I bought this HP-91 calculator recently from a fellow collector. I knew it was non-functional, but I hoped I'll be able to restore it to good working condition.

Well, it proved to be a much tougher cookie than I expected but in the end, it appears that I managed. This machine turned out to be like that proverbial horse at the veterinarians' college (well, there is a Hungarian proverb about such a horse): it had just about every conceivable disease. Except, fortunately, diseases that involve components unique to the HP-91, such as its ROM chips.

Upon opening the calculator, it became immediately obvious that it was tampered with before, by inexpert hands. An NPN transistor in its power supply was replaced by what later turned out to be an NPN power Darlington (i.e., something that definitely doesn't belong there; or, come to think of it, maybe it'd do the job, since the sole function of this transistor is to provide extra current during times of peak current demand, such as printing, but hey, we're not redesigning the calculator, we're here to fix it!) The replacement transistor had the wrong pinout, so its legs were twisted and generally, the whole thing looked ugly. There were further signs that many components were desoldered and then reinserted, but this work was done from the component side (with all the mess that that implies); the "proper" way requires more extensive disassembly of the calculator, to provide access to the solder side.

OK, out with the non-fitting transistor. I have the HP-97 Service Manual from the HP Calculator Museum CD-ROMs, and the power supply circuit is identical between the two models, so I was able to use this manual to diagnose the power supply. The replaced transistor turned out to be a power NPN type, so I replaced it with a 2N3053 which I happened to have in my parts box. I also located a faulty diode by simply measuring the forward bias of all diodes in-circuit using my DMM. I replaced that (apparently a generic signal diode) with an 1N914.

I powered up the unit and was able to ascertain that the power supply was now working properly. Yet I saw nothing on the display. So I further disassembled the unit and began an attempt to diagnose the main logic board.

After wasting a fair amount of time (and needlessly desoldering/resoldering power supply components in the vain hope that the problem is still somewhere there), I came to the realization that you cannot diagnose the logic board while the printer board is not attached. Why? For some inexplicable reason, HP placed an LC circuit, essential for the logic board to generate its clock signals, on the printer board. So if the printer board isn't attached, nothing is working on the main logic board.

This also gave me the idea to check the LC circuit components on the printer board. (This, incidentally, is mentioned in the Service Manual troubleshooting flowchart.) Sadly (sadly, because they'd have been easy to replace), both the inductor and the capacitor appeared to be good.

With the printer board attached, I saw what appeared to be good-looking signals on the logic board but still nothing on the display. Naively following the path in the HP Service Manual, I'd have begun desoldering ROM chips. Fortunately, I am neither naïve nor do I have replacement ROM chips for the HP-91. So in the hope that the problem lies somewhere else (namely in an area where I have replacement components) I decided to diagnose the display board first. After all, if the cathode driver chip has failed, I reasoned, I'd have no display and no response to any keyboard activity (not even PRINT X), and that is precisely what I was seeing.

So I poked around with my scope and saw what were ridiculously weak signals on all the outputs of the 20-pin cathode driver. That doesn't look good. All signals appeared normal on the chip's inputs. So... desoldering time!

Lately, I've come to pride myself in my ever improving skills at desoldering sensitive, old chips without damaging either the chip or the circuit board. Sure enough I was overconfident. This poor chip came out in not one, but possibly a dozen or more pieces. Let's hope that it was a dead chip I ruined, I said to myself, and proceeded (MUCH more carefully) to remove an identical "known good" cathode driver chip from a dead, "for parts" HP-19C.

This time around I was more successful; the chip came out in one piece and within a minute, it was seated in a socket in the HP-91 display circuit board. Power on and... hurray! The display showed 0.00, nice and bright. (Maybe the dead chip was as brittle as it was because it was heat damaged already?)

What next? Well, checking out some basic functions of course. It was then that I found that the entire left side (the "scientific" side) of the keyboard was non-functional. Never mind... disassembling the keyboard is easy, and after staring at it for a while and checking some traces for continuity, it hit me squarely in the eye: the thin metal contact sheet that contains the key contacts is in its place upside down! Of course it won't work. Flip it, reassemble (then disassemble, adjust the power switch that got misaligned, and reassemble again), test... scientific keys were now working correctly.

Time to test the printer. I already had a small problem with it; the paper advance switch, which worked at first, stopped working. I found that the printer motor was not reversing properly. The printer board also showed signs of inexpert repair work, and simply pushing the appropriate driver transistor a little restored printer motor operation. Now I inserted some paper, used the paper advance button to feed it properly, and then hit PRINT X... and got 0.00 on the paper. Or rather, I got something that looked like 0.00, except that three out of the seven dot rows were missing.

What next? The Service Manual suggested that I should measure the print head resistance to ground; if it isn't above 9 Ohms, I should replace the driver transistors. All 7 were above 9 Ohms. Was it the print head, then, that was faulty? Without a tool, it was a difficult task to remove the print head ribbon cable but I managed without damaging it; all 7 print heads still showed the proper resistance. Reinserting the ribbon cable turned out to be a problem; the spring contacts in the 23 year old connector head were anything but springy. Eventually, I decided to take a gamble and soldered small pieces of wire to the ribbon cable. The gamble worked; it appears that the plastic used in this cable is high-temperature plastic, which can withstand careful soldering. Whew.

But I still had three missing rows. The print head is driven by driver transistors that are packaged four each in 14-pin DIP sockets. I traced the circuit to find the pinout for this chip (HP part number 1858-0044 if anyone is interested; the pinout is C-B-E-NC-E-B-C on both sides) and tried to measure the forward bias on all transistors in-circuit. I got what appeared to be abnormal values, with a significant variation from one transistor to the next. Could it be the driver transistors?

These are NPN type transistors, and since they're used as switching transistors, just about any replacement will do... except that they need to carry up to 0.5A of current. I found that I had some 2N2222 transistors in metal cans; those are rated 600 mA, so I figured they'd do the trick. I desoldered the two quad-transistor packages and inserted 8 2N2222 transistors. After fixing one bad solder joint, I found that I now had a fully functional HP-91, complete with working printer!

Reassembly is left. Sounds simple except that it isn't. I mean, geez, these HP guys really made things cramped even when they didn't have to (like, when they built a largish desktop calculator, not a teenie-weenie pocket-sized device!) The socket I used for the cathode driver chip proved too big; the display/keyboard circuit no longer fits. The metal can 2N3053 also proved too large, preventing the bottom of the case from fitting properly with the top. So what needs to be done is desoldering the socket, reinserting the cathode driver without using a socket, and replacing the 2N3053 in the power supply with another NPN power switching transistor that has a more appropriate physical shape. But not tonight... I wasted enough hours already, working on this beautiful calculator. Which goes to prove why it's such a terrible idea to repair vintage calculators for money; had I been spending all this time on my "real" work, I would have earned the dollar equivalent of more than one "New In Box" HP-91, even at today's outrageous eBay auction prices!

* * *

Another day, another transistor. Despite what I said about power Darlingtons above, I ended up using one in the power supply myself, simply because that was the only NPN type I had with the right shape and pinout. Well, it works, so no-one can complain... As it turns out, I didn't need to desolder my cathode driver chip either, it did fit after all (just barely.) On the other hand, that bad contact on the printer board came back to bite me; eventually, I desoldered and properly resoldered those components that my inexpert predecessor felt the need to tinker with. The motor driver transistor actually needed replacing, because one of its legs decided to give up after being bent so many times. No big deal except that next time, I ought to be able to distinguish between a PNP and an NPN transistor before selecting the proper replacement type!

Fully assembled, my HP-91 now works like new. I am, needless to say, a happy camper.