Trellis coding is investigated for a class of

-user multiple-access channels. It is shown that

-user trellis coding, in conjunction with Viterbi or sequential decoding, permits the multiple-access function to be combined with forward error correction. The distance and rate properties are analyzed for linear (convolutional)

-user trellis codes, and the use of nonlinear

-user trellis coding is discussed. A procedure is given for constructing

-user convolutional-code

-tuples with large free distance. Finally the error performance is analyzed, and coding gains of several decibels over uncoded time sharing are verified by simulation results.