Stream FFT and iFFT

[martinvicanek]

Cool update, MyCo! I was thinking that the FIR filter should be windowed rather than truncated. That's actually well known, see for instance here. Below is an implementation (I hope the modules are synced all right this time). The effect is only perceivable at extreme low-pass filter settings, however I think the improvement is worthwhile.

[tester]

Question.

Will it work without issues at higher resolution, i.e. 64kpts?

Because I see, the clock is set to allow to use 4 offsets, which I assume should be then set to 0, 0.25, 0.5 and 0.75.

Currently I configured it into stereo device, but I would add one unit more to get higher resolution.

Input array to mix with spectral peaks - would use 1/4 of sample resolution anyway (about 11kHz), so no problem with green arrays larger than 16kpts.

[Youlean]

Does anybody knows how to make from mag to real and imag? It would be opposite from this?

[KG_is_back]

The schematic contains both real/imag->mag/phase and mag/phase->real/imag conversion modules. Just check in the module called "components".

Note that both amplitude and phase information are needed to reconstruct the complex number properly.

[Youlean]

The schematic contains both real/imag->mag/phase and mag/phase->real/imag conversion modules. Just check in the module called "components".

Note that both amplitude and phase information are needed to reconstruct the complex number properly.

Thanks!

I am trying to make FFT limiter. I tried it by converting real and imag to mag and phase, then limit the mag, and convert back to real and imag. Problem is that it gives me some echo in background. Do you, or anybody else knows how to fix that?

Here is the mp3 example: https://drive.google.com/folderview?id= ... sp=sharing
Regards

[KG_is_back]

Hello, I was wondering if it would be possible to adjust this FFT module, that would read/write data from/to memory. Having an array, that has Real values on even slots and imaginary values on odd slots is not very practical. Maybe having two arrays one for real and one for imaginary would be more practical, as it would let us process the arrays with SSE without rearranging the inputs first.

I tried it myself but I failed miserably.

[tester]

Hmm... There is something wrong with this FFT/iFFT setup (windowing functions). The only windowing function that works correctly (comparing to doing the same thing in cool edit FFT) is the rectangular one (at least at 32kpts). Everything else either does not produces output, or the sound output sounds somewhat half-step modulated, as if only one part was sending out (but windowing steps seem to be sent interchangeably). Am I missing something?

//edit:
I mean - windowing functions should affect only the sharpness of audible details; it should not introduce modulation.

[RJHollins]

testing out v11a ...

Seems WAV files are not playing, only MP3

Why is that ?

[Lex248]

Hi! Sorry, i bad speak english. It is version with a logarithmic scale and Hanning window only.

+ "Log to Lin to FFT" module on Ruby code.