This is a ring modulator in Kosmo format. No, really. A real ring modulator.
There are lots of “ring modulators” out there that are in fact no such thing. They have similar behavior, and may arguably be better in some ways, but they’re more properly called four quadrant multipliers or balanced modulators or something, whereas a ring modulator, in its truest sense, is a specific design based around two transformers and four diodes arranged anode-to-cathode in a ring. Hence the name. It goes way back, predating Moog, used in radio circuits. Then someone realized it could do fun things to audio signals.
Sometimes people say “ring modulator” and “four quadrant multiplier” are synonyms, and sometimes they say a ring modulator is a specific variety of four quadrant multiplier. But in fact the OG ring modulator is not a four quadrant multiplier. A 4QM puts out a signal proportional to the product of its two input signals, similarly to a VCA, but a VCA uses only positive control voltages to scale a positive or negative input voltage. A 4QM takes positive or negative signals on both inputs and outputs their product.
An ideal ring modulator will do that too, but only if one input — the carrier — is a square or pulse wave. When the carrier is high, the output is the same as the other input — the signal — and when the carrier is low, the output is minus the signal. It works because the carrier biases the four diodes such that two are conducting and two aren’t, and which two depends on the sign of the carrier. Then the input signal is sent on either one path, which doesn’t invert it, or another path, which does.
That’s an ideal RM. In the real world the non ideal behavior of diodes makes for a messier transformation. But that’s at least approximately its behavior. With a non pulse wave carrier things get even messier, but to a first approximation it’s still output equals input when the carrier is positive, and output equals minus input when the carrier is negative.
OK, enough pedagogy. You can build a functioning RM with just two transformers and four diodes. It won’t even need a power supply, just connect it to signal and carrier and you get output. But there are problems with that. For one thing, the output level is much lower than the input. For another, non ideal behavior of the diodes — especially if they aren’t well matched — causes some of the signal and/or carrier to bleed through into the output.
Ken Stone’s CGS Active Real Ring Modulator addresses those issues with some active components (drat, gonna need that power supply after all) before and after the transformers and diodes. On the input side, a variable DC bias can be added to the carrier. I think this has little effect when the carrier is a pulse wave, but otherwise it’s similar to pulse width modulating a pulse wave carrier. Also on the input side, Stone’s design incorporates an instrument preamp to allow putting external signals into the module. On the output side, more op amp stages amplify the small output signal, and mix in adjustable amounts of the input and carrier signals to cancel out what bleeds through the RM core.
I chose to design my own PCB for a version of Stone’s design, because I wanted to make some changes. One was to omit the instrument preamp. After all, I’d included essentially the identical preamp in my envelope follower module and I didn’t think I needed to duplicate it here. I can use the EF preamp output if I want an external signal into the RM.
Another change was the transformers. Stone specified M 0222, which is not very widely obtainable. Matthew Skala’s unpowered RM design (PDF) uses the Triad SP66, which is small, hence good for Eurorack, but expensive. I found the much cheaper Triad TY-141P has similar specs to the SP66, though it’s larger, and decided to use that. Its footprint differs from the M 0222, though, so I didn’t use Stone’s PCB.
As for the diodes, Stone basically says use whatever you want, low forward voltage is good and germanium is “traditional”. Skala went with 1N695 germaniums, and so did I. But some Schottky diodes like 1N5817 have even lower forward voltage and might arguably be a better choice. I’m considering building a second RM at some point using Schottkys and seeing which I like better, or if I like both for different purposes. Beyond the transformers and diodes, I did make a couple of tweaks to Stone’s design, but nothing big.
Schematics, KiCad design files, Gerbers, and documentation in the GitHub repo: https://github.com/holmesrichards/arrm.