Making an VCA adaptor card for GSSL

[Soeren_DK]

Implant four VCA in the GSSL

Last night I was reading about how to make your own VCA's. Keith pointed out that this would be to difficult and that not what I want.

I read some notes that the THAT corporation made a 2002N VCA chip a long time ago but its now discontinued. After reading some stuff about some people used four 2181 chips to make the same thing as the 2002 and seeing Igor using the same thing in his mixbuzz I thought about the possibility of making a adaptor card for my GSSL that would have the same pinouts as a dbx 202 can

http://www.groupdiy.com/index.php?topic=1250.0
http://users.telenet.be/Rogy/Paralleling%20VCAs%20notes%20Rogy.pdf
http://i251.photobucket.com/albums/gg291/diy33609/SSMIXBUZZ_500.gif

I can't figure out the schematic but I can draw a PCB layout so anybody wants to help me with this? And is it possible? and is it a good idéer?

Cheers
Søren

 

[Ptownkid]

I'm almost done the layout for this, by all means let's keep this discussion going, and by all means don't let me discourage you from working on it as it is a great learning experience. I will have a pcb and parts kit available soon after I order THAT chips for the SB4000.

As for it being possible, absolutely! Keep in mind that they will not sound like vintage DBX modules. They will be cleaner, lower distortion versions of using a single THAT chip.

Cheers

 

[Moby]

I evaluated mine PCB too, based on the THAT 2002. Yes, discrete VCA is real pain in S, matching, thermal coupling etc.... Clean and transparent , well , why not  ;D

 

[livingnote]

I've been in for this idea many times for my little one here - And if it can go into a DBX-Footprint
adapter, sure why not? Not sure about what exactly to paralleletize how but after Keith telling me
it was basically overkill (considering the sonics of a 2180 A aren't the weak spot anymore - and he
even did the antiparallel thing with them), I'd still be in for a schematic of how it's done by someone
who knows howGoesThat, for sake of own curiosity.

You know, IF doesNoGood; kickItAgain:=true

 

[Kingston]

I was never much fan of a single 218X chip. They sound grainy compared to the (usually awfully noisy) oldschool discrete VCA's I've tried. Stereo image suffers.

So I veroboarded 4X parallel trick some time a go.

218X already has very good noise performance so the 4X parallel thing is pointless if you want to improve that. You're getting 6dB (or in that ballpark) better noise floor. You simply won't hear it. Distortion goes down similarly. This you might be able to hear. The biggest change is the stabilisation of the compression performance. It's not like a 218X chip is bad in that regard, but this parallel thing does wonders to deep compression. I noticed people were saying GSSL sounds aggressive. This parallel trick certainly smooths things down. You can dig deeper without getting the odd "blocky" GSSL thing.

Another similar quality improvement to the VCA is changing the control voltage buffer opamp to something thoroughly modern. Good buffer with the parallel 218X trick easily trumps all the vintage VCA crap people obsess. No noise, no low speed smudge, no chip grain. It's a very modern accurate sound.

But all these differences are subtle. If you don't hear a change in sound with different opamps in text book buffering duties, you won't hear any changes with these VCA's either.

If one was to say this thing is a waste of some perfectly good VCA chips, one would have a valid reasons to do so.

 

[Soeren_DK]

I'm almost done the layout for this, by all means let's keep this discussion going, and by all means don't let me discourage you from working on it as it is a great learning experience. I will have a pcb and parts kit available soon after I order THAT chips for the SB4000.

As for it being possible, absolutely! Keep in mind that they will not sound like vintage DBX modules. They will be cleaner, lower distortion versions of using a single THAT chip.

Cheers

If you have done a layout I will skip this one and wait for yours . When will it be available?

Cheers
Søren

 

[Moby]

Here is mine. Not a big deal

 

[Kingston]

Another important note to add, and I've said this before:

do not make a mistake of adding a 100nF feedback cap on the control voltage buffer. It completely destroys the accuracy of the VCA. It makes the buffer a slow integrator that smooths out sudden changes. But we want all the speed and accuracy we can get.

If you're using a unity gain buffer, this won't matter, but otherwise make the cap 22-56pF.

 

[Moby]

do not make a mistake of adding a 100nF feedback cap on the control voltage buffer

I have 100n but on the input as LPF.

 

[Kingston]

do not make a mistake of adding a 100nF feedback cap on the control voltage buffer

I have 100n but on the input as LPF.

That's exactly what I mean. Do you really want all your control voltage cut away from something like 150hz? (lazy to calculate the exact value). All the sharp attack and accuracy completely gone.

 

[Matthew Jacobs]

do not make a mistake of adding a 100nF feedback cap on the control voltage buffer

I have 100n but on the input as LPF.

That's exactly what I mean. Do you really want all your control voltage cut away from something like 150hz? (lazy to calculate the exact value). All the sharp attack and accuracy completely gone.

I'm a little confused Kingston about which 100nF cap and buffer you mean?
When you say control voltage buffer do you mean the TL072 in the GSSL sidechain circuit or the N5534 in the adapter circuit for the 2080/1? They both could be described as "CV buffers" right?

And you probably mean the 100n cap just after the TL072 and the 100R resistor, right? Would be nice to know so I can try this.

I followed some of your suggestions in another thread on using OPA2604 and OPA604 and it made a big difference. I still got to try your suggestion of putting better ICs in the sidechain, however you said it removes some of the "SSL magic" and makes the comp more transparent and "SSL magic" is what I want these days  ;D Still got to try these things for fun....

Regards
JD

 

[Kingston]

Just say no!  :'(

I speak of the 100n on the feedback loop of course. This is the CV buffer, buffering the CV for the VCA's only and directly. It could just as well be a unity gain buffer, like in the http://users.telenet.be/Rogy/Paralleling%20VCAs%20notes%20Rogy.pdf document. In that one the cap won't matter anyway because it's no longer used.

In the pdf there is a low pass filter on the unity gain CV buffer input. I guess this is what Moby was talking about above. That's just interfacing, and not going to ruin anything of course. But I would not do that either, and use as untouched CV as possible. The 1K/120R voltage divider above should be best for accuracy and you tune that depending on if you use unity gain buffer or some feedback. Same difference.

Oh and for the parallel trick this buffer should be some new fast opamp, like the LME49710 that has better low impedance drive (less distortion) for the four VCA CV inputs than the old NE5534.

I still got to try your suggestion of putting better ICs in the sidechain, however you said it removes some of the "SSL magic" and makes the comp more transparent and "SSL magic" is what I want these days  ;D Still got to try these things for fun....

"SSL magic" in this case is inaccuracy in stereo image during deep compression. Dig 10-15dB into anything on 0.3ms attack and the GSSL (and the original SSL bus) becomes either unusable or to use the kind description "an effect compressor".

LME49740 (modern quad) or maybe OPA404 takes care of this (need one for each sidechain). It's such a difference that your preferred old speed and ratio settings no longer work. It dramatically changes how the compressor reacts on the initial transient, especially apparent on fast attack. Maybe you'll find new preferences, maybe you'll hate this "boringly accurate" sound. I need to investigate what exactly this does with an oscilloscope sometime. I suspect it's the fact a slow opamp will not handle the discontinuations of the rectification as gracefully as a fast one.

 

[Moby]

In the pdf there is a low pass filter on the unity gain CV buffer input. I guess this is what Moby was talking about above.

Not exactly the same but similar filter design.Of course there is no feedback cap since it's unity gain so no changes regarding slew rate.  In the variation I use -3db is at 18khz with 100n. Yes it's possible to make it flat, maybe it's worth of trying but i doubt it will make some big changes in the accuracy.

 

[ruckus328]

I just kicked out a layout for this as well.  Just went with a simple "unity gain" setup for the buffer, 22pF stabilizing cap, too be honest not quite sure why that wasn't done in the original GSSL as it eliminates a couple parts - Kingston am I missing something?  The .1uF Moby is speaking of is the equivalent to the .33uF in the rochy notes, which far as I can tell yes - right around 150Hz LPF if you were to use the 20K/.33uF values in the notes......seems really low to me too.  BTW, by my calculations the 1K/120R voltage divider in the GSSL is incorrect (actually harpo originally pointed that out to me a while back).  The oringal DBX's were expecting to see 50mV/db control signal, whereas the THAT2181's are 6mV/db.  The voltage divider was intended to scale it down to this, however the values are off, it's scaling it down to 5mV/db.  A 1K1/150R or 732R/100R would do the trick correctly.  I suspect this might be one of the reasons people suddenly had to start changing the cap from 120K to 127K over by the TL072, to compensate.  Food for thought.

 

[Kingston]

Both the unity gain vs. feedback buffer, and the voltage divider are extremely non-critical. Basically these will only affect the sensitivity of the threshold pot. No point in calculating elaborate schemes here. A small turn of threshold has the exact same effect.

 

[ptron]

kingston would these new opamps u talk about work if i take one NE out and put one LME in? just like that or i have to change values on parts?

thanks this thread is what i been looking for i want to try this quad vca story.

btw.. other question,  the feedback cap. in the original rev 7 schem says 100p;  calibration`s 100n u say NO and adviceto put 22- 56 pf in its place. this is for the quad vca scheme or for 1 vca too?