DEC troubleshooting

From: Bill/Carolyn Pechter <pechter_at_shell.monmouth.com>
Date: Thu Nov 5 15:34:45 1998

> Date: Tue, 3 Nov 1998 21:28:22 +0000 (GMT)
> From: ard_at_p850ug1.demon.co.uk (Tony Duell)
> To: classiccmp_at_u.washington.edu
> Subject: Re: PDP11 training troubleshooting
> Message-ID: <m0zao01-000IyTC_at_p850ug1>
> Content-Type: text
>
> > > Hmmm.. This actually wouldn't have worked for me, if I'd been allowed to
> > > do what I liked to the machine.
> >
> > Well, the class wasn't 8-(...
>
> You know, I've made some nasty comments about Field Service in the past
> (what DEC enthusiast hasn't...), but now I realise it wasn't all their
> fault. I mean, not being able to pull boards is terrible...

Yeah, but it taught you to think microcode, block diagram and
real troubleshooting instead of pul and swap.

I was a board swapper in the Field, but it made me an informed and
logical board swapper.

>
> God gave us extender boards (and DEC made extender boards) so you could
> run a CPU module outside the case to look at internal signals. It's a
> standard way of figuring it all out. Trying to debug an 11/45 only being
> able to look at backplane signals is horrible...
>

Sure. Tell the customer he/she will be down an extra hour while
you extend the board to find the fault, desolder the chip
and replace the 10 cent bus driver. Nope. Swapping was reality in the field
when I was there. Board level (early 80's DEC) was (unfortunately)
replaced with shotgunning and option swap out by many in the late 80's.



> Were you give (or allowed) a KM11 maintenance card? For those who've not
> seen one, it's a board (acutally 2 boards...) that you stick in special
> slot in some DEC machines. It's got 28 lights and 4 switches on it. In
> the 11/45 it'll let you single-step the microcode (or even the clock
> sequencer), display flag signals, etc. You can get the uPC on the panel,
> of course.

They were used in the 11/70 class, the RK05 maint class and a couple of
others. The 11 new hire used the 11/34's built in programmer panel
for much the same stuff.

> > >> However, in the real world I'm more likely to start with a 'lucky dip'.
> > >> Unplug all connectors/boards, clean contacts, look for bent pins. Check
> > >> all fuses. Power up and check the power rails with a voltmeter.
> > >
> > True... if you know someone's been playing inside. But in Field Service
>
> I'd check edge connector contacts _anyway_. Particularly if the fault
> seems to be totally crazy.
>
> > with machines under maintenance most folks keep their hands out of
> > them. (except for printers and terminals) This means something broke.
>
> Hmmm... OK, pity the poor field servoid who used to deal with the VAXen
> at one of the Universities I was at. He knew that there were DEC hackers
> there - people who read printsets for fun. The system manager was one
> such, and I joined in the fun.
>
> Typically, we'd call out field service with a call of the form 'Oh, it's
> gone again, and this time bring DEC part # <whatever>'. We'd already had
> the thing apart you see, and knew what we were doing. DEC didn't seem to
> mind too much - they'd bring the bit, do a few tests to confirm, and
> install it.


Sure. Had one of those guys at Naval Air Propultion in Trenton, NJ.
Customer: "The te16 won't go on line. U134 should go low, but it's floating.
Can you bring us a driver chip."

Field Service: "We don't have it in stock, but I'll get one in the next
three weeks if I P1 it... Or I could come down and change the LAW board
on the drive (LAW -- TE16 Logic and Write board) in about 15 minutes."

And then there was the site that complained about failing RK07 Dual port
diags on 11/70's when the drives ran flawlessly unde RSX11 and IAS.

The problem was race conditions in the diag. The diag was no longer being
updated. The only fix that got the field service folk off the site
was noop'ing the error calls on the tests with the race condition.
>
>
> > Check power and work from there.
>
> Oh, indeed. And ACLO/DCLO. Those cause more problems that you'd believe.
> Especially on H742 power supplies where C3 (I think) on the power control
> board goes open. ACLO and DCLO then get mains frequency ripple on them,
> and the rest is obvious.
>
> -tony

Absolutely. That's a very common failure along with the diode and cap
on the VT100 video cards that's underrated.

Bill

+---------------------------------------------------------------------------+
| Bill and/or Carolyn Pechter | pechter_at_shell.monmouth.com |
| Bill Gates is a Persian cat and a monocle away from being a villain in |
| a James Bond movie -- Dennis Miller |
+---------------------------------------------------------------------------+
Received on Thu Nov 05 1998 - 15:34:45 GMT