An Easy-to-implement Coding Scheme for Multifrequency PPM

In implementing multifrequency PPM, a naturally arising question is: Let P be a fixed number; among all integer valued processes X1, X2, X3, with E ( X(n+1)-X(n) squared) less than or = P, which has the largest entropy? Earlier work by McEliece and Rodemich answered this question, but there is no obvious way to use this process to implement a code for multifrequency PPM. The present article describes an easy-to-implement process X1, X2, with E ( X(n+1)-X(n) squared) less than or = P, whose entropy is nearly as great as that of the McEliece-Rodemich process.