Epson Geneva PX-8: LCD stopped getting -15v

[halfpress]

Hello!

I stumbled on an old PX-8 that my Mom used to use that was buried in their basement until they found it by accident a week or two ago. I've been documenting my examining and restoration of it on a thread on Mastodon if anyone is interested.

Skipping forward, there was no real or obvious motherboard corrosion despite the typically nasty looking battery covered in white fuzz and the greenish colored liquid I found between the case bottom and communications module (pictures in the thread linked above).

To my surprise, it reliably has fired up on my bench supply from the outset (I cleaned and inspected it before trying, of course). None of the caps look like they have leaked - at least not obviously, so I haven't recapped just yet.

A few days ago I spot welded four new NiMh SubC batteries and made a usable (if not attractive looking) pack and everything was running nicely without external power for a while. I was testing the micro cassette drive, successfully reading and writing files, when I glanced back over and found the LCD was blank. I thought it had locked up or gone to standby in some manner. I quickly realized that it still beeped on power on and, going from memory, I could go into the system menu and operate the tape drive, etc. So it seems to be working, but the display is no longer active.

I started going through the tech guide, read up on the LCD hardware and found the flowcharts. Right off I realized I'm NOT getting -15vdc on pin 15 of the ribbon cable to the display. I'm getting like 70 mv.

That got me digging into the circuit cited on page 2-24 of the tech guide. I see a very clean square wave of 5v at about 32kHz on pin 2 of IC 14D. Moving down to the crucial pin 7, I'm seeing a less distinct square wave hitting about 1.6vdc (maybe a bit under)? The flowchart (page 5-23) says if you see a pattern on pin 7, then just replace transistor Q29. BUT, if I am reading this right, pin 6 of IC 14D should be showing me the inverse of the pattern on pin 7 and, instead, I just see a constant, clean 5vdc there instead. No pattern, no noise. Comparing, other pairs like pins 11/12 on 14D does show me an excellent 5v square wave that is nicely inverted.

I took out 14D hoping that it was the problem, tested it in my Backbit Chip Tester (as a C4049) and it was good. I found another identical hex inverter at 9B (associated with the serial port, I think) and pulled it as well. Also tests fine. After socketing both, I swapped them JUST for the heck of it and no change. So 14D is, I assume, perfectly fine.

I'm curious if someone with more electronics knowledge than I have (and I want to learn!) could give some thoughts here. Could something downstream of pin 6 like the Q29 transistor be causing this (per the flowchart suggestion)? Feels wrong to me.

OR... is something upstream of pin 7 making the signal come in below a threshold the IC requires? Maybe as a result, pin 6 is just outputting what I assume might be reference Vss instead? The oscillator path to pin 7 goes through two resistors and a cap to get there (that's a voltage divider, no?) - so seemingly lower than 5v is expected there... but it's just not a clean looking pattern and it seems lower than I (perhaps wrongly) calculated.

This is me speculating as I try to gain knowledge and would LOVE any insights from those more experienced here and/or already having worked on PX-8s.

I'm glad to test any ideas people might offer and also glad to share screenshots of measurements, etc.

Attaching some images here to show the patterns of pin 2, pin 7 and pin 6 on IC 14D.

Very appreciative of any assistance!
- Aaron

Pin 2 showing a nice square wave at 5v:


Pin 7 showing its weaker, less distinct pattern:


Pin 6 - expected to be inversion of pin 7 - showing a constant 5v:

 

[Istarian]

Looking at figure 2-21 on page 2-26 (of the document): I don't think pin 6 of (IC 14D) is strictly the inverse of pin 7 (IC 14D), at least not in a straight forward 0/1 fashion

It appears that when the output on pin 6 drops below some threshold voltage, pin 7 goes from 0 to a fixed output voltage. And them when pin 6 rises above some threshold voltage (same?), pin 7 drops back down to 0V/Ground.

Without looking at anything else, that strikes as being part of a feedback response that could be used to drive the switching of a transistor.

That seems consistent with the description on page 2-24.

I think a problem with Q29, Q34, or the Zener diode would lead to problems for sure. If the capacitors no longer have the right capacitance that might also be an issue.

Would be nice if the schematic diagrams in the scan were a bit clearer...

PS
The document refers to a clock of ~35 kHz, not 31.65 or 32. So unless your scope is not giving an accurate reading that might indicate a problem.

+/- 3.5 kHz would be a 10% difference

It might be worth the trouble to assume that 14D is bad and just replace it with a new part. Depending on the exact chip used your tester might not be giving valid results.

Is it a 74LS, 74HC, or 74HCT part or something else entirely?

 

[halfpress]

Hello!

Appreciate the reply! I have some questions and possible answers inline.

Looking at figure 2-21 on page 2-26 (of the document): I don't think pin 6 of (IC 14D) is strictly the inverse of pin 7 (IC 14D), at least not in a straight forward 0/1 fashion

It appears that when the output on pin 6 drops below some threshold voltage, pin 7 goes from 0 to a fixed output voltage. And them when pin 6 rises above some threshold voltage (same?), pin 7 drops back down to 0V/Ground.

If so, I need to recalibrate my understanding (which is no shock - I'm learning as I go). 14D is a hex inverter, so my understanding was each pair of pins takes a signal and inverts it. So whatever is coming into pin 7, I thought, should come out of pin 6 inverted. I have examples of that on pins 11/12 and other pairs. The uses of pin 4 and 5, for instance, seem similar to 6 and 7 in that the signal coming into 5 isn't a clean square (much like pin7), but the output of pin 4 is a clean square wave inversion of pin 5.

So per the Circuit Operation text on page 2-24 and my examining other pin pairs, it seems like the output shouldn't be so much conditional as just "signal in, signal out" on each pair... and I assume the output voltage is always pegged to Vss (5v in this case)... ?

That lead to my question of whether there's any chance the voltage of the incoming signal on pin 7 was somehow too low for some required threshold in the IC, so resulting pin 6 output is just flat (and perhaps mirroring Vss reference voltage)?

Without looking at anything else, that strikes as being part of a feedback response that could be used to drive the switching of a transistor.

That seems consistent with the description on page 2-24.

Definitely a switching process going on and, overall, I think the goal is that a steady -15vdc coming "out the other end" when all of this is functioning properly. That -15vdc is definitely missing - but other subsystems doing similar things using this IC and voltage circuit are functioning ok.

I think a problem with Q29, Q34, or the Zener diode would lead to problems for sure. If the capacitors no longer have the right capacitance that might also be an issue.

Yes, I keep probing the board and trying to find a failed component - but not everything can be tested in circuit. I did pop capacitor C27 off and replace it to no avail. It's part of the signal flow being sent to pin 7. The two resistors in that path are both reading properly at 10k when tested in circuit.

Would be nice if the schematic diagrams in the scan were a bit clearer...

When I wrote this post, that was the only copy I could find that I could easily link. I've just now uploaded the PDF version I have (which is very clean and readable) to archive.org. It might not display in-browser due to size, but it is downloadable from there as of now:

Epson Geneva PX-8 Technical Manual : Epson Corporation : Free Download, Borrow, and Streaming : Internet Archive

A full, cleanly readable copy of the Epson Geneva PX-8 Technical Manual.

archive.org

It's 26 MB in size.

PS
The document refers to a clock of ~35 kHz, not 31.65 or 32. So unless your scope is not giving an accurate reading that might indicate a problem.

+/- 3.5 kHz would be a 10% difference

Yeah, this worries me a bit as well and I didn't know how much lifting "approximately" was doing in their wording.

I have a second smaller scope here and will see what it shows me as well. The fact I get a startup beep and can blindly go through the keypresses to operate menus that move the cassette drive makes me think the only thing not functioning is this display (the missing -15vdc being my pursuit). But a lot of other functionality of the system is powered by voltage shenanigans around this same circuit and same 14D IC... so I'm hoping that clock variance is a red herring as I'd expect other functions to not be working as well (but they are). Again, just me theorizing in the dark.

It might be worth the trouble to assume that 14D is bad and just replace it with a new part. Depending on the exact chip used your tester might not be giving valid results.

So there are two of that same IC on the board. I took the time to socket the other one as well. They both pass in the chip tester and I get precisely the same results on the scope with either IC in the 14D socket... so I increasingly feel like the IC is good. Therefore, I keep looking up or downstream in the circuit, testing things and trying to grow my understanding through the process.

I really appreciate your willingness to provide thoughts on this! Every bit of feedback is helpful as I try to solve this and learn more in the process.

 

[halfpress]

Quick followup on my own messages above: I've been discussing possibilities and making measurements with a very generous fellow Mastodon user who has been lending me his time and insights this evening.

In the process, I made an annotated version of the diagram from page 2-24 showing the voltages I am seeing at various points through the circuit leading to pin 7 and exiting pin 6.


I think I was overcomplicating my own theory/question about pin 7 and minimum voltage. Of course it has a minimum and it was right in my face: the square wave coming into pin 7 is peaking at 1.6v, so below the threshold for high... so, logically, pin 6 is a constant 5v since it's the inverse of what is being seen as a constant low.

Hey, I'm learning.

Now, the question is why my this is the case.

 

[halfpress]

Ok - seeming resolution to my issue here, so following up on my own thread in case it's helpful to others in the future.

Today I zeroed in on Zener Diode ZD20 (per the tech guide diagram - same one I annotated in the previous post with voltages) that is associated with transition Q34. First off, I couldn't find a ZD20 anywhere. I looked and looked. Finally, I just traced the circuit per the diagram and the diode I found in that spot is actually labeled ZD10. I'll also note that I think the C15 capacity in the above diagram (toward the bottom right) is actually C16 on the board, if my memory serves. Curious if others have noted the same?

I took it off the board and it seemed to test just fine, including me putting voltage across it with a resistor and confirming its breakdown point at just under 20vdc. Not much more I could ask for, I guess. So I soldered it back in. No change, of course (and none expected).

I then went ahead and removed Q34 so I could test it out of circuit. Finicky little SMD piece, but I managed to do it without flinging it into the nether regions of my workroom by accident. Seemed to test fine with expected 0.6vdc and open while flipping leads around between base, collector and emitter. Seemed fine, so I carefully soldered it back in as well.

With that back, I (for the billionth time) checked all the usual points including IC 14D's pin 7 that was too low of a voltage on its square wave. To my total surprise, it was now 2.8v for the first time. Moved to pin 6 and I had an inverted wave there for the first time. The pin 6 square wave doesn't look precisely inverted relative to pin 7, but it's there! So I checked pins 15/16 on the LCD ribbon connector and, also for the first time since this happened, I had -15vdc rock solid on both.

LCD works fine again now.

So... cold solder joint that developed last week? Something finicky going on in Q34 and I should expect another failure soon? Not sure... but have learned a lot and I'm up and running again... for now, at least.

Open to any and all suggestions here and advice on going forward with the unit!

 

[Istarian]

Sorry, I should probably have said something about the voltage thresholds on any digital logic.

Not all datasheets (modern ones especially) provide any explanations beyond the values of V_IH, V_IL, V_OH, V_OL, etc.

For CMOS logic (not TTL):
- 0V to 30% Vdd is considered a logic LOW
- 70% Vdd to Vdd is considered a logic HIGH
- anything in between is "indeterminate"

In that the indeterminate range, you may see results of HIGH or LOW or wobbling back and forth...

So if Vdd = +5V, the ranges would be (roughly):
- 0.0V to 1.5V -> logic LOW (0)
- 3.5V to 5.0V -> logic HIGH (1)
- 1.5V to 3.5V -> indeterminate

reference:

Logic level - Wikipedia

en.wikipedia.org

for more explicit details:

Logic Levels - SparkFun Learn

 

[halfpress]

Quick followup that it failed once or twice more, each time coming back if Q34 was resoldered. I've since purchased new replacement compatible transistors and outright replaced it today. So far, so good.

 

[Istarian]

That definitely sounds like a flaky transistor, it may be that the component getting hot during resoldering is what is helping it to work okay for a little bit.

If you have the supplies and tools, you might considering removing as much of the old solder as possible. You could have a really poor connection instead of a failed transistor.

 

[halfpress]

Definitely used braid and cleaned the pads of solder, cleaned them with iso and applied all fresh. Fingers crossed it remains stable (both this and other parts) since I'm tired of opening up the case at this point.

 

[daver2]

I would say that transistors generally don't 'fix' themselves when you resolder them.

I would start looking for a micro fracture of a PCB track or a thru-hole-plate.

This has the hallmark of a mechanical failure.

Dave

 

[halfpress]

My theory was more than the transistor perhaps had a fracture at a leg where it meets the package. Nothing in the vicinity that I examined under the scope or cleaned up seemed to exhibit any issues. But I’ll certainly know if things go sideways with it again now…