THAT (was Fet) Compressors

[JohnRoberts]

John, I have been playing with this circuit since you put it in my lap, almost everyday after work.  I'm very familiar with the workings of it now. The breadboard (with power supply) is now known in my family as "the box"... Honey are you working on the "the box" again tonight? ;D

Wondering if there is any way to make it differentiate the steady state and transients more aggressively.  The transients dont seem to have the depth to really charge the smaller cap in a way that it is different from the larger cap. Its almost like the two caps are just in parallel and pinching the current off in the supply diode to the larger cap works perfectly but sounds like its just making the total parallel capacitance smaller.

while you may have been working on this every day, I will need to get back up to speed.

Please post or link to the current schematic.

Is it possible that the state variable filter is not steep enough? I actually tried a typical three op amp filter which is supposed to have a 12db per octave slope but adding the extra cap/op amp just made things all the more complicated to tune.

on reason I think it isn't as aggressive as I would like it to be, is because of the RMS detector is just too slow.  Even with a .01uF cap on it. I tried a regular precision rectifier and I finally heard the transients being caught, but that muddled up the whole workings of the state variable filter.

I also tried the zener approach in the previous schematic to the final one you made.  But again there wasn't enough voltage in the transients to trigger a 3.3v zener.

I had to reverse the steering diodes for attack and release because I have the half wave rectifier (that is in most of the THAT design notes) after the RMS detector. I needed it to have a threshold control.

Any more ideas?

Everything or almost everything is adjustable but please state clearly what you want more/less of.

JR

 

[bluebird]

What I'm looking for is more differential between steady state and transient. Seems like the transients kind of poop out the timing caps, or don't have enough current to fill them quickly enough (if I make them bigger).  I have made sure to check the steady state ripple at the steering diode node was right below 1.2v p-p. I have played with the gain of the SVF extensively.

Attached is what I have so far.

 

[JohnRoberts]

OK I'm soaking that in beer right now, but i'll try to say something intelligent tomorrow.

JR

 

[bluebird]

Soak away

Like I said before it works very well transparently. You said if its working right, you shouldn't hear it working much at all. And that is true.  But I would actually like it to be a little more aggressive with the transients, ducking them as quickly as possible without audible distortion.  Similar to a clipper where your not lowering the level of the program but shaving off the transients.  Could this circuit limit the transients more aggressively as you lowered the threshold, but keep the steady state somewhat the same level?

 

[abbey road d enfer]

But I would actually like it to be a little more aggressive with the transients, ducking them as quickly as possible without audible distortion.

I made a sim of the circuit. As you can see in the attached graph, the resuling voltage has a kink at the beginning, which results in the leading transient not being attenuated as much as the rest. The signal from the RMS is in green, has a leading edge of 20us and a falling edge of 10ms. I have applied an offset that is about typical of normal operation, but changing this offset, as well as changing the overall amplitude, does not change significantly the results.

 

[abbey road d enfer]

Tried to investigate the issue. Find attached waveform at the output of the 1st opamp (junction of D1 & D2).

 

[JohnRoberts]

Thanks... I recognize my contribution, the 2 op amp SVF with steering diodes for multiple different time constants... a "slow" time constant for slowly changing levels where voltage  at my pin 14 is less than +/- a diode drop.  Larger amplitude steps engage either a fast attack time constant through one diode, or "faster" than slow release in the other direction. 

I envision this smoothing circuit connected directly to the RMS output (that pin 7).  The threshold comparator IMO could be placed after the smoothing circuit. My attack and release diodes are backwards because the threshold comparator in between inverts the side chain voltage.

Attack time constant is tweaked with R5 (smaller for faster), but I also included a trick transient attack circuit involving D1 and the 2M pull up resistor.  For brief transient current demands more than what is supplied by the 2M, the op amp will slew the control voltage quickly ignoring the fast time constant and not charging the integration cap (it charges more slowly from the 2M pull-up). This circuit will quickly respond to and clamp transients, without quickly charging up the side chain cap avoiding the gain pumping caused by very fast attack times responding to brief transients.

There are a lot of interrelated moving parts involved to get everything working nice together. Back in the day I would have run my stepped tone burst generator through this to dial in which time constants I wanted to dominate for what size amplitude steps.
When operating as envisioned this will provide 4 different modes...  Very slow for steady state signals, somewhat faster for release,  faster again for attack, then really fast for large transients that does not pump up the side chain cap.

I do not know how to dial this in without a good tone burst generator.

JR

@ Abbey I'm not sure I follow your plots . Are you injecting a square wave at THAT pin 7?
   

 

[bluebird]

Abby those graphs look familiar. The "kink" is the smaller cap being taken over by the larger cap after the initial transient. If you take the larger cap out of the circuit you will just see the spikes.  Thanks for the simulation!

John Ok, I am going to try this with your circuit hooked up directly to the RMS detector. I just couldn't get it working properly like that at first. So how should I control the threshold? Just a pot in front of the RMS detector?

I have been testing with a tone burst to tweak attack and release. Like I said everything works as you describe it, Just not fast enough. When I put a precision rectifier instead of the RMS detector in front of your circuit, it was very fast but it didn't seem to  differentiate between steady state and transient signals, just acted like a typical sidechain.

 

[bluebird]

I labored over this today and came to the conclusion that the circuit cannot be put directly after the RMS detector. Somehow there is an offset voltage going back into the RMS detector. I was getting oscillation from the detector. I think the RMS detector wants to see a low impedance on its output. Whats wrong with having the half wave rectifier between the RMS detector and your circuit? it works very well like that, Just not enough depth from the transient pulses. Not enough voltage above the steady state CV.

 

[abbey road d enfer]

@ Abbey I'm not sure I follow your plots . Are you injecting a square wave at THAT pin 7?

"The signal from the RMS is in green, has a leading edge of 20us and a falling edge of 10ms."

 

[abbey road d enfer]

Abby those graphs look familiar. The "kink" is the smaller cap being taken over by the larger cap after the initial transient.

OK, got it. The desired fast-to-slow transition happens on the positive edge, which is the one that controls attack. Is it really what you want? I would have thought you'd want that on release. OTOH, this would be useful on both atk and release.
I have toyed with the idea of using the dynamic resistance of junctions to make a system of continuously variable time-constants where times get shorter when variations increase in amplitude. I have simulated it but never put that in silicon. The idea is creating a dead band where the time-constant is governed by the leakage of the junctions, and as soon as voltage develops entering exponential current, resulting in log timing.

 

[abbey road d enfer]

I labored over this today and came to the conclusion that the circuit cannot be put directly after the RMS detector.

That may well be the case; it seems the point where the last diode enters into play depends on the actual DC voltage at the input.

Somehow there is an offset voltage going back into the RMS detector. I was getting oscillation from the detector.

I don't think it's possible.

I think the RMS detector wants to see a low impedance on its output.

Unless something has significantly chaged in RMS detectors, that's not my experience. The output is low-Z and can drive any load from a few k to 100's of k.

  Whats wrong with having the half wave rectifier between the RMS detector and your circuit? it works very well like that, Just not enough depth from the transient pulses. Not enough voltage above the steady state CV.

Just to make sure I understand, the half-wave rectifier is the one that prevents negative voltage entering the Ec- port?
If you put it between the rms and the TC, it ignores a large portion of the detection range.

 

[JohnRoberts]

Abby those graphs look familiar. The "kink" is the smaller cap being taken over by the larger cap after the initial transient. If you take the larger cap out of the circuit you will just see the spikes.  Thanks for the simulation!

John Ok, I am going to try this with your circuit hooked up directly to the RMS detector. I just couldn't get it working properly like that at first. So how should I control the threshold? Just a pot in front of the RMS detector?

My smoothing circuit is unity gain over all (it just controls time constants). Move your threshold circuit to the output of the smoothing circuit and it should deliver similar dynamics.

I have been testing with a tone burst to tweak attack and release. Like I said everything works as you describe it, Just not fast enough. When I put a precision rectifier instead of the RMS detector in front of your circuit, it was very fast but it didn't seem to  differentiate between steady state and transient signals, just acted like a typical sidechain.

It should be arbitrarily fast in the "trick" fast attack mode... Making the 2M pull up resistor larger will reduce the step size it responds to. Note: there is no smoothing in this ultra fast mode so this will distort the transient. Sized properly this will squash brief transients and return to normal behavior before any distortion is audible  (I did a lot of work with this designing transparent tape NR back in the late 70s/early 80s).

Tone bursts and scope to inspect side chain waveforms will help dial in the different modes to control different dynamic conditions.  The steering diodes are one set of thresholds, perhaps two diodes in series for the release time constant will deliver a wider window of slow time constant. The current through the 2M pull up defines when it shifts from fast attack to uber-fast attack.

JR

 

[JohnRoberts]

OK, got it. The desired fast-to-slow transition happens on the positive edge, which is the one that controls attack. Is it really what you want? I would have thought you'd want that on release. OTOH, this would be useful on both atk and release.
I have toyed with the idea of using the dynamic resistance of junctions to make a system of continuously variable time-constants where times get shorter when variations increase in amplitude. I have simulated it but never put that in silicon. The idea is creating a dead band where the time-constant is governed by the leakage of the junctions, and as soon as voltage develops entering exponential current, resulting in log timing.

I don't have schematics handy but Paul Buff (RIP) did some work with log domain side chain dynamics. As you can imagine small voltage changes can result in significant impedance (time constant) changes.

JR

 

[bluebird]

OK, got it. The desired fast-to-slow transition happens on the positive edge, which is the one that controls attack. Is it really what you want? I would have thought you'd want that on release. OTOH, this would be useful on both atk and release.

Hmmm, The diodes labeled attk and release on the schematic are reversed, I just cut and pasted that schematic together.  John had them like that because he wasn't planing the inverting threshold stage between the RMS and his circuit. Regardless, its just a matter of adjusting the resistors to the opposite diodes for attack and release right? Or does the CV polarity change the way the SVF works in general?

I have toyed with the idea of using the dynamic resistance of junctions to make a system of continuously variable time-constants where times get shorter when variations increase in amplitude. I have simulated it but never put that in silicon. The idea is creating a dead band where the time-constant is governed by the leakage of the junctions, and as soon as voltage develops entering exponential current, resulting in log timing.

Would this circuit work after the threshold op amp, or right after the RMS detector? I will definitely breadboard this!

My smoothing circuit is unity gain over all (it just controls time constants). Move your threshold circuit to the output of the smoothing circuit and it should deliver similar dynamics.

Ok I will try that. For some reason (when putting Johns circuit directly after the RMS detector)I am having trouble with over all offset voltage reaching the VCA and skewing the overall gain extremely. 

So the schem below should work correctly? I didn't try the threshold circuit after like that.

 

[abbey road d enfer]

Hmmm, The diodes labeled attk and release on the schematic are reversed, I just cut and pasted that schematic together.  John had them like that because he wasn't planing the inverting threshold stage between the RMS and his circuit. Regardless, its just a matter of adjusting the resistors to the opposite diodes for attack and release right?

What I meant is the "kink" happens only on one edge; it should be the one corresponding to release, because that's what you want, variable release, not variable attack.

Or does the CV polarity change the way the SVF works in general? 

No.

Would this circuit work after the threshold op amp, or right after the RMS detector?

It is intended to be in the control voltage path.

I will definitely breadboard this!

Great! I've always been too lazy to do it myself.

Ok I will try that. For some reason (when putting Johns circuit directly after the RMS detector)I am having trouble with over all offset voltage reaching the VCA and skewing the overall gain extremely. 

Have you tried loading the rms detector output with 10k?

 

[bluebird]

What I meant is the "kink" happens only on one edge; it should be the one corresponding to release, because that's what you want, variable release, not variable attack.

ok, so is that because of the polarity of the CV coming into the SVF?

It is intended to be in the control voltage path.

Is that to mean I can just replace Johns SVF circuit with yours in the schematic I posted in reply #101?

Have you tried loading the rms detector output with 10k?

Yes, I went all the way down to 1K. things were still acting funny. Although when I put a 1M resistor from +15v to the input of the RMS detector it stopped the oscillation, but my offset problem at the input of the VCA was still there.

Every single THAT corp design note has the  inverting opamp threshold circuit I have in my schematic after the RMS detector, with the exception of this one:
http://www.thatcorp.com/datashts/dn04.pdf
And that has the RMS detector going straight into the VCA control port. But its input is biased with that 1M resistor.

That's a problem for me to wrestle though. I don't want to bog you guys down with the small details too much.

 

[abbey road d enfer]

ok, so is that because of the polarity of the CV coming into the SVF?

That's because of the polarity relationship between the "kink" and the control port. In sidechain 1, the rms detector signal is inverted first before hitting JR's TC circuit, so the kink is on the positive edge at the output of U1A; with the inverter driving the Ec- port, that corresponds to a gain increase, i.e. release.
In sidechain 2, JR's circuit is connected directly to the rms output, so the kink (= variable TC) happens on the rising edge, which puts the variable TC on the attack path.

Is that to mean I can just replace Johns SVF circuit with yours in the schematic I posted in reply #101?

No. First, big caveat, as I wrote before, I just simmed it, but never actually breadboarded it. Second, it's not meant to replace the TC controls, it's an addition, meant to answer the age-old issue of minimizing LF THD whilst retaining the capability to deal with transients. I had thought that as an add-on to an already defined TC circuit. The response is symmetrical, so the Atk and Release times would be identical with my circuit. I don't think that's what you want.

Yes, I went all the way down to 1K. things were still acting funny. Although when I put a 1M resistor from +15v to the input of the RMS detector it stopped the oscillation, but my offset problem at the input of the VCA was still there.

This seems to indicate the threshold control is not working as expected. You need to measure the DC voltages at various points of the circuit.

Every single THAT corp design note has the  inverting opamp threshold circuit I have in my schematic after the RMS detector, with the exception of this one:
http://www.thatcorp.com/datashts/dn04.pdf
And that has the RMS detector going straight into the VCA control port. But its input is biased with that 1M resistor.

It's a way to adjust threshold that's compatible with the minimum-component-count approach of the design.

That's a problem for me to wrestle though. I don't want to bog you guys down with the small details too much.

The details are probably what would help us help you.

 

[bluebird]

JR's circuit is connected directly to the rms output, so the kink (= variable TC) happens on the rising edge, which puts the variable TC on the attack path.

Sorry if I'm being thick...Wouldn't I have an expander if I had the polarity wrong at the VCA? When I was experimenting with johns TC circuit right after the RMS detector I had to use the non inverting input of the op amp right before the VCA  Ec- port...

Hold the presses...Maybe I need to go into the Ec+ port... Maybe that's why my gain is all messed up.

No. First, big caveat, as I wrote before, I just simmed it, but never actually breadboarded it. Second, it's not meant to replace the TC controls, it's an addition, meant to answer the age-old issue of minimizing LF THD whilst retaining the capability to deal with transients. I had thought that as an add-on to an already defined TC circuit. The response is symmetrical, so the Atk and Release times would be identical with my circuit.

Got it.

 

[abbey road d enfer]

Sorry if I'm being thick...Wouldn't I have an expander if I had the polarity wrong at the VCA? When I was experimenting with johns TC circuit right after the RMS detector I had to use the non inverting input of the op amp right before the VCA  Ec- port...

I think there's some confusion here. I'm not talking about the CV polarity. Obviously you need to make it right, and both your schematics imply the use of Ec-.
What I'm saying is that the TC circuit treats differently positive leading edges than the negative ones.  You must make sure that the edge that has the variable TC is the one that corresponds to a gain increase (Release).