On Apr 18, 2004, at 3:33 PM, Ross Bencina wrote:
//update sum sum -=data[count]; float next= ZXP(in); next *= next; data[count]= next; sum += next;This is not stable for floating point variables. Rounding errors can accumulate in sum.This was discussed in a thread on music-dsp last November under the topic"floating point running sum optimization".
That is why I suggested in that thread to convert to integer first before you store in the delay line. You then avoid the rounding issue entirely.