A Quadrature Power-Summed Compressor Sidechain For "The Uno Compressor"

[mediatechnology]

bruno2000 aka Joe Neil published his Threshold, Ratio and Makeup mods here but I think I was asked to delete them.

[mediatechnology]

I wanted to post a couple of other Ratio curves I simulated with the spreadsheet.

The first is a linear taper rotation versus Ratio. Electrically 50% rotation is 2:1; 75% is 4:1. (The X-axis should start at 0.)



The linear taper curve shown above shows us just how much better the slugged taper in the previous post is.

This is the ratio versus rotation of the FCS P4 on a stepped switch:



The value versus rotation of the FCS P4.



If one were to pick 24 values of slope to suit both mastering and use as a general purpose stereo compressor what would they be?
Using a switch, wired as a potentiometer, we can have virtually any monotonically-increasing taper we want.

I'm finding that control path gain error, though small, makes a big difference between theoretical and actual ratios at higher settings.
The VCAs themselves have a control port scale factor tolerance that ranges from 6.0 to 6.2 mV/dB. That's a potential 3.3% error.
The RMS detector may have a similar error in scale factor.
I may need to put in a control path gain trim that would be adjusted at one of the higher ratios to maintain a fairly-accurate ratio calibration.
This is my next area of investigation.

[mediatechnology]

I found that a ratio trim is necessary to compensate for the tolerance in detector, sidechain and VCA control scale.
When calibrated at 8:1 the lesser ratios remain accurate and near-infinite compression becomes available.

The updated schematic has some additional changes.
The current for the soft knee diode is now set at 10.8 µA by a single resistor from the 0.54V Vbe reference.
A threshold trim was added to accurately set the 0 dBu mark independently of the detector level calibration.

[emrr]

Good work there

[mediatechnology]

Thanks!
What's nice is that the actual ratios now follow the calculated values.
I found that the combined feedback resistance with the trim added was now about 105KΩ or about 5% higher.
Some of it is due to the dynamic resistance of the soft knee diode, Q2 of array 3, but most of it (I think) are the VCA and Detector Gain Control constants.
The VCA can vary from 6.0 to 6.2 mV/dB and the Detector by a similar amount.

I need to revisit using a slugged pot and stepped switch for the Variable Capacitance Multiplier.
The Ratio spreadsheet will get a new tab.

Another area that requires attention is the addition of a low frequency sidechain high pass filter.

[JR.]

any interest in a de-ess side chain filter? Or is that not as much of problem these days?

JR

[Gold]

any interest in a de-ess side chain filter? Or is that not as much of problem these days?

Since this is a stereo compressor a simple de esser like a DBX900 would be inadequate for most audio program. It would need to be a much fancier side chain than a bandpass filter. Floating threshold, reacts to the change in signal level in addition to absolute signal level and is frequency selective. I would certainly welcome a kick ass high frequency limiter but it's a different animal.

[JR.]

any interest in a de-ess side chain filter? Or is that not as much of problem these days?

Since this is a stereo compressor a simple de esser like a DBX900 would be inadequate for most audio program. It would need to be a much fancier side chain than a bandpass filter. Floating threshold, reacts to the change in signal level in addition to absolute signal level and is frequency selective. I would certainly welcome a kick ass high frequency limiter but it's a different animal.

Sorry for the veer... FWIW back when I added de-ess capability to dynamics processors, I used a an under damped HPF in the side chain, so it wasn't just a BP. The shape of that filter could make it pretty effective. I repeat my question is de-ess still a thing? It was a few decades ago.

The de-ess operating on HF signals doesn't need the fancy quadrature smoothing just a fast-attack/fast-release time constant. The de-ess control voltage could be summed in with the existing VCA control voltage for use as little or as much as wanted/needed. The de-ess control could be as simple as a single front panel threshold knob.

Sorry for thinking out load... Never mind.....

JR

PS; At least I didn't suggest using a micro..

[Gold]

I repeat my question is de-ess still a thing? It was a few decades ago.

Yes it's still a thing and very useful. For a dedicated stereo compressor it would have to react favorably to stereo program material and not single instruments, including voice. I can count on less than one hand analog high frequency limiters capable of this. Neumann BSB74, Maselec MDS2, Ortofon 732. Possibly BSS402

[mediatechnology]

I've had an interest in a de-esser/sibilance controller for vinyl and have wanted to wrap my head around the BSB74.
It would need to be a dedicated box.

I decided to take a look at using a 10K linear pot, wired in a poteniometric configuration, to control the attack-release time and see if it were slugged if it could be linearized.
The variable capacitor needs a voltage gain that spans from 0 to -9 to multiply the 2.2µF to 22 µF.
When Av=0 the 2.2µF is at AC ground and the effective value is 2.2µF.
Whwn Av=-9 the 2.2µF is "anti-bootstrapped" by 9X so the effective value is 2.2µF + 9 X 2.2µF.
With an approximate timing current of 8 µA this works out to provide a release rate of 600 dB/sec to 60 dB/sec.

A practical feedback value is 10K. To provide a maximum gain of -9 the input Rin should be 1K1Ω.
1K1Ω loads a 10K linear pot considerably and I was curious to see how it slugged the pot and if it linearized the dB/sec response versus rotation.
As it turns out it linearizes it very well:



The y-axis is the release rate in dB/second. The attack rate varies with the release rate to maintain an RMS response when the variable timing capacitor is changed.
I'll take the above taper as it is quite linear in the most frequently-used range.

For a stepped switch a "rheostat" connection in series with the 1K1Ω input will allow the switch to define the taper.
I'll repost a new drawing to incorporate the change.