Interesting Dual Triode Compressor Circuit.

[SoulFetish]

I came across this circuit earlier today from a November issue of Radio Electronic Engineering. It's a pretty cool, simple, and seemingly useful tube compressor using two triodes. Originally, it was designed for a 6SL7. But it would work equally as shown using a 6CG7, or (what is quickly becoming one of my favorite new tubes) a 6111 - subminiature dual triode.
According to the article, the output of the amplifier remains constant within 1dB for an input variation up to 20dB.
It's pretty compact design, with a relatively small component footprint which might make it a useful addition to an existing circuit. After skimming through the article, I noticed that the one drawback is it apparently has poor low frequency performance. It looks like it was optimized for the human voice, but suffers some kind of transients at low frequencies.
Maybe you guys have some suggestions for improving this after looking over the schematic.

[Tony Bones]

I'm going to admit that I can't follow the signal path.

Are the cathodes tied together? The grids?

Does it work by increased (DC?) current lowering rp which acts as the lower part of a voltage divider?

What?

[SoulFetish]

I pulled this from a parallel stage on another drawing, and looking at it again, I can see why it might be confusing. I'll clean it up a little to be more clear.
But, he cathodes are not tied together and the grids aren't either. Sorry for the lack of obvious separation.

*redrawn - hopefully that is a little better.

[ric]

So the top triode's grid is fed in a familiar fashion. That imposes a signal on the plate that is tied directly to the b+ to the plate load r of the bottom triode. That puts the signal on the grid of the bottom which is tied directly to the k of the top.

Is that the signal path? And interacting in that way holds the level?

[dmeek]

I'll take a stab at it. The audio path is passive. I re-drew it to make more sense. The audio path is C2 to R2 to C3. Triode A is the AC to DC converter, response time set by C4
Tiiode B is the attenuation control by shunting the audio to ground. The problem is it also adds a large DC envelope voltage to the audio and I'm not sure how B is or is not biased or where the threshold would be.
Maybe B needs a small cathode resistor.

https://music-electronics-forum.com/...1&d=1562261328

[Tony Bones]

The signal path is just along the top of the drawing. Neither tube 'passes' signal.

The top tube is biased super-cold with the 25K pot and 10uF bypass cap. An increase in signal level causes an increase in its DC cathode voltage.

The bottom tube is biased super-hot. An increase in cathode voltage of the first tube (caused by increased signal level) causes the second tube to become biased even hotter with the result that its plate resistance drops. This plate resistance forms the bottom part of a voltage divider where the 470K resistor along the top of the drawing is the top part of the voltage divider.

I think rp of the 6CG7 is going to be pretty low in this circuit. Something like 5000 ohms or less, so this circuit is going to result in quite a bit of insertion loss.

[nickb]

I think it's going to have very non linear compression, that is to say the peak signal amplitude to attenuation plot is not a straight line. That might be desirable, or course, in some cases.

[J M Fahey]

Just thinking aloud:

It is passive, tubes are not in the audio/signal path.

Top triode is used as a diode, its grid to cathode *are* a diode anyway.
It rectifies audio signal available at the 470k/470k midpoint.

Rectified voltage charges the 10uF cap.

Some of that voltage (adjusted by the pot) goes to second triode grid.
Second triode works as a variable resistor/attenuator, same as Fets (or even bipolar transistors) are used.

Not sure why first triode plate is connected to +B ; suppose it prebiases the first triode cathode so it´s more sensitive to low level audio signal.

I have seen it done with 1N4002 type diodes: since thay have a 650/700mV "barrier" to start conduction, in some audio detection circuits I have seen a, say, 1M resistor connected to +B (9 to 15V) so it´s already passing a very small current and 700mV barrier has already been "breached" ... suspect something similar here ... might be wrong of course.

So it´s probably a "reasonable quality" compressor, better than nothing.

Not too linear but hey ........

[SoulFetish]

I'll take a stab at it. The audio path is passive. I re-drew it to make more sense. The audio path is C2 to R2 to C3. Triode A is the AC to DC converter, response time set by C4
Tiiode B is the attenuation control by shunting the audio to ground. The problem is it also adds a large DC envelope voltage to the audio and I'm not sure how B is or is not biased or where the threshold would be.
Maybe B needs a small cathode resistor.

https://music-electronics-forum.com/...1&d=1562261328

Sorry if the schematic was still a little confusing the way it was drawn. I redrew it to reflect the format you were using.... 3rd time's the charm.


So I think you're correct about triode A providing the AC to DC conversion. But I'm not sure it's "passive", in the way you and Fahey describe? So triode A is operating as a grounded plate/cathode follower as far as DC is concerned. However, it looks like it would be biased well into cutoff.
I'm wondering if the feedback from the plate of triode B creates a bias point through the RC divider network?
Triode B, on the other hand gets it's grid DC bias through the 25k cathode divider on triode A. This seems a bit strange, but if you look at the datasheets for these tubes, they all have quite high transconductance and linearity with acceptably quite high +grid voltages. So, perhaps the designers purposefully designed this stage in class A₂?

edit: I can say, I've never seen a circuit quite like this (assuming the original didn't have any mistakes). I'll try and find a link to the original article.
Maybe someone can run a simulation. But I may breadboard this when I get a little time. It's simple enough and I have the parts.

[SoulFetish]

*found it!

from:
https://www.preservationsound.com/?p=8272

Constant-Output-Broadcast-Amplifier.pdf

[dmeek]

I wouldn't call the plate of B "feedback". The signal on the grid is not audio, it's a DC voltage that follows the amplitude of the audio input.
The 10uF cap kills any audio.

As JMF said B is similar to the way a FET is sometimes used to attenuate a signal except that a triode needs a higher voltage on the plate to operate.
That voltage is the main problem with this because it will be added to the audio like the old thumping tremolo we hate so much.

Just the audio path is passive, The tubes are the control circuit or sidechain as it's called in compressor parlance.

I'm tempted to breadboard it though I would have to make a breadboard. I have a Heathkit IP-17 power supply (0-400VDC, 6.3VAC)

[SoulFetish]

As JMF said B is similar to the way a FET is sometimes used to attenuate a signal except that a triode needs a higher voltage on the plate to operate.
That voltage is the main problem with this because it will be added to the audio like the old thumping tremolo we hate so much.

Right, this must be LF problems described.

Just the audio path is passive, The tubes are the control circuit or sidechain as it's called in compressor parlance.

Okay, I see what you mean. Interesting. Can you recommend any articles or literature illustrating discrete compressor circuits?
Also, nice use of the word "parlance". Reminds me of The Big Lebowski.

I'm tempted to breadboard it though I would have to make a breadboard. I have a Heathkit IP-17 power supply (0-400VDC, 6.3VAC)

You may get to it before me.
I'm tempted to someday build a nice adjustable, regulated HV bench power supply. Max Robinson has a nice article on his site here:
Bench Power Supplies for Tubes

[dmeek]

The Dude abides!

The Heath IP-17 is like Figure 5 except it uses 2 6L6s paralleled for the pass element.

I can't think of a good article at the moment but a compressor consists of two circuits in the sidechain: an AC/DC converter sometimes called an RMS converter which creates a DC voltage that
is proportional to the amplitude of the input signal, and an amplitude control element which could be a FET, a light source and photoresistor, or a Voltage Controlled Amplifier (VCA) chip or descrete circuit module
which controls the amplitude of the input signal.

These circuits can also be arranged to create an Expander or Noise Gate

[g1]

Is this not a tube version of the 'orange squeezer' (or predecessor maybe a better word) ?

[eschertron]

With the two 470k resistors feeding signal and sidechain to the output, it appears to me that the effect is "mixing" a modulated ground Z with the signal.

edit: so that makes the second triode a voltage-controlled resistor...