(If you want to build this, please check the note at the end…)
My attempt last year to build the Doepfer A-110 heater circuit on a small daughterboard for my 702 VCOs went up in actual smoke. It took me a while to realise I’d used a bad CAN-10 footprint for the ua726 connection, so now we have this
Is it gonna work? Is it gonna fill my room with smoke?
…and nothing is on fire! it works!
It gets vaguely hot – putting my finger on it, it seems warm – and pointing my (newly acquired, slightly shit) laser thermometer at confirms this.
It was a drag to get it in the dodgy socket I’d fashioned out of bits of a standard DIP one. Next time I’ve resolved to build some sort of temporary jig (inevitably out of blu-tack) to make it easier. Getting the legs the same length would make it rather less difficult.
While breadboarding the circuit I tested both surface mount and through hole versions of the 3046, and I’ve gone for surface-mount – partly due because it’s being readily available from the usual places, but also because the SOIC version seemed to get up to temperature more quickly. I suspect this was because the through-hole version is bigger and takes longer to heat up anyway, but also perhaps because those massive legs conduct the heat away into the breadboard, though I can’t prove that.
The components are a bit of a tight fit on the board (which measures 46 x 20mm), but I had to get it small enough to avoid covering over the existing trimpots. I like Dr Korg’s attempt at this but that’s beyond me at the moment.
Bunging my makeshift frequency counter on the output showed that it took longer than I expected to get up to a stable temperature – after about four minutes, it was still slowly dropping. When I returned after wrestling my son about ten minutes later, it had stabilised – a watched pot, and so on.
I’ll write up some numbers another time, and compare it to the original 726s inside my System 100. I’m interested to see if insulating the 3046 might help.
If you’re going to be replacing a ua726 in a circuit (I’ve got an original SH-2 oscillator board which needs a couple) you might need to change a resistor on the feedback of the CV input opamp to give the scaling trim pot enough leeway to tune the oscillator. I had to alter R9 on the 702 from 50k to 68k.
Here’s the schematic – should be familiar enough from the Doepfer version.
If you want to build it, I can’t guarantee it’s going to fit on your board, whatever it is, but I’ve shared it on oshpark, where it’s $7.20 for three. There’s a BOM and further instructions there too.
Update: Feb 2018
I got round to testing this in a System 100 VCO clone, and it didn’t work.
Eventually I realised that the transistors are (sort-of) the other way round in the 702 – input comes into the base on pin 1, and the current comes out on pin 4, but on the System 100, the input comes into pin 2 and comes out on pin 9.
All of which wouldn’t matter but for the fact that we’re treating pin 2 as our ground for the heater board, and it’s pin 1 that is grounded on the System 100 (and 100M, and SH-1 and -2, and probably most other Roland synths from that time).
It’s possible to cut the tracks that lead from pin 2 to the ground, and wire up pin 1 to ground – I’ve got my System 100 VCO up and running like this – but it’s a faff. I’ll do another couple of versions of this board soon.