Tom Scholz's Compressor with two release time constants

[Vivek]

In patent 4627094 of december 2. 1986, Tom Scholz described a guitar compressor.

He also highlighted an issue with the release times of the compressor.

Tom described the ADSR curve of a guitar signal as follows:

"On a guitar, the first sound or pulse that comes out can be a huge peak which is almost always much stronger than the signal that follows after a few milliseconds."

Then he described the problem with his earlier versions of the compressor:

"The operation has not been found to be optimised....When a loud note is followed by a soft note, the transition does not follow (correct compression) curve exactly due to the time constants. The (compressor) tends to act as a fixed gain amplifier when a hard note is quickly followed by a weaker note; there is a tendency for the weak note to start quietly and ride up to the previous note volume"


If the compressor has one slow release time
Lots of hard notes will raise the voltage on the integrator cap
Then if the next note is soft, but the integrator cap is fully charged and discharging very slowly, it will keep the gain of the compressor low even though the current note is low volume.
As the integrator cap slowly discharges, the compressor gain starts to increase and the weak note starts to get louder.

Basically the issue was that a fast release is needed initially, followed by a slow release. Ie a compressor with two release time constants. This would track the amplitude changes in guitar signals in a better way.

Which obviously means two discharge paths for the integrator cap.

Tom solved that problem very elegantly in the Sustainor and Ultimatum by adding various diodes in parallel to the integration capacitor.



Here the output signal of an Opamp goes to R1 (which sets the attack time)

D3 rectifies the signal and charges integration Capacitor C1.

C1 normally discharges via R3

Tom placed diodes D4-D7 as the additional discharge path. He took advantage of the nonlinear VI curve of diodes to basically make a variable resistor, the value of which depends on the imposed voltage.

When C1 has a higher voltage, the diodes D4-D7 appear as lower resistors and discharge C1 at a fast pace.

When C1's voltage drops, the diodes appear to be higher value resistors, and C1 continues to discharge via R3 as usual.

This leads to 2 release time constants, a fast one initially and a slow one later on.



The green curve is a sine wave input signal that starts at t=0.5s and stops at t=1.5s

The red curve shows the exponential discharge of C1 in case there were no D4-D7. It is basically a discharge via R3 with time constant R3.C1

The blue curve shows the discharge of C1 with D4-D7. We can see around t=1.5 till 3.5s, the blue curve shows faster discharge than the red curve. That is because initially C1 is discharging via D4-D7 since they appear to be low value resistors.

However after about t=3.5s, the time constants of the blue and red curves appear to be similar. That is the zone when the diodes D4-D7 appear to be high value resistors.

[Rob Strand]

"On a guitar, the first sound or pulse that comes out can be a huge peak which is almost always much stronger than the signal that follows after a few milliseconds."

It's common problem and there's a few mindsets to think about it.

Suppose you cascade two compressors one with a fast attack and release and another with a slower attack and release.  The first removes the "huge peaks" then the second does the general evening out of the signal.   This way the two processes are separated.   Now, if you think about what's happening to the signal, at any point in time the two compressors are producing a net gain.   So the question is can we use a single gain element with a more complicated control to achieve the same result as the cascaded compressor?   That's where "dual detector" compressors come in.  The Scholtz solution has a single detector which is not linear and morphs its time constants.   You can also separate the rectifiers so the fast and slow actions are separated in the rectifier.   You get more control to tinker with the settings but it may or may not be any better than a finely tuned Scholtz circuit.

Way back in the day the "JoeCheep" and "What" compressors were on the web.  You can see the how the peak control is separate away from the average.   You can put anything you like in the rectifier arms they just need to be merged at the end:
(remarkably they are still on the web!)

http://dt.prohosting.com/hacks/joecheep_description.html
http://dt.prohosting.com/hacks/what.html

Another thing the Scholtz circuit is doing is limiting the negative swing going into the rectifier with the diode clipper string.  This a good idea for many reasons.   Some circuits do this to some degree simply because the opamps clip!   The limiter can stop the cap integrating to silly levels with high peaks, however, it also limits the maximum level going into the control element.   This can stop 'ticks' due to feed-through of the high level control signal.   I found some solid-state gain controls elements produce a large feedthrough at high control input levels, which is disproportional to the control.  They produce objectionable ticks on strong bass guitar signals.

[StephenGiles]

"On a guitar, the first sound or pulse that comes out can be a huge peak which is almost always much stronger than the signal that follows after a few milliseconds."

It's common problem and there's a few mindsets to think about it.

Suppose you cascade two compressors one with a fast attack and release and another with a slower attack and release.  The first removes the "huge peaks" then the second does the general evening out of the signal.   This way the two processes are separated.   Now, if you think about what's happening to the signal, at any point in time the two compressors are producing a net gain.   So the question is can we use a single gain element with a more complicated control to achieve the same result as the cascaded compressor?   That's where "dual detector" compressors come in.  The Scholtz solution has a single detector which is not linear and morphs its time constants.   You can also separate the rectifiers so the fast and slow actions are separated in the rectifier.   You get more control to tinker with the settings but it may or may not be any better than a finely tuned Scholtz circuit.

Way back in the day the "JoeCheep" and "What" compressors were on the web.  You can see the how the peak control is separate away from the average.   You can put anything you like in the rectifier arms they just need to be merged at the end:
(remarkably they are still on the web!)

http://dt.prohosting.com/hacks/joecheep_description.html
http://dt.prohosting.com/hacks/what.html

Another thing the Scholtz circuit is doing is limiting the negative swing going into the rectifier with the diode clipper string.  This a good idea for many reasons.   Some circuits do this to some degree simply because the opamps clip!   The limiter can stop the cap integrating to silly levels with high peaks, however, it also limits the maximum level going into the control element.   This can stop 'ticks' due to feed-through of the high level control signal.   I found some solid-state gain controls elements produce a large feedthrough at high control input levels, which is disproportional to the control.  They produce objectionable ticks on a strong bass guitar signals.

The What compressor was excellent!

[PRR]

Fairchild 660 is not the first 'automatic' dual-constant time control, but arguably the most famous.
https://www.fairchild-recording-equipment.com/FAIRCHILD_Manual/

[PRR]

And out of left field..... 1948's "Speech Blockade" radio commercial blocker. This also uses the two time constant concept to distinguish barky ads from smooth music. Lightly processed speech has lower average than 1940s music of similar processing. Fairchild, and I think G.E. and Gates, used similar networks to hold-down compressor and limiter "pumping".