Using SDK'd RFFT and RIFFT for Filtering

Started by vo1hl March 22, 2003
Hi Everyone!

I am using SDK 2.4.1 with CW 4 on a DSP56F826 platform without
external memory. I have written a program that sucks in a 128 sample
vector at a sample rate of 8000 sps, submits this to dfr16RFFT then
takes the resulting dfr16_tInplace vector and submits it to
dfr16RIFTT to recover the time domain signal. I would have expected
this to result in a nice clean signal but I see what looks like
dropped samples at the boundaries of the sample vectors (ie. every 16
ms). This is apparent when I am sending a tone which is not
commensuate with the size of the sample vector. That is if I submit a
1000 Hz signal (1000/62.5 = BIN 16) then the output of the IFFT is
clean. However if I submit 1100 Hz (say) which will not "fit" into
the sample vector then I see the boundary problem.

Does anyone know if this is a limitation of the process? I would have
thought that all the time domain information would be recoverable
from an FFT frequency domain representation. And there ought not to
be discontinuities at the boundaries because I am sucking out the 128
sample vectors one after another so the edges should align.

What am I missing here?? Any help would be appreciated... 73,