Abstract: A copy of an input signal is clipped and subtracted from another copy of the input signal to generate an error signal corresponding to the clipped portion of the inpur signal. The error signal is filtered to generate a signal that is subtracted from another copy of the input signal to generate a filtered, clipped version of the input signal having a reduced peak-to-average power ratio. The frequency characteristics of the filtering match those of the input signal. For example, when the input signal has distinct frequency bands, the filtering preferably corresponds to a combination of band-pass filters, each corresponding to a different input frequency band. Because only the error signal and not the input signal itself is filtered, the resulting output signal can have a relatively low peak-to-average power ratio, while retaining frequency characteristics that more closely match those of the input signal.