Parallel Signal Path for VCA Compressors (eg S S L)

[gkhebert]

Hi Folks,

At Crusty2's request, I'm jumping in here quickly. The output current from the VCA is inverted with respect to the input. (This is true of all of our VCAs -- 218x, 43xx, etc.) The output voltage from the I/V converter opamp is in phase with the input voltage.

Thus, to keep the "wet" and "dry" signals in phase, the buffer for the dry signal must be inverting if its output is to be summed into the I/V converter summing junction.

Regards,
Gary

 

[Boswell]

Thanks, Gary. I was beginning to suspect as much, although it would be nice if the data sheet contained that information.

It was the only explanation I could think of for the increased compression pumping reported by Chrissugar when he wound up the fraction of dry signal in the output. It meant that what he was hearing was the difference between the wet and dry, i.e. just the compression. Now explain why he reports the effect is not there with a unity-gain buffer in the dry path...maybe he only tried it in the janitorial configuration (wet-dry wipe), where the effect would be less noticible.

 

[chrissugar]

Thank you Gary,
that explains the whole situation.


[quote author="Boswell"]It meant that what he was hearing was the difference between the wet and dry, i.e. just the compression. Now explain why he reports the effect is not there with a unity-gain buffer in the dry path...[/quote]

Boswell
Probably, the confusion was induced by the fact that I didn't specify that I did the initial test with Jakob's description, and it didn't work, and many month later, I tried the parallel compression with buffered dry signal and panorama type pot with an optical comp I was working on. So my conclusions were related to the other compressor (noninverting one), but my logic said that it should work the same way with any compressor.
Now it is obvious if the VCA inverts the signal, even with a buffer the results are similar. So probably the solution in this case (SSL comp) is to use an inverting opamp for the dry signal.

But now with this new info related to the VCA chips, it looks like the SSL comp as is inverts the signal, so probably a solution would be to rewire the ins or the outs.

chrissugar

 

[gkhebert]

[quote author="Boswell"]Thanks, Gary. I was beginning to suspect as much, although it would be nice if the data sheet contained that information.[/quote]

I could have sworn that this was in there, but I searched the document, and it's not specifically mentioned. (It is in all of the 43xx Analog Engine Data sheets in the "Theory of Operation" section.) I'll pass this along to marketing/sales for the next revision.

Regards,
Gary

 

[Crusty2]

Thanks very much, Gary. This forum is really a terrific place!

Another document to pass along is this one: http://www.thatcorp.com/datashts/an100a.pdf :

The 218X produces an output current signal in
pin 8 which is a replica of the input signal, scaled
(in decibels) by the voltage at pin 3....

Which is probably what Boswell remembered reading.

But now with this new info related to the VCA chips, it looks like the SSL comp as is inverts the signal, so probably a solution would be to rewire the ins or the outs.

Yes, not only the GSSL schematic, but also the A$lan SM@art schematic shows what would appear to be an inversion at the outputs. Strange, because this error would have shown itself almost immediately during testing or a session.

 

[chrissugar]

[quote author="Crusty2"]Another document to pass along is this one: http://www.thatcorp.com/datashts/an100a.pdf [/quote]

Now looking at this document to Figure1 and Figure6 it is obvious that the 218x VCA inverts the signal, but this kind of information should be mentioned in the datatsheet and should be obvious.
I'm really curious how many other VCA based comps are inverting the signal.

chrissugar

 

[matthias]

thank you all for your help.

Now I like the idea with a wet/dry mix pot.

My next design approach would be this:

R10 should work like a simple pan pot.

I've found a pan pot schematic here:
http://www.benchmarkmedia.com/pdf/panpot.pdf

do I need the 2 buffers to get the panpot work, or is it enough, when the inputs are buffered.

I'd like to go directly to a bb drv134 balanced line driver ic.


thanks,
mat

 

[chrissugar]

Hey Matthias

You can use for inspiration the schematic posted by Keith (SSLtech) in this thread:
http://www.groupdiy.com/index.php?topic=8360
It is a general purpose parallel comp circuit (not only for the SSL) used externaly but you can adapt it inside the SSL comp.
I extracted the panpot circuit, just omit C1 and C2

chrissugar

 

[matthias]

ok, thank you but at the output where R3 and R4 are connected together,
do I need still need a buffer for the summed signal, or can I go directly to a drv134?

(I would say no, because both signals are already buffered by the inverting amps)

 

[chrissugar]

Matthias,

The schematic posted above is still not good. You invert with OP2 the dry signal but you also invert the already inverted compressed signal with OP1. This will not solve the problem. You should do only one invertion to have dry and compressed signal in phase.

chrissugar

 

[Boswell]

We're back to counting inverters here. Thanks to Gary, we now know to treat the VCA and its output amplifier as a non-inverting unit. If you use resistive summing, you don't therefore need to invert the dry path.

The DRV134 has a relatively low but unknown impedance input in the range 8-12K, so you would need a buffer between it and the pot or resistive summer..

The wet/dry wipe pot can have one end on the wet VCA amp output and the other (dry) on the output of the input buffer. For this application, I prefer the configuration where you take the output from the wiper rather than using the pan pot circuit where the wiper is grounded and selectively shorts out the signals. The wipe configuration gives you no loss of gain (when no compression), where the pan pot would need make-up gain in the buffer to get you back to 0dB through the whole unit.

Note, however, that the DRV134 has a gain of 6dB (single-ended in to differential out), so the overall gain would depend on your input buffer. But if you use an INA137 in its intended G=0.5 configuration as an input buffer, the overall gain is 0dB, and headroom is preserved.

 

[matthias]

regarding tha t would need an extra buffer for the wet/dry control, I'd go for the simpole "dry level control"

from what I know now, I buffer the dry signal with an inverting buffer and sum it to the inverter after the vca.

then it would look like this.

what about the resistors for the inverting "dry-buffer" is that ok so,
or can I leave R22 out, because of the pot..??

thank you,
mat

 

[Boswell]

Still not quite there. Consider what happens if you wind the dry pot to zero - OP4 loses its feedback and its output hits the rails. I'm afraid there's no escaping the need for another amplifier. It's best employed as a buffer for the pot wiper. Then OP4 can be a simple inverter with 20K input and feedback resistors. Also, the resistor at the top end of the pot allows only up to 50% added dry signal - is that what you want?

 

[gkhebert]

Actually, you can just swap the positions of R22 and R19. Then, scale the R22/R23 combination to adjust the loading of R19 and modify its taper to taste.

You should, however, ac-couple into R17 to avoid gain-dependent offsets. Best practice would be to also ac-couple into and out of R19.

Regards,
Gary

 

[matthias]

ok thanks,

@ Boswell:

100% of the dry input level should be mixed to the processed signal. when I use an extra buffer for the wiper, would this version work?

@ gkhebert:

I would use one cap to couple the vca input and the pot input and one for the wiper output. what do you think?

 

[Boswell]

I would leave OP4 (the buffer) in place, although others may suggest sacrificing pot linearity for the sake of simplicity. You only need one coupling cap at the input, but it should be on R1, the 20K input to the VCA, in order to avoid gain-dependent offset changes at the output. It would be wise to include either another coupling cap in front of the final output buffer, or else use a d.c. servo feeding back to the + input of OP1.

So I would go for the first of the two schematics, but with a 10uF in series with R1. Items like an input and output buffer and small feedback capacitor round OP1 are taken as read.

 

[Crusty2]

Still seems overly complicated for such a simple task. Look at the differential idea again: http://groupdiy.twin-x.com/albums/userpics/10071/BLEND.TSC_-_.pdf
A pot and two resistors and you're done...

You can add a cap at the top of the pot like others have suggested. You can easily have more than 50% dry signal; lower the value of the summing resistor the pot is driving. It's just a simple, two-input summing amp...

 

[matthias]

@ crusty: that looks very simple (but effective) ... when I want to have 100% dry signal (unity) , can I simply short R4 ??

 

[Boswell]

Crusty2 wrote:

Look at the differential idea again: http://groupdiy.twin-x.com/albums/userpics/10071/BLEND.TSC_-_.pdf

That circuit has a serious error - the inverting input of the output amplifier is not a virtual earth. It will have the same signal as the potted down dry path.

The VCA will not work correctly with that circuit.

 

[Crusty2]

Posted: Wed Jul 12, 2006 9:34 pm Post subject:

--------------------------------------------------------------------------------

Crusty2 wrote:Quote:
Look at the differential idea again: http://groupdiy.twin-x.com/albums/userpics/10071/BLEND.TSC_-_.pdf
That circuit has a serious error - the inverting input of the output amplifier is not a virtual earth. It will have the same signal as the potted down dry path.

The VCA will not work correctly with that circuit.

I should leave it to the experts!. Sorry Mat, I thought there was a "free lunch" there, oh well...I guess there is no way around at least the one extra inverting stage.

I still like the idea of the fully bypassed pan pot too. More versatile as Chris pointed out.

As simple as this is, the limitation is that you can't approach fully dry. Not sure if that's really a bad thing...