++spectralcompand is a spectral version of the standard expander/compressor. Like ++spectralgate, it divides the frequency spectrum into many bands, and processes each band individually. Each band is processed with a combination expander and compression unit which can go smoothly from a compression ratio of 5:1 to an expansion ratio of 1:5.
This effect can be used in a similar fashion to ++spectralgate. However, since expansion and compression are based on smoothly changing gain, the effect is much more gentle and subtle than ++spectralgate. When in compression mode, it can be used to tame resonant frequencies; to diminish the fundamental; or for radical changes of timbre.
In expansion mode, ++spectralcompand becomes a highly tunable broadband noise remover, capable of removing hiss, hum and machine noise, without damaging the original sound.
threshold – This is the point at which ++spectralcompand turns on and off. When compressing, the effect is active when the sound is higher than the threshold. When expanding, the effect is active when the sound is lower than the threshold (this is known as “downward expansion”). The threshold can be varied from –120 dB to 0 dB. If the shaping line in the spectral display is used to give the threshold a complex shape, the threshold knob represents the average threshold of the shaping line.
ratio – The ratio controls the amount of compression or expansion. The range from 5:1 to 1:1 is the compression range, and from 1:1 to 1:5 is the expansion range. Compression reduces the dynamic range of the sound, resulting in a smaller difference between the loudest and softest harmonic. Compression affects only those harmonics which are above the threshold. Conversely, expansion increases the dynamic range of the sound, resulting in a larger difference between the loudest and softest harmonics. Expansion only affects those harmonics which are below the threshold.
attack and release – These controls specify how quickly the gate will turn on (attack) or turn off (release) after passing the threshold. They both can be varied from 0 to 1 second. A longer value will smooth the transition, making the transition less noticeable. A shorter value will make the transition more abrupt.
tilt – This control gives you an easy way to change the threshold shape, either by raising the high frequencies and lowering the low, or vice–versa. Tilt can go from a 6 dB per octave boost of the high frequencies to a –6 dB per octave cut of the high frequencies.
bands – This control changes the number of bands used for processing from 8 to 8192. This divides the frequency range linearly, with each band having the width of (sample rate x 1/2)/(number of bands). The latency will also change with this setting, if the DAW can respond to changing latency.
learnpeak – This button will set the threshold shape, using the largest (peak) sound between the current input sound and the previous threshold shape for each frequency.
learnrms – This button will set the threshold shape, using the average of the current input sound and the previous threshold shape for each frequency.
reset – This button resets the threshold shape to a straight line.
stereolink – This gives you the same threshold shape for both channels.
invert – This control allows you to invert the compression or expansion process and hear the difference between the processed sound and the input. This is very useful when using spectral expansion for broadband noise reduction, as you will be able to hear the sound that is being removed.
Suggested Uses For ++spectralcompand
The expansion and compression modes in ++spectralcompand have such different sounds, that it is like having two separate effects. The expansion mode is most useful as a broadband noise remover, able to flexibly remove hum or hiss. The compression mode can be used as a spectral flattener or reshaper, changing the spectral profile of a soundfile to another template.
Broadband noise remover – To separate and remove steady spectrum noise (e.g. hiss or hum) from a sound, you first need to play a portion of the noise alone. When this noise is playing, click either learnpeak or learnaverage. This will set the threshold to the level of noise for each frequency. Now set the ratio to 1:2, and raise the threshold until you hear the noise disappear. At this point, it is useful to click invert to hear what you are removing. Increasing the ratio will give you a more abrupt noise reduction. To better preserve transients in the original sound, change the attack and release to match the typical attack and release of the sound being affected. Usually an attack of around 0.0 and a release between 0.3 and 0.5 will sound most natural. Often you will need to perform more noise reduction on hum (50 – 60 Hz) than other frequencies. If you do, put the spectral display in log mode and try to draw a higher threshold for these low (leftmost on the window) frequencies. Clicking learnpeak or learnaverage again will add to the current noise threshold profile.
Spectral flattener and reshaper – Spectral compression can be used to flatten the spectrum, that is, to make all harmonics the same volume. First, set the ratio around 2:1 to 4:1 and lower the threshold slowly. Louder harmonics will be affected first, automatically lowered in volume to match the threshold amplitude. The gain may need to be increased to make up for the reduction in overall sound level. Lowering the attack and release will cause the sound to more quickly drop to the level of the threshold. Excessive flattening will give you nothing but noise, as all the harmonics will drop to the same level. In addition, if you use the learn buttons to grab a spectral shape, or redraw the shape by hand, you can reshape the sound to this learned shape.