Our apps
Sound Spectrum Pro for Windows
FFT (Fast Fourier Transform)
FFT display mode
- Shows the amplitudes (0 to peak) of the frequencies contained in the audio signal.
- Choose it from the Mode menu.
- The graph is scaled with reference to the Full Scale level (0dBFS). It is the maximum numerical value (=32767) for the input signal.
A larger signal will experience clipping and distortion ("CLIPPING" will appear on the graph when this happens).
- An amplitude with numerical value of 1 (1 bit) is at approx -90dBFS (20*log(1/32767) = -90.3).
- The quantization noise level is approx -98dB.
- Select View > Max peak to display the frequency and level of the highest peak.
- It is possible to enable (in View menu) the calculation of THD+N (Total Harmonic Distortion + Noise). It will assume the left-hand cursor is on the fundamental (you can use the internal tone generation). It is calculated from 20Hz to 20kHz.
FFT - the Fast Fourier Transform
- Is an algorithm to rapidly compute the Discrete Fourier Transform, used to perform Fourier analysis of a signal.
- It computes a list of amplitudes of frequencies (frequency domain) from the input signal (time domain - audio waveform in our case).
FFT parameters
- Input sample rate - The app will use the sample rate of the "default" Windows audio input. It's reported in the bottom left corner.
- Number of FFT samples (FFT size) - Determines the frequency resolution of all calculations (from 3Hz at size=16k to 0.05Hz at size=1M). Especially at low frequencies, having more frequency points brings visibly better accuracy.
- High values result in high load of the CPU, both for computing and for display in FFT mode. Depending on the speed and number of cores in your device, this might produce less smooth graph updates and/or slower response to pan/zoom.
- The app auto-adjusts the number of cores used for FFT.
- You can run a speed test in the Help menu to see how fast is your device. Results less than 40ms are needed to maintain the frame rate (depending also on the graphics speed).
- Recommended values are 64k or 128k for normal use. Larger values are useful only when increased accuracy is needed, only for FFT display mode (RTA will not benefit significantly).
- Window type - The FFT accuracy is also affected by the window applied to the audio samples.
You can best see the window effect by starting a sine-wave and enabling loopback. Some hints on usage:
- Rectangular - when the input signal is noise, and in synchronous mode (any signal)
- Hann, Hamming or Blackman - generally good compromise
- Nutall, Blackman-Nutall or Blackman-Harris - good dynamic range (small signals close to large signals)
- Flat-top - good amplitude accuracy
- Smoothing time constant - Higher values give more readable (slow) graphs and reduce noise, lower values give more lively real-time graphs.
- For values up to 10 seconds, an exponential smoothing is applied, coupled with a matching increase of the window width.
- The values from 25 seconds up enable a special accurate mode, with maximum frequency resolution.
In this mode there is no exponential smoothing and the window size is always the full size of the FFT.
The full accuracy is reached after waiting at least the indicated time.
- The values from 50 seconds up use under-sampling to further increase the frequency resolution without increasing the computation time.
The trade-off is the decrease of the maximum frequency.
With a setting of 100 seconds, the accuracy (FFT bin size) is 0.01Hz. Please note that this may be affected by the device's quartz precision and jitter.
See also the measurement page.
Copyright © 2012-2022 Lindentree. All rights reserved.