A Soft-Output Stack Algorithm

Soft output decoding of trellis-based problems, such as BCJR and the SOVA, has advantages in some applications. However these optimal methods are computationally intensive for all but short constraint lengths. A significant reduction in complexity can be obtained by using sequential decoding methods. This paper considers a modification of the stack algorithm to produce soft outputs and compares these with the soft values generated by the BCJR. The soft output stack provides around 1.5 dB of gain over the hard decision stack in some SNR regions. Also the soft values correlate highly with those generated by the BCJR and the Viterbi, especially at high SNR.