Michigan Synth Works MSW-810

General info

This synth is all sweet spot, just like it’s reference point, the CMU810 and its siblings, MC202 and SH101. The Michigan MSW-810 is a super solid machine with massive chassis, high quality parts and very well curated interior (op-amp types etc.). To be honest, first I found that extra VCA/mixer setup lame… But when it dawned on me that this can be used for modulated modifications (FM, feedback, etc.) I realized this is the cat’s feet, as it were!

Listen to filter feedback resulting in different flavours of overdrive…

In the next clip I use the VCA2/Mixer to do linear FM on the MSW VCO – external VCO and extra envelope to VCA2 are used to get this effect.

Two fixes for kinda-bugs in the original design

First off, the two peculiarities described are in the original design. Raf’s rendition is extremely faithful to the original, so, naturally, we finds these also in the MSW.

ADSR plop fix:

As mentioned in the CMU-810 Manual, the ADSR produces a short pulse in some settings. This can be heard very well when the filter self-oscillates w/o any waveform input and VCA set on gate mode. With ADSR sustain and decay at very low settings and VCF ENV slider fully closed, the pulse is audible at the start of your sound. Pulling the attack up delays this short pulse, which only vanishes when decay/and or sustain are pulled up. 

Plop on rapid cap dis/charge looks like this…

Fixing this: Adjusting the +12V trimmer does not eliminate this pulse but placing a 10uf capacitor between ground and the emitter of Q19 softens the amount of timing-capacitor (dis-)charge and rids the pulse plop. One side-effect of this, however, is that the envelope is slightly slewed when charging from zero A and zero D to higher sustain settings, which means with A/D at zero and Sustain up, you get some short whoop now and then.

Additional declick (adapted from MC202 manual): put 100nf cap between the base of Q16 and GND.


Another design peculiarity (again, this is Roland’s doing, not MSWs) that results in unwanted LFO modulation happens when sending control voltages over +4V into the VCA2 CV socket with CV response is set to EXPonential. This can be fixed by adjusting the CV input impedance and increasing the internal prewired CV accordingly. Results between internal/external CV use are now identical, with ratings of zero to +5V (hard limited). Input impedance now also complies with Eurorack ” standard of100k.”

Moreover, this mod also establishes unity gain between SYNTH OUT and MIX OUT when both SYNTH volume slider and OUTPUT volume slider are fully open/full +5V control voltage are sent to VCA2.

What to do: remove R1, increase R8 (470ohm) to 1k, increase R50 (10k) to 100k. Remove D15 and limit right hand terminal of R50 to +5V with new diode. Reduce R10 to 470ohm to adjust EXPO CV span.

Now, okay, folks have not minded these two oddities since that CMU came out in 1983, so, you know, maybe not that much of a biggie. My studio environment is 2020’s though, so in this case I personally prefer going non-original…‚

DIY version and factory build – IR3109 and/or AS3109

I regret having missed the short run of the DIY version, yet luckily I could compare that and my factory build recently when modding and calibrating a DIY build for a friend. What’s immediately apparent: in the DIY version you get DIP/DIL footprints and can place a Roland IR3109 if you have one. Can you spot the difference between the DIY (with Roland chip) and the factory build?

reveal of what’s what is at the bottom of this page.


As for technical documentation, I am not aware of an officially available service manual or schematic, but for most parts the MC202 schematic sorts you out. MSW themselves have good info on their page, and there is a MW thread and an FB group for the DIY version.

So far, as I don’t ant to drill that case yet, I simply made some minor but efficient changes – one switch for changing the filter response to bandpass, and another switch highjacks the second VC/mixer circuit for linear VCO FM or for feedback/external sound input into the filter. These three tricks already massively expand the sonic possibilities at hand. Find them and other possible mods sketched below.

General Pitch

Lower that pitch! – It was only until I could compare the factory made MSW with a DIY version that I noticed there’s some two octaves difference in terms of pitch. So, I lowered the base pitch of the factory version by wiring a 47K resistor between the junction of VR11 (the VCO tune trimmer) and the bottom terminal of R67 (280k) and the -5V test point.

Increase tune knob range: I did simply wire another 1M across R100 (1M) in order to do so.

Fix Octave select intervals: I noticed on two units (after lowering the pitch, but dunno if that’s causal or not) that octave select didn’t give me clean octave intervals. This was easily fixable by wiring a 1M resistor between the left hand terminal of R154 (faces R 153) and the upper terminal of R155. Here before/after (sorry for intermittent noise in the signal… clumsy recording…)

After you’re done, set the tune knob to the middle position and tune your not per VR11 VCO tune trimmer. This way you have enough leeway up/down from here.


The MSW-810 uses the Alfa Par 3340 VCO (datasheet). Like 101/202/810, tri wave is unused and so are couple of other options the 3340 affords. On my unit I simply added a Lin FM input for modulation, but if you go a route of tapping individual wave outputs and are confident to solder at Pin10 of IC11, tap the tri wave too!

Very simple linear FM with untracked external VCO – pretty techno though…

Linear FM: wire a 1M resistor and a 100nf capacitor in series. Wire the fee resistor terminal to the LinFM point labelled above (bottom terminal R126). The free capacitor terminal goes to a switched socket tip: NC lug of that socket goes to ground. On my unit I don’t use an extra input but wire this via the VCA2/Mixer section. Basically – 820k resistor instead of 1M and instead of input socket my cap goes to the point marked MIXOut above (R23 top terminal) — via an A/off/B switch. This way, you can use any input1 for linear FM and utilise VCA2 for cv-controlled modulation depth. More below…

More is possible (soft sync, hard sync etc.), but as long as I don’t fully modularize this thing I won’t get into that – it could be more interesting to place another 3340 in there…

Filter and VCA section

The MSW filter IC is an Alfa Par IR3109 clone, the same type I use in my Model F filter. The setup here is by the (holy) book of Roland with the classic diode clipping and passband level loss compensation.

Extra Filter Input: Connect the lower terminal of C16 (or simply top terminal of either R144 or R143) to a 47k-100k resistor, which is connected to an extra input socket (best via a 10uf electrolytic cap). Adjust value of that resistor to your needs. Here you hear 47k, which gives you quite a hot input (cool for feedback as well!):

Extra Filter CV input: connect your CV in socket via a 150k resistor to the bottom terminal of R70. Cool for FM stuff, for external LFOs and envelopes, abut also rad for using the main output as FM modulator – 303 devilfish style…

Tweak maximum resonance: R145 (15k) sets the maximum resonance level. Reduce it for more (beware screaming!) or increase for less maximum resonance.

Filter Pole Mixing – Bandpass etc.: 4-pole filters based on/related to the IR3109 most often come in lowpass form, yet you can create other filter response types too by simply mixing the outputs of each filter pole in different proportions. In my case, I use a switch to toggle between rso comp lowpass (as is from factory) and bandpass. Easily done: replace R32 and R33 (each 150k) with through hole resistors of the same value but don’t solder in the upper terminals (connected to C22) but tie these together and then to an A/B switch. Then wire in a 100k resistor into the PCB hole labelled -12db in the pic above and another into the hole marked -18db. Tie these tow resistors together and wire them to your A/B switch. The middle pin of your switch goes to the PCB terminals where R32/R33 top used to be (/connect to C22). You can create a whole deal of other responses (check Emilie Gillet’s paper and Electricdruid’s website on Pole Mixing setups)!

Each clip toggles between LP and BP

Increase VCA Input: Reduce R33 and/or R32 (each 150k) – R33 is VCF-VCA, R32 is the resonance passband level loss compensation part. When going ca half of each set value of 150k, things start overdriving. One mod I like quite a bit is to set up a 1M pot as a variable resistor and wire that in series with a 10k resistor – both terminals of your little circuit then goes across R33. Folks also sometimes overdrive the qcomp path (R32) but that’s not so much my thing.

This is R33 reduced with a 1m pot and 47k resistor in series:

This is R32 reduced with a 1m pot and 47k resistor in series – overdrive only happens on resonance sound:

Increase Gate to VCA impact: reduce R171 (27K) to something like 15k. Check sweet spot with higher resonance setting in filter since at one point the VCA will overdrive.

VCA CV input: at the middle pin of the ENV/GT switch for the VCA you can feed in external CV – go via a silicon diode to clip possible negative voltage and use a 33k resistor (adjust latter for your needs).


Looping A-D Envelope: well, this is very easy and should be adaptable to other Roland envelopes. A simple 15k resistor across R151 (100k = MC202 R 182) makes the A and D stage loop as long as S is set to zero. This affects the comparator at IC16A (=MC202 IC25B) so that D fully discharges and resets the envelope to A rather than moving on to the sustain phase. Instead of a 15 K resistor you can also use a simple silicon diode facing upwards (cathode facing label E1) – and you have that on a switch, of course!

Sounds like this:


I’ll toy around with some LFO reset circuit later – was never satisfied with the one going around for the SH101 and want to try how the TR909 VCO reset circuits work in this setting here. In the meantime…

LFO Speed Scale: The min/max LFO speed is determined by C26 (0.1uf) in relation to R104 (150k). If you want to go classic, switch between capacitors of different values (say, 1uf or 47nf). If you like things a bit faster in general, don’t hesitate to reduce R104 – here you hear original vs. 33k resistor across R104….


In this clip, I feed the MSW back into itself but do so via the second VCA. That VCA is modulated by a simple envelope in sync with the CV gate.

Feedback/ext in mod: wire top terminal of R23 via a switch and an 43K resistor to the bottom terminal of C46 (or rather top terminal of R144). Now, use Mixer Fader “SYNTH” and fader “OUTPUT” for dialling in teh right amount of feedback. If you want to use this an an external input, use a signal via “INPUT1” instead of “SYNTH.” As easy as that…

With 43K I find we hit the right spot for feedback but this means sending in eurorack VCO levels will drive the filter *very* hard. This may be desired for that right kind of weird sound but generally, you want to pull levels a bit down – send ext VCO Mixer INPUT 1 and boom, you have level controll – per CV you say? oh….

In this clip I use an external LFO into the MSW (on bandpass filter mode) and simply modulate ext in level per LFO.

In this clip, I use bandpass filter and internal noise. Later I fade in external VCO and then CV over feedback…

Roland/Alfa Riddel reveal

You hear the AS3109 based machine 2x then the Roland IR3109 based machine 2x in each example. Slight envelope differences are on account of my AS-based machine having a modded envelope with smaller timing capacitor for faster zaps. Both machines have the OML clown OTAs for VCAs.