kyma•tweaky . Share . DiscussPsychoacousticCompression

Following along the lines of the Dolby Model 740 Spectral Processor... Ray Dolby had it right for his Dolby-A Noise Reduction Encoding, but not quite right when used for listening to spectral enhancement...

If I may be so bold... I have spent the past 4 years intensively studying human hearing, and especially the loudness response of human hearing. I have been wildly successful in this quest, and I have a number of equations that guide my thinking. These equations are the result of numerous tests of everything from pitch flatting toward higher intensity sound levels, to the measurement of 3rd order and 5th order intermodulation distortion and the consequent nonlinearity of our hearing loudness response.

From these equations we can derive psychoacoustically meaningful compression curves that more closely fit our sense of loudness perception. Matching that sense means dealing with the inherent nonlinearity of our hearing.

We don't hear loudness in the dB amplitude scale. To whit... if you want a sound to appear about twice as loud, you must increase its amplitude by about 9-10 dB, not merely 6 dB. When you compress sound in the amplitude scale you get the familiar sounding compression effects. But when you compress in the Sones scale, which corresponds to how we perceive sound loudness levels, you get what this Sound provides.

Ray Dolby did amplitude compression in a most creative manner, in his Model 740, by adding a hard limited signal to itself and then selecting out various amounts from 3 bands to enhance the spectral components of only the lowest intensity sounds, while leaving the louder portions alone. His is compression in amplitude space and hence, it leaves the user with 2 parameters to adjust -- overall threshold, ranging from -60 to -40 dB, and gain in 3 bands ranging from 0 dB up to about 20 dB.

But now instead of doing that ad-hoc amplitude compression, suppose we did a psychoacoustically matched compression? This sound implements a 128 band Psychoacoustic Compressor (Spectral Enhancer)/Precision Limiter/Soft Clipper. There is now only one parameter to adjust for Spectral Enhancement, instead of two, called Intensity in the VCS. Most musically useful Intensity values lie between 0.2 and 0.5. An Intensity level of zero implies no spectral enhancement, while a level of 1.0 is severe and extreme -- intended only for the Industrial Techno Freaks out there...

What you will notice when listening through this sound, is that there is a fair amount of enhancement of the highest frequencies, so sounds appear much brighter now. Also, you should notice a greater amount of ambience from the recording. Like the Dolby Model 740, the compression curves resemble ski-jumps in their nonlinearity, providing essentially no gain alteration at the loudest levels, and the most extreme gain elevation at the lowest levels.

The other VCS parameters are used to manage the output precision limiter, namely, OutGain?, Limt, and Release. The LimitFB? is a readonly indicator to show you when the limiter is using feedback to further lower the limiter threshold in each of the 128 bands, and the RMS dial shows you an estimate of the current RMS sound level being sent out. The RMS meter uses a 300 ms ballistic, just like a VU meter. Whenever the sound level rises higher than Limit, some feedback attenuation is sent to the precision limiter to further reduce the output level to the specified Limit.

If you find your sounds constantly undergoing feedback compression, then try reducing the OutGain? a bit. No harm in letting your sounds limit to the extreme -- this merely squashes the spectrum and delays its roll-off to higher frequencies, and also generates mild amounts of 3rd and 5th harmonics inside the output soft clipper. You might actually like that sound.

The limiter has an immediate attack, but the release is controlled by the Release parameter, permitting up to 5 seconds of release tail. The precision limiter is quite transparent, having been perfected among the previous collection of Sound submissions.

-- DavidMcClain - 27 Mar 2004

Heh! I just pushed some modern dance music through this sound and... didn't hear any difference between enhancement and bypass mode...

No mistake here... It's just that if your sounds are already overly compressed, they never dip down far enough in loudness to allow this Crescendo Spectral Enhancer to do any work. In order to hear an effect you have to have some reasonable dynamic range in the input sounds. Modern dance music is cut to -0.1 dB and held within about 3-6 dB of that extreme level. Nothing left over to enhance!

-- DavidMcClain - 28 Mar 2004

----- Revision r1.2 - 28 Mar 2004 - 00:52 GMT - DavidMcClain
Copyright © 1999-2014 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding kyma•tweaky? Send feedback.