TOA CL-021 Compressor Limiter

I bought this box about 15-20 years ago (1990?). It was an old piece of junk then. Possibly still is. None the less, here is a page devoted to the beast.
Hand drawn Diagram

The box is a solid state AGC /peak limiter / compressorish thing. It has two parallel gain reduction circuits run from the same side chain, but with different time constants. In fact there are two parrallel fets, one for limiting, one for compression.
The side chain is fed from the output rather than being a feed forward circuit. The 'slow' gain control has an attack time constant is 3.5 sec and release is 14, 26 or 50 sec on the 5,10 or 20 sec from panel setting respectivly. The disparity in the release times is that I have specified the CR time constants, What's on the panel may refer to increasing gain after a hit by a certain number in dB in that many seconds.
The 'fast' gain control has probably around 500us to 1mS attack , mainly limited by the current gain of Q4. The 'fast' release time constant is 1 Sec.
The 'fast' gain control acts as a slowish AGC while the 'fast' acts a a peak limiter.

This is the device hit with a 20db increase signal (voltage 10 times) . The tone is 10 Khz. It takes at least one cycle to bring the gain down to the limiting point . Thats 100 uS - better than I thought.

This is a chart of input level vs output level(in dB). The actual level calibrations don't mean much as I adjusted the input and output pots for useable readings on the test equipment. However the first thing thats obvious is thet the 4:1 setting is pretty weird. Something is wrong here!. The 20:1 setting does what I expected - that is the output is effctively limited at the threshold.
The input can only be held over a 25 to 30dB range, but if thats for catching the occasional peaks then this should be OK. Note that only the 'slow' control loop is plotted by this chart, the fast loop will require some custom test instrumentation.

Transistors are my dsignations, theyre not marked on the board with numbers.
Trimmer on Q10/Q11 source adjusts how high the control voltage on the gates needs to be above 0 volts before there is some gain reduction. Currently 2.63V
1K trimmer to 3x2.2M to Q10/Q11 bases adjusts distortion
Trimmer on base of Q2 adjusts threshold voltage of peak detector, Curently 1.08V
Trimmer on collector of Q3 adjusts slope for 4:1 compression (I think)
Trimmer in PSU adjusts voltage , probably should be 22V ? or maybe 24V

Here are some more plots of compression:
First, below is the 20:1 compression , or limiting:

Below is the initial plot for the 4:1 compression setting:

This is a pretty weird shape. The trimmer resistor on the collector of Q3 is a bit strange. The 8K2 resistor sets the circuit as a limiter, selecting 4:1 shorts this so I'd figure the value of the trimmer should be set to something less than 8K. If this is the case 47K seems a bit extreme. This is backed up by the setting of the trimmer being quite close to the low resistance end. Twiddling the trimmer and plotting the compression results , I decided this was correct and replaced the trimmer with a 4K7. Setting it to about 2.2K results in the chart below:

The compression is more like 2:1, but its a nicer shape. The original trimmer may have been set correctly originally, but it would have only taken a slight nudge to move it from the correct setting. With the 4K7 trimmer its setting is still pretty critical to get the position between no compression, or the initial shape.

Frequency plots

Below is a freqeuncy plot of the line and mic input( I added a pic pre to the 2nd input channel) . The green line is a plot of the freq response of my soundcard looped back. Not too good down low! The red is the mic channel thru the CL-021 with no compresion . Its -1dB at just over 22Khz and 45Hz. The Blue line is the line input with the input pot set at '2' , -1dB at 46Hz and 18Khz. The true low end rolloff is obviously masked by the soundcards low freq roll off.