28 May
2002
28 May
'02
6:57 p.m.
Normally, a machine that damages boards has got PSU
problems (the 5V rail
sitting at 12V or something). Do you know what the outputs of the PSU
should be and can you check them? (use a 'scope as well if you have one
to look for HV ripple on the output -- a common failure mode of SMPSUs is
a high ESR (or open) output capacitor and you can then get 10-20V spikes
on the 5V output...).
I no longer have one to look at, but from memory, the
6K regulators are at the back and the outputs are
shielded by transparent plastic covers. Should be easy
enough to remove these and 'scope away. Note however,
that these are there for protection: the voltage is low
but the current is high - remove metal from body before
trying this out!
One board damaging another is less common unless some
component failure
is shorting one of the pins to a supply line or ground. It can happen (as
here). Do you know what to expect on each pin (I guess not, as AFAIK DEC
never published real technical documentation for the VAX 6000 series).
The 6000 tech manuals tended to concentrate on the
programming aspects. There was an XMI handbook which
covered the bus signals and protocol in some detail.
If there are any recognisable chips on the boards
(especially bus
buffers), check them. And if there are picofuses for the supply lines
(some machines have them, I have no idea if the VAX 6000 is one of them),
check those too. It's not unheard-of for a bus short to blow a fuse and
not damage much else.
The XMI connection was handled by a dedicated chip
(in the "XMI corner") with possibly a few drivers and such
strung around it (much the same concept as with the
VAXBI bus).
I would be surprised if a dead XBIA managed to
also take out an associated XBIB. There are rules about
which boards can go in which slots etc. I had
a few "funny turns" with my 6K but it always turned out
to be a badly seated board, or a missing cable
or the problem being somewhere else and me being
in a rush.
My first step would be to ignore the VAXBI,
remove as many boards as are needed to get the
thing to start at all (which in this case seems
to mean remove the BI-XMI interface boards) and
then add in the XMI end of the interface. Now
try to run diags and see whether the board is
even partially alive. (Obviously it will fail the
tests that go look for the other end of the
interface). If OK so far, empty the VAXBI cage
and add in the BI end of the interface board,
properly hooked up. Pay careful attention to
the transition headers and plugs on the BI
and XMI interfaces. Now see if the diags are
happy or not. If OK, season with further
BI boards.
I did this a few times myself and always
ended up with a full set of working
boards. Maybe I was lucky, maybe there really
were problems with boards being badly seated
or maybe I had a marginal board or two.
Since all my problems happened when I needed
to reconfigure the hardware, I assumed it was
just seating rather than marginal hardware.
Antonio