Matrix-12 control repairs

Once again, a repair documentation just for the sake of having these bits of info written down somewhere. Notebook as good as any ;) .. Just to intro this monster of a synth, here’s what Peter Forrest writes about it in his book “A-Z of Analogue Synthesizers“:

The Matrix-12 is a 61-note synth with three big LCDs, six value knobs and about 60 switches on the front panel. Internally, it is 12-voice polyphonic and 12-part multitimbral, each part responding to its own MIDI channel. Each voice provides two big fat VCOs (offering triangle, saw, pulse, and noise waveforms), one 15-mode VCF (offering various low pass, high pass, band pass, notch and phaser types, all complete with resonant squelch), five LFOs, five envelope generators, 15(!) VCAs, and one FM modulation generator for oscillator sync and cross-mod madness.

There’s also a lag generator (a sort of portamento that can be applied to anything, not just pitch), three keyboard-tracking generators, four ramp generators (a very simple two-level envelope) and, finally, one noise generator. And that’s just one voice. There are 12 inside here! All this sound-creating muscle is useless without an equally powerful control system, and to this end Oberheim came up with the ‘Modulation Matrix’ from which the machine takes its name. This system enables 20 connections per voice between virtually any parameters. For example, to increase the speed of an LFO with time, select LFO speed as the destination and a slow attacking envelope as the modulator.

The Matrix-12 that landed on my desk was in need of lever potentiometer and rotary encoder replacements. As is, the lever potentiometers had problems tracking smoothly (eg. resulting in pitch “warble”) whereas the encoders did a lot of skipping back and forth.. Working but unusable, so to say.

One of the the original Matrix-12 encoders removed from the circuit.

I chose to start repairs from the encoders, since spares were supplied with the synth (how considerate!). These were of Alps type EC11B15202AA, having 30 detents and sending 15 pulses per revolution (DPR & PPR respectively).  I didn’t bother measuring the original encoders more than checking what’s their DPR (=30). For some reason, the middle pin proved impossible to desolder clean off the PCB, so I ended up cutting the that as short as possible and soldering the cable on top. Not exactly ‘proper’ but better this than increasing desoldering temperature and risk damaging the board. As the spares were way smaller, adding a short extension cable was also necessary.

Comparing the ‘new vs. old’ feel, the detent “click” is way more softer on the new ones. Most of the times, they also require two clicks to increment/decrement the value being adjusted. Thus, you do end up doing quite a lot of dialing when adjusting for more than a few digits/steps. If there was support in the firmware for a encoder with a push switch, adding a decade incrementing/decrementing feature to that would be a welcome improvement.

New set of encoders installed. A bit of wire needed between the part and board because of a way smaller part.

Furthermore, I’m thinking a non-detented rotary (with a higher PPR value too) might actually prove better in use than a detented one. But no such spares “in stock”, so I chose not to bother myself with ordering this and that just for the sake of experimenting. If you’ve done a similar repair but with a different encoder, I’d be interested to hear how did that work out. Do leave a comment here with part specs (PPR etc.), if so.

If you have no clue what a Allen set screw looks like..

Ok encoders sorted out, on to the levers!  I somehow wish I would’ve taken detailed step photos during this part of the repair, since explaining the following mech stuff in any KISS way is difficult. And to add to this ‘incomprehension factor’, the repair itself got (sort of) ugly right off the bat.  So try to bear with me and maybe try IDing stuff from the pictures below..

The Matrix-12 has two lever units, both consisting out of similar parts: There’s the lever itself, a 10k potentiometer, a L-shaped bracket (with side flaps), a spring mechanism and a (brass)  calibration locking disc. The flaps on the bracket restrict the lever movement (and thus, potentiometer resistance) range whereas the spring mechanism centers the lever. The resistance range that is actually used measured about 2k with the center (rest) position set to 5k.  4k to 6k, in other words.

The lever assembly without the lever

The “ugly” bit was caused by the weirdo-sized Allen set screws (four per lever assy) for which I was totally unable to find a proper wrench for. These screws are used to fasten the locking disc and the lever to the pot shaft. I’m guessing they must’ve been of some imperial size, since a 2mm tool tip was too big, a 1.5mm was too small AND having a ‘full set’ of .1mm size steps in anything else than drill bits sounds ridiculously pointless.

I then switched to a torx screwdriver set. Here, a T6 tip proved too small and T7 too big, but the overall fit was slightly better. In comparison to the 1,5mm Allen, the T6 actually gripped a little bit when the tool was slightly tilted. This approach turned out a working solution (can you guess already?).. for all but ONE SCREW!

I gave this remaining screw a go with the T7, helping insertion by hammering it in.. Now it was a snug fit for sure, but the screw just stubbornly refused to move. Couple of more attempts and, just for the fun of taking this simple disassembly task to a whole new level, the screwdriver tip announced “fuck off, not putting up with this shit anymore” by SNAPPING in half! I know that eg. if kept in contact (esp. fastened overtight) with other metals over longer periods of time, brass threadings have the tendency to get stuck. But this amount of tightness felt maybe juuust a little bit ridiculous. Anyway, with the tool tip stuck inside the screw, well, not much choices..

..though I gotta admit, fair amounts of drilling was involved prior to “oh hammer, please don’t hurt ’em”. Simply too much mech bits in the way to properly clamp the locking disc to a vice to do any hammering (nor drilling), so the potentiometer shaft + locking disc combo had to come off. First, drilled away the shaft locking clip to detach the combo, then drilled away both the screwdriver bit AND the Allen screw. Believe me, drilling away an entire screw isn’t anywhere near as easy as doing a hole on ‘soft stuff’, like a sheet of aluminium ;)

Lever mech and potentiometer disassembled, locking disc can haz new set screws.

Finally, maybe something like +40 minutes into drilling/milling/hammering, I had all the parts separate. Since the screw had to be drilled away, the resulting hole had to be threaded for a slightly bigger set screw. The ones I ended up using were M4 * 6 with a 2mm Allen insert. Matrix goes metric, yay \o/

Doing this replacement lead me to think that maybe, having a single set screw use a different wrench than all the others would be just downright silly..  Yeah ok fuck it, let’s enlarge and re-thread all the set screw holes to maintain consistency. You might be inclined to think that this bit of extra work is pointless but, in my opinion, this is how you do mech stuff properly.

The locking disc with M4*6 Allen set screws

Gotta admit that this locking disc is actually a quite neat solution to handle the potentiometer calibration. Once you have all the other mech assembled, centering the potentiometer becomes very easy. Just clip multimeter leads to the terminal where wires are soldered, set the potentiometer to read 5k, move the disc to its place and lock it down. Dead-center in no time..  And yes, the pots measured the same despite in-circuit. I do occasionally check for these kind of things, you know ;)

Log taper potentiometer installed for testing. The shaft still needs a bit of trimming (and the white cable some resoldering ;)..

As for the spares: I actually ended up replacing the original potentiometers with ones having a logarithmic taper. Once again, “what is ‘in stock’ and what isn’t” ..   The part type is Radiohm CIP20C. I didn’t bother measuring the taper curve from the original parts, but I guess it’s safe to assume that they were linear type. It just makes more sense considering the use.

Not that I just blindly installed the presumably-wrong-taper-spare either; of course this got compared! Despite the taper curve, to me the resulting pitch bend sounded pretty much the same with both. And verifying things resistance-wise, if calibrated to read 5k in the center position the overall range measured similar +/- 1k regardless of taper curve. Maybe this is because of the limited range used, maybe the log potentiometers I have were of the “cheap” type  (check taper diagram here, if you like).. Who knows.. Not that it matters really; adjustment issues gone, synth in a working condition and shortly after, I was deep into “synth lead solo land” ;)

Phew, this post got a bit longer than intended.. But I like repair stories and think they’re good for learning a variety of things. My contribution to that here should (at least) give you some idea of how even a simple repair could end up taking a way-longer-than-expected detour. Say, if you do electronics repairs, it might be a recommendable habit to say “ok give me time to check things through” instead of promising “come pick this up tomorrow”. You never know when a case like this might land on your desk ;)