• Please review our updated Terms and Rules here

IBM 5110/5100 video Display card repair (possible diode fix)

I know the 5110 has a “print screen” key, just like the later 5150- where the screen content is sent directly to the printer. So I’ve wondered if this print clock was related to that ( timing to avoid sending print request too fast to the printer? ). But I’m curious now, what happens if I print when no printer is attached? Does the system just dead lock or loop forever? I’ll try it out on the good 5110 later today.

I’m also strategizing on how to scope these signals. Swapping cards once the scope is connected is a hassle. So maybe I could just setup two 5110s and manually step them one instruction at a time? ( one with good vs bad displays cards, but otherwise identical )

Plus, adding multiple probes at the same time may influence the system. I had 5 probes setup at once, and the red light came on right away on power up (not waiting till 4th cycle). Could that many probes be adjusting amp distribution ? So may have to prove one pin at a time - might be safer that way anyhow.
 
I’m also strategizing on how to scope these signals. Swapping cards once the scope is connected is a hassle. So maybe I could just setup two 5110s and manually step them one instruction at a time? ( one with good vs bad displays cards, but otherwise identical )

Plus, adding multiple probes at the same time may influence the system. I had 5 probes setup at once, and the red light came on right away on power up (not waiting till 4th cycle). Could that many probes be adjusting amp distribution ? So may have to prove one pin at a time - might be safer that way anyhow.

This seems like a good strategy to me. It's possible that you'll need delicate triggering to see the fault; before you know it, you might also find yourself borrowing or buying a logic analyser :)

But you can always start simple. If you're lucky, maybe just probing individual output lines will reveal something. You might find that one of the output lines has funny voltages: maybe a driver is marginal and isn't quite getting up to +5V or down to 0V on one of the logic lines. Or maybe one of the lines is just switching slowly for some reason, too slowly for the rest of the computer. Good luck!
 
Little tangent from the topic.... Took me awhile to find this again, but it's in the IBM 5100 BASIC Introduction manual:

1655857487004.png


I was disappointed at first when I got an IBM 5110 B-model, I thought the label of the key had worn off. But I was relieved after coming across the above paragraph, since it turns out "that's normal !" - but there is no explanation as to why (maybe to encourage using the keypad * and divide instead? the lower blank key next to P does left/right arrow -- you can still use the keys, just its normal for the label to be blank on them)

CMD+(*/-) (the top blank key) just gives me an ERROR 013 500 (which is simply "printer not attached" -- and I can resume work, no lockup or freeze; I don't even have to ATTN out of the error).

1655858111300.png


Still not sure if the "Printer Clock" pin on the Display card is actually related to the Printer - or why the Display would care (maybe seeing a signal on this pin indicates if the printer is already busy and not available to do the Copy Display operation?)....
 
Ok, I went a single pin at a time, on just the major ones we noted discrepancies on earlier.

Did the BAD display card first, annotated straight into the worksheet as I went (less risk of transcribing error).

I believe the main reason for the 4.8V vs 4.4V discrepancy (on the StorageAddress bit) is that those are actually strobed lines (so the instantaneous DMM is actually taking a sort of "average" voltage, right? so strobing on the line would pull that average down). The o-scope shows the spread of this strobing (will post those later). So all the StorageAddress pins have this behavior, I just monitored B10, D04, and J09 in the set below.

Another thing I realized is -- I stepped a few cycles/instructions past #4, and I think what is happening is the Display is being commanded to BLINK - OFF, ON, OFF, ON, etc. after the first error is encountered. This might explain why in the video posted earlier appears like a Sync issue -- every other cycle that I stepped (on the BAD card after cycle #4), the DisplayOff pin would invert. AND that kind of makes sense -- during syntax errors with the Language ROS, the machine BEEPS, and the display flashes until you press ATTN.

NOTE below when I say "cyclic", the pin value wanders between a range, not just sits at a fixed value.

Mainly what gets my attention now is that D02 Printer Clock pin -- on the GOOD Display card, that "clock" pin is constant (near 0, about 0.1 or 0.2V; at least for the first few cycles) -- as opposed to the BAD card, where it cycles from the get-go. Per the SLM, that's driven by P07/C5 from the H2 (BASE IO card), which I'll monitor in a little bit -- but so far, we are assuming these inputs must be correct, since the GOOD card is working. But P07/C5 goes into the CharacterCount. At the very startup of the system, there is nothing to Display yet - so makes sense to me the CharacterCount would sit at 0? But maybe that's not how it works. I'll probe that pin again also, in Run mode (with the GOOD card) to see it always stays low at ~0.1V.

1655868819058.png

One other thing to note is that P11 (BusOut3) it doesn't seem to get changed during a RESET - it holds the last value it was given. And that seems OK, it's why I marked it as "last" in the table above on the first cycle/instruction.


Now to compile up what these signals look like...
 
Sorry, this will be a little painful - and it'll be hard to see the horizontal scale on all of these (the vertical scale is each tick is 1v, so major tick is 5v).

I mostly present the First Instruction (@1 initial startup), and don't show beyond that unless there is a major change.

D02 is the one that gets my attention the most (and yes both sides are at the same scale for D02 -- 20usec). More on this soon.

display_pins_comparisons.png
 
Ok, so let's dig into this D02 Printer Clock thing some more...

"A1G2" is the Display card in question. "X4D07" I think is on the "X4" connector of the Processor? In any case, it's some division of the ~15MHz clock (I think) that comes in on the BaseIO in slot H2....




1655873478549.png

And now I'm wondering -- maybe the Display isn't really so much flashing, maybe the whole machine is just being RESET ? I guess I could test the Power On Reset pin somewhere to check that... and if so, maybe that's why all those Address and Bus pins are funny -- they're not given a chance to cycle before the RESET?

But I don't see how failure in the Cycle Steal Logic would cause the system to reset. Unless the Processor is detecting that a Cycle Steal is happening at an inappropriate moment??

Ok, that's my next play: see if POR is being somehow triggered by the presence of this Display card.

A1J2 B11, I think?
 
No luck in identifying if anything is explicitly issuing a RESET (by monitoring any POR related signal on the H2/BASEIO or Executive or Display itself - none of them invert, unless I press the RESET switch myself). But, I did find something interesting....

Monitoring A1G2 P02, the Display is being TOLD to turn off.... (more on this in a moment)

As a result of being TOLD to turn off, via A1G2 P05 the Display commands the SPEAKER to turn on....

What's telling the Display to turn off? Monitoring A1L2 D07, the Executive ROS is telling the Display to turn off.

And the interesting part is: this doesn't happen with the GOOD WORKING Display card (i.e. A1L2 D07 never triggers from the Executive to the Display to turn off).


So.... WHY is the BAD DISPLAY inducing the Executive to turn off the Display? (within 4 instructions)

Well... I monitored A1L2 (Executive) P05, P06, D06... None of them ever invert. I zoomed out on the scope, I zoomed in -- thinking maybe I missed the signal. I don't think I missed it (but I haven't learned enough to setup triggers and such on the scope).

So, what does "CR" mean? My working theory is there is logic omitted in the Executive ROS SLM diagram - another signal could be triggering the B07/D07 (display off and process LED), and internally the Executive ROS could be resetting itself back to 0x000A, as an implicit internal reset). Something besides either an Addr or Bus Error is inducing the Display to go off.

So @stepleton I'll have to read up more on what you mentioned about Cycle Steal - that's what I'll need to study next. I assume it's something like: the processor COULD be writing to Display memory (0x0200), but the Display needs to read RWS in order to know what to display. In software, this is a deadlock potential - you have a resource being read/write at the same time. So the Display has to "steal" some time to read from the RWS, in order to know what to render? And if that sequence is getting out of sync, maybe the Executive somehow detects this and induces and internal reset? That's all speculation.... Since I don't think the Executive cares much about Cycle Steal, that's up to the Processor. The Processor A1J2 does have a "Cycle Control" box, with "Display Request Pwrd" as an input. But I'm not seeing how the Processor can induce the Executive to turn the display off (I mean, within 4 instructions; seems it would need a direct path).

In the Processor I checked S08, U09, and I removed the S07/S09 Machine Check Jumper -- no signal change, and no runtime change.



1655880444192.png


So, the next plan then is to monitor B02 and D06 (Stolen Cycle Next and Stolen Cycle) coming out of the Process A1J2, along with the A1G2 D07 (Display Req) coming out of the Display -- to see if their sequence matches how they go with the GOOD card. Maybe the BAD Card is interpreting those outputs differently.

IF that sequence does end up being different between the cards, then -- well, that D02 Printer Clock is one of the inputs into the Cycle Steal Logic box. Maybe just the Character Counter DIP is messed up, causing the Printer Clock signal to go bonkers, which in turn messes up the Cycle Steal Logic, and causes the system to reset itself ?? I wonder if I could test that by somehow fixing the Printer Clock signal to be ~0.1V like it is with the working Display card (in which case, maybe all the output just goes to the first column -- but at least the Alarm would be quiet :D )
 
I wonder if I could test that by somehow fixing the Printer Clock signal to be ~0.1V like it is with the working Display card (in which case, maybe all the output just goes to the first column -- but at least the Alarm would be quiet :D )
Jumping in outside of my usual schedule to say that I wouldn't do this: it seems likely to me that this signal is being driven by the "Character Counter" and probably can't be forced low without straining that IC.

D02 looks very weird to me and I think you are correct to be suspicious of it. The slow rise and fall on the "bad" card does not look right to me at all. On the "good" card, though, I would expect to see more activity than the flat line you've shown (unless the screen is off): is it possible that there are transitions visible if you choose a narrower time base (i.e. "zoom in" on the horizontal axis)?

Just to pass the time, I would recommend two things. First, I'd learn about triggering with your oscilloscope: it's an extremely useful function that you are sure to put to use in this project soon if you know how to use it. Second, I'd trace D02 back into the Display Card and try to find which IC is generating it. It's too early to say that this IC is broken, but at this point I'd already be curious enough to take a look at it.
 
Ah no worries, I won't actually try to drive that Printer Clock (indeed would be some board surgery that is beyond me) - just was curious that it looked rather constant in the working/good case.



Also, I just realized, there is an undocumented 4th line here (see below)... The label correspond to the lines below them, not above. Maybe this extra line at the top is just a mistake.

1655914231190.png
That's what I meant by "what is CR" earlier. From the 5100 MIM:
1655915260048.png


And that's the D07 I monitored when STEPING through the startup (that goes off to A1G2, the Display card). When I press the RESET switch (or even hold it), D05 toggles and does cause the display to go off at that moment (as expected).... But something else (besides the RESET Switch, and besides either P06 or P05) is also causing D07 to toggle (i.e. the Executive being induced to command the Display to turn off).
 
I won't actually try to drive that Printer Clock (indeed would be some board surgery that is beyond me) - just was curious that it looked rather constant in the working/good case.
I'm actually pretty curious about that, because unless there's some detail that the 5110 logic diagram isn't showing, then that line should not be constant on a good board. (Hopefully what I say below makes it clear why this is so.) As above, I'd encourage zooming in on the time scale and seeing if there's actually some activity there.

Also, I just realized, there is an undocumented 4th line here (see below)... The label correspond to the lines below them, not above. Maybe this extra line at the top is just a mistake.
Could be; either way, given that the adjacent input to the diode is connected to the -Machine Check line, it seems like it could just be another copy of -Machine Check.


With regard to your other remarks: there have been a lot of them :) I'm afraid I haven't got time this evening to reply to them all. Here's the thing, though: I'd really consider drilling down into D02/+Printer Clock Pwrd right now. Take a look at where the "Character Counter" sits on the Display Card: it's one of the chambers of the beating heart inside the Card (the other ones being the line and row counters). It takes only a clock input (and so it should always be counting even when the display is disabled), but its outputs go all over the place, including the Cycle Steal logic, the line counter, the display signal generator, and the address bus logic.

You already know that the input to the Character Counter is good: the C5 counter is used all over the computer, and if it weren't working, other Display Cards wouldn't work either, and in fact the machine would probably be a lot more broken. You also know that the output of the Character Counter looks bad: that sine-wavey sawtooth looking trace doesn't even seem digital. It seems like what you would get if you had a square wave and a good chunk of capacitance. (Speaking of that, just to double check against an honest possible mistake: is your oscilloscope ground lead connected to a good ground? Seems like it...)

It's always possible that the Character Counter is fine and that something that it plugs into is bad. But either way I think your problem could be lurking right around here. If you agree with my reasoning, consider taking a closer look at D02 on both good and bad cards. Change the timebase on your scope. Pull the bad card out of the computer and use a continuity tester to find what ICs D02 connects to.


Finally, study the Character Counter and adjacent logic in the 5100 MIM (PDF pages 286, 287), which is more detailed:

1655936136694.png

Notice that what the 5110 logic diagrams show as a line, the 5100 diagrams reveal to be a bus. This is a counter that can count from 0 to 127, and each of its seven bits is a line on the bus. Each of those lines should almost definitely be nice, well-behaved square waves. One of these lines is probably what's going out of D02 as +Printer Clock Pwrd. If that's true, it should look tidy. The fact that it doesn't is deeply weird!


My hope is that you'll wind up tracing D02 to some four-bit counter IC DIP that can be easily replaced. We don't know yet if it's time to replace it, but wouldn't it be nice to know that we could if we needed to? Either way --- I hope we're close now!
 
Last edited:
I measured D02 again on the BAD and GOOD Display Card.

Same results.

EDIT: Good idea about double checking the GROUND. I'm using the metal bracket that the case cover slides into when it's on - verified continuity with multiple other grounds.


On the GOOD Display Card, I "zoomed in" and "zoomed out." The GUI dropdown says it can go down to 1ns, but I don't really believe that on a Windows-based application that I'm using on a laptop -- maybe, but as shown is the 10usec view (same scale as the "BAD" Display card). I zoomed out to 20ms, and this same constant voltage was reported in both STEP at startup, and when I flipped to RUN and "booted" all the way to BASIC prompt. Agree, what kind of "Clock" signal is constant? But as they say, "the results speak for themselves" :) and that Display card works. I guess I could check this against yet another 5110 Display card that I have?

1655944800861.png
 
Here is the ONLY spot that I found continuity with D02.... 4th pin in (from either side). Note the position relative to the "famous" diode mentioned earlier in this thread (I mean, to orient on the relative position of this chip; not to imply the diode has any influence on that chip).

D02_PIN_small.jpg

I read this IC label as:

F 239 2122
OG 765229
7808EJ INDONESIA
 
My hope is that you'll wind up tracing D02 to some four-bit counter IC DIP that can be easily replaced.
I note the following ICs from your pictures:

DIP 2392137 x2: a four-bit binary counter
DIP 2396261 x1: a counter of some kind

Other DIPs are buffers, nand gates, shift registers.

The fact that D02 is flat is a mystery to me, but let's think for a second... C5 is a clock output that could be going as fast as 1/5 of the system clock. That's 3 MHz or 200ns; I see your timebase is 10us, which could still be too big. But I don't hold out much hope yet.

Here is the ONLY spot that I found continuity with D02.... 4th pin in (from either side). Note the position relative to the "famous" diode mentioned earlier in this thread (I mean to orient on the relative position of this chip).

Thanks for checking. A 2392122 is (per this reference) a 7417 open-collector (huh!) hex driver. Maybe the output isn't supposed to do anything... but the input probably is.

Anyway, by "from either side" do you mean that you're detecting continuity to D02 on pin 4 and on pin 11 (as shown in the picture)? I think that would be pretty strange!

You can think of a hex driver as kind-of a digital amplifier or repeater: signals that come in on an input pin are reproduced on a corresponding output pin, sometimes converted to a different signalling method (the 7417 converts the input to open-collector signalling). "Hex" means that there's six of these little repeaters all in the same package. But they're all separate from each other: they're connected to the same power and ground, but they all operate independently. They're just all packaged together for convenience, like eggs in a carton or cans of beer.

If pin 4 and pin 11 are both connected to D02, that means that the input to the fifth driver (pin 11) is connected to the output of the second driver (pin 4), and also to D02. There could be a good reason for things to be this way, but I'm not sure what it is. One of them is: fried chip :)

But let's assume this is actually correct. In that case, continue the search backwards from Pin 4. What's that pin connected to? My hope is that it's one of these counter DIPs; I'd check those chips first.
 
Oh, I just meant from either END, not side. 4th pin from the "left" or "right" - the middle pin, on that side only :D

Here's the same-type of chip from the GOOD Display card ("NK" and different brand symbol, but same numbers). That center pin only beeped with D02 :)
EDIT: The bottom right portion here has a reflection since this was shot thru a magnifying glass.
1655946923674.png
 
Last edited:
Sorry, correction: trace back from pin 3, which is the input to the second driver. And I was wrong: the pin you're probing in the image is pin 4, not pin 11. My bad. In any case, thanks for the clarification: this makes more sense. The hunt continues from the input to the second driver. What's pin 3, the pin closest to the silkscreened letters "EJ" in the "probing" image, connected to?
 
Whew - thanks, I was about to ask if you meant from the corresponding input pin, and what that pin was! Next to the "EJ", got it. A little harder to probe, but I'll trace it.
 
Haven't found it yet. I tried against the other ICs, nothing obvious stood out. Then I realized, maybe I better first trace this using the GOOD Display card instead of the BAD and tried again. Still no luck. It's a lot of pads to try from the backside. Then I remembered I had one of those "electronics" cameras -- basically a web cam that can be mounted on a steady frame to look like a microscope.

Thought some people might appreciate what that looks like.... Here is that IC from the bottom, with the pin 4 connected to D02. So for pin 3 next to it, now I literally follow the trace :D Here goes...

1655949627699.png
 
Uh, well that wasn't very far... It goes "underneath" the chip next to it... So whatever is next to pin 3 of the TI 239 2129 chip that is in the direction towards that diode.
EDIT: the following is rotated 90deg from the above.

Oh, on the top ("other") side of this board, it could then continue to trace.... But it's under a chip. Hmmm. Neat, like a highway - the "bottom" goes "East/West", the "top" (component side) goes "North/South".

1655950022148.png
 
Last edited:
Well, bad news... If we're looking at the right pins, that traces directly over to one of the IBM 22 tin cans (the 2nd one from the edge, where I marked "A")

NOTE: The shot below is from the GOOD Display card (so our tracing starts from a chip labeled NK and I have the mentioned pin3 next to an "11" on the DIP label instead of "EJ").

Sure pin3 is in that direction relative to pin4? I didn't double check.

IMG_2593A.jpg
 
DISREGARD POST #60 - those were all the wrong pins! (think wrong card)

C5 click lines are fine!


I was reviewing the Board Net Listing poster, appreciating its value more and more. And I noticed the following: C3 isn't used. That's fine. But I started monitoring C5 on the other cards, and it looked as-expected, or at least: not flat.


1655966305110.png

D2 = Asynch Communications (click input pins are active even if card not installed)
F2 = Language/Common ROS
G2 = Display Card
H2 = Base IO


The equipment I have may not be sensitive enough to go to these even these modest clock frequencies - so it can't really tell the difference between C2/C4/C5. But the point is: C5 isn't flat.



1655967110936.png

The above isn't to say C2/C4/C5 are the same, it's just a limitation of the equipment (I think) or my setup of it. And this result was consistent when measuring C2/C4/C5 on other cards.


I think this is a relief - since it gives more evidence to that CharacterCount chip being the issue? I'm ready to order a replacement :D

Though it's still a mystery about the A1G2 (Display) D02 output being sawtooth (bad) vs flat (good). I've confirmed that a 3rd time, and I believe that part is correct as previously described. Sorry for the distraction about these Clock pins, they seem OK.

But I have no insight as to how those IC DIP chips can fail. I mean, a power surge or short causes "random" failures (typically starting with the "weakest" component?). Or ambient moisture (or dust) during production may cause some kind of corrosion over time?

And still, this might not be THE problem, or maybe not the ONLY problem.
 
Back
Top