If you had 16 samples you could just do 4 "RR A" instructions on the sum of the samples. That would divide by 16 rather than 17. If there's any way you can work with 16 samples instead of 17, that'd be your easiest way out.

Another way is to divide each 8-bit sample by 17 before adding it to your sum total. Of course, you're going to lose precision that way.

Otherwise, you'll have to do 16-bit by 8-bit division. There are other posts on this board about the subject.

Craig Steiner