SourceForge.net wrote:
Comment By: Mathieu Bouchard (matju)
to save on CPU and memory, a "doubly real" FFT is possible: not only real inputs, but also real outputs. For a N-dimensional vector (block size N), there are 2N outputs. The way that it is normally reduced back to N, is to combine them to produce cosine and sine coefficients. I wonder whether that'd be a more appropriate way of dealing with the problem than using varying block sizes? (Maybe it's not)
but wouldn't combining sine and cosine coeffs into one signal block (instead of 2 half-redundant ones) break each and any patch that uses the fft-objects until now?
while packing the data into 1 signal block saves you memory and cpu, you lose the elegance of having 2 orthogonal data sets separated at hand (which imho is quite elegant)
mfg.asdr. IOhannes