Thanks for your reply, Fred. I'm using these two functions: nspsRealFftNip(xTime, xFreq, 10, NSP_Forw); // forward nspcFft(xFreq, 10, NSP_Inv); // inverse The only reason I am not using nspcFft for the forward transform, thus keep it more consistent, is because my time domain values are real. nspcFft takes complex values.. Unless there's a way to transfer the 1024 real values to complex numbers?> >You didn't say which routine you're using in the Intel Signal Processing>Library. So, assuming that it is Fft specifically, it takes N complex >inputs and generates N complex outputs. Why make it more complicated? > >Fred > > >_____________________________________ Do you know a company who employs DSP engineers? Is it already listed at http://dsprelated.com/employers.php ?
fft and loss of samples
Started by ●March 30, 2007
Reply by ●April 1, 20072007-04-01
Reply by ●April 1, 20072007-04-01
"stefan_k" <stefan@cygnusnj.com> wrote in message news:2-2dnTp-6bmnypLbnZ2dnUVZ_hKdnZ2d@giganews.com...> Thanks for your reply, Fred. > > I'm using these two functions: > > nspsRealFftNip(xTime, xFreq, 10, NSP_Forw); // forward > nspcFft(xFreq, 10, NSP_Inv); // inverse > > The only reason I am not using nspcFft for the forward transform, thus > keep it more consistent, is because my time domain values are real. > nspcFft takes complex values.. Unless there's a way to transfer the 1024 > real values to complex numbers? >Yes. You assign zeros to the imaginary part and match whatever method is used to transfer the data into the routine. Sometimes complex numbers are entered into arrays real_1, imag_1, real_2, imag_2, ..etc. Think of your real time domain samples as complex with zero imaginary part..... Fred
Reply by ●April 3, 20072007-04-03
Yes, that seems to be working out almost fine. I'm assigning 0's to the imaginary part of the input samples, and I do get as many time domain samples at the end of the FFT/IFFT processing as I start with. However, for some reason, I am getting a little glitch about every 4000 values in the time domain values after the IFFT. Seems like an overflow or something is happening every 4 or so times I call the FFT with 1024.. That's a strange artifact, having in mind that I don't have anything happening periodically at 4000 samples time... Maybe it's something to do with the need of normalization of the output.. _____________________________________ Do you know a company who employs DSP engineers? Is it already listed at http://dsprelated.com/employers.php ?
