OTB Mixer Using Current Summation

[mediatechnology]

Nice work John. Very simple using two duals. I think keeping the RC network in there is a good thing.

[JR.]

Ironically to save a few pennies, we could do this with only one opamp per VCA, if that non-inverting node is fed to the CV port unbuffered. However, adding a large shunt capacitor at that NI node to get impedance down would actually increase the noise gain of the inverting opamp to 2x, making the noise with one opamp more than with two opamps per VCA! Lowest noise would be not using any NI buffer or shunt, but then CV port impedance is not as low as possible, so distortion could increase.

I don't have a good handle on the magnitude of that impedance related distortion mechanism. Gary gave me some insight into the internal device HFE (pnp=75, npn=150), so this could be calculated. Crudely the output current of the VCA (audio signal) equals the current change in the VCA devices, this current divided by device beta is the current change at the bases (CV port). This back of the envelope analysis is simplistic and the different beta between P and N a factor.

I'd rather not take any chances so lets buffer both...

JR

[JR.]

I heard back from Gary at THAT. I had sent him a copy of my CV noise canceling scheme. He pointed out that this noise on the CV line is modulated by signal so mainly shows up as N in THD+N measurements. So our improvement will be more for bench test results than audible. I'm OK with that, better is always good as long as it doesn't hurt something else, or cost too much.

For perspective from the AES paper they published when this series first came out the very low noise AD797 driving one VC port exhibits a +N term more than 80 dB down for a +20 dBu signal. Using a TL081 instead as a control buffer that +N noise rises to only -50+ dB down.

I expect our design to be pretty close to the AD797 noise modulation floor perhaps a couple dB worse while we are driving both CV ports.

JR

PS: I also asked Gary about the German article and his reply was that the technique wouldn't help this series of VCA. SImilar to Wayne's comments earlier.

[mediatechnology]

Thanks John. So essentially the schematic here? http://www.picocompressorforum.com/foru ... &start=128

Do you suppose the MC33X78 would be suitable as opposed to an AD797?

I was going to suggest as a "trick" circuit using the 1646 to generate Ec+ and Ec-. The concern would be however the 25 Ohm output series resistors. It would be interesting to try however to see if the Ec modulation from the input back into the Ec ports would cancel with equal value source resistors. That's one thing I wondered about the German article. Of course, if that is a mechanism, he might just come out and say it in the article and my German so poor that I'm clueless.

[JR.]

The only changes to consider now is if we were to scale up to say a 250k pan pot, we could lose the 5V rail and derive control voltages from +/-15V rails. My use of 100k for design was because I have a pile of them laying around, and I was originally influenced by THAT's 5V circuitry, but they were also interfacing for external voltage control. If we develop a control interface a simple micro can crunch all we need. So changing to 250k and tweaking a few resistor values gets us some more KISS.
------
Yes, IMO the MC33x78 (the lower noise one) at 4.5 nV/rt Hz will be adequately quiet and better than what I've seen in professional applications, since my design cancels out 75-80% of the noise from one of those two opamps. I expect modulation noise to be down in the high -70s below signal. If we want to accommodate the "I've got too much money" crowd, perhaps layout for dual opamps, and put the inverting CV drive from two channels in one dual, and the inverting drive in another dual. Those going for the best performance (all their) money can buy could drop a 1 nV/rt HZ dual opamp in that non-inverting slot. I doubt a difference could be measured even on the best bench between a MC33x78 and something better for the inverting CV drive since we are canceling out most of that already modest noise. I guess if I had too much money, I'd put uber low noise opamps even where it doesn't matter, so don't bother parsing the layout like that.. but do stick with duals.
------
You would have to draw up what you have in mind for the 1646, since I don't see any obvious way to use it that would improve things. If we could live with a resistive source impedance we could just grab the NI node from my circuit which is already <90 ohms. But that will generate some equally hard to measure THD in the VCA. I didn't see any plots for THD vs. control port impedance in Gary's paper, but see no reason to investigate that when a buffer is cheap and noise seems more than adeqwuate.

I don't have footprints for VCA, or the sundry other THAT goodies in my schematic capture software. Looking back at my hand drawn master
http://www.picocompressorforum.com/foru ... t=30#p2006
We need to make the following change... DC couple to the bus, cap couple to the input to the VCA across the feedback of the master sum amp, and perhaps add a (simple one opamp) DC servo from the master output back to the bus for DC stability. We could alternately use two resistors and a cap to ground there. The DC servo may be overkill since we will likely cap couple to the output buffer (I think?)

I am inclined to defer to Wayne and Roger to detail the input and output balanced I/O, I have zero experience with those parts (I always rolled my own). The only design consideration is that they be quieter than the VCA at unity gain, which I suspect they are. Use the typical -6dB input gain and +6dB output gain to normal headroom for professional interfaces (or not).

JR

[audiohammer]

Hello to All,

This my first post on the forum.

Is this project still alive? I am very interested in this thread and would like to see a PCB available for testing the concepts covered in the design.

AudioHammer

[mediatechnology]

Thanks audio hammer for both joinging us and posting in this particular thread!

Maybe JR will check in and we can see what the status is. I can build his pan circuit to test if he thinks it's ready.

Wayne

[JR.]

Hello Hammer et all...

I have been a little busy lately and then I had to look up my password again... It seems I need far more passwords to get through any day than I have memory locations left to hold them, so I have lists of passwords as a senior aid.

I have taken the circuit as far as I can take it without melting solder, so if Wayne is willing let me know if he has any unresolved questions.

I feel good about the design, but control values could be scaled (up) for some other popular potentiometer value. I have lots of 100k laying around, but from a blank page design I might use higher.

My friends with the console company, use their VCAs mono in front of the pan circuit so it's not practical for me to convert one of their stock assemblies to check out my bus scheme.

The only aspect of this design I am not completely comfortable with from my paper CAD simulation is the LPF I want to embed in the bus combining VCA. I want this topology to be unconditionally stable for any number of input channels. (I have fed over a 100 signals to L/R before). I don't expect any DIY effort to approach that number, but I want to easily handle tens of inputs if desired. Even for modest numbers of channels I don't want any VCA bus capacitance messing with our in-band phase response.

When we get to bread boarding the actual summing stage we can hang bunch of capacitance on the bus with maybe one or two real inputs and test it for good phase response despite the relatively large simulated loading.

JR

[audiohammer]

Just checking in to see if any new progress on this thread.

Is anyone breadboarding this circuit to prove out the design?

Thanks!

audiohammer

[mediatechnology]

I think that may be my gig once I clear off the bench.

Wayne