The use of non-uniform modulation vector spacing to overcome sequential decoder computational deficiencies is investigated. Traditional convolutional (sequential) decoding techniques can provide arbitrarily low output error rates with a long constraint length code. However, bursts of noise or fades can overburden the decoder, as incoming data overflow the decoders (finite) buffer memory. A shorter constraint length code can be used to enable the decoder to compute through noise bursts, but the penalty is a higher rate of undetected output error. Another consideration of long constraint length codes is that "good" codes are not systematic; no coded bits contain clear information bits. As a result, these codes are not self-synchronizing. The proposed technique for imbedding a short constraint length code, using optimized modulation, promises to mitigate both problems of long constraint length sequential decoding. This is effected by spacing the modulated signal vectors representing the short code relatively far apart, while the signal vectors representing the long code are close together, In this study we propose to: (1) investigate the performance of this technique for various types of communication channels using analysis and simulation; and (2) derive an optimum signal set for this technique.The potential commercial application as described by the awardee: This technique can be used to improve the error performance of satellite, computer to computer, and other types of digital data communications. A standard codec module could be developed to replace existing CODECS.
