Tektronix 4054A HoLLyWoOD Syndrome

[stepleton]

Hi folks,

I have a Tek 4054A whose beam occasionally wanders and jumps around. This yields a phenomenon I call "Hollywood Syndrome" because letters you type as the beam is sliding about look abut as tidy as the Hollywood Sign. This video shows the effect well:

You can really see the cursor jump and slide around at the 30 second mark. Here's a still image of the result:



Usually the wandering is too slow for a noticeable effect on a letter or a shape mid-draw, but it can happen:



Does anyone have any thoughts about how to troubleshoot this problem? Naturally, it's intermittent, and it seems to emerge when the computer gets warm.

It's unfashionable to suspect capacitors in 2022 but I suspect a capacitor somewhere. Don't worry, though, I won't start replacing anything yet. I want to find out what's really wrong and fix it. I've been doing some reading of the 405[24] technical manual and the schematics, and I'm sure the answer can be worked out in time with those resources, but perhaps some likely suspects are known to people on this forum now.

Thanks for any thoughts!

 

[daver2]

You’ve come to the right place!

Let’s get our heads around the problem...

Dave

 

[stepleton]

PS: The second image in my post had me suspect for a while that the Origin Shifter mechanism could be the problem:



That's because the text seemed to displace itself along the 45-degree bottom-left to top-right line, which is how the Origin Shifter likes to shift the origin. But the video shows that the wandering can move in more directions.

 

[daver2]

You have just answered the question was about to ask - is it on both the X and Y axis or only one.

If it is on both - we need to find something that is common to the two.

I was first thinking a faulty preset potentiometer- but not if it affects both the X and Y.

So, how do we isolate the issue to variations in a power supply rail or something on the vector or deflection boards?

What test equipment do you have?

When things warm up does smack of a semiconductor to me though - so a can of freezer spray could help isolate the component.

Dave

 

[stepleton]

More observations about the movement: it's often "more X than Y" or about equal parts both, but it's very seldom pure X or pure Y.

Here is a photo album of the display anomalies illustrated by printing rows of the "@" character:

Tek 4054A HoLLyWoOD Syndrome

15 new items added to shared album

photos.app.goo.gl

You can see that there are other display bugs to fix, including a "residue" from imperfect clearing of the display. I'd like to address this later on, too, but I'm happy now to focus on the alignment problems. Of course I can't rule out that they may be related.

Test equipment is basically this: I have a DMM, an 2-channel oscilloscope/16-channel logic analyser, an ESR meter, plus some digital stuff (a Retro Chip Tester and everyone's favourite USB ROM programmer). I think a friend down the road has a HV differential probe plus a four-channel scope. I don't have proper freezer spray, but I do have an air duster I can hold upside-down...

 

[MattisLind]

Have you checked the voltages from the PSU? I had really bad capacitors in the PSU of a 4051 that made it draw extremly weird. Much worse than yours.

 

[daver2]

An air duster is not cold enough I am afraid.

Does your DMM or scope have the capability to record the minimum and maximum observed voltage?

You could measure each DC voltage rail in turn looking for voltage disturbances. A DMM may be too slow at measuring if we are looking for glitches.

In this case, the scope may be the tool to use. If your oscilloscope doesn’t have the ‘maths’ capability, try setting the trigger level just above and just below the dc voltage level to see if it triggers. Failing that, just monitor the dc voltage whilst playing the Tek lunar lander!

I assume you have spotted Monty’s ‘cache’ of Tektronix programs?

Dave

 

[stepleton]

Thanks, the scope idea sounds like the right next step. I'll have to identify the best place to probe some of the rails.

A few of the voltages in the machine are "unregulated", but I do seem to remember some loose bounds and perhaps also ripple characteristics noted for those somewhere. These tend to be fairly exciting-looking: +490V and +175V. These and the regulated +290V may need some thought to measure safely.

I'll start with the more tame regulated voltages, which I think I can find on the Interconnect Board.

 

[stepleton]

And yes, I have found and enjoy Monty's archive! I'm here on his suggestion in fact. A little while ago we worked together to improve his excellent Flash Drive's compatibility with the 4054A.

It might bear mentioning that I also have Jos Dreesen's recreated diagnostic ROM pack and multi-function ROM/RTC/serial printer module.

 

[djg]

Coming from 4014 I'd send draw commands for a simple line continuously non store. if that showed the problem start with tapping stuff with a non conductive rod to see if that changed behavior. If that didn't put a scope on the last stage of X or Y deflection and use the other probe to work back to see where go you from stable to unstable signal. Also check voltages to see if they jump at the same time. Verify your probes are good to voltage you're measuring.

Simple repetitive patters make troubleshooting easier. Assume you can do something similar with 4054.

 

[daver2]

>>> I'm here on his suggestion in fact.

Ah yes, I remember that now...

A good idea from djg aswell. I was thinking of the storage aspect only, but the 4054 is a different beast isn't it.

Dave

 

[stepleton]

a simple line continuously non store

Thanks for the suggestion. I'm not 100% certain of what this means, but here's what I think you're saying:

This 4054A has the "dynamic graphics" option (Option 31 in this case), meaning that I can show a line without storing it to the display. I should use this option to display a line to the screen and see whether it holds fast to the location where I put it.

I had been thinking to try and keep Option 31 separate from this diagnostic task, but as the dynamic graphics system is interposed between the computer's main logic and the display, there's a chance that it, too, could be the origin of the problem.

The Option 31 in this 4054A does have some hiccups of its own, in fact, but I had also intended/hoped to be able to confront them later. As a preview, though, here's what this particular issue looks like:

New video by Tom Stepleton

photos.app.goo.gl

You can see the refreshed characters jumping around a bit as the Option 30 system attends to changing the location of other items that it is drawing (not visible in the video). This effect settles down once the program stops changing the location of those items.

In any case, if this is what you meant for me to check, I can have this running as I monitor the voltages. But of course I have some afternoon chores to attend to as well, first.

Ah yes, I remember that now...

Right, sorry, I forgot you were on the thread too! Linking names to handles in my mind now...

 

[daver2]

I have some chores now as well...

Yes, that’s what djg was saying.

It looks like you have a few problems to solve, but that one looks definitely like PSU related (bridge rectifier or smoothing capacitor fault at a guess).

The other possibility I thought of (for your random glitch) was a poor interconnection between boards or from the power supply cables.

Dave

 

[stepleton]

I'd be ready to think of PSU for that issue as well, but I'm not so sure.

We did have to replace a shorted primary side smoothing capacitor in the +5V PSU in this machine --- at that time I did in-circuit ESR checks of all the remaining capacitors in the power supply, and all looked OK. Of course there are more voltages than +5V, and of course there's no substitute for real measurements. I hope to have those soon.

The fact that the jitter calms down when Option 30 is less busy with other graphics is also suspicious to me.

 

[djg]

Thanks for the suggestion. I'm not 100% certain of what this means, but here's what I think you're saying:

This 4054A has the "dynamic graphics" option (Option 31 in this case), meaning that I can show a line without storing it to the display. I should use this option to display a line to the screen and see whether it holds fast to the location where I put it.

Yes non store sounds like your dynamic graphics. The thought was to avoid abusing the same spot on the CRT too much. If troubleshooting takes a while wasn't sure if drawing a stored line in the same spot for a long time could cause problems.

Whats option 30 vs 31? Looks like the text was non stored since the characters distored then went back to normal without retained image of distortion. A _ or . for Y jitter or | for X jitter may also be a good simple test case if it shows the problem.

 

[stepleton]

(No voltages yet --- still working on chores!)

Sorry for the nomenclature sloppiness. "Option 30" and "Option 31" are the same thing for the purposes of this conversation, I think. Both are "dynamic graphics", but as I understand it Option 31 entails a different CRT that shows the dynamic graphics in an amber colour, while with Option 30 it would be green.

The amber text in the last video I've shared is indeed non-stored; the green text in all other places is stored.

My plan for the comparison you've called for is something like this: I'll draw a cross twice in the same spot: in storage graphics and in dynamic graphics. If the registration is correct, they should overlay and produce a yellow colour, but if they drift, then predictably you will see a green cross and an amber cross.

I may actually try drawing several crosses: one at each corner of the screen and another in the centre.

 

[daver2]

Sounds like a good plan to me.

Dave

 

[stepleton]

With apologies for the delay. I wound up needing the rest of the day elsewhere, and I still haven't measured everything I'd hoped to measure.

My results are inconclusive except in one respect. I am pretty sure it's a heat issue. My reason for saying so is because I can run the thing for quite a while with the cover off and, it seems, encounter no trouble.

I've only been able to measure some voltages --- access to the ones most directly related to the screen (I assume) is a challenge, and that's just for the sake of making contact at all. Actually clipping leads to voltages so that you can detect transients on a scope seems even more difficult still. I'll see if I can think of something. Thankfully, the heat issue may be bringing us closer to a different diagnosis.

I can at least relate what I've observed so far. +5V out of the logic power supply seems steady, as do +5V, +15V, and -15V on the interconnect board. My oscilloscope shows a healthy amount of ripple on all of these channels, say on the order of 80mV peak-to-peak (much more than what Tek wants!), but given that the ripple's magnitude and frequency seems roughly the same on most all the rails, I'm willing to attribute it to my instrumentation to some degree. These voltages themselves appear to be within their tolerances.

The only obvious anomaly is the unregulated +20V and -20V supply going into the I/O board, both of which have values exceeding +25V and -25V respectively. I suspect that this may have to do with the AC mains here being on the high side: it's about 248V. There's not really much to those rails: the transformer, the rectifier, and some capacitors, and I doubt there's much other explanation than the mains for why the voltage is high. But there is much I don't know about many things.

Supposing we forge on under the assumption that some part is especially heat-affected or marginal... what next?




One last observation. I did make the program that I described earlier --- here is a screenshot:

New item by Tom Stepleton

photos.app.goo.gl

You can see the residue on the screen from imperfect screen clearing, but again, that's probably a topic for another time. The registration of stored and dynamic graphics is pretty good, if not perfect in the top-left corner. Regardless, with the cover off as I've had it during these experiments, there is no drift to speak of... mostly.

I suspect this is also unrelated to the problem we've been discussing here, but if the screen has a lot of stored graphics on it already, then the area addressable by the beam appears to shrink. It's not a lot for this 19" tube: a fraction of a millimetre at each corner, but it is noticeable.

What that means with the test program shown above is this. If I run the program, the registration is fair. The green stored graphics and the amber dynamic graphics overlie each other pretty well. However, if I then fill the screen with "@" characters, then the amber crosses "slide away" slightly off of the tops of the green crosses and towards the centre of the screen (except for the one at the centre, which doesn't move).

Here is a close-up image of the so-so registration at top left under normal conditions:

New item by Tom Stepleton

photos.app.goo.gl

Here is a much more ideal registration at the centre:

New item by Tom Stepleton

photos.app.goo.gl

And here's the top-left corner after I've filled the screen with "@" characters:

New item by Tom Stepleton

photos.app.goo.gl

As further evidence of the problem, you can try refilling the screen once again with "@", and you'll find that they don't manage to rest on the original "@" characters: the screen area where the beam can write has been contracted. As a final demonstration of this phenomenon, you can take this video tour of the four corners of the screen (please excuse the CRT's burn-in!):

New video by Tom Stepleton

photos.app.goo.gl

This could be something that can be remedied via the calibration procedure, which I'll have to try and undertake sometime before long.

 

[stepleton]

I'm continuing to explore the Hollywood heat hypothesis, and right now this means waving a thermal imager around the computer and seeing what I can see.

There are a few toasty connectors, voltage regulators, power transistors, etc. here and there, but I found a pair of (what I'm virtually certain are) NTC inrush-limiting thermistors that take the cake, or that could bake a cake:



I know NTCs work by getting warm, but above boiling seems a bit too zesty to me. They're on the storage board, and it looks like they sit along the flood gun filament power supply. I'm not sure it's related to the heat problems I'm interested in on this thread, but even so... is this normal? 105 degrees C?

Looking ahead, I'm thinking to try discovering our secret Hollywood star by temporarily insulating portions of the display electronics card cage bit by bit... as long as I don't do this for too long and cause unrelated damage, I can identify the card with the problem component by warming cards up and looking to see if the problem recurs.

 

[stepleton]

We may be on to something now. A friend and I hypothesise that the problem might be down to faults in the +15V and -15V power supply, since that's what's feeding the deflection circuitry. What's more, the way the instability feels --- the gradual slides, the big jumps all at once, the lack of predictability --- it just kept whispering "capacitor, capacitor" in our ears...

We opened up the schematics, and there's really not all that many components on the "Low Voltage Power Supply" board's fifteen-volt rails --- they're fairly simple linear power supplies. But we're not really eager to take measurements: the board is buried deep within the machine, and among the "low voltages" it provides are +290V regulated and +490V unregulated. It's an awkward place to be poking around.

So we fabricated a highly specialised diagnostic tool and used it to carry out a sophisticated hardware debugging technique:

The recipient of our generosity is C145, a 47uF electrolytic that seems to be the only bulk capacitor on the +15V rail. Sure enough, this treatment turned a basically steady display into this:



Bothering other nearbly components in a similar way seemed less effective at bringing us views of Hollywood, so it seems worthwhile now to try to extract the Low Voltage Power Supply board and extract C145. If it turns up faulty, then it is easily replaced. Unfortunately, the video shows that there are a lot of wires in the way, and even the Tektronix manuals (whose calibration procedures have you undertaking improbable voltage measurements and moving jumpers that are exceedingly difficult to access) say that getting this board out is a pain. Oh well!