A new approach to the prediction of error rates in digital data transmission systems when the dominant source of errors is given by noise bursts, each having a fiat spectrum across the band of interest, is described. Specifically, a hierarchy of phaseaveraged conditional error probabilities called

th receiver impulse characteristics (RIC) are defined, which depend only on the receiver\´s structure. Error probabilities are then computed by averaging the RICs over the statistics of the impulsive noise which is characterized at the RF input of the receiver. To illustrate the approach in a practical, interesting situation, a detailed analysis is carried out for a phase-shift keyed system employing either a linear or a hard-limiting receiver, and for impulse rates small enough that errors are essentially caused by single impulses.