For the stripboard perverts, here is a version of the System 700 703a/b/c/d 12db state variable voltage controlled filter, which has high-pass, band-pass, and low-pass modes. It should use matched CA3080s, although I haven’t got round to that yet.

According to analogue-heaven lore, it was the only type of VCF included with the first System 700 models, with the 703e/f/g/h 24db low-pass added later.

I spent a fair while wondering why it was so horribly distorted, before I realised that R33 in the feedback loop should be 1K rather than 10K. Here’s a detail from Yves Usson’s photo of the board:

…yes, I have spent a fair amount of time staring at pixels wondering if they’re red or brown, how exciting for me.

The other problem was the orientation of the stereo resonance pot. Having built it up as suggested by the schematic, the level into the filter gets boosted as you turned the resonance up, which makes sense, but it amplifies the input so much that the resonance struggles to be heard over the original signal. The System 100 VCF has a similar arrangement but boosts in a rather more subtle manner. Here’s the relevant bit of the schematic re-arranged a bit to fit into the page.

Flipping the wires round boosts signal as the resonance is turned down and lights the overdrive LED for a 10v peak-to-peak input; not the correct behaviour according to the manual, which states that the red LED should light for an input of 13v p-p.

For now I’ve just bridged the wires going to the boost side of the resonance pot. It’s possible there’s something wrong in another part of the circuit, given that I’ve laid the board out manually in DIYLC, but I’ve been over it several times and been unable to find anything different.

Before I do a PCB for this I’ll experiment to see if it’s worth adding a resistor across the boost side of the pot to tame the amplification somewhat. I’ve stared at Yves’ photo of the 703 trying to work out if there’s anything extra going on but to no avail.

When listening to the first demos for this on headphones I noticed some familiar fizzy distortion on the lowpass output, which turned out to be caused by the connected LED level indicator circuit, just as with the 704 VCA. I re-recorded the demo with the LED disconnected, but fixing this should just be a matter of adding the 1.5nF capacitor from BL+ (the positive supply to the U8) to ground, marked as C17 in the track layout in the service manual.

There are a couple of versions of the 12db 703, one with LM301s and (according to the schematic) 1.5nF mylar capacitors, and another with CA3140s and 470pF polystyrene caps. The 301s need compensation capacitors, and the 3140s don’t. I started out intending to build the LM301 version, but in the course of debugging the distortion I ended up swapping to the CA3140s (The LM301s should work fine, it wasn’t their fault).

Here’s a quick demo of the System 100 square wave into the 703 VCF – first is the bandpass, second is the lowpass, all sequenced by the MC-4, no VCA – these were done using 1.5nF mylar caps.

And here’s the sound of the different modes using the sawtooth from a different oscillator into the 703 VCF, with cutoff initially modulated by a dubious MFB Dual LFO and increasing amounts of resonance, and then some fiddling about with the CV from the MC-4 and the clone System 100 envelope, all with no VCA. They’re hardly the most musical of demos (all in the key of predictability) but it’s more to get an idea of the character of the filter. Warning: nasty high pitched resonant sounds!

First lowpass:

Possibly unpleasant highpass:

And the bandpass:

…not really your classic fruity Roland sound, but then you should expect that from the 12db-iness of it, and to my mind that’s a good thing to have a different flavour to the usual 24db sound. It sounds slower to react to the envelope than the 5-stage (30db?) System 100 VCF.

Here’s the layout out of DIYLC, including the C17 LED noise fix, but with the resonance pot boost as described in the schematic – should you build it you might need to experiment with this. I’ve been over it a few times checking for errors but it’s always possible I’ve missed something, let me know in the comments below if you spot anything.

After all that faffing around peering at photos of circuit boards and the dodgily reproduced schematics, here’s a stuffed and working Roland System 700 VCA clone board.

It turns out that learning Kicad/Osmond was worth the candle, it was so much less tedious to arrive at a layout that mirrors the schematic. And it looks a lot nicer too. The board says “V1” on the silkscreen because there was a later version of the VCA that used a (cleaner sounding) BA662 for the OTA rather than a CA3080.

Speaking of the CA3080, the original schematic (see Florian Anwander’s site) calls for it to be selected, but doesn’t say what for. I’m presuming it should be chosen for low CV feedthrough, but as with the System 100 VCA all I can think to do is to try a few different ones to see which sounds the best.

There are a couple of problems with the schematic. Q1 is written as a 2SC100GR but it’s more likely a 2SC1000GR, and the sketchily annotated Q2 should be a 2SA493GR. R22 should go to -15v rather than ground, and R17 on Yves Usson’s version should be 68k (as on the original schematic) rather than 100k.

I’ve seen a photo of an original 704 VCA which has the DC offset trimmer at VR7 at 100k rather than 50k, so I went with that. Also the board layout shows a 0.1uF mylar cap at C4 rather than 0.01uF.

It kind-of worked straightaway in that it did the VCA thing, responding to a CV input from my clone System 100 envelope, but did so with a set-your-teeth-on-edge distortion. Not the nice kind of fuzz.

I suspected the 2SC1000BL I’d used in the push-pull amplifer/buffer after the 3080 – the schematic specifies a -GR variant – but narrowed it down to something in the LED level detection circuit. Stringing a 15nF capacitor (…as in the service manual) across Vdd and ground of the CD4001/IC3 abated the noise.

The original schematic is confusing when it comes to showing how the CD4001 (which is the brains behind the level indication LEDs) should be powered, being apparently run off both the +15v line and the 14.4v line via a crossed-out diode. When laying out the board I took the crossing out to mean “don’t do this” but it’s probably more like “this is an addition”, or in this case “this is a bodge”.

Here’s how it goes, not that a VCA demo is very exciting, bass going boom on the scope – square wave from System 100 VCO into 703b VCF with the cutoff turned right down so I could check for distortion into the 704 VCA set to linear response, controlled by a System 100 envelope – all clones.

The exponential response is much faster than the linear input for the same envelope settings. I’ve not played with it much yet, but it seems like it’ll make for a really snappy-sounding voice.

My board design was a bit, um, idiosyncratic, let’s say, with wonky haphazardly curved thick traces, no ground pour, and some dodgy spacing like this, although it might helped if I’d used a smaller voltage cap:

But it was my first(ish) go, and I was trying to balance getting it right/making it pretty with finishing it. For future boards, where there’s room I’ll try and put the reference on the outside of the body of the component so it’s visible after soldering when I’m trying to debug.

Here’s the original schematic updated with corrections:

Update: Sep 2016

Via Matrixsynth, a picture of the new style VCA.

Note the BA662 chip replacing the CA3080. The owner of this reckoned he preferred the old style 3080 version.

Here’s my stripboarded attempt at the Roland System 700 706B VC-LFO, which is the one without delay, mostly because I didn’t want to use up a CA3080 or have to finagle a 13700 for a substitution.

It’s based on a not-so-common N13T1 programmable unijunction transistor, which I subbed for a fairly similar 2N6027, and a pair of BC560C in for the dual PNP IT132 expo converter. Actually it turns out that Linear Systems now make this part (IT132 datasheet) but I wouldn’t think it critical. The rest is all the usual period 733/945/1458/2SK30AY.

I used Yves Usson’s redrawn schematic from his System 700 page. Most Roland service manuals are available in good quality scans but the available System 700 schematic is hard to read in places. The only difference from Yves’ drawing is that I’m using a 10K resistor at R61 instead of 470K, cos it’s pretty clear on the original schematic.

When flipping the switch for the first time, it didn’t quite work properly much to my slightly perverted joy, because I’ve come to like trying to work out why these things are broken. The sawtooth offset was slightly low, which led to the sawtooth and sine being mangled. The problem was tracked down to an uncut track (boring!).

While I was testing this I was running at +/-14v, initially intending to put it in my infinitely expanding System 100 expander clone. The voltage levels aren’t quite the same, I’m only getting 8.78v out of the saw rather than the specified 10v.

There’s also quite a spike in triangle and sine, presumably down to the usual reset glitch in the conversion from saw to tri.

It’s only audible at the top of a sweep of an audio oscillator as a slight tick – here’s a demo of it sweeping the System 100 VCO.

If we zoom in a bit on the waveform…

…we see it’s about 840mV high, and lasts for – well, I’m not sure where you would normally count to, but you can see from oscilloscope readout that I’ve measured the whole glitch to last about 100µs.

For a little while I wasn’t bothered about this, but then I got irritated by it. Jacking up the supply voltage to 15v didn’t solve the problem, and neither did using an original N13T1.

Reading around a bit, it turns out that parasitic capacitance is a feature of veroboard/stripboard.

“The stray capacitance can be much easier to deal with, but removing any unused copper can help. That said, …I built a prototype function generator on Veroboard, and couldn’t get the square-wave risetime below a few hundred nanoseconds… When we went [to] the PCB version, the risetime was too fast to be measured with a 20MHz scope!”

(from Bodge-o-tronic: Veroboard! thread on vintage-radio.net)

Following the advice on minimising capacitance problems I removed the flux (with a can of Maplin’s flux removal spray I had in the shed, and never used – whoops), moved the timing capacitor closer to the 2N6027 and cut a couple of unused tracks, which brought the glitch height down to 670mV. Better, but still not great.

Peering at the circuit, it seems C3 (470pF) is there to smooth out the glitch – adding a 1nF cap to ground from pin 3 of IC2 helped, but didn’t entirely get rid of it. Too big a cap would start to noticeably affect the waveform.

I’m wondering if re-doing this as a proper printed circuit board might improve matters, though I suspect it won’t remove the glitch entirely. I’m also wondering if stray capacitance was the thing that made my System 100 VCO build so treacherous.

Going back to looking at the System 100 LFO and MS20 LFO it seems they are tri-core, so don’t suffer the same problem (correct me if I’m wrong, I’m sure you will).

Anyway with all that in mind, here’s the layout from DIY-LC. I added a basic LED rate indicator that only shows one half of the waveform. Can’t imagine anyone will fancy building this but, just in case.