Re: DECstation 2100: power trouble?

[Mouse <mouse%Rodents-Montreal.ORG@localhost>]

I (quoted and) wrote

>> At least replacing output filter capacitors ought to cause no
>> regression.

> I should also put my multimeter in the PSU lines and see what it says
> for both DC and AC measurements, the former for nominal voltage and
> the latter for ripple.

I've now done this.  I did it with all 12 RAM slots populated, as a
worst-case scenario (of the things I've tried, at least).

No joy - as in, nothing that explains problems.

All the measurements below are relative to the power supply's metal
chassis, except for the -9V line, which is relative to the -9V return
line.

DC measurements:

The -12.1V line is at -12.01V
The +5.1V lines are at 5.14V
The +3.5V - +5.25V line is 4.30V
The +12.1V line is at 12.27V
The ground lines are at 18mV
The -9V line is at -9.1V

These are all well within the 5% spec (except the 4.3V line, which is
within its specced range).

I set the meter on AC and repeated the measurements.  Everything came
in under 10mV.  All measurements were taken with a Fluke 87, so I'm
reasonably confident in them.  However, I doubt its AC setting can
notice extremely fast edges such as logic circuits can generate.

The LEDs were stuck at EF when I started and still were when I
finished.  They could in principle have changed during the tests, but I
think I have never seen them change and then later return to EF and get
stuck.

I'm considering adding caps soldered in right at the board's power
connector to see if that helps any.  I don't see any decoupling caps
anywhere near there.  Presumably you can get away with a lot more when
your clock rates are all slower than 25MHz.

I still need to look to see whether I have anything suitable for
replacing the PS output filtering caps.  But if I add caps right at the
board that may not matter.

/~\ The ASCII				  Mouse
\ / Ribbon Campaign
 X  Against HTML		mouse%rodents-montreal.org@localhost
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B