FFTCalc

The FFTCalc AO is used to get frequency content based on band energies.

Input block of length L along with previous N-L samples form a frame of length N (number of FFT points). This frame is multiplied with window function and fed to Real Radix 2 FFT.  The minimum and maximum FFT point supported is 32 and 4096 respectively. Thus, the minimum and maximum resolution for the sampling rate of 48000 is 1500Hz and 11.72Hz respectively.

The FFTCalc provides 4 fixed control outputs giving band information. Three are for band-specific info and one for the entire spectrum.

Design Time Parameters

• FFT size: an additional configuration parameter, should be selected to be power of 2. Other values are not accepted by GTT.
  • FFT size is lower bounded by the block length.
• Block Length: supports all the block lengths supported by the framework upto 4096.

The number of FFT points (N) and sampling rate decides the frequency resolution of the bin. When FFT points selected is lower the frequency bins may not correctly correspond to the designated band and they may fall into adjacent bands because of lower frequency resolution. It is required to correctly choose the number of FFT points. Frequency resolution is computed using the below formula
Frequency resolution = sampling rate/number of FFT points

We recommend choosing FFT points based on frequency resolution requirements.

For E.g if the sampling rate is 48KHz and we want a frequency resolution less than 100Hz  Number of FFT points selected can be greater than or equal to 512

• It is recommended to choose either Hamming/Hanning window for better frequency response.
• Currently, the FFTCalc object supports mono channel only

The AO supports in-place computation based on the core type.

Tuning Parameters