HP 3455A: debugging a dead 7-segment display

[0culus]

So despite the kindness of a forum member in sending me three nice replacement seven-segment displays, it appears that the problem lies in the driver circuits and not the display itself. The display that is nonfunctional is DS2 on the schematic.

I've attached a screen grab of the relevent part of the circuit diagram (page 8-171 in the 3455A operating and service manual). I'm trying to grok how the display operates. U62 and U71 appear to be unique parts on the board, so it doesn't look like that feature on the schematic is duplicated for each display module, which is what I originally thought. How does this work? Additionally, I'm suspicious of Darlington pair Q16, which appears to play a role in turn the dead segment on.

It looks like the TTL logic chips are still available from TI, and there are some places online that have the transistors.

[cncjerry]

I have all those parts if needed for the cost of shipping.  I would need time to desolder from the carcass so if you can get them from mouser, that would be a better option.  If DS2 is out, the problem is most likely in that driver Q16 though it could be the LED, most often they fail by segment.  Most of the time the problem is the cable, though.

Let me know if I can help.


Jerry

[macboy]

Please post the HP part numbers for the ICs and transistors. Maybe I (or someone else) can cross-reference them to standard parts.

This looks like a very standard digit scanned display. All segments of one digit are driven at any time, and the system simply repeatedly scans through all digits fast enough for your persistence of vision to create a steady state display of all 8 digits.

The digit to be displayed at a given instant is determined by U56, operating as a 3-to-8 line decoder, for which a single output is driven LOW. A quick google reveals that the 74xx155 is a very likely candidate for this, as the pin-out and functionality match well. It is a dual 2-to-4 decoder that can operate as a 3-to-8 decoder.

The LOW output for the digit to be displayed will turn on the darlington transistor, providing +5 V power to the common anode of the selected digit (for digit DS1, +12 V). The individual segments are driven by the combination of U62 and U71. Realise that even though these drive the segments for all digits, only one digit is powered at any given time (via U56 and the darlingtons). U62 is a 4-bit BCD to 7 segment decoder, probably with open collector outputs, since the NAND gate U46c is wired directly to one of the outputs (U46c drives the center horizontal segment of the display "-" when a code of 0xC to 0xF is present). For each 4-bit code between 0 and 9, U62 drives the appropriate segments to create the numerals "0" to "9". Since the +/- digit DS1 requires 12 V power, then U71 must have open-collector outputs. U71 is likely a darlington transistor array, or a bus driver with OC outputs.

You are right to be suspicious of Q16. Check whether its output even goes to +5 V like the others do. Also check that pin 10 of U56 ever goes low, and check R65.

[cncjerry]

Given the POV as macboy mentioned, you should put a scope on it to see if it is pulsing at the right level.  You might be able to compare it against those on either side that are working.  The DC will be lower than expected due to the pulse but consistent across those that are working.

[0culus]

Thanks for the input, everyone! I'm of the opinion that the ICs U62 and U71 are probably ok, because otherwise it seems like the whole display would be having issues, and not just the one module DS2.

The Darlingtons have an HP part number 1853-0409.

I did some probing with my 'scope and found that the waveforms from Q16 are not the same as the others.

[nixiefreqq]

wow.  not 100% unobtainium.  but gettin' mighty close.

maybe you saw that our friends in the great white north have some?

[nixiefreqq]

ya' might be able to find a Darlington with similar characteristics.

happy hunting.

[nixiefreqq]

maybe?

https://www.ebay.com/itm/Transistor-TIP107-PNP-Darlington-TO220-100v-volts-8a-amps-80-watts-Qty-1/282948215860?epid=668882514&hash=item41e1072434:g:pH8AAMXQLs5Rx8fC:rk:1:pf:0

[0culus]

I have all those parts if needed for the cost of shipping.  I would need time to desolder from the carcass so if you can get them from mouser, that would be a better option.  If DS2 is out, the problem is most likely in that driver Q16 though it could be the LED, most often they fail by segment.  Most of the time the problem is the cable, though.

Let me know if I can help.


Jerry

I might take you up on that.

[fenugrec]

ya' might be able to find a Darlington with similar characteristics.

Agreed, I wouldn't worry too much. Just make sure you get a PNP darlington. Or why not just a PMOSFET ? Might even perform too well, what's the Vce on the good transistors with all segments lit ?

[0culus]

So I tried swapping the darlington from a working module with the suspect one. Both passed diode tests, and it did NOT fix the dead module. Additionally, it appears that moving the board has introduced further bugs, perhaps aggravated cold solder joints on logic chips, because now the entire display is glitchy and the unit eventually freezes totally.

I'm afraid this new issue might be terminal. Not the end of the world because I didn't pay much for this unit.

[0culus]

Sooo, I fixed it. It was mechanical contact issues in the large ribbon cable all along. That'll teach me to check those first next time...anyway, I determined by tracing with the oscilloscope that the waveform on one particular pin of the "dead" DS2 was massively distorted on the display board side (compared to the matching pin on the next display module), but fine on the main digital board side, coming off the darlington Q16. So I inspected the ribbon cable more closely and discovered that the contact for that wire was mushed in and not springy like the others. I made a crude shim with some hookup wire and, like magic, DS2 lights and everything works. I think the glitches I was seeing were also a contact issue because I have not observed them since seating every cable.

[0culus]

Additional lesson learned: do more probing before soldering. 

[nixiefreqq]

wooooooohooooooo!  you da' man!


ps  you owe us a picture.

[0culus]

wooooooohooooooo!  you da' man!


ps  you owe us a picture.

Ooops, I forgot to take a pic of my quick and dirty fix, but here's one of the display working correctly! I'm going to see about making a small metal shim that will fit under the screwed up contact.

[nixiefreqq]

SWEET!

[0culus]

 

It's a really nice meter, perhaps a little large by modern standards, but it's still accurate. Happy I was able to fix the issue!

[cncjerry]

Great news, and a black face is the nicer in my opinion.

Jerry

[james_s]

Nicely done. My experience suggests that it's usually the simple stuff, especially when some hard to find hard to test part is suspect.

[0culus]

Great news, and a black face is the nicer in my opinion.

Jerry

I agree! I haven't seen too many of these for sale either, so I figure I grabbed a good one.

Nicely done. My experience suggests that it's usually the simple stuff, especially when some hard to find hard to test part is suspect.

Yeah, I'll just chalk it up to more experience debugging. 

[nixiefreqq]

keep thinking about getting a 3456a to use as my reference meter on the bench.

nawwww.  if my 3455 was good enough for Collins its good enough for me.


ps  but i might get a 3490 if i can snag one with a dot led display.  those dot leds are almost as irresistable as nixies.  (and if they ship me a broke one i can always send it to oculus for repair now that he has proven to be a steely eyed meter man.)

[0culus]

Oooo, look at those property tags! Someone peeled them off of mine before I got it. 

I would certainly be willing to try! So long as I can get the service manual (one of the big reasons why I favor old test equipment). 

[nixiefreqq]

sold off all the Collins tube radio stuff when I downsized and moved a couple of years ago.

could not bring myself to scrape off the 3455 Collins cal sticker even though it is pretty crusty looking.

[0culus]

Hey, it's a little bit of history!