Results of optimum time diversity for Gaussian and Rayleigh fading channels subject to pulse-burst interference of large power and duty-cycle λ are presented. For NC-BFSK and DPSK signals the Chernoff upper bound is used to obtain expressions of the optimum diversity and the average bit error rate. Both hard-decision and soft-decision decoding are considered. It is particularly shown that the optimum diversity obtained using exact analysis and the Chernoff upper bound are nearly equal, and that the results are valid for conventional as well as spread-spectrum systems. It is concluded that the optimum time diversity known as the repetition code can be used as an effective countermeasure against pulsed interference of high power. With optimum diversity, the exponential relation between power and bandwidth is recovered for both the Gaussian and Rayleigh channels. Using soft-decision decoding, the combined effect of pulse-burst interference and the noncoherent combining loss is a reduction in the effective signal/noise ratio by factors of (1 – λ0.25) and (1 – λ)2 for Gaussian and Rayleigh channels, respectively. For Gaussian channels, exact analyses show that hard-decision decoding results in an additional loss of 2 dB for λ → 0, increasing to 3 dB as λ → 1. Using the Chernoff upper bound, hard-decision decoding results in an additional loss of 3 dB for all values of λ. For Rayleigh channels with optimum diversity, hard-decision decoding is 3 dB inferior to soft-decision decoding.