Search

A single convolutional code of fixed rate can be punctured to form a class of higher rate convolutional codes. The authors extend this pragmatic approach to the case where the core of the trellis decoder is a Viterbi decoder for a punctured version of the de facto standard, rate 1/2 convolutional code. >

A new method for the performance evaluation of (Viterbi decoded) trellis codes is investigated. The method is based on the fact that the a posteriori probability of individual decoded bits can be computed very efficiently. Unfortunately, the improvement of the new method over ordinary simulation is found to be only marginal, which reveals unexpectedly strong fluctuations of the reliability of individual decoded bits. >

A class of trellis codes is introduced based on a novel trellis structure. Codes are constructed by mapping coefficient sets onto branches in the trellis. Sets are given an algebraic structure by operations in a ring which will generate a rich class of codes and allow linear codes to be defined. Current interest is in finding codes with good performance and a reasonably large data rate. >

This paper presents realizations of entropy-constrained trellis coded quantisation as arithmetic coded trellis coded quantisation. The block diagram of the arithmetic and trellis coded quantisation (ATCQ) is shown. The decoder, encoder and the performance of the ATCQ system are examined. >

In this paper, the construction of rotationally invariant trellis codes with multidimensional signal constellations is discussed. Linear and nonlinear trellis codes with constraint lengths /spl nu/=2-8 are found using a systematic search. Simulation results show that a /spl nu/=8, linear, 180/spl deg/ rotationally invariant code and a /spl nu/=8, nonlinear, 90/spl deg/ rotationally invariant code achieve 0.4 dB and 0.3 dB real coding gain, respectively, compared with the best /spl nu/=6 code, while the decoding complexity is only four times that of the /spl nu/=6 code. By contrast, the /spl nu/=6 code that was adopted for use in the V.34 28.8-kbit/s modem standard requires sixteen times the complexity of the best /spl nu/=4 code, while the gain is also only about 0.4 dB. Thus, the new /spl nu/=8 codes would be a good choice for use in the V.34 modem standard. >

An implementation of a 16 state, rate 8/9 six-dimensional (6-D) 8PSK rotationally invariant trellis decoder for use in a concatenated codec is described, The concatenated codec allows transmission of STM-1 signals (at the 155.52 Mbit's information rate) over a 72 MHz satellite transponder. The inner trellis decoder is used with an outer (255,239) Reed-Solomon (RS) block decoder. The trellis decoder operates at 165.93 Mbit/s and currently has an implementation loss of only 0.2 dB. The concatenated codec achieves a bit error ratio of 10/sup -10/ at an E/sub b/N/sub 0/ of 8.2 dB (assuming an ideal modem and AWGN channel). >

Noting that a trellis code is a union of sub-block codes of different length, the error exponent of the families of trellis codes with known distance distribution of its sub-block codes is determined. The approach allows the analysis of the effect of memory length and of parallel transitions. >

We describe techniques from symbolic dynamics to determine the symmetry system of a trellis code. The construction starts with the partition permutation group, G, and enumerates short sequences over G under which the code is invariant. These sequences are then used to generate a graph. The process terminates when either the image of the symmetry graph is the entire code C (in which case C is known to be geometrically uniform), or the length of the generators exceeds a bound derived from the memory of C. As an example of using these techniques we analyze the V.34 codes. >

A new decoding algorithm for geometrically uniform TCM schemes is proposed, based on the group properties of code sequences and trellis states, Its complexity is almost independent from the number of states of the trellis, and, for a given code, a trade-off between complexity and performance of the algorithm can be chosen. >

We construct a class of noncatastrophic binary (d, k) trellis codes with large free distances/sup 1/, using the structure of a class of (d, k) trellis codes proposed by Lee and Wolf in 1989. However in our design, two disjoint Steiner triple systems are used as frames to select two disjoint subsets from a maximal concatenatable (d, k) block code. >

Maximum likelihood sequence estimation (MLSE) of trellis coded modulation (TCM) signals affected by intersymbol interference (ISI) requires an optimal path search on a trellis diagram usually of a very large size. This optimal search can be performed in principle by a Viterbi algorithm, but is impractical due to the large number of trellis states. Among the various suboptimal alternatives, we concentrate on reduced state sequence estimation (RSSE), in which the trellis diagram is reduced in size by partially representing the ISI characteristics into the trellis diagram itself. A Viterbi algorithm operates on the reduced trellis diagram, taking into account the effects of residual ISI components by means of per-survivor ISI cancellation. This technique entails the cancellation of residual ISI terms, necessary for computing the branch metrics, independently on each survivor path using the code sequence uniquely associated to the survivor path itself. >