Debugging PAL Atari 800XL with scope

[Dr Memory]

I'm trying to fix a PAL 800XL with the black screen of death.  Sometimes the red screen instead.  I've done the vast majority of things I've seen described in the various threads and youtube videos:

 

1. Tried all of the chips in other machines - all worked

2. Tested all the memory in my little Arduino DRAM tester - all good

3. Swapped in chips from other machines to see if it would change anything - no effect, aside from some wierdness with putting NTSC/PAL GTIA/ANTIC in the opposite kind of machine

4. Tried different power supplies and monitor cables, that are known good because they work on other 800XLs - no effect

 

So I've been using various test instruments to dig in deeper.  The problem is, I fear my skills and knowledge just aren't at the level required, and the documentation I'm using is aimed more at NTSC systems than PAL.  For example, the Sams book describes an NTSC system only.

 

I picked up a real Frequency Counter on ebay.  Low-end and used but not one of the Chinese kits.  I've got one of those too but it has limitations that annoyed me, so I sprung for a Simpson 710, which is actually one recommended by Sams (coincidence).  When I use it on the Y1 crystal, I get 3.546 MHz, which matches what it says on the can.  When I use it on the Y2 crystal, I get 886 KHz, which isn't right.  I don't get how that can be - never heard of that kind of failure mode for a crystal, but then again, I'm no expert, that's for sure.

 

So I dug in deeper.  With an oscilloscope I poked around and checked out the various test points in the Sams book.  Generally I see frequencies of 1.773 MHz or so, which doesn't match either crystal.  The Sams schematics give this in terms of delay.  I see delays around 564 nanosec, where I think I should be seeing delays of 0.2 microsec.  So .564 instead of .2, which is wrong.  The waveforms look plausible but the frequency isn't right.  Something seems wrong with time, so I went back to the crystals.

 

This is the really weird part and what finally drove me to post.  When I look at what's going on with Y1, I see ~3.546 MHz on either pin, measured against the ground plane.  When I look at what's on Y2, I see ~886 KHz on one pin and 4.386 MHz on the other.

 

Any of you experts have any insight into what's going on here?  I don't understand why or even how a crystal oscillator can have different frequency measurements on the two pins.  I've been reading up on them but don't get it yet.

 

Also, why would my delays and frequencies be so far off of what is shown on the schematic in the Atari 800XL Sams Computerfacts Technical Service book?  That's the only place I've found that shows expected voltages and waveforms so I've been leaning on it heavily.  But it assumes someone a lot more knowledgeable/experienced than me really.

 

Any help or advice would be welcome.  I just don't seem to have the usual problems with the RAM or OS or BASIC or one of the co-processor chips going bad.  Something odd here and I'd like to understand it, and perhaps eventually fix it.  I've seen several of the elder statesmen mention that if they had a scope they'd go deeper and do things differently, so I'm hoping for some info on just what that means. 

 

Edited February 27, 2021 by Dr Memory
fixed minor typos and spelling

[Rybags]

I doubt your problem is related to the crystals.

~ 4.3 MHz corresponds to the colourburst generation which gets fed to PAL GTIA and isn't used anywhere else.

 

Screen that stays black or occasionally brown or red means the OS isn't starting at all.

Other parts you should consider if not yet tried in another machine are the delay line, MMU and PIA.  Note that the MMU isn't compatible with 1200XL or XEGS.

 

Of course all that aside there's always the chance it's a passive component or the motherboard such as severed trace or solder joint gone bad.

 

[Dr Memory]

The MMU and PIA were two of the many chips I swapped.  Both ways - I tried them in a working machine and it still worked, and tried the ones from that machine in this one and it still didn't.  The only one I haven't been able to try is the wierd delay line thing because I don't seem to have a machine with that socketed.  I've been using a good 800XL for the secondary guinea pig, so no worries on the 1200XL/XEGS front.

 

I agree that it could be some sort of physical damage but I haven't spotted it any issues with a jeweler's loupe.  As for a bad component, maybe?  But my life would be better if I had a clue on where to look.

 

The Y2 crystal still looks suspicious to me.  I'm hoping someone will know what it means to have the frequency so far off as sell as different when measured on the individual pins, or at least tell me if these measurements are normal.  I have no other PAL systems to compare.

[Rybags]

The delay line is used to generate the partial cycle delay important for Ras/Cas in memory accesses.  So a good idea to verify it's operating within specs.

Although... if you get that static background colour that doesn't go black it's indicating the OS isn't even getting a start.

I'm not sure if it can go bad, but maybe the power on/reset circuit?

Also the 74LS138 at U2 - used for memory selection in the IO space.

Edited February 27, 2021 by Rybags

[Mazzspeed]

Bear in mind that if you're still using the original sockets for the IC's, they weren't the best in the 80s and the sockets themselves could very well be the fault.

 

I'd start by removing all the original sockets and replacing them with modern quality components.

[+DrVenkman]

6 hours ago, Mazzspeed said:

Bear in mind that if you're still using the original sockets for the IC's, they weren't the best in the 80s and the sockets themselves could very well be the fault.

 

I'd start by removing all the original sockets and replacing them with modern quality components.

I agree it’s worth looking closely at the sockets for corrosion, bent or broken socket wipes or debris but I strongly disagree with the “just replace them all” approach. That’s a lot of work to replace 5 x 40-pin sockets, plus all the sockets for OS, BASIC, MMU, DRAMs and logic chips. That kind of shotgunning is also intellectually lazy - there are better ways to trace out problems than spending a few hours replacing a bunch components, most of which are certainly fine. Further, that much work if not done by an experienced tech is likely to result in lifted traces or pads on the board.

 

My advice to the OP: start with verifying power and ground on each of the main Atari IC’s, then doing the same for all the DRAM’s. With luck, you’ll find something wrong and can start focusing on that socket. If the socket itself is fine, use the schematics and traces on the board to see where each signal goes and verify continuity to each electrically-adjacent part. Repeat until the problem is ID’d.

 

If all chips have good power and ground, check for the Phi2 clock signal on each main chip using a logic probe or scope - the pinouts in SAM’s or the Atari Field Service Manual will tell you which pin to check. If those are good, then you can know that Phi0 and the related Fast Phi0 and OSC signals between the ANTIC, GTIA and the main crystal are good. If Phi2 is not present, then check for Phi0 input on SALLY and output of ANTIC, and Fast Phi0 input on ANTIC and output on GTIA. While these chips may all be fine physically, if you have a socket or trace problem, checking these signals will help ID it.

 

If all those clock and timing signals are good, start checking for activity on the address and data bus lines for each chip, reset signals, etc. 

 

Yes, this can require meticulous and careful work, making yourself checklists to be sure you’re checking everything but it WILL eventually find your problem rather than simply doing a bunch of physical work on the board and hopping for the best. 

[Dr Memory]

Thanks DrVenkman!  That is very helpful.  I'll probably spend most of the day doing that after I get some caffeine in me.    That's just the sort of advice I was hoping for - something concrete to try that could help me understand stuff.

 

I share your views on replacing all of the sockets, and I can assure you that it would be risky.  I MIGHT be able to do it but it would be ultra-stressful and it is likely I'd make mistakes.  Similar views on things like "replace all the caps" and such.

 

Rybags, the 74LS138 at U2 got the chip swap test too and seemed fine - didn't break other machine and good one from other machine didn't cause me to start getting video.  I'll focus in on the delay line chip - even if I can't swap one in I can certainly measure things.

 

I've attached a couple of the scope pics I've managed to collect.  The procedure isn't great - I'm not all that pleased with Ultra Sigma and UltraScope - but with effort you can get pretty good images.

 

This is what the signal looks like on the nice side of Y2.  It looks pretty clean to me.  I haven't been able to coax UltraScope into capturing the Measurement region along with the data area, but this shows a nice clean 225.4 ns sine wave with a mean p2p of 770.8 mV.  So that's 4.386 MHz, which seems plausible.

 

This is what the signal looks like on the other side of Y2.  I'm really not sure what I'm seeing here.  The other crystal doesn't have a side that looks like this.  Measurement indicates that it's a 1.28 us (886.8 KHz) signal at 3.504V.  I'm not experienced enough with electronics to understand what this means.

 

[+DrVenkman]

1 hour ago, Dr Memory said:

his is what the signal looks like on the other side of Y2.  I'm really not sure what I'm seeing here. 

That's a regular square wave pulse but the oscillations you're seeing at the top and bottom of the pulses are generally described as "ground bounce" - usually it means that you need a better, closer location for the ground connections during your measurement. Also, if your scope has a 1X / 10X switch, be sure to set it to 1X - cranking up the sensitivity to 10X is useful for very weak, low-amplitude signals, but not much on vintage electronics needs that kind of sensitivity, and it can show artifacts like this which aren't real. 

 

Anyway, good luck with your repair efforts. Here's a thread of mine from several years ago that might be helpful or at least prove somewhat inspirational. It started as trying to fix one of my 1200XLs, and then turned into repairing a second, much worse-off 1200XL. I basically figured my through all that you're going through now with your 800XL, but armed only with a multi-meter and logic probe, plus a copy of SAM's for the 800XL (using it by analogy). It took a lot of patience but I ultimately identified several bad sockets and, once I got it to boot, a failing GTIA. But the machine is still my daily-driver 1200XL today, and I've since added an Ultimate 1MB and UAV video mod. 

 

So, patience and meticulous step-by-step work is key rather than some scattershot approach.

[Mazzspeed]

6 hours ago, DrVenkman said:

I agree it’s worth looking closely at the sockets for corrosion, bent or broken socket wipes or debris but I strongly disagree with the “just replace them all” approach. That’s a lot of work to replace 5 x 40-pin sockets, plus all the sockets for OS, BASIC, MMU, DRAMs and logic chips. That kind of shotgunning is also intellectually lazy - there are better ways to trace out problems than spending a few hours replacing a bunch components, most of which are certainly fine. Further, that much work if not done by an experienced tech is likely to result in lifted traces or pads on the board.

It's not so much about lazy diagnosis, it's about replacing OEM poor quality, ageing components. All PCB's of the era have a tendency to lift traces; Be sure to touch up all solder joints with fresh solder first, get the temperature of the iron right, use proper vacuum desoldering equipment and flux, take your time and don't be ham fisted and you'll be fine.

 

 

Edited February 27, 2021 by Mazzspeed

[+DrVenkman]

5 minutes ago, Mazzspeed said:

It's not so much about lazy diagnosis, it's about replacing OEM poor quality, ageing components.

You're suggesting an intellectually-lazy procedure to diagnose a problem that is in all likelihood very simple, and usually has ZERO to do with the actual cause of boot failure on an Atari. Period. Just "replace everything" is terrible practice, especially if someone wants to actually learn what specifically is causing the boot failure and fix it. 

[Mazzspeed]

1 minute ago, DrVenkman said:

You're suggesting an intellectually-lazy procedure to diagnose a problem that is in all likelihood very simple, and usually has ZERO to do with the actual cause of boot failure on an Atari. Period. Just "replace everything" is terrible practice, especially if someone wants to actually learn what specifically is causing the boot failure and fix it. 

No I'm not. I mentioned that the sockets could be the issue and I suggested replacing ageing components as a matter of course, that's all.

 

If you took my comments in a different context I'm sorry you got them wrong.

[+DrVenkman]

1 minute ago, Mazzspeed said:

I mentioned that the sockets could be the issue and I suggested replacing ageing components as a matter of course, that's all.

See, you're right with the part I have quoted. COULD be. But it's far more common on an Atari to be something else. But shotgun-replacing literally a few hundred IC socket pins without actually determining the fault FIRST is bad practice. It's a lot of work, without the proper tools and experience it can be dangerous, and even if somewhere along those several hundred desoldering/re-soldering steps you've corrected the problem, you still end up with no idea - zero, none, zilch! - which component was at issue! ONE bad wipe on ONE socket might prevent the machine from booting. That might be on a single 16-pin DRAM, in which case replacing that one socket is all that's needed. 

If you want to actually learn how these machines work, and how to troubleshoot methodically and carefully to find a problem, the "replace everything" approach is never the best way to do it.

[Mazzspeed]

1 minute ago, DrVenkman said:

See, you're right with the part I have quoted. COULD be. But it's far more common on an Atari to be something else. But shotgun-replacing literally a few hundred IC socket pins without actually determining the fault FIRST is bad practice. It's a lot of work, without the proper tools and experience it can be dangerous, and even if somewhere along those several hundred desoldering/re-soldering steps you've corrected the problem, you still end up with no idea - zero, none, zilch! - which component was at issue! ONE bad wipe on ONE socket might prevent the machine from booting. That might be on a single 16-pin DRAM, in which case replacing that one socket is all that's needed. 

If you want to actually learn how these machines work, and how to troubleshoot methodically and carefully to find a problem, the "replace everything" approach is never the best way to do it.

Totally agreed, its not something you would want to perform without decent experience in doing so and the right equipment, that's something the OP needs to decide for themselves. As to whether it's hard work? it's not that hard, I actually enjoy such work.

 

We replace capacitors due to age, this is really no different, just more pins. All OEM sockets of the era were rubbish and depending on climatic conditions corrosion could be an issue regarding most of the sockets. That's all my comment was suggesting.

[+DrVenkman]

40 minutes ago, Mazzspeed said:

We replace capacitors due to age, this is really no different,

Another retro-myth, at least as applied to Atari-made consoles and computers made before the late-80's. I've had to replace exactly one cap on an Atari 8-bit, a 10uf in an 800XL video circuit that failed quite obviously (leaking goo) when I accidentally shorted something working on the machine. That's it, in over a dozen personally-owned computers, and half a dozen or more personal consoles. 

 

What you say may be true of Commodore machines, or some other make, but not for an Atari of the Warner-era. If it's not broke, don't "fix" it. IDENTIFY THE ACTUAL PROBLEM, then fix that. 

EDIT: Actually, I've also had to replace the green polystyrene chiclet caps on a Warner-era Vader that I bought cheap from a local retro-shop, though not the main power electrolytics.

Edited February 27, 2021 by DrVenkman

[Mazzspeed]

9 minutes ago, DrVenkman said:

Another retro-myth, at least as applied to Atari-made consoles and computers made before the late-80's. I've had to replace exactly one cap on an Atari 8-bit, a 10uf in an 800XL video circuit that failed quite obviously (leaking goo) when I accidentally shorted something working on the machine. That's it, in over a dozen personally-owned computers, and half a dozen or more personal consoles. 

 

What you say may be true of Commodore machines, or some other make, but not for an Atari of the Warner-era. If it's not broke, don't "fix" it. IDENTIFY THE ACTUAL PROBLEM, then fix that. 

Once again, you're implying context in what seems to be an attempt to highlight some form of technical superiority?

 

I never stated that caps on the A8 platform are a problem, in the same way they're not a problem on a C64. In fact I would go as far as to state that it's fairly obvious my comment was a comment in general not aimed at anything in particular. Every time you remove a chip using component substitution as a form of diagnosis (which the OP has stated he has done), it's a good idea to replace the ageing and not too good considering the era socket - Otherwise there is a far chance you're creating problems as you're trying to diagnose the fault.

 

This is pretty straight forward stuff.

Edited February 27, 2021 by Mazzspeed

[+DrVenkman]

Just now, Mazzspeed said:

Every time you remove a chip using component substitution as a form of diagnosis (which the OP has stated he has done), it's a good idea to replace the ageing and not too good considering the era socket - Otherwise there is a far chance you're creating problems as you're trying to diagnose the fault.

Absolutely false. IDENTIFY THE PROBLEM. All of the OP's chips appear functional. Replacing over a dozen sockets is absurd. Read my suggestions to the OP - THAT is how you figure out what is actually wrong. Find what's wrong before you start "fixing" things. Why is that so difficult for you to accept? It seems to me that you just want to argue at this point. 

[_The Doctor__]

I think DrVenkman is talking perfect sense. He is also correct about someones need to argue a point. I personally have started to think it's a person who wants to be some sort of over the top defender of whatever beloved platform gave them a chill up their leg as an adolescent.

 

Since they couldn't help themselves... I say-   At least you don't see warehouses and repair shops stacked with dead Atari's stacked like cord wood like you did some other machines... Since someone wanted to talk of quality and superiority and other BS. New game, search for a link or two showing cord-wood pictures of bread-bins stacked high dead... they won't be hard to find if you carefully search AtariAge.

 

I hope the OP(s) follows the advice of people who know what they are talking about and have years of experience on the machine(s) they are working to repair and not those of people who are not up to speed on it in the least. I wonder if the mouthpiece even has real hardware yet. They were maybe going to sorta um, but using an emulator... ah... but hey they sure are experts aren't they. How many weeks could they have had a machine even if they ever decided to get one at that point anyway...

 

Edited February 27, 2021 by _The Doctor__

[Dr Memory]

The OP is remaining calm and taking it all in.  I tested all the voltages and grounds and they seem reasonable.  All but two chips were at 5.00-5.04V.  Those two were the LM358 and the Hex Buffer CMOS, both of which were at 4.74V.  That doesn't seem far off enough to be problematic but I'm still taking it as a possible clue and will revisit if nothing else is found.

 

The other interesting voltage thing is that very few of the grounds are actually at 0.0 volts.  The ground plane is, but most of the chips have mV or 10s of mV on GND.  I'm guessing this is because of the discrete components and the traces but am not 100% sure of that.

 

So now I'm about to dig into the Phi signals as suggested earlier.  I ended up making this little chart of chips and pins to use a checklist for the voltage check and updated it to add the Phi stuff.  Perhaps someone will find it interesting.  All the info is from known sources but I didn't find anywhere it was organized the way I wanted it.

 

 

Still having fun!

[+bob1200xl]

Make sure that RESET is working properly. It should go inactive about 1 second or so after power on.

 

sequence:

Power ON

RESET is active (at ground) and goes inactive (plus)

6502 wakes up and clears all of memory, including the hardware chips (POKEY and such)

Speaker 'pops' meaning init was successful. Video should also reset to nada. Screen should be black.

 

Bob

[+Stephen]

3 hours ago, DrVenkman said:

You're suggesting an intellectually-lazy procedure to diagnose a problem that is in all likelihood very simple, and usually has ZERO to do with the actual cause of boot failure on an Atari. Period. Just "replace everything" is terrible practice, especially if someone wants to actually learn what specifically is causing the boot failure and fix it. 

Replace all the caps!  If I had a dollar for every time I heard that.

[Mazzspeed]

1 hour ago, Dr Memory said:

The OP is remaining calm and taking it all in.  I tested all the voltages and grounds and they seem reasonable.  All but two chips were at 5.00-5.04V.  Those two were the LM358 and the Hex Buffer CMOS, both of which were at 4.74V.  That doesn't seem far off enough to be problematic but I'm still taking it as a possible clue and will revisit if nothing else is found.

 

The other interesting voltage thing is that very few of the grounds are actually at 0.0 volts.  The ground plane is, but most of the chips have mV or 10s of mV on GND.  I'm guessing this is because of the discrete components and the traces but am not 100% sure of that.

 

So now I'm about to dig into the Phi signals as suggested earlier.  I ended up making this little chart of chips and pins to use a checklist for the voltage check and updated it to add the Phi stuff.  Perhaps someone will find it interesting.  All the info is from known sources but I didn't find anywhere it was organized the way I wanted it.

 

 

Still having fun!

Are you able to post Phi signals? Very interesting.

 

57 minutes ago, Stephen said:

Replace all the caps!  If I had a dollar for every time I heard that.

Well there was a period there where caps were a real problem. The worst thing is they're as damaging as leaking Varta NiCads...

Obviously when it comes to 8bit machines they're not really an issue unless you're talking about BBC power supplies and it was never implied they were.

Edited February 28, 2021 by Mazzspeed

[+Stephen]

50 minutes ago, Mazzspeed said:

1 hour ago, Stephen said:

Replace all the caps!  If I had a dollar for every time I heard that.

Well there was a period there where caps were a real problem. The worst thing is they're as damaging as leaking Varta NiCads...

Obviously when it comes to 8bit machines they're not really an issue unless you're talking about BBC power supplies and it was never implied they were.

Edited 47 minutes ago by Mazzspeed

This wasn't really directed at you.  I am into vintage audio, and people will have you believe that if you don't just bulk replace every cap in every device more than 5 years old, shit will blow up when you try to use it, or it won't sound as sonically pleasing as possible.  100% complete and utter bollox.  I have 2 Crown DC 300A Series 2, a D75, a Comtech 800, two Comtech 1610s as well as 5 other amps of the late 60s early 70s era.  Not a single cap has been replaced in any of them, and I have had all of them on my test bench, proving they can do over rated power for hours on end.  It's a MAJOR pet peeve of mine, but I didn't mean to let it spill over into this thread.

[Dr Memory]

Sure.  The phi signals look plausible to me but I'm obviously no expert.  I saw no big differences between the source and sinks of the various phi signals, just some voltage drop.  No open circuits and no frequency changes.  But, I can't rule out the voltage drops being too high or the freqs being wrong or whatever, so I logged it all.

 

Phi 2 source (Sally) 1.77 MHz 2.19V

Phi 2 destinations (ANTIC, GTIA, POKEY, PIA, Delay Line) 1.77 MHz 1.75V on all except Delay Line, which measured 2.19V

 

Phi 0 source (ANTIC) 1.77 MHz 2.48V

Phi 0 destination (Sally) 1.77 MHz 1.99V

 

Fast Phi 0 source (GTIA) 3.55 MHz 2.46V

Fast Phi 0 destination (ANTIC) 3.55 MHz 2.42V

 

Phi 1 source (Sally) 1.77 MHz 1.73V

 

I'm pulling the frequencies and the voltages from the scope.  The voltages are "Avg" after I let is settle for a bit.  Obviously Avg isn't all that meaningful when you have signals with a large voltage swing, but it was the best summary number I could find for voltage.

 

I'm a little surprised to see that GTIA and ANTIC are the sources of two of the clock signals - would not have guessed that.  I guess GTIA is actually driving things with "Fast Phi 0" and ANTIC divides it by two and passes that along to Sally?  Huh.

 

I do not know if the frequencies are correct for a PAL 800XL as I haven't been able to find anything like the Sams book for PAL systems.  I did find a nice set of reverse engineered schematics that include the PAL circuitry though, so at least I can trace things out.  I just don't have example voltages and waveforms.  That would be cool.

 

I'm hoping that posting all this detail will help the next person that needs to debug the same sort of problem.  If I had been able to find a thread with this info in it, I'd be reading that and comparing it to my machine rather than doing all this work! 

[Mazzspeed]

Just now, Stephen said:

This wasn't really directed at you.  I am into vintage audio, and people will have you believe that if you don't just bulk replace every cap in every device more than 5 years old, shit will blow up when you try to use it, or it won't sound as sonically pleasing as possible.  100% complete and utter bollox.  I have 2 Crown DC 300A Series 2, a D75, a Comtech 800, two Comtech 1610s as well as 5 other amps of the late 60s early 70s era.  Not a single cap has been replaced in any of them, and I have had all of them on my test bench, proving they can do over rated power for hours on end.  It's a MAJOR pet peeve of mine, but I didn't mean to let it spill over into this thread.

The quality of caps in the 70s wasn't too bad, the issue is mostly isolated to a period around the mid 90s to early 2000's. My uncle used to have a DC300A with the factory Oak case, a beast of a power amp. Played Dire Straits MTV really loud.

[+kheller2]

21 hours ago, Dr Memory said:

 The only one I haven't been able to try is the wierd delay line thing because I don't seem to have a machine with that socketed.  I've been using a good 800XL for the secondary guinea pig, so no worries on the 1200XL/XEGS front.

 

 

Hmm. Have we verified that is working correctly?