Coco Las

I am beginning to study the fascinating world of gen ~ inside pfft and fft~ and I find that bin = freq / (samplerate / fttsize) +- corresponds to the bin (float) to that frequency. What is the simplest way to, given a frequency, produce the corresponding sinusoid sound using ifft (inverse discrete Fourier transform)?

I am beginning to study the fascinating world of gen ~ inside pfft and fft~ and I find that bin = freq / (samplerate / fttsize) +- correspon

ifft-frequency-oscillators

I've had this idea. I thought "Wouldn't it be cool if I could play an arbitrary number of sinewaves using a single ifft" Well I've asked this around on the comp.dsp forums. They basically say you'd be crazy to try it. And, in reality, all you have to do is play a super odd frequency sinewave into an fft and look at the pattern to realize it'd be really hard to create a function that looks like that.

I've had this idea.  I thought "Wouldn't it be cool if I could play an arbitrary number of sinewaves using a single ifft"  Well I've asked this around on the comp.dsp forums.  They basically say you'd be crazy to try it.    And, in reality, all you have to do is play a super odd frequency sinewave into an fft and look at the pattern to realize it'd be really hard to create a function that looks like that.  

Indeed, an ifft is akin to a set of sine wave oscillators (inverse FFT with FFT size 1024 is equivalent to a set of 512 sine wave oscillators).

To synthesize a given frequency, you put energy in the corresponding frequency bin (give some amplitude in the complex spectral domain). Then, you need to adjust the "phase" part of the complex spectrum. 2 PI covers the whole range of the frequency bin, and you add the number from -PI to PI proportionally to the offset between the frequency you want and the center of the frequency bin.

I recommend you read "The Phase Vocoder: a Tutorial", an article by Mark Dolson published in the Computer Music Journal in 1986.

Indeed, an ifft is akin to a set of sine wave oscillators (inverse FFT with FFT size 1024 is equivalent to a set of 512 sine wave oscillators).
To synthesize a given frequency, you put energy in the corresponding frequency bin (give some amplitude in the complex spectral domain). Then, you need to adjust the "phase" part of the complex spectrum. 2 PI covers the whole range of the frequency bin, and you add the number from -PI to PI proportionally to the offset between the frequency you want and the center of the frequency bin.
I recommend you read "The Phase Vocoder: a Tutorial", an article by Mark Dolson published in the Computer Music Journal in 1986.

IFFT frequency Oscillators