Trying to fix Mikro's wrecked TT board

[dml]

Haaaaaaa!

Bonus remote debugging points awarded someone's recent post. :p :p :p

[dml]

Ok long story short - I was busy trying to narrow the short with the DMM by looking at one side of all the decoupling caps and I thought I found an interesting one at 12R instead of 28.

Poking around I found other values at 33R. But then noticed all the places where I had measured 28R were now 33R.

Fiddled with the DMM leads, made a better tie to the power connector with some contact cleaner, tried again - same 33R. Something has changed.

Moved the board around a bit and back down to 27R.

Hmm.

So I put the DMM across the GND/5V rails at 27R and took a grinding pen to the broken corner of the board. Initially nothing, then it drops to 5 ohms.... and then back up to 33... and then 0.

A bit more of this and it is back up to 33. So there is definitely a significant fault at that corner. It might be the whole problem or it might be interrupting some other problem I'm measuring but either way it's still bad so I'll need to deal with it before anything else.

(bonus points -> BW, with frank.lucas in second place with the multilayer reminder)

The short is somewhere in this inner sandwich. I have cleared it down to 0.08R and it is very touchy - I'll need to get the microscope on it but that's a bit tricky since the board needs to be end-on to the lens.

Screenshot 2025-04-08 at 13.31.03.png (2.19 MiB) Viewed 89 times

[Badwolf]

Ooh, yeah. that doesn't look pretty. Look how wiggly up and down the internal layers are. Could easily touch on any of those direction changes.

I was going to suggest you may be better off chamfering it and a really shallow angle: instead of a 90 degree cut that has 5V and GND constantly within one inner layer thickness of each other vertically, try to grind it down at, say, a 30 degree angle so, in theory at least, you'd introduce some horizonal separation between them.

But looking at that, I'm not sure it'll help and may make things worse.

At least it'd be easiler to scope!

Perhaps you could try to grind away one layer at a time and stop at the first power plane? Would only need to step back four or five mills from the edge, you'd think.

Note for self: don't cut through power planes!

BW

[mikro]

the PSU Jookie refers to (MeanWell RQ-65B) is 5, 15, -5, -15.

Is it? From their website: AC-DC Quad output enclosed power supply; Output 5Vdc at 8A +12Vdc at 3A -5Vdc at 1A -12Vdc at 1A.

@dml I wonder how could I get the machine running then? I must have been very lucky, at one point I even saw a white screen, i.e. the TT nearly booted.

[dml]

I was going to suggest you may be better off chamfering it and a really shallow angle: instead of a 90 degree cut that has 5V and GND constantly within one inner layer thickness of each other vertically, try to grind it down at, say, a 30 degree angle so, in theory at least, you'd introduce some horizonal separation between them.

I was thinking exactly the same and have been looking for the dremel... it might not help but I'm thinking it won't necessarily be worse than it already is.

Also getting a little concerned that I was able to nudge the short between 0.08R and 33R but not higher than that... suspiciously capped there -_- maybe there are more shorts waiting....

[dml]

Grinding the broken edge at an angle did help stabilise things a bit and makes the inner planes easier to see. But I still have a low resistance of about 35R across +5/GND everywhere. It won't go higher than that while messing with this cut. It can drop lower but won't go above 35R.

Screenshot 2025-04-08 at 14.40.58.png (1.54 MiB) Viewed 77 times

I'll continue with it a bit more but I think this is not the only problem.

[edit]

Got the edge ground fairly clean and at a shallow angle. The wobbly internal planes look straight. I don't see any shorts there.

Time to look elsewhere!

[dml]

@dml I wonder how could I get the machine running then? I must have been very lucky, at one point I even saw a white screen, i.e. the TT nearly booted.

I *think* some of the other voltages are needed for audio, comms/ports and parts of video output only. It may partially work without all of the voltages. I'm not sure about that for TT yet but it is true for some other machines.

[Badwolf]

Grinding the broken edge at an angle did help stabilise things a bit and makes the inner planes easier to see. But I still have a low resistance of about 35R across +5/GND everywhere. It won't go higher than that while messing with this cut. It can drop lower but won't go above 35R.

Well done!

Is the 35R the same in both polarities?

BW

[dml]

Is the 35R the same in both polarities?

Suspiciously yes-but-no. It's 35R in one direction, 31R in the other direction (for +5V/GND). So there's a semiconductor involved somewhere.

The -5V rail reads open.

Already tempted to try 5V with infrared camera. Either the 35R is a dead IC or it's an apparent resistance for the ohmmeter across a bunch of junctions.

Unless I think of some other way to narrow it down, that's probably next.

[Badwolf]

Is the 35R the same in both polarities?

Suspiciously yes-but-no. It's 35R in one direction, 31R in the other direction (for +5V/GND). So there's a semiconductor involved somewhere.

I think that's alright. There will be chips and bits and pieces across the rail with diode-like behaviour in there, so you probably want it to have a little bit of an apparent resistance in one direction and be open in the other. This sounds like you've got a fixed low impedance path then the normal lower resistance in the other direction (this is assuming the lower value is with positive on the 5V and negative on GND, of course!)

Already tempted to try 5V with infrared camera. Either the 35R is a dead IC or it's an apparent resistance for the ohmmeter across a bunch of junctions.
Unless I think of some other way to narrow it down, that's probably next.

Less than a watt of extra dissipation? I'm not sure you'd pick it up and it runs a risk.

I think I heard you mention a seven digit meter. Inject 1.5V in one corner and measure values across the board? All those decoupling caps should actually help you measure local variations stably, I'd have thought.

BW