Images of my 9100B


Picked up on TAS a few weeks ago...

Man, they knew how to make ROMs in those days. Check each individual bit with a Radio Shack multimeter!

The beauty of discrete components...

CRT deflection boards. Note how they're screwed directly into the cast aluminum case body.

Nothing seems to be leaking, but we're gonna replace those caps anyway. In England, Tony Duell cringes.

The keyboard fascia and labels are in excellent condition.

Board closeup...pretty!

The unit was advertised as "working", and aside from the doubtless gummy card reader wheel, the only problem we've noticed so far is this "comet" artifact in the display. It grows shorter, reducing to a tiny dot, as the digits in the first line extend farther to the left. Hopefully this will be an easy fix, but even if it remains, it doesn't affect the usability of the machine.


Looking at your pictures I can imagine the puzzled faces of the engineers when Dave Packard told them to pack the functionality into a shirt pocket sized device. :-)

"It's impossible!"

"I know, just do your best!"


"We've done it!" (Very big grin!)


You'd think those integrated circuits would make all the difference! But I have the 9100's successor, the 9810, stuffed full of ICs and it's even larger! Of course it does a lot more, too...


Actually, the base model of the 9810 does less than the 9100. The trig and log functions are an optional module for the 9810, admittedly one that fits entirely within the machine.

That big board of diodes in the 9100 is not a ROM. It's often thought to be one, but it's the AND gates and OR gates that interconnect the flip-flops and the other parts of the machine. One odd feature is that none of the power lines are used on this board (they are routed through it between the plug-in boards, but none of the resistors or diodes are connected to any power line or ground).

The ROMs are the 'core on a rope' microcode on the leftmost plug-in board (visible in one of your photos) and the inductively-coupled PCB ROM that is under the gating board (the board of diodes). For the latter, the ROM itself is the PCB in the middle of the assembly, the 2 boards down the left and right sides are the sense amplifiers (32 bits each), the board across the back is the address decoder.


Very nice sample you got there, Ramsey.

The close up pictures are amazing and no IC anywhere. Would like to have one but I have no place on my desktop.

I was very amused by the desciption of the development of the 9100 where Bill Hewlett demanded a machine which would fit on his writing desk. The engineers tried to place the final calculator there but the result was sobering as it was a little bit too big. Wisely, they did this secretely while Bill Hewlett has been off and engaged a carpenter who managed to enlarge the opening in the desk so that the specifications were met.

I wonder if something similar happened with the proverbial shirt pocket-size calculator years later. Maybe someone changed secretely Bill Hewlett's shirts with pockets beeing able to fit the HP35.


and no IC anywhere.

The 9100 actually does contain one IC.


Are you thinking of the main ROM PCB (which is, I suppose an
integrated circuit of inductively coupled tracks), the resistor/capacitor modules used to couple transistors in the flip-flops, or the op-amp ICs (8 in total, all the same type) used on the card reader PCB?

Or something else that I've forgotten?


I was thinking about the op-amps, but had forgotten that there are eight of them.


(From Puerto Rico)

Your machine is very clean and in good shape cosmetically, mines is somewhat neglected condition as appears that it got hard use. Yesterday as per your posting, I powered it up and the display comes on in 3 minutes. It is a bit dim, the 2 right digits position (exponent) are with less contrast than the others, but it worked fine. For a machine that old, the HP9100s are really robust.
I am very happy about your wonderful example, enjoy.



Some friends and I have been discussing the feasibility of building a micro controller-based LCD replacement display for the 9100. The consensus is that it probably would not be that hard. Wonder how many we could sell?


Wonder how many we could sell?

Probably none. The major cool feature about the 9100a/b is the CRT. However it would be a fun project to read the X-Y analog values and map them to pixels on the LCD. I agree that this should be pretty straightforward using a micro with ACD inputs and a serial LCD with a decent enough (64 x 128?) number of pixels.


Of course, some of those CRTs don't work any more, and replacements are unavailable. And you wouldn't have to wait 30 seconds or more for the CRT to warm up every time, either.

Nothin' worse than a bored engineer or programmer!


And you wouldn't have to wait 30 seconds or more for the CRT to warm up every time, either.

But it resumes where it left off when power was turned off (even when this was done in the middle of a running program). My PC takes several minutes until it is ready for use. In that respect the 9100 was way ahead of time.

BTW, I'm still seeking for a 9102A Buffer Box for my 9100B, just for the case someone has a spare one :-)


But it resumes where it left off when power was turned off (even when this was done in the middle of a running program)

I was amazed when I discovered for the first time, that one could turn the HP9100A off in the middle of a running program and be resumed when the unit is turned on.

If you want a fast modern machine, buy one :-). For me, the warm-up time is part of the charm of the 9100.

That said, yes, the CRT is a custom part and is unavailable. It's odd in that it's an electrostatically-deflected CRT with a fairly wide deflection angle (this would lead to non-linearity of the display which would matter in a 'scope but not in a numeric display).

I think if either of my CRTs ever failed, I'd find some other CRT that could be driven by the 9100's deflection boards and hack the PSUs and z modulation to get it to work. I've got plenty of calculators with LCD displays after all...


Hi Tony. Your comments triggered an old memory for me. I'm presently in the broadcasting industry and back in the mid-70s all the test gear we used had electrostatic deflection for their CRT displays. As I remember, electrostatic deflection was much more precise and could hold a calibration better. Also it was easier to draw discrete characters directly on the display since it's basically vector graphics. Displays with electromagnetic deflection (like home TVs) were less precise and used for raster-style displays.

We used Tektronix test equipment back then and they built oscilloscopes that would actually display the time-base and vertical sensitivity directly on the screen. You could also separately adjust the character display brightness from the waveform brightness. When I first saw that I was blown away, it was so neat for the time.

I am assuming, that since HP built o'scope based test gear as well, they had a lot of electrostatic CRTs around and could build vector-based characters directly on the CRT that were bright, easy to read and stable (they didn't wobble around on the display). It was probably the most cost effective way to have a bright, clear display that was easy to refresh with parts used in other equipment.

Also, when I looked at the pictures on inside the 9100, it reminded me of how equipment was built back then. A lot of this stuff had to meet military specifications (I was in the Air Force at the time) and they could be difficult to repair given all the circuits needed to maintain temperature stability for calibration. I had to troubleshoot and repair this equipment down to the component level but it was such a pain. I'm so glad it's now done to the board level; it's a lot less time consuming but you have to be more proficient in system troubleshooting with individual parts being so small.


Edited: 2 Nov 2011, 3:17 p.m.


Electrostatic deflection is more precise because there there is a liner relationship between the charge on the deflection plates and the amount of deflection, because of this the electronics to drive electrostatic deflection are also simpler. The down side is that the angle of deflection for electrostatic tubes is quite low so the length of the tube is quite high relative to the size of the screen. Electromagnetic deflection on the other hand has a non-linear relationship between field strength and deflection of the electron beam so the electronics driving it has to be designed to compensate for this to get a constant sweep rate, making the electronics driving it more complicated. Magnetic deflection likely won out because much sharper deflection angles could be obtained making large screens practical.


One of the main advantatges of electrostatic deflection over electromagnetic deflection is that it's a lot easier to change the voltage on a small capacitor (the CRT deflection plates) quickly than to change the current through a sizeable inductor (a deflection coil) quickly.

To generate a steady ramp, as is needed for a raster display, is quite easy in an elecrtomagnetic display, you make use of the inductance. But to move the beam around 'at random' requires a current source with enough voltaeg compliance to overcome the back EMF of the deflection coil. It can be done, but it means a big PSU, particularly if you want much bandwidth from it.


Oh don't get me started on board-swapping. There is absolutely nothing sensible (IMHO) about replacing parts essentially at random until the device seems to work again. Note I said 'seems to work again' -- if you don't really know what the fault is, you can't know you've fixed it.

And yes I've seen plenty of examples where board-swapping has appeared to cure the problem only for it to come back a little later. Either because the problem was actually a bad connection and replacing a part -- any part -- disturbed it enough for it to work again, or because the part that was replaced wasn't the real problem in the first place, but the new one will just about work with another part that really is failing.

And of course board-swapping makes no sense at all if you don't have a known-perfect set of boards _of the right revisions_ to swap in.

It's odd, but every one of the books I have on electrical and electronic faultfinding stresses that the only way to find a fault is to make measurements, think about what those measurements mean, and then (and only then) replace the faulty component. I do that on all my computers, calculators, instruments, etc, which may be why they work :-)

Getting back to the HP9100 CRT display, certainly HP had a lot of experience with electrostatically-deflected CRTs in 'scopes, etc. I don't think they had as much experience of magneticallly-deflected CRTs, but I am darn sure the engineers could have made such a device if they had wanted to. But for a numeric display as on the 9100, the vector approach saves a lot of electronics (the beam essentially scans a '7 segment raster' and is unblanked as appropriate), and it's a lot easier to make an electrostatically-deflected vector display than an electromagnetic one.

Incidentally, one reason the HP9100 display is stable is that the 'display task' is synchronised to the mains. So a bit of ripple on the power lines may cause digits to be displaced slightly from where they should be, but they will be in the same position each time, so they don't wobble.


If someone is randomly swapping boards in a piece of equipment to try to fix it, they are asking for trouble and will probably get it. Having an understanding of how a piece of equipment works as well as technical documentation is critical to finding a problem quickly and safely. In my case, I have qualified engineers on staff to troubleshoot and repair our equipment and keep the facility functioning. The reason we are doing more board-swapping is based upon how OEMs have designed their equipment and the economics of repair.

First, at a systems level, broadcasters are consolidating their facilities to take advantage of the economics of scale by doing more with what they have. That means adding more revenue streams by having more cable channels that are little more than another automated playlist and a simple master control(M/C)switcher added to an existing facility and expanding the infrastructure.

Second, facilities like these have large items like a router that can be expanded (in our case) to 1024 x 1024 (that's the chassis limit) and we add I/O cards as necessary to handle more sources and destinations. Some manufactures have added extras into simple routers like M/C switchers on an output card, multiviewers (allows one large LCD display to show many router sources at once with a browser-based software application for configuration) and shufflers (which allow reconfiguring of embedded audio in A/V sources among other things)to add more 'flexibility' to their product. It works fine when it works but when it breaks it's usually not at a convenient time. So, spare boards and maintenance agreements are used to reduce downtime. In our case downtime can literally translate into millions of dollars. A Super Bowl spot can cost up to $2 million but that's an extreme, if often quoted example.

This brings me to my point (finally!). My staff will troubleshoot a problem and if necessary call the OEM to narrow down where the problem is and then they send us a replacement part. (For critical systems we do maintain a set of spare cards but that can be a quite expensive way to tie up capital money that's not doing anything until that one critical failure occurs that was anticipated.) 90% to 95% of the time that fixes the issue. But if not, then Engineering management will step in to expedite the issue and, if necessary, and OEM rep will arrive to figure out what's going on.

What I'm saying in a long-winded way is that boarding swapping is a compromise between keeping a facility running and keeping costs down. It's great fun to sit down and spend the time troubleshooting an piece of analog gear with a scope and multimeter, but in this day an age board-swapping is the only economic way to go. And that's why OEMs build their equipment that way, it's what broadcasters want to save on costs.

When commodity items like LCD displays or PCs act up for any reason, we just replace them and dispose of the broken stuff, it's just not economically feasible to repair them; though it does break our hearts. We all hate to see repairable equipment go to waste.

Anyway I've blabbed on long enough. I haven't even mentioned trying to replace surface-mounted parts that you need a magnifying glass to read or the desoldering station needed to replace ICs on a bad board. The horror stories I could tell about boards being butchered when someone thought they could just fix it. BTW, if we damage a board trying to fix it, the OEM won't accept it for repair and we have to buy a NEW board. It's no fun going to the VP of Engineering and explaining why we have to spend $20,000 for a new board. I happen to like my job so being conservative about repairs helps to keep me employed.

On a personal note, no matter the challenges when trying to repair an item, there is no greater sense of satisfaction when you fix something that was broken by replacing a some small part. My wife has the mindset of if it's broken it has to be replaced and she sees dollar signs. So when I say I fixed it by replacing a $1 part, she is always amazed and grateful for a husband that can fix stuff (usually). It helps to make up for other stuff I can't do, like dance.



IIRC you can grab the digits in BCD (+ a few other codes) from the expansion connector. There are 2 other signals which indicate which register is being displayed, basically yopu wait for those to change to a different state, then start grabbing digits (Least significant first IIRC). Look at the HP9125A plotter schematics (which does this to get the X and Y values to plot).

Of course the expansion bus levels are totally non-standard...


Very nice unit, and very nice photos. Thanks!

It's prettier than my non-working HP 9820A. I never saw a 9100 anywhere at Georgia Tech from 1970 to 1974, but I did see and use a 9820A. There were some
WANG 360SE units around, whose nixie-tube digital displays were almost as neat as the 9100's CRT display.

What's the first few characters of the serial number? It's amazing that 40 year old units like this still work, especially since a rather high voltage power supply was needed for the CRT. I can understand wanting to replace electrolytic capacitors in the CRT HV supply, but unless an exact replacement can be obtained, I personally would be reluctant to alter the original and now almost historic condition of such units.


What's wrong with your 9820? I would assume a totally blank display at power-on, which means a fault just about anywhere in the unit :-(. However, if you want to have a go at fixing it, I can talk you through it. In many ways the 9800 machines are easier to work on than the 9100, in that it's mostly TTL logic using voltage levels you can handle, but being a bit-serial machine, a logic analyser is very useful....

As regards replacing old electrolytics (in 40 year old computer/calculator) : I must have 20 or 30 machines of that date (certainly not all HP desktop calculators!), and, yes, I've had to replace the odd electrolytic capacitor. But certainly not all of them. Quite a few of my old machines are still running on all their original capacitors. I feel that this 'witch hunt' on old electrolytics is totally unjustified.

There is, of course, the issue of what will happen if the capacitor fails. If the only effect is that, say, the machine stops working, or the display goes unstable, then I'd leave it. Replace the capacitor when it fails. If, as in one much more recent HP I repaired, the result of a bulging capacitor actually 'blowing its top' would be that it would hit the neck of the CRT and probably let the vacuum out, then I'd replace it (as, indeed, I did).


Hi David,

why check with a Radio Shack multimeter when I see a Fluke on your bench ;-)

nice find!

hpnut in Malaysia


Because that's not my bench; it's Larry Atherton's bench.


And here is an image from the HP9100A I still power-on some time.
The original magnetic card protruding from the card reader has been kindly provided by a forum member (I leave him to disclose himself if he like to). I also tried to reproduce DIY magnetic cards with some success using VCR tape and carboard. These were useful e.g. to store a complete diagnostic program (see list in appendix A of the 9100A/B service manual).




Your register labels to the right of the display seem to be lit from above. I thought they were backlit...(bear in mind that as I write this I have never seen a 9100 "in person")


Yes, there is a filament bulb behind the upper portion of the registers display.
It is a little too bright though, and the silkwork mask is also a bit weared out so that it is overshining.

I suspect originally there were two bulbs, one on the lower portion now missing or blown, but I didn't checked it yet.




Those bulbs can still be bought, e.g. on TAS.


No need for TAS, here they are from Mouser.

I no longer have my 9100a, but used these bulbs and they are perfect!


Thanks for pointing to an alternative source. But, for heaven's sake, why did you give away your 9100A? I would sell my mother-in-law (even on TAS) to get one ;-)


why did you give away your 9100A?

I didn't "give" it away, I sold it to someone that thought he could repair it. I bought it broken and I had made some progress is repairing it myself, at one point getting the X and Y registers working along with basic arithmetic. I never got past that point however. It was very frustrating since I didn't have a full schematic and nothing to test against -- just Tony's hand drawn 9100B schematic.

I did learn a lot about it in the process of doing some repairs (I needed the bulbs too).

Possibly Related Threads…
Thread Author Replies Views Last Post
  Sending little images to the Prime (...and Program name bug?) Erwin Ried 19 7,319 12-10-2013, 05:35 PM
Last Post: Erwin Ried
  HP-Prime CAS nice images CompSystems 9 2,705 08-28-2013, 12:00 PM
Last Post: Juergen Keller
  [41CL]Updating ROM images with a Clonix Dan Grelinger 14 4,354 02-13-2013, 10:41 AM
Last Post: Ángel Martin
  Newest 41CL ROM Images Dan Grelinger 2 1,197 01-16-2013, 01:25 PM
Last Post: Monte Dalrymple
  Some errors about NOMAS, (c) of 71B ROM images Christoph Giesselink 4 1,804 06-13-2012, 04:10 AM
Last Post: Valentin Albillo
  9100B restoration progress David Ramsey 6 2,124 11-27-2011, 10:33 AM
Last Post: Etienne Victoria
  9100B progress David Ramsey 11 3,156 11-17-2011, 08:30 AM
Last Post: Joel Setton (France)
  9100B progress David Ramsey 6 2,078 11-04-2011, 11:32 AM
Last Post: David Ramsey
  Nerdvana & mass production of FOCAL rom images Geir Isene 11 3,245 10-10-2011, 04:52 AM
Last Post: Egan Ford
  Updated list of ROM images in the 41CL - help needed with links to quick reference guides for any! Gene Wright 1 981 06-02-2011, 07:20 PM
Last Post: Geir Isene

Forum Jump: