SCI Pro One – another approach to MIDI

Back in 2007 I developed a replacement PCB for the Pro One’s CPU, not knowing of parallel work on the other side of the ocean.
I planned it to be mounted below the main PCB, just where the old CPU was, but on the other side, but accidently made the PCB too big so that I had to use flat cable to install it.

Some five years later I got another Pro One in need for a new CPU but also for MIDI, so I re-designed the PCB, still with a DIP CPU so that it still needs to be mounted below the mainboard, but now with the correct dimensions.
Keeping in mind that there a much smaller plug-in replacements on the market, I left it this way to allow for DIY building of those daring to solder SMD resistors, but not fine pitch processors.

Here’s the result:

Except from emulating all of the original Pro One’s functions, it can…

Normal and Retrig Mode:

  • use keyboard and MIDI IN in parallel
  • send every key press via MIDI OUT (also in Normal mode, where only lower notes will sound due to low note priority)

Arpeggiator Mode (which has priority over Sequencer Mode in this version, by the way):

  • use keyboard and MIDI IN in parallel during normal and latch mode
  • send the note on/off’s via MIDI OUT as they are played by the arpeggiator, including additionaly hold notes in latched mode

Sequencer Mode:

  • use keyboard or MIDI IN for sequence entry while recording, transpose with keyboard or MIDI IN while playing
  • sends notes via MIDI OUT when local keys are pressed while recording and the notes as they are played while playing
  • offers two independent 100 note memories
  • sequencer memory is non-volatile (saved once leaving the record mode)

The MIDI channel is set by holding down one of the first 16 keys while turning the Pro One on. Every other key will set omni mode.

For the MIDI note assignment I relied on the user’s manual of the Pro One showing the keyboard scale from C0 to C3.
Therefore, the lowest C corresponds to midi #24.

While working on a Rev.1 Pro One one should always keep in mind that the PCB-mounted mains tranformer is much too heavy for the PCB and that the track clearance on the primary side is much too low to comply with any safety standard or electrical code. In this particular case, the transformer was already damaged and had to be replace anyway. This photo shows a possible solution: screw terminal for power cord, protective earth securely bolted to case, fuse in series with mains, double-pole switch with double isolation, toroid tranformer bolted to chassis. The original power connector to the main board has been rewired and re-used.

Sliding again

As if disassembling and cleaning all linear pots  of this ARP 2600 is not enough to deal with, several levers were also broken and missing, together with the caps.

Amadeus from made a 3D CAD model from the parts of a broken slider pot, evaluated the best production method and had some new levers sintered for me.
The caps could be made on a typical 3D printer.

All parts cleaned, waiting for assembly

Eveything lubricated and put back together

Good sliding action and no dropouts anymore. Mission accomplished 🙂


Wave Thinking

Today I analyzed a PROZ84 board from a 2.3 with some “odd” sounds.
There was audible distortion, varying with the pitch and the partial wave number.

The oscilloscope revealed (sorry, no image here) that in cases of highest distortion levels only every other sample appeared on the DAC outputs, while the others were at ~0 volts. Half of the partial wave samples having a value of zero is hardly a problem of the wavetable playback, but most likely to be caused by wrong data in the wave RAM.

It turned out that one of the 74S257’s multiplexers outputs in the write circuit had a stuck low output on RAM address line A0, so only even samples were actually written into RAM, with the odd locations containing what is left in the DRAM’s capacitors – in this case, 0.


Update for the Wave 2 modern display firmware

Lately LCD modules showed up that do not comply with the unwritten standard that returning the cursor to home position on a non-shifted display will not take longer than 40 microseconds. The datasheet of the HD44780 controller chip clearly says 40µs up to 1.64ms, with the latter being common for un-shifting shifted displays. A display taking longer than approx 53µs for executing the home command, erratic behaviour will occur.

As the controller originally used in the Wave 2 does not know about shifting, the problem was solved by using a command to explicitely set the cursor position to 0 (zero) instead of the home command.

Here’s the result:


A big one: Arp 2600

This Arp is here for several reasons. Although a basic repair has already been done before it came in (S&H replaced and some other typical tasks for the 2600), but there is a lot of work to be done. Here’s a first impression of the Arp

There are may modifications, some not yet fully understood. The main problem are several broken potentiometer levers – original parts are not available, nor aren’t rebuilds of similar quality. Therefore we are evaluating levers and knobs made using 3D printing techniques. This work is carried out by and the parts are now waiting for installation.

Here are some photos of the major modifications and additions:

This is a simple 2 transistor astable multivibrator which is actually not connected to any parts of the Arp.
The output signals are routed to separate jacks and may be used for modulation purposes.


A 2nd order RC filter with switchable corner frequency. The filters are passive with an buffering op amp in between.


One of the two transpose circuits connected to each of the two VCOs. Two switches expand the transpose possibilites by fourths and fifths.

During repair, all modifications will be reverse engineered and documented. For now, here’s an overview of the modules in the Arp:

Another 360A leaves the lab

The next PPG 360A has just left the workshop. It seems that all 360A that have been sitting around for years or maybe decades will need several hours of troubleshooting and repair and a bunch of 15..20 ICs.
Fortunately most of the parts are easily available, and the next one is already in work.

As you can see, two boards are missing – between IO and TONRAM there is one of my RAMPROM replacement boards hiding. Less ICs to fail, less power, less heat.

Korg Trident – Intermittent Problems

After restoring the volume potentiometer from contact cleaner abuse and replacing a bad SSM2044, the Trident still showed some intermittent problems.
One time, the attack on one voice has gone, the next time another voice was lacking release. After some measuring and finally swapping the double transistors, the fault eventually disappeared, but new problems came up.
It turned out that the solder joints of the rivets used as vias on the paper laminate board showed microsopic cracks. Resoldering brought all ADSR features back, needless to say that all rivets of this board need to be reworked.

An example of the affected vias – resolder all of them, there are probably 50+ on this board good for all kind of intermittent trouble.

A side note: this synth uses quite interesting ADSR generators, a 1µF electrolytic is charged and discharged by controlled current sources built around double transistors which are selected by the MOSFETs in a CD4007 package.
The generated envelope controls a VCA built from two selected standard transistors – not always easy to troubleshoot and repair, but the parts are much easier and cheaper to obtain than the typical CEM3310/3360 combo.

CS50 – CS for CraftsmanShip

After all those plastic synths and heavy metal profile construction with no chance to operate in disassembled state this Yamaha CS50 was a real pleasure to work on. Nice condition and just a minor fault: the external modulation input did not work anymore. There is not much in between the input jack and the modulation source switch – just an OP amp and a capacitor, and unfortunately no protection resistor. This means the OP amp is blown when a high level signal is applied to the modulation input with its level pot turned fully clockwise. After replacing the OP amp and adding a series resistor to protect its input the external modulation was back again.



Some trouble evolved within this Evolver – the barrel jacks permanently failing in laptop computers don’t do better in a synth. Once used on stage, a broken jack, plug, cable… is almost certain. A quick repair on the kitchen table brought it back to live, but a serious improvement would require another connector.

Testing the ELTEC

The PPG Waveterm A was built around an industry processor board named EUROCOM II V7 made by ELTEC. Based on a 6809 CPU it contains two 6821 PIAs, a 6850 ACIA with baud rate generator, a 1793 FDC controller, 64k of dynamic RAM and two 4k ROM sockets. One out of three 16k memory pages can be mapped to the integrated discrete monochrom graphics controller.

Due to the complexity it can become rather complicated to trouble shoot those boards. As lots of counters and other TTL ICs are required for the memory to work properly, chances are good that an application’s firmware won’t run due to memory problems.

For the software I’ve created only the CPU, ACIA and baud rate generator need to work. The first routines run completely within the CPU registers and do not require any RAM.
Once started, a welcome message should appear on the connected terminal. If not, basic 6809 system troubleshooting needs to take place.

As soon as the terminal comes to live, the first 4k of RAM will be checked by a AA/55 pattern. If a mismatch occurs, the routine will stop and the original and read bit pattern is displayed.
Assuming the first memory page is in good condition, the stack pointer is initialized and a menu appears.

For now, four tasks are available:
1. Memory test. The whole memory between 1000 and EFFF will be tested, an error will be displayed with the actual address, the written and the read bit patterns. The test will run forever.
2. PIA output. All 8 bit PIA ports show a square wave with a frequency of 19.2kHz on PA0/PB0 decreasing to 150Hz on PA7/PB7.
3. PIA input. Binary patterns of all four 8 bit ports are displayed continuously.
4. Screen test. A test image will be show.

To Do’s:
– add two more test screens, switchable using the page select lines
– add a test for the hardware scrolling functions
– keep away from the FDC, this will have to be troubleshooted in the target system if necessary

ELTEC Eurocom II V7 board under test

A first impression of a Eurocom II on the emulator