How to diagnose dead Videl?

[dml]

Did I understand correctly that you read a continuous 3.8v on U11 pin #2 (XRESET?) on the bad Falcon but 5v on the good one?

If so, you can find every IC using XRESET and measure at each IC. You will need a precise reading so make sure the scope probes are on maximum impedance (select x10 mode - or use a digital voltmeter with 5 digits or more).

Look for the lowest reading and if you find one measurably lower than the rest, you could lift that pin and re-measure.

I think that's what the other guys are steering you towards - probably with a more methodical route - but I think just finding an IC with XRESET reading lower than everywhere else would be very interesting in itself.

[edit]

A10 also goes to U44 which is the socketed DSP GAL. That's one easy thing to eliminate on that net.

[mikro]

Did I understand correctly that you read a continuous 3.8v on U11 pin #2 (XRESET?) on the bad Falcon but 5v on the good one?

Correct.

If so, you can find every IC using XRESET and measure at each IC. You will need a precise reading so make sure the scope probes are on maximum impedance (select x10 mode - or use a digital voltmeter with 5 digits or more).

Look for the lowest reading and if you find one measurably lower than the rest, you could lift that pin and re-measure.

Interesting. How does this work? I mean, I measure on U11#2: 3.87V. EXP#17: 3.87V. CPU#64: 3.87V (as average / rms readings). So what I'm expected to see? Even *lower* reading on some IC?

A10 also goes to U44 which is the socketed DSP GAL. That's one easy thing to eliminate on that net.

I removed it and measured again, same resistance. I guess it makes sense to eliminate the low voltage problem first and then focus on this one. But I'll follow @exxos's recommendation of identifying the resistor pack / lifting the pins but I want to do it really as a last resort. Me , SMD ICs and lifting pins, that's really a hazardous game.

[Badwolf]

Sorry, been out of circulation for a day on this, so here's a melange of possibly out of date thoughts:-

D0 - D7: 9.5K
D8 - D15: 8.5K
Hmm. The good Falcon had all of those at about 10K. And when I got back to the bad Falcon, and specifically D8-D15 were also 10K. So perhaps this is OK (?)

I've just been doing this exact thing on my broken Falcon board and I hit exactly 10.0 on a number of occasions. 8.5 seems a bit low, so that *could* be an indication of a problem.

A1: 2.3K
A2 - A5: 2.8K
A6 - A7: 3.2K
A8 - A9: 2.8K
A10: 0.8K (!)
A11: 2.8K
A12 - A23: 3.2K
Again, when I went to check this with the good Falcon, I could see the values are supposed to be around 2.8K

Now that's odd. All my address lines were about 4.7k. Are there signs of rework on the address pull-up resistor packs? Can you read a figure on them? If your good Falcon is similar then perhaps there was a change of packs between revisions. My board is a rev A.

A10 looks utterly borked, though.

Then the ill-fated reset circuit. Its various red flags were hidden by all that /AS debugging. I re-measured everything again without ROM/RAM and for my final measurement I even used the same PSU because the level of my paranoia was raising after every measurement (all values are Uavg/Urms):

You've had AS asserting so the CPU is coming out of reset. 3.8V isn't correct and probably indicates a problem somewhere, but I think it may be symptom rather than cause.

U11#2 - good: 5.08V - bad: 3.83V (!)
...
As I replaced U11 a few months ago perhaps we can rule out the possibility that it's broken? Do you see something else in the schematics which could drag U11's pin 2 down so much?

Silly question, but I'm a silly man: are you completely sure the replacement is a 5V part?

Look for the lowest reading and if you find one measurably lower than the rest, you could lift that pin and re-measure.

Interesting. How does this work? I mean, I measure on U11#2: 3.87V. EXP#17: 3.87V. CPU#64: 3.87V (as average / rms readings). So what I'm expected to see? Even *lower* reading on some IC?

You need really good voltage resolution, but in theory: yes. With good enough resolution you would be able to pick up the voltage drop caused by the current flow in the traces. The lowest one sinking the most current.

My suspicion is that *might* work for the low voltage on A10 as it's so extreme, but I suspect you'd need something fancy like a ten digit bench meter to pick up the reset drop. I may be completely wrong on that and it's easy enough to do so why not? I'd be a legendary save if it works :D

The only other thing I'd add to all the above suggestions is take out everything that's socketed when you're doing any of these measurements. GALs, memory boards, ROM, codec (if yours is), PSU (if you're measuring resistances rather than voltages) etc. Even unplug floppies, keyboards, everything from the back.

Simplify the problem as much as is feasible.

BW

[mikro]

Trouble is that I don't trust my meter (or my skills when using it) too much, especially when measuring resistance. As mentioned, sometimes I'm seeing weirdly different values when re-measuring the same pins, I have no explanation for that.

However, voltages are a much safer domain and I remembered about @exxos's post here: viewtopic.php?p=29463#p29463. That definitely confirmed some of the bad things and if I were to make a guess, this must be related to either the pull-up resistors or some other, system-wide disaster ... still not sure whether Videl can play a role in this (measured with ROM/RAM out):

A1: 3,3V
A2-A5: 3,8V
A6-A7: 5,14V
A8-A9: 3,8V
A10: 1V (!)
A11: 3,8V
A12-A23: 5,14V

Data lines are even worse:

D0-D1: 5,14V
D2: jumping between 0V / 4V (with ROM out?!)
D3-D4: 5,14V
D5-D10: jumping between 0V / 4V (with ROM out?!)
D11-D12: 5,14V
D13: jumping between 0V / 4V (with ROM out?!)
D14: 5,14V
D15: jumping between 0V / 4V (with ROM out?!)

What the hell. Naturally, I checked the same pins in the good Falcon, all Ax/Dx pins 5+ volts.

@Badwolf the question about 5V IC was a good one, that's what I like, questioning even the most basic assumptions. Unless the store cheated on me, they were supposed to deliver SN74LS06D by TEXAS INSTRUMENTS with data sheet linked to https://www.ti.com/lit/ds/symlink/sn74ls06.pdf.

Any other idea what to check while I'm trying to find courage to desolder the resistor packs? ;-)

EDIT:

My suspicion is that *might* work for the low voltage on A10 as it's so extreme, but I suspect you'd need something fancy like a ten digit bench meter to pick up the reset drop. I may be completely wrong on that and it's easy enough to do so why not? I'd be a legendary save if it works :D

Do I understand correctly that you are suggesting to do the check not only on the XRESET/XHALT voltages but also on various chips with A10?

And don't worry, I have only the bare PCB here, not even FPU. The GALs are from a working Falcon, so that shouldn't cause trouble. And ROM/RAM is out even in the working Falcon, exactly for that reason: to minimise influence from other parts.

[dml]

Interesting. How does this work?
I mean, I measure on U11#2: 3.87V. EXP#17: 3.87V. CPU#64: 3.87V (as average / rms readings). So what I'm expected to see? Even *lower* reading on some IC?

Yes - the traces taking a signal between ICs are effectively low value resistors. The longer they are, the higher the resistance.

For a net to be brought down from 5v to 3.8v, some current is being sucked away by one of the ICs, through those traces. So there must be a voltage drop across the trace, where current is flowing.

Problem is, the drop will be very small because the resistance is small. Two decimal points won't be enough. A minimum of 5 digits and maybe more - or two probes in subtract mode on a good scope might be better than a basic DVM. Also needs good contact with the traces using sharp probes (to keep contact resistance lower than what you're trying to measure).

This could help avoid lifting pins one at a time - but at some point you will need to lift at least one to confirm anyway.

[edit]

(as BW said - chances are better with A10 because the drop is huge)

Oh - and measure from the *signal* side of the pullup resistor to each IC, not from the ICs to each other. The current is sourced from the resistor and sunk by the ICs.

You might hack things further in your favour by reducing the pullup resistance = more current sinking = bigger drop to measure.

[Badwolf]

@Badwolf the question about 5V IC was a good one, that's what I like, questioning even the most basic assumptions. Unless the store cheated on me, they were supposed to deliver SN74LS06D by TEXAS INSTRUMENTS with data sheet linked to https://www.ti.com/lit/ds/symlink/sn74ls06.pdf.

Any other idea what to check while I'm trying to find courage to desolder the resistor packs? ;-)

I missed Exxos pointing out the LS06 is an open collector. In that case I think it can be disregarded that as it's not actually a voltage source anyway.

BW

[mikro]

So, back to this one. Armed with new energy and multimeter (ironically, I bought this one in Beijing back in 2007 (!) and thought that I was never going to need it again since I have got the newer "more digital" one bought in Australia in 2017...).

I re-measured everything again. Started with the resistance against VCC. Previous results:

D0 - D7: 9.5K
D8 - D15: 8.5K

A1: 2.3K
A2 - A5: 2.8K
A6 - A7: 3.2K
A8 - A9: 2.8K
A10: 0.8K (!)
A11: 2.8K
A12 - A23: 3.2K

New results:
D0 - D15: 9.5K ~ 9.96K (should be 10K)

A1: 2.92K (should be 4.7K)
A10: 0.77K (should be 4.7K)
Ax: 4.63K

I checked also a few other signals: /AS, FC0-2, /BR, /RESET, /HALT, /BERR whether they match the pull-up resistors in the schematics. Except /AS and /BERR (which were 0.953 and 0.986 respectively while they should be 1.2K) I didn't notice anything out of ordinary.

Clearly, A1 and A10 are pulled down by some IC and especially A10 is by no doubt the best candidate for further checks.

I also checked voltage, just to decrease chance of error in my previous measurement:

A1: 3,3V
A2-A5: 3,8V
A6-A7: 5,14V
A8-A9: 3,8V
A10: 1V (!)
A11: 3,8V
A12-A23: 5,14V

D0-D1: 5,14V
D2: jumping between 0V / 4V (with ROM out?!)
D3-D4: 5,14V
D5-D10: jumping between 0V / 4V (with ROM out?!)
D11-D12: 5,14V
D13: jumping between 0V / 4V (with ROM out?!)
D14: 5,14V
D15: jumping between 0V / 4V (with ROM out?!)

New results (reset released / pushed):
A1: 3.66V / 3.36V
A2: 3.83V / 3.83V
A3-A5: 0 / 3.83V
A6-A7: 0 / 5V
A8-A9: 0 / 3.83V
A10: 0 / 1V
A11: 0 / 3.82V
A12-A23: 0 / 5V

Dx lines were basically identical, I just couldn't replicate the jumping part, i.e. the signals were either 5V or 0V. And as we can see, A1, A2 and A10 are our main suspects here.

So... taking Badwolf's advice:

My suspicion is that *might* work for the low voltage on A10 as it's so extreme, but I suspect you'd need something fancy like a ten digit bench meter to pick up the reset drop. I may be completely wrong on that and it's easy enough to do so why not? I'd be a legendary save if it works :D

I decided to measure it just by my scope, for starters at least. So what is connected to A10?

EXP 39
CART 26
EPROM 35
VIDEL 31
COMBEL 16
SDMA 18
CPU 48
U44 10

The first three are not important / present. So that leaves Videl, Combel, SDMA, CPU and U44. I checked that all of them are getting +5V on their VCC pins and measured resistance between the A10 and GND/VCC and the actual voltage on the A10 pin.

I also tried to remove U44 just because it's easy to do so but no change on A10 unfortunately.

The resistances were nearly the same (0.702 for A10/GND and 0.800 for A10/VCC) but there was a hope! VCC at the A10 were quite similar at all chips (1.03V - 1.08V) except ............... the Videl! (0.985V)

It took some courage but since A10 on Videl is pin 31 which is in the top left corner I was like OK, let's do it. I lifted the A10 leg and .... we're getting 4.6K on EXP between VCC and A10 ladies and gentlemen!

I didn't do anything else (too late) but I guess it doesn't make sense to continue before I ("I" ;-)) replace the Videl. Of course, I'll definitely try to boot the diagnostic cart without Videl, to see how it behaves.

[dml]

\o/ woohoo

[mikro]

Update time!

It took some convincing but @jookie agreed to help me out with the Videl. So I paid him a visit this morning and in less than two hours he was able to do the job:

jookie.jpg

Pretty interesting experience to observe his techniques, from cutting the IC out to fine-soldering every leg in a pretty efficient manner. We verified that DE is active (15.6 kHz) and /RESET is reaching full 5V. Jookie also noticed that pin 14 (EXTCLK) seems to have (deliberately?) severed connection to the actual pin on the monitor connector, so he created a solder bridge there. Despite my expectations, Jookie's room was anything but ready for further debugging, so I decided to thank him for his work and test everything in my office afterwards.

The good news: all pull-up and voltage measurements have been fixed. /AS and /BERR are still 0.953K and 0.986K (should be 1.2K), hard to say how significant this is. I have grabbed 8 words on the scope, I can confirm that I see $602E, $0404, $00E0, $0030 , $46FC, $2700, $3038, $8006.

Cool! Now the bad news: /ROM3 & /ROM4 is still HIGH. And what's worse, it seems like /RESET isn't triggered at all. I'm a little bit suspicious about the data signals, though. This is how a typical trace looks like (yellow - data, blue - /AS):

weird_spike.jpg

Notice how weird the 4th edge looks like (it is supposed to be zero). Even if there are all zeroes:

all_zeros.jpg

it seems to be aiming to go high and then dragged back to zero. Maybe significant, maybe not. However. When I tried to see the /RESET (yellow):

reset_missing.jpg

We can see that there's really nothing happening. The red "R" marks the RESET instruction. My assumption would be that at most four data fetches later we should see its effect (it's the 9th word in the series, i.e. just after the verified 8 words I saw) and we see about ~2us gap because of ... what?

RESET takes 518 cycles * (1/8000000) = 0.00006475s = 64.75us, i.e. way, way longer. On the other hand, an exception takes about 20 cycles = 2.5us so ... maybe? And maybe not, 20 cycles can be any CPU instruction accessing memory or so. Then, when I set the trigger to /RESET:

full_reset.jpg

We can see that CPU lives its life (no idea about that 10us gap) but definitely nobody is resetting anything. I measured /RESET directly on CPU pin 64, also nothing.

Jookie says that he would swap the CPU right away. I guess I'll try to see whether CPU makes it really to the cartridge checking code (without any /BERR or /HALT) and if so, I can try two things:

- lift CPU pin 64 and see whether there's something happening (however this will disable the power up reset, hmm..)

- mount the Mighty Sonic 32 back, which has its own 68030 CPU (one has to lift pin 16 & 17 then)

Any ideas or theories welcome. ;)

[dml]

>/AS and /BERR are still 0.953R and 0.986R (should be 1.2K)

Do you mean 0.95K? or 0.95R? I will assume that's a typo and it's meant to be K :)

So if I followed that news correctly, swapping the VIDEL has fixed the shorted address line but the machine is still dead - and XRESET seems to be not responding to the CPU when it executes the reset opcode.