I’ve built two buffered multiples modules in the past: A Eurorack one by Horstronic and a Kosmo one by Cory Torpin, both of which are based on a design by Dave Jones.
A design I considered for my first one, but decided against, was the Barton Eurobuffer. At the time I distrusted that design, because you (normally) want a buffered multiple to be unity gain, and while the Barton is designed to be unity gain, the Jones design uses trim pots which allow you to calibrate it to unity gain. I liked having that control. And it turned out the calibration feature was useful to me because it allowed me to work around the volts per octave problem I was having with the Mother-32: I could miscalibrate one channel to compensate for the deviation from 1 V/oct I was getting.
More recently there’s the LMNC buffered multiple.
Since my Eurorack build I’ve learned more about synth electronics and I have views.
On designs
The Jones design uses an inverting input stage for each input, fanned out to an inverting stage for each output. It has 100k input resistors and in-the-loop 1k output resistors (see below for discussion).
The Barton design uses a voltage follower on each input fanned out to the outputs. It has no input pulldown resistors and out-of-the-loop 1k output resistors.
The LMNC design fans out each input to a voltage follower on each output. It has 1M input pulldowns and out-of-the-loop 1k output resistors.
On impedances:
The reason for the Mother-32 V/oct problem is that, while I don’t have schematics, I am sure the Mother-32 CV output has an out-of-loop 1k output resistor while the Befaco VCO definitely has a 100k input resistor. These form a voltage divider that turns a the 1 V change produced by going up an octave into a 0.99 V change, leading to an “octave” that’s an eighth of a semitone too narrow, significantly out of tune. A cure for that would be to buffer the CV with a module having a larger — say 1M — input resistor and in-the-loop output resistors.
You’d need both, otherwise you once again get a 1 over 100 voltage divider. So any of the above three designs, unmodified, would fail to preserve a V/oct signal coming from a circuit with 1k output resistor and going into a module with 100k input resistor.
On inverting:
The Jones design’s inverter on each input followed by an inverter on each output has some distinct advantages. An inverting amplifier keeps the op amp’s inputs at zero volts (unless the output is saturated), which is ideal. A voltage follower’s op amp inputs follow input voltage, potentially to values close to the voltage rails, which isn’t.
On the other hand, there are some drawbacks. With a quad op amp chip, you can make a voltage follower buffer with four outputs, but with inverting stages you can only get three. You need more resistors — five for a 4-output voltage follower buffer, eleven for a 3-output inverter buffer.
Moreover, for accurate unity gain, three of those eleven resistors have to be trimmers, or eight of them have to be precision resistors or hand matched. The accuracy of the gain is limited by how accurately you can adjust the trimmers, or match the resistors, or by how much you can spend on precision resistors. A voltage follower buffer is unity gain pretty much by definition, to maybe something like ten parts per million.
There are two circumstances under which the voltage follower’s connection to the input voltage is particularly problematic. One is if that input voltage is out of range. Anything much beyond the rail voltages is out of spec and potentially damaging to the op amp. With an inverting amplifier, even when the output saturates there’s some way to go before the input pin gets beyond the rail voltage, and even if it does, the 100k input resistor may help the op amp survive. So it’s more robust.
The other is relevant for JFET op amps like TL07x. There’s a phenomenon called phase reversal. With the TL07x it happens when the input is below about -8 V (with ±12 V power). Then the output, instead of matching the input, slams up to the positive rail voltage. This doesn’t damage the op amp, but it certainly isn’t the behavior one normally wants. There are op amps that don’t have this problem, but they tend to have other unsuitable characteristics such as high bias current or an inability to run on ±12 V.
On output resistors:
In-the-loop current limiting resistors are pretty much necessary if you want to send an accurate V/oct signal into a module with 100k input impedance. But their drawback is that they do nothing on their own to suppress op amp instability which can occur when driving a capacitive load. To stabilize the op amp it’s necessary to add a capacitor in parallel with the resistor — not a big deal, except I’ve seen three different explanations for how to determine the best capacitor to use and they give results that diverge quite a lot.
Out-of-loop resistors, on the other hand, do suppress instability with no added capacitor required. But they add output impedance you can’t afford for V/oct signals. Granted, you could use, say, 100Ω resistors instead of 1k, and that would solve the problem. Except: The whole point is to limit current if the output is shorted to ground or a voltage rail. If the op amp is (say) putting out 0 V and the output is shorted to a rail, then you have a 12 V drop across a 100Ω resistor — so 120 mA current. Realistically a TL07x can’t source that much, but even at 70 mA the power being dissipated in a 100Ω resistor is half a watt. Normally I use 1/4 watt resistors, and finding room for eight 1 watt resistors on the PCB would be a problem. (If it really were 120 mA, that would be 1.44 watts!)
The other thing about out-of-loop resistors is they allow you to fan the output of a single op amp stage to multiple outputs. You could use a single op amp chip on each input instead of a quad. But the space savings there would be small.
Choices:
The larger multiplicity and reduced parts count of the voltage follower approach are significant. With through hole parts and 1/4″ jacks on a Kosmo size panel, I can get a pair of four-output voltage follower buffers onto a single board behind a 50 mm front panel. I’d get more real estate for the circuit with inverter buffers, since I’d have to limit myself to three outputs each, but I’m not sure I’d pick up enough to accommodate all the added parts — trimmers especially. If not I’d have to go to a more complicated two PCB design.
The vulnerability to overvoltage damage and phase reversal are concerning, but not very much so. I have no power supply voltages larger than ±12 V in my synth, so it’d be very hard to get higher voltages onto a patch cable. -8 to -12 V signals certainly are possible, but normally any audio or control voltage signals fall into the -5 to +10 V range.
There are ways to add protection against overvoltage or -8 to -12 V inputs, preventing op amp damage or phase reversal behavior… but once again, one would need a second PCB to accommodate the added parts. It doesn’t seem worthwhile to do so when the voltages being protected against are either virtually impossible, or highly unusual and not damaging.
In-the-loop resistors really are needed. As for stability capacitors, they probably aren’t required. It’s mainly a consideration when the capacitive load is large, and I think that would probably only happen if I were trying to send the output over a really long patch cable… like maybe 20 meters.
So: 1M input resistors, a voltage follower for each output, in-the-loop resistors, no stabilization capacitors, that’s my design. And we’ll see if it works out well.
GitHub repo for the design:
Analog Output 