Great work. It is almost inspiring me to look at restoring my HP9820A.
I was goign to say the 9820 is my least favourite of this family (but
only just), but... OK, it uses algebraic ntoation, which is a minus-point
compared to the 9810, but it does properly hadnle interruptsm and in
theory it can support a DMA periperhal (although AFAIK the only DMA
peripheral for this family was the 9880 disk system, and the software to
run that on a 9820 never existed). It does support the 59405 HPIB
itnerface (the 9810 doesn't)
Anyway, it's a fine machine!
As for restoration, there are preoavbly 2 aspects to this. The first is
gettign it working. It's a bit-serial machie built mostly from TTL with 9
PROMs in the CPU (7 for the microcode, 2 for the ALU tables). THe
processor iteself is 4 boards, they link to the 'memory box' which
cotnains another 3 boards of logic  (address register/decoder ; data
register ; memroy timing), along with RAM (103 DRAMs) and ROMM (2 boards
f HP custom 512 byte ICs).
 I beleive there awas a version which I have never seen which
essentially used the memroy system fo the 9821. This has 2 boards of
memory cotnrol logic and, iIRC static RAM.
I cna talk youthrough any part of the debugging/repair. Just to give you
some idea what you are letting yourself in for, I would start by checking
the PSU with no logivboards fitted. Asusmiugn that's OK, fit the CPU
clock board, check it's running and the 2 main clocks look right. Then
fit the basic set of boards (all 4 CPU boards, memroy box, display) that
will give a display. If you get _anything_ on the display than a lot of
the mahcine is working.
Msot lilely the display will be blank. The display ius entirely
software-drven in this machine. So you first check to be sure the CPU si
not tryign to drive the display (you would feel a right idiot if you
spend time going trhoug hthe CPU and memory logic only to find the actual
fault was the display driver itself!). Assuming the processor is not
driving the display, I generally do a quick check for stuck bits of the
memory address and data registers (avaialbe on a test conencotr on the
memroy box backplane), then put a logic analyser on the mirocode program
counter (available on a trst conenctor o nthe PCU control PCB) and see if
(a) it's attempting to ru nthe fetch-execute loop, and (b) if so, does it
seem to bre reading instruions or does it do an ADA every time (this has
an opcode of 0 and a machien that always seems to be executing thsoe
probably isn't reading the ROMs correctly).
Of coruse a lot mroe can be wrong, but in general knowing roughly what
the machien is and is not doing is a big help in figuring out where to look.
The secodn aspect of the restoration si mechanical. There arr 3 areas here.
Teh gooling fan. Don't replace it, the original HP fan was a lot better
made than the ones you get today. Take it apart, oil it with light
machine oil, and it should be good for another 20-40 years :-)
The pritner platen roller. This turns to the most horrible black tar-like
goo imaginable. It drives everywhere!. Cleaning it of is the worst part.
You ahve to strip the printer, then you can repair the roler with 'Cold
Shrink'. the hard part is cutting the new roller to length, and gettign
The card reader roller. This also fails, fortuantely it turns to acrimbly
dust that is relatively easy to clean up. The rpoble is that the old
roler was moulded ont othe hub. O-rigns will work, but ensurign they stay
in place is a problem. I remvoe the old hubs from the spindle and machien
and fit a new brass hub to take O-rings. That works, the only problem
being not everyone has a lathe.
Anyway, if you do feel like gettign started on your 9820, posthere, and I
will almost certainly reply.