How to diagnose dead Videl?

[mikro]

I'm really getting senile. While investigating this oscillating RESET signal, it occurred to me that I vaguely remember doing the same in the past. And thanks to Messager messages with Jookie I could verify that indeed, I had the same problem months ago, in this very Falcon.

Back then I found at least one offender -- the R2 (by U2, on the back side) was showing 100+ Ohms while it should be 10K. I even verified it on the second Falcon that it indeed shows such a value. So I ended up replacing R2, U1, U11, all the caps there... and told Jookie that the reset signal is fine now.

... which was unlikely (still a noob when it comes to the scope...) because imagine my surprise when I measured R2 and it was again 100+ Ohms. Fortunately, I had a few spares already so I replaced it again and *now* imagine my surprise, when it was again 100+ Ohms !!!

I didn't even switch the computer on, I literally just measured the resistor before soldering - 10K Ohms, soldered it in - 100+ Ohms. Considering the simplicity of that circuit:

r2.png (73.14 KiB) Viewed 87 times

I just don't get it. Could the pads be somehow shorted?

[dml]

> I just don't get it. Could the pads be somehow shorted?

Well in these machines (Ataris generally) there are usually a couple 100 ohms between VCC-GND, or less in some cases (remember my fun with the SCSI termination resistors in the TT?). So the 100k can become 100R if the low side of the resistor is at GND.

Which would make U2 or the reset switch suspect.

It's late for me to go poking around in Ataris but I can check tomorrow and see what it measures on mine.

...

Just re-absorbed that a bit more carefully.

Has there been rework near that area before? It suggests there is a problem either with U2 or the PCB itself - a low resistance path from U2 pin 1/4 to VCC, which would be bad. It might depend on what kind of inputs U2 has but if you have compared already with another machine and saw different results, it is suspect.

The RES switch toggles GND between those two inverters in U2, which looks like a debounce thing. The 555 does the delay timer. If you're seeing 100R across R2, U2 might be damaged or some shorting in the PCB layers in that area, so the reset circuit might not be working properly. But you can confirm what the reset circuit output looks like by checking the state of C7 (which charges over the reset delay period) and the output of U1 to U11. If that's switching after a short delay and XRESET from U11 is going high as a result, the bigger problem is not the reset circuit. But the 100R reading is very suspect.

[mikro]

Sorry, I have just noticed your answer...

I had a long-ish reply prepared but then I went to re-measure some things again to give the answer some fact-checking and I discovered some new things.

The R2 is still 100+ Ohms, that hasn't changed. Also, if I may ask you, could you rephrase this:

But you can confirm what the reset circuit output looks like by checking the state of C7 (which charges over the reset delay period) and the output of U1 to U11. If that's switching after a short delay and XRESET from U11 is going high as a result, the bigger problem is not the reset circuit. But the 100R reading is very suspect.

into a step-by-step guide for dummies? First I don't know how to "check the state of C7". Second, I'm not sure whether you mean comparing C7 vs U1-U11 (i.e. U1's = LM555's OUT pin?) or comparing C7 vs XRESET and U1 vs XRESET and what exactly am I even comparing.

As you say, the U2 is definitely a good candidate to replace and maybe I'll just do that although I really hate working with those SMD parts (as mentioned, I already replaced U1 and U22 = hell on earth but since the behaviour hadn't changed, I'm hoping that I didn't do more damage there).

I tried to do some measurements against the GND plate on both Falcons: the reset switch's pin 3 for example shows 0 Ohms on the working one and 10K on the broken one. Hm. When measuring resistance between reset switch's pins 1 and 2, it gave me 10k on the working one and exactly 1/10th = 1k on the broken one. Hmm. Pin 2 = GND so now the measurements are reversed. But I'm unable to draw any conclusions from that (too stupid). Other measurements were either inconclusive or the same on both machines.

But, my new discovery. Unsurprisingly, I didn't interpret / set the scope correctly it seems. I could swear that I saw a series of oscillating voltage levels on the reset signal but I couldn't reproduce it today at all, just a couple of them. However I did notice something else.

My test case: Videl's pin 1 (XRESET), holding the reset button, the trigger set to the raising edge around 4V, scale set to 200us (this seems to also influence how much I'm able to record? <----- total scope newbie), releasing the reset button.

On the working Falcon:
- 0V until the treshold
- 5V for a very short time
- 0V for about 1100us
- 5V for a very short time
- 0V for about 60us
- 5V for the rest

On the non-working Falcon:
- 0V until the treshold
- 5V for the rest

Perhaps what confused me yesterday was that if I repeat this test with 1us scale, I get different results again:

On the working Falcon:
- 0V until the treshold
- 5V for about 1.5us
- 0V for the rest, perhaps running out of storage space for showing the other peaks

On the non-working Falcon:
- 0V until the treshold
- 5V for a very, very brief moment (like 0,3us?)
- 0V for about 1us
- 5V for about 0.75us
- 0V for about 1us
- 5V for the rest

Wow. What is even weirder that sometimes, and I'm unable to pinpoint it, I could see three of those 5V peaks in series on the non-working machine.

[mikro]

I have a hard time interpret those 5V levels, frankly, on both machines.

AFAIK, there's a RESET instruction quite early in the TOS code. So if the trigger waits for the raising edge (basically going *from* the reset state as 5V = non-reset state), I'd expect to see:

- 5V for releasing the reset button
- a short 5V line before RESET instruction
- 0V for RESET instruction; as RESET takes about 518 cycles, 518 * (1/16 MHz) = ~32us? (EDIT: actually it's 8 MHz yet, so yes, this must be the 60ns phase...)
- 5V for the rest

So I'm not sure where the 1100us is coming from? And that's the WORKING Falcon.

Then there's the other thing, the NON-WORKING Falcon seems to shortly oscillate (two or three peaks) before switch to 5V and doing nothing else. So that would imply that something is wrong with the reset circuit but I don't know, there were at least two 1us reset pulses, that isn't enough to reset the circuits, even if TOS ROM and its RESET instruction wasn't reached?

[dml]

The normal expectation is that XRESET (the final reset signal which actually goes direct to the chips) should be LOW from power on - there should be a short delay (I'm not sure exactly of the delay but it is 100s of milliseconds to perhaps up to 1 second or so) and then it should go HIGH.

There should not be lots of transitions taking place, just one. At least, not anywhere near the timeout. Maybe some noise at the start on an old board but that's all. The purpose is to keep the ICs alseep until the power supply is up to speed then wake them all together.

It is possible that you will see many spikes on the switch side of the reset circuit, but should not be on the XRESET line itself which goes to the ICs. It is not supposed to go HIGH before the ICs are all up to power.


If you have the scope set in DC mode, positioned near the bottom of the trace, with the probes on 10x and volts-per-division at 0.1v, you should be able to see it move from LOW to HIGH a short delay after releasing RESET.

The circuit responsible for the delay is the 555 timer chip, with a reservoir capacitor, I think C7 (it's a small electrolytic near the back of the board under the PSU). This cap's charge should change slowly (100s of ms) in a curve when you release reset, and the 555's output should change when that cycle completes.

There is also a debouncing thing before the 555 involving a couple of back-to-back inverters, so when you press/release the reset button it starts the timing cycle from a single pulse. That part doesn't matter so long as you see XRESET switching a short time after you release the button.

Since it is possible for a bad IC to hold XRESET low, its worth checking the level of this signal at each IC to make sure they all read the same, near 5v. If any read significantly lower, it may be pulling the signal down at that location.

[dml]

> AFAIK, there's a RESET instruction quite early in the TOS code.

That's a good point - the CPU can assert RESET on the same net.

However this would not affect anything upstream of that net (i.e. the reset timer itself) so you could confirm that.

In other words, check that the reset circuit makes a single output transition LOW-HIGH, then separately confirm the XRESET signal shared by the ICs is toggling an extra time, after the ICs are awake. That would tie with what you are seeing.


[edit]

I think generally, if you find the XRESET signal is HIGH for all the ICs using it, after some delay (ignoring spikes and possible self-induced resets) then the system is not being held in RESET and you can probably look elsewhere.

If you are seeing 100R across a 10K resistor, with the machine cold/off, it could point at an issue with U1 but not necessarily (semiconductors make reading resistances etc. in-circuit difficult). If it reads differently on a working Falcon then it should be suspected though.

[mikro]

@dml yes, that's how I understand the RESET signal levels as well. Basically it seems that my working (!) machine has a small peak of 5V before doing the "proper" (1100us) reset (and then everything follows as it should) while my non-working machine just goes 0-5-0-5-0-5 and gives up, never waits for that time period. So it seems that someone is confusing the 555 to do its job and HOPEFULLY the fact that the XRESET is held 0V for such a short time is making the Videl not waking up.

When you say "This cap's charge should change slowly (100s of ms) in a curve", you literally mean connecting the probe to C7's pin and watch how quickly it triggered on +5V?

[dml]

When you say "This cap's charge should change slowly (100s of ms) in a curve", you literally mean connecting the probe to C7's pin and watch how quickly it triggered on +5V?

Yes, you should be able to capture the charging cycle on your scope and also watch the output pin of the 555, which should switch once. I'm pretty sure it is C7 - could be C3 but I think that is a ceramic. I'll find a board and have a look.

Yep, C7, just below U1.

[mikro]

Indeed, your memory serves you well. I replaced this one along with R2, U1 and U11 (and a couple of other caps) so it should be working as you describe.

Perhaps I'll try what you suggests (i.e. observe 555's output vs C7 charging) and report back. There can be basically two results:

- C7 is slowly charged but 555 doesn't give a s**t and does the two or three peaks independently from that

- C7 is for some reason badly charged and that makes the 555 confused

However as you say, if the R2 is giving wrong values... there's hardly someone else to blame? (except some catastrophic scenario where PCB gets damaged...)

[dml]

I'm looking at the schematic just now and the reset button/debounce thing, 555 timer are completely isolated from the XRESET network, via U11 (74LS06 logic). The 555 feeds two inputs to that IC (pin1, pin3 - both inputs to separate buffers) and should not respond to any additional reset asserted by the CPU on the XRESET line, which emerges from pin 2 of U11 (output from one of the buffers).

So whatever the main ICs might do with XRESET, they can't feed signals back to the 555 before U11 - unless U11 is faulty.

In other words - the 555 OUT pin should switch one time only (I think the 555 goes from HIGH to LOW, but the U11 buffers invert that from LOW to HIGH)