Psychoacoustic Compression

* PsychoacousticCompression.kym (There are more of David's compressors on the C&L Share)

It turns out that you can use the Kyma Compressors for psychoacoustic compression. This kind of compression corresponds more closely to the way we perceive loudness in music. It has many features similar to the classical compressors of yore that always receive so many rave reviews. These are characterized by compression that is strongest near the threshold, and tapers off to essentially no compression at the loudest levels.

This kind of compression is also similar to Dolby B and C encoding where preemphasis is applied only at the faintest levels in the sound. But unlike Dolby compression which sometimes uses a hard limiter in parallel with the direct signal, we use a compressor that has a negative compression slope of -1.67. That means that for every dB rise in signal level above threshold, the output level actually drops by 0.6 dB. Quite unlike typical compressors. Using the Dolby style hard limiter instead of negative compression results in a compression sound that is too harsh for our ears. Negative compression more closely matches our sense of musical sound.

When you combine this negative compression in a mixer with a delay compensated direct signal you get the classical ski-slope shaped compression on your signal, but this time, we have the curvature of that ski-slope matched to our human hearing which carries an exponent of nonlinearity of around 0.6 for sounds above 40 dBSPL (most all of what we normally listen to is above that level).

The Kyma Compressors are tricky to tune up, but they are wonderfully adaptable once you understand how the inner workings are mapping sound levels to gains. This simple circuit uses a negative value for compression equal to the negative inverse of that human exponent of nonlinearity.

-- DavidMcClain - 15 Sep 2004

[Note: I left the post gain adjustment up to your ears. But if you want to know what value to use for some specified gain at the threshold level, the expression for the gain value to use is as follows:

gain_to_use = (10 ** (gain_desired_at_threshold / 20) - 1) log * 20

So for example, if you want a boost of 12 dB at threshold, independent of threshold setting and compression ratio, you should use about 9.5 dB gain in the negative compressor. For a desired gain of 6 dB use 0 dB in the compressor.]

What utter nonsense!

It turns out that one gets far better results just by using conventional compression with a positive compression ratio, mixed into the direct sound. These negative slope compressors are even more harsh, often times, than using a hard limiter.

Interesting compression settings to try, mixed with direct sound, are

R = 2.76, Threshold = -50 dB, postGain = 0 dB,

along with fixup gain of 3 * (Threshold * (1 - Ratio inverse)) dB.

That fixup gain just ensures that we have postGain boost at the threshold.

By amplifying or attenuating the sidechain input to the compressor you can vary the apparent brightness at the low end of the loudness range. At high loudness levels of input, there is very little boost applied. All of the action occurs at lower loudness settings.

-- DavidMcClain - 23 Sep 2004

* PsychoacousticCompression.kym