CA2058450C - Treillis precoding for fractional bits/baud - Google Patents

Treillis precoding for fractional bits/baud Download PDF

Info

Publication number
CA2058450C
CA2058450C CA002058450A CA2058450A CA2058450C CA 2058450 C CA2058450 C CA 2058450C CA 002058450 A CA002058450 A CA 002058450A CA 2058450 A CA2058450 A CA 2058450A CA 2058450 C CA2058450 C CA 2058450C
Authority
CA
Canada
Prior art keywords
sequence
code
trellis code
trellis
digital data
Prior art date
Legal status (The legal status 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 status listed.)
Expired - Lifetime
Application number
CA002058450A
Other languages
French (fr)
Other versions
CA2058450A1 (en
Inventor
Vedat Eyuboglu
Michael Pin-Li Chen
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
General Electric Co
Original Assignee
Codex Corp
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
Application filed by Codex Corp filed Critical Codex Corp
Publication of CA2058450A1 publication Critical patent/CA2058450A1/en
Application granted granted Critical
Publication of CA2058450C publication Critical patent/CA2058450C/en
Anticipated expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Classifications

    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/25Error detection or forward error correction by signal space coding, i.e. adding redundancy in the signal constellation, e.g. Trellis Coded Modulation [TCM]
    • H03M13/256Error 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
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/25Error detection or forward error correction by signal space coding, i.e. adding redundancy in the signal constellation, e.g. Trellis Coded Modulation [TCM]
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00Baseband systems
    • H04L25/02Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/03Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
    • H04L25/03006Arrangements for removing intersymbol interference
    • H04L25/03343Arrangements at the transmitter end
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00Baseband systems
    • H04L25/38Synchronous or start-stop systems, e.g. for Baudot code
    • H04L25/40Transmitting circuits; Receiving circuits
    • H04L25/49Transmitting circuits; Receiving circuits using code conversion at the transmitter; using predistortion; using insertion of idle bits for obtaining a desired frequency spectrum; using three or more amplitude levels ; Baseband coding techniques specific to data transmission systems
    • H04L25/497Transmitting circuits; Receiving circuits using code conversion at the transmitter; using predistortion; using insertion of idle bits for obtaining a desired frequency spectrum; using three or more amplitude levels ; Baseband coding techniques specific to data transmission systems by correlative coding, e.g. partial response coding or echo modulation coding transmitters and receivers for partial response systems
    • H04L25/4975Correlative coding using Tomlinson precoding, Harashima precoding, Trellis precoding or GPRS
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L27/00Modulated-carrier systems
    • H04L27/32Carrier systems characterised by combinations of two or more of the types covered by groups H04L27/02, H04L27/10, H04L27/18 or H04L27/26
    • H04L27/34Amplitude- and phase-modulated carrier systems, e.g. quadrature-amplitude modulated carrier systems
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L27/00Modulated-carrier systems
    • H04L27/32Carrier systems characterised by combinations of two or more of the types covered by groups H04L27/02, H04L27/10, H04L27/18 or H04L27/26
    • H04L27/34Amplitude- and phase-modulated carrier systems, e.g. quadrature-amplitude modulated carrier systems
    • H04L27/3405Modifications of the signal space to increase the efficiency of transmission, e.g. reduction of the bit error rate, bandwidth, or average power
    • H04L27/3416Modifications of the signal space to increase the efficiency of transmission, e.g. reduction of the bit error rate, bandwidth, or average power in which the information is carried by both the individual signal points and the subset to which the individual points belong, e.g. using coset coding, lattice coding, or related schemes
    • H04L27/3422Modifications of the signal space to increase the efficiency of transmission, e.g. reduction of the bit error rate, bandwidth, or average power in which the information is carried by both the individual signal points and the subset to which the individual points belong, e.g. using coset coding, lattice coding, or related schemes in which the constellation is not the n - fold Cartesian product of a single underlying two-dimensional constellation
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00Baseband systems
    • H04L25/02Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/03Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
    • H04L25/03006Arrangements for removing intersymbol interference
    • H04L2025/03777Arrangements for removing intersymbol interference characterised by the signalling
    • H04L2025/03802Signalling on the reverse channel
    • H04L2025/03808Transmission of equaliser coefficients

Abstract

A digital data sequence is mapped into a signal point sequence e(D) for data transmission over a channel h(D) (40) to produce a channel output sequence y(D)=e(D)h(D), e(D) is chosen so that the signal points in the sequence y(D) belong to a class of possible sequences based on the digital data sequence, the class being based on a time-varying treillis code derived from an N-dimensional time-invariant treillis code (44) by using different transformed versions of its underlying lattices for respectively different N-dimensional symbols.

Description

2f~~i~~;'r~
WO 91/15900 _ 1 - PCT/US91/02384 Trellis Precoding for Fractional Bits/Haud Hackqround of the Invention This invention relates to modulation systems for sending digital data via a channel.
In a quadrature amplitude modulation (QAM) system, for example, two key parameters are the baud rate 1/T and the number of bits per baud f3. The bit rate R of the QAM system is given by R=f3/T, measured in bits per second (b/s).
Transmission equipment, such as a voiceband modem, operates at one of a number of possible bit rates (e.g., in the range 9.6 to 19.2 kb/s) depending on the line quality or the user's requirement. Therefore, for a given baud rate such a modem must be able to operate at different values of f3, which may be a fraction, and the different values of f3 may not differ simply by integers. For example, if the baud rate is 2743 symbols/s, then to send at conventional modem rates of 19.2, 16.8, and 14.4 kb/s requires (3=7, 6.125 and 5.25, respectively.
When t3 is a rational number of the form f3=n+d/2m, where n and 0<d<2m are integers, a higher-dimensional (higher than two-dimensional) QAM constellation.of dimension L=2m+1 with 2BL/2 points can be used to represent t3L/2 bits over L
dimensions. Higher-dimensional constellations save in average power when their signal points are chosen to lie within hyperspheres instead of hypercubes. .This saving, called shaping gain, can be as high as 1.53 dH.
Because the size of the signal constellation increases rapidly with n and m, mapping from bits to signal points can be cumbersome. For example, in an uncoded system (one in which there is no interdependence between selected signal points), to send 7.5 bits/baud, a four-dimensional constellation with 215=32,768 points is needed, which may be too large to -:
handle. However, such a 4D constellation can be formed by the o Cartesian product of simple 128-point and 256-point WO 91/15900 ~ ~ ~ ~ ~~ ~ ._ - 2 - PCT/US91/023$4 two-dimensional constellations. More generally, a simple way of sending f3=n+d/2m bits per two-dimensions is to define a series of frames each encompassing the bits in a group of 2m bands and then for each frame send n bits/baud for (2m-d) bands and send (n+1) bits/baud for d bands. The average power per two-dimensions for this scheme is approximately P =Pn(1+d/2m), where Pn is the average power of a ' 2~-point two-dimensional (2D) QAM constellation (The 2D
constellations are scaled to have the same minimum distance between adjacent points.). This approach is approximately as efficient as a conventional QAM system sending an integer number of bits/baud. However, the imbalance in size between the two 2D constituent constellations creates a large peak-to-average ratio (PAR) in two-dimensions, making the system susceptible to other impairments such as nonlinear distortion or phase fitter.
Generalized cross-ccnstellations, described by i~rney and Wei, "Multidimensional Constellations - Part I:
Introduction, Figures of Merit, and Generalized Cross Constellations," IEEE JSAC, pp 877-892, August, 1989, offer an alternative way of generating higher-dimensional constellations to represent fractional bits/baud. Here, the higher-dimensional constellation is a, subset of the signal points in a Cartesian product of ider~tical expanded two-dimensional constellations; simple coding rules are used to construct the subset. A generalized cross-constellation has a low 2D constellation expansion ratio (CER) and a low PAR in two-dimensions; however, it also generally has a low shaping gain.
Higher-dimensional constellations with simplified bit mapping and with significant shaping gain can be constructed using the Voronoi region of a lattice n' as~the boundary of ' the signal point set, Forney, "Multidimensional Constellations - Part II: Voronoi Constellations," IEEE JSAC, pp 877-892, August, 1989. If signal points are chosen from some lattice A, then A and A' must be scaled such that the Voronoi region of A' contains 2~L~' point:s from A; that means the order ~A/A'~ of the lattice partition A/A' must be 2pL~'. The mapping of bits to sign<~.l points involves a decoding operation whose complexity is that of a minimum-distance decoder for A'. The inverse mapping is accomplished by a simple hard-decision decoding operation. A Voronoi constellation generally is more complex to generate than a generalized cross constE~:Liation. For given lattices A and A', the Vorono:i constellation can typically support only certain fractional value:a of (3, and that by using different versions of A' .
An important expansion of the Voronoi constellation technique is trellis shaping. In trellis shaping, the signal. con:~tellation is constructed on a sequence basis rather than on a block basis. Tnstead of the Voronoi region of a. lattice, the decision regions associated with trellis codes are used to form the signal constellation boundary. In analogy to Voronoi constellations, i.n trellis shaping the mapping from hits to aignal points involves a decoder for the trellis code. Trellis shaping offers a better performance/compl.exi.ty trade--o.ff than Vorono:i constellations; but, it also can support only certain fractic>nal values of (3.
An extension c>f: Voronoi constellations and trellis shaping to non-ideal channels with linear distortion or correlated noise has beerG disclosed in United States patent number 5,048,054, issued on September 10, 1991, assigned to the same assignee as this invention. This extension, called lattice or trellis precod.ing, pro~rides the performance of ideal decision feedback e~.-~ualization (DFE), preserves the coding gain of known coded modulation schemes designed for ideal channels,, and, in addition, provides shaping gain.
Trellis precod_ing is similar to trellis shaping, however, it involves a decoder for a filtered trellis code rather than for the ordinary trellis code. Tre1_lis precoding is a practically important scheme, but, like trellis shaping, is limited in the fractional values of (i that it can support.
Summary of the Invention The invention cam support values of (3 (bits per baud) that differ by fractions without significantly modifying the underlying trellis codes used in the previously disclosed trellis precoding scheme. This is achieved by using a shaping t~rel.~.is code Cg which is derived from some conventional trellis code C (A//1' , C) b~~ using different versions of the underlying lattices A and A' at different times, in a periodic fashion. For example, if C
is an N-dimensional tre~ll.is code, to send (N/2) (3==n+d/2m) (N/2) bits/N~-dimensions, where Osd<2m, we define a frame (period) of length L/2=: (N/2) 2m bands, and use the code ~~ for (N/2) (2m-d) bands, and its sca_Led ver~~ion RC for (N/2)d bands..
In general, in one aspect, the invention features mapping a digital data sequence into a signal point sequence e(D) for data transmission over a channel h(D) to produce a channel output: sequence y (D) =a (D) h (D) . a (D) is chosen so that the signal points in the sequence y(D) belong to a class of possible sequences based on the digital data sequence, the class being based on a time-varying trellis 4a code derived from an N-dimensiona~ time-invariant trellis code by using different tr°ansformed versions of its underlying lattices for respectively different N-dimensional symbols.

WO 91 / 15900 - S _ PCT/ US91 /02384 Preferred embodiments include the following features.
In some embodiments, the class comprises a signal space translate of a time-varying trellis code Cs, where the translation is a code sequence from a translate of a trellis code C_c determined based on the digital data sequence, and the time-varying trellis code Cs is derived from an N-dimensional time-invariant trellis code C(.1/A', Cs) by using different transformed versions of its underlying lattices A and A' for respectively different N-dimensional signal points.
In some embodiments, the class comprises a signal space-translate of a trellis code Cs, where the translation is a code sequence.from a translate of a time--varying trellis code Cc determined based on the digital data sequence and the time-varying trellis code C_c is derived from an N-dimensional time-invariant trellis code _C(L/L', Cc) by using different transformed versions of its underlying lattices L and L' for respectively different N-dimensional signal points.
In some embodiments, the class comprises a signal space translate of a label translate of a time-varying trellis 2p code Cs, where the label translate is based on a portion of the digital data sequence and the signal space translation is based on another portion of the elements of the digital data --sequence, and the time-varying trellis code Cs is derived from an N-dimensional time-invariant trellis code C(A/A', Cs) by using different transformed versions of its underlying lattices A and A' for respectively different N-dimensional signal points.
In some embodiments, the class comprises a signal space translate of a label translate of a trellis code Cs, where the label translate is based on a portion of the digital ° data sequence, where the signal space translation is based on another portion of the elements of the digital data sequence 2~5~~
w~ 9/15900 - 6 - PCT/US9i/02384 and is a code sea_uence from a time-varying trellis code Cc which is derived from an N-dimensional time-invariant trellis code C(L/L', Cc) by using different transformed versions of its underlying lattices L and L' for respectively different N-dimensional signal points.
In some embodiments, the class of possible sequences is identified by some sequence of subregions into which N-space has been partitioned, the sequence of subregions being specified by a label sequence which belongs to a coset of a l0 convolutional code Cs and the coset is determined based on a portion of the digital data sequence and the signal points in the subregions belonging to a coset of a time-varying trellis code Cc, where the sequences from the trellis code Cc are chosen based on another portion of the elements of the digital data sequence, and the time-varying trellis code Cc is derived from an N-dimensional time-invariant trellis code C(L/L', Cc) by using different transformed versions of its underlying lattices L and L' for respectively different N-dimensional signal points.
In some embodiments, the class, of possible sequences is identified by some sequence of subregions into which N-space has been partitioned, the sequence of subregions being specified by a label sequence which belongs to a coset of a ~~
convolutional code Cs, the signal points in the subregions belonging to a coset of a trellis code Cc, where sequences from the trellis code CC are chosen based on another portion of the elements of the digital data sequence and using w different approximately scaled versions of the subregions for respectively different N-dimensional signal points.
In general, in another aspect, the invention features choosing e(D) so that-y(D) belongs to a signal space translate of a label translate of a trellis code Cs, the label '~
translate being based on a.portion of the digital data 2~~~~r~~;
WO 91/15900 _ 7 _ PCf/US91/023&1 seo_~uence, and the signal space translation is specified based on another portion of the elements of the digital data sequence and is a code sequence from a trellis code Cc.
In general, in another aspect, the invention features choosing x(D) so that the signal points in the sequence y(D) belong to some sequence of subregions into which. N-space has been partitioned, the sequence of subregions being specified by a label sequence which belongs to a coset of a convolutional code Cs, the.coset being determined based on a portion of the digital data sequence, the signal points in the subregions belonging to a coset of a trellis code Cc, and where sequences from the trellis code Cc are chosen based on another portion of the elements of the digital data secx_uence.
In preferred embodiments, the versions axe respectively used periodically (e. g., two different versions used alternatingly) wherein each period encompasses multiple N-dimensional symbols. The transformed versions comprise rotated and/or scaled versions of the underlying~'.attices. The trellis code Cs is filtered to. form a filtered trellis code _CS' whose sequences are cs'(D) = cs(D)Q(D) where g(D) is the formal inverse of a response related to the channel response h(D), and cs(D) is any code sequence in the trellis ..
code Cs. The signal point sequence selection tends to minimize the average power of the signal point sequence e(D) _ y(D)g(D). The signal point sequence e(D) lies in a fundamental region of the filtered trellis code Cs'. The signal point sequence is selected based on reduced state sequence estimation with respect to the filtered trellis code Cs'. The reduced ' state sequence estimation uses no more states than the number of states of the original trellis code Cs'. The fundamental region of the filtered trellis code.comprises approximately a Voronoi region of the filtered trellis code Cs': The. digital data.sequence is recovered from a possibly noise-corrupted v~ i) ~ Ic) WO 91/15900 - g - PCT/US91/02384 version of the signal point secruence, including decoding the signal point sequence to a secruence of estimated digital elements and forming a syndrome of fewer digital elements based on a portion of the estimated digital elements using a feedback-free syndrome former HT. ' The signal point sequence is selected by mapping the digital data sequence into an initial sequence belonging to and representing a congruence class of the original trellis code Cs, and choosing a signal point sequence belonging to and l0 representing a congruence class of the filtered trellis code Cs' and which has no greater average power than the initial sequence, and wherein the mapping includes applying a portion of the elements of the digital data sequence to a coset representative generator for forming a larger number of digital elements representing a coset representative sequence. The coset representative generator comprises a multiplication of a portion of the elements of the digital data sequence by a coset representative generator matrix (H 1)T which is inverse to a syndrome-former matrix HT for the code.
The trellis code Cs may be a linear trellis code or a non-linear trellis code. The linear trellis code -CS is a 4-state Ungerboeck code. The trellis code Cs is based on a partition of binary lattices, or a partition of ternary or quaternary lattices.
The step of selecting the signal point sequence is further constrained so as to reduce.the peak power of the -.signal point sequence where the peak power represents the maximum energy of the signal point sequence in some number of dimensions N. In some embodiments, N = 2 or N = 4. The step of mapping the digital data sequence into the signal point sequence is arranged to ensure that the digital data sequence can:be recovered from a channel-affected version of the signal point sequence which has been subjected to one of .a number of WO 91/15900 _ g _ PCT/US91/02384 predetermined phase rotations.
Other advantages and features will become apparent from the following description of the preferred embodiment and from the claims.
Description of the Preferred Embodiment We first briefly describe the drawings.
' Fig. 1 is a block diagram of trellis precoding.
Fig. 2 is a framing scheme.
Fig. 3 is a block diagram of an alternatively sealed trellis code transmitter.
Fig. 4 is a trellis diagram.
Fig. 5 is a reduced state trellis.
Fig. 6 is a diagram of the steps of a Viterbi algorithm.
Fig. 7 is a trellis for a Wei code.
Fig. 8 is a general block diagram of a transmitter and a receiver used in precoding.
Fig. 9 is a block diagram of a transmitter.
Fig. 10 is a block diagram of a receiver.
Fig. 11 is a block diagram of binary encoding for trellis precoding.
Fig. 12 is a block diagram of a Wei binary convolutional encoder.
Figs. 13A, 13B are diagrams~of quadrants of signal sets .
Figs. 14A, 14B are diagrams of quadrant shifting.
Fig. 15 is a diagram of portions of the trellis precoder of Fig. 9.
Fig. 16 is a block diagram~of a binary syndrome former.
Fig. 17 is a block diagram of the binary portion of the trellis decoder. ~ w .
' Fig. 18 is a block diagram of a binary inverse syndrome former. -- 1~ - ' Fig. 19 is a block diagram of a constellation dormer.
Fig. 20 is a diagram of a riuadrant labeling scheme.
Fig. 21 is a diagram of a phase-invariant labeling scheme.
Fig. 22 is a diagram of an encoder for trellis precoding based on label-translates of trellis codes.
Fig. 23 is a diagram of a decoder for trellis precoding based on label-translates.
Fig. 24 is a. diagram of an encoder for trellis precoding based on regions.
Fig. 25 is a diagram of a decoder for trellis precoding based on regions.
Fig. 26 is a diagram of regions derived from a 4-way partition.
Trellis Precodinq First we will summarize the principles of trellis precoding. Certain terminology and principles which underlie the invention are set forth in Agpendix A.
Referring to Fig. 1, trellis precoding is an encoding technique based on a discrete-time channel 40 operating at a baud rate 1/T. (If the actual channel is continuous-time, or discrete-time operating at a higher sampling rate, the discrete-time channel at the baud rate can be constructed by.
appropriate filtering, sampling or sampling-rate conversion operations, as described in the trellis precoding patent cited above).
We assume the following input-output relationship for the channel:
r(D) - e(D)h(D) + n(D) where e(D) is the channel input, h(D) is a causal, stable channel response and nCD) is a noise sewence that can be modeled approximately as Gaussian and white.
Let Cs(As/~.s';Cs) be a time invariant 72987-E>

trellis code based on N-dimensional lattices AS and AS' and a rate-k;;/ (k9+rs) binary convolutional code CS. We assume that CS has the property that: all possible symbols assigned to trellis branches leaving any state sk of the encoder belong to a coset Aos + a (sk) of some lat=tice Ass, call.ed the time zero lattice of CS.
The following theorem, which is proven in the Appendix, plays a central role in trellis preceding:
Theorem 1. Let R~ be any fundamental region of the N-dimensional time-zero lattice /10s of CS, and let x (D) E (Re) °°
and cs (D) sCs be unknown sequences whose sum is y (D) - x (D) +
cs (D) . Then x (D) and c;; (D) can be uniquely determined from Y (D) -This important: theorem implies that if x(D) is an information-bearing signal that lies in the region (Ro)~, then instead of transmitting x(D), one may transmit e(D) -[x (D) +cs (D) ] g (D) , where g (D) - 1./h (D) , and still be able to recover x(D) at the receiver, provided that cs(D)sCs. Any desired criterion may be used to select the code Cs and the code sequence c5 (D) . F':i.rst, we consider strictly an average power criterion, in which case the code CS and the code sequence CS(D) are chosen to minimize the average power of the transmitted sequence e(D) (the mean-squared 'error').
In Fig. 1, t:he first step in trellis preceding is to map the incoming bit stream 42 into the information-bearing signal x(D) usi.ng a trellis code 44 to map to wards and simple mapping 46 tc map from words to x(D). For simplicity, the fundamental region R~ in N-space can be lla chosen as the Cartesian product of simple two-dimensional rectangular regions.
To obtain coding gain we choose x(D) from a coset of some trellis code CS iA~/A~' , C~.;? . W_i.thout much loss of generality, we assume that A~ and A~' are both ~f~'3~'~~;~ ti WO 91/15900 - 12 - PCf/US91/02384 N-dimensional binary lattices and that Cc is based on a binary rate-kc/tkc+rc) convolutional encoder. It is essential that Cs be a subcode of Cc, which is assured if As is a sublattice of Ac'. This implies that if cs(D) is a sequence from'Cs and x(D) is a sequence from Cc, then cs(D) + x(D) is also a sequence from Cc.
To send f3 bits per two dimensions, we will initially require that the fundamental volume of nOs be 2N~/2+rc times larger than the fundamental volume of .~c, so that l0 2N3/2+rc points from Ac can be placed in RO (to accommodate the Nf3/2 information bits and rc redundant bits);
i.e., A /~ =2N3/2+rc. Now.
c Os~ ~ given Cc and Cs, we can determine the values of 1i that trellis preceding can support. First, we write the partition ,1c/~los more explicitly in terms of the partition chain ~c/AOc/~s/'~Os' We have ~Ac/~Oc~=2rc ~AOc/Ac'~=2kc . ' ~As/AOs~=2rs.
Furthermore, if we define ~Ac'/As~=2~, where j is some integer, then it follows that log2~~c/~Os~=rc+kc+rs+j, and therefore since also log2 ~Ac/AOs~ = Nf3/2+rc,Nfi/2=kc+rs+j;
that means trellis preceding can support f3=(2/N)(kc+rs+j) bits per two dimensions (i.e., per baud in a QAM system).
Suppose we replace the code Cs by its scaled and rotated version RCs that~is based on the partition R.~ls/RAs', where.R is the N-dimensional rotation operator defined in G.D. Forney, Jr. "Coset codes--Part 1:
Introduction and Geometrical Classification," IEEE Inform. Vol.
Theory IT-39, pp. 1123-1151, Sept. 1, 1988. Then I~c./~S'=2~~N/2 and hence f3=(2/N)(kc+rs+j)+1. ' This illustrates that for binary lattices, trellis preceding can be - 1~ -scaled to support integer increments in 3, by simply replacing the code Cs by its rotated, scaled versions. A similar effect can be achieved by replacing the code Cc by its scaled and rotated version RCc.
Having discussed the scaling and the mapping of bits into the information bearing sequence x(D), we rlow proceed to describe the remaining operations in trellis precoding.
Referring again to Fig, 1, once x(D) is determined, the trellis precoder 48 'tries to find a code sequence cs(D)eCs such that the sequence e(D)=[x(D)-cs(D))g(D) has as small an average power as possible. The "error sec_ruence' e(D) is then transmitted.
The received sequence is giver. by r(D) - e(D)h(D)+n(D) - y(D)~n(D) - x(D)-cs(D)+n(D?.
Since x(D) is chosen from a translate CC+a(D) of Cc, y(D)=x(D)-cs(D) is also a code sequence in C~,+a(D).
Therefore, a conventional decoder 50 fox Cc can estimate y(D) with essentially the same error probability as if x(D) were detected with an ideal DFE which cancels the ISI due to the tail of the impulse response. If the decoder can correctly, recover y(D)=x(D)-cs(D), then it follows from uheorem l that x(D) can be correctly extracted using the simple hard-decision decoder 52 correspanding to the region (RO)'~. In order to prevent an error in yk from triggering error propagation, the syndrome-former methods, disclosed in the trellis precoding patent oited above, may be used. The original bit stream 42 may then be recovered by inverse mapping 54.
Fractional bits per baud As demonstrated above, using trellis codes based on binary lattices, trellis precoding can support values of f~, the transmitted bits/baud, that differ by integers. In the invention, trellis precoding can also support fractional differences in (3 without significantly changing the shaping code CS .
This is achieved by deriving the shaping code CS
from some time-invariara.t code C by using different (rotated, scaled) versions of its underlying lattices, A and A', at different times in a periodic fashion. (Again, a similar effect can be achieved by deriving the coding code C~ from some time invariant code C(L/L', Cc) by using a different rotated, scaled version of its underlying Lattices L and L' ) . As we saw above, i.f C (A/A' , C) can support k~+rs+j bits/N-dimensions, then it can be made to support k~+rs+ j + (N/2 ) bits/N-dimensions, by replacing ~'1 and A' by RA
and RA' , respectively. 'Io send (N/2) ~i=k~+rs+j + (d/2m) (N/2) bits/N-dimensions, where 0<_d<2"', we define a frame (period) of length L/2 - (N/2)2'° bands, and use the code C for (N/2) (2'°-d) bands, and its scaled version RC for (N/2) d bands"
Using RC and C at different times will create an imbalance in :?D conste:.L:Lation sizes and result in a PAR that is higher than that of t:.rellis precoding with time-invariant trellis codes. However, using constellation constraints, this :imbalance wi:l1 be substantially eliminated.
In the preferred embodiment, we use the 4D 16-state Wei code (described in the trellis precoding patent) as the code f'~r fundamental coding gain, and we use the 2D
4=state Ungerboeck code (described in the trellis precoding patent) as the code for shaping gain. With this combination N=4, k~=3, and rs=:2, arad we can :.end at rates n+U.5 bits,/baud 14a using time-invariant sca~.ing, where n>2 is an integer. To transmit at 19.2 kb/s using a baud rate of 1/T=2743 symbo:ls/s, however, we need to send 7 bits/baud.

~~~~J~
WO 91/15900 - 15 - >'CT/LS91/02384 This may be done as follows.
The 4D 16-state Wei code Cc is based on a rate-3/4 convolutional encoder and the four-dimensional partition Z4/2Z4. This code adds one redundant bit every four dimensions (N=4, rc=1). On the other hand, the code Cs is derived from the mod-2 trellis code C_ which is based on a rate-1/2 binary convolutional encoder and the two-dimensional partition Z2/2Z2. Equivalently, C can be described by a rate-2/4 convolutional encoder and the four-dimensional partition Z4/2Z4(N=4, ks=2). In this form, C has the 4D time-zero lattice RZ4. To send 6.5 bits per symbol using these codes, we would pick Cs = 8C such that it is based on the partition 8Z4/16Z4, with time-zero lattice 8RZ4. Then ~Ac/:10s~ ~Z4/BRZ~~=214=22x6.5+1=22fi+rc, On the other hand, to send 7.5 bits/symbol, we would pick Cs - 8 RC_ such that it is based on the partition 8RZ4/16RZ4, with time-zero lattice 1624. Then ~Ac/AOs~=~Z4/16Z4~=22x7.5+1=216, Thus, referring to Fig. 2, we can send an average of 7 bits/symbol (i.e., 7 bits/baud) by using a frame of 4 bauds or L = 2x4=8 dimensions, and by sending 13 and 15 bits in alternate 4D blacks using respectively the 8Z4/16Z4 and the 8RZ4/16RZ4 based codes for shaping. Referring to Fig. 3, the alternating of codes is achieved by including in the transmitter a framer-blocker 60 which assembles the bits into frames and organizes the frames into blocks 62.
Every two bauds, the coding code Cc-will output the two symbols (x2k and x2k+1) and these would be processed by the precoder which will operate on a 2D trellis representation gp of the trellis code _Cs. Referring to Fig. 4, for the first two bauds of the frame, this trellis corresponds to. the trellis ' of the code 8C, whereas for the next two bauds, it corresponds to the trellis of the code BRC. The~decoder will..search through this time-varying trellis to find a sequence cs(D).
Filtered trellis codes The time varying trellis code discussed in the preceding section is subjected to a fitering process to form a filtered trellis code in the game manner as described in the trellis precoding patent cited above. This is accomplished in the shaping/precoding operation 48 in Fig. 1.
More specifically, we seek a cs'(D) that minimizes ~~e(D) ~~2 - ~~x' (D) - cs' (D) ~~2, where cs ' (D) C 58 in Fig. 1) is any sequence of the form cs(D)g(D), cs(D)eCs. We call the set of all such cs'(D) the filtered trellis code Cs' - Csg. To implement trellis precoding, we need a minimum mean-squared error (MMSE) decoder for Cs' .
Even if Cs can be represented by a finite-state trellis, however, Cs' in general cannot. Indeed, if sk is the state of an encoder that generates a code sequence cs(D)e Cs at time k, then the state of the sequence cs'(D) - cs(D)g(D) at time k is sk' - (sk;pkl, where pk - ~cs,k-1' cs,k-2' w J represents the state of the 'channel' whose response is g(D). Tre number of possible values of pk is not finite, ever. if g(D) has finite degree.
because the set of passable code symbols cs,k-i' 1=1,2, ...
is a coset of a lattice and thus hasv nfinitely many elements.
Therefore' in general, an I~iSE decoder for Cs' cannot be realized by a trellis search.
Reduced-state decoders for filtered two-dimensional trellis codes We now show how to construct a class of near-optimum 'reduced-state' trellis decoders for Cs', using reduced-state sequence estimation (RSSE) principles.
In this section, we shall assume that Cs is a two-dimensional trellis code derived from a scaled, rotated ~~'.~~~~~~;
WO 91/15900 _ 17 _ PCT/L'S91/02384 version of some 2D code C_ which is based on the partition A/A' - Z2/Rks+rsZ2; these codes can be represented by trellises that have 2ks transitions per branch where each branch is associated with a unique coset of a scaled, rotated version of Rks+rsZ2. In what follows, we will denote the partition used at time k as As,k/A~s,k ' k = 1,..., L/2. In the next section, we will show how the method may be extended to higher-dimensional trellis codes. It will be apparent that two-dimensional codes l0 are the most natural ones to use when g(D) is a complex-valued response.
' Two relevant principles of reduced-state sequence ' estimation are, first, to keep track only of the last K code symbols cs,k-i'1_i<K, even when the filter response g(D) has degree greater than K, and second, to keep track of cs,k-i only with respect to its membership in one of ~As,k-i/Ak(1)~ - 2~i = Ji cosets Ak(i)+a(cs,k-i) of some lattice Ak(i). The sequence of lattices' Ak(i) must satisfy the property that ji' i=1'...'K' is a non-increasing sequence. so' as a code symbol cs'k-i becomes less recent, the information that ~is kept about it becomes coarser and coarser, until it is forgotten altogether (at i = K).
A code symbol cs,k-i' 1<i<K; in As,k-i belongs 2~ to one of Ji cosets of Ak(i). we denote this coset by a(cs~k-i)' Then, the 'coset state' of a code sequence c's(D) = cs(D)g(D) at time k can be defined as tk =
La(cs,k-1)~ ..., a(cs,k-K))' ' Next we concatenate coset states tk with encoder s;~ states sk to obtain 'super-states' vk:
~k = Lsk;tkl - Lsk:a(cs,k-1)' ...'a(cs,k-K)]' Note that given the current state vk and the code-symboh zt~~~~.~s ~

cs,k, the next state vk+1 is unicruely determined. The uniqueness is guaranteed by the condition that Ji is a non-increasing sequence.
Referring to Fig. 5, the super-states define a reduced-state trellis, which we sometimes call a ' super-trellis. In this trellis, signal points associated with all branches 27 leaving a super-state 25 belong to a coset of the time-zero lattice AsO,k' If As, k' is a sublattice of '1k(1), then as in the ordinary trellis there are 2ks 1U branches leaving any super-state, each associated with one of the 2ks cosets of As,k' whose union is the given coset of AsO.k' If Ak(1) is a sublattice of As,k', on the other hand, then there is a branch for each of the 2~1 rs cosets of Ak(1) whose union is the given coset of -~so,k' l5 The number of possible values that vk can take is finite. In fact, if Ak(i) is a sublattice of '~s,k~(~i~ks+rs) for all i < K, and Ak(K) is a sublattice of AsO,k (~K~rs)~ then the total number of super-states is given by 20 S' - S nl<i<K 2J1 rs°
Having described the reduced-state trellis, we now show how the Viterbi algorithm is used to search these trellises. Let g(D) = gK(D) + g~(D), where gK(D) = 1 +
g1D + ... + gKDK is the polynomial corresponding to the 2S first K+1 terms of g(D), and g~(D) is a residual component of possibly infinite span. Let g~(D) = gN(D)/gD(D), where gN(D) and gD(D) are both finite-order polynomials.
Then gN(D) has the same delay as g~(D), whose delay is at least K+1.
3o Suppose that the input to the reduced state decoder is x'(Dy. Referring to Fig. 6, let us define the Euclidean distance path metric of a code sequence cs'(D) at a given time k as the cumulative sum ~i~~'~c~
WO 91/15900 _ 19 _ PCf/US91/02384 rk Lj<kYj of the branch metrics Yj,j<k, computed (31) according to Y,(D>=IIx' (D>-cs' (D) 112 =IIx (D)-cs(D)gK(D)-w(D)II2 where w(D) = cs(D)g~(D), which means that w(D) satisfies w(D) =-w(D)LgD(D)-1]+cs(D)gN(D), with wk=0, for k<0. Note that fgK(D) - 1], gN(D), and fl - gD(D)] all have a delay of at least 1; the Viterbi algorithm (VA) can proceed in a recursive manner 37 by 1u comparing metrics (33) of merging paths and retaining 35 the surviving sequences with minimum path metrics. Therefore, the VA needs to store the wk's for every surviving path. It should be noted that even though only K terms of the filter response g(D) are taken into account in the trellis 15 construction, its entire memory is included in the branch metric computations.
Among these reduced-state decoders, there is one that particularly stands out in terms of its tradeoff of performance against complexity. This is the special case of parallel 7o decision-feedback decoding (PDFD), where the reduced-state trellis is simply the trellis of the original code Cs. This is obtained by setting K=0. The PDFD decoder is the simplest in this class, and its performance is. the poorest, although it often performs close to the optimum decoder.
A practical VA has a finite decoding delay M. We can assume that it effectively makes a decision on the kth symbol cs~k based on observations xk, xk+1' " " xk+M-' .
assuming the correct node at time k is vk. (Under this assumption the VA will always select a true code sequence.) ::c: For M = 0, such a decoder becomes a simple decision-feedback decoder (DFD),,.where in every symbol interval~'the decoder performs filtering using past decisions, and-operates like a hard-decision decoder by selecting the closest signal point Z~~~iv ~i WO 91 / 15900 - 2 0 - PCT/ L~S91 /02384 ck in an, appropriate cosec of :~s0,~ using a MMSE decoder for ~s0,k' Clearly in this case, regardless of the channel response, the error sequence will always lie in ~RV('1s0,k))~' where RV(ns0,k) is the Voronoi region of the time-zero lattice .~SO,k' and the average ' power of these regions will determine the average power of the transmitted symbols. Since AsO,k may be time-varying, the ' size of the signal constellation boundary may be time-varying also. This may translate into a higher peak-to-average ratio.
However, this will be controlled by putting constraints into the decoder as will be explained later.
The error sequence e(D) = x(D) - cs(D) will lie in some decision region of the RSSE decoder for the filtered trellis code Cs'. The average power of this region will essentially determine the shaping gain that can be achieved.
This region, and thus the shaping gain will generally depend on the interaction between the shaping code Cs and the filter g(D), as we have seen with MMSE decoders.
In practice, we can measure the shaping gain of an error region as follows. We take an input sequence x'(D), which is uniformly distributed in some simple fundamental region of Cs'. We decode it and obtain an error sequence e(D). This sequence will be uniformly distributed in the common error region of the decoder. .The output mean-squared error (MSE) gives the shaping gain. For M = 0, we get the shaping gain of the regions Rv(~s0,k)' k = 1,..., L/2.
As M is increased towards infinity, the error region will change and the output MSE will monotonically decrease towards a limit which depends on the reduced-state trellis being used.
Reduced-state Decoders for Filtered Multidimensional Trellis Codes .
We now extend_these~reduced-state decoding methods to higher-dimensional (N >.2.)_.trellis codes Cs, such as the dual ' - 21 -Wei codes. The dual wei codes are duals of the wei codes of the kind disclcsed in Wei, "Trellis-Coded Modulation with Multidimensional Constellations." IEEE Trans. Inform. Theory, Vol. IT-33, pp. 483-501, 1987-These codes are 'mod 2' trellis codes, which can always be regarded as being based upon the lattice partition ~s~~s~-ZN/2ZN~ and on some rate-k/N binary convolutional code _C that selects cosecs of 2ZN in ZN.
These cosecs can be specified by N/2 cosets of 2Z2 in Z2 . ' It follows that the N-dimensional trellis code can be represented by a two-dimensional trellis, which is periodically time-varying with period N/2. At times nN/2, the trellis ras S
states and 2m branches per state as in the original N-dimensional trellis, whereas at intermediate times (kN/2) +
j, j - 1, 2, ..., (N/2 - 1), the number of states may be larger, where the exact number of states depends on the specific code in use.
As an example, consider the 4D 8-state dual Wei code, which can be viewed as a period-2, timervarying trellis code based on the partition Z4/2Z4 and a rate-1/4 convolutional encoder. This trellis is illustrated in Fig. 7.
Note that there are 8 states at even times, with 2 distinct branches leaving each state, and there are 16 states at odd times, with one branch leaving each state.
Once a two-dimensional trellis is obtained, reduced-state decoders can be defined as in the previous section. The operation of the VA is then similar, except it has to account for the additional time-varying nature of the trellis.
Construction of the Discrete-Time Channel Model:
So far, we have assumed a canonical discrete-time channel model described by ~U~p ~':~ i~ _ WO 91/15900 ~ _ 22 _ PCf/US91/02384 r(D)=e(D)h(D)+n(D) where e(D) is a complex input sequence, h(D) is a complex, stable, causal response with h0=1 and n(D) is a white Gaussian noise component which is independent of e(D). For best results we will choose h(D) to be minimum phase. In ' practice, the physical channel will rarely obey this canonical model; the channel response is often nonminimum-phase, the noise can be correlated and, furthermore, the physical channel is often continuous-time. Therefore, referring to Fig. 8, in practice, the physical channel 201 is often augmented by linear transmitter and receiver filters 202,204 in the transmitter and in the receiver, respectively, to construct a discrete-time channel 205 that obeys the canonical model.
In addition to providing the canonical discrete-time channel, it is desirable that these filters also help optimize the performance of the trellis precoder. It has been shown in the case of Tomlinson precoding that the optimum transmitter filter has a brickwall (or flat) spectrum within the Nycxuist band (for continuous-time channels this result holds under certain mild conditions (Price, "Nonlinearly Feedback Equalized PAM versus Capacity for Noisy Filtered Channels," Proc. ICC-72, June, 1972, supra)); the optimum receiver filter, on the other hand, consists of a matched filter sampled at the baud rate (and at the optimum sampling phase), followed by a discrete-time noise whitening filter that produces a minimum-phase combined response h(D) (Price, s,upra). This noise-whitening filter can be described by the cascade of a zero-forcing linear equalizer and a~ minimum-phase linear prediction (error) filter for the residual noise sequence appearing at the output of the linear equalizer. The prediction .filter represents the channel model h(D).
In practice, it may sometimes be preferable to allow ' the overall discrete-time channel to deviate from.the canonical WO 91 /15900 - 2 3 - PCT/L'S91 /02384 model to arrive at a better trade-off bet~.~een ISI and noise.
For example, the selection of the filter may be based on the output MSE. Of course, due to the presence of ISI, minimum MSE
does not guarantee minimum error prabability. Nevertheless, this criterion is often used in practice because of ease in adaptive implementation, and because it is believed that often (if not always) it leads to lower error probability, particularly at low SNR. Under the MSE criterion, the optimum transmitter spectrum is different; it has the 'water-pouring characteristic' found in information theory (J. Salz, "Optimum mean-squared error decision feedback equalization," BSTJ, vol.
52, pp. 1341-1373, 1973). For the moderate or high SNR's found on telephone lines,.however, a 'water-pouring spectrum' can be closely approximated by a brickwall spectrum. The minimum MSE
receiver filter also consists of a matched filter sampled at the baud rate and a discrete-time filter that produces a white residual error sequence. The whitening filter can now be represented by the cascade of a MSE linear equalizer and a prediction error filter for the error sequence (ISI+noise) that appears at the output of that equalizer, Under the MSE criterion, the received sequence can be written as r(D)=e(D)h(D)+n'(D), where the error sequence n'(D) is now signal-dependent and possibly non-Gaussian; the trellis precoder can still operate in the same manner. Here, it may be helpful to note that with an optimum transmitter filter, the overall response h(D) is independent of the SNR (in fact: it is the same filter that we obtain under the no-ISI:criterion.) In practice, the combination of the matched filter and the MSE linear equalizer can be implemented as a digital transversal equalizer with a fractional tap-spacing of T/M, where T is the baud interval and M is chosen sufficiently large to avoid.aliasing. Since the characteristics of the physical channel is often unknown, an adaptive training algorithm is >~~~~~t~
WO 9i/i5900 _ 24 _ PCf/US91/0238~1 needed to learn this equalizer. This can be accomplished by transmitting a known training sequence such as a pseudo-noise (PN) sequence modulating a QPSK signal structure prior to data transmission, and then using a least-mean square (LMS) algorithm (J. Proakis, "Digital Communications," McGraw-Hill, 1983). Once the equalizer is learned, an adaptive minimum MSE
linear predictor can be realized. This predictor has a tap-spacing of T and it whitens the residual error sequence.
Its steady-state coefficients form the desired channel response h(D) of the model. So far, we have implicitly assumed that all filters are infinitely long. Of course, in practice, the linear equalizer and the predictor are implemented with finite length filters: nevertheless, when these are sufficiently long the general discussion presented above will approximately apply.
After a sufficiently long training period, the information about h(D) is passed back to the transmitter for use during trellis precoding. During data transmission an adaptive algorithm may continue to adjust~the coefficients of the linear equalizer to track small variations in the channel response. However, the prediction filter is kept fixed. It is conceivable to update the predictor as well, provided that this information can be relayed back to the transmitter using a 'service channel' and proper synchronization between the transmitter and the receiver can be maintained. Alternatively, the receiver may monitor the true current values of the prediction filter coefficients and may request a new training signal if the.discrepancy becomes excessive.
Modem Implementation One application of trellis precoding is to a voiceband modem operating at a baud,rate of 2743 symbols and a bit rate of 19.2 kb/s;. thus sending 7 bits/baud.. The transmitter and the receiver are implemented mostly in digital form.
The transmitter:102 and the receiver 104 of such a ~~5~~~~
W091/15900 _ 25 _ PCT/US91/023$4 modem are shown respectively in Figs. 9, 10, Based on the optimality discussion above, the transmitter filter 105 is chosen to have a square-root-of-raised-cosine characteristics with a small excess bandwidth (< 120), thus approximating a brickwall spectrum. During training, a known pseudo-random four-point QAM sequence 107 x(D) is transmitted via train/data switch 105 for a sufficiently long period of time to allow the receiver to learn first its adaptive equalizer 108 and subsequently its adaptive prediction filter 110. The complex output of the transmitter filter is modulated to a carrier frequency we (radians/s) and the real part of the modulated signal is D/A converted, filtered by an analog filter, and transmitted over the channel, all by unit 107. In the receiver 104, the received signal is filtered by an analog filter 109 and then A/D converted at a sufficiently high nominal sampling rate M/T (the sampling phase is controlled by a timing recovery circuitry). The sequence of digital samples in is fed into an equalizer delay line 108 with a tap spacing of fi/M, where n is the sampling index.
To describe the operation of tie adaptive receiver, we denote the complex coefficients of the linear equalizer as cn, -N1<n_<NZ (N1+N2+1 taps). The output of the equalizer is computed once every baud; then, it is demodulated to obtain the sequence rk'. We assume that all coefficients are initially set to zero. The initial values may also be determined using a fast initialization method such as the scheme described in (Chevillat et al., "Rapid Training of a Voice-band Data Modem Receiver Employing an Equalizer with Fractional T-spaced Coefficients, IEEE Trans. Commun., vol.
COM-35, pp. 869-876, Sept. 1987). In any case, the coefficients of the linear equalizer are adjusted according to cn.k+1-cn,k c'.sk zn*' n=-N1'~..,0,....N2 where -ek=rk'-xk is the error between the equalizer output v rk' and the known transmitted symbol xk, and k is the index for the baud interval. If the step size a is sufficiently small, then the coefficients will converge to values which minimize the MSE. Furthermore, if N1 and N2 are sufficiently large, the filter converges to the optimum MSE
linear equalizer that we described earlier.
Once the equalizer has converged, its outputs rk' ' are fed into a T-spaced prediction (error) filter with coefficients h0=1 (fixed), hl, and h2. This filter produces r(D), the output of the channel model. Its coefficients are updated according to hi, k+1-hi,k-~e*k-iE'k, i=1.2 where e'k=rk-xk - xk-lhl,k xk-2h2,k is the final error after prediction and f3 is another small stec_ size.
Again, when f3 is sufficiently small, these coefficients will converge to their values which minimize the MSE. If the overall channel model can be represented by a three-coefficient model as we assumed here, we expect the residual error sequence to be approximately white, as explained earlier.
After convergence, the coefficients hl and h2 are sent back to the transmitter for use during trellis precoding.
This can be accomplished, for example, by simply encoding the coefficients into binary words and transmitting these using 4-QAM signaling. These words are typically sent in a small frame which has a flag for recognizing the start of the frame and an error control check to detect transmission errors: In . what follows, we denote the estimated channel response by h(D), and its formal inverse by g(D).
During transmission of actual data from a data terminal equipment (DTE) 120, the prediction filter 110 is kept fixed,, but the linear equalizer 108 is continually adapted in a decision-directed mode to track small.channel.variations,. (It is also possible to update the prediction filter along,with the ~~58~ ~~;
WO 91 / 15900 - 2 ~ _ PCT/LS91 /02384 precoder coefficients. The latter ~,rould be updated using a side channel:) Because of trellis precoding, it is not possible to generate the error signal ek directly at the equalizer output, since decisions are available at the output of the prediction filter 110 h(D). Therefore, to reconstruct the equalizer error, the final prediction error e'k=rk-yk' between the output rk of the prediction error filter and the estimated channel symbol yk' is passed through g(D) 122. The estimates yk' can be obtained with no delay. using a simple slicing operation, or they can be obtained from the Viterbi decoder, albeit after some delay. Experiments indicate that this structure operates reliably even in the presence of decision errors.
Referring again to Fig. 9, in the trellis precoder 124 we use a four-dimensional 16-state Wei code for coding (denoted as Cc) and a two-dimensional 4-state Ungerboeck code for , shaping (denoted as Cs), and we send 7 bits per two dimensions (or baud). The code Cs is derived from trellis code C which is based on a rate-1/2 4-state binary convolutional encoder and the two-dimen~ianal partition Z2/2Z2. Equivalently, C can be described by a rate-2/4 convolutional encoder and the four-dimensional partition Z4/2Z4. We use a four baud frame to define Cs, where ~ .
in the first two bands Cs = 8C and im the next two bands _Cs - 8RC, as described earlier. Thus, the time-zero lattice is 8RZ4 during the first two bands and 1624 during the next two bands of the frame. The code C_c, on the other hand, is based on a rate-3/4 binary convolutional encoder and the 4D.
partition Z4/2Z4. This code adds one redundant bit every four dimensions (or equivalently, every two bands). Note that a fundamental region of the time-zero lattice 8RZ4 will contain exactly 22x6.5+1 _ 214 points from any coset of the:lattice 24; i.e., ~Z4/8RZ4~=214, and a " , _.
WO 91 / 15900 ~ - 2 8 - PCT/US91 /02384 fundamental region of the time-zero lattice 152' will contain 22x7.5+1 __ 216 points from any coset of 24.
A small buffer 130 is filled with bits received from the DTE at the rate of 7 bits per (2I>) signaling interval.
Referring also to Fig. 2, during the two bauds, 4j and 4j+1 ' [the two bauds 4j+2, 4j+3] a scrambler 132 takes 7 [8] and then 6[7] bits from this buffer, respectively. The scrambled bits are delivered to a binary encoder in groups of 13 [15]
so-called I-bits, labeled I6(4j) through IO(4j) and IS(4j+1) through IO(4j+1) [I7 (4j+2) through IO(4j+2) and I6(4j+3) through IO(4j+3)], at two successive times 4j and 4j+1 [4j+2 and 4j+3].
Referring to Fig. 11, the I-bits I2(4j)I1(4j) [I2(4j+2) I1(4j+2)] are taken to represent an integer mod 4 and are differentially encoded by a coding differential encoder 150 to produce t'~ao differentially encoded bits ID2(4j)ID1(4j) [ID2(4j+2) ID1(4j+2)] according to the relation ID2(4j)ID1(4j)=I2(4j)I1(4j) 0+4 ID2(4j-2)ID1(4j-2) [ID2(4j+2)ID1(4j+2)=I2(4jt2)I1(4j+2)~4I~2(4j)ID1(4j)]
where ~4 represents mod-4 addition. , The bits ID1(4j)~ and IO(4j)~[ID1(4j+2) and IO(4j+2)]
enter a rate-2/3, 16-state binary corrvolutional encoder 152, which produces one redundant bit YO(4j) [YO(4j+2)]. Referring to Fig. 12, this encoder includes a sequential circuit 246 having four delay-2 shift registers 248 and three mod-2 adders 250. (This encoder can also be realized using other circuits.) Referring again to Fig. 11, the bits ID2(4j)ID1(4j) IO(4j) [ID2(4j+2)IDl(4j+2)IO(4j+2)] and the redundant bit y0(4j) [YO(4j+2)] enter a bit converter 154 that generates four output bits Z1(4j)ZO(4j)Z1(4j+1)ZO(4j+1) w-[Z1(4j+2)ZO(4j+2)Zl(4j+3)ZO(4j+3)] according to the following ..: ::

~~~~3~.:~ ~' WO 91/15900 - 29 - P~1'/(JS91/02384 table. Note that the corresponding bracketed headings for ~he table are (ID2(4j+2)ID1(4j+2)IO(4j+2)YO(4j+2)Z1(4j+3)ZO(4j+3)Z1(4j+2)Z0(4j+
2)J:
ID2(4j) ID1(4j) IO(9j) YO(4j) Z1(4j+1) ZO(4j+1) Z1(4j) ZO(4j) 0 p 1 1 1 1 0 0 I 0 0 0 0 1 : 0 1 1 0 1 . 0 0 0 0 1 1 1 0 0 1 ' 1 1 1 1 1 1 0 1 0 1 _ 1 These output bits are used to select one of 16 cosets of 2Z4 whose union is the four-dimensional grid Z4+(.5, .5, .5. .5). Each such coset is represented by a pair. of cosets of 2Z2 in the two-dimensional grid Z2+(.S, .5). _ 72987-~6 These are selected individually by Z1(4j)ZO(4j and Z1 (4j+1) ZO (4j+1) [Z1 (4j+2) ZO (4j+2 i and Z1 (4j+3) ZO (4j+3) ] .
The remaining I-bits I3(4j) through I6(4j) [I3(4j+2) through I7(4j+~~)] pass through unchanged and are 5 relabeled as Z2(4j) through Z5(4j) [Z2(4j+2) through Z6 (4j+2) ] . They are combined with Z1 {4:j ) ZO (4 j ) [Z1 (4j+2Z0 (4j+2) ] to form the si.x-tuple Z5 (4j ) Z4 (4j )...Z0 (4j ) [seven-tuple Z6 (4j+2) Z5 (4j+2)...ZO (4j+2) ] . Similarly, the I-bits I2 (4j+1) through I5 (2j+1) [I2 (4j+3) througr. I6 (4j+3) ]
10 together with Z1 (4j+1) ZO (4j+1) [Z1 (4 j+3) ZO (4j +3) ] form the six-tuple Z5 (4j+1) Z4 (4j-+-1)...ZO (4j+1) [seven-tup.le Z6 (4j+3) Z5 (4j+:3)...ZO (4j+3) ] .
The :C-bits I1 (4 j+1) IO (4j+1) [I1 (4j+3) IO (4j+3) sequentially enter a rate-1/2 inverse syndrome former (H-1)T
15 156. H is a syndrome former of the binary code C defined above . Referr~.ng to Fig . 14 , (H '~) T produces two pairs of S-bits S7_ (4j ) SO (4j ) [S1 (4 j-h2) SO (4j +2) ] and S1 (4j-~-1) SO (4j+1) [S1 (4j+3) SO (4j+3) ] . As discussed earlier, (H-~) T .is the left inverse of (H)T discussed below, and is included t:o limit 20 error propagation.
Referring to Figs. 13A, 13B during the two bauds 4j and 4j+1 [4j+2 and 4j+3] , the :1.6-bits [18-bits] generated in this manner are mapped into two initial signal points r(4j) and r(4j+1) [r (4j+2) and r;4j+3) ] chosen from a 25 256-point [512-point] two-dimensional square signal set which lies on the grid Z2-~{.5, .5). First, the six-tuple [seven-tuple] Z-bits are used to select a signal point from the 64-point [128-point] region-0 signal set 261. 'this signal set is divided into four subsets which are cosets of 30a 2Z2 (indicated by different shadings in Figs. 13A and 13B).
The signal points are labeled such that Z1 and ZO together select ~~~8~~a~

one of the cosets according to the labeling shown in the lower left corner of Fig. 13A. The labeling of the remaining bits Z5,Z4,Z3,22 is shown only for coset a. For other cosets, the labeling can be obtained by the rule that 90 degree rotation of a point around the center (4,4) [8,0) of the region-0 signal points result in the same labeling.
The Wei encoder 1S2 (Fig. 1S) insures that the minimum distance between allowable sequences of region-0 points is kept large.
The S-bits Sl,SO determine the final region of the initial point according to the rule:
S1S0 Region of 1 Referring to Figs. 14A, 148, the signal points in region-0 are moved to the region chosen by the S=bits in such a way that they remain in the same coset of 8Z2 [8RZ2].
This is done by offsetting the region 0 point by (0,0),(0,-8), (-8,-8), or (-8.0) [(0,0), (-8,-8), ('-16,0), or (+8.-8)l. It follows that the translated final point is in the same coset of Z2 as the region-0 point. Therefore, the minimum di's'tance between allowable sequences remains unchanged..~The signal points obtained in this manner form the initial sequence x(D).
Thus the S-bits can be viewed as a coset representative sequence which determines the coset of 1622 in 8Z2 [16RZ2 in 8RZ2J for.the elements of initial sequence ..
x(D)., Referring to Fig. 15, the initial sequence.x(D) is ~~:~J ~-WO 91 / 15900 - 3 2 _ PCT/L'S91 /02384 transformed into the transmitted sequence e(D) using parallel decision-feedback decoding (PDFD) 294 (which, along with the circuitry of Fig. 3, is part of the trellis precoder 124 of Fig. 9). It is possible to replace PDFD by a more powerful RSSE decoder, or by some other reduced-complexity decoder for the filtered trellis code C '.
s As we saw earlier, the optimum decoder selects from the shaping code Cs the code sequence cs(D) which minimizes the mean-squared error between the sequences x'(D)=x(D)g(D) and cs~(D)=cs(D)g(D). The transmitted sequence is the error sequence e(D)=x'(D)-cs'(D). The code sequence selected by PDFD will not generally result in minimum MSE, however, experiments indicate that the excess MSE will often be small.
For example, for a channel with two coefficents, PDFD can achieve 0.5-0.85 dB shaping gain, depending on the values of hl and h2.
Now, we consider the operation of PDFD in some detail.
First, we describe the shaping code Cs.
Referring again to Figs. 14A and 14B, the two-dimensional symbols of this code arP chosen from the integer lattice 8Z2 [8RZ2] 299. These symbols are partitioned into four subsets which correspond to the (four) cosets of 1622 in 8Z2 [16RZ2 and 8RZ2.] and are labeled as A, B, C and D as shown in the lower-left corner of Fig. 14A.
All possible sequences cs(D) in the shaping code Cs are represented by the trellis diagram:of F-ig. 4. Here, from any state there are two branches 302, each branch representing a two-dimensional subset. For example, state 0 has two branches labeled A and B.
PDFD-operates recursively using the Viterbi algorithm and in synchronism with the four-dimensional Wei encoder such .
that every other baud it releases two delayed error symbols WO 91 / 15900 - 3 3 - PC'f / US91 /02384 e(4j-M) and e(4j+1-M) (e(4j+2-M) and e(4j+3-M)], where M (an even number).is the decoding delay measured in number of bauds. Every baud interval, the VA has in storage four path metrics ~i(4j+Q)(~i(4j+Q+2)], i=0,1,2,3, and Q=0,1, one for each state in the trellis diagram. The VA
also has in storage a finite history of previously hypothesized error symbols ei(4j+2-m) (ei(4j+2+2-m)], i=0.1,2,3.
2=0,1 and m=1,2,...,M, one for each state. In each iteration (every baud), the paths are extended by incrementing path to metrics by the branch metrics; the branch metric for the path branch p(p=1,2) from the i'th state is described by bi~p(4j+SL)=~-ei(4j+Q-1) hl-ei(4j+Q-2) h2+x(4j+Q) - ci p(4j+2)~2 Ibi~p(4j+Q+2)=~-ei(4j+Q+1) hi-ei(4j+2)h2 +x(4j+Q.+2) -15 ci~p(4j+Q+2)~2]
where c. (4j+9,) (c. (4j+Q+2)] is a symbol from the coset of~i6Z2 (16RZ~]passociated with this branch~that gives the minimum value for the branch metric. For each state 20 i' after the transition, the accumulatefd metric of the two merging paths are compared and the path with the smaller metric is declared a survivor; its path metric becomes the new path metric for state i' and the error symbol for the surviving transition (i*,p*), which is given by ei,(4j+2)=-ei*(4j+Q-1)hl - ei*(4j+Q-2.)...h2+x(4j+Q) - ci*,p*(4j+Q), (ei,(4j+Q+2)=-ei*(.4j+Q+1)hl - ei*(4j+Q) h2+x(4j+Q+2) - ci*,p*(4j+~+2 becomes its most recent error symbol stored in its path history for subsequent use. Every.other baud, or when 2=1, the _ algorithm determines the state with the minimum path metric, traces back its error symbol history .for M symbols, and."
releases the error symbols for,timee 4j-M and 4j-M+1 (4j-M+2 y WO 91 / 159m0 ' - 3 4 - PCT/US91 /02384 and 4j-M+3] as outputs.
Note that the error symbols generated by the precoder 124 lie inside the Voronoi region of the sublattice 1622, [16RZ2]
which is a square of side 16 [a rotated square of side i6 J2], centered at the origin, independent of the channel response h(D). When hl and h2 are both non-integers and h(D) is FIR, then the error symbols can take infinitely many values within that square.
In trellis precoding using trellis codes that are based on partitions of binary lattices, the elements ej of the transmitted sequence belong to a 2D constellation that lies within a scruare boundary. Furthermore, trellis precoding expands the size of the 2D constellati.~:... Moreover, all tzese effects, and in addition, the further unbalance in the sizes of the Voronoi regions of the time zero lattices :1 s0,k' increase the peak-to-average ratio of the transmitted symbols, which may be undesirable. It is possible to reduce the P ~R and obtain a more circular boundary by applying constraints to PDFD
(Fig. 7). For example, PDFD can-be constrained in such a way that it only selects code sequences cs'(D) (from the filtered code Cs') which result in transmitted sequences e(D) whose elements e~ have magnitudes no greater than some predetermined radius Rc; i.e., (ej~<Ry. When Rc is not too small, the decoder can proceed without violating the constraint.
This type of constraint can be incorporated into any RSSE-type decoder very easily by deemphasizing=those branches of the trellis whose mean-squared error (MSE) metrics exceed R~, by multiplying them with an appropriately chosen number Q. The larger Q gets, the harder the constraint becomes: This reduces the probability of choosing points lying outside the constraint circle.- When Q is sufficiently large, - and Rc is. not very. small, points outside the constraint-W091/15900 _ 35 _ PCT/LS91/02384 circle will never be selected. Note that if Rc is too s;nall, one should not disallow branches, because this may force the decoder to choose an illegal sequence as a code sequence, and then the initial sequence x(D) cannot be correctly recovered in the receiver.
When the transmitted sequence e(D) is filtered before transmission over the channel, it may be beneficial to apply the constraint in higher dimensions (on groups of successive elements of the sequence e(D)), in order to minimize the PAR
after the filtering. For example, the constraint may be , applied in four dimensions by comparing the sum of the two most recent branch metrics against some squared radius R~. The decoder could force.the condition ~e2j~2+~e2j+1~2~Rc~' Rc being the radius of a 4D square. This constraint may be applied blockwise or sequentially in a sliding window fashion. _ The VA is further slightly modified to insure that the , code sequence cs'(D) it selects is an allowable sequence from C_s', since otherwise the initial sequence x(D) cannot be correctly recovered. Whenever the VA traces back the path history to make a decision, it traces beck other paths as well to determine whether they end in the same state; when they don't, those paths are pruned by setting their current path metric to a very large value. This insures that only allowable code sequences are selected.
The precoded sequence e(D) is passed through the transmitter shaping filter 106 (Fig. 10), D/A converted, filtered by the analog front-end 107 and subsequently transmitted over the channel.
At the receiver, the received signal r(t) is first filtered by the analog filter 109 (Fig. 10) to remove out-of-band noise, and A/D converted into a digital stream (zn) at_.a rate of M/T. This digital stream is input into the adaptive linear equalizer 108 with N1+N2+1 taps and a J ~~c .i t tap-spacing of T/M (not to be confused with the case of M for delay elsewhere). The output of this filter is sampled at the baud rate 1/T, then demodulated and subsequently filtered by the prediction filter h(D) 110 to approximately produce the received sequence r(D)=e(D)h(D) + n'(D)=x(D)-cs(D)+n'(D)=y(D)+n'(D).
As we discussed earlier the elements of y(D) lie in the same coset of 2Z4 in Z2 as x(D), and thus the same coset of 2Z2 in Z2. Therefore, the minimum distance between possible sequences y(D) is no smaller than the minimum distance between~possible input sequences x(D) and the sequence y(D) can be decoded with a Viterbi decoder 133 for x(D), except the decoder must be slightly modified to take into account the potentially different signal set boundaries for the elements y~ of y(D).
It can be shown that the shape and size of: the signal set boundary for yd's depend on the channel coefficients hl and h2. For example, if hl and h2 are both real, then the boundary will be a square of side length 2(1+hl+h2), again centered at the origin. If the output constellation size is of concern, it may be beneficial to limit the magnitudes of the coefficients during training. This may be accomplished using a constrained adaptation algorithm.
It is possible to make the decoder 133 transparent to the constellation expansion that is caused by the channel filter. If we take the received sequence y(D) and translate it by some sequence a(D), where the elements of a(D) are elements of the lattice 1622 (16RZ2), then the modified sequence y(D)-a(D)=x(D)-cs(D)-a(D) is still equal to x(D) mod Cs.
since the component cs(D)+a(D) is an allowable code sequence, It follows that the received sequence can. be '~~;~~~r3~
WO 91 / 15900 _ 3 ~ - PCC/US91 /02384 translated inside the Voronoi region of 1622 (16RZ2].
The Viterbi algorithm for the (coding) code Cc can now operate on this translated sec_ruence assuming a square boundary of side 16 during the first two bands of a frame and a rotated square of boundary of side 16 J2 during the next two bands of ' the frame and output a delayed estimated sequence y"(D).
Every other baud, the VA will produce two delayed decisions y"(4j) and y"(4j+1) [y"(4j+2) and y"(4j+3)]. To extract the I-bits from these, we first obtain the Z-bits by translating the decisions into region-0, in such a way that the 2D elements remain ir. the same coset of 8Z2 (8RZ2].
Then, the Z-bits can be extracted using an inverse mapping according to Figs. 15A and 15B.
To extract the S-bits, we first label the regions (as before) according to Region SQ1SQ2 Then, the regions of y"(4j) and y"(4j;-1) determine the labels -SQ1(4j)SQO(4j) [SQ1(4j+2)SQO(4j+2)] and SQ1(4jt1)SQO(4j+1)[SQ1(4j+3)SQO(4jt3)). It is easy to show .
that SQ1(i)SQO(i)=S1(i)SO(i)~2B1(i)BO(i), where i=4j or 4j+1 [4j+2 or 4j+3],~~92 represents exclusive-or operation and.-- , B1(i)BO(i) is a binary two-tuple represeting the coset (A=00, B=11; C=O1 and D=10) of 1622 [16RZZ] for the code symbol cs(i) chosen during precoding by PDFD.
.. Ref erring to Figs. 16 and 17, the labels SQ1(i)SQO(i) are then sequentially passed through a feedback-free~:_ y~;,~~ ~r~:s ..
,~U~~.)'v WO 91/15900 - 38 - PCT/US91/023$4 T
syndrome; former H' 256. The syndrome-former has the property that any allowable sequence of coset labels B1(i)BO(i) will produce an all-zero sequence at its output. Furthermore, (H-1)THT=I, the identity matrix. Therefore, a sequence which first passes through (H 1)T and then through HT
will remain unchanged. Thus, as shown in Fig. 17, at the output of the syndrome-former we obtain the transmitted bits I1'(4j+1)IO'(4j+1) [I1'(4j+3) IO' (4j+3)], as long as the estimates y"(i) are correct. When there are occasional errors, they do not create catastrophic error propagation, because HT
is feedback-free.
Referring again to Fig. 17, the remaining information bits can be recovered by an inverse bit converter 286 and a coding differential decoder 288 which operates according to the relation:
I2'(4j)I1'(4j)=ID2'(4j)ID1'(4j) 94 ID2'(4j-2) ID1'(4j-2) [I2'(4j+2)I1'(4j+2)=ID2'(4j+2)ID1'(4j+2)84ID2'(4j)I'D1'(4j)]
where A4 is a modulo-4 subtraction. Referring again to Fig. 11 the I-bits are then descrambled in a descrambler which forms part of unit 133 and delivered to the DTE via a buffer 135.
Differential Coding _for Shaping Suppose the channel introduces a phase rotation of 90k degrees, k=1, 2, or 3. The translation of. the error point into regiont-1 is then rotated around the point (4,4,) [8,0] by the same amount. Since the Wei code is transparent to 90k degree phase rotations, the mapping used in the transmitter guarantees that the Z-bits can be correctly recovered. Unfortunately, this is not true for the region labels SQ1SQ0.
To remedy this situation,.the:label. sequence SQ1SQ0 can ' ' be generated according to the phase-invariant labeling shown in ~;;~~~~ ~'J
WO 91 / 15900 _ 3 g _ PC'T/US91 /02384 Fig. 18. In crder to guarantee that the relationship SQiSQO=S1S0~2B1B0 still holds, a differential encoding operation is necessary. This is done as follows: Referring to Fig. 19, for each signal point obtained through region-0 mapping 258, we extract its subregion label SQ1SQ0 with a subregion extractor 306 according to Fig. 20. A shaping differential encoder 308 and a map into offsets 310 use the bits SQ1SQ0 and S1S0 to offset the region - 0 point into two new points x0(i) and xl(i), where i = 4j [4j+2] or i = 4j+1 [4j+3]~ such that they remain in the same coset of Z2. This mapping can be described by the following two tables:

0 . 0 0 1 1 0 0 1 0 1 0 0 0 1 0 : 1 1 "~1 0 1 0 1 1 1 Di0 Dil 0.i (offset);
p 0 . 0.0+j0:0 (0.0-j0.0]
0 1 0.0-j8.0 [-8.0-j8.0]
1 0 -8.0+j0.0 [8.0-j8.0]
1 1 -8.0-j8.0 [-16.0-j8.0]

~~r~~y:::~
IJ :J L% . .~ J --w0 91 / 15900 ~ - 4 p - PCT/G'S91 /U2384 In the PDFD decoder, the point x0(i) is used ~or branches corresponding to cosecs A and B, whereas xl(i) is used for branches corresponding to cosets C and D. It can be shown that if SQ1SQ0 is the subregion label of the received point, then SQ1SQ0=S1S0~2B1B0. Therefore, by passing the "
subregion label (which is rotationally invariant) through the syndrome-former HT we can recover the bits IO(4a+1) and I1(aj+1) [IO(4j+3) and I1(4j+3)l even in the presence of 90k degrees phase offset.
Fiexaqonal 2D Constellations If we choose the underlying shaping trellis code C(A/A';C) as a code based on a partition A/A' of so-called ternary or quaternary lattices whose constituent 2D
lattice and sublattice A2 and !12' are versions of the hexagonal 2D lattice A2, then the elements ej of the transmitted~sequence will be bounded within a region Rv(A2') which is hexagonal rather than square. Such a constellation has the advantage of being more nearly spherical.
Alternative formulation based on. label translates:
In the preferred embodiment of trellis precoding, we have used a linear (mod-2) code CS. This allowed us to use the syndrome-former mapping methods for preventing error propagation. It is possible to give an alternative formulation of trellis precoding, based on label~translates of trellis codes, such that the syndrome-former methods can be applied with nonlinear trellis codes, as well.
~ As before, for coding we use an N-dimensional trellis.
code Cc(Ac/Ac';Cc), where Ac/A~' is a binary lattice partition of order 2nc and CC is a binary, rate-kc/nc convolutional-code (nc=kc+rc). For shaping and precoding, we use a,trellis code _Cs(As/As'; Cs) based on a binary rate-ks/ns convolutional-code Cs(ns=ks+rs).~
I
' l ,' l s ~~ 1 ~ ~~ ~~3 '~i WO 91/159()0 - 41 - PCTIL'S91/02384 As shown in Figure ?.2 , a sequence of rs-tuo_ies s(D) enters an inverse syndrome former (H 1)T for the convolutional code CS to produce a sequence of ns-tuples t(D), which specifies a label-translated code Cs(As/'1s~' Cs~2t(D)). The trellis precoder will select the channel output sequence y(D) from a coset Cs(As/As'; Cs82t(D))+d(D) of this label-translated code, where d(D) is a coset representative .
sequence identifying cosets of AS in some translate AC+a of Ac.
A sequence of kc-tuples i(D) enters the convolutional encoder Cc to produce a sequence of nc-tuples b(D). The sec_ruence b(D) together with an uncoded sequence of nf-tuples v(D)specifies the coset representative sequence , d(D). Then a channel output sequence y(D) that lies in Cs(As/As', Cs+t(D)) + d(D) can be identified by .
d(D) which specifies a sequence of cosecs of ;vs in .1c +
a, a sequence z(D) = cs(D)+t(D) with cs(D)eCs, which further specifies a sequence of cosecs of As' in .1s +
d(D) and a sequence a(D)eAs' which selects a particular sequence in that coset. A decoder will find the sequences cs(D) and a(D), such that the resulting y(D), after filtering by the channel inverse g(D), has as small an average power as possible. The transmitted sequence e(D) is the filtered sequence: e(D)=y(D)g(D).
In the receiver, after appropriate front-end filtering, we obtain the observations r(D) = e(D)h(D) + n(D) _ y(D) + n(D). Referring to Figure 23, the sequences i(D), v(D) and s(D) can be recovered as follows. First, since As is a sublattice of Ac', the sequence y(D) is a legitimate code sequence from _Cc. Therefore, a conventional decoder for Cc can be used to find an estimate y(D), as before, with the same decoding complexity and probability of error as-that of the ( i_; ~ fl ' , h cs L. d ~_~ "J ,.
WO 91 / 15900 - 4 2 - PCTlUS91 /02384 code Cc on an ideal channel, h(D) = 1. As before, it is possible to translate r(D) to a fundamental region of .\s', thus making the decoder essentially independent of the channel response h(D) without affecting the performance. In the absence of decoding errors, the.sequence d(D), representing the cosets of As' in Ac + a, can be obtained directly from y(D). This can be used to recover the sequences b(D) and v(D). The information sequence i(D) can be found using a decoder for the convolutional code Cc. The sequence z(D) _ cs(D ~+t(D) representing the coset of As' in As can be recovered also directly from y(D). The estimate of the sequence s(D) is obtained by passing z(D) through a syndrome-former HT; i.e., s(D) = z(D)HT - cs(D)HT ~+
t(D)HT - s(D)(H-1)THT - s(D), where we used the well-known fact that cs(D)HT=0. If HT is chosen to be feedback-free, then s(D) can be recovered without catastrophic error propagation.
The selection of the sequence e(D) can be:done, as before, using RSSE techniques, with the slight modification that trellis branches will now correspond to cosets of :1s' in As+dk as identified by the label-translates z(D) _ cs(D)Q+t(D), the coset representative sequence d(D) and the sequence a(D), and the branch metrics are computed according ~~
to ~Y(Zk = tk + cs,k' dk, ak,) -j > 1 ek-j hj ~ . _ Here we have used the notation y(zk, = tk ~+ cs,k~ dk' ak) to indicate a channel output symbol yk specified by zk, dk and ak.
With this method, we can send 13N/2 = kc + nf_+
rs bits per N-dimensions, where of = - ...

WO 91 / 15900 - 4 3 - PC,'T/lJS91 /023&t log2~nc'/ns~. As before, we can increment of by N/2 (or equivalently, by 1 bit per two dimensions) by replacing As by Rns, because iog2p~,/Rns~ = log2p~'/ns~ +
N/2. (Or similarly, we can increment of by N/2 by replacing AC by Rnc.) To send (N/2)13 = kc+rs+j+(d/2m)(N/2) bits/N dimensions, where O<d<2m, we can again define a frame of length (N/2)2m bauds, and use the code Cs for (N/2)(2m - d) bauds, and its scaled versions PCs for (N/2)d bauds. (A
similar effect can be achieved by replacing the coding code Cc by its time-varying versions based on the codes CC ad RCc.) The implementation given above for the time-invariant case remains essentially unchanged, except one now has to take into account the variations in scaling. Again, constellatation constraints can be applied to control PA.R.
B. Alternative formulation based on re ions:
It is possible to formulate trellis preceding using convolutional codes and N-dimensional regions, instead of trellis codes. This formulation is more general in some respects. In this section, we describe this formulation, first using time-invariant regions, and then with regions having time-varying scaling to support fractional rates.
As before, for coding we use an N-dimensional trellis code _Cc(nc/Ac'' Cc), where nc/nc' is a binary lattice partition of order 2nc and C~ is a binary, rate-kc/nc convolutional code. For shaping anc preceding, instead of a trellis code, we use a binary, rate-ks/ns convolutionai code Cs and a.set of N-dimensional regions R_(j), j=1;...,J, where J may be infinite. Each region is partitioned into 2ns subregions Rn( j ) , j=1, . . . , J.:. ~~.'such. a way that every subregion contains an equal number 'of.points from each of the 2nc cosets of 1~' in a translate nc,.

~~i~8~~'~
WO 91 / 15900 - 4 4 - PCTlIJS91 /02384 + a of A,c. To transmit at a rate f3, where fl > (2/N) (kc + rs), the regions must be properly scaled such that each subregion contains 2''c+nf points from Ac + a, where of = fiN/2 - kc - rs.
We select the elements of the channel output sequences y(D) from cosets of Ac' in Ac+a, as specified by the sequence of nc-tuples b(D) produced by the convolutional encoder Cc, from within one of the subregions Rn(j). Any signal point yk from Ac + a, in any one of the regions R(j), can be identified by a quadruplet: y(bk, vk, zk, jk)~ where an index jk selects a region R(j), an ns-tuple zk selects a subregion _Rn(j), an n~-tuple bk specifies one of 2nc cosets of Ac' in Ac+a, and an nf-tuple vk picks one of 2nf points of that coset in Rn(j). As illustrated in Figure 24, the convolutional encoder produces i5 the sequence of nc-tuples b(D), and encoded symbols form the sequence of nf-tuples v(D). The sequences j(D) and z(D) which identify the subregions are selected as follows. First, a sequence of rs-tuples, s(D), is mapped into a sequence of n -tuples t(D) according to t(D) = s(D)(H 1)T, where s (H-1)T is a right inverse to the nxr syndrome-former matrix for Cs. Then, in the key step of trellis precoding, a decoder selects j(D) and z(D)et(D)$-2C~ such that the transmitted sequence e(D) = y(D)g(D) has as small an average power as possible.
. Since y(D) is a code sequence from ~c(Ac/Ac'; Cc), it can be decoded, as before, by a decoder for Cc. Again the decoding complexity and the probability of error will be the same as if the code _Cc was .
used on an ideal channel. Referring to Fig. 25, if y(D) is the estimated channel output sequence, then z(D) can be extracted by identifying the subregions Rz(j) in which the elements yk lie. The estimates b(D), v(D) can be found directly-from ~,'U5~45~
WO 91/i5900 - 45 - PCf/US9i/02384 y(D) with an inverse map. The sequence s(D) can be extracted directly from z(D) with a syndrome-former HT according to s(D) = z(D)HT, because cs(D)HT - 0 and (H 1)THT -I, where I is the rxr identity matrix. If HT is chosen to be feedback-free, then s(D) can be recovered with no error propagation.
Having discussed the operational principles, we now turn to the selection of the index sequences j(D) and z(D). To minimize the output average power, we need to search a trellis with states sk' - [sk; jk-1~ zk-1)~ where jk 1 - (jk_1," " lk-K) and zk 1 - (zk-1," " zk-K)' If J and K are both finite, then this trellis will have a finite number of states. But'.
even in that case, it is often necessary to reduce complexity to make the search feasible. Again, this can be accomplished by an RSSE-like shaping decoder. An RSSE trellis can be constructed by applying set partitioning principles to the past values of jk-i and zk_i, i=1,...L<K. Then, the VA can be applied to search the reduced-state trellis with the branch metrics ~ek~2 - ~Y(bk, vk, zk = tk 9-2 cs.k' )k - ~j~lek_~hj~2 where cs~k is an ns-tuple corresponding to the branches of the convolutional code, and the minimization is over all possible values of jk. Here the minimization is over all possible indices jk, 1<jk<J; for the regions R(j). Notice that, for large values of J, this minimization can be very time-consuming, if one has to compute the above metric for all possible values of jk. Now we will show how this problem is.
avoided, when the regions are chosen as fundamental regions of lattices.
But first we should note that when the channel is ideal (h(D) =~1), and the number of regions is one (J=1), then ~~3~~~~~,__ WO 91 / 15900 ' 4 6 - PCT/US91 /02384 the method described above reduces to trellis shaping on regions. Notice that in that case, the minimization step in the branch metric computations is absent.
Now let As/As' be an N-dimensional lattice partition of order 2ns. Next, let R_(1) be a fundamental ' region of As', and then form an exhaustive partition of N-space 'into regions R(j) - _R(1) + aj, were aj is an element of As', with Al = 0. Then partition each _R(j) into 2n.s subregions Rn(j), such that each subregion is a fundamental region of As, and furthermore, each subregion contains an equal number of points from the 2nc cosets of Ac' in Ac + a. Notice that the regions R_(j) differ by points on the lattice As', and hence the minimizing step in the branch metric computations is now only as complex as decoding As'. An example is shown in Figure 26 for a 4-way partition o~ the type Z2/2Z2.
We may point out that although we have used the syndrome-former methods to generate the coset representative sequence t(D) in the extensions we described above, other mapping methods can be used also. However, in that case, to recover t(D) from z(D) a binary dscoder~for Cc associated with the set of coset representatives must be used to extract cs(D) and then farm t(D) = cs(D)~z(D);
Finally, we should also point out that when Cs is a linear code, the original formulation based on fundamental regions of time-zero lattices is identical to the formulation based.on label-translates, and they both.can be posed as a special case of the formulation based on regions.
With this method, we can also send 13N/2 = kc + of .
+ rs bits per N dimensions. As before, we can increment of by N/2 by replacing Ac by R1c, to send (N/2)ti = kc +
rs + j + (d/2m)(N/2).bits per N dimensions, where .
O<d<2m. We can again define a frame of length (N/2)2m bands, and use the code Cc for (N/2)2m - d0 bands, and WO 91/15900 - 47 - PCT/L'S91/0238A
its scaled versicn PCc for (~1/2)d bauds. A similar effect can be achieved by varying the scaling of the subregions. The implementation given above for the time-invariant case remains essentially unchaged except one now has to take into account g the time-varying nature of the code or the scaling, Again, constraints can be applied to reduce the peak to average ratio.
Other embodiments are also within the following claims. For example, other reduced-complexity decoders (e. g., M algorithm) could be used.

~~~~.'.3~~~t1 WO 91 / 15900 - 4 8 - PC?1US91 /023$4 Appendix A
Terminology and principles we first discuss terminology and principles which underlie the invention.
Trellis Precodinq A lattice A is a collection of N-dimensional vectors c, which form a group under ordinary vector addition. A .
fundamental region R(A) of A is a region of N-space which contains one and only one point (a cosec representative) from each distinct coset (translate) A + a of A. All fundamental regions of a given lattice have the same volume, V(A), and their translates R(A) + c, ceA, form a tesselation of N=space. This induces the following coset decomposition: Any N-vector y can be uniquely decomposed as y = c + e, where ceA and eER(A); that means, y is equivalent to a coset representative eeR(:~) modulo :1.
The translates R(A) + c may be viewed as the decision regions D(c) of a decoder D(A), whose error regions D(c)-c are all equal to R(A) (to show this correspondence explicitly, in what follows, we will sometimes write R(A, D) or D(~1, R)).
_ _ -Thus, if eeR(~1) is a coset representative, but otherwise an unknown information-bearing vector, then it can be recovered from y = c + e, without any knowledge about c, by first finding the lattice point c using the decode); _D(~, _R) and then-forming the error a = y - c.
The Voronoi region RV(A) of a lattice A is defined as the set of points that are at least as close to the origin as to any other lattice point. The Voronoi region is a fundamental region corresponding to a minimum-distance decoder of A (we ignore the effect of ties at the boundary points).
A rate-k/n convolutional code C is the set of sequences c(D) of q-ary n-tuples, which can be generated by a kxn generator matrix G as its input ranges over all sequences of q-ary k-tuples. In analogy to the fundamental region of a ~~~vJ~a~
W091/15900 - 49 - PC?lUS91/02384 lattice, we can define a set R of sequences of q-ary n-tuples, which contains one and only one sequence (a coset representative) from every coset C8~2a(D) of C, where represents modulo-q addir_ion. Then, any sequence of q-ary n-tuples z(D) has the unique decomposition z(D) = t(D).~2c(D), where t(D)eR and c(D)eC. An unknown coset representative sequence t(D)eR_ can be recovered from z(D) = t(D)~2c(D), by first finding the code sequence c(D) using the decoder D(C,R) for the set of coset representatives k, arid then forming t(D) =z(D)~2c(D).
If A' is a sublattice of :1, then A is the union of ~A/A'~ = v(A')/v(A) cosecs of A' in A. Then we say A/A' is a lattice partition of order ~A/A'~, and each fundamental region R(:~') of A' will contain ~A/A'~ points from A.
In general, a signal-space trellis code C(,1/A';C) is based on an N-dimensional lattice partition ,1/.~' of order qn and a rate-k/n, q-ary convolutional code:C. Each coded output n-tuple bk selects one of the qn cosets A' + ~(b~) of A' in A, or some translate of A, according to some labeling function ~(bk). The trellis code _C(A/A';C) is the set of all sequences c(D) that belong to a sequence of cosets of A' that could be selected by the code sequences in C. To simplify exposition, we shall often consider two-dimensional Ungerboeck-type trellis codes based on 2n-way two-dimensional lattice partitions Z2/RnZ2, where Z2 is the lattice of integer pairs and R is the two-dimensional rotation operator defined by the 2x2 matrix {1.1). (1,-1)}. In sequence space, a lattice A can also ~30 be viewed as a trellis code _C(A/A; C) _ (A)~, which is based on the trivial lattice partition A/A and a trivial rate-1 convolutional code, whose generator matrix G is an identity. ,.

~~J~~=~~u .For all known codes C, given any encoder state sk, the set of all possible next outputs is a coset of some lattice .10, called the time-zero lattice. The fundamental regions _R0 of the time-zero lattice play a central role in trellis precoding. For a two-dimensional code C_(Z2/RnZ2; C), ' the time-zero lattice is AO = Rn-kZ2.
If C is a linear code, i.e., a group under sequence .
addition, then it will carry the key properties of lattices.
For example, we can define a fundamental region _R(C) of _C as a set of sequences that contains one and only one sequence (a coset representative) from each distinct coset C + a(D) of C.
Also say any sequence y(D) can be uniquely decomposed as y(D) _ c(D) + e(D), where c(D)eC and e(D)eR(C). Thus, a coset representative sequence e(D)e R(_C) can be recovered from y(D) - c(D) + e(D), without any knowledge about c(D), by first finding the sequence c(D) using the decoder D(C,R) for the fundamental region R(C), and then forming the error sea_ue:.ce , e(D) = y(D) - c(D).
A trellis code is linear, if and only if the labeling function ~(b) is linear; i.e., ~(bAb') _ ~(b) +
~(b'), where b, b' are two q-tuples. Most known trellis codes, although nonlinear, exhibit sufficient regularity such that their geometric properties such as distance profile or the shape of their Voronoi regions are independent of the code sequence c(D). All trellis codes with an isometric labeling function exhibit such geometric uniformity. For such codes;
the decision regions for some given decoder are not simple translates of each other. But they can be transformed into another by isometric transformations, which consist of translations, rotations or reflections. The following theorems play an important role in trellis precoding.-Theorem 1: Let Cs be a trellis code,.possibly nonlinear, with a time-zero lattice AOs, then any sequence y(D).can be uniquely decomposed as y(D) = e(D) + cs(D), where , r WO 91 / 15900 - ~ 1 - PCT/ C.'S91 /023$4 cs(D)EC_s and e(D) is a cosec representative sequence whose elements belong to some fundamental region ROs of '~Os' An unknown coset representative sequence e(D) can we recovered from y(D) by First finding the code seque~ce cs(D) with a hard-decision decoder, based on R~s and then forming the error sequence e(D) = y(D) - cs(D).
Proof: Suppose at time k we are in state sk of Cs. Then the next allowable symbols cs,k belong to some coset AOs+a(sk) of the time-zero lattice ~10s. Hence, we can use a decoder D(ROs) for AOs+a(sk) to obtain the decomposition yk = ek+cs,k, where cs,k~AOs+a(sk) and ekeR~s. Next, we let cs k determine the next-state sk+1 and then repeat the same argument for time k+1, and so on. Thus, starting from a kown state s0, the sequences e(D) and cs(D) in the decomposition y(D) = e(D) + cs(D) will be uniquely determined. Q.E.D.
Theorem 1 implies that if e(D) is some information-bearing signal whose symbols ek are constrained to lie in the fundamental region R~s, then we can add an unknown code sequence cs(D) from Cs to e(D), and we will be able to recover e(D) from the resulting sequence y(D) = e(D) +
cs(D) perfectly without the knowledge of cs(D). In fact, the above proof shows that the symbols ek can be recovered -with no delay from the symbols yk up to time k by a simple 'hard-decision' decoder for CS. Typically, ROs is chosen as a simple rectangular rega~an, and then decisions on cc,k may be made coordinate by coordinate.
If C(A/A'; C) is a trellis code, then its label translate C(A/A'; C+a(D)) is obtained by replacing the .
convolutional code C by its coset Ct~2(D). When C(A/.1';
C) is geometrically uniform, then C(A/:1' ; Ct~2a(D) ) has the same geometric properties as C(.1/A';C) and it is related to C(A/A'; C) by an isometric transformation (for linear codes that isometric transformation is a simple translation).

WO 91/15900 - 52 - PCT/L'S91/02384 Theorem 2: If cc(D) and cs(D) are code sequences in N-dimensional codes Cc and Cs, respectively, and .1S is a sublattice of :1c', then y(D) = cc(D)+cs(D) is a code sequence in Cc.
Proof: Every sequence cs(D) is a sequence of elements of .
As and therefore of nc', which means that y(D) _ cc(D)+cs(D) is a sequence with elements in the same .
sequence of cosets of ~c' as cc(D), so it must be in Cc. Q..E.D.
_ i

Claims (34)

CLAIMS:
1. A method of mapping a digital data sequence into a signal point sequence e(D) for data transmission over a channel with impulse response h(D) to produce a channel output sequence y(D)=e(D)h(D), comprising the steps of:
defining a class of possible sequences based on a time-varying trellis code derived from an N-dimensional time-invariant trellis code C by using different transformed versions of its underlying lattices for respectively different N-dimensional symbols, and choosing e(D) so that the signal points in the sequence y(D) belong to said class of possible sequences based on said digital data sequence.
2. The method of claim 1 wherein said class comprises a signal space translate of a time-varying trellis code Cs, where the translation is a code sequence from a translate of a trellis code Cs, determined based on said digital data sequence, and said time-varying trellis code Cs is derived from an N-dimensional time-invariant trellis code C(A/A', Cs) by using different transformed versions of its underlying lattices A and A for respectively different N-dimensional signal points.
3. The method of claim 1 wherein said class comprises a signal space translate of a trellis code Cs, where the translation is a code sequence from a translate of a time-varying trellis code CC determined based on said digital data sequence and said time varying trellis code Cc is derived from an N dimensional time-invariant trellis code C(L/L', Cc) by using different transformed versions of its underlying lattices L and L' for respectively different N dimensional signal points.
4. The method of claim 1 wherein said class comprises a signal space translate of a label translate of a time-varying trellis code Cs, where the label translate is based on a portion of said digital data sequence and the signal space translation is based on another portion of the elements of said digital data sequence, and the time-varying trellis code Cs is derived from an N-dimensional time-invariant trellis code C(A/A', Cs) by using different transformed versions of its underlying lattices A and A' for respectively different N-dimensional signal points.
5. The method of claim 1 wherein said class comprises a signal space translate of a label translate of a trellis code Cs, where the label translate is based on a portion of said digital data sequence, where the signal space translation is based on another portion of the elements of said digital data sequence and is a code sequence from a time-varying trellis code Cc which is derived from an N-dimensional-time invariant trellis code C(L/L',Cc) by using different transformed versions of its underlying lattices L and L' for respectively different N-dimensional signal points.
6. The method of claim 1 wherein said class of possible sequences is identified by some sequence of subregions into which N-space has been partitioned, said sequence of subregions being specified by a label sequence determined based on a portion of the digital data sequence and said signal points in said subregions belonging to a coset of a time-varying trellis code Cc, where the sequences from said time-varying trellis code Cc are chosen based on another portion of the elements of said digital data sequence, and the time-varying trellis code Cc is derived from an N-dimensional time-invariant trellis code C(L/L', Cc) by using different transformed versions of its underlying lattices L and L' for respectively different N-dimensional signal points.
7. The method of claim 1 wherein said class of possible sequences is identified by some sequence of subregions into which N-space has been partitioned, said sequence of subregions being specified by a label sequence, which belongs to a coset of convolutional code Cs, based on a portion of the elements of the digital data sequence and using different scaled versions of said subregions for respectively different N-dimensional signal points, said coset being determined based on another portion of the digital data sequence, where the signal points in the sequence y(D) are chosen based on a portion of the elements of said digital data sequence, and wherein said subregions are scaled versions of each other.
8. A method of mapping a digital data sequence into a signal point sequence e(D) for data transmission over a channel h(D) (h(D) not a constant) to produce a channel output sequence y(D)=e(D)h(D), comprising the steps of:
defining a label translate of a trellis code Cs based on a portion of said digital data sequence, and choosing e(D) so that y(D) belongs to a signal space translate of said label translate, said signal space translation being specified based on another portion of the elements of said digital data sequence and being a code sequence from a trellis code Cc.
9. A method of mapping a digital data sequence into a signal point sequence e(D) for data transmission over a channel h(D) (h(D) not a constant) to produce a channel output sequence y(D)=e(D)h(D), comprising the steps of:

defining a label sequence based on a portion of the elements of the digital data sequence, specifying a sequence of subregions into which N-space has been partitioned based on said label sequence, signal points in said subregions belonging to a coset of a trellis code Cc, and choosing e(D) so that the signal points in the sequence y(D) belong to said specified sequence of subregions, sequences from said trellis code Cc being chosen based on another portion of the elements of said digital data sequence.
10. The method of claim 6, 7 or 9 wherein said label sequence belongs to a coset of a convolutional code Cs.
11. The method of claim 1, 2, 3, 4, 5, or 6 wherein said transformed versions are respectively used periodically.
12. The method of claim 11 wherein each period in which said transformed versions are used periodically encompasses multiple N-dimensional symbols.
13. The method of claim 1, 2, 3, 4, 5, or 6 wherein only two said different transformed versions are used alternatingly.
14. The method of claim 1, 2, 3, 4, 5, or 6 wherein said transformed versions comprise rotated or scaled or rotated and scaled versions of said underlying lattices.
15. The method of claim 2, 3, 4, 5, 8, or 9 further comprising the step of filtering said trellis code Cs to form a filtered trellis code Cs, whose sequences are cs'(D)=cs(D)g(D) where g(D) is the formal inverse of a response related to the channel response h(D), and cs(D) is any code sequence in the trellis code Cs.
16. The method of claim 15 in which the signal point sequence selection tends to minimize the average power of the signal point sequence e(D)=y(D)g(D).
17. The method of claim 16 wherein said signal point sequence e(D) lies in a fundamental region of said filtered trellis code Cs'.
18. The method of claim 15 wherein said signal point sequence e(D) lies in a fundamental region of said filtered trellis code Cs'.
19. The method of claim 18 wherein said fundamental region of said filtered trellis code Cs' comprises a Voronoi region of said filtered trellis code Cs'.
20. The method of claim 15 in which the signal point sequence is selected based on reduced state sequence estimation with respect to said filtered trellis code Cs'.
21. The method of claim 20 in which the reduced state sequence estimation uses no more states than the number of states of the original filtered trellis code Cs'.
22. The method of claim 15 wherein said signal point sequence is selected by:
mapping said digital data sequence into an initial sequence belonging to and representing a congruence class of said trellis code Cs, and choosing a signal point sequence belonging to and representing a congruence class of said filtered trellis code Cs' and which has no greater average power than said initial sequence, and wherein said mapping includes applying a portion of the elements of said digital data sequence to a coset representative generator for forming a larger number of digital elements representing a coset representative sequence.
23. The method of claim 22 wherein said coset representative generatar comprises a multiplication of a portion of the elements of said digital data sequence by a coset representative generator matrix (H-1)T which is inverse to a syndrome-former matrix HT for said trellis code Cs.
24. The method of claim 23 further comprising the step of recovering the digital data sequence from a possibly filtered and noise-corrupted version of the signal point sequence, including decoding the signal point sequence to a sequence of estimated digital elements and forming a syndrome of fewer digital elements based on a portion of the estimated digital elements using a feedback-free syndrome former HT.
25. The method of claim 1, 2, 3, 4, 5, 6, 7, 8 or 9 further comprising the step of recovering the digital data sequence from a possibly noise-corrupted version of the signal point sequence, including decoding the signal point sequence to a sequence of estimated digital elements and forming a syndrome of fewer digital elements based on a portion of the estimated digital elements using a feedback-free syndrome former HT.
26. The method of claim 2, 3, 4, 5, 8, or 9 wherein said trellis code Cs is a linear trellis code.
27. The method of claim 25 wherein said linear trellis code C s is a 4-state Ungerboeck code.
28. The method of claim 2, 3, 4, 5, 8, or 9 wherein said trellis code C s is a non-linear trellis code.
29. The method of claim 2, 3, 4, 5, 8, or 9 wherein said trellis code C s is based on a partition of binary lattices.
30. The method of claim 2, 3, 4, 5, 8, or 9 wherein said trellis code C s is based on a partition of ternary or quaternary lattices.
31. The method of claims 1, 2, 3, 4, 5, 6, 7, 8, or 9 further comprising the step of selecting said signal point sequence in a manner which is constrained so as to reduce the peak power of said signal point sequence where said peak power represents the maximum energy of said signal point sequence in some number of dimensions N.
32. The method of claim 31 wherein N=2.
33. The method of claim 31 wherein N=4.
34. The method of claim 1, 2, 3, 4, 5, 6, 7, 8, or 9 further comprising the step of mapping said digital data sequence into said signal point sequence in a manner arranged to ensure that said digital data sequence can be recovered from a channel-affected version of said signal point sequence which has been subjected to one of a number of particular phase rotations.
CA002058450A 1990-04-06 1991-04-08 Treillis precoding for fractional bits/baud Expired - Lifetime CA2058450C (en)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
US505,418 1990-04-06
US07/505,418 US5214672A (en) 1990-04-06 1990-04-06 Trellis precoding for fractional bits/baud
PCT/US1991/002384 WO1991015900A1 (en) 1990-04-06 1991-04-08 Trellis precoding for fractional bits/baud

Publications (2)

Publication Number Publication Date
CA2058450A1 CA2058450A1 (en) 1991-10-07
CA2058450C true CA2058450C (en) 2004-06-29

Family

ID=24010227

Family Applications (1)

Application Number Title Priority Date Filing Date
CA002058450A Expired - Lifetime CA2058450C (en) 1990-04-06 1991-04-08 Treillis precoding for fractional bits/baud

Country Status (4)

Country Link
US (1) US5214672A (en)
EP (1) EP0476125A4 (en)
CA (1) CA2058450C (en)
WO (1) WO1991015900A1 (en)

Families Citing this family (38)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1992017971A1 (en) * 1991-03-28 1992-10-15 British Telecommunications Public Limited Company Tcm scheme with fractional bit rates, framing signals and constellation shaping
CA2106848C (en) * 1991-03-28 1998-12-01 John David Brownlie Frame synchronisation for qam
US5216694A (en) * 1991-09-12 1993-06-01 At&T Bell Laboratories Trellis coding for fractional bits
US5280489A (en) * 1992-04-15 1994-01-18 International Business Machines Corporation Time-varying Viterbi detector for control of error event length
US5388124A (en) * 1992-06-12 1995-02-07 University Of Maryland Precoding scheme for transmitting data using optimally-shaped constellations over intersymbol-interference channels
US5424881A (en) 1993-02-01 1995-06-13 Cirrus Logic, Inc. Synchronous read channel
US5373460A (en) * 1993-03-11 1994-12-13 Marks, Ii; Robert J. Method and apparatus for generating sliding tapered windows and sliding window transforms
US5418531A (en) * 1993-10-20 1995-05-23 At&T Corp. Structured quantizer for sources with memory
EP0677967A3 (en) * 1994-04-12 1997-07-23 Gold Star Co Viterbi decoder for a high definition television.
US5559561A (en) * 1994-07-15 1996-09-24 Lucent Technologies Inc. Coded modulation with shaping gain and tomlinson precoding
US5548600A (en) * 1994-08-12 1996-08-20 International Business Machines Corporation Method and means for generating and detecting spectrally constrained coded partial response waveforms using a time varying trellis modified by selective output state splitting
US5742643A (en) * 1994-11-22 1998-04-21 Mikros Systems Corporation Quantizing and decoding of phase-amplitude modulated signals in hexagonal code
FR2741215B1 (en) * 1995-11-14 1998-01-23 Matra Communication METHOD FOR TRANSMITTING A SEQUENCE OF INFORMATION BITS WITH SELECTIVE PROTECTION AGAINST TRANSMISSION ERRORS, CODING AND CORRECTION PROCESSES WHICH CAN BE IMPLEMENTED IN SUCH A TRANSMISSION METHOD
JP3140974B2 (en) * 1996-03-31 2001-03-05 富士通株式会社 Judgment method and precoder device
DE19635813A1 (en) * 1996-09-04 1998-03-05 Johannes Prof Dr Ing Huber Process for reducing the peak value factor in digital transmission processes
DE69720712T2 (en) * 1997-08-05 2004-02-26 Sony International (Europe) Gmbh Circuit for regression of QAM symbols
US6563877B1 (en) * 1998-04-01 2003-05-13 L-3 Communications Corporation Simplified block sliding window implementation of a map decoder
US6269131B1 (en) * 1998-05-28 2001-07-31 Glenayre Electronics, Inc. Physical channel estimator
US6452978B1 (en) * 1999-02-11 2002-09-17 Tioga Technologies Inc. Combined shaping and peak-to-average reduction for single carrier modulation systems
KR100318912B1 (en) * 1999-04-28 2002-01-04 윤종용 Method and apparatus for state value normalization of component decoder in mobile communication system
US6408419B1 (en) 1999-07-01 2002-06-18 Infineon Technologies North America Corp. Trellis code for extended partial response maximum likelihood (EPRML) channel
US6415415B1 (en) 1999-09-03 2002-07-02 Infineon Technologies North America Corp. Survival selection rule
US6680980B1 (en) 1999-09-03 2004-01-20 Infineon Technologies North America Corp. Supporting ME2PRML and M2EPRML with the same trellis structure
AU7036100A (en) * 1999-09-13 2001-04-17 Fernway Limited A method for transmitting data between respective first and second modems in a telecommunications system, and telecommunications system
US7251270B2 (en) * 2000-06-20 2007-07-31 Paradyne Corporation Systems and methods for fractional bit rate encoding in a communication system
FR2830384B1 (en) * 2001-10-01 2003-12-19 Cit Alcatel DEVICE FOR CONVOLUTIVE ENCODING AND DECODING
US7502418B2 (en) * 2001-12-18 2009-03-10 Agere Systems Inc. Method and apparatus for joint equalization and decoding of multilevel codes
US6891902B2 (en) * 2002-07-02 2005-05-10 Intel Corporation System and method for adjusting a power level of a transmission signal
US7613985B2 (en) * 2003-10-24 2009-11-03 Ikanos Communications, Inc. Hierarchical trellis coded modulation
US7958425B2 (en) * 2004-02-19 2011-06-07 Trelliware Technologies, Inc. Method and apparatus for communications using turbo like codes
US7673213B2 (en) * 2004-02-19 2010-03-02 Trellisware Technologies, Inc. Method and apparatus for communications using improved turbo like codes
US7188302B2 (en) * 2004-04-14 2007-03-06 Realtek Semiconductor Corp. Parallel decision-feedback decoder and method for joint equalizing and decoding of incoming data stream
US7725807B2 (en) * 2006-01-23 2010-05-25 Vixs Systems, Inc. Error checking using a field of a frame
US8788921B2 (en) * 2011-10-27 2014-07-22 Lsi Corporation Detector with soft pruning
US8787477B2 (en) * 2012-09-18 2014-07-22 Texas Instruments Incorporated Methods for assigning codewoods to points in hexagonal constellations in digital communication systems
US9246510B2 (en) 2013-10-01 2016-01-26 Texas Instruments Incorporated Apparatus and method for multilevel coding in communication systems
US9467164B2 (en) 2013-10-01 2016-10-11 Texas Instruments Incorporated Apparatus and method for supporting polar code designs
US10135460B2 (en) 2013-10-01 2018-11-20 Texas Instruments Incorporated Apparatus and method for multilevel coding (MLC) with binary alphabet polar codes

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CH609510A5 (en) * 1976-06-18 1979-02-28 Ibm
US4586182A (en) * 1984-02-06 1986-04-29 Codex Corporation Source coded modulation system
US4713817A (en) * 1985-04-25 1987-12-15 Codex Corporation Multidimensional, convolutionally coded communication systems
US4959842A (en) * 1988-04-13 1990-09-25 Codex Corporation Signal constellations
US5048054A (en) * 1989-05-12 1991-09-10 Codex Corporation Line probing modem

Also Published As

Publication number Publication date
WO1991015900A1 (en) 1991-10-17
CA2058450A1 (en) 1991-10-07
US5214672A (en) 1993-05-25
EP0476125A4 (en) 1992-06-24
EP0476125A1 (en) 1992-03-25

Similar Documents

Publication Publication Date Title
CA2058450C (en) Treillis precoding for fractional bits/baud
US5159610A (en) Trellis precoding for modulation systems
US5249200A (en) Device and method for combining precoding with symbol-rate spectral shaping
US5455839A (en) Device and method for precoding
US5150381A (en) Trellis shaping for modulation systems
Eyuboglu et al. Trellis precoding: Combined coding, precoding and shaping for intersymbol interference channels
Duel-Hallen et al. Delayed decision-feedback sequence estimation
US5095497A (en) Technique for achieving the full coding gain of encoded digital signals
Padovani et al. Coded phase/frequency modulation
CA2164174C (en) Rotationally invariant trellis coding incorporating transparent binary convolutional codes
EP0485108B1 (en) Coded modulation with unequal levels of error protection
EP0566331B1 (en) Coded modulation using a combination of convolutional and block codes
US5953376A (en) Probabilistic trellis coded modulation with PCM-derived constellations
CA2142846C (en) Device and method for precoding
US4646305A (en) High speed data modem using multilevel encoding
CA2012914C (en) Trellis precoding for modulation systems
EP0383632B1 (en) Mapping digital data sequences
US7003716B2 (en) Method and apparatus for using multi-dimensional trellis codes over multi-path channels
EP1151588B1 (en) Termination of coded or uncoded modulation with path-oriented decoder
Carlisle et al. Performance bounds for trellis-coded modulation on time-dispersive channels
Akay et al. Bit-interleaved coded modulation: low complexity decoding
Shivakumar TCM: Code Design and Improved Sequence Estimation on Channels subjected to ISI
Fagan et al. Performance comparison of detection methods derived from maximum-likelihood sequence estimation
Clark et al. Detection of coded and distorted QAM signals
Barry et al. Signal-Space Coding

Legal Events

Date Code Title Description
EEER Examination request
FZDC Discontinued application reinstated
MKEX Expiry