Technical discussions related to Analog Devices DSPs (including Blackfin, TigerSHARC, SHARC and ADSP-21xx DSPs).
|
Hi I'm presently trying to get the radix-4 fft algo to work on my 21160 processor. I've downloaded the sourcecode from analog device, but the results I get is not what I expect. When having a DC input I get a AC spectral component, and when I use a sinewave as input it has an offset about 16 bins. Has anyone worked with this sourcecode and know what the problem is? Any comment appriciated Thomas K. Stoltz |
|
|
|
hi Thomas , while using the FFT on the 21160, have u checked for the address of the input vectors. There is a constraint on the bit-reversed access of the data. suppose u doing a N-point FFT, the address of the input should be a integral multiple of 0,N,2N etc. Then only data access using the i0 and i8 register will be proper. If this is not done u will get junk as the output. If u r using SIMD, make sure that the simd flag is set. we in our project have used the FFT routine and is working fine. I think u must be having the doucment describing the FFT routine. That contains the conditions which must be followed to use the FFT routine given. see if works and get back if there are any problem. with best regards sundi.. Sundararaman Sivaraman Sasken Communication Technologies Limited Bangalore, India. --- wrote: > Hi > > I'm presently trying to get the radix-4 fft algo to > work on my 21160 > processor. I've downloaded the sourcecode from > analog device, but the > results I get is not what I expect. When having a DC > input I get a AC > spectral component, and when I use a sinewave as > input it has an > offset about 16 bins. Has anyone worked with this > sourcecode and know > what the problem is? > > Any comment appriciated > > Thomas K. Stoltz > > _____________________________________ > /groups.php3 ===== __________________________________________________ |