Smoking IBM 5154 EGA Display - intensity problem

[NickyDoes]

I picked up another XT 5160 a few weeks ago. The XT had a bad tantalum cap on the MB. Thanks to @modem7 awesome site, that was an easy fix. The config is pretty common: 64-256K motherboard + a Quadboard. The rest is IBM. I have a few brick-sized hard drives from Seagate and Miniscribe, but this one came with an ICI, which I hadn't seen before. Better still, it works.

On to the salacious title, which is accurate - the monitor's power supply Y class caps smoked shortly after powering up the monitor. Specifically, they smoked when I switched to graphics mode for the first time.

Before the caps smoked, I was able to use the monitor for a bit. Regular text starts out moderately bright, while high intensity text is proper bright. After 2-3 minutes, the regular text fades to be unreadable. It stays white (gray). The brightness and contrast controls work normally, including pull-to-activate on the contrast control. Text always stays white, and does not lean toward a color.

Some of the electrolytic caps had leaked and had corrosion under their shrink film. I replaced all + the Y-class film caps. Power supply filter caps did test bad, so parts were exposed to ripple.

I freeze-sprayed quite liberally, and saw no change at all. Given that the display starts moderately ok, then deteriorates, I expected some thermal relationship, but haven't found one.

The first image is of FreddyV's Picomem boot screen, which many of you have seen firsthand. Notice the ellipsis is high intensity, whereas the rest of the text is quite dim.

The second image is a CheckIt menu. By now, the entire top portion should be a pull-down menu, but it is unreadable. Only the bottom line is readable - in high intensity white.

I'm finding this to be an interesting problem. I have the SAMS for this monitor. I expect to have time later this week to trace through and understand more functions of the board. I'll be sure to post updates if I learn something interesting.

Edit: I have also tried a 2nd IBM EGA display adapter in an attempt to rule out the intensity signal coming in. That adapter has RAM errors, so shows a pattern of backslashes. Those backslashes are also bright and dim.

 

[vwestlife]

EGA has secondary RGB pins to make the colors darker. Maybe something is shorting those pins high or low?

 

[MiaM]

There is also a circuit that makes only one of those pins be active, controlling all three colors, if the monitor is in the 15kHz CGA compatible mode.
(This limits the palette to 16 colors in 640*200, even though an EGA card itself can output 64 colors at any mode).

 

[NickyDoes]

That's an interesting observation. The 5154 EGA display uses a frequency discriminator to monitor H-Sync: ~15.75 kHz for CGA, ~21.85 kHz for EGA, and sets a mode line.

In CGA mode, four signals can be low or high: Red (3), Green (4), Blue (5), and Intensity (6) can be low or high. Intensity intensifies R, G, and B for that pixel. Possible pixel colors are 2^4 = 16.

In EGA mode, 6 signals can be low or high. 3, 4, 5 are still RGB, but are now "primary RGB" for regular intensity. Pins 6, 7, 8 are now: Red Hi (6), Green Hi (7), Blue Hi (8).

Pin 7 has a different job for CGA vs. EGA mode. I wonder if the R-C discriminator is not discriminating.

Then, there's the slow dimming of the low intensity pixels after the monitor is powered. Maybe these are two different problems. I'll have to think if this makes sense, and how to devise a test.

In CG

 

[Eudimorphodon]

The 5154 EGA display uses a frequency discriminator to monitor H-Sync: ~15.75 kHz for CGA, ~21.85 kHz for EGA, and sets a mode line.

I don't think the EGA uses frequency *per se* to tell the difference between 200 and 350 line mode; the monitor switches modes based on the polarity of the vsync signal; it's positive-going in the 200 line mode, aka Mode 1, negative-going in Mode 2. I suppose you might still call this a "frequency discriminator" but the choice is made on the basis of the percent of time vsync is high/low, not the timing of hsync. There's a SAMS manual out there for the 5154, but A: the goings-on beyond the input jacks of CRT monitors are scary voodoo for me, and B: the manual doesn't have any explicit explanation of how the monitor chooses modes. This is going to sound really stupid, but I wonder if it's not a lot more complicated than for the active area it's in "mode 2" when Vsync is high and "mode 1" when Vsync is low. (in terms of how it interprets color.) Someone should page @Hugo Holden , this is right up his alley.

Do you have BASIC (gwbasic/basica, this would be a lot easier to use for this test if you're having trouble seeing things than qbasic) on the machine? After booting can you get into gwbasic and type "SCREEN 7" (or SCREEN 8)? That will put you into 200 line mode, you can see if the color interpretation changes compared to the 350 line mode you're getting into on boot. (Or, alternatively, you could set the switches on the EGA card to lie and say you have a 5153 attached, that'll give you a 200 line text mode at startup.)

 

[modem7]

In case it becomes relevant later:

The mode1/mode2 (CGA/EGA) detection circuitry is the 'integrator' in the centre of the diagram at [here], a resistor, a capacitor, and an exclusive NOR gate.

The resistor and capacitor form an RC integrator.
- The positive-going sync waveform is mostly at LOW (i.e. the average voltage is LOW), and pin 8 of IC505 will see a LOW.
- The negative-going sync waveform is mostly at HIGH (i.e. the average voltage is HIGH), and pin 8 of IC505 will see a HIGH.

The resulting signal (mode1/mode2) (CGA/EGA) out of IC505 then goes off to:
- The ROM for colour interpretation; and
- Various circuitry to cater for the differing sync.

 

[modem7]

One good basic check is to see if the 5154's PSU output voltages are close to what they should be.
The diagram at [here] shows the voltages for my functional 5154.

( Safety warnings at [here]. )

 

[NickyDoes]

@modem7 Good point out. Thanks for that. The DC power supplies are all ok.
I left the monitor on for hours. When I came back to it, it is working much better. I don't have a good explanation - I replaced all the electrolytic caps with new ones.

Better means I can get a reasonable display of low and high intensity pixels in both sync frequencies. The geometry is good - size, centering, linearity, pincushion. The colors are off. The SAMS guide on adjusting is written for someone who is familiar with CRTs and their adjustment. There are cutoff pots for all three colors, and MSB and LSB for green and blue.

I can keep adjusting them until I find a good picture across all brightness and contrast values. I would be grateful for a more scientific procedure.

 

[MiaM]

If the colors are off as in the wrong hue, then you might just have accidentally had a magnet too close to the CRT? If so you could for example take any beefy enough mains frequency transformer (say a large "wall wart" that isn't switched, like the PSU for a 1990's printer or modem) and move it around close to the front of the CRT until the tint is corrected. Always move the transformer away from the CRT before powering it off though, as otherwise you might end up with the same problem again.

 

[NickyDoes]

@MiaM as in 'degaussing' the CRT? I don't think that's the problem. The color intensities are off, and probably a bit of convergence. Focusing (pardon the pun) on the intensities, I'm not sure if I should perform the adjustments with the brightness and contrast full, low, middle? Is there a clear preferred CRT test pattern program? So far, I'm just using CheckIt.

 

[NickyDoes]

It is behaving well enough that I will put it together. Thank you everyone for your support.

The monitor is in beautiful condition, inside and out. It shows no evidence of hot/burning at the video amplifier, or anywhere. It was repaired in 1991 - a power supply fix, evidenced by the cable ties where the rivets were.

This is my first color EGA monitor ever. When these PCs were contemporary, I had a Magnavox branded CGA monitor that could switch to composite video. It was my PC monitor for schoolwork and play, and my TV for entertainment.

 

[NickyDoes]

Here are some photos. Notice the EGA card's video memory isn't healthy. It's the better of the two I have, though.

 

[MiaM]

@MiaM as in 'degaussing' the CRT? I don't think that's the problem. The color intensities are off, and probably a bit of convergence. Focusing (pardon the pun) on the intensities, I'm not sure if I should perform the adjustments with the brightness and contrast full, low, middle? Is there a clear preferred CRT test pattern program? So far, I'm just using CheckIt.

In general you shouldn't have to adjust things unless you have replaced parts that affect the adjustments.

As it seems to be getting better it might had been an electrolytic capacitor that needed reforming. If it's in a circuit where only a rather low current is pushed through a leaky capacitor then it has a decent chance at just reform itself without damage and then just start working. However if it's for example a smoothing cap for a power supply / rail then it's way more likely that it goes bad if it needs reforming, as there is nothing limiting the current (and thus it might overheat at the particular spots needing reforming, causing damage).

I would say that the picture looks good enough that I wouldn't do any work on the monitor at least until having correctly working EGA cards.

Note that for testing purposes you can use it with a CGA card, obviously only with CGA display modes. That would be good enough for testing 16 color text mode.

Also, in general before VGA it was very much a personal preference on how to set the knobs and thus people might not agree on what is a "correct" picture. In particular I think that the normal intensity colors are way too dark in almost every VGA or newer display.

 

[modem7]

Edit: I have also tried a 2nd IBM EGA display adapter in an attempt to rule out the intensity signal coming in. That adapter has RAM errors, so shows a pattern of backslashes.

IBM EGA cards.

Notice the EGA card's video memory isn't healthy.

If that is the case, then the draft information at [here] may help you.

 

[andy]

Slightly off topic, but I was repairing a 5154 a few months ago, and I was tearing my hair out trying to get good color balance. Every time I adjusted the RGB drive, and bias controls for correct neutral white, or dark gray, the other end of the brightness scale would be tinted. I was at the adjustment limits for the controls, and it still looked bad. By luck, I connected it to another system, and things looked different. I started from scratch, and everything adjusted like it should. It turned out that my video card (Everex Viewpoint) was the problem. I still haven't looked at its TTL output, but I suspect some of the signal levels are out of spec. You wouldn't the video card on a TTL monitor could cause that issue, but it did. My point is that, I wouldn't judge the monitor until you have a known good video card.