Oberheim Matrix 6 – SUNKEN SHIP rescue

I was approached by a client if i could fix Matrix 6 with missing transformer. He said he bought in on Ebay and some other tech he knows failed to get it running. I agreed to have a look and see. The closer I looked the more obvious it became the reason for missing transformer is that the instrument was in a flood or sunk with a ship.

There was massive corrosion everywhere, keybed pcb was really bad and you still could smell the mud combined with cigarette smoke.

I informed the client this is beyond economical repair, but I might adopt it as a pet projet. I was curious if I could get it going.

I started by attaching it to lab power supplies +-12 and +-5 and it looked like the CPU is going and that it has some chance. The snag came with powering up the display. Service manual says there is 9V AC, I connect it to 9V AC and nothing happened. I was kind of suspicious because these VFD displays usually need much higher voltage to operate but the publicly available service manual was so blurry i couldnt really properly see whats there. Luckily there came one friend to the rescue and scanned his paper copy of the manual in appropriate resolution.

I was surprised there really was clearly written 9V AC C.T. but following the schematic it was apparent you can’t make -42V DC from 9V AC C.T. There was clearly error and it meant to be 90V AC C.T. which is 45V effective and some 63V DC after rectification. This is then dropped by regulator to 42V DC stabilised. I only had 2×18 transformer at hand so I tried to connect 36VAC and suddenly it all started up but display was showing garbled data.

However when i booted up to service mode (press store button while powering up) synth sprung to life and I was able to play some sounds

Keybed was in really poor shape, about 30% of the keys wouldnt work. I wanted to test it over midi but I was unable to list through the menus, buttons were all rusted solid and I had to replace them all.

In the mentime I have ordered custom wound toroidal transformer and replacement 3D printed key.

Another issue was when i plugged in the headphones, synth would stop making sound including the main outputs. This was caused by totally corroded contacts in the headphones jack shorting against ground. I replaced all the jacks front and back with new ones becase others were in the similar shape.

When I had it wired up over midi, I focussed on why it wont boot on it’s own but only in the service mode and why even when i loaded the patches it would forget them right away. I suspected the battery but that sill had healthy 3.2V, ram was also good but at the end the culprit was eventually diagnosed to be faulty IC socket under the CPU, one of the legs wasnt making proper contact. I decided to swap the CPU and ROM sockets and viola problem solved.

Last but not least came the keybed, there was severe corrosion on so many traces it was frustrating but eventually one after one I managed to get it going reliably.

Now I had fully working keybed, new buttons and new jack sockets, new transformer, reliable memory keeping the sounds as it should.

There was last thing to deal with and that was the volume slider pot. The corrosion there was so extensive I havent seen anything like it before, there was no way to repair it and it had to be replaced. To my surprise such part was completely unobtainable. I ended up ordering a custom made batch of 100 and sold some of them to my synth tech colleagues throughout the world.

After calibration I noticed cutoff on one of the voices is way out and again this was caused by bad IC socket so I decided all the remaining sockets need to go despite having to desolder so many IC legs

Last thing I did is I installed Tauntek firmware made by amazing Bob Grieb and I gave the panel proper wipe with paint restorer and touched up the rust spots to stop it rusting further. When the owner seen it he couldnt believe it’s his machine, it looked really well.

Later on few more issues surfaced with the leaver pots being too crackly to function properly (again the flood damage) and the aftertouch sensor wasnt really working all that well, I couldnt revive it fully because part of the function requires the felt to be intact and apply pressure and that felt was completely ruined by water so I was only able to gain partial results.

My right hand fell off

There goes the saying you don’t know what you have until you lose it. One day my Tektronix Oscilloscope started to misbehave. It is one of the last of its kind. It is analogue at heart but computer controlled. Sort of hybrid between the old days and the days to come.

I haven’t really appreciated how good tool it still is despite it’s age up until it broke down. While the digital scope I have is super powerful and packed with intelligence this machine is just so much quicker to use and the reading gives you immediate idea about what is going on.

I was forced to take a break from all my day to day repairs and dwell into the world of unknown. I never worked on oscilloscope before and I never worked on anything with CRT screen where several kilovolts are present. While the service manual is exceptionally thorough in test procedures and calibration steps it doesn’t really tell you where to look when tests fail or something is not as expected.

I was very fortunate to find internet group of enthusiasts on groups.io who live and breathe vintage Tek scopes. The level of knowledge in that group is mind blowing. Some of the original design engineers are there and continue improving and supporting those ancient beauties.

There were two possible causes for my problem. One was leaked capacitors on the A5 computer logic board cause corrosion and key reference voltages end up all over the place. My oscilloscope had these leaked but the damage was not yet severe. It was caught the last minute I’d say.

I had to remove all the smd components in vicinity, thoroughly clean the PCB with demineralised water and activated flux in several rounds to ensure there is no residue of the electrolyte. Next was inspection under microscope, removal of the mask where traces already started to corrode, repair of the traces, new application of solder mask and finally put back the components.

It was solid 11 hours of surgical work but the outcome was good.

Since Tek used tantalum caps in later revisions which don’t tend to leak or go bad I replaced the original SMD aluminium electrolytics with tantalum caps.

When all this was completed I to my surprise found out the problem remains the same . The root cause was failed hybrid chip in charge of mixing signals from all 4 inputs as well as readout generator together before it’s sent to final vertical amp stage and CRT.

I’m very happy to have the Tek 2445B back in shape once more.

Story about Emu who wanted to be Theremin

It all begun as a straight forward job. When the EMU Emulator II HD plus arrived it was working and booting correctly and only one channel was not making any sound. Customers request was to replace all the buttons and sliders, install DREM Winchester HDD Emulator, fix the silent voice and put new display retrofit in there. I started with pots, sliders and display to have any further work eased by having reliable controls. All went well.

Next step was to find why the voice isn’t producing any sound. There was nothing on the output, I was releved there was also nothing at the input of the super rare switched capacitor filter chip as well as the rather expensive VCF chip. There was also nothing coming from the DAC and nothing coming in. I could see data at the input of the latch but there was no acknowledge signal triggering the latch. When I looked into how is the request generated I found one clock missing coming from Intel 8254 timer. Swapped the 8254s around and confirmed the culprit.

At that point all that was needed was to install the DREM and jobs done. I 3D printed the bracket, mounted it according to the instructions and I was hoping for it to work.

When the installation was done I was getting format error while I was able to run even HDD tests and they would return no errors, but as soon as I tried to access the data WRONG format message would appear.

cut the long story short there were two issues, I used 1:1 cable for the MFM data cable but it required the connector to be crimped the other way around. That took some head scratching to discover. However when I tried to access it I would still get wrong format message.

EMU II HD + has one of the earliest SCSI implementations from the time the standard was just being formed. Its not using full SCSI implementation to communicate with ADAPTEC ACB4000 Winchester HDD controller for its missing parity bit and ATN attention signal. Because of this it is not possible to just use SCSI to SD or SCSI HDD directly connected to the interface but the solution requires this ancient and rare Adaptec controller. With no documentation to hand and several custom Adaptec chips I was not giving it much hopes to be able to repair it but when I probed the HDD MFM balanced data lines with scope instead of the logic analyser I noticed one line is a bit dodgy. This was driven by chips very known to me from my endeavors with SSL 5000 series mixing console the RS422 drivers by Motorola. I replaced this and finally the whole thing was working and I took EMU to friend who has all the sound libraries so we could load this onto DREM SD card.

There I noticed the new display customer provided for installation suddenly lost few lines but it only did it after solid warm up. I ended up fatching different display but it wouldn’t work at all in this installation. I asked about this on Facebook group specialized on music gear display retrofits and luckily enough Ray Bellis, bless his cotton socks, chimed in with an idea to swap one of the IC’s in the display driver circuit from HCT version to LS version. Suddenly new display was working correctly. This has something to do with display pulse timing alteration. I do not fully understand why such a small speed variation can make it or break it but I’m happy it works.

Now with everything working it was finally time to close the lid, tighten the screws and give a good news to the owner.

Here came the trick, it would boot and work reliably with the lid open but it wouldn’t boot with the lid closed and it would even stop working whenever I tried to close the lid down. I tried all the possible things and connections which could be moving during the process but I could move press bend the PCB and all the connectors and cables as much as I wanted but it would still play reliably. When the lid was closed things would stop and the machine would freeze. I couldn’t understand what’s going on, why would just a proximity of the lid cause the CPU to crash.

Several hours of head scratching and brainstorming later I checked clock signal for the Scanner z80 CPU and I didn’t like what I’m seeing

Voltage was correct, frequency was correct but there was this ghosting as if it was jittering like mad. Then I noticed I could modulate the signal with getting my hand closer and further and that was finally something to focus on and track down. It was digital Theremin, I could make it dance like mad and I could even make the clock disappear just by waving my hand above the circuit boards in specific manner. This particular behavior was then tracked down to bad contact in one of the IC sockets in charge of the clock. Finally I was relieved I can finally close the machine down.

Beware Ostriches, EMU’s can be very mischievous and this one didn’t want to leave me. I closed the lid, turned it on and nothing!!!!

Display was just showing full block of squares and the boot sequence would hang after 3 clicks through the hard drive boot. I suspected while I fixed the one socket problem other may have happened elsewhere. After consulting the service manual I knew the problem is the Scanner CPU related so I checked all the data lines, address lines, clock, reset as well as all the other pins and all was behaving normally. There was nothing to suggest something is wrong, after a bit more sniffing around the components associated with scanner CPU boot process I noticed the CTC Z8430 is suppose to generate 3 timers on the output pins which are used as interrupts. I could only see one signal and others were all doing nothing. I once again checked this time on the CTC whether it has all the data lines all the address lines chip enable and all the other signals it needs to do the job and to my surprise all were OK. At this point it was not funny anymore, I tested different CTC but it was doing the same, all the input signals were correct but there was wrong output! The only explanation could be that there is something in the boot sequence before the CTC and its failure makes the CPU not to initialize this peripheral. The only way to find out was to use the analyzer and de-compile the Z80 instructions.

The boot sequence was quite long and because I only have 16 channel logic analyzer I could either decode instructions or addresses, I didn’t have enough channels to do full disassembly. After digging through about 500MB of data and going through the boot sequence step by step I finally found the problem. Somehow between the last working test and the lid closure contents of 2 bits within the Scanner EEPROM got corrupted and returned to FF state. How utterly unlikely! I burnt new Scanner EEPROM v 3.1 and everything is working reliably since.

Well fought mighty EMU, but you were defeated after all!


it’s been very busy year for me especially since I became father in March. It is one of the reasons who so few of the repairs made it to this blog. I had quite a impressive selection of instruments on my table.

Oberheim OB8 which initially looked like it’s all working and is in good shape but it ended up being very time consuming excercise to get it to 100% reliable condition. All keys needed new bushes and J wires, nearly all trim pots needed to be replaced and I’ve learned one lesson there. I replaced the filter tracking trimpots with temperature stable BOURNS multi turn trimpots. They are quite expensive at about EUR 6 each and the outcome was actually less stable filter tuning then before. Changing this for regular cheaper multiturns with higher temperature coefficient made things better. Lesson learned was that sometimes temperature drifts compensate each other partially and therefore better components do not always bring better results. Another lesson was a weird behaviour of the demux circuits. Filter wasn’t tracking well. There was always some offset. It all went away after factory reset.

Shortly after OB8 came another Oberheim this time OBXa. While it was expected to be in rougher shape then the OB8 it ended up being better off. Most of the upgrades were already there such as higher spec rectifier bridge in the PSU and multi turn trimmers at critical positions. It was actually pretty smooth ride, there was one faulty TL084 so I had to recalibrate the offsets but rest was just working great.

Prophet 5, very little needed to be done to this one, mostly cleaning job and calibration.

And in the row of Oberheims there came two Xpanders and two Matrix 6Rs and one Matrix 6.

Each had a different fault. One had a dead CEM3372 what I’ve learned is the suffix does matter. This Xpander had CEM3372B and when I tried to use CEM3372D there, this voice won’t pass the autotune procedure. Basically if I would replace all 6 with D all will be fine but different suffixes cannot always be mixed and matched as you like. It is safe to say it’s recommended to have them all of the same suffix.

Second Xpander had a problem with demux chip in charge of panorama so some of the voices couldn’t be reassigned.

There was a funny error with the Xpander rotary encoders. I suspected dirty or worn out encoders itself I suspected the debounce circuit isn’t right for the replacement encoder but all suddenly was fixed with replacement of Schmidt trigger IC. I later verified it would work even with original encoders.

Matrix 6 all 3 had a same problem which was just contact on the CEM chip socket and needed general cleaning.

Minimoog original D needed mainly thorough chemical cleaning of keyboard contacts to get tracking back in shape.

I thought after dealing with Farfisa Soundmaker as described in earlier post that next time I encounter that instrument it will be easier because I already had quite decent understanding of how the circuitry works for most part. What I did not expect was how big of a nightmare it can still be. I have spent over 150 hours of troubleshooting to restore this one. There was no single difficult fault but neverending row of problems throughout. I started on the main board for Mono /poly generation where there were the usual suspects 74C85 comparators faulty. There were two faulty 3080s in the VCA VCF mono section, about 6 bad small signal transistors, poly section had multiple shored tantalum caps in the sustain modules. Keybed had a funny fault where one of the contact springs was pulled and sticking out a bit. When the corresponsing key was pressed the sticky out wire shorted against chassis. Last but not least was a impossibility to create decent sounding flute. Some of the transistors in the DA converter on the main sound generation board had a very low beta and things improved dramatically when they were replaced with modern BC549/559 equivalents with plenty current gain. I had to make new circuit board for PSU because it got shattered during transport so if you need one give me a shout. Its really cool to have a proper FR4 material.

There was too many other instruments to count them all. I’ll add some mention of them if I can think of something with useful lessons learned.

Kawai K5m display replacement

Its a rather common issue with most of the vintage gear that all of the backlit LCDs fade with time or the backlight dies altogether. Same fate met this Kawai unit, the step up converter was just fine but the backlight foil just went past it’s useful life.

One way to deal with the problem is to replace the foil with a new one, the other is to use LED backlit display.

I have in this case opted for white on black LED backlit LCD. The exact part number is LCD-AG-240064A-FHW K/W-E6

It wasn’t without issues. First of all the cable is connected upside down. So you’ll have to put nr1 wire on nr20 position. Secondly the display quality is so so. Far from perfect but it’s fair enough.

RME Fireface 400

This sound card came in for repair with issue reported as not being visible to the computer and oddly freezing up. Customer also admitted he was “poking around with a voltmeter and may have shorted something”

Initial evaluation of the PSU section discovered few things. He reported initially it was working fine and lately it was only working when powered from the FireWire interface. External PSU wouldn’t work.

The way the PSU circuit works is there is a LM2596 5 switch mode converter which makes 5 V out of the external PSU 9-21V. At this stage selector switch decides if this power is used or the 5V from the FireWire interface. This 2596 was running on very odd frequency way lower hen the specified 150kHz. Also the associated filter cap was totally gone. ESR of 100Ohm and the same for the other cap after the second coil. This 5V Is then converted to +-12V by yet another switch mode regulator LM2588.

This one had better caps used (Nichicon specified for high ripple application) and these were all ok.

There are few more regulators one tiny switch mode making 48V phantom and two more linear regulators making 3.3 and 5V for the ADC and DAC chips.

Once the PSU was all sorted with new LM2596 and new low ESR high ripple caps, all voltages were stable and without ripple. Following this repair sound card was stable without any freezing up.

The FireWire HOST issue was caused by ESD which zapped the ESD protection on first port as well as it took down the FireWire controller IC.

I am really not well equipped for SMT. All I have is $30 hot air station, simple basic soldering iron, microscope, flux and solder wick.

It took fair amount of heat to get the controller chip off. It is a 4 layer PCB with extensive cooling ground planes. I slowly warmed the whole PCB to try to lower the heat expansion stress on the board and then focused the heat on the controller chip. I usually set the hot air station to 335 Celsius for 40/60 leaded solder or to 350 for unleadded. This time however I had to raise the temperature to 370 to be able to melt it and remove the controller chip. I was worried it would damage the PCB but it lasted well without any damage.

Main culprit

Soldering the new one back in was even harder. Its on the edge of my hand soldering skills.

Professional equipped for SMT would probably diss this but for me I’m happy with the outcome. Its soldered on, there’s no damage to the traces and the card worked ISH.

When I swapped the controller I noticed only one FireWire port works. Further evaluation and measurements narrowed this down to the ESD protection device. When I removed it for the purpose of testing both ports worked just fine. I ordered new part and it’s going back where it belongs. ESD is a serious problem for these. This time the protection wasn’t enough and it took down the controller as well.

Dr Böhm – Digital Drums dual restoration

It does not happen very often for two rare vintage drum machines to turn up at the same time and both suffered exactly the same fate. Internal NiCd backup battery cell leaked out its potassium hydroxide contents and eaten away several of the traces.

One case was rather mild and only suffered localized corrosion. I had to replace about 5 smaller IC sockets, thoroughly clean it, neutralize with Vinegar and the only reconstruct few traces and mainly the vias connecting top and bottom side of the PCB. The other instrument was in much worse situation, it was repaired at some point in time but because the leak wasn’t neutralized but probably just wiped away it spread the corrosion across entire length of the PCB. I was quite desperate because this would require massive surgery, I would have to re route dozens of tracks and i was worried even that wouldn’t last for long given how severe the corrosion was.

Lucky enough my friend Miro Ruml offered he would clone the circuit board, not the task for the faint hearted but to quote his words “I prefer it to crosswords or Sudoku! “

The first prototype.

After some 50 hours of design and fault finding tedious work the unit sprung to life. There were only 8 routing errors made but I had the new version 2 of the PCB made. (if you’d like to buy one, please contact me, price is 70EUR per piece, still beats the time one would spend reconstructing heavily corroded PCB).

Its quite rare to see the version with accompaniment fitted and in this case both of the units had it. It is a whole separate CPU with its own ROM and RAMS, in a fact the accompaniment is much more complex circuitry then the drum machine itself. Those two computers are linked via serial interface and allow for quite interesting features.

To the dustbin and back again, the story of Roland TR-707

Original state as found

I was contacted by the customer that he found 707 by the bins. He sent me picture and I knew Id need rubber gloves to work on this one but gladly accepted the job. I was quite surprised apart of the missing battery tray lid everything was there and undamaged.

following a proper bath in soapy water the case looks simply brand new. Awesome condition

When I had the unit clean enough to touch it thorough inspection revealed there was a crack in the PCB where the power connector resides. It is the most frequent place to get fractures as the unit falls on the back while the connector is plugged in.

cracked PCB
tidy up
reinforcement added
final repair before the crack itself was reinforced with glue.

Following the broken trace repair unit sprung to life and booted alright. Unit was however acting very oddly and most of the sounds were crackly and distorted. It was caused by open circuit or intermittent circuit on the slider pots. I took all of them, apart and carefully cleaned and put them back together. This brought all but two of them back to life. Final one to tackle was rimshot/cowbell. Carbon track looked ok but it behaved as if the leg of the slider pot came detached from the carbon track underneath the plastic. I had to carefully remove the plastic and then i applied conductive paint. It restored most of the connection but there was another part of the track I had to fix with carbon paint.

Sadly it is not possible to source replacement sliders for those anymore. It would have been much easier just to fit a new set of sliders and be done with it.


DPA4099 suddenly died within recording session. Input dual transistor labeled 3Ft was the culprit. Had to order it from Mouser. There are two types used throughout 1Ft and 3Ft, they are BC847/857 respectively.

I wouldn’t normally post about such a small repair but it was my first SMD repair. I recently acquired microscope and hot air station.


I need bit more practice but I hope it’s acceptable outcome. It works and that’s important.


For those new to SMD I learned few things.

  1. use loads of flux, more is better then not enough
  2. add leaded solder before removing the lead free, it makes things easier
  3. solder wick is your friend
  4. 340 degrees is enough
  5. kapton tape helps protect surrounding devices from displacing
  6. excess solder can easily be removed afterwards as long as there’s a plenty of flux



My customer bought them while he lived in Ireland and upon return to Prague, Focal refused to repair them. I never seen such a poor customer service to refuse post warranty service.

I investigated and the problem was with the soft start circuit for the internal 800VA toroid transformer. It consists of triac and rectifier bridge. Both the triac and bridge were shot so I replaced them with beefier spec units and added additional heat sink on both.  

Audio Repairs Synthesizer Repairs Synth Restoration