My re-do of the stripboarded 706 LFO I did last year, only this time I found something else to fret about.

Last time I was kvetching over the reset pulse; this time it was the fact that it drifts all over the place. With my test BC560C transistors as a replacement for the IT132 in the expo converter, after setting the LFO to an initial frequency of 44Hz it drifted to 33Hz over three minutes, when it finally stabilised.

Yep, it’s an analogue synth and I expected some general general temperature movement with a pair of unmatched transistors but… even so, that’s a bit drifty even for me. I went back to my stripboarded 706 and it did the same, so uhhhh… I just hadn’t noticed it previously.

Next I tried some Vbe matched BC557B transistors, and they were better, but still drifted 3Hz or so.

Micross in the UK distribute the Linear Systems version of the old IT132 dual PNP transistor, so I had that to fall back on, but as luck would have it I found someone selling some original ICs in the US.

They arrived yesterday to my pathetic joy – once I’d finished the near-endless pre-Christmas cleaning, I plugged in the new (very old) transistors to test.

When being set to 20.23Hz (swept down from the maximum preset frequency of 33.58Hz), after three minutes it had only drifted to 20.28Hz.

Trying again from the minimum frequency (got too bored to to count, least a minute per cycle) sweeping up to maximum, it hit 33.32Hz and drifted down to 33.30Hz.

There must be some property of the old IT132 that makes it so much more stable – by matching my BC557s I thought I’d removed that from the equation, but it’s possible that they weren’t matched closely enough.

Looking at other Roland schematics, they used the IT132 as the expo converter in the CV clock speed control input of the System 100 104 sequencer, but with an SDT-1000 thermistor to provide some sort of temperature compensation.

…obviously they didn’t think it was worth it in the 706.

I’ve got enough of the old transistors now to make a small pile of 706 LFOs, so I’m satisfied for now. I’ll have a go at the functionally similar 100m LFO at some stage for comparison.

Here’s the 706 schematic with the (minimal) corrections derived from the track layout.

Diverting from the System 700 obsession for a bit, I thought I’d try my luck at surface mount soldering with the OpenMusicLabs BA662 clone, in this case Altitude909’s version with the DC offset trimmer.

I’d already bought a few already built up from Synthcube, but they’re $12 each, and as I (possibly…) needed a big pile of them I was considering getting the boards made and soldering the parts myself.

The first attempt was going so well until I plugged it in, and nothing happened. Then I realised I’d got one of the transistors the wrong way round.

Next I yanked two of the pads off the board when trying to extract the transistor with my shovel sized soldering iron. Then I used tiny wires to reattach the legs to the connected transistor, which was tortuous, and on plugging it in all the smoke came out anyway.

It went from shit to worse when I found one of the Synthcube 662s, plugged that into my breadboarded SH-2 VCA test circuit and that promptly smoked.

Hmmm. All of a sudden now I’m not sure why I’m writing all this out ‘cos it makes me look like a total idiot – anyway, I needed a test circuit that didn’t actually fry the chip.

I went back to my stripboarded System 100 VCA, pulled the CA3080 out, wrote a pin mapping out…

and soldered up a new 662 (one hour with my fat fingers and vast soldering iron, more or less). Wobbly photo of the tiny fucker:

I plugged it in and it worked, in that the LFO varied the volume of the AS RS-95 oscillator. Whoop.

Not really sure if hand-soldering a shitload (however many that is – 8 x 4 plus a few more, possibly) is really do-able if I want to retain my eyesight. Despite smashing up a $12 “chip” and my £4-ish version, I’m still fairly chuffed to have managed a surface mount board for the first time.

SOT-363 is possible but feels like a step too far for doing it by hand on a regular basis, at least with my tools and skills as they are. I half-wonder if it might be possible to do a version of the 662 with a bigger surface mount package size, just to make it less of a nightmare. Having just done a test with an LM3046 I’d be tempted to use SOIC versions of chips in original boards.

Here’s my silly test setup in the meantime. This doesn’t use the 662 buffer, so that bit might still be broken. But I’m optimistic.

The other alternative might be to use LM13700 OTAs – in their SOIC-14 format they’re 65p in quantity of 10 and above from Farnell. There’s already another 662-pinout IC which uses it, I’m only wondering if it’s as simple as that, is there anything else going on under the IC on the synthnstuff version?