In this paper, we present a capacity analysis for broadband power line communication (PLC) channels, impaired by the Bernoulli–Gaussian impulsive noise, exploiting both orthogonal frequency division multiplexing (OFDM) and singlecarrier frequency-domain equalisation (SC-FDE) techniques. First, we investigate point-to-point communications and formulate the continuous signal model for OFDM and SC-FDE techniques in PLC environments and find the tight bounds derived for differential entropy of the non-Gaussian noise samples. Capacity bounds and a certain capacity result for pointto- point PLC systems are also computed. Then, we analyse the capacity in the presence of a relay terminal between the transmitter and the receiver. In this part, we assume correlated noises between the relay and the receiver and derive general upper bound and achievable rate using partial decode-and-forward coding strategy for PLC channels. We also show that the obtained bounds coincide in degraded and reversely degraded special cases. Finally, using two statistical divergences, we compare the distances of the Bernoulli–Gaussian and Middleton class A noise models from the normal distribution of their background part and show that the Middleton class A noise is nearer to the Gaussian background noise than the Bernoulli–Gaussian. Therefore, the capacity curve for Middleton class A noise is between the capacity for the Gaussian background noise and the obtained capacity for the Bernoulli–Gaussian model. The obtained capacity theorems are also illustrated numerically. Copyright © 2014 John Wiley & Sons, Ltd.

Keywords