A simple method to improve the performance of convolutional vector symbol decoding with small symbol size

Vector symbol decoding (VSD) is a channel code decoding technique that works with nonbinary symbols. It has many good features and performance, but one main limitation is the need to use large symbol size. Prior works usually use at least 24-bit symbols and in most case the assumption is 32-bit symbols. This paper shows the effect of symbol size and also a simple method to improve the performance of relatively much smaller symbol size of 16-bit symbols. The result shows that the improvement in the decoding failure probability of 16-bit symbols is noticeable with the proposed method. However, it is still more suitable to use 24-bit or 32-bit symbols whenever possible since their performances are significantly better as expected.