Stereo Compressor Link

[ruffrecords]

Recently I built a simple optical compressor:

http://groupdiy.com/index.php?topic=44550.msg731329#msg731329

So that  it acted on both halves of the signal waveform, I used a pair of Vactrols with the LEDs connected in antiphase. It means you have two opto resistors connected in parallel with slightly different characteristics so overall you get some sort of average both in terms of attack/decay and law.

I have always shied away from tackling a stereo compressor because of the problem of ensuring similar characteristics in each compressor when the two side chains are linked. It occurred to me the other day that having a pair of opto resistors in each channel might provide a way round this. Suppose that instead of linking the side chains you swapped one pair of optos so each side chain operated on both channels? Here's an outline schematic of what I am suggesting:

S1, the 'Link' switch is shown in the off position. The left signal comes in via R1 and the two otpo resistors R2 and R3 attenuate it. Similarly, the right signal enters via R4 and is attenuated by opto resistors R5 and R6. Notice how the opto LEDs are wired back to back so they respond to both halves of the waveform.

When the 'Link' switch is activated, R3 is now connected across the right channel audio and R6 is connected across the left channel. In this way the control signal from each channel affects the signal level in the other. The overall attenuation should be something like the average of the attenuation created by each signal. If you have a big left signal it will cause a biggish attenuation in both left and right channels. If, at the same time you have a small signal in the right channel it will create a smallish attenuation in both left and right channels. So each channel has a combination of a biggish and a smallish attenuation; in other words pretty much the same.

This feels right in my head. Is it going t work or have I missed something? Has this been done before?

Cheers

Ian

 

[Panorama]

I guess the effectiveness with two opto couplers //  goes down. It could be compensated by a larger control voltage, don't know if that gives any problems. Why not mix the control voltage and use one optocoupler per channel? Perhaps even using a pan pot?
(How appropiate to mention a panorama pot in my first post!)

 

[JohnRoberts]

That could give some modest benefit.

I did some bench work back decades ago trying to linearize (log-ize?) the gain law of LDRs. I used a LDR with a single center-tapped resistance element as two independent but closely tracking resistances when the center tap was grounded. Feeding DC current into one half and using NF I could force that one resistor to be a desired target resistance. Since the other half should track with a similar resistance I now had a voltage controlled LDR resistance.  8)  By definition two or more of these should track each other.

FWIW I dismissed this as not competitive with VCAs. even back a few decades ago.

JR

 

[ruffrecords]

I guess the effectiveness with two opto couplers //  goes down. It could be compensated by a larger control voltage, don't know if that gives any problems. Why not mix the control voltage and use one optocoupler per channel? Perhaps even using a pan pot?
(How appropiate to mention a panorama pot in my first post!)

LOL, many years ago I was involved in the development of a daisy wheel printer. The company set up to fund its development was called Panorama!

I think the effectiveness of the optos might actually be increased, There are two in parallel so their combined resistance is half of a single one. For a given series resistor, the maximum attenuation will therefore be doubled (6dB extra). The minimum attenuation will be almost unchanged because the off resistance of the optos is really high (around 1 megohm). Certainly I had no problem achieving 20dB of gain reduction which is more than enough for most purposes.

Cheers

Ian

 

[Panorama]

Ian, you are right.

I was thinking about the switch "down" situation, where I assumed that R2 (for instance) has a "fixed"  resistance value, lowering the impedance at the "Left signal out" point, so that the effectiveness of R6, and as such of the  "Right control voltage" becomes less effective.  Anyway, perhaps this leads exactly to the desired behaviour of the compressor. So there is a mechanism where the left channel control voltage is influencing the effectiveness of the right channel control voltage on the output of the left channel, but thinking of it: this might be exactly what one wants! 
It is a parameter  that is not adjustable, but perhaps that is not a problem but an asset!

EDIT: the more I think about it, the more I think the basis idea is  just fine!  My comments are superflous, more noise than signal so to speak, it is time to build the thing a do "real space" experiments.  Please keep us posted!

 

[L´Andratté]

That´s maybe a brilliant idea...

Needs to be listened to wrt stereo image, but with vactrols it´s not too much about linearization I guess, even center tapped units are rumoured to be like 20% tolerance... Still they sound good especially 5c3/2

I think it will be a very good sounding stereo compressor anyway!

If I may ask, how did you arrive at the 47k value?

 

[ruffrecords]

If I may ask, how did you arrive at the 47k value?

If you look at the 5C3 data sheet you will see its resistance is about 10K at 2mA LED current. Two of them in parallel would be close to 5K which gives 20dB attenuation with a 47K resistor. Also, I know my SRPP output stage will happily drive a 2K4 load. In the prototype I used a 2K series resistor to drive the LEDS. The LEDs turn on at about 1.5V so when passing 2mA they look like 0.75K so this, combined with the 2K series resistor is fine for the driving stage. For 20dB attenuation the output stage needs to drive 2mA into 2.75K which is 5.5V peak (equivalent to about +14dbu) which can do with ease. The threshold is at about 1.5 V peak or around 2.7dBu. The means the driver stage only needs to provide 22.7dB of gain for the threshold to be -20dBu referred to the output.

Cheers

Ian

P.S. You are right that the proof is in the stereo imaging. It was quite straightforward to build on of these so I'' put another one together and try out the stereo linking idea.

 

[pvision]

Having two vactrols per channel should average out the response / tolerance variations of the units

So why not use four, or even eight, per channel? You could switch four between channels for stereo, or switch two to reduce the channel interaction

I am beginning to think that he sidechain is the heart of a compressor and that sidechain eq is essential

Nick Froome

 

[joaquins]

In tube optical compressors it tends to be more about the element and less about the said chain, having a couple of elements only adds 2 vactrols, or eve one if the dual unit could be get, which the value is not much more than a single one. If I would be doing this in solid state I would agree with you, to take more attention to the SC, but in tube compressors side chain tends to be oversimplified to avoid adding tube stages. Would be good to have some equalization there, that's true, but as long as it doesn't add loss than would harm later.

The API 525 has variable time depending on frequency and level, so peaks attenuations are fast and low frequency intense material will make the timing of the compressor slower. This is a pretty old design which I find amazing, now you could use a micro controller or small DSP to work on the side chain and do all sort of wired stuff there, multiple RMS timings, variable ratios and thresholds, etc. with no extra hardware, if you want to do that in analog, even worst in tube, the circuit would be over complex, and the simplicity of the tube compressors is something we may be finding nice, or why we like so much LA2A, it couldn't be simpler, well implemented and good sounding compressor.

JS

 

[ruffrecords]

Just been thinking about wiring the audio between two modules to implement the link switch.  The audio level across the opto resistor can be quite low so I am reluctant to send it outside the module. Obviously this is no problem if both compressor channels are in the same box. It then occurred to me that we could achieve the same linking  effect by swapping the control voltages to a pair of LEDs, one in the left and the other in the right channel. so they are operated by the control signal from the other channel. This means there would be no need for audio signal to pass from one module to another. I think I will try it this way first.

Cheers

Ian

 

[djgout]

I like the link idea, but the way this is drawn with the LEDs in anti parallel, the linked CV is only the negative peaks. 

I don't know if that's actually going to matter though... It's an opto comp.

What about using an incandescent bulb instead of the LEDs? You could still double them up.  I don't think you'll need nearly the drive for an el panel.

 

[ruffrecords]

I like the link idea, but the way this is drawn with the LEDs in anti parallel, the linked CV is only the negative peaks. 

I don't know if that's actually going to matter though... It's an opto comp.

Yes, it is inevitable. You could wire it so that the link works on +ve peaks but you have to choose one or the other. As wired now, when linked, the left channel gets +ve peaks of left signal and -ve peaks of right. Similarly right channel gets +ve peaks of right and -ve peaks of left. It all depends on the the relative size of +e and -ve peaks in normal signals.

What about using an incandescent bulb instead of the LEDs? You could still double them up.  I don't think you'll need nearly the drive for an el panel.
[/quote]

I suspect the reason is they will be way too slow.

Cheers

Ian

 

[abbey road d enfer]

Just been thinking about wiring the audio between two modules to implement the link switch.  The audio level across the opto resistor can be quite low so I am reluctant to send it outside the module. Obviously this is no problem if both compressor channels are in the same box. It then occurred to me that we could achieve the same linking  effect by swapping the control voltages to a pair of LEDs, one in the left and the other in the right channel. so they are operated by the control signal from the other channel. This means there would be no need for audio signal to pass from one module to another. I think I will try it this way first.

Cheers

Ian

Apparently you have based your primary design on driving the LED's with alternate polarities of the audio signal. I think it is not the right approach. Indeedyou can use several optocouplers in parallels to somewhat average their characteristics, but my personal experience with optocouplers tell me that the disparity is so high that it would not ensure performance, in particular pairing, in a repeatable way.
Proper operation of optocoupler based comp/lim relies on two aspects: feedback operation and/or adjustability of the drive circuit. Even this does not prevent the need for selection of optocouplers.
As I mentioned earlier, the only notable exception I know of is Ted Fletcher had a feedforward opto-based comp that involved servoing the optocouplers. His drive circuit was VERY elaborate.
Since efficient stereo coupling requires perfect control of the attn vs. control voltage law, i don't see how you could get away with including a drive circuitry that provides adjustment of the slope and origin of the control curve.

 

[G-Sun]

BTW Ian: I know your a tube-purist and I'm not skilled enough to know if what I'm saying is relevant,
but imaging an IC-based in/out board providing balanced insert in/out for all stages in a stereo pre/eq/comp-channel, sidechain-filter and whatever for the sidechain, eq pre/post comp, even m/s encoding ?

 

[ruffrecords]

Just been thinking about wiring the audio between two modules to implement the link switch.  The audio level across the opto resistor can be quite low so I am reluctant to send it outside the module. Obviously this is no problem if both compressor channels are in the same box. It then occurred to me that we could achieve the same linking  effect by swapping the control voltages to a pair of LEDs, one in the left and the other in the right channel. so they are operated by the control signal from the other channel. This means there would be no need for audio signal to pass from one module to another. I think I will try it this way first.

Cheers

Ian

Apparently you have based your primary design on driving the LED's with alternate polarities of the audio signal. I think it is not the right approach. Indeedyou can use several optocouplers in parallels to somewhat average their characteristics, but my personal experience with optocouplers tell me that the disparity is so high that it would not ensure performance, in particular pairing, in a repeatable way.

It was not as sophisticated as that. I had seen a number of Vactrol based designs using a single LED (and sometimes not even an anti-phase diode to prevent it acting as a charge pump). it was a simple, obvious  step to add an anti-phase Vactrol LED and parallel its opto resistor with the existing one.  It seemed to me to be better than ignoring one half of the waveform altogether, despite mismatches between Vactrols.

Proper operation of optocoupler based comp/lim relies on two aspects: feedback operation and/or adjustability of the drive circuit. Even this does not prevent the need for selection of optocouplers.

Agreed in part. Feedback is a common, near ubiquitous means of mitigating optocoupler variations. On the other hand, there are so many time variant parameters to an optotcoupler that there's no certain way of determining  measurable selection parameters

As I mentioned earlier, the only notable exception I know of is Ted Fletcher had a feedforward opto-based comp that involved servoing the optocouplers. His drive circuit was VERY elaborate.
Since efficient stereo coupling requires perfect control of the attn vs. control voltage law, i don't see how you could get away with including a drive circuitry that provides adjustment of the slope and origin of the control curve.

Absolutely. This was the basis of my degree thesis back in 1973. I build a compressor which used a piece-wise linear approximation of the required input/output characteristic to create a compressor whose performance was independent of the parameters of the gain controlling element.It use d about a dozen 741 chips.

Cheers

Ian

 

[gyraf]

Ian,

Any chance of seeing a scan/copy of said thesis? (I'll swap you for mine, about AI in psychology - in Danish, unfortunately)

I'm currently fighting some control-nonlinearities in vari-mu'ing

Jakob E.

 

[ruffrecords]

Ian,

Any chance of seeing a scan/copy of said thesis? (I'll swap you for mine, about AI in psychology - in Danish, unfortunately)

I'm currently fighting some control-nonlinearities in vari-mu'ing

Jakob E.

I am not sure if I have it any more - it was over 40 years ago. I will have a hunt for it.

Cheers

Ian

 

[abbey road d enfer]

It was not as sophisticated as that. I had seen a number of Vactrol based designs using a single LED (and sometimes not even an anti-phase diode to prevent it acting as a charge pump).

That I call bad practise. I've seen it in FX pedals and I think this is where it belongs (just).

it was a simple, obvious  step to add an anti-phase Vactrol LED and parallel its opto resistor with the existing one.  It seemed to me to be better than ignoring one half of the waveform altogether, despite mismatches between Vactrols.

Agreed. That is acceptable for an individual unit that does not need to be paired in a stereo signal path.

Proper operation of optocoupler based comp/lim relies on two aspects: feedback operation and/or adjustability of the drive circuit. Even this does not prevent the need for selection of optocouplers.

Agreed in part. Feedback is a common, near ubiquitous means of mitigating optocoupler variations. 

Let's be clear; I'm referring to the global operation of the limiter, i.e. the control signal is derived from the output of the limiter. I'm not talking about building a drive circuit using NFB from a dual cell.

On the other hand, there are so many time variant parameters to an optotcoupler that there's no certain way of determining  measurable selection parameters

Agreed. There are limits to what good practise can achieve; linearization of the attn vs. CV law is one reachable goal, that answers your basic requirement. Remodeling time-constant goes much deeper than that.

Since efficient stereo coupling requires perfect control of the attn vs. control voltage law, i don't see how you could get away with including a drive circuitry that provides adjustment of the slope and origin of the control curve.

Absolutely. This was the basis of my degree thesis back in 1973. I build a compressor which used a piece-wise linear approximation of the required input/output characteristic to create a compressor whose performance was independent of the parameters of the gain controlling element.It use d about a dozen 741 chips.

Then you know how hard it is...

 

[ruffrecords]

Let's be clear; I'm referring to the global operation of the limiter, i.e. the control signal is derived from the output of the limiter. I'm not talking about building a drive circuit using NFB from a dual cell.

Yes, we are talking about the same thing.

Then you know how hard it is...

Oh yes, which is why I am still searching for an elegant solution 40 years later!

Cheers

Ian

 

[pvision]

I believe Ted Fletcher used another approach in some of his units - he converted the stereo signal to MS before the opto compressors and back again afterwards

Microprocessor control of the sidechain is used in the RNC and others

Nick Froome