Bounds on distances and error exponents of unit memory codes

Binary unit memory codes, originally introduced by Lee, are investigated. A few examples of constant unit memory codes are given and bounds on the distance profile and the free distances are discussed. For time-varying codes asymptotic lower bounds on the distance profile and the free distance are given. The error probability for the codes, used on a memoryless binary-input, output-symmetric channel, is asymptotically upper bounded. The asymptotic results for the free distance and the error probability, which are in some respects better than for conventional convolutional codes, are interpreted by Forney's inverse concatenation construction.