[Beowulf] Re: Emergency Power Off

David Mathog mathog at caltech.edu
Tue Mar 20 13:25:42 PDT 2007

> At 04:20 PM 3/19/2007, David Mathog wrote:
>In any case, the current and dc voltage on these Tripp-Lite EPO
>lines isn't very high.  I'm thinking for the few that we have
>in our computer room this circuit would probably due the trick
>(sorry about the ASCII art):
>   +-[9V+]-[9V+]---SWITCH -+  (<--switch is normally open)
>   |                       +-10K-- UPS1 -------+
>   |                       +-10K-- UPS2 -------+
>   |                       +-10K-- UPS3 -------+
>   |                       +-10K-- UPS4 -------+
>   |                       +-10K-- UPS5 -------+
>   |                       |                   |
>   +-----------------------+                   |
>   |                                           |
>   +-common ground ----------------------------+

On further consideration, that common ground might be
a problem.  On the one hand it's effectively running from one
serial port to another and generally that's ok.  On the other
hand, if for whatever reason one or more of the grounds is 
a volt or two higher than another one, that could result
in quite a current running on that signal wire.  This would
probably be safer:

   +-[9V+]-[9V+]---SWITCH -+  (<--switch is normally open)
   |                       +-10K-- UPS1 -1K----+
   |                       +-10K-- UPS2 -1K----+
   |                       +-10K-- UPS3 -1K----+
   |                       +-10K-- UPS4 -1K----+
   |                       +-10K-- UPS5 -1K----+
   |                       |                   |
   +-----------------------+                   |
   |                                           |
   +-common ground ----------------------------+

Whatever the slight variation in the grounds are
they'll more or less average out at the point "common
ground", without putting much current down the signal
lines. Since the impedance across the UPS input is
usually quite high, most of that 18V will still be seen by the
UPS input circuitry.  I'll have to experiment with
this a bit, it's possible that on some units this is
too much resistance.  Hmm, replacing the 1K resistor with
a diode in each line ground line, only allowing current to flow
towards the common ground point, should work too.  

APC sells a device for adding EPO (sort of) to their
smaller units that don't already have it.
But this box costs about $100 and each UPS is supposed to
have its own.  Here's the manual:



David Mathog
mathog at caltech.edu
Manager, Sequence Analysis Facility, Biology Division, Caltech

More information about the Beowulf mailing list