Subject: Re: NetBSD kills sparc 10 PSU
To: None <port-sparc@netbsd.org>
From: Greg A. Woods <woods@most.weird.com>
List: port-sparc
Date: 11/21/1999 14:51:15
[ On Sunday, November 14, 1999 at 14:27:30 (-0800), Jonathan O'Brien wrote: ]
> Subject: Re: NetBSD kills sparc 10 PSU
>
> Anyway, disconnect it and see if they'll turn on without anything
> attached.

I've not looked in detail at the specs for more recent Sun power
supplies, and indeed earlier models (eg. those in Sun-3s) were OK, but
it's not always safe to run a switching power supply with no load.  Some
are designed for this; some are designed to shut themselves down; and
some will just croak for good with all the smoke getting out of the
wires and important parts (eg. cheaper PC-clone supplies).

For power supplies that can be controlled by a low-voltage signal, such
as those for ATX motherboards and presumably those in newer Suns, all
kinds of things could potentially get "stuck" in the power control
circuitry which might cause them to exhibit this behaviour.  Many such
power supplies have a trickle current available on one or more outputs
even when "powered down" so that things like "wake-on-*" services can
still function.

Disconnecting from the AC side for a few *minutes* may sometimes allow a
power supply to completely discharge (another no-no with cheap switching
supplies is to touch their PCBs and capacitors even with no wires
attached until after you've manually discharged them with an appropriate
resistor) and thus reset their electrical state to a normal expected
state where they'll then power up normally.  I had some AT&T gear once
that would get fooled into thinking that it was getting 230vac after a
minor power glitch (maybe it wasn't that minor after all! ;-) and they
had to be unplugged for about five minutes before they'd work with
115vac again.

The very first computers I ever saw with CPU control over the power
supply were AT&T 3B2s (way way back in the mid 1980's! ;-).  Their power
switches would generate a "powerfail" signal to the motherboard, which
would be passed as SIGPWR to init and a final uadmin() system after all
shutdown procedures had completed would signal the power supply to
finally actually turn itself off.  Since then SGI, Sun, Digital, and now
Intel PC-ATX systems have begun to mimic this feature.....

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>