Spectrum Filter

Menu Command: Effect | Filter | Spectrum Filter

The Spectrum Filter is a general purpose audio filter similar to a parametric equalizer, but with much greater control. Instead of using individual bands, the entire frequency spectrum is controlled using a shape line that controls the gain. This allows many kinds of filters to be designed, such as lowpass, highpass, bandpass, bandstop, notch, peak, comb, and more. Filtering is performed in the frequency domain using Fast Fourier Transforms (FFTs).

A spectral analysis window is displayed with a shape line and several other controls. The X and Y coordinates are updated when you click-and-drag a shape point. The X coordinate is the frequency in Hertz and the Y coordinate is the magnitude in decibels. The time of the spectral analysis shown is given in the Time box. If you move the time scroll bar, located below the analysis window, the graph changes to show the spectral analysis of a different part of the sound. The Master gain controls the overall gain of the filter.

Initially the shape line is horizantal at 0dB, which means that no changes in gain are made at any frequency. Alter the shape line up or down to increase or decrease the gain at a particular frequency. In technical terms, the shape line represents the frequency response function.

Graph Range
Min and Min control the range of values for the y-axis in decibels. Setting Min to -5 and Max to 5, for example, sets the graph to show a narrower range between -5dB to 5dB. That allows shape points to be set more precisely with the mouse within that range.

FFT Settings
The FFT size determines the detail of the spectral analysis and the resolution of the filter. Higher settings provide a higher resolution, allowing the filter to follow the shape more accurately, with sharper cutoffs. When processing high sampling rate files, such as 88kHz or 192kHz, the FFT size must be set higher for the filter to follow the shape. Using too high a value may cause overshoot and oscillations in the gain. The Overlap value specifies the amount of audio to overlap from one calculation to the next. The lowest value gives the fastest processing and generally works well.