Yamaha QX3

I recently won an eBay auction for a "spares or repair" QX3 sequencer. It looked bad from the photos, but was very cheap. It turned out to be worse than expected. The case was held together with gaffer tape as and the PCBs were loose inside as almost all of the plastic screw posts had crumbled to nothing. It was also absolutely filthy, but a quick test showed it at least booted.

These used to be screw posts

The keyboard. Filthy and missing a few keycaps

Cherry key switches - and 30 years of crud!

I stripped the case of all electronics and gave it and the keycaps a bath and cleaned up the key frame and PCBs. Copious amounts of epoxy were used to attach new plastic standoffs to replace the broken screw posts. I managed to break one off on reassembally, but the rest seem solidly attached.

New screw posts glued in place

The floppy drive was full of dust, so this was cleaned up and reassembled, before being reinstalled.

Electronics cleaned up and reinstalled

Further testing has found that the sequencer basically works, including the floppy drive, but the encoder is jumpy and only works in one direction and a few of the keys are intermittent. I have removed, opened and cleaned the encoder and the worst of the key switches, but will probably end up doing every switch.

Still to do:

• Reinstall the encoder and check that it now works.
• Check the rest of the key switches and clean as required - or just do them all.
• Replace tact switches - these are the same as the ones in the Juno 106 and I still have about 70.
• Replace the mains flex.
• Replace missing key caps - I'm unlikely to find the correct ones.

Juno 106 tact switches

I've just spent the last hour replacing all of the tact switches on the Juno 106 panel board. I also took the opportunity to remove the remains of the foam dust guards that had become so brittle that they were themselves the major source of dust, and give the front panel a good clean - much easier without the sliders getting in the way.

The remains of the dust guards. They literally fell apart when I touched them.

Dust guards removed, board cleaned up and new tact switches installed.

All of the sliders were still functioning perfectly, so I left them alone. I'm currently waiting for the switch caps to dry after giving them a wash, and will reassemble the synth this evening.

MKS-50 finally repaired

Where did I leave off? Having replaced the BBD drivers and the 4051 that de-multiplexes the control voltages for chorus rate and volume, the noise remained. Replacing the underrated and likely half dead bypass capacitors improved the noise that the BBDs were injecting into the rails, but still didn't solve the problem.

I was on the verge of putting the MKS-50 back on the broken pile when I nudged something near the 4051 and the noise stopped! Initially I expected to find a broken trace or cold solder joint. When I first acquired it, the synth did not respond to MIDI messages due to a number of broken traces. Thanks to all but one of the mounting screws being missing, the PCB had flexed and cracked. The underside is a mess of bodge wires as a result. I had also found and reworked a few suspect solder joints.

Poking around with a meter revealed nothing, but a closer look at the component side found this.

Look closely at the capacitor array. Something isn't quite right.

The corner isn't supposed to be chauffeured and the common pin is missing!

Could this be the cause of the noise problem? The capacitor array is used to sample and hold the control voltages coming out of the 4051. These control the overall volume VCA as well as the VCAs and VCFs for each voice and the chorus rate. Without the common ground pin I'm amazed there weren't more problems. Rather than providing sample/hold the damaged array would capacitively couple all of these control voltages, meaning the chorus CV was being modulated by and indeed modulating filter cutoff, etc.

As the synth is already quite a mess (cracked PCB, bodge wires, missing case lid...) I made a repair horrible hack.

Awful hack, but it works!

As a point of interest, the construction of the MKS-50 is awful. A single layer board with a lot of jumpers, ICs seated at odd angles and some very poor soldering. The case design also makes it difficult to remove the PCB without flexing it. Roland gear tends to be fairly well made, so I'm not sure what happened here.

The MKS-50 lives and sounds great! Next up will be another horrible hack involving a sheet of aluminium and some extrusions to fashion a top for the case.

MKS-50 and Juno 106 update

Just a quick post. Having now restored the Juno 106 to working order, I've noticed several of the front panel tactile switches are rather intermittent. Unlike those on the MKS-30, these are of a standard and widely available design, so I'll replace them all.

The MKS-50 continues to annoy. Swapping the MN3101s had no effect and neither did changing the 4051 or the TL072 that feed them. Further probing reveals the power supply rails in the chorus area are significantly noisier than the rest of the synth. A close look at the board reveals that most of the bypass caps on the + and - 15V rails have the rubber bungs in their bases displaced. They are only rated for 16V, which really can't be a good idea on a 15V rail. In fact the whole construction of the synth is pretty poor. Roland really seem to have been penny pinching by 1987.

Juno 106 Resurrection

In my previous post, I mentioned that the 80017A VCA/VCF ICs in the MKS-30 will die at some point. I am expecting this because about two years ago, two of the six in my Juno 106 failed in the space of a few weeks. The 80017A is not actually an IC, it is in fact a "hybrid module" - a collection of SMD ICs and capacitors and printed resistors on a small ceramic PCB, which is then coated in epoxy. It is also notoriously unreliable. The theory goes that either some reaction occurs in the epoxy, causing it to become slightly conductive, or moisture becomes trapped inside it. One solution that many people have had success with is to remove the coating with a solvent. This seems to restore some modules to full health, but not all. Some also require rework due to poor soldering.

After my Juno expired, I removed all six 80017As, intending to give them a bath in Acetone and soldered sockets in their place to make future replacement easier. One of them had already been replaced by a previous owner and the solder pads had suffered. Then I got a Jupiter 6 and basket case Emulator II and fixing the Juno slipped down my list of projects.

Original 80017As. Two are dead and the others are on borrowed time.

On to late 2014 and having spent the last few months on a big project (to be revealed as soon as I get around to writing a post) and repairing the Emulator II (well almost, there are still some issues), I finally decided to tackle the Juno. I couldn't be bothered spending several days stripping the 80017As with no guarantee of success, so I opted for the alternative - the D80017A clones from Analogue Rennascince.

Bring on the clones!

Lots of passives.

One of the many calibration procedures.

Clones installed. Probe on one of 12 test points.

There isn't much more to write really. The clones arrived on a scored PCB panel and installation was simply a case of separating them and plugging them in to the sockets I installed earlier. An hour with a scope had everything calibrated according to Roland's original service manual and I finally have a working 106 again. Many people say the Junos are overrated as they are fairly limited, simple synths. This is true, but they do sound good.

Roland MKS-30 modification

I intend on getting a Kiwitechnics patch editor at some point, but the latest revision of it no longer supports emulation of the Roland PG-200 programmer, so cannot control a stock MKS-30. I believe the MKS-30 was the first MIDI module produced and it's MIDI implementation is severely limited - basically note on/off, no sysex or CC. The only way to backup patches is with an expensive and rare memory cartridge. Editing via the front panel is a painful process and the original PG-200 programmers go for as much as the synth itself, but thankfully Johannes Hausensteiner came up with a rewritten firmware and minor hardware modification that allows saving and loading patches over MIDI as well as realtime control. There is also a mod to control the speed of the chorus at the expense of being able to adjust the high pass filter, but I decided to forgo this.

Lid removed to expose the guts of the MKS-30. The mod comprises a replacement EPROM and a single wire from pin 12 of IC 40 on the main board to R12 on the chorus/PSU board.

Closeup of the main board changes. Rather than solder directly to the IC pin, I used a Molex connector so that the boards can be separated easily when I inevitably have to replace the VCF/VCA ICs. There are two conveniently positioned vias here with the right pitch for a 2.54mm header.

The other end of the wire. R12 is disconnected from the MIDI in circuitry and the free end connected to the wire, gunked down with silicon to stop it wandering. This turns the MIDI thru port into a MIDI out.

The infamous Roland 80017A filter/amp hybrids. Two of them have failed in my Juno 106 and it is only a matter of time before these fail too.

Replacement battery installed last week.

In other news, the intermittent front panel button (MIDI channel) is now behaving itself after some massaging. I haven't taken the panel board out to have a look, but I now believe these to be the same as the tact switches in the Jupiter 6, i.e. totally unobtainable.

September 2014 update

I haven't posted anything in nearly two years, so I guess it's time for an update. On the vintage front, I've done a few more repairs and will endeavour to post details at some point. The current state of my synth collection is:

Roland Jupiter 6: All good. Previous repairs: Replace volume pot, resolder some front panel tact switches, clean all pots and sliders.

Roland Juno 106: Still dead. I haven't got around to removing the coating on the 8017As and will probably end up buying clones instead. Update: Clones purchased and installed 28/9/14. Now back in the living, but tact switches intermittent. Replacements ordered 01/10/14. Tact switches now replaced. All good.

Roland MKS 50: Works, but chorus clock bleeds into the output even when "off". Likely to be failing MN3101 BBD drivers. Still missing the top half of it's case - I doubt I'll ever find one. Update 29/9/14: Changing the 3101s had no effect. Probing with a scope reveals the BBD clock CV rises and falls in time with the noise. I now suspect a 4051 CMOS demux IC. 02/10/14: It wasn't the 4051, or associated TL072. Power rails noisy in the chorus area. 09/10/14: Recapping reduced the noise on the rails, but didn't solve the problem. Eventually found the sample/hold capacitor array CA1 to be cracked and missing it's common pin. Now repaired.

Roland MKS 30: A few temperamental tact switches, but otherwise OK. The 80017As haven't expired (yet). Update: Replacement EPROM and MIDI out mod installed 23/9/14. This makes the synth much more usable.

Roland U 220: No problems apart from a missing volume knob. Previously replaced a damaged mains lead.

Roland TR707: Perfect. Previously replaced power jack.

Oberheim DPX1: Perfect.

Yamaha TX81Z: No problems.

Yamaha DX7: No problems. Previously replaced damaged mains lead and  several broken keys.

Roland MC500: Working fine.

Roland SBX-80: Recently replaced tempo pot. Now good.

Roland SH1000: Working to the extent that it produces sound. All pots scratchy, keys intermittent and front panel severely rusted. No work done so far.

E-MU Emulator II: Finally working for the most part. The non-booting problem was cured by replacing a few ICs in the clock divider. Two RAM chips were also faulty, leading to corrupt samples, and two voices were silent. One was down to a faulty CMOS IC, but I have yet to diagnose the other. I have also cleaned all the switches and pots, replaced one floppy drive with an HxC floppy emulator and installed a new LCD backlight.

Yamaha QX3: New acquisition November 2014. Intermittent switches and encoder. Badly mangled case. Very dirty. Cleaned up and reassembled. Still much to do, but it mainly works.

I've spent the last few days replacing the batteries in most of the rack mounts. None of them were dead, but the CR2032 in the MKS50 had started to leak.