PDA

View Full Version : IJ Matrix fun and games



AskJacob
7th June 2009, 10:22 PM
Hi Guys,

Thought I'd log this, man what an interesting fault this turned out to be!

Symptoms: on startup, GND SHORT ROW1 error

Fix:

well well, this wouldn't be interesting if it was JUST U20 - the ULN2803. I managed to convince myself it was not the right area, as the fault changed when the row connector was removed from the CPU board.

To make matters more interesting, there were also a little more to the errors, the diagnostics only mention row1 but row2 was also shorted too.

Combine that with some other matrix oddities, which was hard to tell if it was related to the row1/2 short, and I was really confused on where the problem really was.

We spent a very long time (hours) going over the wiring, following from switch to switch along the row wires, looking for damaged wire, shorts and anything else.

Every time we were certain we had isolated the wire run "this one is definately it" the bloody short would up and move on us.

Tiredness and frustration settled in, and a bit of sanity - I decided to take the CPU board home and try it in my IJ.

Took it home, and fair enough, the CPU reported GND SHORT ROW1. Buuut, only row 1, row 2 was ok.

OK, time to have a close look at the CPU. I could see the ULN28003 had been replaced before (without a socket :realmad:) and both LM339's had also been replaced, using some (seemed crappy) sockets.

I removed the old ULN28003, put in a machine socket and put in a new one.

Fired it back up, passed the startup test. Woohoo, look at the switch test, and it looks good too.

Headed off for a coffee, thinking OK that was it, and then though I'd better go back into the switch test, and confirm the whole matrix is behaving...

New Symptom: Column 2 and 3 all out, not responding

Fix:

Dammit, what happened? OK, spent some time probing around the other sockets (for U18 and U19 that were already there, as the work looked a little shonky). I could not find any connectivity faults, so pulled the LM339s to swap them over, and saw the bloody cheap socket had 3 sockets that were "worn or bent" so that 3 of the IC pins would not connect properly.

I removed and replaced both sockets. I also replaced 1 of the LM339's at U18 as it was still acting rather odd.

Put it back in, and passed all tests with flying colours.

At this stage, thinking it was beat, I upgraded the CPU code from L6 to L7. The sound was already up to date on L3.

Took the CPU back today, thinking a quick plug, handshake and she's all good.

Haha no...

New Symptom: GND SHORT ROW1 (and ROW2)

Fix:

aaaaaaaaargh what the hell? there DEFINATELY was a problem with the CPU board, and it has been fixed, so what is going on here? My first worry was that some playfield fault was still there, and had undone all the work.

Unplugging the row conenctor cleared the fault this time, so phew the CPU was still good.

Now the next thing to look at again was the opto boards. The row 1 and 2 wires diasy chain all aver the place, but DO end up at all the opto boards on the machine - The 10 switch opto pcb, the 3 switch opto PCB and the 7 switch trough PCB.

Disconnect all, the fault clears.

Reconnect the 10 switch, and the fault returns.

Disconnect all, the fault clears.

Reconnect the 3 switch, no fault

Reconnect the 7 switch trough, fault returns.

OK, looks like the 10 switch and trough pcbs are causing problems... Makes sense if 20 or 50v belted through the switch matrix at some point, all of these boards (and the CPU) have ICs connected to these two wires...

Luckily we had access to a spare trough PCB, and swapping them cleared the fault (as long as the 10 switch pcb was disconnected). Nice, we are nearly there.

I pulled the 10 switch opto pcb, a quick check (too quick) of the schematic showed Row 1 and 2 feed to U1, an LM339. Ok, a quick meter check shows the outputs are LOW and unchanging (pin 13 was 0.6v, pin 14 was 2.6v - compared to around 6.7v for the other outputs which is the multimeter's average of the pulsed 12v).

Clip of the old IC, install a socket, pop in a new LM339, put it all back together, fire it up and...

New Symptom: GND SHORT ROW2

Fix:

gah! I now perform a FULL check of the schematic for the 10 sw opto, and guess what, the row 1 and 2 connector ALSO goes to U3. D'oh!

I remove, socket and replace U3 (another LM339).

Test, and now for a real WooHoo moment.

All working!

This was a real test! multiple levels of failure at multiple locations. I am so glad we nailed it, it was a very satisfying fix indeed.

My suspicion is that something blasted 20 or 50v on either the row 1 or 2 wiring on the playfield. During all the switch checking we didn't find any culprits like pinched, stripped wires or mashed switches, so I don't know what caused it, but the damage was pretty extensive, almost every IC that was connected to those row wires was toast, except on the 3 switch opto (must have been a tough little lm339 there).

Fortunately now those parts are now socketed with nice quality sockets, which will make any future trouble shooting a lot easier.

Cheers
Jacob

Skybeaux
8th June 2009, 07:59 AM
What an annoying chain of faults , i'm glad you figured it out:cool:

Would have had me pulling a few hairs for sure!

Beast
8th June 2009, 08:35 AM
Good stuff Jacob.....

I would have needed something stronger than coffee!!

David_AVD
8th June 2009, 09:36 AM
Well done AJ. Nice to fix a complex fault. Still annoying that you didn't find the "smoking gun" that caused it all in the first place I bet.

AskJacob
8th June 2009, 09:38 AM
Well done AJ. Nice to fix a complex fault. Still annoying that you didn't find the "smoking gun" that caused it all in the first place I bet.

mmm I have my suspicions ;)

Arcade King
8th June 2009, 10:06 AM
Glad to see you got it sorted mate.
Thanks for sharing your log :)