To Michael Coyle: re: single-digits



Somehow I lost track of this thread, and it slipped ignominiously into Archives. After reading your response, I did the same as you--- searching for parts which might be pressed into service. Unfortunately, I, too, have come to the conclusion that NO CURRENTLY AVAILABLE PART will do what I hoped could be accomplished.

I feel like Emily Lattela (sp?) from the old Saturday Night Live: "Oh. Never Mind..."

Not having a 55 or similar calc at hand, I had to dig for info about what display it had. It was, indeed, unusual. The characters were .11" high. This alone is unique: I might have expected, though, to follow the lead of finding the supplier (prob chinese) of the displays for the Retro LED watch made by Fossil just last year. (Fossil "RoadTrip" JR-7749 is one model). Digit size, color and brightness and encapsuation in "bubbles" made it seem close.

But HP Classic displays were unique in quite another way, too: unlike most "seven-segment" displays which actually have 7 segments to make the digit, then also a decimal point located either to the left (LH) or right (RH) of the digit elements, the displays used by HP held the decimal point INSIDE the lower digit "square". This implies, firstly, that NO OTHER DISPLAY WILL LOOK RIGHT, and also, that HP displays took what normally would have been a digit position to display the decimal alone. Very unique.

So, assuming that no display will work but for the originals, we can see WHAT and WHY they are what they are: a substrate on which the LED structure is built, with 5 digit position areas (the 5 common cathode plates), and layered or printed on top is the set of anode segments; these are tied together with clever circuit tracing so that all top horizontal segments are common, all top left vertical segments are common, and so on. I would expect that building all this in sets of fives would make yield and brightness-matching fairly easy. So digit selection was by activating the cathode, and segment selection chose the corresponding anodes on all digits. You could run through all the positions pretty quickly. All the same, this "multiplexing" would be a constant thing: the 40 segments (including the decimals) of each 5-digit display would be run through, lit or not, with each given a bit less than 1/40th of the "display cycle". I don't know how the classic calcs were wired, but it is MOST likely that the 3 5-digit displays were operated simultaneously, rather than in sequence, so that, at any given moment, three segments of the overall display were being tended to.

Now. Could a display be custom built as a "Classic" LED replacement? Sure. If yield were unimportant, you could chance it as a garage operation; this I do NOT recommend as the MINIMUM number of wire-bonding operations (to run a little wire from anode to connecting pads) would be 240 for the 15-digit display... and so your success rate is drastically increased if you start with a Clean-room environment and closely controlled bonds.

You'd create a multi-layer "thick-film" substrate, with connected cathode pads for each segment of each digit, and pads beside each segment which run to each segment position across five adjacent digits. At this point, I think you see that it could be done as single digits, five at a time, or all fifteen at once. The difference is entirely in how you want to make the "wiring together" of these things. The description of each digit itself is a "common cathode, 8-segment" device. If *I* were doing it, the "five at a time" or fifteen at a time would be too complex-- I'd go single digit and use a circuit board outside of them to tie them in fives...

Then, of course, you'd buy "LED die", in hi-efficiency red... (One major supplier of these is Marubeni). Die are square, and not particularly sized right, at 350 micro millimeters a side. But, gluing them down with conductive epoxy and wire-bonding to the top anode side with a tiny gold wire, they would be bright enough so that one per segment is all you'd need.

Dark epoxy poured between die (50 micro-millimeters high) would secure the wires and provide some optical isolation between them. Over that, a laser-cut "mask" would provide the segment "slots" that the LEDs would shine through. If the mask is somewhat thick, and "filled" with clear epoxy, you effectively have created a diffuser/light-pipe to the outside, which will make the individual segments bars instead of points. If this clear epoxy is not just syringed in to make a light-pipe, but overflows the mask face and is shaped into an over-digit "bubble", then you have done as close a job as you can probably do to creating an HP-equivalent part without going into the Ga-As business yourself.

Weell, that's a WAY to do it, maybe there are others. If you want to do this, it would be best to use a Class what? 10000? cleanroom environment, automation for bond-welding and encapsulation, and each individual die pre-screened for brightness characteristic. Wow.

Unless the market for new displays for Classics is bigger than I'm thinking, this is not how I plan to spend MY summer. Maybe I'll just scavenge a while and try to find an undiscovered stash of HP 5082-7405's lurking out there.

Sorry now that I "off-the-cuff" figured single-digit displays existed that could be used to make a replacement display. My! The display business is certainly chock-full of suppliers, all making the exact same BIG displays. Wonder how they all stay in business?


you have created a great word-picture of something that has been lurking in the corners of my mind. Whenever I consider the mechanical complexity of a Teletype, I think, "Boy, they must have wanted really badly to send text!", almost - but not quite - wondering why they didn't just use a microprocessor - of course, they didn't have one! But while I can just use a microprocessor (theoretically), it's only because of all the steps vastly more complicated and precise than building a Teletype that have been performed to make a microprocessor.



Thanks much for the appreciation. One thing that I *wished* to convey as I wrote was: that everything involved in making a workable substitute display was NOT magic, but DO-ABLE.

Oh, I was stupid in suggesting that somebody out there likely still made the parts. They CAN, of course, but apparently decided the market did not support making such a bird and keeping any stock of it. (I imagine Agilent, at the time they pared their parts catalog, never suspected that calculator COLLECTORS would exist to look for such things.)

I myself hadn't suspected that HP was using a design in their displays that was so deviant from the layout of OTHER seven-segment displays that it would make a difference in how the device using it sent the information to be displayed (the decimal point requiring a full digit-space).

And I plainly muffed it when I thought that somebody would still be making a seven-segment display so small (ha!! and I thought .2" WAS small!!), when about the only manufacturers that would care to buy such a product would be calculator makers, who had abandoned LED displays for LCD long ago. (The main advantage of LCD of course is its low power requirements, making LED seem way too hungry and wasteful.)

So, the simple, easy method of component replacement is no longer available, unless we stumble into a box in some old dusty warehouse, with long-overlooked tray upon tray of 5082-7405s. It could happen. Still worth looking out for.

So much for the "easy way". I wrote about the possibility of a more complicated option, and maybe ended on too skeptical a note, primarily because in my mind, the economics of it were poor as an "enterprise", even considering the eBay price of Classics. But economics can change, and even such a "Frankenstein" display as my proposal might SOMETIME be worth doing, should fully-working Classics be so desired that collectors begin to seek extra-HP solutions.

But I intended to point out that while what HP produced was a showcase example of their technological superiority (as most everything they produced back then WAS-- they showed off because they got business that way, and because THEY COULD do what others often couldn't!), that you could still investigate, understand and finally, replicate the result, if not the elegance, of the original.

Some of the tools and the tech once available only to a major concern such as HP are now the realm of small organizations in your hometown.

Agilent may still possess vast powers of "magic", but while once upon a time you'd have HAD to build your own LEDs from raw chemicals and using arcane processes to layer and etch a wafer that, hopefully, would emit light at the end of it all, NOW there are companies willing to sell you the bits and pieces that can, with some toying around, be cobbled into what you need.

These days ARE different, and though true "magicians" will always be at the sharp bleeding-edge, doing what we can't, there are a lot of good, underutilized, underemployed techies, possibly some with kit they bought at bankruptcy auction, and they've worked on the factory floor, seeing what steps get you where. Technology "booms" and "busts" do something useful; they spread last year's models around and take suppliers' smug disapproval of dealing with small businesses down a peg.

There are design and automation tools that would really have BEEN "magic" only a few years ago. There's the internet, which allows us to bounce around techniques past all, from professional to tinkerer. There are businesses whose mission is to support and supply esoteric materials to "the rest of us". There is a lot of business devoted nowadays to providing and supporting various scales of production. And stuff that makes production work efficiently often has the effect of making entry-level, minimum-wage workers able to take over tasks that once required extensive training.

So, while I agree that some of what I described sounds daunting, and it IS a lot easier NOT to do it if you don't absolutely HAVE to, nothing I've described would be impossible (or even improbable) for one who was looking for the way to get the job done.

Heck, you can RENT the equipment to produce wire bonds and drop glue down and encapsulate in epoxy and cure it-- in any reasonably-sized city in most places in the world, a trip to the phone-book will find the resources needed to get the job done. You can MAKE it more complex than is absolutely necessary-- and sometimes you'll want to, when it would make the work go faster, easier, more fruitful.

But Ellis, in about two hours peering down the microscope, making practice wire-bonds and testing them, you'd have the technique pretty well down. You'd be on your way to making a display, one segment at a time.

Placing the right-size drops of glue and getting the die right side up and in the spot you want them to be-- if you can play video games reasonably well, you can be a pro at this WITHOUT a degree or years invested in learning all about it. It is developed and applied technology now, and whether doing it one die at a time with micro-tweezers or going someplace where they do it a hundred times faster with automated suction placement gear, it's NOT "rocket science" and not any more complex than the same equipment used in many modern medical labs for automated sample testing.

That's why I looked at this way of doing it: making your own GaAs wafers (usefully) is a bit beyond the "lone techie" or the small club without a couple of years and a fortune to waste; but "hand" assembly of an array of die sourced from a carefully controlled wafer-fab IS WITHIN REACH, if the economics make sense enough to make it worth your while.

So, if enough owners of HP "Classic" calculators want for displays, and they want them BAD enough to pay for the costs, and a techie wants to shepherd the process and make this his "project", it CAN be done. Without running to Agilent and begging on your knees. Without a lot of silicon-doping and vapor-deposition and advanced chemical knowledge. Without building an empire of technical, marketing and financial genius.

I once watched, and marvelled at the sheer madness involved, as a friend of mine molded, carved upon and painted, using the tiniest of brushes, a small part of a model-railroad station. Using bright lights and magnifying glasses, lilliputian details were sculpted and finished to the point where, using a camera with a macro lens, you would swear you were there. Signs advertised on the wall, and they were fully readable and professional. The figures wore plaid and denim and had hair and pearlescent buttons on their shirts and real, rosy expressions on their faces. The wood all had grain and was weathered. A rocking-chair was poised ready to tread on the tail of a peacefully-napping dog. Norman Rockwell would have been proud to have portrayed such a scene, if only Mr. Rockwell had been about 80mm tall. My friend used dental tools and picks and almost daily made revisions and reworked some part of it, gradually bringing the scene closer to his vision of what the passengers of his train would see as it pulled in to this station. The engine and train of cars was equally painstakingly detailed.

I had always thought that you just bought some machine-made, pre-printed station, plunked it down beside a pre-built track, and ran a pre-built train past it. Seeing this little diorama evolve over a couple of years, I had to admit: hobbyists and enthusiasts can be stark-raving nuts, yet what they can do when they are totally involved in their passion is astounding and admirable.

What someone determines if it is worth doing? Only the do-er.

But many things I would not have thought do-able on a limited budget and within a reasonable time have BECOME so, for both corporations and individuals who still strive to create.

Take care-- g.


... just complicated. A teletype is very complicated. But making a microprocessor is more complicated, and requires more precision, taking into consideration all the steps. Once the microprocessor is built (and memories and logic etc.), making a terminal is easier than building a teletype.

I once helped salvage some "first silicon" ASICs that had a metal layer inadvertently connected to an analog output. We passed some current between two pins to burn out the metal, and the designer was able to test the output! (on some of the chips)


Hi Glynn!

Thanks for your excellent post. I have, as usual, many comments.

1. I'm sorry you couldn't find suitable LEDs either. And I doubt Fossil will be forthcoming about their supplier. (It looks like their Web site has no LED watches any more -- I guess they were discontinued.)

2. ...the displays used by HP held the decimal point INSIDE the lower digit "square". This implies, firstly, that NO OTHER DISPLAY WILL LOOK RIGHT, and also, that HP displays took what normally would have been a digit position to display the decimal alone.

You are correct. The decimal point is inside the digit and does indeed use an entire digit position to display. (In the Classics only -- see below.)

3. is MOST likely that the 3 5-digit displays were operated simultaneously, rather than in sequence, so that, at any given moment, three segments of the overall display were being tended to.

This is not correct. Each of the fifteen digits were turned on just one at a time, and in fact it appears that each of the segments in the digit is also driven one at a time (to reduce cathode driver current). HP designed a very clever circuit which involved charging an inductor and then discharging it through the LED. The charge time limited the current, which is more efficient than a current-limiting resistor. This is all described in detail in the HP-45 patent, but I have not analyized the circuit in detail.

4. Your description of LED stick manufacturing is correct as far as I can tell. I have a similar display in my junk box, an HP HDSP-6508, an 8-character LED with segments in a "starburst" pattern (and two dots for making a colon or decimal point). All the LEDs and bond wires are clearly visible. I can't tell if the segments are one-piece or made of shorter bars, though.

5. I don't think garage maunfacturing is likely. Not by me anyway!

6. Different HP calculators use different LEDs. The Classic LED displays are as described above. Woodstocks used a 12-digit display. From the MoHPC photos, they look like one big assembly of 12 digits in one long molded package. They also have the more common right-hand decimal points. Spice machines (34C, etc.) apparently have bigger digits, but fewer of them -- the picture makes it look like they are one-piece 11-digit displays (limiting the display to 7 digits in scientific mode). There is also a single dot in the left-hand position just above the "-" sign, used as a low-battery warning. I also seem to remember reading that the decimal points had tails so they could display commas, European style.

It's too bad these displays are no longer available.

- Michael


Howdy Michael--

Re: 1. You are right-- LED watches had a short season in retro-land. I think everyone realized that, no matter the nostalgia, you STILL had to press a damn button to read the time!

Re: 3. Aha!! Okay, so scanning 120 segments singly instead of 120 as 3 groups of 40. This makes a difference in the timing, of course, not in the construction of the display itself-- but it is really important to know how much of a duty-cycle and how fast it runs... the LED on for 1/120th of a cycle puts out a lot of light in that short of a time, and...

BUT WAIT!! Did you say there was an inductor in there? Hmmm.

Y'know, it IS possible that the purpose of that inductor was twofold: to both limit the current, and to stretch out the timing/duration of the pulse delivered to the segment... wait, that might not be right-- was there an inductor PER SEGMENT, or one PER DIGIT?

Re: 4. Yeh, my proposal was a bit of a kluge, since HP can do vapor-dep and etching to get bars, circles, stars or whatever they dang well please as an anode. I would content myself with a point-source (well, a square) that was available for purchase, then "form" the light from it by blocking some of it with a mask and making the light INTO a bar shape with diffusion. Not elegant, but would do the job. What you see under the magnifier is a layer of LED anode, the light-producing bit, and for long-bar or stripe versions of this they would have segmented it so it was, as it were, multiple parallel LEDs. I imagine that larger displays have to do this sort of thing more than a .11" version would have. In any case, they were not relegated to gluing down bits of die, as I would be; well, not unless they did so for an economic reason.

Re: 5. Dang. Your garage was PERFECT. We only had to move the car out for a while-- say, three or four months? ;-)

Re: 6. Yes, it's funny, though: I hear of Classics with display problems more frequently than of any other type. Maybe the others die memory-chip death before they can develop display problems, I dunno. I know that if I could select three things for HP to manufacture as replacement parts (assuming not calcs, but parts for them) I'd put in requisitions for battery doors for 41's, 28's and Spices, displays for Classics, and rubber anythings for anything HP has ever used rubber for. They were bad, bad guys in the rubber department. (I have a 9144a tape drive which turned completely to goo, ruining my one valid OS tape; a 9100a with its card-reader wheel, cracked and swelled; a 97 with a card reader that SMOKED rubber the first time I tried it... etc etc etc)

Hey, if not your garage, a spare room perhaps?


Hi Glynn,

I hate to break the news to you but there are no spare rooms at my house. I have my father and his wife here, with all their possesions, and together we're using about 110% of the available space. (Good thing I'm single, I guess. :) )

Anyway my mechanical skills are minimal. Maybe you should get your model railroad friend to put together LED displays in his spare time.

According to the patent, one diagram showed one inductor per segment. If you are really interested, you can download the patent (US Patent No. 4,001,569) from Peter Monta's Web page at It's written in English, not patentese, and makes for interesting reading.

Displays of the type we're looking for could be made by just about any LED manufacturer. We wouldn't have to go begging if we were to put up all the required money up front. It might be a lot of money, though. Hybrid IC houses might be able to do it too.

Nothing will happen on this front for a while, as we aren't sufficiently desparate yet. If calculator prices keep going up, I think you will start to see small garage junk dealers appear. At some point it will become profitable to buy busted calculators, strip them down to their component parts, and sell the pieces. It works for cars -- if I need a water pump for an '86 Ford, my mechanic knows about 10 junkyards and dealers to call.

This doesn't create new parts but it does get the existing parts into circulation more efficiently.

- Michael

Possibly Related Threads...
Thread Author Replies Views Last Post
  Prime: Placing more than 1 item on the RPN stack in a single program? John Colvin 4 801 11-19-2013, 08:59 AM
Last Post: Miguel Toro
  Obtaining More Decimal Digits (50g) Eddie W. Shore 3 605 09-13-2012, 05:38 PM
Last Post: Gilles Carpentier
  Single Level X^2 for HP-35...Any ideas Matt Agajanian 0 368 08-11-2012, 10:01 PM
Last Post: Matt Agajanian
  OT: The first 4,000,000 digits of Pi, visualized in a single image Jeff Kearns 36 3,669 04-29-2012, 11:06 PM
Last Post: Jeff Kearns
  Classic repair - ghosting digits - What's your experiences? McAllan 11 1,222 03-11-2012, 08:25 PM
Last Post: Geoff Quickfall
  Tiny Challenge: significant digits Werner 70 4,010 01-21-2012, 05:53 PM
Last Post: Paul Dale
  # digits displayed by HP 15c LE Masao Kinoshita 6 753 09-27-2011, 09:05 AM
Last Post: Paul Gaster
  10 or 12 digits? Kiyoshi Akima 2 423 08-09-2011, 03:22 AM
Last Post: Kiyoshi Akima
  WP 34S: Another poll about single letter labels Marcus von Cube, Germany 31 2,484 08-05-2011, 05:27 AM
Last Post: Marcus von Cube, Germany
  Reciting digits of pi Palmer O. Hanson, Jr. 14 1,219 04-04-2011, 02:37 PM
Last Post: Ed Look

Forum Jump: