CODEWITHSUNDEEP
iosobjective-cfftcore-audiovdsp
vlad.grb
vlad.grb

Reputation: 339

vDSP and FFT calculation

I am interesting in audio processing and I want perform FFT calculation. Many sources and tutorials says that vDSP is great. But I don't understand why size of real and imaginary part after calculation is N/2 ? How I can obtain 1024 real and imaginary values on 1024 samples. For example this service for calculation perform FFT correctly.

Upvotes: 0

Views: 283

Answers (1)

Jason B
Jason B

Reputation: 12975

When the input to the FFT is real-valued (like audio), the output is symmetrical about the midpoint (N/2+1), and therefore the second half of the output buffer is redundant. Most FFT algorithms worried about speed take advantage of this by not actually calculating those output samples, thereby reducing computation time by nearly half.

If you look at the vDSP reference, you will see that there are quite a few FFT variations. Those that take complex input will calculate all of the output samples since there is no such redundancy.

Upvotes: 6

Related Questions