order in which the forward engine 652 processes sub-blocks to compute forward state probabilities and release soft outputs in this example is the same as the physical order of the sub-blocks.
  • the latency in this example in terms of traceback lengths, is equal to four. This is the delay between the time the system began acquire processing of a sub-block and the time the system completed the release of soft outputs for that sub-block. For example, consider sub-block L 1 -L 2 . The system began acquire processing of this sub-block at time T 0 . Yet, it did not complete the release of soft outputs for that sub-block until time T 4 . Since each time period is assumed equal to a traceback length in this example, the total latency is four traceback lengths.
  • FIG. 7A A first embodiment of a decoding system according to the invention is illustrated in Figure 7A.
  • m TCM encoded symbols identified in the figure with numeral 704
  • a priori probabilities for k source bits identified in the figure with numeral 706, where both m and k are integers of one or more.
  • the m symbols may result from passing the k source bits through a TCM encoder which, for purposes of this disclosure, is the combination of a rate k/n convolutional encoder and a bit to symbol mapper which maps the n output bits from the convolutional encoder into m symbols.
  • TCM encoder is the combination of inner encoder module 314 and mapper 318 in Figure 3 , but it should be appreciated that a TCM encoder is not limited to serving as the inner encoder in a SCTCM encoder, and that other applications are possible, including where the TCM encoder operates in a standalone mode.
  • the symbols may also have been transmitted over a wireless or wireline communications channel prior to being input to the system.
  • branch metric logic 702 within the system computes branch metrics for one or more of the branches in a trellis which corresponds to the TCM encoder which gave rise to the m symbols input to the engine.
  • the branch metric for a branch involves a measure of the correlation between the k inputs bits corresponding to the branch and the k a priori probabilities input to the engine, and also the correlation between the m symbols corresponding to the branch and the m symbols input to the system.
  • the number of states in the trellis at a particular point in time may be represented as p, where p is equal to two raised to the lth power, where 1 is the number of storage registers in the corresponding TCM encoder.
  • p is equal to two raised to the lth power
  • 1 is the number of storage registers in the corresponding TCM encoder.
  • p modules are provided, one for each of the p states in the trellis. These p modules are configured to compute in parallel the state probabilities of each of the p states in the trellis.
  • p modules are provided, one for each of the p states in the trellis. These p modules are configured to compute in parallel the state probabilities of each of the p states in the trellis.
  • less or more than p modules are provided, where the modules compute in parallel less than p of the state probabilities, or where some or all of the state probabilities are computed in parallel.
  • the system may function as either a forward engine or a backward engine.
  • ⁇ / ⁇ switch 710 is provided to indicate the particular mode of operation in which the engine is functioning, whether forward recursion mode or backward recursion mode.
  • each of the p modules 708(0), 708(1), . . . 708(p-l) may be configured to recursively compute forward or backward state probabilities depending on the state of ⁇ / ⁇ switch 710, which may be indicated to each of the modules over one or more signal lines 714. If the switch 710 indicates the backward (or ⁇ ) mode of operation, the modules may respond by computing backward state probabilities. Conversely, if the switch 710 indicates the forward (or ⁇ ) mode of operation, the modules may respond by computing forward state probabilities.
  • the modules recursively compute, in parallel, a backward state probability for one or more of the p states.
  • the backward state probabilities may be computed responsive to one or more of the branch metrics and one or more of the state probabilities of successor states.
  • the modules recursively compute, in parallel, a forward state probability for one or more of the p states.
  • the forward state probabilities may be computed responsive to one or more of the branch metrics and one or more of the state probabilities of predecessor states.
  • Estimation/extrinsic output logic 716 computes estimates or extrinsic outputs for the k underlying source bits responsive to one or more of the forward and backward state probabilities, one or more of the branch metrics, and, possibly, one or more of the a priori probabilities. These estimates or extrinsic outputs may be output over one or more signal lines 718.
  • the estimation/extrinsic output logic 716 may be enabled only when the system is functioning as a forward engine.
  • the mode of operation of the system whether forward or backward, may be provided to the logic 716 over one or more signal lines 714. If the switch indicates a backward mode of operation, the logic 716 may be disabled. If the switch indicates a forward mode of operation, the logic 716 may be enabled.
  • the backward state probability for a state may be computed by grouping the branches exiting the state according to the state at which the branches terminate, determining for each group a group value based on the branch metrics for the branches in the group, and then determining the backward state probability for the state responsive to the group values for one or more of the groups and one or more of the backward state probabilities of the successor states of the groups.
  • the group value for a group may be derived from the MAX* of the branch metrics for all the branches in the group.
  • the MAX* of two values A and B may be represented as follows:
  • the MAX* operation is associative, the MAX* of three or more values in a group may be performed through successive application of the MAX* operation on pairs of values.
  • the MAX * of A, B, and C may be computed as follows:
  • MAX * (A,B, C) MAX * (MAX * ⁇ A, B),C) (2)
  • MAX * (A,B,C,D) MAX * (MAX * (A,B,C),D) (3)
  • the backward state probability for state m' at time k+1, ⁇ k+ ⁇ m'), the backward state probability for state m" at time k+1, ⁇ k + ⁇ (m"), and the backward state probability for state m'" at time k+1, ⁇ k+ ⁇ (m'") are all assumed known.
  • the branches originating from state m at time k are first grouped according to the state at which the branches terminate. Thus, the t branches terminating at state m' are grouped together; the u branches terminating at state m' ' are grouped together; and the v branches terminating at state m'" are grouped together.
  • a first group value may then be computed responsive to one or more of the branch metrics in the first group, ⁇ , 0 ⁇ i ⁇ t-1.
  • a second group value may be computed responsive to one or more of the branch metrics in the second group, ⁇ k 1 , t ⁇ i ⁇ t+u-1.
  • a third group value may be computed responsive to one or more of the branch metrics in the third group, ⁇ , t+u ⁇ i ⁇ t+u+v-1.
  • the first value group may be derived from MAX*( ⁇ k ", Vi, 0 ⁇ i ⁇ t-1); the second group value may be derived from MAX*( ⁇ , Vi, t ⁇ i ⁇ t+u-1); and the third group value may be derived from MAX*( ⁇ k ', Vi, t+u ⁇ i ⁇ t+u+v-1).
  • the backward state probability ⁇ k (m) may then be derived from the first, second and third group values and the backward state probabilities of the successor states, ⁇ k+ ⁇ (m'), ⁇ k+ ⁇ (m"), and ⁇ k+ ⁇ (m'"), respectively.
  • the backward state probability ⁇ k (m) may be derived from MAX*[( ⁇ k+ ⁇ (m') + MAX*( ⁇ i , Vi, 0 ⁇ i ⁇ t-1)), ( ⁇ k+ ⁇ (m") + MAX*( ⁇ k ! , Vi, t ⁇ i ⁇ t+u-1)), ( ⁇ k+ ⁇ (m' ' ') + MAX*( ⁇ k j , Vi, t+u ⁇ i ⁇ t+u+v-1))].
  • the forward state probability for a state may be computed by grouping the branches entering the state according to the state at which the branches originate, determining for each group a group value based on the branch metrics for the branches in the group, and then determining the forward state probability for the state responsive to the values for one or more of the groups and one or more of the forward state probabilities of the predecessor states of the groups.
  • the group value for a group may be derived from the MAX* of the branch metrics for all the branches in the group.
  • the forward state probability for state m' at time k, ⁇ k(m'), the forward state probability for state m" at time k, ⁇ k (m"), and the forward state probability for state m'" at time k, ⁇ k (m'"), are all assumed known.
  • the branches terminating at state m at time k+1 are first grouped according to the state at which the branches originate. Thus, the t branches originating at state m' are grouped together; the u branches originating at state m' ' are grouped together; and the v branches originating at state m'" are grouped together.
  • a first group value may then be computed responsive to one or more of the branch metrics in the first group, ⁇ ', 0 ⁇ i ⁇ t-1.
  • a second group value may be computed responsive to one or more of the branch metrics in the second group, ⁇ , t ⁇ i ⁇ t+u-1.
  • a third group value may be computed responsive to one or more of the branch metrics in the third group, ⁇ 1 , t+u ⁇ i ⁇ t+u+v-1.
  • the first value group may be derived from MAX*( ⁇ k , Vi, 0 ⁇ i ⁇ t-1); the second group value may be derived from MAX*( ⁇ k ', Vi, t ⁇ i ⁇ t+u-1); and the third group value may be derived from MAX*( ⁇ k , Vi, t+u ⁇ i ⁇ t+u+v-1).
  • the forward state probability ⁇ k+ ⁇ (m) may then be derived from the first, second and third group values and the forward state probabilities, ⁇ k (m'), ⁇ k(m"), and ⁇ k (m'"), respectively, of one or more of the predecessor states to the three groups.
  • the forward state probability ⁇ k+ ⁇ (m) may be derived from MAX*( ⁇ k (m') + MAX*( ⁇ k i , Vi, 0 ⁇ i ⁇ t-1), ⁇ k (m") + MAX*( ⁇ J , Vi, t ⁇ i ⁇ t+u-1), ⁇ (m'") + MAX*( ⁇ k , Vi, t+u ⁇ i ⁇ t+u+v-1)).
  • the estimates or extrinsic outputs in this implementation may be computed one bit at a time.
  • the branches for a particular portion of the trellis may first be grouped according to the origination and termination states, and whether the branch implies release at the corresponding bit position of a logical "0" or logical "1".
  • a group value may then be computed for each of the groups based on the branch metrics for the branches in the group.
  • the estimates or extrinsic output for the bit may then be determined responsive to the group values, the forward state probabilities for the predecessor states of the groups, the backward state probabilities for the successor states of the groups, and, possibly, the a priori probability for the bit.
  • the group value for a group may be derived from the MAX* of the branch metrics for all the branches in the group.
  • the forward state probability of the origination state for the group and the backward state probability for the termination state for the group may both be added to the group value to form a second group value for the group.
  • a soft estimate for the bit may then be derived from the difference between the MAX* of the second group values for all those groups which imply a release of logical 1, and the MAX* of the second group values for all those groups which imply a release of logical 0.
  • An extrinsic output for the bit may then be derived by subtracting the a priori probability for the bit from the soft estimate of the bit. This process may then be repeated for each of the k bits.
  • branch metrics for the q branches from state m at time k to state m" at time k+1 which imply a release of a logical "0", ⁇ k 1 , 0 ⁇ i ⁇ q-1, are known; that the branch metrics for the r branches from state m at time k to state m" at time k+1 which imply a release of a logical "1", ⁇ k ', 0 ⁇ i ⁇ r-1, are known; that the branch metrics for the s branches from state at time k to state m" ' at time k+1 which imply release of a logical "0", ⁇ k , 0 ⁇ i ⁇ s-1 , are known; that the branch metrics for the t branches from state m at time k to state m'" at time k+1 which imply release of a logical "1", ⁇ k 1 , 0 ⁇ i ⁇ t-1, are known; that the branch metrics for the t
  • the branches may be grouped in groups according to the origination and termination states, and whether the branch implies release of a logical "1" or "0".
  • the branches are grouped into eight groups, with the first group comprising the q branches from state m at time k to state m" at time k+1 which implies release of a logical "0"; with the second group comprising the r branches from state m at time k to state m' ' at time k+1 which implies release of a logical "1"; with the third group comprising the s branches from state m at time k to state m'" at time k+1 which implies release of a logical "0"; with the fourth group comprising the t branches from state m at time k to state m'" at time k+1 which implies release of a logical "1”; with the fifth group comprising the u branches from state m' at time k to state m" at time k+1 which implies release of a logical "0"; with
  • a first group value may then be computed responsive to one or more of the branch metrics in the first group, ⁇ , 0 ⁇ i ⁇ q-1.
  • a second group value may be computed responsive to one or more of the branch metrics in the second group, ⁇ k ', q ⁇ i ⁇ q+r-1.
  • a third group value may be computed responsive to one or more of the branch metrics in the third group, ⁇ , q+r ⁇ i ⁇ q+r+s-1.
  • a fourth group value may then be computed responsive to one or more of the branch metrics in the first group, ⁇ k 1 , q+r+s ⁇ i ⁇ q+r+s+t- 1.
  • a fifth group value may be computed responsive to one or more of the branch metrics in the second group, ⁇ , q+r+s+t ⁇ i ⁇ q+r+s+t+u-1.
  • a sixth group value may be computed responsive to one or more of the branch metrics in the third group, ⁇ , q+r+s+t+u ⁇ i ⁇ q+r+s+t+u+v-1.
  • a seventh group value may then be computed responsive to one or more of the branch metrics in the first group, ⁇ k 1 , q+r+s+t+u+v ⁇ i ⁇ q+r+s+t+u+v+w-1.
  • An eighth group value may be computed responsive to one or more of the branch metrics in the eighth group, ⁇ , q+r+s+t+u+v+w ⁇ i ⁇ q+r+s+t+u+v+w+x-1.
  • the first group value may be derived from MAX*( ⁇ k ', Vi, 0 ⁇ i ⁇ q-1); the second group value may be derived from MAX*( ⁇ k ', Vi, q ⁇ i ⁇ q+r-1); the third group value may be derived from MAX*( ⁇ k , Vi, q+r ⁇ i ⁇ q+r+s-1 ); the fourth value group may be derived from MAX*( ⁇ k , Vi, q+r+s ⁇ i ⁇ q+r+s+t-1); the fifth group value may be derived from MAX*( ⁇ k ⁇ Vi, q+r+s+t ⁇ i ⁇ q+r+s+t+u-1); the sixth group value may be derived from MAX*( ⁇ k , Vi, q+r+s+t+u ⁇ i ⁇ q+r+s+t+u+v
  • the estimate or extrinsic output may then be derived from these eight group values and the forward state probabilities of the states from which the groups originate, ⁇ k (m') and ⁇ k (m'), the backward state probabilities of the states at which the groups terminate, ⁇ k + m") and ⁇ k+ ⁇ (m" !), and, possibly, the a priori probability ⁇ j for the bit in question.
  • the extrinsic output for a bit, BITj may be derived from MAX*[ ⁇ k (m) + MAX*( ⁇ k i , Vi, 0 ⁇ i ⁇ q-1) + ⁇ m' ') - ⁇ i5 ⁇ k (m) + MAX*( ⁇ k i , Vi, q+r i ⁇ q+r+s-1) + ⁇ k+ ⁇ (m"') - ⁇ i, ⁇ k (m') + MAX*( ⁇ k J , Vi, q+r+s+t ⁇ i ⁇ q+r+s+t+u-1) + ⁇ k+ ⁇ (m") - ⁇ ;, ⁇ k (m') + MAX*( ⁇ k J , Vi, q+r+s+t+u+v ⁇ i ⁇ q+r+s+t+u+v
  • FIG. 7B A second embodiment of a system according to the invention is illustrated in block diagram form in Figure 7B.
  • Four dimensional quadrature symbols identified in the figure with numeral 720 and represented as l ⁇ s Qx, I ⁇ , and Qy, are input to branch metric logic 724.
  • the symbols may comprises two 8-PSK symbols as produced from a TCM encoder comprising the series combination of a rate 5/6 convolutional encoder and a bit to symbol mapper for mapping each 6 bit output of the convolutional encoder into 2 8-PSK symbols.
  • Such an encoder is an example of the encoder illustrated in Figure 4 and previously described.
  • a priori probabilities for the 5 bits which gave rise to the two 8-PSK symbols, identified in the figure with numeral 722 and represented as ⁇ , are also input to the branch metric logic 724.
  • the branch metric logic 724 is configured to produce, in parallel, branch metrics for each of the branches in the portion of the trellis corresponding to the inputs 720 and 722.
  • branch metrics for each of the branches in the portion of the trellis corresponding to the inputs 720 and 722.
  • the corresponding portion of the trellis may be represented as illustrated in Figure 8D.
  • state 0 and state 1 reflecting the fact that the constraint length of the corresponding encoder is two (one plus the number of shift registers), and the number of states in the trellis is given by two raised to the power of the constraint length minus one.
  • Each branch represents a transition of the encoder from one of the states at time k to one of the states at time k+1 and therefore corresponds to a particular combination of the 5 input bits, represented in the figure with u, and the resulting two 8-PSK symbols, represented in the figure with x, y.
  • a total of 32 branches originate at state 0 at time k, representing the 32 possible combinations of the 5 input bits, and 32 branches originate at state 1 at time k, also representing the 32 possible combinations of the 5 input bits.
  • 16 terminates at state 0 at time k+1 and 16 terminate at state 1 at time k+1.
  • 16 terminates at state 0 at time k+1, and 16 terminates at state 1 at time k+1.
  • the branch metric logic 724 computes in parallel the 64 branch metrics for the 64 branches illustrated in Figure 8D.
  • the branch metric for a branch represents (1) a measure of the correlation between the 5 bits corresponding to the branch, which may be represented as u, with the 5 a priori probabilities input to the branch metric logic 724, which may be represented as ⁇ ; and (2) a measure of the correlation between the 2 8-PSK symbols corresponding to the branch, which may be represented as x and y, with the 2 8-PSK symbols input to the branch metric logic, which may be represented as I ⁇ , Qx, I ⁇ , and
  • QY- ⁇ / ⁇ switch 730 indicates whether the system is functioning as a backward engine ( ⁇ mode) or whether it is functioning as a forward engine ( ⁇ mode).
  • the branch metrics produced by branch metric logic 724 are provided to ⁇ / ⁇ logic (state 0) 726(0), ⁇ / ⁇ logic (state 1) 726(1), and extrinsic output logic 728 over one or more signal lines 734.
  • the indication of whether the system is functioning in ⁇ or ⁇ mode is provided to ⁇ / ⁇ logic (state 0) 726(0), ⁇ / ⁇ logic (state 1) 726(1), and extrinsic output logic 728 over one or more signal lines 732.
  • ⁇ / ⁇ logic (state 0) 726(0) is configured to update the backward or forward state probabilities of state 0
  • ⁇ / ⁇ logic (state 1) 726(1) is configured to update the backward or forward state probabilities of state 1. Both logic 726(0) and logic 726(1) are configured to operate in parallel.
  • ⁇ / ⁇ logic (state 0) 726(0) computes the backward state probability for state 0 at time k, ⁇ k (0), responsive to the backward probabilities for states 0 and 1 at time k+1, ⁇ k+ ⁇ (0) and ⁇ k+ ⁇ (l), respectively, and selected ones of the 64 branch metrics between the two states, which may be represented as ⁇ , 0 ⁇ i ⁇ 63, where ⁇ k 1 , 0 ⁇ i ⁇ 15, represent the branch metrics for the 16 branches between state 0 at time k and state 0 at time k+ 1 ; ⁇ k ", 16 ⁇ i ⁇ 31 , represent the branch metrics for the 16 branches between state 1 at time k and state 0 at time k+1; ⁇ k ', 32 ⁇ i ⁇ 47, represent the branch metrics for the 16 branches between state 0 at time k and state 1 at time k+1 ; and ⁇ k 1 , 48 ⁇
  • ⁇ / ⁇ logic (state 1) 726(1) computes the backward state probability for state 1 at time k, ⁇ k (l), responsive to the backward probabilities for states 0 and 1 at time k+1, ⁇ k+ ⁇ (0) and ⁇ k + ⁇ (l) > respectively, and selected ones of the 64 branch metrics between the two states, ⁇ k , 0 ⁇ i ⁇ 63.
  • the processes occur by first grouping the branches into 4 groups of 16 branches each according to their origination and termination states.
  • the first group represents the 16 branches between state 0 at time k and state 0 at time k+1;
  • the second group represents the 16 branches between state 1 at time k and state 0 at time k+1;
  • the third group represents the 16 branches between state 0 at time k and state 1 at time k+1 ;
  • the fourth group represents the 16 branches between state 1 at time k and state 1 at time k+1.
  • a group value is formed for each of the 4 groups equal to the MAX* of the branch metrics for the 16 branches in the group.
  • the first group value is set to MAX*( ⁇ k J , Vi, 0 ⁇ i ⁇ 15); the second group value is set to MAX*( ⁇ k j , Vi, 16 ⁇ i ⁇ 31); the third group value is set to MAX*( ⁇ k , Vi, 32 ⁇ i ⁇ 47); and the fourth group value is set to MAX*( ⁇ k ', Vi, 48 i ⁇ 63).
  • the backward state probability for the terminating state of a group at time k+1 is added to the group value for the group.
  • the backward state probability for state 0 at time k, ⁇ k(0) is then derived from the MAX* of the resulting values for the first and third groups, MAX*[( ⁇ k + ⁇ (0) + MAX*( ⁇ k , Vi, 0 ⁇ i ⁇ 15)), ( ⁇ k+ ⁇ (l) + MAX*( ⁇ 1 , Vi, 32 ⁇ i ⁇ 47))].
  • the backward state probability for state 1 at time k, ⁇ k(l), is then derived from the MAX* of the resulting values for the second and fourth groups, MAX*[( ⁇ k+ ⁇ (0) + MAX*( ⁇ k i , Vi, 16 ⁇ i ⁇ 31)), ( ⁇ k+ ⁇ (l) + MAX*( ⁇ k i , Vi, 48 ⁇ i ⁇ 63))].
  • ⁇ / ⁇ logic (state 0) 726(0) computes the forward state probability for state 0 at time k+1, ⁇ k+ ⁇ (0), responsive to the forward probabilities for states 0 and 1 at time k, ⁇ k (0) and ⁇ k (l), respectively, and selected ones of the 64 branch metrics for the branches between the states at times k and k+1.
  • branch metrics may be represented as ⁇ , 0 ⁇ i ⁇ 63, where ⁇ , 0 ⁇ i ⁇ 15, represent the branch metrics for the 16 branches between state 0 at time k and state 0 at time k+1; ⁇ k 1 , 16 ⁇ i ⁇ 31, represent the branch metrics for the 16 branches between state 1 at time k and state 0 at time k+1 ; ⁇ , 32 ⁇ i ⁇ 47, represent the branch metrics for the 16 branches between state 0 at time k and state 1 at time k+1 ; and ⁇ k 1 , 48 ⁇ i 63, represent the branch metrics for the 16 branches between state 1 at time k and state 1 at time k+1.
  • ⁇ / ⁇ logic (state 1) 726(1) computes the forward state probability for state 1 at time k+1, ⁇ k+ ⁇ (l), responsive to the forward probabilities for states 0 and 1 at time k, ⁇ k(0) and ⁇ k (l), respectively, and selected ones of the 64 branch metrics, ⁇ , 0 ⁇ i ⁇ 63.
  • first grouping the branches into 4 groups of 16 branches each according to their origination and termination states The first group represents the 16 branches between state 0 at time k and state 0 at time k+1; the second group represents the 16 branches between state 1 at time k and state 0 at time k+1; the third group represents the 16 branches between state 0 at time k and state 1 at time k+1; and the fourth group represents the 16 branches between state 1 at time k and state 1 at time k+1.
  • a group value is formed for each of the 4 groups equal to the MAX* of the branch metrics for the 16 branches in the group.
  • the first group value is set to MAX*( ⁇ k i , Vi, 0 ⁇ i ⁇ 15);
  • the second group value is set to MAX*( ⁇ k ', Vi, 16 ⁇ i ⁇ 31);
  • the third group value is set to MAX*( ⁇ k , Vi, 32 ⁇ i ⁇ 47);
  • the fourth group value is set to MAX*( ⁇ k ! , Vi, 48 ⁇ i ⁇ 63).
  • the forward state probability for the originating state of a group at time k is added to the group value for the group.
  • the forward state probability for state 0 at time k+1, ⁇ + ⁇ 0) is then derived from the MAX* of the resulting values for the first and second groups, MAX*[( ⁇ k (0) + MAX*( ⁇ k i 5 Vi, 0 ⁇ i ⁇ 15)), ( ⁇ k (l) + MAX*( ⁇ k j , Vi, 16 ⁇ i ⁇ 32))].
  • the forward state probability for state 1 at time k+1, ⁇ + i(l), is then derived from the MAX* of the resulting values for the third and fourth groups, MAX*[( ⁇ k (0) + MAX*( ⁇ k' ' , Vi, 32 ⁇ i ⁇ 47)), ( ⁇ k (l) + MAX*( ⁇ k ', Vi, 48 ⁇ i ⁇ 63))].
  • Extrinsic output logic 728 determines extrinsic outputs for each of the underlying 5 input bits. In one embodiment, extrinsic output logic 728 is only enabled if the ⁇ mode is called for, i.e., if the system is functioning as a forward engine. In this embodiment, one or more backward engines compute backward state probabilities for a portion of a trellis, and then one or more forward engines compute forward state probabilities and concurrently generate extrinsic outputs in parallel with the calculation of forward state probabilities. However, it should be appreciated that other embodiments are possible, including embodiments where extrinsic outputs are generated by one or more forward engines in parallel with the calculation of backward state probabilities.
  • extrinsic output logic 728 is able to produce extrinsic outputs only after the forward state probabilities for time k have been produced, the backward state probabilities for time k+1 have been produced, and the 64 branch metrics for the branches between the states in the two time periods have been produced.
  • This process occurs through 5 iterations, one for each of the 5 underlying input bits.
  • the process begins by dividing the 64 branches into 8 groups, according to the originating and terminating states, and whether the branch implies release of a logical "1" or "0" for the underlying source bit. These 8 groups may be depicted as shown in Figure 8E.
  • a solid line represents a branch which implies release of a logical "1" and a dashed branch implies release of a logical "0".
  • the first group comprises the 8 solid branches from state 0 at time k to state 0 at time k+1; the second group comprises the 8 dashed branches from state 0 at time k to state 0 at time k+1; the third group comprises the 8 solid branches from state 1 at time k to state 0 at time k+1 ; the fourth group comprises the 8 dashed branches from state 1 at time k to state 0 at time k+1; the fifth group comprises the 8 solid branches from state 0 at time k to state 1 at time k+1; the sixth group comprises the 8 dashed branches from state 0 at time k to state 1 at time k+1 ; the seventh group comprises the 8 solid branches from state 1 at time k to state 1 at time k+1; and the eighth group comprises the 8 dashed branches from state 1 at time k to state 1 at time k+1.
  • Group values may then derived for each group from the MAX* of the branch metrics for the branches in the group.
  • the branch metrics for the first group may be represented as ⁇ , 0 ⁇ i ⁇ 7; those for the second group as ⁇ , 8 ⁇ i ⁇ 15; those for the third group as ⁇ k ', 16 ⁇ i ⁇ 23; those for the fourth group as ⁇ k ", 24 ⁇ i ⁇ 31 ; those for the fifth group as ⁇ , 32 ⁇ i ⁇ 39; those for the sixth group as ⁇ , 40 ⁇ i ⁇ 47; those for the seventh group as ⁇ k 1 , 48 ⁇ i ⁇ 55; and those for the eighth group as ⁇ , 56 ⁇ i ⁇ 63.
  • the first group value may be set to MAX*( ⁇ k , Vi, 0 ⁇ i ⁇ 7); the second group value may be set to MAX*( ⁇ k , Vi, 8 ⁇ i ⁇ 15); the third group value may be set to MAX*( ⁇ k , Vi, 16 ⁇ i ⁇ 23); the fourth group value may be set to MAX*( ⁇ k , Vi, 24 ⁇ i ⁇ 31); the fifth group value may be set to MAX*( ⁇ i , Vi, 32 ⁇ i ⁇ 39); the sixth group value may be set to MAX*( ⁇ k , Vi, 40 ⁇ i ⁇ 47); the seventh group value may be set to MAX*( ⁇ k i , Vi, 48 ⁇ i ⁇ 55); and the eighth group value may be set to MAX*( ⁇ k ', Vi, 56 ⁇ i ⁇ 63).
  • the forward state probability of the originating state, and the backward state probability for the terminating state may be added to the group value for the group.
  • the MAX* of all resulting values from groups which imply release of a logical "1" is then formed.
  • this value may be represented as MAX*[( ⁇ k (0) + MAX*( ⁇ k j , Vi, 0 ⁇ i ⁇ 7) + ⁇ k+ ⁇ (0)), ( ⁇ k (0) + MAX*( ⁇ k ', Vi, 16 ⁇ i ⁇ 23) + ⁇ k+ ⁇ (l)), ( ⁇ k (l) + MAX*( ⁇ k J , Vi, 32 ⁇ i ⁇ 39) + ⁇ k+ ⁇ (0)), ( ⁇ k (l) + MAX*( ⁇ k i , Vi, 48 ⁇ i ⁇ 55) + ⁇ k + ⁇ (l))].
  • this value may be represented as MAX*[( ⁇ k (0) + MAX*( ⁇ k j , Vi, 8 ⁇ i ⁇ 15) + ⁇ k+ ⁇ (0)), ( ⁇ k (0) + MAX*( ⁇ k i , Vi, 24 ⁇ i ⁇ 31) + ⁇ k+ ⁇ (l)), ( ⁇ k (l) + MAX*( ⁇ k i , Vi, 40 ⁇ i ⁇ 47) + ⁇ k+ ⁇ (0)), ( ⁇ k (l) + MAX*( ⁇ k i , Vi, 56 ⁇ i ⁇ 63) + ⁇ k+ ⁇ (l))]- The difference between these two values, MAX*[( ⁇ k (0) + MAX*( ⁇ k !
  • This implementation is a system, which may be configured as a forward engine or a backward engine, which comprises two core modules, one for state 0 and the other for state 1, and an extrinsic output logic module.
  • the two core modules function and are configured similarly to one another, and therefore need not be shown separately.
  • Figure 9 illustrates the core module
  • Figure 10 illustrates the extrinsic output logic module.
  • the core module receives as inputs two 8-PSK symbols, typically after transmission over a wireless or wireline communication channel, X and Y, and a priori probabilities ⁇ for the 5 underlying source bits, and determines a state probability for the state corresponding to the module.
  • the core module comprises branch metric logic 912 and state probability logic 936.
  • Branch metric logic 912 comprises cross correlation symbol X logic 902 for computing, for each of the 32 branches originating or terminating at the state associated with the module, the correlation between the input 8-PSK symbol X and the 8-PSK symbol X associated with the branch in question (both of which may be represented in quadrature form as I ⁇ and Qx); cross correlation symbol Y logic 904 for computing the correlation between the input 8-PSK symbol Y and the '8-PSK symbol Y associated with the branch in question (both of which may be represented in quadrature form as I ⁇ and Q Y ); and A Priori Sum logic 906 for computing the correlation between the input a priori probabilities ⁇ for the 5 underlying source bits, and the 5 source bits associated with the branch in question.
  • Adder 908 is also provided for adding the three correlation values to form a branch metric for the branch in question.
  • branch metric logic 912 determines the 32 branch metrics which correspond to the 32 branches which terminate at the state associated with the logic 936.
  • branch metric logic 912 determines the 32 branch metrics which correspond to the 32 branches which originate at the state associated with the logic 936.
  • Figure 8D if the logic 936 is associated with state 0 and is part of a forward engine, the 32 branches are those identified with numeral 802. Similarly, if the logic 936 is associated with state 1 and is part of a forward engine, the 32 branches are those identified with numeral 804.
  • the 32 branches are those identified with numeral 806.
  • the 32 branches are those identified with numeral 808.
  • the 32 branch metrics from branch metric logic 912 form two groups of 16 each, with the grouping depending on whether the core module is functioning as part of a forward or backward engine and whether the logic 936 within the module is associated with state 0 or state 1.
  • the two groups are those identified with numerals 810 and 812.
  • the two groups are those identified with numerals 814 and 816.
  • the two groups are those identified with numerals 810 and 814.
  • the two groups are those identified with numerals 812 and 816.
  • the two groupings of 16 branch metrics are each input to separate renderings of MAX* 16-> 1 logic 914a, 914b which determine in parallel the MAX* of all the branches in a group.
  • adders 916a, 916b add to the MAX* value for the group a previously determined state probability.
  • the state probability is an ⁇ value for the originating state for the group
  • the state probability is a ⁇ value for the terminating state for the group.
  • the state probability logic 936 is functioning as part of a forward engine, and is seeking to compute ⁇ k + ⁇ ;(0), for the group identified with numeral 810, the previously determined state probability will be the value ⁇ k(0), and for the group identified with numeral 812, the previously determined state probability will be the value ⁇ k (l). If the state probability logic 936 is functioning as part of a forward engine, and is seeking to compute ⁇ k + ⁇ (l), for the group identified with numeral 814, the previously determined state probability will be the value ⁇ k(0), and for the group identified with numeral 816, the previously determined state probability will be the value ⁇ k (l).
  • the state probability logic 936 is functioning as part of a backward engine, and is seeking to compute ⁇ k (0), for the group identified with numeral 810, the previously determined state probability will be the value ⁇ k + ⁇ (0), and for the group identified with numeral 814, the previously determined state probability will be the value ⁇ k + ⁇ (l). If the state update logic 936 is functioning as part of a backward engine, and is seeking to compute ⁇ k(l), for the group identified with numeral 806, the previously determined state probability will be the value ⁇ k+ ⁇ (0), and for the group identified with numeral 808, the previously determined state probability will be the value ⁇ k+ ⁇ (l).
  • the outputs of adders 916a, 916b are then input to MAX* 2->l logic 924. (One of these outputs is coupled to input 925 of logic 924; the other is coupled to input 926 of logic 924).
  • MAX* 2->l logic 924 computes the MAX* of its two inputs, and this value is then input to normalization logic 930 on signal line 931, which normalizes this value in relation to the corresponding value (output of logic 928) from the core module for the other state in the trellis. This other value is provided to input 932 of normalization logic 932.
  • the output of logic 928 in this other module is coupled to input 932 of normalization logic 930, which may be shared between the two core modules.
  • normalization logic 930 normalizes the two values relative to one another. It outputs the normalized value of the input 931 on signal line 935, and outputs the normalized value of the input 932 on signal line 934.
  • the normalized value on signal line 935 is stored in storage logic 918.
  • the other normalized value on signal line 934 may also be stored within the storage logic 918 assuming it is shared between the two modules. For implementations in which separate storage logic is maintained in the core modules for the two states, the other normalized value on signal line 934 may be stored in the storage logic 918 for the other core module.
  • Figure 10 illustrates the extrinsic output logic module. As illustrated, the extrinsic output module comprises branch metric logic 912, storage logic 918, and extrinsic output computation logic 1000.
  • This module is configured to compute extrinsic outputs for each of the 5 underlying source bits once the state probabilities and branch metrics for the corresponding portion of the trellis have been computed.
  • branch metric logic 912 provides the branch metrics for the 64 branches that are present between the states in successive time periods.
  • Storage logic 918 provides the two forward state probabilities at time k, ⁇ k (0) and ⁇ k (l), and the two backward state probabilities at time k+1, ⁇ k+ ⁇ (0) and ⁇ k+ ⁇ (l).
  • Logic 1002 provides one of the five a priori probabilities ⁇ j.
  • the 64 branch metrics are divided into eight groups, according to originating and terminating states, and whether the branch implies release of a logical "1" or "0" for the bit position in question (there are five).
  • the four groups which imply release of a logical "1”, identified in Figure 8E with numerals 820, 822, 824, and 826, are each input to MAX* 8->l logic 1004, which separately determines the MAX* of each group.
  • the four groups which imply release of a logical "0”, identified in Figure 8E with numerals 828, 830, 832, and 834, are each input to MAX* 8->l logic 1006, which separately determines the MAX* of each group.
  • the MAX* values for the four groups which imply release of a logical "1" are input to adder logic 1008 which separately adds to each of these values the forward state probability of the state from which the group originates at time k, ⁇ k (0) and ⁇ k (l), and the backward state probability of the state at which the group terminates at time k+1, ⁇ k+ ⁇ (0) and ⁇ k+ ⁇ (l).
  • adder logic 1008 which separately adds to each of these values the forward state probability of the state from which the group originates at time k, ⁇ k (0) and ⁇ k (l), and the backward state probability of the state at which the group terminates at time k+1, ⁇ k+ ⁇ (0) and ⁇ k+ ⁇ (l).
  • the shifted value of the a priori probability ⁇ j is subtracted from each of these values, so that the outputs of the module will truly be extrinsic outputs.
  • the MAX* values for the four groups which imply release of a logical "0" are input to adder logic 1010 which separately adds to each of these values the forward state probability of the state from which the group originates at time k, ⁇ k (0) and ⁇ k (l), and the backward state probability of the state at which the group terminates at time k+1, ⁇ k+ ⁇ (0) and ⁇ k u(l).
  • the shifted value of the a priori probability ⁇ j is subtracted from each of these values, so that the outputs of the module will truly be extrinsic outputs.
  • the resulting four values from adder logic 1008 are input to MAX* 4->l logic 1016, which determines the MAX* of these four values.
  • the resulting four values from adder logic 1006 are input to MAX* 4->l logic 1018, which determines the MAX* of these four values.
  • Adder 1020 subtracts the MAX* value of the groups which imply release of a logical "0" (output from MAX* 4->l logic 1018) from the MAX* value of the groups which imply release of a logical "1" (output from MAX* 4-> 1 logic 1016). The result is an extrinsic output of the bit in question.
  • Figure 11 A illustrates an implementation of cross correlation symbol X logic 902 (as well as cross correlation symbol Y logic 904 since the two are configured identically).
  • This implementation takes accountof the fact that each symbol in an 8- PSK symbol constellation rotated counter-clockwise by ⁇ /8 radians (22.5°) (illustrated in Figure 1 IB) can be represented in quadrature form as either ( ⁇ cos( ⁇ /8), ⁇ sin( ⁇ /8)) or ( ⁇ sin( ⁇ /8), ⁇ cos( ⁇ /8)). Since cos( ⁇ /8) ⁇ 15/16, and sin( ⁇ /8) ⁇ 6/16, the representations may further by simplified as ( ⁇ 15/16, ⁇ 6/16) or ( ⁇ 6/16, ⁇ 15/16). Ignoring the divisor of 16 which is common to all the values, the representations can even further be simplified as ( ⁇ 15, ⁇ 6) or ( ⁇ 6, ⁇ 15).
  • multiplier 1106 multiplies the I component of the input symbol, identified with numeral 1102, by the appropriate value, ⁇ 15/ ⁇ 6, which represents the I component of the 8-PSK symbol which corresponds to the branch in question
  • multiplier 1108 multiplies the Q components of the input symbol, identified with numeral 1104, by the appropriate value, ⁇ 6/ ⁇ l 5, which represents the Q component of the 8-PSK symbol which corresponds to the branch in question.
  • Adder 1110 adds the outputs of the two multipliers.
  • the resulting value, identified with numeral 1114 represents the cross-correlation between the input symbol and the symbol associated with the branch in question.
  • cross correlation symbol X logic 902 (and cross correlation symbol Y logic 904), please refer to U.S. Patent Application Serial No. 09/815,149, filed March 22, 2001, which is hereby fully incorporated by reference herein as though set forth in full.
  • FIG. 12 An implementation of a priori sum logic 906 is illustrated in Figure 12.
  • Figure 13 illustrates an implementation of MAX* 16->1 logic 914a, 914b appearing as part of state probability logic 936 illustrated in Figure 9.
  • this logic may be implemented as a hierarchical tree of multiple renderings of MAX* 2->l logic, with the first level 1302 consisting of 8 separate renderings of the MAX* 2->l logic; with the second level 1304 consisting of 4 separate renderings of the MAX* 2->l logic; with the third level 1306 consisting of 2 separate renderings of the MAX* 2->l logic; and with the fourth level 1308 consisting of a single rendering of the MAX* 2-> 1 logic.
  • Figure 14 illustrates an implementation of MAX* 8->l logic appearing in blocks 1004, 1006 as part of the extrinsic output logic illustrated in Figure 10.
  • this logic may be implemented as a hierarchical tree of multiple renderings of MAX* 2->l logic, with the first level 1402 consisting of 4 separate renderings of the MAX* 2->l logic; with the second level 1404 consisting of 2 separate renderings of the MAX* 2->l logic, and with the third level consisting of a single rendering of the MAX* 2->l logic.
  • Figure 15 illustrates an implementation of MAX* 4->l logic appearing as blocks 1016, 1018 within the extrinsic output logic illustrated in Figure 10. As illustrated, this logic may be implemented as a hierarchical tree of multiple renderings of MAX* 2->l logic, with the first level consisting of 2 separate renderings of the MAX* 2->l logic; and with the second level consisting of a single rendering of the MAX* 2->l logic.
  • Figure 16A illustrates an implementation of the MAX* 2->l logic which appears directly as block 924 in Figure 9, and which, as described above, forms the basis for the MAX* 16->1, MAX* 8->l, and MAX* 4->l logic appearing in Figures 9 and 10.
  • logic 1602 receives the two operands, A and B, forms the difference, A - B, between the two, and outputs a signal representative of this difference value.
  • Logic 1604 receives the signal representing this difference value, tests whether it is greater than zero, and then outputs a signal indicative of the results of this test.
  • Logic 1606 receives this signal, and the two operands A and B, and outputs the operand A if the signal from logic 1604 indicates that the difference value A - B is greater than zero; if this signal indicates that the difference value A - B is less than or equal to zero, it outputs the operand B.
  • the output of logic 1606 is a signal representative of the maximum of A and B, or MAX(A, B).
  • Logic 1610 receives from logic 1602 the signal representative of the difference value A - B, and from logic 1604 the signal indicative of whether the difference value A - B is greater than zero, and outputs a signal representative of the value A - B if the signal from logic 1604 indicates that the difference value A - B is greater than zero; if the signal from logic 1604 indicates that the difference value A - B is less than or equal to zero, logic 1610 outputs a signal representative of the value -(A - B).
  • the output of logic 1610 is a signal representative of the absolute value of A - B or ABS(A - B).
  • Logic 1614 receives the signal from logic 1610 (which is representative of ABS(A - B)) and accesses a lookup table (LUT) using the value ABS(A - B) as the argument.
  • the LUT which is illustrated in Figure 16B, associates a value corresponding to ln(l + exp(- ⁇ A - R
  • logic 1614 retrieves a value corresponding to ln(l + exp(- - R
  • Logic 1614 outputs a signal representative of or corresponding to the value .ln(l + ex ⁇ (-
  • Logic 1616 receives the signals output from logic 1614 and logic 1606, and adds the two to form a signal representative of the value MAX (A, B) + ln(l + exp(- Referring to equation (1) above, it can be seen that this value is the MAX* of the operands A and B.
  • Figure 17 illustrates an implementation of the normalization block 930 illustrated in Figure 9. As illustrated, logic 1702 determines the maximum one of the two input values provided respectively on signal lines 931 and 932, and logic 1704 and logic 1706 subtract this maximum from each of the two input values. Consequently, one of the outputs of logic 1702 will be zero, and the other output will be negative. The results comprise normalized values provided respectively on signal lines 935 and 934.
  • any of the logic referred to above can be implemented in hardware, software, or a combination of hardware and software.
  • the logic may be implemented in hardware as an asynchronous integrated circuit chip (ASIC), or a DSP or other processor configured to execute a series of instructions embodying a method according to the invention.
  • ASIC asynchronous integrated circuit chip
  • DSP digital signal processor
  • Figure 18 is a flowchart of one embodiment of a method for determining state probabilities according to the invention. As illustrated, the method begins with step 1802, where k a priori probabilities ⁇ j, 1 ⁇ i ⁇ k, and m symbols Sj, 1 ⁇ i ⁇ k, are input. Step 1802 is followed by state 1804, where branch metrics are computed for the branches in a portion of a trellis responsive to the a priori probabilities and symbols input in the previous step.
  • the branch metric for a branch is computed by determining the correlation between the k a priori probabilities with the k bits associated with the branch, determining the correlation between each of the m symbols and the corresponding symbol associated with the branch, and then adding all these correlation values together to arrive at the branch metric for the branch.
  • step 1804 the method branches into p different branches, where p is the number of states in the trellis, which are all performed in parallel. The same steps are performed in each branch, so only the branch at the far left for state 0 need be explained in detail, it being understand that this description applies to each of the other branches.
  • the first step within the branch for state 0 is step 1806(0), in which the branches terminating or originating at the state are divided into groups. If forward state probabilities are being computed, this step involves dividing the branches terminating at the state into groups according to originating state; if backward state probabilities are being computed, this step involves dividing the branches originating at the step into groups according to terminating state.
  • Step 1806(0) is followed by step 1808(0), in which a group operation is performed on each of the groups to determine a group value for each group.
  • the group operation is the MAX* operation performed on the branch metrics of the branches in a group
  • the group value for the group is the MAX* of these group values.
  • Step 1806(0) is followed by step 1808(0), in which the group values for the groups are updated responsive to the values of previously determined state probabilities. If forward state probabilities are being computed, the forward state probability of the originating state for the group is added to the group value for the group; if backward state probabilities are being computed, the backward state probability of the terminating state for the group is added to the group value for the group.
  • Step 1810(0) is followed by step 1812(0), in which a group operation is performed on all the updated group values to determine a state probability for the state.
  • the group operation is the MAX* operation.
  • step 1812(0) the method waits if necessary until all the p branches are completed. At that point, optional step 1814 is performed, in which the p state probabilities are normalized relative to one another. After optional step 1814, the method may branch back to step 1802 for another iteration.
  • FIG 19 is a flowchart illustrating one embodiment of a method for determining extrinsic outputs according to the invention.
  • the method begins with step 1901 , in which k a priori probabilities ⁇ are input.
  • Step 1901 is followed by step 1902, in which an index i is initialized to zero.
  • Step 1906 follows step 1902.
  • the index i is tested to determine if it exceeds k, the total number of input bits which correspond " to a branch. If so, the method jumps back to step 1901 for another iteration of the method with a new set of k a priori probabilities; if not, the method proceeds to step 1908.
  • step 1908 the branches in the corresponding portion of the trellis are divided into groups according to originating and terminating states, and whether the branch implies release of a logical "1" or "0" for the input bit in the ith position
  • Step 1908 is followed by step 1910, in which a group operation is performed, and a group value obtained, for each of the groups.
  • the group operation is the MAX* operation performed on the branch metrics for the branches in a group
  • the group value is the MAX* of these branch metrics.
  • Step 1910 is followed by step 1912, in which the group values are updated by adding to each group value the forward state probability of the originating state for the group.
  • Step 1912 is followed by step 1914, in which the group values are further updated by adding to the group value the backward state probability of the terminating state for the group.
  • Step 1914 is followed by step 1916, in which a group operation is performed on all the group values for groups which imply release of a logical "1" for BITj. In one implementation, this group operation is the MAX* operation.
  • Step 1918 follows step 1916.
  • a group operation is performed on all the group values for groups which imply release of a logical "0".
  • this group operation is the MAX* operation.
  • Step 1918 is followed by step 1920.
  • step 1920 the value from step 1918 is subtracted from the value from step 1916 to determine a soft estimate of BITj.
  • Step 1922 follows step 1920.
  • the a priori probability ⁇ j is subtracted from the soft estimate computed in step 1920 to determine an extrinsic output for BITj.
  • the a priori probability ⁇ j for BITj may be subtracted from the group value for each group as part of steps 1912 or 1914.
  • Step 1924 follows step 1922.
  • the index i is incremented.
  • a jump is then made to step 1906 for another iteration of the method for another bit.
  • processor readable medium including without limitation RAM, ROM, PROM, EPROM, EEPROM, floppy disk, hard disk, CD-ROM, DVD, etc.
  • One advantage which may be realized in one embodiment is a reduction in computational complexity due to the order in which group operations (such as MAX*) are performed on the branch metrics in the process of determining state probabilities.
  • the forward state probability for a state may be computed by grouping the branches terminating at the state into groups according to the originating states of the branches, performing a first group operation (such as MAX*) on the branch metrics in each group and determining a group value (such as MAX*) for each of the groups, updating the group value for each group by adding the (previously determined) forward state probability of the originating state of the group, and then performing a second group operation (such as MAX*) on the updated group values to determine the forward state probability for the state.
  • a first group operation such as MAX*
  • a group value such as MAX*
  • the backward state probability for a state may be computed by grouping the branches originating at the state into groups according to the terminating states of the branches, performing a first group operation (such as MAX*) on the branch metrics in each group and determining a group value (such as MAX*) for each of the groups, updating the group value for each group by adding the (previously determined) backward state probability of the terminating state of the group, and then performing a second group operation (such as MAX*) on the updated group values to determine the backward state probability for the state.
  • a first group operation such as MAX*
  • a group value such as MAX*
  • a second advantage which may be realized in one embodiment is a reduction in computational complexity due to the normalization of state probabilities which is performed. More specifically, by normalizing state probabilities, the magnitude of the state probabilities is reduced. Hence, the process of recursively computing state probabilities and the process of computing extrinsic outputs using these state probabilities is simplified.
  • a third advantage which may be realized in one embodiment is an efficiency in the design of the MAX* 2->l logic which occurs through the sharing of the output of the test logic, which indicates whether the value A - B is greater than zero, by the MAX logic and the ABS logic.
  • a fourth advantage which may be realized in one embodiment is simplicity in the design of the MAX* 16->1, MAX* 8->l, and MAX* 4->l logic each of which are formed using the MAX* 2->l logic as constituent building blocks.
  • a fifth advantage which may be realized in one embodiment is design efficiency in that a forward and backward engine may each be configured from essentially the same core module.
  • a sixth advantage which may be realized in one embodiment is design efficiency in that the state update modules within the core module may each be implemented from essentially the same logic.
  • a seventh advantage which may be realized in one embodiment is computational efficiency in that a branch metric can be computed for a branch which represents both the level of correlation between one or more symbols and one or more symbols corresponding to the branch and one or more a priori probabilities and one or more bits associated with the branch.

Landscapes

  • Physics & Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Error Detection And Correction (AREA)

Abstract

La présente invention concerne des systèmes et des procédés associés permettant (1) de déterminer une ou plusieurs probabilités d'état concernant un ou plusieurs états dans une représentation en treillis, (2) de déterminer une entrée estimée ou extrinsèque concernant un ou plusieurs bits utilisant une représentation en treillis, (3) de déterminer un élément métrologique de branche pour une branche de représentation en treillis, (4) d'effectuer un calcul de MAX* 2-⊃1, (5) d'effectuer un calcul de MAX* 2p-⊃1 dans lesquels p est un entier au moins égal à deux, dans un arrangement hiérarchique de calculs de MAX* 2-⊃1 et, (6) de calculer par déroulement les probabilités d'état dans un mode de déroulement de calcul. Cette invention concerne aussi des combinaisons des éléments susmentionnés.
PCT/US2001/048238 2000-12-15 2001-12-14 Systeme et procede de decodage de codes en treillis WO2002049219A2 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
AU2002232576A AU2002232576A1 (en) 2000-12-15 2001-12-14 Decoding of trellis codes

Applications Claiming Priority (8)

Application Number Priority Date Filing Date Title
US25579700P 2000-12-15 2000-12-15
US60/255,797 2000-12-15
US32725801P 2001-10-04 2001-10-04
US60/327,258 2001-10-04
US1268401A 2001-11-13 2001-11-13
US10/012,684 2001-11-13
US10/013,492 US6865711B2 (en) 2000-12-15 2001-12-13 System of and method for decoding trellis codes
US10/013,492 2001-12-13

Publications (2)

Publication Number Publication Date
WO2002049219A2 true WO2002049219A2 (fr) 2002-06-20
WO2002049219A3 WO2002049219A3 (fr) 2003-05-08

Family

ID=27486218

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2001/048238 WO2002049219A2 (fr) 2000-12-15 2001-12-14 Systeme et procede de decodage de codes en treillis

Country Status (2)

Country Link
AU (1) AU2002232576A1 (fr)
WO (1) WO2002049219A2 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7154965B2 (en) 2002-10-08 2006-12-26 President And Fellows Of Harvard College Soft detection of data symbols in the presence of intersymbol interference and timing error

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5541955A (en) * 1992-11-06 1996-07-30 Pericle Communications Company Adaptive data rate modem
US5933462A (en) * 1996-11-06 1999-08-03 Qualcomm Incorporated Soft decision output decoder for decoding convolutionally encoded codewords

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5541955A (en) * 1992-11-06 1996-07-30 Pericle Communications Company Adaptive data rate modem
US5933462A (en) * 1996-11-06 1999-08-03 Qualcomm Incorporated Soft decision output decoder for decoding convolutionally encoded codewords

Non-Patent Citations (11)

* Cited by examiner, † Cited by third party
Title
BENEDETTO S ET AL: "Iterative decoding of serially concatenated codes with interleavers and comparison with turbo codes" GLOBAL TELECOMMUNICATIONS CONFERENCE, 1997. GLOBECOM '97., IEEE PHOENIX, AZ, USA 3-8 NOV. 1997, NEW YORK, NY, USA,IEEE, US, 3 November 1997 (1997-11-03), pages 654-658, XP010254740 ISBN: 0-7803-4198-8 *
CHERUBINI G ET AL: "Concatenated coding for binary partial-response channels" COMMUNICATIONS, 1994. ICC '94, SUPERCOMM/ICC '94, CONFERENCE RECORD, 'SERVING HUMANITY THROUGH COMMUNICATIONS.' IEEE INTERNATIONAL CONFERENCE ON NEW ORLEANS, LA, USA 1-5 MAY 1994, NEW YORK, NY, USA,IEEE, 1 May 1994 (1994-05-01), pages 1789-1794, XP010126581 ISBN: 0-7803-1825-0 *
GROOHYUN PARK ET AL: "An implementation method of a turbo-code decoder using a block-wise MAP algorithm" IEEE PROCEEDINGS OF THE 2000 VEHICLUAR TECHNOLOGY CONFERENCE, vol. 6, 24 - 28 September 2000, pages 2956-2961, XP010525119 *
JUNG P ET AL: "Comprehensive comparison of turbo-code decoders" VEHICULAR TECHNOLOGY CONFERENCE, 1995 IEEE 45TH CHICAGO, IL, USA 25-28 JULY 1995, NEW YORK, NY, USA,IEEE, US, 25 July 1995 (1995-07-25), pages 624-628, XP010167017 ISBN: 0-7803-2742-X *
PIETROBON S S: "IMPLEMENTATION AND PERFORMANCE OF A TURBO/MAP DECODER" INTERNATIONAL JOURNAL OF SATELLITE COMMUNICATIONS, JOHN WILEY AND SONS, US, vol. 16, no. 1, January 1998 (1998-01) - February 1998 (1998-02), pages 23-46, XP000856961 ISSN: 0737-2884 *
RAGHUPATHY A ET AL: "A transformation for computational latency reduction in turbo-MAP decoding" CIRCUITS AND SYSTEMS, 1999. ISCAS '99. PROCEEDINGS OF THE 1999 IEEE INTERNATIONAL SYMPOSIUM ON ORLANDO, FL, USA 30 MAY-2 JUNE 1999, PISCATAWAY, NJ, USA,IEEE, US, 30 May 1999 (1999-05-30), pages 402-405, XP010341252 ISBN: 0-7803-5471-0 *
ROBERTSON P ET AL: "Extensions of turbo trellis coded modulation to high bandwidth efficiencies" COMMUNICATIONS, 1997. ICC '97 MONTREAL, TOWARDS THE KNOWLEDGE MILLENNIUM. 1997 IEEE INTERNATIONAL CONFERENCE ON MONTREAL, QUE., CANADA 8-12 JUNE 1997, NEW YORK, NY, USA,IEEE, US, 8 June 1997 (1997-06-08), pages 1251-1255, XP010226956 ISBN: 0-7803-3925-8 *
TRI NGO ET AL: "Turbo codes on the fixed point DSP TMS320C55x" 2000 IEEE WORKSHOP ON SIGNAL PROCESING SYSTEMS (SIPS 200), 11 - 13 October 2000, pages 255-264, XP010525235 Lafayette, LA, USA *
VOGT J ET AL: "Comparison of different turbo decoder realizations for IMT-2000" IEEE PROCEEDINGS GLOBECOM'99, vol. 5, 5 - 9 September 1999, pages 2704-2708, XP010373441 *
ZHIPEI CHI ET AL: "Iterative decoding of space-time trellis codes and related implementation issues" IEEE PROCEEDINGS OF THE CONFERENCE RECORD OF THE THIRTY-FOURTH ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS AND COMPUTERS, vol. 1, 29 October 2000 (2000-10-29) - 1 November 2000 (2000-11-01), pages 562-566, XP010535429 pacific grove CA (US) *
ZHONGFENG WANG ET AL: "VLSI implementation issues of TURBO decoder design for wireless applications" SIGNAL PROCESSING SYSTEMS, 1999. SIPS 99. 1999 IEEE WORKSHOP ON TAIPEI, TAIWAN 20-22 OCT. 1999, PISCATAWAY, NJ, USA,IEEE, US, 20 October 1999 (1999-10-20), pages 503-512, XP010370879 ISBN: 0-7803-5650-0 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7154965B2 (en) 2002-10-08 2006-12-26 President And Fellows Of Harvard College Soft detection of data symbols in the presence of intersymbol interference and timing error

Also Published As

Publication number Publication date
AU2002232576A1 (en) 2002-06-24
WO2002049219A3 (fr) 2003-05-08

Similar Documents

Publication Publication Date Title
US6865711B2 (en) System of and method for decoding trellis codes
US6145114A (en) Method of enhanced max-log-a posteriori probability processing
JP4741154B2 (ja) ターボ復号器に対する軟判定入力のメトリックを計算するための装置及び方法
RU2263397C1 (ru) Устройство и способ снижения коэффициентов битовых ошибок (кбо) и коэффициентов кадровых ошибок (кко) с использованием турбодекодирования в системе цифровой связи
Duman et al. Performance bounds for turbo-coded modulation systems
US6393076B1 (en) Decoding of turbo codes using data scaling
EP1119110A1 (fr) Procédé de transmission numérique de type à codage correcteur d'erreurs
GB2395097A (en) A decoder apparatus and method of decoding therefor
US6760390B1 (en) Log-map metric calculation using the avg* kernel
EP1314254A1 (fr) Achevement d'iteration a l'aide de criteres d'indices de qualite de turbocodes
Tong et al. VHDL implementation of a turbo decoder with log-MAP-based iterative decoding
US6973615B1 (en) System of and method for decoding trellis codes
WO2002049219A2 (fr) Systeme et procede de decodage de codes en treillis
Fowdur et al. Performance of Turbo coded 64-QAM with joint source channel decoding, adaptive scaling and prioritised constellation mapping
Kim et al. A Memory‐Efficient Block‐wise MAP Decoder Architecture
Song et al. The implementation of turbo decoder on DSP in W-CDMA system
Mathana et al. FPGA implementation of high speed architecture for Max Log Map turbo SISO decoder
Papaharalabos et al. Modified log-MAP algorithm for simplified decoding of turbo and turbo TCM codes
GB2418109A (en) Approximation of the logarithm of the sum of exponentials correction term in log-MAP decoding using Padé approximants or continued fractions
Osman et al. Performance of multilevel turbo codes with group partitioning over satellite channels
Verma et al. An FPGA realization of simplified turbo decoder architecture
Jung et al. Design and Architecture of Low‐Latency High‐Speed Turbo Decoders
Sybis et al. Simplified Log-MAP decoding with max* approximation based on the Jensen inequality
Adamu et al. Performance Analysis of Turbo Codes for 5G Massive Machine-Type Communication (mMTC)
Savitha et al. Performance comparison of hard and soft-decision turbo coded OFDM systems

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TN TR TT TZ UA UG UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase in:

Ref country code: JP

WWW Wipo information: withdrawn in national office

Country of ref document: JP