Using cascaded VCAs in compressor designs?

[abbey road d enfer]

I'm not looking for justn20 db of compression or so - this is for much higher range,

OK, so what is the expected gain range?
Anything more than 100dB is going to be challenged by noise.
The problem with your proposed schemo is that you may have stability problems because the VCA has a definite phase shift: two in cascade may result in excessive phase-shift and oscillation. The other problem is loading the output of the second one with a resistor is not good for distortion. These VCA's are really designed to be loaded with a VE input stage. And anyway one single opamp is hardly capable of providing 100dB gain over the audio range (I assume, maybe wrongly, it's an audio application). I would share the gain between two cascaded stages.
Using the THAT VCA in the NFB of an opamp is not simple since you need to reverse polarity, which implies another opamp in the loop, with its associated phase-shift.
I'm not saying it's impossible, but I'm afraid you're on your own here.

 

[JohnRoberts]

Thanks for all the useful posts. I remember a high end board that used driving both VCA control ports via an inverter and I was never sure why that added complexity was included. Must have missed that it improves noise in the app notes. Thanks for sharing that bit.

My understanding is that it improves linearity (not noise) by cutting the maximum control voltage in half. This is mainly used in fader automation where large attenuation commands occur.

JR

 

[mr coffee]

@Abbey

Thanks for the feedback. I'm estimating the maximum total gain will need to be 50-70db. This will need to have a maximum limit set to prevent acoustic feedback to the guitar pickups. This is a design for a specific guitar with low impedance, low noise pickups.

I was thinking, perhaps wrongly, that the inverting input of the op amp in the schematic is a virtual ground/earth, and the resistor acts as another inverting input. IIRC, the 2081 data sheet says that multiple inputs can be summed at the VE point, so I assumed that would be essentially the same thing. No?

I have also considered doing two cascaded stages as you suggest, although the problem of the phase inversion in the VCA without adding an additional op seems confounding. I know virtually nothing about applying negative feedback to an op amp applied to the non-inverting input. Anyone here?

@JR

When you say "Linearity" re: using both control ports, do I assume correctly that you mean linearity of the control response, or are other linearities involved?

Thanks, guys!

 

[JohnRoberts]

@JR

When you say "Linearity" re: using both control ports, do I assume correctly that you mean linearity of the control response, or are other linearities involved?

Thanks, guys!
[/quote]

[/QUOTE]

Linear transfer function, means low distortion...

When you try to put 70-100 dB of gain change into just one control port the DC voltages relative to other internal nodes can become significant. Driving the two control ports differentially cuts that maximum voltages in half.

If you look at what is going on inside current ratioing VCAs we are dealing with Vbe junctions. It is not good for a control port to rise to a major fraction of a Vbe junction forward drop.

I repeat it is difficult to measure VCA THD at -100 dB and this phenomenon may be subtle, but when in doubt do what is right.

JR

 

[mr coffee]

Thanks for clarifying, John. I agree that that kind of non-linear distortion is extremely hard to measure when high attenuation is commanded from a VCA. I get the sense that It often gets lumped into distortion + noise.
That's part of why I am considering a cascaded design of two VCAs operating in their more common attenuation range.

Any wisdom you can share about applying negative feedback sent to the non-inverting input for this application using a single (inverting) VCA in two cascaded stages?

Thanks for your reply.

 

[JohnRoberts]

Sorry it does not pique my interest.

JR

 

[abbey road d enfer]

I'm estimating the maximum total gain will need to be 50-70db

OK.

low impedance, low noise pickups.

Most of the noise comes from external magnetic fields. Only humbucking pu's are less noisy than single_coil. There is almost no difference between a High-Z and a low-Z pu of same construction. Don't put too much hope in low-Z.

I was thinking, perhaps wrongly, that the inverting input of the op amp in the schematic is a virtual ground/earth

Well it should -then the resistor to ground is useless- but the inverting input carries a signal that's identical to the input signal, which may be an issue in terms of distortion.

, and the resistor acts as another inverting input.

I don't get it...

IIRC, the 2081 data sheet says that multiple inputs can be summed at the VE point,

That is true.

so I assumed that would be essentially the same thing. No?

No.

I have also considered doing two cascaded stages as you suggest, although the problem of the phase inversion in the VCA without adding an additional op seems confounding.

That's an issue I mentioned.

I know virtually nothing about applying negative feedback to an op amp applied to the non-inverting input.

TBH I really never thought about it. Don't seem easily feasible without an other active stage.

 

[FIX]

The general rule for noise with a VCA is 2 in parallel will drop the noise about 3-4 dB (as it's random noise, so 6 won't happen), 2 more will drop is about 1.5-2 dB, and double that and maybe another dB. The fanciest ones I saw were the DBX ones that had 8 in parallel. I used 4 in parallel when I did the 2500 compressor.

Since it's an impedance thing and not actual noise cancellation, I wouldn't thing series would help.

You can double the speed of the attack and release by driving the current ports differentially, which doubles the attack time, allowing the RMS detector to be slower, lowering distortion, but you have to be REALLY CAREFUL of the location of the pop-amp driving it. Always use a dual, and one that can drive a very low Z source and remain stable. And locate it as close to the E+ and E- pins of the VCA. Good decoupling caps everywhere, as close to the op-amp and VCAs as you can. THAT recommends that you don't do that, but after speaking to one of my friends that is an engineer there, it can be done, if you are careful.

 

[abbey road d enfer]

You can double the speed of the attack and release by driving the current ports differentially, which doubles the attack time,

So doubling speed doubles time also? C'mon...

allowing the RMS detector to be slower, lowering distortion,

Using the RMS voltage differentially with equal negative and positive excursion results in doubling ratio. This is not usually the intended goal.
The Control Voltage has to be dropped by half before being split differentially. The net result is exactly the same speed/time (and ratio).

but you have to be REALLY CAREFUL of the location of the pop-amp driving it.

Of course.

Always use a dual,

Why would it be so?

and one that can drive a very low Z source and remain stable. And locate it as close to the E+ and E- pins of the VCA. Good decoupling caps everywhere, as close to the op-amp and VCAs as you can. THAT recommends that you don't do that,

What part do they don't recommend? Decoupling? Locating the opamp close to the control pins? I've never seen such a warning anywhere in THAT's litterature...

 

[radardoug]

OP, dont you think you should do some more work to refine your concept? Current technology tells us that more than 20 dB of gain reduction is not often needed. So why would you want 70 dB? Are you pulling up the tail of the guitar signal? What happens when the noise then becomes louder than the signal? Going to sound like ****.