Error detection in operating digital systems: influence of error distribution

Different statistical distributions of errors (isolated and in bursts) were provoked on digital systems. Binary errors and errors estimated by `coding rule violations¿ were compared for the HDB3 and 4B3T codes. In the case of the HDB3 code, 85% of isolated binary errors are detected. However, only 15% of burst errors are detected. In the case of 4B3T with an asynchronous scrambler, it may be considered that one code error corresponds to 10 binary errors, on average. Results with burst errors are quite different from theoretical evaluation for isolated errors. These results must be considered when evaluating `degraded minutes¿ and `severely errored seconds¿ of CCITT Recommendation G.821 for operating digital systems.