Probability of error versus SNR for random data in an AC-coupled baseband receiver

Long runs of one and zeros input to an AC-coupled baseband communications system can significantly degrade the signal level and therefore increase the bit error rate. Suitable coding or bit-stuffing can ensure that long run-lengths do not occur, but the data rate is decreased as a consequence. In this paper, the use of a randomizer as an alternative to coding is investigated. Assuming exponential decay in signal level with respect to run length, probability of bit error is derived as a function of input SNR and probability of run length. The results are compared to the theoretical best case. It is concluded that a randomizer results in very little degradation for commonly used receiver constants.<
>