System and method of data communication using turbo trellis coded modulation combined with constellation shaping with or without precoding

Abstract
A technique that combines a turbo trellis coded modulation (TTCM) coding scheme with constellation shaping and precoding schemes to implement a binary coded communication system and method that can achieve high performance (high coding gains achieved in combination with shaping gain, and when necessary, also with high performance in ISI-channels via preceding).
Description


BACKGROUND OF THE INVENTION

[0002] 1. Field of the Invention


[0003] This invention relates generally to constellation shaping, precoding and coding, and more particularly to a method and system of combining turbo trellis coded modulation (Turbo-TCM/TTCM) with constellation shaping with/without preceding.


[0004] 2. Description of the Prior Art


[0005] Constellation shaping is a means for selecting the stream of transmitted symbols in a communication system, according to a criterion of minimum average power under the constraint of a given minimum distance between neighboring points. By doing that, the constituent two-dimensional constellation is “shaped” into a certain form (with a certain probability distribution between the constellation points). Thus, a symbol sequence with a lower average power can be used (or alternatively a sequence with a high inter-symbol distance between neighboring points for a given power constraint). This results in “shaping gain” which may cause the communication system to become more robust to noise and channel impairments, or result in higher achievable data rates.


[0006] Two known methods of constellation shaping are trellis shaping and shell mapping. On Gaussian channels with intersymbol interference (ISI), it is often desired to use preceding methods to mitigate channel distortion. Tomlinson-Harashima (TH) precoding is a well known precoding scheme for ISI-channels.


[0007] When constellation shaping is used in ISI channels, it is desired to use a combination of shaping and precoding. A method of combining trellis shaping with TH-precoding is called trellis precoding. Combining shell mapping with preceding can be done by incorporating Laroia precoding. It is known in the art that lattice codes can also be combined with trellis shaping and trellis precoding. Thus, trellis coded modulation (TTCM) schemes can be combined with shaping and preceding to achieve coding gain together with shaping gain, and to have high performance, even in ISI-channels.


[0008] Lattice codes (e.g. TTCM) can also be combined with shell mapping and Laroia preceding to achieve coding gain together with shaping gain, and to have high performance, even in ISI-channels. This is done for example, in the ITU V.34 standard for voice grade modems.


[0009] Recently, a new coding technique, “turbo coding” was introduced, which can be used to achieve higher coding gains than other coding schemes; and when used in communication systems, may result in better robustness to noise and other impairments, or in higher (closer to capacity) data rates. Whereas the original work in the field of turbo coding addresses the case of binary transmission, new schemes have recently been developed for incorporating turbo coding when using larger constellations. These schemes are referred to as Turbo trellis coded modulation (TTCM).


[0010] In view of the foregoing, a need exists for a technique that combines a TTCM coding scheme with shaping and preceding schemes to implement a system and method that can achieve better performance (higher coding gains achieved in combination with shaping gain, and when necessary, also with high performance in ISI-channels via precoding) than existing methods and systems.



SUMMARY OF THE INVENTION

[0011] The present invention is directed to a technique that combines a turbo trellis coded modulation (TTCM) coding scheme with constellation shaping and preceding schemes to implement a binary coded communication system and method that can achieve high performance (high coding gains achieved in combination with shaping gain, and when necessary, also with high performance in ISI-channels via preceding).


[0012] According to one embodiment, a method of binary coded data communication comprises the steps of:


[0013] providing a transmitter having a turbo trellis coded modulator (TTCM) encoder and constellation shaping elements; and


[0014] generating a plurality of signal points in response to a partitioned binary coded symbol sequence that is processed via the TTCM encoder and constellation shaping elements.


[0015] According to another embodiment, a binary coded data communication system comprises:


[0016] a transmitter having a turbo trellis coded modulator (TTCM) encoder and constellation shaping elements; and


[0017] a receiver having a turbo decoder and constellation shaping elements, wherein the transmitter is operational to generate a plurality of signal points in response to a partitioned symbol sequence that is processed via the TTCM encoder and constellation shaping elements, and further wherein the receiver is operational to receive the plurality of signal points over a transmission medium and recover the partitioned symbol sequence in response to the plurality of signal points that are received and processed via the receiver turbo decoder and the receiver constellation shaping elements.







BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Other aspects and features of the present invention and many of the attendant advantages of the present invention will be readily appreciated as the same become better understood by reference to the following detailed description when considered in connection with the accompanying drawings in which like reference numerals designate like parts throughout the figures thereof and wherein:


[0019]
FIG. 1 illustrates simulation results for a shaped constellation using trellis shaping compatible with TTCM when using R=5 information bits per 2D symbol;


[0020]
FIG. 2 is a simplified block diagram illustrating a transmitter using TTCM combined with trellis shaping according to one embodiment of the present invention;


[0021]
FIG. 3 is a simplified block diagram illustrating a receiver using TTCM combined with trellis shaping according to one embodiment of the present invention;


[0022]
FIG. 4 is a simplified block diagram illustrating a transmitter using TTCM combined with trellis precoding according to one embodiment of the present invention;


[0023]
FIG. 5 is a simplified block diagram illustrating a receiver using TTCM combined with trellis preceding according to one embodiment of the present invention;


[0024]
FIG. 6 is a simplified block diagram illustrating TH-precoding;


[0025]
FIG. 7 is a simplified block diagram illustrating a transmitter using TTCM combined with shell mapping according to one embodiment of the present invention;


[0026]
FIG. 8 is a simplified block diagram illustrating a receiver using TTCM combined with shell mapping according to one embodiment of the present invention;


[0027]
FIG. 9 is a simplified block diagram illustrating a transmitter using TTCM combined with shell mapping and Laroia precoding according to one embodiment of the present invention;


[0028]
FIG. 10 is a simplified block diagram illustrating a receiver using TTCM combined with shell mapping and Laroia precoding according to one embodiment of the present invention;


[0029]
FIG. 11 is a simplified block diagram illustrating Laroia precoding;


[0030]
FIG. 12 illustrates a 64 QAM constellation with square shaping regions, designed for trellis shaping and that is compatible with rate-½ code;


[0031]
FIG. 13 illustrates a 64 QAM constellation with spherical shaping regions, designed for trellis shaping and that is compatible with rate-½ code; and


[0032]
FIG. 14 illustrates a 32 point constellation with spherical shaping regions, designed for trellis precoding and that is compatible with rate-½ code.







[0033] While the above-identified drawing figures set forth alternative embodiments, other embodiments of the present invention are also contemplated, as noted in the discussion. In all cases, this disclosure presents illustrated embodiments of the present invention by way of representation and not limitation. Numerous other modifications and embodiments can be devised by those skilled in the art which fall within the scope and spirit of the principles of this invention.


DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0034] The present inventor recognized that although constellation shaping and precoding have been implemented in the past and also combined with TTCM, it has never before been combined with Turbo-TTCM. This combination of TTCM, constellation shaping and precoding is believed by the present inventor to be better than the use of TTCM alone since it is believed to provide additional shaping gain, in addition to the coding gain achieved by the TTCM. It is also believed to be better than conventional TTCM combined with shaping with or without precoding since TTCM achieves better coding gain than TTCM; thus the overall gain is higher, and better performance (e.g. data rates or robustness) can be achieved.


[0035] For channels with ISI, it is sometimes desirable to use precoding to mitigate channel impairments. Such precoding can be implemented in the present method, as discussed herein below. The present inventor has found shaping and preceding results for cases suitable for use with TTCM, using codes of rate ½ and rate ⅔. Particular embodiments presented herein below combine TTCM with shaping, and with/without preceding, subject to the following conditions: 1) Known TTCM coding schemes may be used; and the embodiments presented herein below deal with TTCMs of rate ½ and of rate ⅔ resulting in dividing the 2D constellation into 4 or 8 cosets respectively; 2) Known constellation shaping methods that may be used are trellis shaping or shell mapping; and 3) Precoding may also be used together with the shaping and Turbo-TTCM system. FIG. 1 illustrates simulation results for a shaped constellation using trellis shaping compatible with TTCM when using R=5 information bits per 2D symbol.


[0036]
FIG. 2 is a simplified block diagram illustrating a transmitter 200 using TTCM combined with trellis shaping according to one embodiment of the present invention. The input sequence can be seen to be divided into three parts (xJ, wJ, sJ). The first part is a binary kc-tuple, and is an input to a TTCM encoder 202 for a rate kc/nc TTCM code. The middle part is an uncoded binary nu-tuple. The last part, a syndrome rs-tuple, is an input to an rs input, ns output coset representative generator (Hs−1)T 204 for a rate ks/ns convolutional shaping code, where ks=ns−rs. The procedure used is then as disclosed in section III(A) of Trellis shaping, G. D. Forney, IEEE Trans. Inf., Vol. 38, March 1992, which is incorporated by reference herein, but implemented instead with the channel code being TTCM.


[0037]
FIG. 3 is a simplified block diagram illustrating a receiver 300 using TTCM combined with trellis shaping according to one embodiment of the present invention. In similar fashion, the procedure used in association with receiver 300 follows that disclosed in section III(C) of Trellis shaping, G. D. Forney, IEEE Trans. Inf., Vol. 38, March 1992, incorporated by reference herein, but with the channel code decoder 302 being a TTCM decoder such as that disclosed in Bandwidth efficient Turbo Trellis-coded Modulation using Punctured component codes, P. Robertson, T. Woerz, IEEE Journal on selected areas of comm., Feb. 1998, also incorporated by reference herein.


[0038] A method of using the trellis shaping is now set forth herein below with considerations for selecting the shaping regions for compatibility with the TTCM. In trellis shaping such as disclosed in Trellis shaping, G. D. Forney, IEEE Trans. Inf., Vol. 38, March 1992, incorporated by reference herein, one uses a “shaping code” (Cs of rate ks/ns) and then uses the redundancy (resulting in constellation expansion) to select such a code sequence that minimizes the overall power of the transmitted sequence. This is done by implementing a decoder for the code Cs (e.g. Viterbi algorithm (VA)) in the transmitter. The constellation is divided into 2ns “shaping regions”, each having the same number of points (when coding based on set-partitioning is used also, each region must have the same number of points from every coding subset). All input bits enter the decoder that selects the signal sequence (i.e. the shaping region sequence) which results in minimum energy (other criteria may be easily implemented by replacing the metric calculation in the VA).


[0039] The receiver requires reconstruction of the rs=ns−ks information bits per symbol that enter the shaping coder. This is done with a syndrome former for the code Cs (i.e. passing the sequence through a ns×ns−ks transfer function matrix HT) as stated herein before. As also stated above, trellis shaping can be easily implemented together with TTCM by using appropriate shaping regions. Issues that should be considered are now discussed with reference to a 4-state rate-½ convolutional shaping code which has been tested by the present inventor.


[0040] Trellis shaping was simulated using different constellations and shaping regions with the results shown in Table I below. These results are provided for schemes compatible with rate-{fraction (1/2)} and rate-⅔ codes where * indicates square region shaping was not checked, but is expected to give similar results as that provided by spherical region shaping.
1TABLE IDate rate RConstellationTheoreticalActualShapingInformationwithout →shaping gainShaping gainRegionsBits/symbolwith shaping[dB][dB]used ****Remarks24PSK →1.250SphericalNo coding8QAM38QAM →1.031.54Square/Compatible16QAMSphericalwith rate-1/2 code416QAM →0.960.68SphericalCompatible32CRwith rate-1/2 code532CR →0.97/0.950.69/0.67Square/Compatible64QAMSphericalwith rate-1/2 code664QAM →0.960.89 (0.79)***Spherical **Compatible128QAMwith rate-1/2 code7128QAM →0.970.79Square/Compatible256QAMSphericalwith rate-1/2 code8256QAM →0.960.95 (0.76)***Spherical **Compatible512QAMwith rate-1/2 code416QAM →0.850.57SphericalCompatible32CRwith rate-2/3 code532CR →*/0.90*/0.62Square/Compatible64QAMSphericalwith rate-2/3 code664QAM →0.960.89 (0.79)***Spherical **Compatible128QAMwith rate-2/3 code7128QAM →*/0.97*/0.78Square/Compatible256QAMSphericalwith rate-2/3 code8256QAM →0.970.95 (0.77)***Spherical **Compatible512QAMwith rate-2/3 code


[0041] The double ** shown in Table I indicates that for non-square constellations, it should be determined whether a scheme similar to square region shaping is possible, since this may be important for combination with TH-precoding. The triple *** indicates that actual shaping gain was calculated relative to a square constellation. This was calculated in accordance with




γ


s


=S


x
⊕(R)/Sx=2R/6Sx  (1)



[0042] where γs is defined as the ratio between the average energy per two dimensions of the transmitted sequence and the baseline average energy, that of a conventional square QAM constellation supporting the same data rate per two dimensions. Using a trivial 2D “constellation shaping” (i.e., using a circle constellation rather than a square) yields the relative shaping gain in shown parenthesis. The quad **** indicates that constellations with square shaping regions, and spherical shaping regions, designed for trellis shaping and that are compatible with rate-½ code are illustrated in FIGS. 12 and 13 respectively. As can be seen in the simulation results shown in Table I, a shaping gain of almost 1 dB may be achieved by using a 4-state rate-½ convolutional shaping code. Further, the shaping gain is achieved only with sufficient decoding delay in the decoder. Regarding the 4-state rate-½ code, a delay of 20 symbols was found to be sufficient. The results shown in Table I also indicate the shaping scheme depends on the shaping regions and on the constellation size. FIG. 1 illustrates simulation results for a shaped constellation using trellis shaping compatible with TTCM when using R=5 information bits per 2D symbol.


[0043] The complexity of trellis shaping lies primarily in the transmitter (e.g. 200 in FIG. 2), whereas the receiver (e.g. 300 in FIG. 3) is kept simple. Shaping also requires expanding the constituent 2D constellation by a factor of two.


[0044] The complexity of the transmitter involves implementing a (4-state) VA which requires eight metric calculations per symbol iteration (each metric calculation requires bits to symbol mapping and energy computation). The VA must be implemented in such a way that will result in a legitimate shaping code sequence, since other implementations will not work. Reducing complexity of the decoder (i.e. by using estimated symbol power rather than accurate symbol power in the branch metric) might result in degradation in the shaping gain, but will not harm the receiver. The information shaping bits (rs per symbol) in the transmitter 200 also pass through a ns−ks×ns=1×2 transfer function matrix (H−1)T, enumerated 204 in FIG. 2. One element of (H−1)T may be zero when implementing the present method of combining TTCM and constellation shaping. The information shaping bits in the receiver 300 are reconstructed by passing through a ns×ns−ks=2×1 transfer function matrix HT, enumerated 304 in FIG. 3.


[0045] Peak-to-average (PAR) constraints can be employed in the VA such that certain (high power) constellation points will yield an infinite metric and will never appear. When using spherical shaping regions, for example, all points in the outer region can be assigned infinite metrics (practically this means deleting these branches from the trellis). Prior art simulation results have shown that the shaping gain is reduced by only a few hundredths of a dB when employing such PAR constraints. In view of the foregoing, it can be seen that choosing appropriate shaping regions may have a substantial influence on the shaping gain.


[0046] Since a FIR is implemented in the receiver 300 to reconstruct the shaping information bits, caution most preferably should be taken as to the initial state of the trellis (or the first few bits may be incorrect). Starting from a predefined state will result in no noticeable degradation in performance.


[0047] Other shaping codes may also be implemented to achieve a higher shaping gain, but the 4-state rate-½ code that is used in Trellis shaping, G. D. Forney, IEEE Trans. Inf., Vol. 38, March 1992, incorporated by reference herein, was found by the present inventor to provide the best performance/complexity tradeoff. It will readily be appreciated that the present invention is not so limited however, and that multi-dimensional shaping schemes may also be implemented.


[0048] In summary explanation, trellis shaping may achieve a shaping gain of almost 1 dB in addition to other schemes, and may run smoothly together with them, using a simple 4-state code. Since most of the complexity is centered in the transmitter, it is an especially appropriate method for broadcast systems. Caution must be taken as to choosing the appropriate shaping regions for the scheme employed (constellation, code, etc.). No severe error propagation is induced by the receiver when implementing trellis shaping.


[0049]
FIG. 4 is a simplified block diagram illustrating a transmitter 400 using TTCM combined with trellis precoding according to one embodiment of the present invention.


[0050]
FIG. 5 is a simplified block diagram illustrating a receiver 500 using TTCM combined with trellis precoding according to one embodiment of the present invention.


[0051]
FIG. 6 is a simplified block diagram illustrating one embodiment of a TH-precoder 600.


[0052] The method of combining trellis precoding with TTCM is similar to that described herein before for combining trellis shaping with TTCM, except that now, precoding via TH precoder 600 is performed on the output symbols of the mapper 402. The decoder 404 for the shaping code uses the output of the TH precoder 600 in order to select the desired bit sequence ys,J and the transmitted symbol sequence. A folding step 502 is required in the receiver 500 to reconstruct the symbols at the precoder 600 input (symbol sequence w), based on the product of the turbo decoder 504 (which recovers the symbol sequence w up to a modulo operation). It can be readily appreciated that some enhancements easily implemented by those skilled in the art might be needed in the TTCM decoder 504 to support this larger (“folded”) constellation. Once the symbols at the precoder 600 input are recovered, mechanisms similar to those described herein before regarding combining trellis shaping with TTCM can be used to recover the original transmitted bits. Methods for using trellis preceding with TTCM, as well as some novel schemes for using TH preceding with non-square constellations are now set forth herein below.


[0053] Trellis precoding, as stated herein before, means implementing trellis shaping together with Tomlinson-Harashima (TH) precoding. FIG. 4 illustrates a block diagram of a transmitter 400 that employs a TH precoder 600. The transmitter 400 resembles the one for trellis shaping combined with TTCM, depicted in FIG. 2; only now branch metrics are calculated according to the precoded symbols. This means implementing a separate precoder for each state in the trellis of the VA in the transmitter 400. Generally, the trellis has an infinite number of states due to the modulus operation within the TH precoder 600; but implementing a reduced complexity PDFD (parallel decision-feedback decoder) having the same number of states as the decoder for trellis shaping usually results in a small degradation in performance (although this may depend on the specific channel and precoder employed). The receiver 500 is similar to the one for trellis shaping combined with TTCM, shown in FIG. 3; only modulus operation is most preferably implemented (or using an infinite grid in the slicer) due to the TH precoding scheme. The present inventor performed trellis preceding combined with TTCM simulations using different constellations and shaping regions for two channels: H1=[1 −0.2 0.1] and H1=[1 0.4 −0.2]. The simulation results are set forth below in Table 2.
2TABLE 2TheoreticalShaping gainDate rate RConstellationand actualover nonShapingInformationwithout →shaping gainprecodedRegionsbits/symbolwith shaping[dB]*system [dB]usedRemarks38QAM →(0.75/0.59)/(0.47/0.31)/Square/Compatible16QAM(0.77/0.58)(0.49/0/30)Sphericalwith rate-1/2 code532CR →(0.88/0.70)/(0.60/0.42)/Square/Compatible64QAM(0.89/0.69)(0.61/0.41)Sphericalwith rate-1/2 code7128QAM →(0.93/0.74)/(0.74/0.54)/Square/Compatible256QAM(0.93/0.73)(0.74/0.53)Sphericalwith rate-1/2 code532CR →(**/**)/(**/**)/Square/Compatible64QAM(0.86/0.71)(0.58/0.43)Sphericalwith rate-2/3 code7128QAM →(**/**)/(**/**)/Square/Compatible256QAM(0.92/0.74)(0.73/0.54)Sphericalwith rate-2/3 code


[0054] The single * indicates that shaping gain is given for both channels tested, in the form [square regions: (shaping gain for H1)/(shaping gain for H2)/(spherical regions: (shaping gain for H1)/(shaping gain for H2)]. The double ** indicates that square region shaping was not checked, but is expected to provide similar results to that provided with spherical region shaping. It can be seen from the simulation results shown in Table 2 that using trellis preceding results in a slight degradation in performance in comparison to the no precoding case. This degradation is channel (and hence precoder) dependent due to the use of the PDFD. Using more sophisticated decoders will result in achieving shaping gains similar to the no precoding case. Still, a shaping gain of almost 1 dB may be achieved by using a 4-state rate-½ convolutional shaping code as in Trellis shaping, G. D. Forney, IEEE Trans. Inf., Vol. 38, March 1992, incorporated by reference herein. It can be appreciated that other issues such as shaping regions, decoder delay, and the like discussed herein before also apply for trellis precoding.


[0055] The complexity of trellis precoding is higher than that of trellis shaping due to the implementation of precoding for each state in the VA as stated herein before. For long precoders, the complexity may be reduced by using only a few (large valued) taps within the VA. This might result in a slight degradation in shaping gain. The PDFD used in the simulation is 4-state, equivalent to the one used for the no preceding case discussed herein before with reference to Table I. Those skilled in the art will readily appreciate that more complex decoders (such as 8 or 16-state) may result in better performance, especially for “difficult channels.” As in the non-precoder case, the receiver 500 is kept simple, and the only difference is the use of modulus arithmetic, or the use of an infinite grid in the slicer. It can also be appreciated that other issues discussed above regarding trellis shaping are applicable to trellis preceding.


[0056] When using non-square constellations, some adaptation is most preferably done to the TH precoder 600 since using the standard precoder will result in constellation expansion according to the square of minimum side that bounds all constellation points; and thus, a “negative” shaping gain. Introducing dependency between the real and imaginary parts of the signal points in the precoder 600 results in unnecessary complexity. A “rotated square constellation” such as the one illustrated in FIG. 14 (for 32 point constellations), as well as modulus arithmetic on the diagonals (i.e. real+image and real−image) are most preferably employed. In this way, trellis preceding can be implemented using both square and spherical regions. FIG. 14 illustrates a 32 point constellation with spherical shaping regions, designed for trellis preceding and that is compatible with rate-½ code.


[0057] In summary explanation, trellis preceding may be used to combine trellis shaping and precoding and can be combined with TTCM. Some performance degradation is generally caused in comparison to trellis shaping without preceding (especially for the case of difficult channels) due to the use of a reduced complexity decoder (PDFD); whereas some extra complexity is introduced in the transmitter due to the implementation of a precoder in each state of the VA. Importantly, a shaping gain of almost 1 dB may still be achieved. The low complexity of the receiver is left unchanged. Special constellations are most preferably employed for scenarios that use an odd number of transmitted bits per 2D symbol.


[0058]
FIG. 7 is a simplified block diagram illustrating a transmitter 700 using TTCM combined with shell mapping 702 to implement a binary coded modulation system according to one embodiment of the present invention. Notations are described in detail herein below. The input bit stream 704 is divided into blocks of N*b bits each, resulting in N consecutive 2D symbols. Each block is divided into three parts. The first part 706 consisting of K bits is used to select N shells using shell mapping. The second part 708 consisting of N*k bits is passed through the encoder 202 of the TTCM, resulting in N cosets. One method that can be used to select a constellation point in each of the 2D symbols includes 1) selecting the shells according to the output of the shell mapper 720, 2) selecting the coset within the shell according to the output of the TTCM encoder 202, and 3) selecting between multiple cosets in each of the shells (if there is more than one coset in a shell) using the rest of the bits 710 in the block.


[0059]
FIG. 8 is a simplified block diagram illustrating a receiver 800 using a TTCM 302 combined with shell de-mapping 802 to produce hard decisions based on the noisy symbols received, according to one embodiment of the present invention. For each block of N symbols 804, the original K bits 706 are produced by the shell de-mapper 802; whereas the rest of the bits 710 are produced by the symbols to bits mapper 810. When no errors occur, the original N*b bits are recovered.


[0060]
FIG. 9 is a simplified block diagram illustrating a transmitter 900 using TTCM combined with shell mapping 702 and Laroia precoding 902 according to one embodiment of the present invention.


[0061]
FIG. 10 is a simplified block diagram illustrating a receiver 1000 using TTCM combined with shell de-mapping 802 and Laroia preceding according to one embodiment of the present invention.


[0062]
FIG. 11 is a simplified block diagram illustrating a technique 1100 for implementing Laroia precoding according to one embodiment. The technique 900 is similar to the technique discussed herein before regarding shell mapping except that precoding 902 is now performed on the output symbols of the mapper 720 as seen in FIG. 9. In the receiver 1000, an additional step 1002 is required to reconstruct the symbols at the precoder input (symbol sequence U in FIG. 11), based on the product of the turbo decoder (which recovers the symbol sequence Y in FIG. 11). Once the symbols at the precoder input are recovered, similar mechanisms such as discussed herein before regarding shell mapping can be used to recover the original transmitted bits.


[0063] Shell mapping concepts as well as other related issues are now set forth in detail to further clarify and explain the particular embodiments presented herein above in order to provide those skilled in the arts of trellis coded modulation, turbo trellis coded modulation, constellation shaping and precoding, with the information needed to apply the novel principles and to construct and use such specialized components as are required. In 2N-dimensional shell mapping, each 2D constellation is divided into M shells. For every N symbols, the MN possible shell combinations are sorted from the lowest-power combination to the highest power combination. K bits enter a shell mapper that implements a one-to-one map between the 2k possible bit combinations and the 2k<<MN lowest power shell combinations. A precoding scheme that de-couples shaping and preceding is described in A simple and effective precoding scheme for noise whitening on intersymbol interference channels, R. Laroia, S. A. Tretter, N. Farvardin, IEEE Trans. Commun., October 1993, incorporated by reference herein. This precoding scheme may suffer from error propagation in the receiver due to the implementation of an IIR in the receiver that reconstructs the original signal (although in the presence of channel coding, this may be done after the channel decoder).


[0064] In V.34, described in ITU-T Recommendation V.34, September 1994, incorporated by reference herein, 2N=16 dimensional shell mapping and preceding are implemented, and work together with TTCM. The present inventor employed this scheme to simulate shell mapping together with TTCM or TTCM using different constellations and shaping parameters with results shown in Table 3 below; wherein the complexity of shell mapping is similar in the transmitter and the receiver, and results from the storage of some vectors used by the shell mapper, and from the actual mapping procedure.
3TABLE 3ShapingData rate RConstellationTheoreticalActualparametersinformationWithout →shapingshapingused K,M,Lbits/symbolwith shapinggain [dB]gain [dB]**Remarks38QAM →0.671.188,3,12Compatible with12QAMrate-1/2 code38QAM →0.701.218,4,16Compatible with16QAMrate-1/2 code416QAM →0.840.5616,6,24Compatible with24QAMrate-1/2 code416QAM →0.850.5616,7,28Compatible with28QAMrate-1/2 code532CR →0.880.60Compatible with48QAMrate-1/2 code532CR →0.900.6224,14,56Compatible with56QAMrate-1/2 code664QAM →0.880.82 (0.71) *24,12,96Compatible with96QAMrate-1/2 code664QAM →0.930.86 (0.75) *24,14,112Compatible with96QAMrate-1/2 code7128QAM →0.900.7124,12,192Compatible with192QAMrate-1/2 code7128QAM →0.940.7424,14,224Compatible with224QAMrate-1/2 code416QAM →0.380.1 8,3,24Compatible with24QAMrate-2/3 code532CR →0.730.4516,6,48Compatible with48QAMrate-2/3 code532CR →0.760.4816,7,56Compatible with56QAMrate-2/3 code664QAM →0.860.79 (0.68) *24,12,96Compatible with96QAMrate-2/3 code664QAM →0.900.84 (0.73) *24,14,112Compatible with96QAMrate-2/3 code7128QAM →0.910.7224, 12,192Compatible with192QAMrate-2/3 code7128QAM →0.940.7624,14,224Compatible with224QAMrate-2/3 code


[0065] The single * means that actual shaping gain was calculated relative to a square constellation. Using a trivial 2D “constellation shaping” (i.e. using a circle constellation rather than a square) yield the relative shaping gain shown in parentheses. The double ** means that K is the number of bits that enter the shell mapper for every block of eight 2D symbols (16D symbols). The letter M is the number of shells per 2D constellation and L is the enlarged 2D constellation size. The ratio L/M gives the number of 2D constellation points per shell. For rate-½ codes, this should be a multiple of 4, and for rate-⅔ codes, this should be a multiple of 8 (necessary but not sufficient requirements. For compatibility with a code based on constellation partitioning, there should be in each shell the same number of points from every subset. Looking at the results in Table 3, a 0.3 dB degradation in shaping gain can be seen when trying to transmit 4 bits/symbol compatible with a rate-⅔ code. This seems to occur due to the “large” shells in comparison to the constellation, whereas symbols within the shells are uniformly distributed.


[0066] Other related issues concern peak-to-average ratio (PAR), constellation expansion ratio (CER), other dimensionalities and multi-dimensional (MD) shaping schemes, among other things. Desired constraints on the PAR and CER are easily met by appropriately choosing the shell mapping parameters (constellation size, number of shells). On optimal shaping of multidimensional constellations, R. Laroia, N. Farvardin, S. A Tretter, IEEE Trans. Inf., Vol. 40, July 1994, incorporated by reference herein, notes that in general, for a given PAR and CER, no other shaping scheme results in a higher shaping gain and smaller delay than shell mapping. As stated above, the present inventor used a 16D shell mapping as described in the V.34 recommendation referenced herein before, but in combination with TTCM. This results in a shaping gain of almost 1 dB with a “shaping delay” of eight 2D symbols, also described herein before. The reference cited immediately above demonstrates achievable shaping gains, CER and PAR for different dimensionalities. It can be appreciated that using other dimensionalities will of course affect the algorithm complexity. Those skilled in the art will also appreciate that multi-dimensional shaping schemes may also be implemented, although this was not investigated by the present inventor.


[0067] In summation, 16D shell mapping may achieve a shaping gain of almost 1 dB in addition to other schemes, and may run smoothly together with them. It can be combined with TTCM as described above. Specifically TTCM of rate-½ and of rate-⅔ has been tested. The complexity is mainly due to storage which is similar in the transmitter and receiver, whereas the computational complexity is lower in the receiver.


[0068] Table 4 below illustrates a comparison of the trellis shaping/precoding and shell mapping schemes discussed in detail herein above.
4TABLE 4CriterionTrellis shapingShell mappingRemarksShaping gainApproaching 1dBApproaching 1dBIn Tables 1 and 3, trellisshaping is slightly superiordue to the higher CER (exceptfor 4-bits/sym rate-2/3 code)Complexity in transmitter4-state, 4-transition VARelatively high storageShell mapping superiorcomplexity, mediumcomputational complexityComplexity in receiverVery lowRelatively high storageTrellis shaping superiorcomplexity, low computationalcomplexityDelay in transmitterAbout 20 symbols for8 symbol blocksShell mapping superiorsufficient performanceDelay in receiverNo delay8 symbol blocksTrellis shaping superiorPAR and CER constraintsMay be introduced in VA,Inherent to algorithmShell mapping superioraffects complexityCombination with code (e.g.Possible (rate-1/2 and rate-2/3)Possible (rate-1/2 and rate-2/3)EquivalentTTCM)Combination with precodingTrellis precoding-higherLaroia precoding- introducesWhen error propagationcomplexity (transmitter), noerror propagationdisastrous-trellis precodingerror propagationsuperiorLaroia precoding-introduceserror propagation


[0069] Since shaping tries to reduce the transmitted energy, it results in larger use of constellation symbols with small energy, relative to symbols with high energy. It can be shown that the asymptotic distribution of the symbols for high dimensional shaping approaches Gaussian distribution. This fact slightly reduces the performance of the proposed combined scheme, especially for small constellations. This happens because the outer symbols (those with large energy) have better probability of error than the inner symbols. If they are transmitted with lower probability, the total probability of error is enlarged.


[0070] This degradation (which is usually less than 0.5 dB) can be improved by using the non-equi probable symbol distribution in the TTCM decoder. The decoder computes, in each iteration, the likelihood ratio for each data bit, which equals the ratio between the probability that this bit is ‘1’ and the probability that this bit is ‘0’. If one prepares a table of the probability distribution of the data symbols, this table can be used in the calculations of these probabilities. Such a table can be prepared by simulating the transmitter and generating a histogram for the symbol distribution over a long period of time. This method can reduce the degradation by 0.2-0.3 dB.


[0071] In view of the foregoing descriptions, it should be apparent that the present invention represents a significant departure from the prior art in construction and operation. However, while particular embodiments of the present invention have been described herein in detail, it is to be understood that various alterations, modifications and substitutions can be made therein without departing in any way from the spirit and scope of the present invention, as defined in the claims which follow. For example, although various embodiments have been presented herein with reference to particular shell mapping/de-mapping elements and associated characteristics, the present inventive structures and methods are not necessarily limited to such a particular architecture or set of characteristics as used herein.


Claims
  • 1. A method of binary coded data communication, the method comprising the steps of: providing a transmitter having a turbo trellis coded modulator (TTCM) encoder and constellation shaping elements; and generating a plurality of signal points in response to a partitioned binary coded symbol sequence that is processed via the TTCM encoder and constellation shaping elements.
  • 2. The method according to claim 1 wherein the step of generating a plurality of signal points in response to a partitioned binary coded symbol sequence that is processed via the TTCM encoder and constellation shaping elements comprises the step of generating a signal constellation having square shaping regions capable of use in association with trellis shaping that is compatible with rate kc/nc TTCM, such that a binary kc-tuple portion of a desired symbol sequence can be processed via the TTCM encoder for a rate kc/nc TTCM code to generate a nc-tuple.
  • 3. The method according to claim 1 wherein the step of generating a plurality of signal points in response to a partitioned binary coded symbol sequence that is processed via the TTCM encoder and constellation shaping elements comprises the step of generating a signal constellation having spherical shaping regions capable of use in association with trellis shaping that is compatible with rate kc/nc TTCM, such that a binary kc-tuple portion of a desired symbol sequence can be processed via the TTCM encoder for a rate kc/nc TTCM code to generate a nc-tuple.
  • 4. The method according to claim 1 wherein the step of providing a transmitter having a turbo trellis coded modulator (TTCM) encoder and constellation shaping elements comprises the step of providing trellis precoding elements and TH-precoding elements capable of use with non-square constellations.
  • 5. The method according to claim 1 wherein the step of generating a plurality of signal points in response to a partitioned binary coded symbol sequence that is processes via the TTCM encoder and constellation shaping elements comprises partitioning a signal constellation into cosets and shells in compliance with shell mapping and TTCM.
  • 6. The method of binary coded data communication according to claim 1 wherein the step of generating a plurality of signal points comprises the steps of: dividing a desired symbol sequence into a binary kc-tuple, an uncoded binary nu-tuple, and a syndrome rs-tuple; processing the kc-tuple part of the desired symbol sequence via the turbo trellis coded modulator (TTCM) encoder for a rate kc/nc TTCM code to generate a nc-tuple; processing the rs-tuple part of the desired symbol sequence via a coset representative generator for a rate ks/ns convolutional shaping code, where ks=ns−rs to generate a ns-tuple; processing the uncoded binary nu-tuple part of the desired symbol sequence, the output of the rate kc/nc TTCM encoder, and the output of the coset representative generator for a rate ks/ns convolutional shaping code via a decoder to generate a desired bit sequence; processing the desired bit sequence and the output of the coset representative generator for a rate ks/ns convolutional shaping code via a combinational element to generate a selected bit sequence; and mapping the selected bit sequence, the uncoded binary nu-tuple part of the desired symbol sequence, and the output of the rate kc/nc TTCM encoder to generate the plurality of signal points.
  • 7. The method according to claim 6 further comprising the steps of: providing a receiver having a turbo decoder and constellation shaping elements; and processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements to recover the partitioned symbol sequence.
  • 8. The method according to claim 7 wherein the step of processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements to recover the partitioned symbol sequence comprises processing the plurality of signal points via the turbo decoder using a non equi-probable symbol distribution.
  • 9. The method according to claim 7 wherein the step of processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements comprises the steps of: receiving the plurality of signal points via a transmission medium and processing the received plurality of signal points via the turbo decoder to generate estimated signal points; and processing the estimated signal points via an inverse mapper to generate an estimated binary kc-tuple part of the desired bit sequence according to the rate kc/nc TTCM code, an estimated uncoded binary nu-tuple part of the desired bit sequence, and an estimated binary rs-tuple part of the desired bit sequence according to the rate ks/ns convolutional shaping code.
  • 10. The method according to claim 7 wherein the step of processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements further comprises the step of processing the estimated binary kc-tuple part of the desired bit sequence according to the rate kc/nc TTCM code to recover k bits based on n bits and to generate an estimated binary kc-tuple part of the desired symbol sequence.
  • 11. The method according to claim 10 wherein the step of processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements further comprises the step of processing the estimated binary rs-tuple part of the desired bit sequence according to the rate ks/ns convolutional shaping code to generate an estimated syndrome rs-tuple part of the desired symbol sequence.
  • 12. The method according to claim 1 further comprising the step of processing the plurality of signal points via a trellis precoder to generate a coded symbol sequence.
  • 13. The method according to claim 12 further comprising the steps of: providing a receiver having a turbo decoder and constellation shaping elements; and processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements to recover the partitioned symbol sequence.
  • 14. The method according to claim 13 wherein the step of processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements to recover the partitioned symbol sequence comprises processing the plurality of signal points via the turbo decoder using a non equi-probable symbol distribution.
  • 15. The method according to claim 12 wherein the step of processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements comprises the steps of: folding the coded symbol sequence to generate a folded constellation; processing the folded constellation via the turbo decoder to generate estimated signal points; and processing the estimated signal points via an inverse mapper to generate an estimated binary kc-tuple part of the desired bit sequence according to the rate kc/nc TTCM code, an estimated uncoded binary nu-tuple part of the desired symbol sequence, and an estimated binary rs-tuple part of the desired bit sequence according to the rate ks/ns convolutional shaping code.
  • 16. The method according to claim 15 wherein the step of processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements further comprises the step of processing the estimated binary kc-tuple part of the desired bit sequence according to the rate kc/nc TTCM code to recover k bits based on n bits and to generate an estimated binary kc-tuple part of the desired symbol sequence.
  • 17. The method according to claim 16 wherein the step of processing the plurality of signal points via the receiver turbo decoder and the receiver constellation shaping elements further comprises the step of processing the estimated binary rs-tuple part of the desired bit sequence according to the rate ks/ns convolutional shaping code to generate an estimated syndrome rs-tuple part of the desired symbol sequence.
  • 18. The method of binary coded data communication according to claim 1 wherein the step of generating a plurality of signal points comprises the steps of: dividing a desired symbol sequence into a first part having K-bits, a second part having N*k-bits, and a third part having the remaining bits; processing the first part of the desired symbol sequence via a shell mapper to generate N shells; processing the second part of the desired symbol sequence via the TTCM encoder to generate N cosets; and mapping the third part of the desired symbol sequence, the N shells, and the N cosets to generate a number N of transmit symbols.
  • 19. The method of binary coded data communication according to claim 18 further comprising the steps of: providing a receiver having a turbo decoder and constellation shaping elements; and processing the N received symbols via the receiver turbo decoder and the receiver constellation shaping elements to recover the desired symbol sequence.
  • 20. The method according to claim 19 wherein the step of processing the N received symbols via the receiver turbo decoder and the receiver constellation shaping elements to recover the desired symbol sequence comprises processing the N received symbols via the turbo decoder using a non equi-probable symbol distribution.
  • 21. The method of binary coded data communication according to claim 19 wherein the step of processing the N received symbols comprises the steps of: decoding the N received symbols via the turbo decoder to generate N hard symbols; de-mapping the N hard symbols into the first part of the desired symbol sequence via a shell de-mapper; mapping the N hard symbols into the remaining part of the desired symbol sequence via a symbols to bits mapper; and combining the first part of the desired symbol sequence recovered by the shell de-mapper with the remaining part of the desired symbol sequence recovered by the symbols to bits mapper to recover the desired symbol sequence at the receiver.
  • 22. The method of binary coded data communication according to claim 18 further comprising the step of processing the N transmit symbols via a Laroia precoder to generate a precoded symbol sequence.
  • 23. The method of binary coded data communication according to claim 22 further comprising the steps of: providing a receiver having a turbo decoder and constellation shaping elements; and processing the coded symbol sequence via the receiver turbo decoder and the receiver constellation shaping elements to recover the desired symbol sequence.
  • 24. The method according to claim 23 wherein the step of processing the coded symbol sequence via the receiver turbo decoder and the receiver constellation shaping elements to recover the desired symbol sequence comprises processing the coded symbol sequence via the turbo decoder using a non equi-probable symbol distribution.
  • 25. The method of binary coded data communication according to claim 23 wherein the step of processing the coded symbol sequence comprises the steps of: decoding the coded symbol sequence via the turbo decoder to generate N hard symbols; processing N hard symbols to reconstruct symbols appearing at the Laroia precoder input; de-mapping the reconstructed symbols into the first part of the desired symbol sequence via a shell de-mapper; mapping the reconstructed symbols into the remaining part of the desired symbol sequence via a symbols to bits mapper; and combining the first part of the desired symbol sequence recovered by the shell de-mapper with the remaining part of the desired symbol sequence recovered by the symbols to bits mapper to recover the desired symbol sequence at the receiver.
  • 26. A binary coded data communication system comprising: a transmitter having a turbo trellis coded modulator (TTCM) encoder and constellation shaping elements; and a receiver having a turbo decoder and constellation shaping elements, wherein the transmitter is operational to generate a plurality of signal points in response to a partitioned symbol sequence that is processed via the TTCM encoder and constellation shaping elements, and further wherein the receiver is operational to receive the plurality of signal points over a transmission medium and recover the partitioned symbol sequence in response to the plurality of signal points that are received and processed via the receiver turbo decoder and the receiver constellation shaping elements.
  • 27. The binary coded data communication system according to claim 26 wherein the turbo decoder employs a non equi-probable symbol distribution.
  • 28. The binary coded data communication system according to claim 26 wherein the constellation shaping elements comprise trellis shaping elements.
  • 29. The binary coded data communication system according to claim 28 wherein the turbo trellis coded modulator (TTCM) encoder is configured to generate a rate kc/nc TTCM code in response to a kc-tuple part of a desired symbol sequence, and further wherein the transmitter constellation shaping elements comprise: a coset representative generator configured to generate a rate ks/ns convolutional shaping code, where ks=ns−rs, in response to a rs-tuple part of the desired symbol sequence; a shaping code decoder configured to generate a desired bit sequence in response to an uncoded binary nu-tuple part of the desired symbol sequence, the rate kc/nc TTCM code, and the rate ks/ns convolutional shaping code; a combinational element configured to generate a transmit symbol sequence in response to the desired bit sequence and the rate ks/ns convolutional shaping code; and a mapper configured to generate a plurality of signal points in response to the transmit symbol sequence, the uncoded binary nu-tuple part of the desired symbol sequence, and the generated rate kc/nc TTCM code.
  • 30. The binary coded data communication system according to claim 29 wherein the turbo decoder is configured to receive the plurality of signal points via a transmission medium and generate estimated signal points therefrom, and further wherein the receiver constellation shaping elements comprise an inverse mapper configured to receive and process the estimated signal points to generate an estimated binary kc-tuple part of the desired bit sequence according to the rate kc/nc TTCM code, an estimated binary nu-tuple part of the desired symbol sequence, and an estimated binary rs-tuple part of the desired bit sequence according to the rate ks/ns convolutional shaping code.
  • 31. The binary coded data communication system according to claim 30 wherein the receiver constellation shaping elements further comprise a bit recovery element configured to process the estimated binary kc-tuple part of the desired bit sequence according to the rate kc/nc TTCM code such that k bits can be recovered based on n bits to generate an estimated binary kc-tuple part of the desired symbol sequence.
  • 32. The binary coded data communication system according to claim 31 wherein the receiver constellation shaping elements further comprise a transformation element configured to process the estimated binary rs-tuple part of the desired bit sequence according to the rate ks/ns convolutional shaping code and therefrom generate an estimated syndrome rs-tuple part of the desired symbol sequence.
  • 33. The binary coded data communication system according to claim 29 wherein the transmitter further comprises a trellis precoder operational to generate a coded symbol sequence in response to the plurality of signal points.
  • 34. The binary coded data communication system according to claim 33 wherein the receiver further comprises a folding element operational to generate a folded constellation in response to the coded symbol sequence.
  • 35. The binary coded data communication system according to claim 34 wherein the turbo decoder is configured to generate estimated signal points in response to the folded constellation, and the receiver constellation shaping elements further comprise an inverse mapper configured to generate an estimated binary kc-tuple part of the desired bit sequence according to the rate kc/nc TTCM code, an estimated uncoded binary nu-tuple part of the desired symbol sequence, and an estimated binary rs-tuple part of the desired bit sequence according to the rate ks/ns convolutional shaping code in response to the estimated signal points.
  • 36. The binary coded data communication system according to claim 35 wherein the receiver constellation shaping elements further comprise a bit recovery element configured to process the estimated binary kc-tuple part of the desired bit sequence according to the rate kc/nc TTCM code such that k bits can be recovered based on n bits to generate an estimated binary kc-tuple part of the desired symbol sequence.
  • 37. The binary coded data communication system according to claim 36 wherein the receiver constellation shaping elements further comprise a transformation element configured to process the estimated binary rs-tuple part of the desired bit sequence according to the rate ks/ns convolutional shaping code and therefrom generate an estimated syndrome rs-tuple part of the desired symbol sequence.
  • 38. The binary coded data communication system according to claim 28 wherein the trellis precoder comprises a Tomlinson-Harashima precoder.
  • 39. The binary coded data communication system according to claim 26 wherein the constellation shaping elements comprise shell mapping elements.
  • 40. The binary coded data communication system according to claim 39 wherein the partitioned symbol sequence comprises a first part having K-bits, a second part having N*k-bits, and a third part having the remaining bits, and further wherein the transmitter TTCM encoder is configured to process the second part of the partitioned symbol sequence to generate N cosets.
  • 41. The binary coded data communication system according to claim 40 wherein the transmitter constellation shaping elements comprise: a shell mapper configured to process the first part of the partitioned symbol sequence to generate N shells; and a mapper configured to process the third part of the partitioned symbol sequence, the N shells, and the N cosets to generate a number N of transmit symbols.
  • 42. The binary coded data communication system according to claim 41 wherein the receiver turbo decoder is configured to process the N receive symbols to generate N hard symbols and wherein the receiver constellation shaping elements comprise: a symbols-to-bits mapper configured to map the N hard symbols into the third part of the partitioned symbol sequence; a shell de-mapper configured to de-map the N hard symbols into the first part of the partitioned symbol sequence; and a combinational element configured to combine the first part of the partitioned symbol sequence recovered by the shell de-mapper with the third part of the partitioned symbol sequence recovered by the symbols-to-bits mapper to recover the partitioned symbol sequence at the receiver.
  • 43. The binary coded data communication system according to claim 41 wherein the transmitter further comprises a Laroia precoder operational to generate a precoded symbol sequence in response to the N transmit signals.
  • 44. The binary coded data communication system according to claim 43 wherein the receiver turbo decoder is configured to process the coded symbol sequence to generate N hard symbols and wherein the receiver constellation shaping elements comprise: a symbol processor configured to process the N hard symbols to reconstruct symbols appearing at the Laroia precoder input; a shell de-mapper configured to de-map the reconstructed symbols into the first part of the desired symbol sequence; a symbols-to-bits mapper configured to map the reconstructed symbols into the remaining part of the desired symbol sequence; and a combinational element configured to combine the first part of the desired symbol sequence recovered by the shell de-mapper with the remaining part of the desired symbol sequence recovered by the symbols-to-bits mapper to recover the partitioned symbol sequence at the receiver.
  • 45. A method of binary coded data communication, the method comprising the steps of: providing a receiver having a turbo trellis coded modulator (TTCM) decoder and constellation shaping elements; and processing a plurality of signal points via the TTCM decoder using a non equiprobable symbol distribution to recover the partitioned symbol sequence.
  • 46. The method according to claim 45 wherein the step of processing a plurality of signal points via the TTCM decoder comprises the step of computing in each iteration, the likelihood ratio for each data bit, which equals the ratio between the probability a bit is a ‘1’ and the probability the bit is a ‘0’.
  • 47. The method according to claim 46 wherein the step of computing in each iteration the likelihood ratio for each data bit comprises the step of preparing a table of the probability distribution of associated data symbols.
  • 48. The method according to claim 47 wherein the step of preparing a table of the probability distribution of associated data symbols comprises the step of simulating a transmitter and generating a symbol distribution histogram over a long period of time.
RELATED PATENT APPLICATIONS

[0001] This application claims the benefit, under 35 U.S.C. §119(e)(1), of U.S. Provisional Application Serial No. 60/217,767, entitled Method And System Of Data Communication Using Constellation Shaping, Precoding And Coding, filed Jul. 12, 2000 by Lior Ophir.

Provisional Applications (1)
Number Date Country
60217767 Jul 2000 US