Trying to figure out if a PLASMA/Neon display/driver is broken (74 series logic)

[shapirus]

Posting, since there's no other way to follow.

I want to see this beauty lit up.

[pcprogrammer]

And as someone who seems to have ADHD, I know that I'm obsessed with getting this to run only while the parts for this are splurged all over my apartment's floor. The moment I clean it up into a box or something — like that Nakamichi I was posting on this forum a year or so ago — I will forget about it's existence for an eternity until I need to free up space for something else and be like "Oh... Well that was one stylish waste of money"

My problem is more often that my brain is constantly coming up with new fun projects causing the ones I started to be lying around for long times. 

I did to some extent. While I don't have any idea on how to identify glass diodes, there is one huge polar EL capacitor across that line, which explains the polarity quite clearly (I don't think they designed it to explode :-), and it's also rated at 250V. The ceramics in that area also have no underline under the value code, which means they are rated at 500V (all the smaller ones have an underline). There is also a 2SC1473 on the drive board in there, the purpose of which I don't yet fully understand, but it's also rated at 250V. So I think if not 40-60V then 100ish volts might be that probably. The logic ICs that input into the display's side boards I expect to be isolated by the Mitsubishi module, as we've discovered in the schematic above.

With this you at least know to keep it below 250V. 

In the one I found with problems the voltage difference is from -125V to +62V so 187V.

For it to work it is most likely that the drive board has to be loaded with some data, but maybe the ram will startup with random bits and you will see some nice glow on the display when the voltage hits the sweet spot.

I want to see this beauty lit up.

Me too. 

[factory]

Hmm, if this is a plasma that would explain the lack of a heater. This could have a connection to how most LED displays in Japanese trains are orange, being an homage to their plasma predecessors? (even though sadly they're being replaced by LCDs nowadays) But in that case transistors capped at 15V make no sense, neither the usage of PC standard molex connector. Then again, when did the PC caome around? In 1984+, after the Macintosh? Given the ROM was labeled 1983, they might have not yet known what the connector would mean for us here in the future indeed.

But also given this was likely in an automotive setting, I don't expect the other power rail to be less than 12 and more than 24 volts.

I'm too drunk today and I have to run an event tomorrow, so maybe next week I'll get some schematic drawing done.

That being said I have a board off a 1970s cash register which has a whoppin' Panaplex display, and a bunch of similar looking Mitsubishi modules... which I couldn't get working either... let me dig it out and compare next week as well.

One of the boards is branded "Morio Denki", a web search reveals they make passenger information displays for trains. And a little later than your board, in 1988 they started making three color LED versions. http://www.morio.co.jp/english/201product-eg.html

I don't know about Japanese trains, but over here nominal 24Vdc or 110Vdc supplies are commonly used for the electronics, including the destination & passenger info displays, internal DC-DC converters changed this to whatever was required (or resistor dropper & Zener based supplies in the ancient stuff).

We had LCD displays over here in the past, backlit with fluorescent tubes, with inverters for the lighting.
I have a few of the unused LCD panel sections in my collection of junk, I bought from work when we stopped repairing electronics. The newer displays are LED.

David

[akasaka]

We have somewhat of a breakthrough!

While waiting for my bench PSU and HV source, which should be here Wednesday or Thursday, I've decided to turn my attention to another project.

Now, during a long boring meeting in Zoom, I've desoldered a nice Panaplex display from another scrap board I've had. I've attached a picture of that to the message :-)

I was eager to try it though, and digging through my parts bin, I've found a bridge rectifier and an old EL wire inverter, that takes 12VDC and turns them into 130VAC. One full bridge rectifier later, that turns into 160VDC. Which I happily solder to some pins on the panaplex I assumed to be the anode and brush the other wire on what I think is the cathodes. No dice. Well, the Matsushita catalog says this panel requires 200VDC to light up, so I'll wait for my HV supply board.

And then I thought, what if I do something stupid... I soldered the negative to one of the vertical pins (where the transistors are) on the big display of this thread. Power on. The inverter starts whining. Well, now I either get a revelation, or a bunch of magic smoke and a reason to abandon the whole project.

Swiping the positive wire on the horizontal pins (where the weird modules are)... And I see an orange dot traveling down the screen!

I've recorded a video in excitement right away:

 

So it is, indeed, over 100 volts on that pin of the Molex. Now I'm not sure if I can join the negative of the bridge rectifier to the ground of my primary switching power supply (it's an ATX PC supply), in order to have the two voltages with the same reference ground. Given it's an inverter and I don't know it's circuit, I'm worried about damaging the PSU. Otherwise I would be able to test it with the controller board — although I doubt the tiny EL wire transformer will have enough juice to push this huge panel through.

And, I guess, it's not a VFD indeed!

Thanks to everyone who helped, I'll post more once I get it running with the controller.

P.S. Whatever we've been calling Mitsubishi modules here — it's actually Matsushita modules :-) (aka Panasonic, National, Technics, you name it — the three triangles thing is their very old logo)

P.S.2. I tried merging the grounds of the inverter and the PSU, which gave me a +208V rail if my DMM is to be trusted. Connecting that to the driver board makes a noise of the EL wire struggling, but nothing is on the screen. Trying to randomly short control pins of the data port to the ground doesn't do much either. I'll wait for something less jank (the HV supply off Amazon) to check it for longer as I'm afraid 208V is too much and something like 160V would be reasonable.

But also not sure if I should make my entirely own control boards, utilizing the full 192x32 screen rather than the 96x16 one. Then just scale text fonts to use 2x2 pixel groups, but make icons in full resolution.
Hard to imagine how to drive such an enormous number of rows and columns in multiplex.

[inse]

Cheers!
I am sure its the Mitsubishi logo, Matsushita is triangular too, but different.
By the way, EL wire needs AC drive voltage, so your series connection will deliver an odd supply voltage.

[pcprogrammer]

 

Wow, you are more daring than I am. 160V brushing along the pins on the display, but it is clear that it works and displays in orange. Right side for the top pixels, left side for the bottom pixels.

Making new boards for it could of course be done, but controlling 160V needs some proper design, unless the boards connected to the display still work then it would be making the matrix lines go high or low in the right order. As stated before a ATmega2560 should be able to pull it of.

 

[factory]

I agree, the hybrid modules have the Mitsubishi logo, some of the 74LS ICs on one of the logic boards have the Matsushita logo, the difference should be obvious.

David

[akasaka]

Making new boards for it could of course be done, but controlling 160V needs some proper design, unless the boards connected to the display still work then it would be making the matrix lines go high or low in the right order. As stated before a ATmega2560 should be able to pull it of.

The boards connected to the display are grouping the rows and columns it seems, so that way I cannot utilize the "subpixel" resolution of the panel. If not doing that, then of course reusing the existing ones sounds much more interesting. But also I don't think if I feel like figuring out a spaghetti with 38 chips worth of 74 series logic, or instead having a crash course on HV driving.

[pcprogrammer]

That "subpixel" thing is true. The way they did it, is that in a square both top and bottom are turned on due to the three diodes connected to a single drive line?

But indeed HV driving is also a thing. Looking for driver IC's that can handle the 160V will be a bit of a search.

It is interesting though.

[shapirus]

But indeed HV driving is also a thing. Looking for driver IC's that can handle the 160V will be a bit of a search.

Level conversion is fairly easy to do (but may require a lot of transistors, one per output pin), if no HV driver can be found.

[akasaka]

I wonder how the columns are grouped — not by controlling multiple by a single transistor, so I wonder if the columns are already separate. Then if we can find out how the transistors are multiplexed, the column boards could be reused — and 32 bit row boards would be easy to do with something like Microchip's HV5622.

[PA0PBZ]

The problem when using all the dots is that the distance between them is not the same, but how that would look in practice I'm not sure.

[akasaka]

That would allow to create virtual shades for the cells by using dithering, pretty much that's the only benefit

[pcprogrammer]

I wonder how the columns are grouped — not by controlling multiple by a single transistor, so I wonder if the columns are already separate. Then if we can find out how the transistors are multiplexed, the column boards could be reused — and 32 bit row boards would be easy to do with something like Microchip's HV5622.

A matrix based on emitter and base control by the looks of it. ~17 groups of 12 transistors. Try to follow a base to the drive board and see if it is connected to a TTL output and do the same for the emitter groups.

[akasaka]

So, I tried to connect everything together and run it with my janky 200V supply for now. But it seems like it's not powerful enough — the voltage on the HV line stabilizes at 116V, which is probably too low for the panel, as I don't see anything, unless the digital side is blocking it.

That being said one SN74107 is getting pretty toasty — they are all warm, but most seem to be within 42ºC, one goes as high as 55ºC or so, hot to the touch. That is while there is no activity on it's pins whatsoever, and no matter whether the panel/HV are connected or not. I wonder if that might be a dead chip?

Will also post the waveforms that I find suspicious later today.

[akasaka]

So, a few suspicious waveforms...

The nibble select on the input mux:



This is somewhat "squashed"? The dashed line on the top is the 5V level for reference, and the center is 0V.
Or is that normal for a discrete logic circuit to do that when the signal is "inhibited" e.g. by not being let through an AND gate? I've never worked with discrete logic closely before.

One of the 2114 SRAM output lines also look kinda weird:



The rising edges are somewhat slow it seems?

This does not look like real logic signal to me entirely 



If anyone has more experience with discrete logic, your input is much appreciated!

[pcprogrammer]

Ai, the first signal looks indeed suspicious, even though the edges are fast, the low is a bit high and the high is very low. What are the IC's connected to this signal.

The RAM output at least reaches high, but seems to be capacitive loaded. Might be that it needs additional pullups. The third one is even weirder with the two steps in it.

Did you figure out the transistor connections yet?

[PA0PBZ]

If this is supposed to be TTL logic then 1 and 3 are definitely broken, the high should go a lot higher and the low (picture 1) should be lower, in my experience it should not be more than 0.5V from the rail. The SRAM picture looks fine to me, I've seen this a lot of times on memory lines in working equipment.

[akasaka]

If this is supposed to be TTL logic then 1 and 3 are definitely broken, the high should go a lot higher and the low (picture 1) should be lower, in my experience it should not be more than 0.5V from the rail.

Given the era I wonder if it might be CMOS. But also I wonder if it might be a "blocked" signal "leaking", e.g. imagine an AND gate where one input is this signal, and the other is 0 to "block" it, so the output kinda resembles the signal anyway but not enough to go into the definitive high logic level.

The power supplies should arrive tomorrow so maybe guesswork will show some result too. Should have tossed some buttons in the order now that I think of it — the data bus is covered by two hex spinny things, but I only have bare pins for the control signals.

For now I am currently trying to draw a schematic of the drive board, and this is anything but easy...



Though I wonder if this bit means that the ExtCtl1 pin of the input connector is something like a "transfer request" pin? (The ExtInEnable signal goes to the OE of the switcher that switches the nibbles of the input data bus onto the main 4-bit bus)
Seems that it would also require the RAM to be unwritable at the same time though. But then the RAM would interfere with the input signal, given it's #CE is always to ground?

I need sleep I guess

[pcprogrammer]

The designer most likely made a setup where there is some dead time either between each data fetch or in a dead period at the end of each line or full screen.

With buffers like 74244 or 74245 they can separate the buses in these periods. Timing depends on how fast the display needs to be handled and how fast the RAM is.

[akasaka]

OK, I tried with the proper HV supply, I can see when the thing is assembled there is some tiny glow behind the top right corner of the screen visible area — I assume that's some keep-alive to make the thing trigger pixels faster. I also assume it must be scanned in rows, but there is just a single dot and that's it. Also found the blanking logic gate and shorting that out makes some pixels appear in the rightmost two columns, but that's about it. Either there must be something to make it start scanning horizontally, or something's dead!

Either way, I've looked at the pinout of the display connectors on the display board and came back with this:

Left/Right board:
1. TTL Data
2. TTL Data
3. TTL Data
4. TTL Data
5. TTL Sync? Outputs from 74LS14 on the Left board, directly linked to the same pin on the Right board, where that weird transistor circuit is
6. TTL Vcc
7. COM Ground
8. HV_ High Voltage supply
9. ? Blanking? (NC on Left, via Zener to GND on the mainboard for the Right)
10. ? Just linked between L and R

For the top/bottom boards it's a bit harder. There are 3 connectors named P, Q and B.
P12 is the +150V supply, Q2 and P11 are GND. Q11 is NC.
P10 is linked to Q1 and both are linked to pin 9 of the LR boards, so something related to blanking as well?

All other P and Q lines go to a positive side of a diode for each line. After that, on the negative side of the diode there is a pullup to 5V, and then it goes to a pin on an Open-Collector BCD decoder, SN74145.

All of the B lines are linked in pairs, each pair has a pullup and goes to another Open-Collector inverter, SN7416. There is a leftover pattern for another SN7416 and pullup array, but they are not used.

Questions, as always: if I choose to throw in the towel on the original board and switch to an Atmega or something, would just putting SN7416s in front of the Atmega pins, as well as the diodes where they were — basically copying the original circuit up until it becomes TTL — be enough to protect them?

I will try tomorrow to run the thing without an original board using this info, and also have a look at those chips with the weird waveforms — if they are also those I found to be suspiciously hot, I'll have to order replacements I guess. There is zero response from the board no matter what signals I input on the external data bus and I don't think that's expected, especially given there seems to be no scanning going on either.

Upd: another approach I might take before giving up is to just go through all the logic chips one by one with a scope to see if anything too obvious stands out. Maybe just a loose gate not gating properly or a trigger not triggering somewhere is wrecking all havoc.

[pcprogrammer]

Sure mimicking the existing TTL outputs, either open collector or push pull including the diodes, to the display should protect the Arduino pins to some extend.

Since you succeeded in making single pixels glow with the high voltage you can try a static test first. Connect the display up to the 150V in the way it is supposed to, and use 5 volt to control the "data" pins.

The 4 data lines on the left and right display board select the Y pixel, so making one X pixel high on the base and low on the emitter should show a pixel on in that row. With the left and the right two pixels can be done in the same time. One up top and the other down low.

Then changing the base line should move the pixel in the X direction. Changing the emitter group should give a bigger shift.

[akasaka]

OK, I tried doing something and it didn't light up, maybe one signal or something was missing. Either way I tried brushing the tube pins directly again, without realizing that my proper HV supply can give up to an amp at 160V, compared to my old EL wire inverter which barely gave a few mA. A loud bang ensued, and now there is a scorch mark on one of the pixels inside the tube 
Otherwise it seems to work fine, even the affected pixel lights up (after promptly adding a resistor to the test setup), just there is now a big chunk of charred something obstructing it.

Before I break it even more I turned my attention back to the logic board. I hooked it up to the bench supply, and it draws a whooping 0.85A, which is like 25mA per chip on average, and I think that is a bit unreasonable for 74 series logic. So I went over the whole thing with a scope — aside from suspicious signals here and there all devices seemed to be fine, inverters invert, nands nand, yada yada. But the extremely hot 74107 trigger went over 60º during the debug time, the ~1Q out was too low, and 1Q too high, 1CLR squashed into the middle and 1CLK kinda low too.

I turned it off and noticed the 1J input is shorted to the 1Q output. 
Haven't taken it off yet, but with a strong flashlight I don't see a trace doing that, so it's clearly inside the chip — no other 107s on the board do that.

If that's the case I'll buy a 107 from a guy I know to carry known good NOS parts and fingers crossed that makes it run!


Completely different sidenote, the planar nixie I've shown before won't light up even after giving it 50V more than the spec. Is there a way to find out the pinout using a multimeter? Or find out if it's broken somehow? I tried dunking it in (distilled) water to see if bubbles rise from the vacuum being compromised, but no, it's totally sealed.

The only piece of spec I found for it is this catalog entry, no datasheet or anything anywhere.

[pcprogrammer]

OK, I tried doing something and it didn't light up, maybe one signal or something was missing. Either way I tried brushing the tube pins directly again, without realizing that my proper HV supply can give up to an amp at 160V, compared to my old EL wire inverter which barely gave a few mA. A loud bang ensued, and now there is a scorch mark on one of the pixels inside the tube 
Otherwise it seems to work fine, even the affected pixel lights up (after promptly adding a resistor to the test setup), just there is now a big chunk of charred something obstructing it.

Whoops. 

But hey, that is how we learn. 

Hopefully the driver starts to work with a new 107.

[akasaka]

I think I might have found something in common on all the chips that have weird signals. Look at this:



This is another suspicious 74107, which is not shorted, but measures 2k\$\Omega\$ between 1J-1Q (same as the other one which had a dead short) — the rest of the 107s have a few Megs between those (which makes all the sense, given it's input to output).

The solder on the Gnd pin is slightly darkened, and there's a lot of it making like a tiny mountain. All other supposedly working chips have equally looking solder joints on all of the pads.

The 107 I've removed looked like that too, but I didn't pay too much attention to it — solder is solder who knows how it flows  and now it seems that the current at some point was so high that it melted the solder and caused it to flow down.

I've found a shop which sells really old and odd components NOS in here, so I'll go ahead and take out the suspicious chips, confirm the issue is in those chips and buy new ones as well as sockets for them (in case it's just an easy to kill design):

* 107-1: dead short in to out (1K to 1Q, pin 4 to 3). Recorded symptoms from my logbook: HOT!, output level ~1Q too low, 1Q too high, CLR and CLK squashed towards 3.4V.
* 107-2: 1K to 1Q is 1.3kOhm, 1K to ~1Q is 4kOhm. Symptoms: output 1 signal low, output 2 is nonsense mismatching the datasheet.
* 107-3: low resistance in to out (1K [or 1J, they're parallel on the PCB] to 1Q or ~1Q), is 2k\$\Omega\$ in both cases.
* 393-1, 393-2: 1A input to Vcc is 2k (2A, or 1A on 393-3, to Vcc is nearing MOhms)
* 32-1: 1A short to 1B (Needs checking, maybe a short in 08-2 instead — but 08-2 doesn't have the Solder Mountain Syndrome), 3A has a weird pulse, 4A is also weird
* 04-2: 1A kinda low, Vcc to 1B is 13kOhm needs checking, has the solder mountain too, but 1B is an output to the "blanking" transistor base via a resistor, and base to 5V resistance in total is around the same...

Others so far seem to be fine... but what the heck happened to this thing?
Also I guess I need some more solder wick or some other tool to take out the chips 


Oh and to add about the 8080 board... The PSU was going into protection because I determined the input polarity wrong 
Before hooking it up to my bench PSU I looked at it more closely and oh shoot, the capacitors are backwards! Wait a second... 
Not that like it's of any use with the dead ROM (the ROM was dead even before I powered it up for the first time, so unless someone did the same before me it died of some other cause. There are also pretty toasty chips on there too, despite the CPU clocking and all that now)