The multiple stack algorithm (MSA), devised by Chevillat and Costello,
is an efficient algorithm for erasurefree decoding of long constraint
length convolutional codes. rn the MSA, substack size and the number
of transferred survivors (or successors) are assumed to be small. Lowe
r error probabilities can be achieved by increasing the first stack si
ze and/or increasing the computational limit. A large storage capacity
for survivors is required to prevent memory overflow and achieve a lo
w error probability. The authors present a modified MSA, in which the
storage capacity for survivors is kept as a constant, while the substa
cks are arranged in a ring-like structure to handle the overflow probl
em of storage for survivors. In addition, the substack size and the nu
mber of transferred survivors are made large to improve performance. T
he performance of the modified MSA in decoding a convolutional code wi
th constraint length m = 23 is investigated and compared with the perf
ormance unmodified MSA.