Tuesday, December 6, 2016

Korg Trident Mk II - My first service job

Time to get started

So, like I said in my previous post, my blog here and posts elsewhere got a few people asking if I could fix stuff for them. Long story short: I registered a business and now accept service jobs reparing synthesizers and various electronic stuff. I'm not gonna bore you with repeating all that again though. This post will be about the first job: a Korg Trident Mk II.

I don't have any photos of it closed, so enjoy the guts :)
This is a long post (Seriously long! I MEAN IT!), probably partly because the repair was a bit complicated and partly because it's my first "real" post in a long time :) I'm sure I'll make them shorter and less boring soon after this one.
If you're still interested, read on...

Can you look at it?

First off, this job was started back in 2014 so my memory is a bit vague. I did take notes while working on it and I hope they will come in handy now when I over 2 years later try to write down what actually happened. Let's see :)

I was asked to look at a Korg Trident that had been unusable for a long time. The problems reported, the way I have it noted at least, were:
  1. Synthesizer section was dead.
  2. Keyboard tracking was horribly off.
I started by doing a visual inspection of it. Sometimes the eye alone can give you clues a lot quicker than spending time probing around with a multimeter and oscilloscope. However the exterior looked fine, so I powered the thing up and tried it.

I immediately noticed that the tracking was indeed horrible. Listen to this :)

Further testing showed that the LED for the "Synth" section enable button wouldn't toggle when you pressed the button and it didn't enabled it either (so it wasn't just the LED). So yes, the synthesizer section was dead.

PSU

I did a quick check of the PSU to see if the voltages were stable and reasonably correct. They were, in fact they were only off by less than 50mV, which is actually not bad. This was probably because another tech had been looking at it previously, and I believe he had done some PSU calibration.
There was no visible ripple or noise either, so the PSU was fine.

Battery leak, battery leak. It's always the battery leak...

A common issue with Korg Polysix and Poly-61 is leaking RAM backup battery. I should know because I have both synths, and they both had problems with leaking batteries. I was curious to see if the Trident, being from the same era, could suffer from the same problem.
Here's a picture of the Programmer CPU board, KLM-380.

Synth CPU board
Look at the area around the battery and the CPU, which for some reason is always placed next to the battery in the old Korg poly synths. That doesn't look good!

I started probing the CPU pins to see if it was running.
There were no signals anywhere, not even on the XTAL pins. This thing was dead.

I removed the CPU, and this is what I saw.
CPU socket area
There was obvious corrosion damage as well as scratch marks that look like somebody had been trying to scrape some of that crud off, presumably when the battery was replaced last time.

So what did the battery look like?
RAM backup battery
Not too bad actually! I mean the battery itself. This is not a leaky battery. A previous battery, likely an old NiCd type, had leaked and been replaced by a new NiMH battery, but the corrosive electrolyte hadn't been cleaned up. Doesn't matter.
This area needs to be stripped and cleaned.

Something to consider when dealing with leaking batteries is "How was the keyboard stored while the battery was leaking?".
If the keyboard has been standing on its side, you can often find that the corrosion has moved sideways and sometimes gathers at the edge of the board and causes a lot of damage there, seemingly mysteriously far away from the battery itself.
In this case, since the CPU board and battery is actually mounted on the front panel (and lifts up when you open it), if the synth is sitting flat, where does the electrolyte go? It probably follows gravity, so down, on the board below.

There is a board in the bottom of the Trident, and this is what some traces on that board looked like, just below where the battery is.
Board directly below the battery.
Something to keep in mind when looking for further problems and dealing with leaky batteries!
Well, the CPU/battery area needs to be cleaned in any case.

Here's what the board looked like under the CPU socket after I had desoldered it (which was not much fun since most of the solder had corroded and was basically not solder anymore...)
Bad corrosion under CPU socket
Seeing that, it was no surprise this thing was dead. In fact, I guess the CPU itself was probably fine (though the pins needed a bit of cleaning from corrosion as well), but the actual problem was the board and socket.

I removed the battery and the components next to it, that also seemed to be affected by the corrosion.
This is the area where the battery was (and you see the CPU area in the background).
Under battery and nearby components
In the photo above you see the footprint for a diode D3 right next to the battery.
That one was my favourite :)
D3... We'll replace it too, ok?

Cleaning time!

I used a "Dremel" (not an actual dremel, but a cheap shit kind of the same tool) and basically sanded the damaged parts of the PCB to get rid of as much corrosion as possible. A crappy little toothbrush with soft plastic bristles isn't of much use here.

I'm actually not sure if what I did next is good or bad, but I cleaned all the holes with vinegar and let it sit for a few minutes. The electrolyte is an alkaloid, and any leftovers still hiding in the holes would be difficult to get rid of mechanically, so I wanted to let whatever remained react and become neutralized. I used a pure diluted vinegar, which is basically just watered down acetic acid.
I washed it off properly using lots of water and finally I gave it a good cleaning with isopropyl alcohol.

Now there were bare copper traces exposed on the PCB, and they should probably be sealed to prevent oxidization and more corrosion, in case there were leftovers hiding somewhere.
After letting the board dry (I can't remember for how long, but it was a few days I think) I sprayed some PCB protective lacquer on the area and let it dry.
It creates a very shiny surface so it's hard to take photos of, but here's what it looked like.
Nice and shiny!
A serious improvement compared to earlier! The lacquer is designed to melt from soldering so you don't really need to worry about that, though it will probably create some weird fumes and may not be the best spa treatment for your solderin iron tip.

I started by soldering a new socket for the CPU in place and then began checking continuity between the socket pins to the expected destinations. It appears some traces had been so badly corroded they were open circuit, so a few patch wires had to be added to the bottom to restore contact.
Some patches to fix bad traces
Apparently I stopped taking photos at this point, so I'll show you one of the last pictures I took.
Sorry for the lack of photos in the upcoming sections...

Here's the Synth CPU board with all the removed components replaced (except the CPU itself).
Cleaned with parts back!
The keen observer has now noticed I did't put one of the resistors back (R29, between the CPU and D3). D3 isolates the entire +5V supply from being powered by the battery when power is switched off. R29 is used to limit the charge current to the battery when power is switched on.
As I recall I installed a 2032 lithium coin cell holder on this board before returning it to the owner. In order to prevent it from charging, R29 must be replaced by a diode to block incoming +5V from entering the coin cell. So, that's what I did (but I have no photos of it).

But, the synth section was now working!

However there were still a bunch of remaining issues. My notes say:
  • Synth section is now producing output, but pitch is unstable and tracking is bad.
  • Storing patches doesn’t work.
  • Write LED permanently on.
  • Unable to switch off Synth section once switched on.
  • VCO2 detune doesn’t work.


Synth Output Enable

I did manage to toggle the synth section once by just hitting the Synth output enable button repeatedly. It didn't seem like the switch was bad. Something else was going on.
I removed KLM-381 from the panel and started looking at the button circuitry.
The Synth output switch is actually connected to a small toggle circuit based around IC2 (14011B quad NAND) and couple of resistors, caps and a diode.

I checked the passive components in the circuit, and while I couldn't get a definitive result measuring in-circuit, I didn't find anything obviously wrong (like shorted cap or something).
While probing the NAND, I noticed pin 10 (output from the first inverter circuit) was sitting at around 3V when the synth output button was pressed. I wasn't sure if this was normal since it's not a "healthy" level for 5V TTL, but started suspecting IC2.
A big clue came when I switched the scope probe from x1 to x10, which actually made it possible to occasionally switch the output on and off.

One thing I've learned is, that if the state or stability of your logic circuits is easily affected by the oscilloscope probe (in particular for CMOS, though this was TTL), then you may have a problem with floating inputs, bad pullup/pulldown or bad outputs that are too weak to drive the signal to the desired level.

I took a guess that IC2 was bad and replaced it with a new 4011B.
Problem solved. Synth output toggle switch now worked fine! Well, except that the switch itself was giving me some trouble being a bit worn out, but I saved that for later. One issue less.

Write LED is always on

Next up, the write LED stayed permanently lit and it was also located on KLM-381.
The LED is connected to an output of IC5 (74LS75) and since the LED was obviously working, the next step was to check if the IC outputs had the expected values when considering the inputs.
In my notes it says that pin 3 (D2) is always low when the enable pulse on pin 13 fires, but Q2 remained high and /Q2 remained low. If the input (D2) is low, you'd expect the output (Q2) to be low.
I suspected IC5 was bad, but didn't replace it yet. More investigations were needed around the same area. Keep reading...

Storing patches doesn't work

Next problem, storing patches doesn't work.
I started probing around the write pulses originating from the write button.. and...
I'll keep this story short by copying my final notes regarding this issue.

Write Enable switch on back of unit was set to “Disabled”. RTFM.

Saving patches works.
Very professional! Let's never speak of this again :)

Manual LED is always on

I hadn't made a note of this before, but apparently it was a problem too.
Now, the manual LED is connected to an output of IC5, just like the Write LED I just mentioned above. In fact, the Write LED is connected to /Q1.
I could see the state change on the input (D1) before the enable pulse was fired, but the output remained fixed.
At this point I had two indications that IC5 was bad so it was time to remove it.

For some reason I placed it in a breadboard and made a simple function check (I don't own an IC tester) and came to the conclusion that of the 4 gates, gates 1 and 2 were dead and kept the outputs high constantly no matter what. Since gates 1 and 2 were responsible for the Manual LED and the Write LED, the pieces fell into place.
I put a socket in for IC5 and put a new 74LS75 in it and tested it. Now both LEDs were working as intended.

Oh, I found another photo of the Trident:
A photo of IC2 and IC5 socketed and replaced. Good? Well...
Is it a good idea to socket ICs below the front panel?
I have a habit of using sockets when replacing ICs. I'm not sure it's always necessary.
On one hand, if the IC died once, the design may cause it to be prone to failure so perhaps it's good to leave it socketed.
On the other hand, it may now be the last IC to fail since it's the most recently replaced and the socket itself may cause more problems instead (though I try to use milled sockets).
In this case though, the socket had caused the IC to sit a bit higher than normal. While this isn't usually any problem with boards sitting in the bottom of a synthesizer, it can be a problem when the IC is mounted on the front panel, facing the panel...
OK, I was stupid (again). I had to remove the socket and solder IC5 directly to the PCB :)
IC2 seemed to fit anyway so I left it in the socket.

VCO2 detune doesn't work

Now this is a bit of a mystery... I don't actually have any clear memory of a problem with VCO2 detuning, but apparently there was a problem according to my notes :)

Checking the voltage from the detune knob for a start showed that the potentiometer output a varying voltage and checking the output from the mux IC1 (4051) on KLM-380 showed that it was varying its output as expected depending on which potentiometer was being changed, including detune.

Since all the other pots work and the muxing worked, there's no problem with the ADC so the problem must lay at the other end of the chain, i.e. the output to the oscillator circuit.
I might have been able to come to that conclusion sooner if there were any stored patches to test with, but since I just had the programmer CPU and battery removed, patch memory was just garbage so I didn't know exactly where the problem was with the detune; programming or output.

Looking at KLM-379 (signal generator board), the detune CV comes in to IC4 (4558), and I see I have made a capslock warning in my notes:

On KLM-379, IC4 has WRONG PIN NUMBERING in schematic. Pins 5 and 7 are swapped.
Future references are to actual pins, not ones in schematic.

Pins 5 and 7 correctly being the non-inverting input (pin 5) and the output (pin 7). Good to know if you're poking around there some day :)

In any case, I noted that the detune CV was present at pin 5 and was varying between roughly -5V to +5V, but the output (pin 7) and inverting input (for negative feedback) does not change. Hmmm...
The other half of IC4 (since the 4558 is a dual op amp) was involved in the antilog circuit for the pitch CV. So if there was an issue with IC4, it could potentially affect both the antilog pitch CV circuit and the VCO2 detune, which is exactly the two problems this unit had! So you know what happened next :)

...or you thought you did...
I removed IC5 (!) by mistake. It was also a 4558 and I wasn't paying enough attention... I put a socket in and put a new 4558 in there (and before anybody asks, no that wasn't included in the parts cost).
THEN I removed IC4, put a socket in for that and placed a new 4558 in it.
I also replaced a nearby diode (D4) sitting next to IC4. Not sure why but I assume it looked bad.
You can actually see IC4 and the designator for D4 on the photo above showing the corrosion "dripping" down on the board below. So the battery leak happened to also affect KLM-379 in the area right by IC4 and D4. Coincidence?

Now both VCO2 detune and keyboard tracking worked just fine, so all the issues had been resolved!

Finishing

Once I was satisfied everything was working I replaced some switches that weren't responding properly, for instance the Synth output enable switch I mentioned earlier, as well as Flanger strings, Preset 2, Brass 16’, Brass 8’, Strings 4’, Synth Bank A, Synth Bank C and Synth Program 2.

Next up was the calibration, which actually took a while, but since I had been messing around with a lot of CV related stuff I figured I might as well do the whole thing. Calibration involves programming 32 specific patches and storing them in the internal memory and using them throughout the calibration procedure, but there's not really much to write about that since I was just following the instructions.

One thing I actually regret about not taking better care of (as far as I can remember) was the corrosion on KLM-379 (the stuff that wasn't on the battery board). It didn't seem bad at the time and I didn't want to start messing around too much with it. If I remember correctly those traces that had some corrosion were power rails and if they go open circuit all sorts of bad things will happen. I'm hoping though that it's just superficial and not as bad as it was on KLM-380 next to the battery...
Maybe I was a bit lazy. Maybe I didn't want to fix what I thought didn't need fixing...

In the end I managed to restore it to a fully working and calibrated Korg Trident Mk II.
Good as new! (Well.. almost)

I'll leave you with the second recording I made with it. This seems to have been made after I got the synth section and keyboard tracking working but I apparently don't have an IC in one of the sockets on the panel board you see hanging loosely :) I also don't think it had been calibrated at this point.



Now there's a chance the events in this blog don't reflect the order in which I actually did things (like one of the ICs on the panel might not have been put in until after I got the keyboard working properly), but who cares? I fixed it! It felt good. The owner was happy.

Now don't expect all my blog entries to be this long and boring. Sometimes I don't take a lot of notes when fixing stuff. Sometimes I do. Sometimes there just isn't a lot to write about what I did, and I honestly think most (if not all) blog posts of the jobs I've done will be shorter than this one. ;)

No comments:

Post a Comment