Non-linear signal conditioning is provided by signal processing in at least 2 consecutive steps (10). In each step (10) an insertion source signal (41) is provided, which is treated non-linearly to fulfil certain restrictions in bandwidth and/or spectral mask. The insertion signal is combined with a main signal (21) using a reinsertion factor (k1-kn), as a correction of the main signal (21). The reinsertion factor of at least one step is larger than the reinsertion factor of a previous step. In particular embodiments, the reinsertion factor increases for each of the at least 2 steps. This principle can be applied to most non-linear signal processing, and for example in PAR reduction systems, dynamic drain bias PA systems, Doherty and Chireix PA systems and in Dynamic Gate Bias systems.