This is a Kosmo format module inspired by the Befaco A*B+C. It can do four quadrant multiplication as well as serve as a VCA or an attenuverting 2 channel mixer.
Ringer and A*B+C both are based on the AD633 chip, but other than that the circuit designs are entirely different: Ringer uses a different attenuverter circuit, and includes offset trimmers not included in the Befaco module.
A few boards and panels will be on Tindie soon.
definitely need a video on this one!
Demo
Didn’t know I needed one, but I need one.
I made even simpler version of that some time ago:
Actually, there’s still three resistors that could be removed with basically the same results.
Yeah, attenuators are a good deal simpler than attenuverters (especially Befaco’s weirdly complicated attenuverters). Inversion is nice to have especially for low frequency applications, though.
Speaking of AD633, this just popped up:
“apparently fake” multiplier chips “that were prone to overheating”, you say? Click through to the schematic and, yes, they’re AD633.
It’s painful to pay Analog Devices prices for genuine AD633 chips, but at least they work.
(Could probably make an interesting module out of that circuit, by the way… but it uses two AD633…)
A cheaper alternative to the AD633 is to use an OTA. I used this circuit for my “Ring Mod,” and I’m happy with it. It’s adapted from Ray Marston’s circuit in Nuts and Volts.
It does require some calibration. Use an oscilloscope while adjusting R1 to balance the signal around 0V, and an audio spectrum analyser (there are lots of free, open source options) while adjusting R8 to make sure that the carrier frequency is rejected. I used Friture as a spectrum analyser.
Here is a Falstad version:
Note: Error in simulated circuit invalidates much of what is said here!
I’m not familiar with Falstad but I tried simulating this in LTspice:
What I see is that with a slow ±5 V triangle on the Y1 (blue line) (and pot settings as shown), the corresponding control current (red line) goes from 0.64 to 1.04 mA. So the OTA gain never goes to zero, let alone negative. (In fact AFAIK the OTA gain simply cannot change sign.) With a fast ±5 V ramp wave on the X1 input, the output is the green line, and you can see it does not flip sign when the Y1 signal goes negative.
So assuming I’ve done this right, this is sort of a 2-, not 4-quadrant multiplier — sort of, because the gain does not go to zero when the Y1 input is zero. Which doesn’t mean it’s unuseful, but it’s not equivalent to an AD633 based circuit.
Barton does have an OTA based 4QM. It uses two OTAs, one controlled by Y1 and the other by -Y1. The latter output is inverted and the two are summed. It’s a more complicated circuit than the AD633 one, though definitely cheaper.
OTA:s aren’t exactly free either. If LM13700 costs 5 eurodollars and AD633 is 15, the price difference isn’t that big. Especially if OTA-based needs more components and fine-tuning.
Here’s a link to the Ray Marston article:
Understanding and Using ‘OTA’ Op-Amp ICs
He explains that: “zero carrier output is available when the modulation voltage is at zero volts, but increases when the modulation voltage moves positive or negative relative to zero. When the modulation voltage is positive, the carrier output signal is inverted relative to the carrier input, and when the modulation voltage is negative, the carrier is non-inverted.”
The value of the rejection resistor is important (hence the trimpot). I notice that in your sim, you’ve set the wiper at 0.5, which would give 25k. For decent rejection, it needs to be close to 18k. I think that if you set it at 0.35 (or maybe 0.65, not sure which ;)), you’ll see it working as a decent four quadrant multiplier.
I’m not suggesting that it’s as good as an AD633. If you were building a proper analog computer and required a high level of precision, there’s no question that a chip designed for the purpose would be the better choice. For a ring mod or experimental Lorenz Attractor in a modular synth, though, the OTA version is more than adequate.
You get two OTAs in an LM13700, so even at 5 bucks, it’s a lot cheaper, especially when looking at a project that requires two multipliers.
LCSC sell the XD13700 for $0.38.
This is very strange.
I did try varying that, and this morning I tried again with 0.35 and 0.65. Neither gave 4QM behavior.
To get around any ambiguity with the pots I replaced them with two resistors and set their values to match what you used. I also changed the power voltage from ±12 V to ±15 V just in case that had anything to do with it. As far as I can tell what I have now is identical to your simulation:
And this is what I get:
I managed to get your Falstad to use the same waveforms and frequencies as mine and the results are entirely different:
I suppose there could be a problem with the LTSpice LM13700 model, but I’ve used it with no evident problems before.
I could try breadboarding this but I have something else on the breadboard right now.
Edit to add: I see the Falstad behavior is fairly sensitive to the pot setting; at 29.5k on top it looks more like what I see
But in LTSpice I’ve scanned the full range of that pot at small intervals and it never looks like the Falstad 19.1k result.
I haven’t seen any user reports on the XD13700. Have you tried it?
Just noticed that you’ve wired the pot incorrectly. You’ve set it up as a voltage divider, but it needs to be wired as a variable resistor (alternatively, you could replace it with a single 18k resistor).
As I say, I’ve built it, and it works as per the Falstad sim. The rejection isn’t as complete as in the sim, of course, but it’s good enough that it sounds like ring modulation, not amplitude modulation.
I haven’t used the XD13700, but given that LCSC is the main supplier for JLCPCB, whom I’ve found to be very reliable, I would be surprised if they were selling dodgy parts. This post on the SynthDIY reddit says that they are real.
Gaah, of course. Thanks. I compared both multiple times and never noticed. I changed it (needed more like 17.1k to get best results, it’s pretty sensitive) and it looks pretty much like yours:
Note: I’m seeing an average control current around 780 µA and maximum about 1 mA like this. Turning the pot down to about 8k gets you to around 1.6 mA average and around 2.3 mA maximum, which exceeds the safe limit for the 13700, so don’t do that! Perhaps a 10k fixed resistor in series would be a good idea.
