Trellis codes for transition jitter noise

We present a trellis code which reduces the probability that transition jitter will cause one allowed sequence to be decoded as another allowed sequence. This code requires that transition times mod 4 be in the code of a 4-state convolutional encoder. We also exclude adjacent transitions (d=1). Sequences satisfying this constraint have capacity of approx. 4/9. This code can be implemented using a finite state encoder, Viterbi detector and sliding-block decoder. In the presence of pure jitter noise this code realizes a gain of approximately 3 dB over uncoded recording at the same user density.