Saddlepoint approximations of lower and upper bounds to the error probability in channel coding

Abstract

Saddlepoint approximations of the meta-converse and random-coding union bounds are derived. These bounds accurately characterize the channel coding minimum error probability for symmetric memoryless channels in a wide range of system parameters. The proposed approximations are simple to compute and yield a unified analysis of both hypothesis-testing lower bounds and random-coding upper bounds.