FPGA Central

Hi all,

I've been using matlab to analyse musical sounds using the Short-Term
Fourier Transform based on Matlab's fft implementation and also using
the CQT implementation by Brown
<http://web.media.mit.edu/%7Ebrown/cqtrans.htm>

I have been trying to determine the best set of parameters for the
analysis such than I can modify the frequencies in the sound and
recreate a new sound by invertnig the transform. I believed that as the
CQT, like the human ear, has a logarithmic frequency scale, the CQT
would produce a more accurate reproduction, but the STFT always seems to
produce better results. Am I doing something wrong? Is this a problem
with CQT itself? Is this a problem with the CQT implementation I am
using, and do you know of a better one for Matlab?

Regards,
Craig Nicol.