Negative Feedback Compressor

[separateness]

Fellows,
Are there compressors that work by varying the amount of negative feedback fed back into the input?
A rudimentary mockup of the sort of thing I mean:

Ignore the values for now, I just wanted to indicate the basic topology I am describing. I am not very well versed in compressors but most I have thus far seen seem to either vary a bias point or shunt signal to ground.
If there are no such compressors, is there some obvious reason this should not work or is not used?

Thank you.

 

[zamproject]

Are there compressors that work by varying the amount of negative feedback fed back into the input?

For what I know that's how VCA chips work... ? more advanced fellas around might tell more.

But are you talking about -gain cell- or overall topology ?
Some comp have feedback, some have feed-forward, from audio to side chain

 

[separateness]

I believe VCAs have a bunch of different methods of acting. Of the schematics I have looked at and somewhat understood some use OTA's or varying how much signal is sent to the two inputs of an opamp. There are probably many, many ways of skinning that cat and some might be doing something isomorphic with what I am trying to describe but I have not seen that or if I have seen it, I did not understand it to be the case.

I was thinking more in terms of the method of compressing the signal being a modulation in the amount of local negative feedback around a tube gain stage. It would probably require a high gain, small signal pentode like an EF86.

 

[abbey road d enfer]

Are there compressors that work by varying the amount of negative feedback fed back into the input?

Many opto compressors do it that way. The photo resistive element is placed between the output and inverting input of an opamp.
Resistive opto couplers lend themselves to this topology because they are fully floating.
FET's not so much because they may interfere with the opamp's operating point. It could probably be done, but I don't think there would be any advantage compared to the usual shunt topology.

 

[separateness]

Ah, I see now that many, many opamp optos do indeed do this. I must have missed this because the only opto I have really taken a look at is the LA2A.

Still, I believe there is potential for a serviceable one tube comp here and I have yet never seen the negative feedback around a triode manipulated like this for compression.

 

[Bo Deadly]

Are there compressors that work by varying the amount of negative feedback fed back into the input?

It probably wouldn't work very well. And your circuit won't work at all as-is because there is no path to 0V on the grid. The tiniest current through the mosfet (?) into the high impedance of the tube will servo the tube. If you added a resistor from the mosfet / grid to 0V, then you will get the voltage divider you need to get variable feedback..

But in practice distortion will be quite high with this scheme because the mosfet will be using a large part of it's ohmic region which is non-linear. Although if you used a JFET, there is a very neat trick that can very nicely cancel out the distortion. See the Urei 1176 circuit.

 

[Bo Deadly]

For what I know that's how VCA chips work... ? more advanced fellas around might tell more.

Traditionally a "VCA" refers to a Blackmer cell which does not vary the amount of feedback to control gain. Blackmer VCAs use the fact that gain through a transistor varies with current. So a Blackmer VCA is essentially an amplifier that also has ports that control the amount of current following through it. More current, more gain. Less current, less gain.

Some comp have feedback, some have feed-forward, from audio to side chain

I think we're talking about VCAs and not compressors. So the feedback / feed-forward of a compressor is not applicable to this circuit.

 

[separateness]

I think an opto resistor would work better than a MOSFET. I mostly put the FET in the above example as a generic stand in for any variably resistant component. There are perhaps other possibilities which might work even better.

 

[JohnRoberts]

I am not a tube guy but know a little about compressors.

It is a fairly common topology to locate the variable gain element in the negative feedback path. An apparent benefit of this when dealing with large signals, the gain element is always seeing the compressed output. This was useful back when early inexpensive IC gain elements had limited signal handling capability.

JR

 

[Script]

in the above example

NFB around a tube is there for a reason. Changing that dynamically would change THD and IMD, I'd think. And it can be quite dramatically so (up to 7th harmonic and higher). Is that acceptable ? How much attenuation in dB ?

 

[separateness]

NFB around a tube is there for a reason. Changing that dynamically would change THD and IMD, I'd think. And it can be quite dramatically so (up to 7th harmonic and higher). Is that acceptable ? How much attenuation in dB ?

Ah, see this is what I was really wondering: why wasn't this method used around tube gain stages, especially since it seems so simple and cheap to implement? And I think you may have presented a satisfying answer.

Probably also the fact that the feedback components present an additional AC load on the tube (and one that is changing constantly at that) would lead to some undesirable instability/unpredictability. Perhaps this is just another phrasing of what you have just said.

 

[abbey road d enfer]

NFB around a tube is there for a reason. Changing that dynamically would change THD and IMD, I'd think. And it can be quite dramatically so (up to 7th harmonic and higher). Is that acceptable ? How much attenuation in dB ?

There are two issues in applying NFB around tubes.

• Around a single stage: voltage-to-current NFB (from anode to grid) reduces the input impedance significantly, and what's more, in a dynamic way
• Around two stages: voltage-to-voltage NFB can be used (from the 2nd anode to the 1st cathode), but there are stability issues, which would probably limit the attenuation range

In both cases, whatever hypothetical advantages could be claimed hardly justify the extra effort.

 

[Script]

Hmm, a guitar amp with a 3 dB leveller maybe (?) Tuned to when input gets loud and comp kicks in, sound gets slightly more linear... The opposite of rock'n'roll

 

[separateness]

Hmm, a guitar amp with a 3 dB leveller maybe (?) Tuned to when input gets loud and comp kicks in, sound gets slightly more linear... The opposite of rock'n'roll

Or use a phase shift oscillator for the control voltage for a particularly bizarre sort of tremolo.

 

[plexi]

Hi, long time member but never post....There is a design like this from way back, it is the GE BA5A. It uses a push pull line amp with NFB applied from output plates to input cathodes. The variable device is a pair of triodes with a (sidechain derived) neg voltage applied to grids. I actually have one of these - I had to build a supply for it finally after 20 years sitting in the corner of the shop so its running now.....I was able to look at it and try understand what is happening here.

I have built a proto w 12AU7's subbed and (partial) mosfet sidechain, It works well over 12-14 dB range varying the neg voltage at the control tube grids. It gets wierd because it is AC and DC feedback at the same time so at some point it starts to cut off the input tube. But the last triode is set to limit the amount of swing so it keeps it out of the cutoff range.

But it does sounds pretty cool, and it's a different method of GR

 

[separateness]

@plexi That is a fascinating circuit. Thank you for posting it as I am completely unfamiliar with the unit. With the aid of your explanation I am almost able to follow what is going on in there.

As an aside, when this notion originally came to me, I had thought of using a triode as the variable element but could not think of a way to make it behave properly.

 

[Script]

Is that a somewhat 'standard' variable mutual conductance configuration (DC sidechain) -- with additional fixed NFB (AC) wrapping around the audio path from Out to In to help linearize the tube ?

Or does the AC go thru the timing network as well?

-- Sorry for asking, my understanding and reading skills of tube circuits are elementary at best and the schematics look quite complex, to me. --

I like the case and faceplate. Thanks for showing

 

[abbey road d enfer]

Is that a somewhat 'standard' variable mutual conductance configuration (DC sidechain) -- with additional fixed NFB (AC) wrapping around the audio path from Out to In to help linearize the tube ?

No. The vari-mu tubes operate in the opposite direction of a standard "vari-mu" compressor.
Here, under threshold, the vari-mu tubes have almost no conduction, and the output amp operates at nominal gain.
Over threshold, the vari-mu tubes conduct, which results in increased NFB and hence reduced gain.
The big advantage is that the tubes in the normal path always operate at their optimum operating point, whence in a standard vari-mu, the more compression the more the tubes are running at low quiescent, resulting in diminished out capability and thus increased distortion.

 

[Script]

Thank you for the rundown. I need to print out those schematics.

 

[Script]

So roughly like this ?