Despite its unassuming calculator looks, the HP 9825 is a powerful 16 bit, all-in-one desktop scientific computer and instrument controller. Well, powerful for the time, which was 1976. It is actually far faster and has more memory than the HP-85 from the 1980s.
That's because it had a 16 bit multi-chip hybrid processor, an early version of the one that would be used to make the famously powerful HP 9845 workstation. However, with its one-line display and efficient but cryptic HPL language, it is much less approachable than the BASIC and CRT endowed HP 85. It was also far more expensive, listed at a whopping $5,900 for the base model in 1977. And that's without the many expensive option ROMs and additional RAM you'd need for a functional system. The fully equipped 9825T model was listed at $8200 in the 1981 catalog.
After the heroic repair of our HP 98035 clock interfaces, we gave made a nice demo of the HP 9825T in action.
The HP 9825 also makes a few cameo appearances in my other videos, like here in my HP DC100 Cartridge restoration video at 18:03:
https://youtu.be/cTZCD4OXETk?t=1083
or in my 100k subscriber video, at the very beginning:
My 9825 suffered a catastrophic power supply and overvoltage shortly after filming the demo video above. Just about everything on it went bad - power supply, RAM, ROM, keyboard and display interface. Not to worry! We just need to repair it all. What a great excuse to dive into the guts of this amazing machine and understand how it works.
The integrated HP Mini Cartridge tape drives are essential to the functionality and enjoyment of the HP 9825. Unfortunately, if you have a 9825, your chances are pretty close to 100% that the tape will not work. Here is a two parts series on how to repair them (and also repair the cartridges).
One of the interesting plug in peripherals for the HP 9825 is the 98035 real time clock module. I have 6 of them, and none of them work. But in this epic repair-a-thon, we get them all going!
The Hewlett-Packard Journal edition of June 1976 gives the best overview of the HP 9825:
Steve Leibson has an entire web site dedicated to the HP 9825, recounting its history from inside HP and its lineage starting with the quintessential HP 9800 calculator and the HP 1000 16-bit mini-computer:
The HP Museum has good documentation although naming is hard to follow. Here is the whole series in the context of 98x5 computers:
My own curated version of the manuals is further below. It is a renamed and organized list of all the manuals I could find on the net, with my own additions of a few missing manuals I scanned.
Here is a collection of all the manuals I could find, plus two exceptional engineering resources: the HP Journal Article and the US4075679 US Patent. The HP Journal is a great overview of the engineering of the machine, and the quintessential, 606 pages long patent is the one and only in-depth reference we have about the details of the engineering of the 9825A. It is complete with the firmware source, a detailed description of the processor, and explanation of sections of the schematics.
For info, this is the 16 bit binary word code as it appears to the processor on the data bus, suitable for disassembly. It is not how the data is arranged on the 8 bit ROMs themselves, so the following is not suitable for burning ROMs. Use the files above for burning replacement ROMs.
The HP 9825 patent has a complete commented listing of the 9825A code. It's not exactly what the production code of the HP 9825T is, but is quite close. Craig Ruff was able to match the two and recreate a commented version of the production code, which was invaluable when we debugged our very sick 9825T.
In this video, I explain how I used Logisim Evolution to simulate the mysterious RAM control logic used in the 9825.
Download Logisim Evolution from here:
https://github.com/reds-heig/logisim-evolution/releases
Download the simple counter example here:
Dowload my 9825 sim project file from here:
My HP 9825 was crippled after the power supply main transistor shorted and the whole thing went into an overvoltage. The main regulating transistor in the buck converter shorted, and the 5V for powering the TTL section went up to 13V, destroying the entire machine. This resulted in a great video series about the repair of the machine, forcing us to understand and reverse engineer every board and every function in the machine, which I don't think had ever been done before. About 5 TTL circuits per board were fried on average, as well as all the LED display chips and some ROMs. Needless to say, if it happens to your 9825, it is no fun at all.
HP apparently knew about the problem and provided a crowbar protection add-on, which is a little board that installs easily on the original supply. Fellow HP restorer Rik Bos found a crowbar on his two European-built HP 9835's, which share almost the same power supply. The PCB number is 09825-66552, which indicates it was originally meant for the HP 9825. As installed on Rik's 9835 machines it looks like this:
From that photo, I was able to figure out what the schematics were and made a reproduction. I even sourced the exact same vintage HP-branded thyristor as in the original, just for kicks. A modern thyristor equivalent is also provided in the schematics, and will be smaller and do just as well of course. See below, the replica board solders right onto the original supply with no modifications.
Documentation including the full KiCAD project and PCB gerbers is provided below. I ordered my PCBs from PCBWay, you can order them right on the link below, but you should be able to get a PCB made from the gerbers below from any vendor of your liking:
I have finally reached a point where I am able to recover old software tapes for the HP 9825, thanks to a modified tape drive with a modernized head amplifier circuit.
HP developed many verification tapes for its test instruments. They are impressively thorough and require a lot of test equipment. But it turns out I have most of it. Seeing it run for hours and put an instrument through its paces is quite impressive!
I have finally reached a point where I am able to recover old software tapes for the HP 9825, thanks to a modified tape drive with a modernized head amplifier circuit.
HP developed many verification tapes for its test instruments. They are impressively thorough and require a lot of test equipment. But it turns out I have most of it. Seeing it run for hours and put an instrument through its paces is quite impressive!
In this video, we restore and use the HP 9895 8" drive, which is far more potent and reliable than the finnicky tape drive included in the 9825. And it lets us play Colossal Cave!
The HP 9825 supports two 8" floppy disk drives, the HP 9885 and the HP 9895. These are far faster, more reliable, and larger capacity than the DC-100 tape drive included in the unit.
The HP 9885 is a single drive, single sided, single density floppy, and connects via a parallel 16-bit IO interface.
The HP 9895 is a dual drive, supporting both single and double sided, single and double density floppies, for up to 1.2MB of storage. It connects via the HP-IB interface plug-in (which must be either a Revised A or a B model, the original A won't work).
But in order to use either of these drives, you'll need an HP 98288 Mass Storage ROM. And these are as rare as hen's teeth. And they were thought to be difficult to duplicate, since HP used proprietary mask ROMs with some very unusual address decoding, bank switching and power logic embedded in them.
But that all changed when Paul Berger got his hands on an equally rare SSS Mass Storage ROM, which my friend Jack Rubin had acquired. SSS was non-HP, but they were obviously former HP engineers from the 9825 development team with very deep insider knowledge. They made an upgraded Mass Storage ROM that supported the newer 5.25" and 3.5" HP-IB drives, at the expense of dropping support for the older HP 9885 drive.
Paul reverse engineered the boards and found that SSS had used standard ICs and two Intel 2764 EPROMs to make an HP ROM equivalent. The only slight difficulty in the design is that it also includes a small bipolar 82S129 PROM, which does not contain code, but is instead used to implement address decoding. So you have to program 3 ROMs to make this work: the two 2764 EEPROMs that contain the same code as the original HP ROMs or the SSS ROMs, and the little 82S129 with a small binary image.
Once Paul had reversed engineered the SSS design, a copy of the SSS PCB was made by Adrian Nicol, and a make-your-own Mass Storage ROM clone solution became available to the community. You could flash the two 2764 EPROMs with the SSS or the original HP 98288 images. It was soone realized that if you used 27128 instead of the 2764, and slightly modified the board with a switch and a resistor, you could burn both images on the larger PROM and select which one you wanted to use with the switch. That's what I did on my board, and I include explanations on how to do it in my notes in the files below.
The info to make the clone is available below, except for the Gerber of the PCBs, which I could not find. But KiCAD schematics are included as well as images of both sides of the original SSS PCBs, so making a new PCB should be relatively straightforward.
I listed all the individual files below so you can browse around, but I also included a .zip file that has all the files in it so you can download the whole package with one click.