Friday, 15 March 2019

Yamaha QY20

After fixing the MS-20 I was having a lot of fun with it, but really wanted some kind of sequencer. I would have preferred not to have to rely on a laptop and was looking for something small and cheap.

I found this Yamaha QY20 on sale from a Japanese eBay seller, listing as non-functional. It was a portable sequencing device from the early 90s, with some workstation-style sounds and MIDI I/O. The service manual was available online, it was very cheap and it looked like a decent form-factor so I took a chance.

Yamaha QY20, with carry case

I applied 12V from a bench power supply to the DC jack - it pulled around 135 mA but didn't show any sign of life. However, when I connected the headphone jack to a speaker and pushed some keys I could hear a piano. I could also connect up the MS-20 by MIDI and trigger notes using the QY20 keys. It looked like the there was just an issue with the display.

Usually LCD displays have a bias voltage supply that is much more positive or negative than the supply for the logic chips. I expected a bad switching power supply. The service manual is very comprehensive and has a complete schematic, it shows a MAX680 inverting charge pump generating a negative supply of -9.3 to -9.6V from the 5V rail. Either the charge pump has failed, or the contrast pot has gone open circuit.

LCD bias supply schematic, from service manual.

It's fairly easy to take apart the QY20, but there is a copper sheet for shielding that needs to be desoldered.

Shielding has to be desoldered at the MIDI and line output connectors.
Digital and audio PCB
The insides are fairly typical surface mount construction for early 90s, particularly for a portable unit.

Funnily enough, the microcontroler is a H8/520, the same family used in the MS-20 20+ years later. The main soundchip is a YMW-258-F "AWM & FM Tone Generator", which is likely to be some variant on the chips Yamaha were doing for synths and video games at the time.

I measured the Vee bias voltage for the LCD at only -5 V. It should be closer to -10V, as the MAX680 charge pump is an inverting and voltage doubling converter. I would guess that one switching stage inside the MAX680 is bad and it is only inverting.

I removed the MAX680 (looks like I forgot to take a picture of this, it's a SOIC 8 chip on the backside of the board) and the 4 22 uF switching capacitors (C26,C27,C28,C29).

C26-C29 removed
After installing new parts, Vee was now -9.2V - a little bit low, but certainly a change. When I re-attached the LCD board, the display worked again.

Reflective LCDs are surprisingly difficult to photograph,
I can change the cheesy electric piano to 100+ other throwback workstation sounds! Time to learn how to use this as a MIDI sequencer.

Tuesday, 12 March 2019

Korg MS20 Mini

Korg MS-20 Mini Repair:

I have wanted one of these since they were first released, but ended up waiting until a broken one came along. I probably overpaid a little for this, but the fixes weren't too bad.

This came from eBay, the seller had bought it as faulty with a non-functioning headphone output. When they received it they found it wouldn't power up, so they sold it on again. I am expecting some kind of power fault and something else wrong with the headphone output.

Despite the popularity of this synth and the number of mods documented online, there isn't a lot of information on some of the parts used, so hopefully this will be useful to others.

There are 3 main PCBs, all mounted to the sheet metal case and connected together with cables.
  1. Digital board - power entry and power supplies, the microcontroller, MIDI and USB ports.
  2. Analog board - all synthesizer circuits, knobs, switches and jacks.
  3. Keyboard PCB - breaks out the keys to a connector. I didn't look at this.

Digital Board (KLM-3163C):

The digital board is the largest difference between the Mini and the original MS-20, and unfortunately there is no available manual or schematic.


There is a switching power supply on the left hand side of the image, the large transformer/dual inductor is a give-away. The microcontroller (IC3) is right next to the keyboard connector and all the pins are routed right to it, so it is doing keyboard-scanning as well as USB and MIDI.

I applied power and found that there was no voltage at the switch mode power supply inputs, and therefore nothing powering the analog board.

There is a component marked "F2" that is connected to both the SMPS input and the incoming 9V. Despite the silkscreen, I was pretty sure this was a P-channel FET and not a fuse. IC50 is a CD4011 Quad NAND which is powered by 9V - I am fairly certain this uses some surrounding resistors, capacitors and transistors as a timing circuit to turn on the P-FET gate a short amount of time after power is applied. This works as a "soft-start" and limits the inrush current.

Temporary fix to get things working

To test this, I just shorted across the FET and connected power to the switching power supply chip. This worked, I now had +14.5V and -14.5V rails, and when I connected everything back together the synth worked when using the main output jack, but not the headphone jack.

I later confirmed the part number of F2 and replaced it.

Here are part numbers for all the ICs and transistors, from my notes:

F2 (?) - RRL025P03 - Pch -30V -2.5A Power MOSFET - used for power supply polarity protection. Amusing numerated "F2".

IC1 - R1154H036B - 3.6V voltage regulator. MCU (IC3) runs at 3.6V.

DT1 - DT4 - 2DTC114 digital transistors, marked "24" - used in other Korgs.

IC2 - BU4227 - marked YU, used in other Korgs. This is an under-voltage detector, it resets the MCU if the supply voltage drops below 2.7V.

IC3 - H8S/2210C - main microcontroller, covers keyboard scanning, USB & MIDI.

IC9 - JRC 4558 - opamp, not sure what this is doing.

IC44 - 74LVC1G126DCKRG4 - Single Bus Buffer Gate, marked CN5, also used in Kronos.

IC47 - TPS54240 3.5-V to 42-V Step-Down DC - DC Converter With Eco-Mode™ - uses an onboard transformer to generator the analog voltage rails, which are regulated down further by IC49 and IC53, probably to clean up the switching noise.

IC49 - TPS73801 1.0-A Low-Noise Fast-Transient-Response Low-Dropout Regulator - regulates the +14.5V rail.

IC50 - CD4011 - Quad 2 Input NAND gate

IC 53 - TPS7A340 1–20-V, –200-mA,Low-Noise Negative Voltage Regulator - regulates the -14.5V rail.

DT1 - DT4 - 2DTC114 digital transistors, marked "24" - used in other Korgs.

PC1 - Toshiba TLP2368 Optoisolator - isolates the MIDI input.

Analog Board (KLM-3162C):

I also wanted to fix the headphone output, if possible.

The analog guts are all on one large PCB that has all the pots and jacks mounted. The jacks and pots are not actually panel mounted with nuts, they just poke through the panel. This does not give the greatest tactile response - everything wobbles a little bit - but it probably explains how Korg managed to keep the price so low. There is at least a large sheet metal cover to stiffen the PCB.
KLM-3162C with shield

KLM-3162C uncovered

The headphone output is in the upper right. The original MS-20 has a very simple headphone amplifier, just an opamp driving each side. The Mini is pretty different, and seems to add transistors to buffer the opamp outputs.

Headphone output circuit. C251 was removed for testing and later repopulated.

I held down some keys and probed around for a signal. The headphone output seemed to disappear at one side of a 10 ohm resistor (R314 & R315) for both left and right outputs. I desoldered these and they confirmed they were open circuits. I replaced them with new 0603 10 ohm parts and the headphone worked. It's possible these were killed by someone patching the headphone jack into somewhere strange and pulling too much current (?)

Offending 10 ohm resistors

Korg Service Manuals:

There are a lot of Korg schematics and service manuals online. Many are publically accessible on the portal, even though the page requires a dealer or repair center account. I used some of the following as reference material:

MS20 Original service Manual

MicroKorg Service Manual

MicroKorg XL Service Manual

Korg Kronos Service Manual

Korg Volca Bass Service Manual:

Monotribe schematic:

Monotron delay schematic:

Thursday, 14 February 2019

Another Ibanez ES2 Echo Shifter repair

I took another look at an Ibanez Echo Shifter that I had previously repaired, but had come back to me. This time it wouldn't light up, but did pass a clean signal. I suspected a bad power supply for the digital section.

I opened it up, and I could see ~8.5V volts powering the opamps but no voltage at any of the digital parts, where I would expect 3.3V.

Way back in July 2017 a commenter asked if I knew what the part number for U14 was, likely to be a buck-converter, and that they suspected it was a Texas Instruments TPS62056. The package, function and chips marking all looked like a perfect match so I ordered a few this week to see if it was right.


Desoldering was straight-forward, I used hot air and kapton for protecting parts I didn't want to overheat.


The new chip has near identical markings to the old one. Re-soldering was a little harder as there isn't a lot of space, I ended up removing C73 temporarily to get better access. The bridged pins are fine, all those pins are connected together at the PCB.


Everything works again with the new chip. Thanks to Shane Bussiere for doing the research and sharing the part number, sorry I didn't help out.

All working again.

Looking at the TPS62056 datasheet, I can guess why this failed. The buck converter chip has a maximum Vin of 10V, the Echo Shifter runs the power jack through a series Schottky diode for polarity protection and then to the TPS. If you use a 9V power supply the chip gets around 8.6V, which is fine, but using a 12V power supply or higher will probably kill it. I couldn't find a compatible chip from TI with the same footprint and pinout but higher maximum input voltage, let me know if one exists.

I would really like to modify this to add a modulation rate control. I hunted around for modulation signals and unfortunately there doesn't seem to be a LFO onboard, it looks like the modulation is done in software in the ADAU1701 DSP. A modulated square wave is run out of the first audio DAC on pin 46.

While I had this open I desoldered the 24AA128 serial EEPROM and dumped the contents, it can be downloaded here. Afaik the ADAU1701 instruction set is not publicly documented, so I don't think the firmware can be easily modified.

Friday, 8 February 2019

2 70s Electro-Harmonic Small Stone phasers

2 vintage Electro-Harmonix Small stone phasers. The artwork for both of these is almost identical to the 90s reissue (which uses LM13700 OTAs) but the lack of any LED indicators gives away that these are older - probably late 70s or 80s.

Small Stone #1

I think this is the pedal I have owned the longest with repairing. I started buying broken stuff  to repair from eBay in 2011, and picked up a couple of vintage Big Muffs. One was a lot that came with this pedal, a Small Stone, pretty much as pictured - no knob, no pot, no switch and a lot of broken wiring.

The Small Stone has 4 phase shifting stages, using 1 OTA as an LFO and 4 more as variable RC filters that give varying phase shifts. Early versions of the Small Stone used CA3094 OTAs branded as "EH1048", a house-marking for Electro-Harmonix. This version is an "Issue J" and has the 5 EH1048 chips, dated to 1977.

Issue J

EH1048 - 1977, week 32.

I bought a new 24mm reverse log pot, installed a 3PDT switch and re-wired the pedal (true bypass, why not). It didn't pass an effected signal. I think I put it aside at this point, I suspected that the OTAs might be bad and didn't have any replacements.

I came back to this recently. The oscilloscope showed that there was no LFO signal anywhere on the board, even though most of the voltages on oscillator OTA looked reasonable. I bought an RCA metal can CA3094 and replaced it - now I had an LFO, but still no wet signal. Looking at the input and output pins of each OTA, I could see that the first stage was phase shifting, but the second had no output. I shorted together the input and outputs of the second stage, and now I had a phase-shifted signal. It wasn't quite as deep as it should be, with only 3 or 4 stages active, but it verified that the other two ICs were good. I ordered one more CA3094 to replace the dead EH1048, and complete things.

Before re-housing
I sourced a new hockey put knob that fit the new 24mm pot. Originals are hard to come by, but it fits the right aesthetic.

This was inspected by... Elsa? Cheers, Elsa.

There is usually some foam behind the PCB on the back panel of the housing. The PCBs just hang off the back of the rate pot, the foam is to prevent it from shorting out on the back panel. This foam had perished, so I taped some card down to insulate the PCB instead.

Small Stone #2

This second unit is a similar vintage, I picked it up hoping it might help repair the first one.

This is a slightly different PCB (with a phenolic substrate instead of fibreglass?), but looks to be the same circuit more-or-less. It also has 5 EH1048s, dated to 1979.

This one actually worked despite being sold as faulty (this is not that unusual). It just had a couple of quirks. It was a big help in verifying the switch wiring on the other unit, and for taking reference voltage readings off the OTA chips.

Phenolic PCB, instead of fibreglass?
The first was that the rate pot had some odd damage, the casing was partially open. I guess this could have been caused by dropping the pedal onto the knob, or by pulling on the PCB while it was still attached to the enclosure. Or during factory assembly, this was EHX after all. This was easy to close and re-crimp with pliers.

Opened pot housing...

...closed again

The second was that it would start to oscillate with the "color" switch in the up position and no input. It works fine with a guitar connected, or a buffered pedal in front of it, but with no cable or dangling unconnected cable it will start to ring at the top of the phaser sweep. As far as I can tell this is just something that this revision does, to fix it I would have to switch to a shorting input jack or modify the pedal to reduce the positive feedback when the color switch is in the up position, neither of which I really want to do.

These turned out really well. Feels good to have them done. They sound almost identical, the only change that jumps out is that the two different brands of pots don't quite match up - the plastic shaft CTS pot physically rotates further than the new Alpha pot, so the rates are slightly different when the pointers are matched by eye.
I don't have a modern Small Stone, or a Sovtek, to compare with. They definitely sound smoother than JFET phasers I'm used to.

One or both of these will probably hit Reverb in the next couple of days, get in touch if interested.

Electro-Harmonix Stereo Memory Man (EH-7811)

The classic, basic, no-frills BBD delay. This is an EH-7811 revision, dating from around 1980 based on IC codes. This version is main powered (240V), runs at +/-15V internally, Panasonic MN3005. There is no LED, I think this was the last version without one. There is an Echo/Chorus switch which probably reduces delay times, and in-phase and out-of-phase outputs for a "stereo" effect.

This is another one that I have had a for a while, and later came back to. I bought this a couple of years ago and nearly got it working, then hit a dead-end.

It was pretty dirty on first inspection, and missing a knob for the blend control. The power cable had been shortened to a ridiculous length, about six inches, making it awkward to work on.

Before cleaning...

The original eBay picture shows this off:


The insides show that the PCB is complete with no obvious damage. It does anchor everything off of board-mounted pots which are only on one side of the large PCB - the other end floats and tends to cantilever.

PCB as received

The delay/chorus switch is almost entirely missing, just the frame left.

Interior of case, Echo/Chorus switch.

Closer inspection of the PCB found that the Blend potentiometer's pads had all craclked off. The pot was still hanging onto the board, but nothing was electrically connected. I ran some small jumpers from the pot back to the nearby traces. I also installed a new sliding switch. I referenced a schematic for the later EH-7811B at David Morrin's excellent site. The main difference (apart from the LED) seems to be that this version has an extra 741 opamp to invert the delay signal for the out-of-phase output.

At this point, I had some signal coming thorough, but hugely distorted. All output opamps were saturated, sitting at ~ 13 or -13 volts. I socketed and replaced some of the opamps with no change. There was a DC offset being introduced somewhere.

I had a few ideas:
  1. dead opamp, or opamp feedback network. No changes when swapping opamps and measuring feedback resistors.
  2. leaking AC-coupling capacitors. I replaced some 1uF caps of a type I had seen fail before with modern film caps, no changes
  3. Missing ground node somewhere...
This went back into the "fix later" box for a while. I dug it out and went over some of schematics for other revisions and noticed that one side of the blend knob should be connected to ground. My blend knob had been cracked off the board, I could barely see a small track below the pot's pads that should have been connecting to ground.

I ran another jumper wire to ground, and now all the outputs were sitting at 0V.

I fitted a new mains cable so that is actually usable. There is no internal fuse, so I changed the cable fuse to a 3A part. I tried the original opamps in the sockets, but the outputs got noticeably more noisy. Maybe semiconductor processing has improved to the point where new 4558s and 741s are less hiss-y.

There was some serious clock whine, especially at long delay times. Fortunately I was able to completely trim this out.

PCB after repairs.

PCB after repairs, parts replacement.

I've said before I haven't noticed huge differences between analog delays based on BBD types. I had the Aqua Puss at hand for comparison between a V3205 and MN3005. The Memory Man sounds cleaner, if that makes sense? Less distortion on each repeat, a bit closer to the original signal. Still sounds like analog delay, just not as overblown.


Monday, 4 February 2019

Digitech PDS 1700 Chorus/Flanger

I picked this up out of curiosity, I'm fond of the PDS series and the off-the-shelf design. This is a digital Chorus and Flanger, where the traditional BBD design has been replaced with an 8-bit digital delay line (i.e. not DSP or modelling) and the delayed signal mixed in the analog domain. This one worked but didn't switch very well.

PDS 1700

There aren't too many surprises when comparing the insides to the PDS delays. There is the same ADC (ADC0820) but instead of DRAM it uses a single 2k 6116 SRAM. This is probably because the required delay times for a Chorus and Flanger are much shorter (this pedal maxes out at 51 ms) they could spring for the more expensive SRAM. This also means all the DRAM refresh circuitry isn't needed, and there are fewer logic chips overall.


PCB Backside

The switches were pretty flaky. The mechanical design DOD/Digitech used at the time has a poor reputation for reliability. I find that if they are maintained they work well, but I don't think they stand up to force. The classic problem is that switches that won't switch on the first try get a harder stomp the second time, and things deteriorate.

PDS hardware. Input jack looks non-original.
The paddles press down on momentary push buttons on PCBs inside the pedal. The paddles bear against two bolts that pass through slightly oversized holes in the enclosure. There is a preloaded spring on the bolts on the insides to push them against the enclosure, with nuts holding the springs under tension. There is another return spring on the switch PCB to bounce the paddle back.

I replaced the switches, I'm using these. There was some play in the paddles, the bypass one would rotate slightly in place, meaning it would not always hit the button switch. I tightened down the nuts on the inside until this play was removed, then it switched every time. You need a 3/32" hex key and a 1/4" socket wrench to make the adjustment.

The effects are cool. There is interesting play between the delay time and depth controls, the flanger side can go from covering high to lower-frequency ranges, kind of like the difference between a Boss BF-2 and a HF-2. The chorus does odd (cool) detuning effects at high delay times, modulated chorus in the middle and traditional chorus at minimum delay.

I have this one listed on Reverb if anyone is interested.

Thursday, 31 January 2019

Another Electro-Harmonix Freeze

I got another Electro-Harmonix Freeze fairly cheap, it sounded like it had been killed by an incorrect wrong power supply. I took it apart and found that it was a newer revision than I had seen before, the board was EC-D68 Rev C (the last one was Rev B). The main difference I notice was an AK4558 codec instead of PCM3052A.

Freeze Rev C PCB, flash ROM temporarily removed.
Main PCB, LM317 temporarily removed.

There was no sign of life, I found that the series Schottky diode at the 9V input had failed open-circuit and that the LM317 had an internal short between it's input and output pins. The fast/slow/latch mode switch also fell to pieces when the board came out of the enclosure.

I replaced the bad diode with an SS14L (it was the correct size and I already had some at hand) and installed a new LM317. I thought that the LM317 was supplying 3.3V to the DSP and this would fix everything. Instead, the pedal would pass a clean signal but with a clicking sound once or twice per second. I found out that LM317 is actually used as a 1.25V regulator, which is only used for the analog VCC of the DSP56374.

I thought the clicking may be the DSP watchdog timer firing, possibly because it was missing program code, so I removed the flash memory and dumped it. It was a perfect match for the dump I made of the first pedal I repaired, so no problem there.

Measuring power at every IC I found that the 5V rail was high at around 5.9V and the 3.3V supply was sitting at 3.9V. I thought that U5 and U8 (both SOT-89 packages) were also voltage regulators and that maybe they had been damaged as well. I also noticed that U6 is connected to the reset pin of the DSP, and it was resetting the DSP every second or so. My guess is that U6 (also connected to the 3.3V supply) is some kind of voltage supervisor and it is resetting the DSP because of an incorrect 3.9V at VCC.

I found a great picture at freestompboxes (thanks to Steven_M!) showing that on previous version, U5 was 78L05 (5V regulator) and U8 was BA932 (who knows, but from context it has to be a 3.3V regulator).

U5 & U8 on another revision.

The Rev C board in front of me had "BA420" on both chips. Possibly EHX decided to run all 5V parts off of 3.3V, and doubled up on regulators? I don't

To test this out, I decided to remove both chips and apply 5V and 3.3V from  external power supplies.

U5 & U8 removed, external power applied

This set up was a little awkward, but everything work correctly. Current draw looked totally reasonable, 19mA from the 5V supply and 21mA from the 3.3V. No resetting, no clicking, and the pedal could freeze audio in all 3 modes.

Current draw on 5V and 3.3V supplies.

I ordered L78L05ABUTR and MCP1804T-3302I/MB as replacement 5V and 3.3V SOT-89 regulators. I also used 2MD3T2B2M2RE as a replacement switch. This replacement is not threaded for a nut, but recent version of the Freeze don't look like they are threaded either. Pedal works like new.