#FAST FOURIER TRANSFORM SERIES#

the number of samples has to be a power of 2), because it uses some clever tricks to drastically reduce the amount of calculation performed on the sample buffer. It imposes some constraints you have to work with in your implementation (e.g. So in this case the 1024 digital 'oscillators' should actually be measured after 1/44th of a second, given the audio is sampled at the rate of 44kHz.įast Fourier Transform is an algorithm to perform a Discrete Fourier Transform that's pretty easy for computers to run on an incoming signal. Of course in explaining the DFT, the speakers part is not really appropriate since you have to work on sampled input.

#FAST FOURIER TRANSFORM SERIES#

Instead of visualising the sound as amount of air pressure caused by the waveform, changing in time slots, you visualized it as a series of intensities of the frequency ranges. As a result you get a discrete fourier transform, meaning you get a chart of how much each of the frequency ranges contributed to the sound coming from the speaker. After the said second you read how much every oscillator is resonating.

Oscillators begin to resonate to the sound coming from the speaker. You then set up, let's get a nice round number here, 1024 harmonic oscillators that resonate to specific frequency ranges. Say you have a sound coming from the speaker.

As for the first part of the question, a former Physics professor, Bartosz Milewski, has a very nice explanation, what FFT is and how it works.Īlso, Mastering The Fourier Transform in One Day is worth reading as well.

Fast fourier transform