Latest rescue...

Tony Duell ard at p850ug1.demon.co.uk
Wed Jun 9 15:47:38 CDT 2010


> 
> Ok, I found it on EBay, but I consider it a rescue. :)
> 
> This is an HP 9836 Model 200, a machine I've been looking to find for
> quite some time.  It's a Motorola 68k-based workstation that was
> targeted (I believe) at the scientific market.  It runs a proprietary
> OS and has several languages available for it.  My next goal is to try
> and generate diskettes for it, which is not going to be easy until I
> finish getting my IBM PC set up for use at home.


They are nice machines. I have 2 of them (a 9836A and a 9836CU) along 
with the closely-related 9826. And a couple of other machines in the 
range -- the 9816 (built into the monitor case)  and the 9817 (a 'shoebox').

Anyway, back to the 9836. To get inside, there are 4 large screws on the 
underside. The top cover then lifts off. There's a motherboard flat in he 
bvottom of the case, which as well as connectors for the other boards, 
contains trhe HPIB, keyboard controller and some of the address decoder 
circuitry. Plugged into the motherboard are the expansion backplane (8 
DIO slots), the CPU board, the floppy controller and the PSU regulator. 
There are 2 more PCBs under the left-hand disk drive which are the video 
display circuitry (the top board is for the graphics display, the lower 
board for the text display).

Several CPU boards have been used in these machines. Most of them contain 
a 68000 CPU and some RAM (which is auto-addressed [1]. And the boot ROMs. 
If at all possible, get a machine with boot ROM 3.0 or later. Those can 
boot from any device (internal drives, HPIB drives, SRM network, even 
bubble mmemory (!)) whereas older versions of the ROM can only boot from 
the internal drive. I believe there is a boot disk (it's part of the SRM 
system) that allows older machines to boot from anything, though. And 
note that alas the later ROMs are larger than the older ones, so it's not 
simply a matter of getting the images and burning a couple of EPROMs to 
upgrade an older machine.

[1] In these machines, RAM starts at the top of the address space and 
builds down. The expansion RAM cards have to be addressed by DIP 
switches. At part of the power-on self-test, the machine scans down 
throught the RAM. When it fails to get a DTACK/, then hardware on the CPU 
board automatically sets the address of the RAM on the CPU board to fit 
immediatately below all the RAM it's found.

The latest CPU board contains no RAM, but it does have a 68010 processor 
and an MMU circuit. This is fitted to the -U machines (U == Unix, these 
machines can run HP-UX). If you have this board, you need at least one 
expansion RAM card.

The floppy cotnroller uses a Western Digital IC. A couple of points. 
Firstly, there are spearate cables from each drive to the controller 
board. The drives have their MX jumpers fitted, so the outputs are always 
enabled. The controller monitors things like the WrProt signal from each 
drive to detect disk changes. The cable for drive 0 is soldered to the 
board, the one for drive 1 is plugged in. This is because the same 
cotnroller is used in most single-drive 9826s, whereupon the drive 1 
cable is not fitted. There is a 256 byte sector buffer on the disk 
controller board. It's a pair of either 2112s or 2114s (with 2 address 
pins grounds). I mention this becvuase if you hve the latter RAM and have 
internal disk problems, you know where to look :-) (Yes, I did have this 
problem on my 9836CU).

The power supply is quite friendly in some ways, but a paiu to test on 
dummy load. The basic design is that there's a transformer, a little PCB 
at the back containing the voltage selector switches and a rectifier (the 
TO3 can is a double diode, nothing more), and a smoothing capacitor. This 
gives about 30VDC to the regulator board which contains swithcing 
regulators to produce +5V, +12V and -12V. 

The problem comes in testing it. The regualtor board has 2 sets of edge 
figners. One plugs into a chassis-mounted connector carrying the DC input 
from the smoothing capacitor. The other plugs into the motherboard and 
carries the outputs. Now, since the mothrboard contains considerable 
circuitry, including a programmed 8041 microcontroller (for the keyboard 
interface), you don't want to use that as a dummy load.

What I did was make up a test box. A die-cast box with a couple fo edge 
connectors on top, load resistors, lapms and voltage test sockets on all 
the outputs and a pair of input sockets to power it from my bench supply. 
Then I can just plug the PSU regualtor board in that and run it out of 
the machine. But hten I have 3 of these machines to maintain. It's 
probably too much work to do if you only have one.

A few comments now on the 9826 in case anyone has one. It's very similar 
inside. Early models had a floppy controller that could only handle one 
drive and a lower-current PSU, but I have never seen these. Most of them 
have the same boards as are used in the 9836. The video system is very 
different, of course, (a monitor PCB -- which contains just the 
horizontal and video amplifier cirucits, the vertical circuit is on the 
text video PCB) mounted in the base of the machine in front of the 
motherboard, a text video PCB standing vertically behind the CRT and a 
graphics board on top of that, screwed to the top of the CRT cage. 

A word of warning. The CRT in the 9826 is _not_ rimbanded (the ones in 
all 9836s are). If you have to work on the CRT area of the machine, take 
great care, an impolosion is likely to be unpleasant.
 
As for expansion, there are 8 DIO slots. These uise the same connectors 
as the S100 bus, but different signals (obviously) and smaller PCBs. The 
signals are essentially the 68K bus with a few extras. 4 of the slots can 
take boards with external connectors (interfaces, etc), the 4 between 
them can take things like RAM, ROM, bubble memory, DMA controller, etc.

These machines have a built-in HPIB interface, but nothing else. I think 
an RS232 interface (HP98626 or even better HP98628, which is 
Z80-controller and does intenral buffering) is almost essential. An 
HP98622 GPIO card (16 bit parallel in and out) is useful too. There are 
many other interfaces and add-ons, inclduing ROM boards containing BASIC 
or HPL, more HPIB interfaces, and even a bubble memory board.

> There's more information available at
> http://www.hpmuseum.net/display_item.php?hw=3 and I have a photo

There's an HP serivce manual which is a boardswapper guide, but is 
probably worth reading. The only thing I really disafgree with in it 
(other than boardswapping :-)) is the procedure fro removing the CRT from 
a 9826. I find it a lot easier to remove the keyboard and then the bezele 
assembly, then put the latter face down on the bench, remove the CRT 
mounting band, and lift the CRT out. It takes alittle longer than the HP 
method, but it's much less likely to damage the CRT.

There's also a set of unofficial schematics which may be useful if you 
have to fix the thing.

You also want to read the manauls for the language you are using 
(probably HP BASIC 5.x). There are a lot of interesting features...

On Bitsavers, you want to look at the _pascal_ system manuals. Some them 
cotnain a lot of low-level programming information. And the 'breadboard' 
interface manual which describes the DIO bus.

As for software, be warned that he disk images on the Australian museum 
site are geernally for 3.5" disks. You can't dump them to 5.25" disks for 
the internal drives. If you have a HP9122 or similar (and boot ROM 3.0 or 
later) then there's no problem, you can boot from that and transfer the 
software to the interal drives later.

Do watch the power-on tests. The error messages are not always taht 
clear, but they will pick up a lot of problems. If something eems 'not 
right' then you need to debug the hardware.

Let me know if you need any help. I've done a bit with these machines.

-tony




More information about the cctech mailing list