I wonder what the rules of thumb are for scaling input & output to the fft routine in CLM. My friend here tells me that this scaling normally is taken care of inside the routines - whether in the fft or in the inverse-fft. Any guidelines? Anders Vinjar