Abstract: In a method and an amplifier for the compensation of unlinearities e.g. of the class D type, wherein an audio signal is pulse-width modulated, e.g. with a carrier wave signal in the form of a triangular signal to provide a pulse-width modulated small-signal, the so-called multiplicative error signals, which occur prior to the provision of a pulse-width modulated great-signal (7), are detected in a detector (10)). It is noted that the carrier wave signal could be analog as well as digital. The signal from the detector, which is derived on the basis of differences between the pulse widths of the small-signals and the pulse widths of the great-signals, is used for changing the carrier wave signal so that the amplifier gets a constant gain in the entire audio range and is thereby linearized.