getting started with your new HP 21xx
jwest at classiccmp.org
Mon Jan 8 16:04:31 CST 2007
> From my own 2117F experience, one if the first things to check is
> whether or not the 21MX power supply has the battery backup option.
Glen is spot on... that's the number one most common reason a system won't
power up (or I should say, won't run - it'll still power up). Number two
reason is probably incorrectly configured memory (controller, memory boards,
or MEM module), including the very likely possibility that the large ribbon
cable which connects memory to the controller isn't seated well or making
good contact. Third reason is missing or improperly configured microcode
(FAB, FEM, WCS, UCS, or mainboard jumpers), including the possibility that
the FAB/mainboard/FEM cable isn't seated well or making good contact. If the
system was delivered exactly as it was previously used without having been
scrounged for parts or poked at by non-HPaware playbabies, the above list is
wrong, there are more likely causes of failure.
> If it does and you don't have a battery backup box attached the CPU
> will not start without a battery box eliminator plug, which is just an
> 820ohm resistor across two of the pins in the battery connector plug.
Viewing the battery connector plug from the rear, there are three rows of
three pin sockets. You want your 820 ohm resistor across the outer pins of
the middle row. Another thing I should mention... remove the power supply
before you power up the machine. Very simple/quick to do, it was made for
quick/easy replacement. Inside the supply was the ubiquitous black foam. It
has probably crumbled and/or turned to goo. Take off the cover of the power
suppy after removal and get rid of all that foam with a vacuum cleaner.
Clean any messes (contact cleaner, toothbrush, etc.) up. Some advise just
closing it back up at this point. I don't - that foam was glued to the top
lid for very good reason. Those 4 or 5 power supply cards inside the
supply - you do NOT want them touching the top of the metal cover :) Less
impressive but equally destructive is if they come loose which the foam also
prevents. I cut a piece of foam from air conditioner filter material (either
1/8 inch thick or 1/4 thick or maybe 1/2?, I forget what the power supply
foam is but I think it's the thicker stuff) to replace it. I scrape off the
old foam on the cover with a putty knife and clean to bare metal with
goo-gone. Then I tape off the right size area and spray with adhesive spray.
Remove the tape then press the air filter on. After it dries, button up the
supply and reinstall it - you're good for another 20+ years :)
> I believe the symptom is that all of the front panel LEDs are lit and
> state lit on power on with the battery backup option installed and
> with no battery box connected.
I'd need to dig into the microcode test routines, but there MAY be other
diagnostic selftests that could display before the battery backup check
error (which is exactly as Glen states).
> If the CPU does power up ok and there is memory installed, the low
> bits of the display should count for every 32kw of memory that is
> installed and tested. One of the manuals explains that, and what to
> toggle in to run the test again and in loop mode.
21MX M or E reference guide, which I can't find on bitsavers :) However,
here's the scoop. In microcode there is a power up self test, and
(separately) tests for FPP and SIS microcode. Let's deal with the self test
microcode first. There's 3 self tests. Test 1 tests basic registers and
functions, no memory. If this test fails, all display register bits, all
indicator bits, and the over flow register will be on. Test 2 is a basic
(read, doesn't catch many errors) memory test but ONLY the first 32kw of
memory. This test is nondestructive. Errors usually show a parity error
light, set all display registers and indicator bits, and clears the overflow
register. A register is expected data, B register is bad data, M register is
failing address. Note that the cpu front panel lights do NOT increment/count
up like Glen mentioned during this test. Test 3 is a more comprehensive
memory test, and it tests all memory installed not just 32kw. Each 32kw
passing the test gets the display register incremented by one. Errors are
reported the same as test 2, except the S register shows which 32kw block
failed. Note that upon cold powerup, tests 1 and 3 are run automatically.
That's it. No other tests. Note also, that kinda behind the scenes - anytime
you press the IBL button to load a loader rom into ram, tests 1 and 2 are
automatically run first. That's why it's labled IBL/TEST. You can also run
tests 1 and 3 manually without powering on/off the machine. Set P register
to 0, A register to 100000. Set the S register to whatever value you want
memory filled with. Press PRESET. If you want to loop the diagnostic, set
the LOCK/OPERATE switch under the front panel to LOCK, otherwise continue
on. Press INSTR STEP and the diagnostic runs (and loops if lock switch set).
If diag is looping, stop it by putting LOCK/OPERATE switch to OPERATE. To
test the FPP, set 105004 in A register. Put 0 in P register and press
PRESET. If overflow is on now, stop and go directly to the FPP
installation/service manual. If overflow is not on, press INSTR STEP. A
display of 102077 means all is well with FPP. To test the SIS, set 105477 in
A register, 0 in P register. PRESET, then INSTR STEP. 102077 is good pass.
FYI - these addresses/procedures aren't magic, they are just loading A
register with an instruction from the user instruction group (used to call
microcode). Then setting the P register to address 0 - the A register is
addressible as memory address zero, and starting execution.
> Jay, if you do get a collections of diags on CD I would very much like
> a copy. I have some diags on 2645A tapes, but they have almost all
> fallen about now.
Actually, I have this project all done eons ago... except for just one or
two or three diagnostics aren't in absolute format! There's just 3 or less I
think that are actually in relocatable format. Yes, I know how bizarre that
is. But what it means is that the assembler listing for just those few
diagnostics isn't particularly helpful. On all other diags, you can look at
the halting address (or single step) and then look as the asm listing and
see exactly what's going on. Not so if the program is relocatable. One of
the listmembers took the diag source for these oddball relocatable ones and
assembled it for me and produced a load map. This makes it fairly easy (with
the extra step of adding the map's offset to the relocatable (from address
zero) listing) to figure out where you are. So, this compilation of "the
ultimate diagnostic cd reference" should be available soon.
More information about the cctech