MC-4 power switch horror

Alright, horror is a bit dramatic. Bear in mind I love my MC-4, maybe not quite to the extent that I want to be buried with it when I die, but you know.

The softly glowing power switch had gone wonky, with one side slightly pushed in. I didn’t really think anything of this until the other day when I went to switch it off and it fell in completely, and then started flipping off and on as if it was panicking…

…and then I started to panic, and lunged for the mains switch.

Then the switch fell out completely – here’s the button with the two broken bits.

Broken MC-4 power switch, showing the pivot points having snapped off

Clearly one of the plastic (not delicious) orangey bits had fallen off at some stage, so it was just a matter of time until the other one broke. The bobbley bits are the parts that the switch face pivots on, they clip into holes in the on the inside of the body of the switch:

Bosted power switch

…which is looking pretty manky after forty years or so.

I glued the pivot points back onto the switch in the hope of getting it working again, but I couldn’t balance the contacts and springs and squeeze the switch face into the body it without hearing it all coming adrift.

I didn’t fancy taking a guess at how it should go back together when 230V would be bouncing around inside, so I started looking for a new switch, more out of hope – and found this from RS for a panel cut-out of 30x22mm:

Brand new replacement power switch - shame it's not amber-colour, but can't have everything

Not the right colour, but maybe the right size?

Time to pull the old switch body out, and it was a bit of a fight. I took a photo of the markings on the side of the original in case anyone comes across a warehouse full of the things – marked T100 16A 2WI XII, and also T 65 C UND.LAB JNC LIST.

MC-4 original power switch markings

I made a note of all the connections before desoldering:

Making a note of the connections on the old bosted switch

After carefully heatshrinking the connections, and doing a bit of testing with a multimeter, it fitted in place perfectly:

Replacement switch in place - fits perfectly

Flip the switch, nothing goes bang, it glows…

New power switch glowing on in the murkiness

and success. (Still a shame about the cracked screen, nothing to be done about that really).

Switched-on and now working ok, phew

I’ll keep hold of the original amber switch in case I manage to finagle it back together.

While we’re here, let’s have a look inside. First time I’ve actually had it open, most things usually succumb to the screwdriver in the first week or so – here’s the transformer/noise filter/power supply board.

Transformer, noise filter, and power supply board

Here’s the main board, the Sharp version of the Z-80A in the middle-top, 16K of Mitsubishi-flavour memory on the left and 10K of EEPROMs including the system program on the right.

Roland MC-4 main board, including Z-80A CPU, memory, EEPROMs, and supporting circuitry

I went off into a little dream for a bit thinking, ooh actually, it seems like there’s actually 6K of memory address space free for extra code – maybe we could add some extra functionality.

Even if that’s technically possible, I realised it wouldn’t be much fun actually finding somewhere to put the extra ROMs and wiring up the additional select lines from the ROM address decoder at IC60.

And then desoldering the EEPROMs, installing sockets, reading the EEPROMs, disassembling the code, re-learning how to write Z-80 assembly (not that I was ever very good at it) and then about a billion other things.

It would be easier just to recreate it, like I’ve been talking about for the last ten years or so. Maybe this scare will spur me on.

Here’s the extra 32K of memory on the expansion board, with the connector on the right leading from the main board, and the connector on top leading to the digital cassette interface at the back.

MC-4 RAM board, including 32K of extra RAM and the digital cassette interface circuitry

This MC-4 has a serial number that works out to being built (according to the handy Roland/Boss serial number decoder site) in September 1981, and the date codes on the ICs on the main board all line up with that, but expansion board ICs are dated 1982/83, which is weird.

I would have wondered if mine was an MC-4A (with 16K of memory) later expanded to an MC-4B (with 48K and the digital cassette port) but for the fact that the serial number plate clearly says it’s a B. Maybe the extra RAM board chips were failing and another one was subbed in.

Here’s the back of the front with the lovely period wiggly lines.

Wiggly lines: MC-4 front panel circuit board from the back

Now it’s all back together and working happily.

So anyway, if you’ve got an MC-4 still in use, go easy on the power button.

2 Comments

fixing the armstrong 626

I’d been using my dad’s old hi-fi tuner-amp without any problems for nearly thirty years, but recently it developed a super-loud low-pitched hum on the outputs, of the type that makes you go – shit, turn it off quick.

It’s been knocked about a bit over time so no photos of the outside, just the inside – if it was in decent nick it’d look like the example here. Here’s the inside from the top:

Inside the Armstrong 626

Lots of brown resistors! Barely any ICs! Some weird retro-futuristic green transistors! Complicated wiring!

Once I’d worked up the confidence to turn it all back on again, I pulled the output from my Soundcraft mixer all the way down: still humming. Turned the volume down on the Armstrong: slightly quieter, but still loud.

Straight-away I was thinking it would be due to a capacitor somewhere, but it’s easy to jump to conclusions, so thought I’d try and take measurements here and there.

The service document/schematic could be a lot clearer for a basic idiot such as me, but I managed to find the stereo amplifier inputs to check with my multimeter, and measured a load of AC there (with the output level from the external mixer set to 0).

The main internal rail voltage on the Armstrong is a chunky 82V, and I measured some ripple on that. Here’s the actual power supply board – the black slightly pointy blobs on the outside of the pair of green blobs are the diodes in the rectifier:

Armstrong 626 power supply circuit board

Actually it would have been a lot easier just to measure the voltage on the main smoothing capacitor; speaking of which, what’s this here?

Ancient 3300uF 100V DC capacitor in-place in Armstrong 626, leaking crystallised electrolytic

That brown-ish crystalline blob that is growing out of the top of the capacitor just next to the blue wire looks very suspicious. OK, given that we’re seeing ripple on what should be a nice flat DC rail and the capacitor looks dicey, let’s just replace it.

Here’s the old one versus the new one – it’s definitely not as pretty a colour:

The massive old 1974-era 3300uF capacitor vs a much smaller 2022-era device of the same rating

…but that sorted it.

Not the most exciting, but I was glad that this one was a boring easy fix.

The two fat 4000uF output caps are obviously the same vintage and will also need replacing. The cap I changed was apparently the first component change it’s needed in nearly fifty years, which is good going.

Really I should celebrate by listening to Radio 3 FM stereo broadcasts, which is what I think it was mostly made for, but recently it’s just been pure Intergalactic FM disco fetish all the way. Poor old Armstrong.

No Comments

rotary 100M rampage

Here’s my second attempt at the slightly-boring-but-I-like-it-anyway 100M. The last (mostly stripboard!) one I made is somewhat delicate but works really well, and is permanently attached to the MC-4, but I wanted to do some of the other modules outside of the D-set.

And I wanted to do a closer clone of the style of the 100M, just not with bastard slide pots. And also to do all that wiggly track bullshit again, as if it makes any difference. Here’s a couple of 182s (one missing the step number switch) and a 110, the panel is just paper and card while I test for now:

System 100M 182 sequencer and 110 voice clones

Just poking out of the right of the middle 182 is my rough attempt at replacing the 2SA798 dual transistor with a BCM857, which works better here. Hopefully I’ll just about get away with the height. I should’ve really just put a footprint in for the surface mount transistor, I can’t see there’s much point in trying to get hold of the original.

Replacing the 2SA798 dual transistor with a surface-mount BCM857 on my 182 sequencer clone

The 182s are really basic (wot no reset, etc) but have actually proved to be loads of fun, which I wasn’t really expecting. I was never that bothered by the 104 sequencer, and the SQ-1 I’ve got somewhere hasn’t seen much use either. Maybe it’s the portamento, it makes everything a bit techno.

On the 110 I had a boring problem with the green LED not coming on in the VCA section, with the red LED just barely flickering, which turned out to be a duff transistor at Q18.

Should be polystyrene caps in the filter here, I know, the other 110 I’m building up will have those, just wanted to compare them. For some reason I’ve a feeling the film caps will sound a bit nicer.

110 voice board detail

Maybe too much authenticity here; the original has the diodes and resistors limiting the filter resonance tacked on the back of the pot board. The service manual refers to cutting tracks to include these, which rather suggests there wasn’t any technical reason on the original beyond “whups, left those out”. Cover your eyes, this is a crime.

board hack to include resonance limiting diodes on the 110 VCF

I’d intended to include footprints for these on the board, I just…also forgot.

For the 140 envelope, finding the right sort of 2P3T switch for the trigger source selector was proving to be a pain in the arse so I ended up using my usual DPDT 3 position ON-ON-ON and having that switch a 4066. It will eventually get some push button manual gate switches, I didn’t have any at the time.

Also I added a couple of opamps and some LEDs to give a visual indication as to the state of the envelope, ‘cos it seemed wrong that Roland didn’t have those on the original.

The extra components made routing the control panel so that all the connectors are over on one side a bit more tricky.

140 dual envelope/LFO clone

The envelope and LFO seem to work without any fuss apart from the second envelope being pinned high until I replaced the buffer opamp on the output, must’ve been a dead one.

Brilliantly I got the orientation of the Alpha rotary switch completely wrong on all of these boards, so a D-type knob will point off at a funny angle. There’s probably something horrific that I can do with a hacksaw to bodge it, I don’t really want to do another run just for that.

Another shot of the 140 dual envelope/LFO clone

There’s a couple of my versions of the OpenMusicLabs 662 clone/clown on the LFO, one for voltage control over the LFO rate and another for the LFO delay.

I had a load of dead ones after that Jupiter-4 attempt I made, so I’ve worked out a brutal way of reviving them, which essentially consists of connecting the dead 662 to power, then seeing which transistor starts to smoke, and then replacing that one and testing again, until it no longer makes the power supply panic and it works in my VCA test board.

140 dual envelope and 130 dual VCA clones from the front

All the main boards are as close as I can get them to the originals without driving myself round the bend, with the only major changes being the power/internal routings connector and more regular spacing of the connections at the front of the main board PCBs. Wonky tracks!

I tried to keep to the original track layout for the boards; this shows the wonky tracks on the 130 VCA and 140 envelope/LFO

The control boards are completely different because of the rotary pots and my desperate need to avoid manually wiring everything up as I did on the System 100.

Currently the control panels are just attached to the main board through the friction in the connectors and good intentions, so I added some holes to allow for some for some sort of mini-bracket to fix the two boards together. But then again, it might actually be easier just to use a zip-tie. Which feels wrong, but would mean I can avoid doing any potentially self-maiming metalwork.

Here’s the 172 phase shifter/audio delay/gate delay, and the 132 mixer/voltage processor:

172 phase shifter/audio delay/gate delay and 132 CV/audio mixer and voltage processor from the front

MN3004s are hard to find as NOS and no longer made, so I’ve gone with an MN3007 instead with the intention of changing the timing capacitor to overclock it to get closer to the original specs, but I quite like it as it is. I struggled to measure it at the shorter end but the delay time pot looks to give a range of about 1.7ms to 66ms, with clock noise creeping in at longer delay times. The original 172 delay is specified as between 0.3ms and 7ms.

Because the old Toshiba 4013s work differently to modern versions, I again used Fitzgreyve’s tactic of using a 4049 to get a modern 4013 to oscillate and clock the 3007 correctly (nuff respect, hat tip).

The 132 is super basic but I still managed to mess something up on the control panel; I swapped the first and third pins, which are the input and output pins for the first mixer. Another thing to bodge.

I spent a bit of time trying to work out why the red LED wasn’t lighting with input, but as should’ve been obvious, this is just to show an overload condition, which is the standard in the rest of the system.

Here’s a few of them lined up for a test in half-darkness, mostly trying to show off using three 182 sequencers at once, and struggling a bit with the tuning (but that’s techno, right?)

The 182 on the far right is connected to the 140 and the 110 and occasionally going through the audio delay on the 172, while the other two 182s are playing my original stripboard 100M and the barely-working remains of my dodgy clone System 700. I’m also putting the 606 through the phase shifter here and there, took me ages to match those transistors.

I’ve still got some more to do, such as the control boards for the more recent modules (165/173/174), as well as both boards for the 131 output mixer, and the control board for the 121 VCF with its annoying 4 position switch for the high-pass filter.

Seeing as though Roland never did a multimode filter for the 100M, and Alfa have got their IR3109 in production, I’m tempted to do a dual version of the Jupiter 6 VCF in this format.

And they all need actual metal panels as well, the annoying bit. I quite like them in white, but it’d be good to get them done in something like the original grey.

9 Comments