Re: Multiple motherboards from a single PSU

The message <gAI7UhW4DD8DFwCG@xxxxxxxxxxxxx>
from Mike Tomlinson <mike@xxxxxxxxxxxxxxxxxxxx> contains these words:

In article <1139778240.752134.158280@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
scott.leckie@xxxxxxxxxxx writes

So, I reckon I'm going to do this;

(PSU) PSON---------->|------pin 14 (Mobo1)
|
|--->|------pin 14 (Mobo2)


(PSU) PWROK--------->|------pin 8 (Mobo1)
|
|--->|------pin 8 (Mobo2)

My first thought is that you'd be OK just paralleling PWROK (this used
to be called POWER GOOD) from the PSU to each board without the diodes.
It's just a logic level signal telling the motherboard that the PSU
voltages have stabilised and that the PSU's functioning. I suspect this
is simply wired straight from +5V on many cheapo PSUs nowadays. You
could even try leaving it out altogether - dunno if present-day boards
bother monitoring it.

PSON, usually a green wire: why not take this to one board (let's call
it the master), omitting the connection from the other three, and
connect the front panel power button to the master board? Pushing the
power button then turns on power to all four.

Disclaimer: it's early and I've only had one coffee!

Well, that's a very good suggestion, assuming the psu has sufficient
power capacity for the job and the four MoBo combo is to be treated as a
'single unit' wrt to powering up and shutting down.

The PG signal is a lot more important than you might think. Early
personal computers (pre-IBM PC) had special 'power on' circuitry to hold
the CPU in a reset state for a short delay after application of the DC
bias voltages in order to guarantee a safe and predictable starting
condition. The IBM design shifted this circuitry to the PSU itself. The
PG input line on an IBM (clone) PC motherboard is effectively an
alternative reset line.

The big advantage of letting the PSU generate the PG signal is that it
can also provide for an orderly shutdown when power is removed, planned
or otherwise (especially otherwise! :-).

Now, fairly obviously, the PG is generated from the voltage monitoring
circuitry and, on power up, will be asserted a few hundred milliseconds
after the voltages have stabilized. More importantly, however, the PG
signal is used to force a reset condition on power down when the
voltages are just about to drop below minimum spec.

The benefit of this being that short interruptions in mains supply need
not cause a shutdown for very short interruptions that failed to cause
any voltage dropouts. Longer interruptions in mains supply will negate
the PG signal and cause the PC to reset. In short, if the PC didn't
reboot in response to any such interruptions of mains as might be
indicated by the room lights flickering off, then you can be sure it's
operation wasn't compromised.

When the short interruptions of supply are just long enough for the PSU
to negate the PG line, the resulting reset is a far better option than
allowing the logic circuitry to start behaving erratically for several
thousands of clock cycles as the bias voltages fall below minimum spec.
You might say, "Discretion is the better part of valour".

This is especially true of hard disk drives which might have been in
the middle of writing out a sector to disk. The disk controller will
recieve such a reset and complete the current writing operation before
halting, thus obviating any false bad blocks due to the CRC error that
would otherwise have occurred.

It's true that such events will cause FS errors but these, at least,
can be readily corrected post-mortem.

HTH

--
Regards, John.

To reply directly, please remove "buttplug" .Mail via the
"Reply Direct" button and Spam-bots will be rejected.

.