0

I am trying to calculate an the DFT of an audio sample. I run into problems when I attempt to calculate the fft using non log2(N) when N is the number of sample pieces. I am currently using vdsp_fft_zripD(). In the solution is basically what I have. FFT using Accelerate. I need to replicate the FFT values from Matlab.

I know that Matlab uses FFTW3 library but would like to avoid using it if possible because I am writing an ios app using swift. I've created a wrapper for the vDSP in obj-c to make it a little easier. It does compute values correctly for all log2(N). I tried padding the input array with zeroes at the end but I don't get the correct results towards the end.

Community
  • 1
  • 1
Kendall
  • 17
  • 3
  • 1
    What value of N? What differences are you seeing? – hotpaw2 Apr 01 '16 at 05:49
  • So it turns out I've been overthinking the problem. I was seeing differences in the non abs() value version. I also realize that the magnitude would not be the same however the proportions did turn out the same for the frequencies. I still don't know exactly how Matlab gets the values for the untrimmed array but it is not important for me anymore. – Kendall Apr 04 '16 at 05:03

0 Answers0