• Please review our updated Terms and Rules here

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

What happens if you just let the machine start up normally?

Excellent question, I did forget to describe that for future reference....

Here is the comparison below.

On the left is the "good" Display Card inserted, "normal" startup. The 013 D80 error is related to the disk system not being attached or available - it is an error that can be by-passed (normally I just press ATTN a few times and RESET, or if you just RESET a few times with power already on, it'll by-pass this error and boot up to the language ROS as normal).

On the right, with the "bad" Display Card inserted, it depends on if I leave it in regular RUN mode or STEP mode. In RUN mode, the red status light at the front still comes on, and you see a scramble of content on the screen for about 4-5 seconds (BUP sequence in progress), then you get to a static image like shown below (center image). It's not quite as bad as the flickering we saw in the previous video, but as you can see, it is a scrambled jumble of dots. Now the more intense dots at the top left and bottom right, interestingly I think those do correspond to the normal intended text (of the BUP checks, and the 013 D80 error status). I can't really explain the intermediate sets of dots - and the display is fairly static, this isn't wobbling or flickery. But if I move the switch from RUN to STEP - then you get what you see below on the far right. The content of the screen is "double'd up" (again the first 8 characters are incorrect, but the next 8-characters are correct).

comparison_after_7417_replacementA.jpg



Haven't had time yet to probe the pins - but I suspect we're going to find another 7417 with a stuck pair of In/Out, and then hopefully that'll be all we need to get this up and going again :D
 
Good progress.

It is a shame that your CRT in the VDU has had the phosphor damaged in the screen center area. This is because the VDU did not have turn-off spot suppression. It can be added to the VDU fairly easily and the CRT could be replaced.
 
Well, I do have an extra CRT that is in another 5110. It's not showing anything at all - I haven't started investigating that (power connector issue? CRT PCB issue?), it's a future project. Nothing seems burnt or broken, so I'm hoping it's as simple as adjusting those internal brightness knobs :)

The external BNC still works as a display, so the system can still be used.
 
@Hugo Holden can you say more about this? It might be useful to add this suppression to machines that haven't encountered this problem yet.

At the moment the VDU is switched off, the H & V scan stages stop fairly abruptly and there is little CRT beam deflection. But the EHT is still there because the CRT's glass bulb acts like a capacitor, storing charge on the external and internal conductive coating, the glass bulb being the dielectric. The Silicon EHT rectifier feeding the CRT bulb's EHT connection (internal coating and final anode connection on the gun) has low reverse leakage so it can take some hours for the charge on the CRT bulb to bleed away. The CRT's heater is still hot, emitting copious electrons. So the high beam energy, rather than been distributed evenly across the phosphor screen, is applied as a high intensity spot to the center area. This, over time, relentlessly and cumulatively damages and de-sensitizes the screen phosphor there. It physically darkens and fails to emit much light when hit by the usual scanning beam. There is no fix for this once it is damaged, though in the past there were companies that could re-gun and re-screen CRT's, but they are all gone now. The last one was in France over a decade ago.

It is an age old problem with CRT's in TV's and VDU's. The PET VDU has the same problem, unmodified that is. One thing that can help a little, is if the user turns the CRT brightness to full low before switch off, but that is very easy to forget.

The general way the problem is remedied is to ensure, that at turn off, the grid (in the CRT's gun) is held negative with respect to the CRT's cathode (Or the cathode positive with respect to the grid, which amounts to the same thing) for a long enough period for the cathode to cool down and stop emitting electrons. This voltage repels the electrons back to the cathode and extinguishes the beam current for long enough for the cathode to cool. (although other methods had been tried including transiently increasing the beam current to help deplete the EHT initially, but that idea wasn't very good).

The usual way it can be done as a modification is to either alter an existing circuit value, or add a few parts. In the PET VDU for example is was just a matter of increasing a capacitor value which stores charge on the brightness control and grid circuit to hold the grid negative for a longer period after turn off. In other cases, the capacitor feeding the CRT's cathode B+ circuit can be increased, to keep the cathode more positive after turn off for longer. In other cases a diode, capacitor and resistor arrangement can be added to the CRT's grid circuit to pull the grid negative when the power supply rail in the VDU collapses, this method is shown on the very last page of this article:


The better method depends on the particular VDU design, I will look at the schematic of your VDU........can you post the link to your VDU's schematic. Depending on the CRT part number, I might know where to get a replacement or equivalent type.
 
Last edited:
It's the same CRT mentioned in this thread:

Actually @stepleton would be more interested in the older version of that CRT (which the "extra" I have is also that same kind, as used on the IBM 5100). I haven't taken it out of the system to really explore and look for more specific model number information.

Neat info about the PET - I still have my PET. I haven't seen any with "blown out screen" likes I've seen with these IBM 5110's. Maybe the 40XX PETs had more protection built in about this, compared to the earlier PET 20XX's?

And good idea on turning brightness down before power off. I've wondered if on a "correctly calibrated" IBM 5110, if flipping between NORMAL and REVERSE video should need no adjustment to the brightness? Whenever I flip that mode, I seem to need to either turn-up or turn-down the brightness. I haven't messed with the knobs inside on the CRT, but I've wondered if there is some 'perfect' setting where no brightness adjustment when switching between NORMAL/REVERSE (black background vs white background) is necessary.
 
Ok, I think I need help on what address lines to examine for the next step.

The following chart may be hard to read, but I'll try to explain it... While not quite yet an actual circuit diagram, I wanted to map out where all the 24x4 = 96 pins of the Display card when to, on the Display card. So yep, using the DMM and tinfoil, I started exploring. Many were obvious traces, but I did end up needing that tinfoil to help isolate down the area of quite a few connections that didn't go just straight up/down or across the board.

The Display card has two main groups of components: "tincans" (the silver square IBM chips), and 14-pin or 16-pin IC's. (then a few resistors, diode, and whatever those "915" chips are called - a capacitor of some type).

Anyhow, I'm concentrating on IC's and "tincans" and I labeled them like this (1-7 on the tincans, 1-15 on the IC DIPs). The set of 96 pins at the bottom of the board connect (and are exposed from) the A1 board, so I'll call those "A1-pins" or "external pins" (from the perspective of this Display card). It turns out that tincan 2 and 7 aren't referenced at all from any of those 96 pins, so they are for "internal processing only" (likewise with several of the DIPs). I didn't concentrate yet on the internal interconnections between these chips, just on there the A1-pins go. (the numbers are "backwards" in the image below, because I was working from the perspective of the solder pads on the backside of the board, so I was never really even looking at the component side)

IMG_2469A_singleA.jpg

So, here's the tentative result.... On the left in each box is the DB GJ MP SU from pins at the bottom of the Display card, then I just tested continuity to find the "first major component" that these pins had continuity with; most of them had several intermediate pad connections along the way to that first-component; I do have another diagram that marks those, but the following was the most concise way I could come up with. On the right side of the boxes below: the IC chips are standard pin number arrangement, but for the tincans I just numbered them in-sequence from top to bottom of the pins they had. It doesn't especially matter which pin on those, since we don't know inputs vs outputs on those chips anyway.

"Rx" is that the pin connects to a resistor in parallel, not that the signal goes through a resistor.

"951" is code for these little "monolith" components marked with "9" in white background and "51" in black background, which I think are a small regulating capacitor.



ibm5110_display.png

I wasn't sure the best way to group these resulting notes, but I grouped by them by major IC chip (tincan or DIP).

There are some pins that are truly NC ("not connected") and I removed those from this diagram. But there are a few that per the SLM that were "not connected" and yet seem to still have a connection (these are marked in black background).

So my question is: to find this "stuck address pin", should I be looking at "SA" pins? But those are outputs from the Display card (Storage Address Bits).

What are the Even/Odd bits? Those are inputs that go to the Display Data Register. I assume its related to the characters on a row being in some Even/Odd column assignment? Or maybe it relates to being in REGISTERS mode, and showing portions of bytes vertically in Even/Odd Rows? Or something else entirely.

Or should I be focusing on the BO (BusOut) pins? Notice that two of these go straight to an IC (#12) instead of the tincan5 like the rest of the lines.
 
Last edited:
Cross-referencing the IBM-chip reference here that @stepleton reported a few pages back... Here is what I get for the DIP chips on this Display card:

Code:
IC1 = 1582601 / 767396 ==> 74157 (DATA SEL/MUX)
IC2 = 1582601 / 767396 ==> 74157 (DATA SEL/MUX)
IC3 = 2392101 / 767368 ==> 7402  (2-IN NOR)
IC4 = 2392102 / 767368 ==> 7404  (HEX INVERTER)
IC5 = 2396261 / 721153 ==> 1231  (COUNTER)
IC6 = 2392137 / 851888 ==> 74193 (COUNTER)
IC7 = 2392137 / 851888 ==> 74193 (COUNTER)
IC8 = 2392129 / 760506 ==> 7495  (SHIFT REG)
IC9 = 2392129 / 760506 ==> 7495  (SHIFT REG)
IC10= 1582601 / 767396 ==> 74157 (DATA SEL/MUX)
IC11= 1582601 / 767396 ==> 74157 (DATA SEL/MUX)
IC12= 1589415 / 851041 ==> 74298 (QUAD 2-IN MUX)
IC13= 2392166 / 763776 ==> 7437  (2-IN NAND)
IC14= 2392122 / 766229 ==> 7417  (HEX DRIVER)
IC15= 2392129 / 760506 ==> 7495  (SHIFT REG)
 
It's the same CRT mentioned in this thread:

Actually @stepleton would be more interested in the older version of that CRT (which the "extra" I have is also that same kind, as used on the IBM 5100). I haven't taken it out of the system to really explore and look for
I couldn't make any more remarks on it unless I could see the exact schematic. I have not been able to find it yet.

It is great it has been figured out what all the DIP IC's are, that will drastically increase the chances of being able to fix the card. It also means that there is a trick that can be performed now. HP made a logic comparator that you can clip onto an in circuit IC for testing. You put the reference known good IC on a card that fits into the comparator. The device is powered from the PCB you are testing. All of those 74 series IC's can be tested in circuit now without having to remove any because you now know what they are:

 
Last edited:
So my question is: to find this "stuck address pin", should I be looking at "SA" pins? But those are outputs from the Display card (Storage Address Bits).
I would think that these are the pins to probe. The Display Card should use these pins to indicate which data it wants to collect from memory to display on the screen. If one of these pins is stuck, then it can wind up repeating patterns of characters we see on the screen as the Card effectively requests the same data twice.

In our case, I would expect that the culprit could be D04, storage address bit +B, although since the data bus is 16 bits wide, it could be one of the adjacent adress lines. I'm a little surprised to see that these lines are marked +, meaning that they are ordinary non-inverted logic. I expect the pin will be stuck then at around +5V.

I see that D04 is connected to TINCAN4, but never fear --- let's not rule out that there might be some other issue in play. Maybe there's another component attached to this same line, or perhaps some fault outside of the chip.

What are the Even/Odd bits? Those are inputs that go to the Display Data Register. I assume its related to the characters on a row being in some Even/Odd column assignment? Or maybe it relates to being in REGISTERS mode, and showing portions of bytes vertically in Even/Odd Rows? Or something else entirely.
Those bits are the data coming in over the storage data bus: after the Display Card has requested an address via the address bus, it listens on the storage data bus for the data it asked for. It should use that information to display text on the screen.

It looks like the data bus is two bytes wide, so I think "odd/even" just refers to the two separate bytes: the odd byte will be the byte that had the odd address in memory (e.g. $1235) and the even byte is its neighbour (e.g. the byte at $1234).
 
The better method depends on the particular VDU design, I will look at the schematic of your VDU........can you post the link to your VDU's schematic. Depending on the CRT part number, I might know where to get a replacement or equivalent type.
@Hugo Holden thanks for this detail.

@voidstar78 has found two models of VDU in his 5110s, but the most common ones so far as I'm aware were made by Ball. Based on these photos from an earlier post, the Ball VDU is the TV-5

1658251690017.png

which leads us to this manual at Bitsavers. Comparing the diagram of the VDU's circuit board on PDF page 20 and voidstar's photos, they seem to be the same --- IBM didn't substitute a custom board.

I think the schematic we want is on PDF page 23.

The other of voidstar's VDUs is made by NEC. I did not look long, but I don't think Bitsavers can help us out here.

The CRT's heater is still hot, emitting copious electrons. So the high beam energy, rather than been distributed evenly across the phosphor screen, is applied as a high intensity spot to the center area. This, over time, relentlessly and cumulatively damages and de-sensitizes the screen phosphor there. It physically darkens and fails to emit much light when hit by the usual scanning beam

This is interesting: so even though you cannot see a lingering dot as you do on e.g. the unmodified PET display, there is still damage being caused?
 
@Hugo Holden


This is interesting: so even though you cannot see a lingering dot as you do on e.g. the unmodified PET display, there is still damage being caused?
No, there has to be an intense visible dot, past a certain energy threshold to damage the phosphor. The aluminium layer on the rear of the phosphor, provides some protection. In CRT's that don't have this the phosphor is even more vulnerable.

In the case of the PET, it primarily had a late relatively low intensity spot, compared to the initial ones that normally burn the screen immediately at turn off in other VDU's. It appeared because there was insufficient charge storage (filter capacitor value) in the CRT's brightness control circuit. So initially the spot was suppressed immediately at turn off, but then after the capacitor discharged into the brightness control pot and the grid's negative voltage depleted, the beam current came up again and a lower energy longer duration spot turned up after some seconds and stayed until the CRT's cathode cooled down. Luckily this did not cause too much damage in the PET CRT, but still, it is better suppressed to zero.
 
Looking at the VDU schematic, there is very little charge storage on the -160v negative voltage supply to the brightness control and grid circuit. The first move would be to increase C110 from 0.047uF to around 4.7 to10uF, with an electrolytic capacitor of >160V rating.

Also, very oddly, they created some charge storage on the +ve side of the brightness control with a series diode and filter cap C119. This is counter productive, because after turn off, it(the charge on C119) holds the brightness control (and CRT grid) on the average, a little more positive than the collapsing +34V rail powering the video output stage and cathode circuitry, for a while, this would be expected to aggravate the turn off spot, so I would link out CR107 (short out the diode CR107) should help. But you could just try initially tacking in the 4.7uF cap first and see if the turn-off spot is eliminated or reduced.

Bit it might be hard to see on your damaged CRT, as the phosphor is so badly damaged in the center area.

There might be a CRT part number on the CRT somewhere to help identify the type.
 
Last edited:
Ok, I think I've found the next discrepancy between the GOOD and BAD Display cards....

I monitored all the SA (Storage Address Bits), which are outputs from the Display card over to the J-Card (the Processor). They all looked and behaved basically the same (as shown below, a "shark fin" spaced out). Nothing really stood out as stuck or inconsistent within this group of pins (we thought SA+B might be an issue; why -5 and -6 are marked negative in the SLM is weird, maybe the -/+ there don't mean what I think they do). These are all output lines (from the Display card).

The only oddity to note here is that U02 (SA+C) behaved slightly different than the rest in this group (during STEP mode -- after a hesitation of about half a second, the signal would settle to a flat 5V; only for this U02 pin, until the next STEP). Just noting this, may come back to it later. But in RUN mode, there was no obvious visual difference between any of these Storage Address Bits.


The more obvious finding, I'll describe below...




IBM5110displaycardissue2.jpg


So then I compared the BO-pins (Bus Out). These are inputs from Card-H, which is the IO Controller that is next to the Processor. The behavior of all these pins is basically as imaged above, with a kind of "long" 5V signal. But on the BAD display card, BO-3 of pin P11 displayed quite a different behavior. This is the bottom image of the above, where it is a more periodic ~2.0V signal.



@stepleton pointed out the following from the IBM 5100 MIM (I never found this chart in the 5110 MIM or other documentation, but I think it's safe to assume it's the same levels between them).
1658288284331.png

So this BO-3/pin 11 is effectively stuck on unreliable region, which is what we expected to find "somewhere" - although I can't yet explain why it is this pin. And this pin being "stuck" I assume would make BO-P (the expected parity of this data) to fail. As an input pin, it occurs to me now that I should check this corresponding pin (J10) on the Base IO card (H2 card). Or I can trace on the board to see if this same pin splits off to anything else on the board.

Also, when have more time, I'll examine those Even/Odd bits to see if anything looks out of place there as well. These pins are all accessed at the back ("top") side of the A1 board, so pretty easy now.
 
These guys?

IC1 = 1582601 / 767396 ==> 74157 (DATA SEL/MUX)
IC2 = 1582601 / 767396 ==> 74157 (DATA SEL/MUX)
IC10= 1582601 / 767396 ==> 74157 (DATA SEL/MUX)
IC11= 1582601 / 767396 ==> 74157 (DATA SEL/MUX)

I haven't yet. Setting up to probe the pins on any of the DIPs takes more setup work (no riser card), but I can focus on that later this week.

I did probe through the Even/Odd pins and didn't see anything in particular that stood out (in terms of differences between the BAD and GOOD cards and what the signal along those pins looks like).



I did find some additional pins connected to P11 (Bus Out -3) on the Display card, that has the suspicious ~2V signal described above. It is connected (has continuity with) one pin on two different IBM silver tincans (labeled A and B). The unmarked square at the bottom is just an onboard intermediate pad connection - not an actual component. Then it also has continuity with pin 6 (labeled C below). This is....
IC12= 1589415 / 851041 ==> 74298 (QUAD 2-IN MUX)
Rotating as pictures below,
1658390800334.png

So, I'm not really sure where to go from there... These D1 and D2 (pin 6) are both inputs, right?

1658390841164.png




1658391241317.png
 
It is interesting but not unheard of for an input to behave differently if a faulty component is present. What can happen is that a fault in an IC introduces a low-resistance path between the input and ground, and this causes any signal on the input to get dragged down. If it were a dead short, then the signal would never climb much above 0V, but since we see +2V, there's obviously some resistance present still.

There is not much that seems obvious to do if the problem is in the IBM cans, so in the spirit of "looking for your car keys under the streetlamp", it might make sense to focus first on the 7429B IC.

Actually, scratch that: perhaps a good first thing to do is rule out problems on the board itself. I know you've already looked over this thing with a fine-toothed comb (and then some!), but could there be any place where a piece of debris has bridged any of the vias or any legs of any component? Could there be a small strand of wire or metal fragment trapped underneath one of the ICs? I wouldn't desolder anything to check, but if there's a way you could peek underneath or maybe give it a blast of air from a duster, perhaps that might be worthwhile. It's a long shot but sometimes you never know.

Assuming that's all OK, it's a bit of a puzzle. Just for fun, could you measure the resistance to ground on the 7492B's pin 6 and compare it with some of the other signal input pins to that chip, pins 1-5, 7, and 9?

It's not a really informative measurement: the resistance will depend on everything else that those pins connect to, plus variation inside the IC itself. I'm honestly not really sure what we're looking for: it wouldn't easily determine that a low-resistance path to ground exists in this IC versus the IBM cans. But maybe we'll think of something as we poke away at it.
 
I'm still reviewing over the "HP 10529A Logic Comparator" option. Haven't finished the full manual about it, making sure I understand how it works first (still not fully clear to me). I have to find and order good-copies of those DIPs, so that may take a little while (can't always just order just 1, seems places have minimum batch orders).

I liked the Hunton Tracker idea - just for the original models, I couldn't find a digital (PDF) copy of its manual, and the newer versions (which do have easy to find digital copies of their manuals) are a bit expensive. I do have a ATF20B function generator, which maybe is good enough to essentially replicate the Hunton Tracker features?


Not a bad idea about blast of air - especially after wiping tinfoil all over it, which snagged on pins and little bits of the foil crumble off.
 
The HP logic comparator is a nice tool, but... has a shortcoming. If the outputs of the chip are connected to a shared bus then they will show as failed. The comparator doesn't know how to handle those other signals.
 
Back
Top