Reply by Rick Lyons June 5, 20132013-06-05
On Sun, 02 Jun 2013 19:05:40 -0500, "rigor" <95223@dsprelated> wrote:

>Hi DSP people, > >Is there a way to do complex FFT (with imaginary components) with a real >only FFT alogarithm ? Thanks ! >
Hello rigor, perhaps the following will be of use to you: http://www.dsprelated.com/showarticle/97.php Good Luck, [-Rick-]
Reply by robert bristow-johnson June 3, 20132013-06-03
On Jun 3, 11:46&#4294967295;am, "rigor" <95223@dsprelated> wrote:
> Hi r b-j, > > Thanks ! > > I combine the outputs per the below and got the correct result. Is the > program I am having the real-input FFTs you called ?
i dunno, are you running MATLAB?
> A = fft(real(x)) > B = fft(imaginary(x)) > result = (real(A)-imaginary(B)) + i * (real(B)+imaginary(A))
... because in MATLAB, this should also work: result = fft(x) r b-j
Reply by rigor June 3, 20132013-06-03
Hi r b-j,

Thanks !

I combine the outputs per the below and got the correct result. Is the
program I am having the real-input FFTs you called ? 

A = fft(real(x))
B = fft(imaginary(x))
result = (real(A)-imaginary(B)) + i * (real(B)+imaginary(A))

Reply by robert bristow-johnson June 2, 20132013-06-02
On Jun 2, 8:05&#4294967295;pm, "rigor" <95223@dsprelated> wrote:
> Hi DSP people, > > Is there a way to do complex FFT (with imaginary components) with a real > only FFT alogarithm ? Thanks !
uhm, we need to talk a bit about what you mean by a "real only FFT". there are FFT programs that take N real values, x[n], going in and tightly packed (no zero-padding the imaginary part) and returns N/2 complex values, X[k], where X[0] is packed with the real part of X[0] and the real part of X[N/2] (the imaginary parts of those two components are known to be zero for real x[n]). for X[1] ... X[N/ 2-1], these values are understood to have complex conjugates that are not redundantly computed since x[n] is given as real. now, you can do a complex FFT with that alg by running it twice, once with the real part of x[n] and another with the imaginary part of x[n]. combining the outputs is no sweat. but here's the dumb thing about it: normally these real-input FFTs (i would not call it "real only" FFT) are *built* outa a regular complex FFT of half the size. so you separate the even-n samples of x[n] from the odd-n samples. the former goes into the real part and the latter goes into the imaginary part of N/2 complex samples and then you send that to a size N/2 FFT and fix the output. so you might already have a complex FFT that your real-only FFT is using. r b-j
Reply by rigor June 2, 20132013-06-02
Hi DSP people,

Is there a way to do complex FFT (with imaginary components) with a real
only FFT alogarithm ? Thanks !