Hi,

 I have tried some FFT programs (Split-Radix FFT for real-valued
inputs, C program) from a book:  "Signal Processing with Lapped Transforms",
writen by Henrique S. Malvar. The Split-Radix FFT algorithm is interesting
as it operates faster than other FFT algorithms I know.  However, I found
that the Inverse FFT program ( sub-rountines: void rsifft( float *x,
int logm ), static void rsirec( float *x, int logm ), etc. ) does not
work when the input vector length is greater than 8 ( eg. 16, 128, 256 ).
I have studied the source codes but could not locate any bugs from there.
Can anybody enlighten me on what has gone wrong with the programs?
Thank you.


22 Mar 1994