PC18 from Jamma+ forum dropped me 2 UMK3 boards around a while back in the hope we could get one of them working.
Both had sound issues.

The first board failed its self test flagging the ROM U2 as bad, when one fails none of the others are checked and the game boots.
The second board just had a feedback sound playing all the time and no game sounds were there even though all self tests pass.

I check the ROM’s out first and they were all fine
Checked the voltages on both boards and they were all fine.

Probing around at the sound ROM’s on the first board I could see that none of the chip select lines were ever being activated. I traced this back to the protected GAL22V10 chip, fortunately I had another GAL from the second board but got the very same fault. Refusing to believe that both chips were bad and in the same way I started looking elsewhere.
The ADSP-2105 chip is the sound processor on this board and could see that all the address and data lines were completely dead.
At this point I wrote the board off and focused my attention on getting the second board running.

I know that if the -5v is missing then the you will get the very noise that I was getting with this one. Only trouble is all the voltage levels were fine to all points I could trace.
These boards have A LOT of small surface mounted capacitors and resistors which makes tracing things out a huge task and as there are no schematics available I was getting a bit frustrated.

Out of frustration I decided to probe the sound processor on the board and found that on start up all the address and data lines are in fact tri-stated just like the first one so it got me thinking that the other board was maybe OK.

Looking around on the underside of the board I found this

This actual mark was tiny until I starting prodding at it and missed it on my initial inspection but there was no continuity between the 2 ends of it so it was cut all the way through. Even better news was it did go to that GAL chip so probably would affect the sound circuit.

I patched the trace with some kynar

Fired the game up and was greeted with a 1 tone sound which actually means the sound hardware is OK, previously I had 2 tones which means there is a fault.

Due to the reason mentioned above I’m not going to try fixing the other board up but I have confirmed that all the processors and stuff are working so there are spares should he need them.

Got this from a Jamma+ forum member.
Sold as having graphics problems that went away when touching one of the MASKROMs.
On testing it had lines running vertical down the screen and the fault did appear to go away when touching K27 MASKROM. This however was not dry joint, it was a floating output meaning the MASK was knackered.

I removed the MASKROM and programmed a replacement 27C400 EPROM.
Game now works fine and all self tests pass.

The game does appear really washed out on my test bench but plays fine in a cab. Not sure why this is but I cant actually see anything wrong with the output signals so ill ignore it for now.

UPDATE:
This board also had another fault where button 2 for players 1, 2 and 3 didn’t work.
Turtles actually has an IO test within its test menu so it was easy to check all inputs.

I traced the signal through a resistor network and could see the state toggling with a button press. Traced it back a bit further to a 74LS253 chip at location C26. Replaced this chip now all working again.

Been after a genuine board of this for a bit now and this one came along. I thought it might be in full working order but once again eBay lets me down.
This one had jailbars through all the main sprites.

More often than not this is a bad ROM and this was ended up being just that.

In the corner of this board there is a bank of 8 MASKROMs that contain the graphics. Unfortunately none of these are in sockets and as I couldn’t really see anything hugely wrong on the scope.
This game doesn’t have a test mode nor does it carry out any RAM/ROM tests on start up so I wrote a little self test program myself in assembly that tested the sprite RAM at location 0x0c2000 – 0x0c3fff in the address map and simply would display a “P” for a pass or “F” for a fail.

This passed no problem so my hope is for one of the ROM’s to be dodgy as opposed to one of those custom chips.
I set about desoldering them one at a time and checking along the way. After the fourth one I found my problem with a bad ROM dump.

My dump is on the right and the MAME set dump is on the left. As you can see, bit 0 is stuck on.
I fitted a socket and reinstalled the ROM but with data line D0 lifted and checked it out with the scope.

On a closer inspection it shows that the logic level isnt dropping below 2v.
The MASKROMs are pin compatible with 27C080 EPROMs of which I have none. I have ordered some so will hopefully receive it early next week

EDIT: EPROM came today, the game is now fully working