HP 9830A desktop programmable calculator - An exploration

[TerraHertz]

Here's an item from the Australian Computer Museum Society storage dispersal. An HP 9830A desktop programmable calculator.

I actually got two 9830As:
* Unit 1. SN 1303A01320 Just the main unit. Nothing else. Very dirty as received.
* Unit 2. SN 1303A02375 Very clean & complete set. Has dust cover, printer & cable, manuals, rolls of paper.
* 9866A printer. SN 1234A02133  Came with Unit 2 above.

The clean unit went straight to my storage space in the house. At some point I'll see if it works, but there's no rush. The filthy unit otoh, must be cleaned before it goes in the house. So it begins.

First, does it run? After a slow power-up on a variac, no smoke, then normal switched power-up, the answer is no, it's dead. Fan runs, a light in the cassette drive lights, the 'rewind' button makes the cassette motor run, but that's it. Except, at start there's a brief flicker of a random character on the display. Which means at least some logic power, and a working power-on reset.  Hopeful.

Next, disassembly for cleaning and evaluation.

More in next post

[TerraHertz]

It's a strange feeling to see a 'CPU' with no micro. Just 4 cards of not very much logic. I was looking forward to seeing the schematic. At this point I hadn't tried hunting that down.  Oh, naive.
Also amusing that none of these boards shows any sign of HF digital design rules. No ground planes, no controlled impedances. Just tracks.

My apologies for the poor pics of the cards. At this stage I was hurrying. I'll do better photos later when the cards are out again. They WILL be.

[TerraHertz]

The display is pretty neat, for the early 1970s. 32 characters of 5x7 LED matrices. Ceramic substrates with glued-on glass windows. Must have cost quite a bit back then. 

The keyboard. Now that's what I call a keyboard controller. We don' need no steenkin keyboard micro (because we don't have any micros at all.)

All the keytops were removed, cleaned in detergent, air-blown to get the water out of cavities, then sun dried for a while. That metal mesh is an old worn-out garden sieve. It occurred to me that local birds might think keytops could be edible. I didn't want to find any mysteriously missing.

An interesting kludge under some of the keys. There were two different types - some with thick mounting posts, and some thin. It seems they discovered the thin ones were prone to splitting. So most of those have spring circlips fitted, to reinforce the plastic. But not all. A few thin ones are without the circlips, and are showing signs of splitting. I need to find some of these exact size circlips.
Also their assembly was slack - someone put a lot of the circlips too far down (up!) the posts to be any use. I moved those ones to be near the end, like the ones in the pic. First time I've needed to use circlip pliers to reassemble a keyboard.

Last pic - the keyboard nice and clean. But then the fun began.

The front facia panel has a small filament bulb, that is the power light next to the mains switch. Naturally it's blown. I'll replace it, preferably with an orange LED if possible. What's the supply voltage?

Rather than just power the machine up and measure it, I thought this would be a good time to web hunt for the service manual. Will need it anyway.

First google found https://www.hpmuseum.net/exhibit.php?hwdoc=55
Great, looks like quite a lot:

9830A - Documentation
9830TechData-5952-2348-6pages-Jan73.pdf 0.352 MB
9830TechData-5952-2348-4pages-May74.pdf 0.263 MB
9830ROMSTechData-5952-2325-6pages-Nov72.pdf 0.35 MB
9830PricingInfo-5952-9092-2pages-May75.pdf 0.087 MB
9830TechData-5952-9074-16pages-Jul76.pdf 1.092 MB
HPJournal1976Feb-9830ATerminal-7pages.pdf 0.208 MB
9830QueenslandFloodStory-4pages.pdf 0.075 MB
KeyboardVol6No1-9830inBusinessApplications-4pages-Apr74.pdf 0.143 MB
9830A-SystemTestInstructions-09830-90027-25pages-May75.pdf 0.396 MB
9830ACalculator-ServiceManual-09830-90030-140pages-Jan76.pdf 1.724 MB
9830A-DataBaseRoutinesPac-09830-76500-152pages-Jun74.pdf 1.424 MB
9830A-MathPac-09830-70000-103pages-Jul75.pdf 1.041 MB
9830-OperatingAndProgrammingManual-09830-90011-171pages-Jan73.pdf 1.761 MB
9830-SchematicsByTonyDuell-89pages.pdf 1.53 MB
9830_SurveyingPrograms_Vol1_157pages_1973.pdf 1.973 MB
9830_BAMP-30-PAC_Volume1_09830-71102_173pages_Oct74.pdf 2.699 MB
9830A_BatchTerminalPAC_09830-76510_117pages_Aug74.pdf 1.143 MB
9830A_SurveyingPAC_Volume1_09830-74009_146pages_Oct74.pdf 1.519 MB
9830_DesktopComputingPower_5952-9083_16pages_Aug76.pdf 1.795 MB
9830_AccountsReceivable_BillingPacSupplement_09830-73026_117pages_May74.pdf 1.323 MB
9830_OrderEntry-Invoicing_103pages_1974.pdf 2.023 MB
9830_QuickReference_09830-90012_5pages_May73.pdf 0.806 MB
9830A_SystemTestInstructions_09830-90032_31pages_Feb73.pdf 0.343 MB
UtilityProgramsForSorting_11141-99006_31pages_Oct74.pdf 0.208 MB
9830A_AccountsReceivable-BillingPac_09830-73024_210pages_1973.pdf 2.201 MB
9830A_SimplifiedOperatingInstructions_09830-90000_75pages_Feb73.pdf 2.888 MB
5952-8963_9830A_BAMP-CAD_System_3pages_Feb75.pdf 0.457 MB
09830-70800_9830A_StatPacVol1_86pages_Sep73.pdf 0.945 MB
09830-70850_9830A_AnalysisOFVariancePac_102pages_May74.pdf 1.017 MB
09830-70875_One-Sample_AnalysisPac_121pages_Apr74.pdf 1.275 MB

Some downloading and looking at the results later...

Oh god. All the usual abysmal stupid, incompetent & lazy scanning flaws. F-ing unbelievable. Some examples:

9830ROMSTechData-5952-2325-6pages-Nov72.pdf  Pages horribly skew and out of order.
9830ACalculator-ServiceManual-09830-90030-140pages-Jan76.pdf   Schematics broken up, 2-tone, images 2-tone.
9830A_SimplifiedOperatingInstructions_09830-90000_75pages_Feb73.pdf  B&W (original is colour), lost page #s, skewed.
9830A-SystemTestInstructions-09830-90027-25pages-May75.pdf Pages totally out of order, lost colour, so lazy.
HPJournal1976Feb-9830ATerminal-7pages.pdf  Ha ha ha! What is that on page 20? Paper was moved, culprit didn't care.
Almost all are crap. Sigh. The ones of interest are:

Operating and Programming Manual  09830-90011  171 pages  Jan73    1.761 MB
Service Manual                    09830-90030  140 pages  Jan76    1.724 MB
System Test Instructions          09830-90032  31 pages   Feb73    0.343 MB
System Test Instructions          09830-90027  25 pages   May75    0.396 MB
Quick Reference                   09830-90012  5 pages    May73    0.806 MB
Simplified Operating Instructions 09830-90000  75 pages   Feb73    2.888 MB

But the main point is, the schematics in the service manual are fragmented. (I hadn't spotted the true horror yet.)

Artek manuals has nothing for the HP 9830A.
Ditto Bitsavers.
The bama site seems to be down atm.

The plan was to stitch together the schematics into useable form. At least they have OK resolution though it is fax-mode. It's also apparent that I have some paper manuals that are not online yet, so scanning those goes on the todo list. Plus scanning the ones I have paper copies of, that ARE already online, but doing it with decent quality.

First though, the schematics. I extracted the images from the PDF with photoshop CS6, and started stitching.
First discovery: Oh good grief. Whatever auto-scan process they used, produced results AT DIFFERENT SCALES, depending on the extent of the material on that part of the foldout page. Result, see scan_fail_variable_scale.png
That's two 'pages' that should be one schematic, just abutted in photoshop.
I don't know why anyone bothers to even pretend to be scanning technical docs, if this is the stupid crap that results. It makes me laugh even harder at the "we don't need physical copies, it's all online" crowd. They are so delusional.

Also, note that the two parts don't have any overlap. In fact there's a small sliver missing. Anyway, some rescaling later, this is what I consider an almost acceptable schematic. 125-126.png

So, all good?
Ha ha ha! Aaaagh! It turns out the 'service manual' only has schematics for the kiddie-stuff PCBs. The PSU and the cassette tape interface, and minor others. All the important cards are listed as "HP blue or red stripe replacement", ie swap cards back to factory, no schematics for you.
And there isn't even a diagram of the backplane listing board numbers for slots.

Oh great. So if I want to get this unit working and nothing else turns up (I did more searching, no luck yet) I'll be debugging a custom, discrete-chips 'processor', with no doco at all about the architecture, instruction set, etc.

Well this will be fun. (And I do have another unit, that just might be working (I hope) but probably not.)

In the meantime, I want to put the case back together, so next will be just measuring the voltage to that light bulb.

To be continued.

[TerraHertz]

That dead bulb turns out to be running on +5V, via a 22R resistor. So substituting an LED was easy. An orange LED, to keep the power light looking authentic. One slight issue was that the bulb was side-illuminating the clear plastic bezel. Solved by sanding a 45 degree flat on the LED, and gluing on a little bit of al-foil as a reflector. I used pink nail varnish as glue, to get the color a bit warmer.  Swapped the original 22R resistor for a 470R, to get about the right brightness. (I'm guessing, since I've never seen the original bulb running.)

The next issue to solve before reassembly, was that back in the '70s they had quite different ideas from today about mains safety clearance and shrouding. That front mains switch's terminals are only a few mm from the chassis, keys, and the wiring for the LED power light (that runs off system 5V.) I thought it would be a good idea to put an insulating shroud around the bare mains switch terminals, authenticity be damned.

It was a bit of a relief to get the keyboard structure back together and screwed into the chassis, without having dropped and broken anything, lost any parts, etc. The machine certainly looks much nicer clean.

Something cosmetic still to do - the lettering on the keyboard fascia, and at the power switch, is all paint in shallow indentations in the plastic. Typical, except that this paint has lost strength, and in some places the indentations are are barely there at all. So some of the lettering erased at the slightest touch. Not sure how to deal with this.  It needs some experimentation (not on this machine) to see what works before I try to do anything about it.

I'd been faintly hoping the cleaning and effectively 'wiggling all the connectors' might make it spring to life. Nope.
All the power rails are good. And the problem isn't any of the optional ROM packs hanging the bus - it still doesn't run with them all removed.

Well at least the power light works now. And the machine can be moved to my 'museum' in the house, now it's clean. I need the workbench space back.

Next will be some more efforts to find tech info (schematics please!) on the processor cards. If that fails, some PCB reverse engineering.  Plus I have one suspect small group of components (bus-drivers on the main system ROMS card) that have obviously been overheated.

Then there's trying the second unit. If it works, and once I've satisfied myself there are no contagious bus faults on the dead machine (which requires having identified what buses exist, and their backplane pinouts), then it's card swapping time.
Also, logic analyzer - what do the address and data bus do in the cycles coming out of reset?

[precaud]

Have you checked out the Vintage HP computer group?

https://groups.io/g/VintHPcom/topics

Some of the guys who designed the 9830A hang out there.

[TerraHertz]

Have you checked out the Vintage HP computer group?

https://groups.io/g/VintHPcom/topics

Some of the guys who designed the 9830A hang out there.

Ah! Thanks for the link. I had been told to check out "cctalk" and "vinthp" for contacts with HP engineers.
For cctalk found:
  http://www.classiccmp.org/  ClassicCmp - enthusiasts of classic computer hardware, software, and docs.
  http://www.classiccmp.org/lists.html
  cctalk  http://www.classiccmp.org/cctalk.html  on & off-topic posts members:~650 Posts:50-100 per day
  cctech  http://www.classiccmp.org/cctech.html  on-topic only        members:~150 Posts:50-75 per day

But hadn't yet found https://groups.io/g/VintHPcom/topics
Inquiring on both those is next.

Other things to do:
* Extract the binary of all ROMs in the system. Both the system ROMS and option ROMs. Not going to desolder the ROMS, just make a hotwired adapter for the card edge connectors and extract that way. To be posted to bitsavers, etc.

* Apparently the software architecture allows for programs from HP provided on the tapes, to be 'hidden'. Both assembler and BASIC software.
  Let's see if that can be broken, to extract and publish binary images. If that hasn't already been done.

* There's no built-in HP-IB interface, and I don't have a HP-IB plugin. Find or make one.

* Convenient file transport between PC and HP 9830A.

* It's really surprising there seem to be no manuals online for some of the ROM options. More hunting required.

[TerraHertz]

A lucky break! The second unit runs.
Though it has some damages to the keyboard fascia. A big chunk broken off the left hand rim, an internal mounting post snapped off on the right hand side, and an adjacent post stressed to near-breaking.

Does anyone know what plastic this would be, and what glue would work?
This was made before the days of marking plastic type on injection molded parts.
I tried a small spot of model glue on an inside surface, and it doesn't soften this plastic or adhere. So NOT that glue.

It doesn't look like this one saw much use. Little dust inside and the filament power light still works.  Also the area on the system ROMs PCB that is heat-browned on the non-functional Unit 1, shows no sign of heat stress on this one.

[rrinker]

CA would probably work.

It may be ABS or Butyrate. Model cement, at least that type, won't touch it, but other sorts of solvent cements can handle it. Plastruct Plastic Weld would work, but the chemicals it's made up of are usually available locally (not that you need a lot, but as someone who builds a lot of models, buying the generic chemicals in larger quantities is a lot less expensive). I get confused about which is which but they commonly have MEK as a main ingredient. Styrene like most model kits is easier to dissolve and you can get away with less aggressive chemicals to join it.

[rdl]

What kind of model glue was it? It used to be that almost all model kit "glue" used MEK as the active ingredient, but maybe that's changed since I last built a plastic model. MEK will dissolve most non-thermoset plastics. If MEK or acetone doesn't dissolve it, then CA or epoxy may be your only hope. It's hard to figure why HP would use anything but ABS though.

[TerraHertz]

Cyanoacrylate doesn't seem like a good idea, since the surface area is quite small. And if it fails, then the fracture plane has CA contamination that would prevent any other glue attempt from working well.

I have plenty of MEK and acetone. I'll try making up a slurry of ABS and MEK, and do a small adhesion test on the underneath of that panel.

What kind of model glue was it?

It's in that photo above. Surprisingly, there is no mention of the solvent used. The smell is different to MEK. Sharper tang.

It used to be that almost all model kit "glue" used MEK as the active ingredient, but maybe that's changed since I last built a plastic model.

Something to do with MEK being declared carcinogenic? And probably replaced with something worse, I bet.
Found an interesting thread here:
    http://www.theminiaturespage.com/boards/msg.mv?id=236559  "Best Glue for Plastics?"

MEK will dissolve most non-thermoset plastics. If MEK or acetone doesn't dissolve it, then CA or epoxy may be your only hope. It's hard to figure why HP would use anything but ABS though.

The material seems a bit tougher than the usual consumer appliance cases. Hard to quantify. And I suppose ABS comes in a range of strengths.

[rdl]

I didn't enlarge that one photo and missed the glue. Reviews on Amazon for it often mention the applicator tube clogging so apparently there's something in it besides just solvents. I'd still bet that a solvent is the active ingredient though. MEK is not too bad, it's very similar to acetone which was actually delisted as a VOC here by the EPA. If you can get the pure solvents it might be worth a try. I'm pretty sure PVC pipe cement is mostly MEK, so if you have any it could be tried also.

[xrunner]

Wow very cool project! 

Now I've got it bookmarked I'll be keeping up.

[TerraHertz]

Reviews on Amazon for it often mention the applicator tube clogging so apparently there's something in it besides just solvents.

Yeah, after not using it for a while it's always clogged. But the cure is simple - pull out the tube, and heat it in a lighter flame. The tube is stainless steel, tiny and thin-walled. All material inside vaporizes, apparently leaving no residue. Blockage gone.

It's very quick to do. Don't even need to hold it in pliers, since the stainless has so small a thermal mass and conducts heat so poorly that you can hold one end in your fingers, heat 2/3 of the tube to near red heat, blow on it briefly to cool, flip it round, and do the other end.


Bah. A while ago I threw out a great pile of assorted plastic cases scrap, from disassembled street tossed e-waste. Now do you think I can find some scrap ABS to shred for this fix? No.  Not so far anyway. But it can't take long to find something.

Actually this exercise will be useful for something else. I also have a Facit 4047 paper tape punch/reader with a quite badly broken casing. Also probably ABS. http://www.bitsavers.org/pdf/facit/Facit_4046_4047_Punch_Reader_Technical_Description_Edition_3_Oct86.pdf

Edit to add: Ah the joys of mass-produced e-waste. Structures in PC CD/DVD players are plastic and marked with the plastic type. 1st one I checked has a CD tray of ABS, frame of PPE, and head carrier of PPE+PS. Fine for now. I'll also buy some ABS sheets on ebay.

===========================

March 2019   Edit to add:
This HP 9830A restoration got shoved down my project queue by an avalanche of new equipment arrivals later in 2018, and other urgent 'non-maskable interrupts' from Life. So it's on hold but not forgotten.   See http://everist.org/NobLog/ for progress on some of those, and eventual resumption of the 9830A project. The machines are all fully reassembled, clean, and under dust covers in my 'museum'.

[dfnr2]

Oh great. So if I want to get this unit working and nothing else turns up (I did more searching, no luck yet) I'll be debugging a custom, discrete-chips 'processor', with no doco at all about the architecture, instruction set, etc.

Actually, Tony Duell Created wrote a fantastic monograph on 98xx theory of operation and repair, available here

While reverse engineering the machines, he created some schematics also on hpmuseum.net.

And Brent Hilpert has a very pretty set of schematics on his site.

The GO9800 emulator available in the files section of the vintHPcom group has dumps of all the ROMs