🚀 go-pugleaf

s41758@zeus.usq.EDU.AU (Brian Mesken) writes:


>I have a server (486) running 3Com's 3+Open connected to two Epson DFX-
>8000's. It appears that when the printers are turned off and on they
>create power spikes and cause the server to crash, sometimes. I have a
>power monitor recording these. There are also spikes from other sources
>not identified yet, that cause crashes.

>I have put a passive filter and a active tracking filter in series in
>the power line, but still the server crashes. I monitor the output of
>the wall socket (a), the output of the passive filter (b) and the output
>of the active filter (c). The spikes appears on (a) only, not (b) or
>(c). (BTW the power monitor only measures spiked greater than 50 volts)
>the spike voltage on (a) is measures a 60 volts to 120 volts, with
>durations up to 400 micro seconds.

>Question,
>Can the effect of the spike be transmitted over the Parallel printer
>cable, or ethernet?
>Has anyone else had these type of problem?
>Any suggestion on how to solve this problem?

Hmm, seems to me like there is an e-potential difference causing problems...
As a test, I'd suggest trying these cases:

   Disconnect Printer entirely from System, except for the power.
   Connect Printer to System, power Printer via Power Conditioner.
   Similarly, connect Printer to System, power Printer by a separate
     supply-line circuit.
   Ensure Common Ground between System and Printer.  In fact, make sure
     that the two are grounded via the Parallel Printer Cable.
   (Far shot) Move Printer far away from the System, but using the same
     supply-line circuit. 
   
Another remote possibility is that the Server System's power supply is 
operating at limit (i.e., marginal) and that the spikes are enough to
deprive the CPU of needed power.  (Remember, it only takes 1 bad opcode
taking less than ~100ns to crash.)

Could the other spikes you mention be a result of, say, an air conditioner?
Out here at the Caltech Campus, all major computing rooms have conditioned
power going into the machines with individually filtered (passively) outlets.


-- 
Joseph Chiu, Dept. of Computer Science, California Institute of Technology
1-57 California Institute of Technology, Pasadena, California 91126