Re: Power Button on PWS

sms_at_antinode.org
Date: 02/29/04 

Date: Sun, 29 Feb 2004 01:31:18 -0600 (CST)

From: "Tom Linden" <tom@kednos.com>

> It appears to be a momentary switch. I looked at the wires and I couldn't
> conclude anything from that ther were 6 with a carbon choke on them.

   That's a ferrite core, not carbon, unless yours is peculiar.

> I tried holding the switch down, but that didn't work which is why I
> concluded that it was a momentary.

> Not sure why anyone would make a switch like that. I see no functional
> reason for it, but I am sure there must be one.

   Ever notice the last of the SHUTDOWN options?

Shutdown options (enter as a comma-separated list):
 REMOVE_NODE Remaining nodes in the cluster should adjust quorum
 CLUSTER_SHUTDOWN Entire cluster is shutting down
 REBOOT_CHECK Check existence of basic system files
 SAVE_FEEDBACK Save AUTOGEN feedback information from this boot
 DISABLE_AUTOSTART Disable autostart queues
 vvvvvvvvv vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
 POWER_OFF Request console to power-off the system
 ^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

It works, and it's hard to achieve that with a hard-wired power switch.

   On my PWS (old-style, no on-board SCSI, et c.), the power switch
seems, according to me and my trusty Simpson 270, to be a momentary
SPDT (Red = NC, Black = Common, White = NO).

   I did a bit of Googling to learn a bit about a typical ATX (or
similar) PC power supply, about which, up to now, I've never cared much.

   There's an always-on stand-by 5V supply output ("+5VSB") to run the
logic for the power control (manual switch, wake-on-LAN, wake-on-modem,
et c.), and a (TTL) control input ("PS_ON#"), which, when pulled low,
causes the other outputs to come on.

   As both power button switch contacts appear to be used (unlike the
simple NO reset button switch), I'd guess that it's being nicely
debounced, and then used to control a toggle-flip-flop which drives the
PS_ON# input. As the console can also switch it off, it might feed a
set/reset input on the same flip-flop, or it could be a bit more
complicated.

   If the flip-flop were waiting for an appropriate clock edge, a
permanently pressed power button would not do the job.

   It would take a bit of research to puzzle out the details, and I'm
not eager to rip that far into my system just now. If my guesses are
correct, I don't see an easy way to modify the current switch to provide
an automatic-on capability. As it needs to be in the stuff run by the
stand-by power supply, I also don't see how the console firmware could
control such a feature.

   It would have been fairly easy to build it in, controlled by a
jumper, say, which determines whether a power-on reset pulse (easy to
generate) reset or set the magic flip-flop, but without a lot of poking
(or a real diagram), it's not obvious to me where to start looking.
(There's a lot of miscellaneous/analog stuff on the riser card. In easy
sight, there are an NE555 timer and an LM339 quad comparator, for
examples.)

   If the capability was designed in, it should be easy to control (only
knowledge required). If not, adding it could be pretty tough.

   If I had a pair of apparently identical machines which acted
differently, I'd take a careful look for a suspicious jumper difference
(and then report my findings back to the group).

   Anyone with a diagram of the riser card is, of course, welcome to
contribute.

------------------------------------------------------------------------

   Steven M. Schweda (+1) 651-699-9818
   382 South Warwick Street sms@antinode-org
   Saint Paul MN 55105-2547