A modified version of the stack algorithm of sequential decoding particularly suitable for the decoding of high rate convolutional codes is presented. The modification consists in using a discarding threshold to eliminate the least likely branches for each path extension performed by the decoder. Computer simulation results with rates 2/3 and 3/4 codes over hard and soft-quantized memoryless channels

show that substantial stack memory savings and computational complexity can be achieved at hardly any cost in either error performance or computational variability

making sequential decoding a very attractive and practical alternative for the decoding of high rate codes. Furthermore

a simplified stack algoritm which exploits the periods of noiseless reception is described briefly. Finally

a new approach to the analysis of the average computational effort is presented.