Multiple motherboards from a single PSU
scott....@btclick.com
thought I'd run a theory past the electronically minded amongst you.
I'm going to power 4 mini-ITX boards from a single ATX PSU and I intend
parallel cabling all of the power leads, apart from PWROK and PSON#.
It seems to me that I need to feed both of these to each of the mobos,
but that I should protect each of them from back-feed with a diode. I'm
not sure which way round they should go, tho, because PSON# is active
low, from mobo to PSU and PWROK is active high, from PSU to mobo.
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)
Any thoughts, advice?
Cheers,
Scott
Frazer Jolly Goodfellow
news:1139778240.7...@o13g2000cwo.googlegroups.com:
Your diodes look to be the right way round, based on the signals
involved. You need to check the current passed when active, in
order to choose the right rating of diodes.
However, can I check your requirements: will there be one 'on'
switch for the system or one per motherboard? [so you can power
them on selectively]
scott....@btclick.com
Ideally, I'd like one per motherboard but I'm not sure what will happen
when you signal PSON to ground, starting a shutdown, but the power
lines stay active (because the other mobos are still running).
I have some sacrifical 400Mhz mobos so I guess I'll start with them!
Cheers,
Scott
Frazer Jolly Goodfellow
news:1139789960....@g14g2000cwa.googlegroups.com:
> Hi, thanks for your response.
>
> Ideally, I'd like one per motherboard but I'm not sure what will
> happen when you signal PSON to ground, starting a shutdown, but
> the power lines stay active (because the other mobos are still
> running).
>
On further reflection, the diodes in the PWR_OK connections are
redundant: once any one of the PS_ON#s is asserted (active low),
the PSU will assert PWR_OK (active high) and all 4 mobos will see
it, diodes or not.
Now comes the tricky bit. Under normal circumstances:
1) The power on button of an ATX PC provides a momentary low signal
to the motherboard.
2) The motherboard responds to that signal for its own purposes.
3) The motherboard then asserts PS_ON# to the PSU.
4) The PSU starts up and asserts PWR_OK.
In your 4-way independent power button scenario, the first mobo you
power up will see the right signals in sequence. But the other
three will all see PWR_OK go TRUE without having had a power on
momentary input. So the issue is: what does a mobo do in that
circumstance?
a) Do nothing until the power on button is pressed - the required
action.
b) Power up and run normally - unlikely, and not the required
action.
c) Power up in an undefined state and misbehave because step 2) was
missed out. *Definitely* not the required action.
Mike Tomlinson
scott....@btclick.com 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!
--
(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.
Daniel James
wrote:
> thought I'd run a theory past the electronically minded amongst you.
That rules me out, but ...
> I'm going to power 4 mini-ITX boards from a single ATX PSU ...
I had that idea, and came to the conclusion that it wasn't a good solution.
I'm now tending to think that giving each mini-ITX board its own DC-DC PSU
(such as one of these http://www.mini-itx.com/store/?c=10#picoPSU maybe?)
and running those from a single mains-to-12VDC PSU makes more sense ...
though the relatively high price of the pico PSU makes having for separate
ATX PSUs and damning the bulk/noise/heat/inefficiency tempting.
I'll be interested to hear how you get on, though, if you do follow the
single ATX-PSU route.
Cheers,
Daniel.
Johnny B Good
from Mike Tomlinson <mi...@NOSPAM.jasper.org.uk> contains these words:
> In article <1139778240.7...@o13g2000cwo.googlegroups.com>,
> scott....@btclick.com 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.
skotl
I see the point about not needing to diode-protect the PWROK line; it
should go to all of the boards.
As for what happens when PWROK hits a board that didn't realise it was
getting turned on; good point! I need to experiment with a couple of
old boards.
I think I also need to expirement with the PSON# to see whether it
makes sense to have 1 "master" board control power to the others. I'll
let you all know how I get on.
Cheers,
Scott
Dorothy Bradbury
o Do a Google for multiple SETI systems
---- some people ran 2-3 P2 boards off 1 PSU
o Use industrial N+1 power splitters
---- these are a backplane type board for ATX PSUs
---- you plug N+1 PSUs in to power N+1 PCs
---- a lot cheaper than N+1 power supplies
o Use DC-2-DC convertors with Industrial PSU
---- ensure the DC-2-DC can handle any drives you need
---- use a caged industrial 12V PSU to power of enough Watts
Overall...
o Cost it out carefully
---- DC-2-DC boards nor big wattage 12V PSUs are not cheap
---- DC-2-DC boards & mini-itx have a history of "tweaking"
o Multiple PCs off 1 PSU is a large single point of failure
---- PSUs are not so expensive as to make this idea economic
o 1U PSUs provide low size if needed
---- open-frame 350W 1U for 15ukp (new), fully-enclosed 30ukp+
---- 101x42x190mm typical size, so if size critical these help
There are solutions to run 1 PC off 2 PSUs - a simple board with
MOSFETs to handle voltage differentials between the two. Used
by Dell to provide dual-1U redundancy on their 1U servers.
Not many solutions would want to run 2 PCs off 1 PC PSU, since
the failure of 1 causes the loss of both machines re increased risk.
If you go with a DC-2-DC solution, Google carefully for Forums
that have proven solutions and know the caveats. The main one is
ensuring a sufficient wattage PSU - re startup loads & such like.
--
Dorothy Bradbury
www.dorothybradbury.co.uk for NMB-MAT & Panaflo (Panasonic Industrial) fans
Daniel James
> ---- DC-2-DC boards & mini-itx have a history of "tweaking"
Can you expand on that, please? What do you mean by "tweaking"?
> If you go with a DC-2-DC solution, Google carefully for Forums
> that have proven solutions and know the caveats. The main one is
> ensuring a sufficient wattage PSU - re startup loads & such like.
Mini-ITX systems typically have very low power requirements -- especially
if using notebook drives -- so although DC-DC supplies can be quite low
power (though some are 200W or more) they should easily be capapble of
powering typical mini-ITX systems with sensibly-chosen components.
I'm thinking of VIA-based systems here, systems using S478 P4s (M or
otherwise) will draw more.
Even so, it says here http://www.mini-itx.com/store/?c=2#pentiumm that
1.4GHz Celeron-M systems can run on 60W (peripherals not listed).
Cheers,
Daniel.
Dorothy Bradbury
>
> Can you expand on that, please? What do you mean by "tweaking"?
o Identify which one you believe is suitable
o Check through the SFF/Mini-ITX forums & elsewhere
There are certain limitations...
o Desktop boards impose quite high demands
---- requiring sufficient DC-2-DC wattage rating
---- requiring sufficient PSU wattage rating to the DC-2-DC unit
o Verify compatibility of the DC-2-DC unit with the board/CPU
---- some may be suitable re headline wattage
---- however verify it actually works
o Desktop drives impose high startup demands
---- 3.5" pulls 2A at startup on the 12V rail for example
Going by wattage etc may not provide a working solution.
Some sites indicate which will work with a desktop board & which will not.
> Mini-ITX systems typically have very low power requirements -- especially
> if using notebook drives -- so although DC-DC supplies can be quite low
> power (though some are 200W or more) they should easily be capapble of
> powering typical mini-ITX systems with sensibly-chosen components.
o If using a VIA C3 or Skt370 SBC board then by wattage yes
o If using desktop drives, check carefully - there are exceptions
o If using desktop boards check carefully re compatibility
Most likely the convertor wattage rating per rail needs considering,
and that is not always stated. 200W "should" be ok for desktop boards,
but I know the 120W versions do not work with many desktop boards.
If you are going for P-M Cel-M solutions, multiple 1U PSU are minor.
Check the reviews on the boards re performance (if going for C3) and
general Video or I/O performance (P4 Insight) & reliability (all of them).
Industrial Skt 370 boards can be had for 45ukp...
o Matrix-Three I think the Ebay name
o 5.25" SBC, i815 chipset, PCI slot, USB, Skt370
---- used 45ukp
---- new 95-115ukp depending on VIA or i815 chipset
o MIGHT take power off a single 5.25" connector
---- in which case easy to do several boards off a single PC PSU
o Cheap to add a low power Tualatin 1.2Ghz or fast 1.0-1.4Ghz P3
---- far cheaper & much faster than the VIA C3 things
www.bvm-store.com is one useful stop re Mini-ITX/Industrial-SBC.
--
Dorothy Bradbury.
Jeff Gaines
> Even so, it says here http://www.mini-itx.com/store/?c=2#pentiumm
> that 1.4GHz Celeron-M systems can run on 60W (peripherals not listed).
A slight change of direction - what would that be like as a file server
on a home network? Literally just storing files and dishing them put
and running an automated backup to a NAS each night.
--
Jeff Gaines - Damerham Hampshire UK
Using XanaNews 1.17.6.6
Bernard Peek
<jgaines...@yahoo.co.uk> writes
>On 14/02/2006 Daniel James wrote:
>
>> Even so, it says here http://www.mini-itx.com/store/?c=2#pentiumm
>> that 1.4GHz Celeron-M systems can run on 60W (peripherals not listed).
>
>A slight change of direction - what would that be like as a file server
>on a home network? Literally just storing files and dishing them put
>and running an automated backup to a NAS each night.
It's overkill. On a 100Mb network you probably don't need that much
horsepower to keep up, one of the Via C3 boxes would do the job
perfectly well. Make sure you give it lots of RAM though to help with
disk caching.
--
Bernard Peek
London, UK. DBA, Manager, Trainer & Author.
Jeff Gaines
> It's overkill. On a 100Mb network you probably don't need that much
> horsepower to keep up, one of the Via C3 boxes would do the job
> perfectly well. Make sure you give it lots of RAM though to help with
> disk caching.
Thanks Bernard, lot cheaper as well! How much is 'a lot' of RAM, 1GB?
Bernard Peek
<jgaines...@yahoo.co.uk> writes
>On 14/02/2006 Bernard Peek wrote:
>
>> It's overkill. On a 100Mb network you probably don't need that much
>> horsepower to keep up, one of the Via C3 boxes would do the job
>> perfectly well. Make sure you give it lots of RAM though to help with
>> disk caching.
>
>Thanks Bernard, lot cheaper as well! How much is 'a lot' of RAM, 1GB?
That sounds about right. It's probably overkill too but if you are
fitting out a new machine I'd say that it would future-proof it.
I've just replaced my Celeron 400 server with an 800MHz box and that's
fine as a file-server. It's also a database server too and that doesn't
seem to overtax it. Of course a lot depends on exactly what you intend
to do with the server. A file server might be delivering a few word
documents or feeding multiple video streams.
One of the advantages of the C3 systems is that they can operate with no
fans and no moving parts other than the disks. That should make them
more reliable and less likely to be affected by dust. Dust could be a
problem if you put a machine in a loft or cupboard.
If you do have a server hidden away somewhere it is less likely to get
nicked if Burglar Bill pays a visit.
Jeff Gaines
> > Thanks Bernard, lot cheaper as well! How much is 'a lot' of RAM,
> > 1GB?
>
> That sounds about right. It's probably overkill too but if you are
> fitting out a new machine I'd say that it would future-proof it.
Many thanks :-)
skotl
Think I'm getting there, after a bit of R&D and soldering. There are 4
VIA boards, and their combined rail requirements (plus the disks) are
more than met by the 450w supply I'm putting in.
I could have put 4 ATX supplies, or fed DC-DC supplies (although the
overall power consumption would be higher, given the losses in the
DC-DC tx's) but I'm after a neat, pretty and over-engineered solution.
I bought a brand new 6U server chassis and everything (including the
KVM switch and USB switch for the CDROM) is going in here.
I have also figured out the diode shenanigans; PWROK is just going to
be fed to every board, but PSON# has a 1n4148 flowing *from* the PSU to
each board. This means that as long as one of them is still connecting
their PSON# to ground, current flows thru the diode and the PSU
supplies juice.
When the last of the four raises PSON# (actually, allows it to float)
then PSON#' at the PSU will float up and the power will be cut.
Worst effect on the other boards that were waiting is that they sit
with the OpSys saying "it is OK to remove power".
Final trick is that I couldn't figure how to split all the power cables
without using veroboard (doubts over current capacity) or splicing them
together (yuck), so I have invented my own bus with eight 2.5mm rails,
and the cables are looped and soldered to the rails.
There are eight cos that is the number of discrete supplies / signals
from an ATX PSU.
All works, with 2 boards on the bench. I'll let you know when I tie 'em
all up for real!
Cheers,
Scott
Mike Tomlinson
<scott....@btclick.com> writes
>Final trick is that I couldn't figure how to split all the power cables
>without using veroboard (doubts over current capacity) or splicing them
>together (yuck), so I have invented my own bus with eight 2.5mm rails,
>and the cables are looped and soldered to the rails.
Any chance of a photo?
skotl
I'll post the URL then.
Cheers,
S.
Johnny B Good
from "skotl" <scott....@btclick.com> contains these words:
> Thanks Dorothy,
> Think I'm getting there, after a bit of R&D and soldering. There are 4
> VIA boards, and their combined rail requirements (plus the disks) are
> more than met by the 450w supply I'm putting in.
> I could have put 4 ATX supplies, or fed DC-DC supplies (although the
> overall power consumption would be higher, given the losses in the
> DC-DC tx's) but I'm after a neat, pretty and over-engineered solution.
> I bought a brand new 6U server chassis and everything (including the
> KVM switch and USB switch for the CDROM) is going in here.
> I have also figured out the diode shenanigans; PWROK is just going to
> be fed to every board, but PSON# has a 1n4148 flowing *from* the PSU to
> each board. This means that as long as one of them is still connecting
> their PSON# to ground, current flows thru the diode and the PSU
> supplies juice.
If the PSU 'Power on' input uses TTL logic levels, the diode volt drop
will compromise the "noise margin". However, it's very likely to be
using cmos input with pullup (hopefully cmos that doesn't try to emulate
the equal to or less than 0.8v is logic 0 and 2v or greater is logic 1
levels of real TTL :-).
It might be worth testing with a string of three 1n14xx diodes to
confirm use of classic cmos input thresholds (45 and 55% Vdd). If the
input activates the PSU using three diodes, you'll know for sure that a
single diode volt drop is not going to cause any problem.
If you can't get even a two diode string to switch the PSU out of
standby, you'll have to use either low resistance germanium diodes or
shotky barrier types in place of the humble 1n14xx series you're
currently using.
> When the last of the four raises PSON# (actually, allows it to float)
> then PSON#' at the PSU will float up and the power will be cut.
> Worst effect on the other boards that were waiting is that they sit
> with the OpSys saying "it is OK to remove power".
Sounds an optimal solution to me.
skotl
Someone else had suggested schottky diodes to me, so I think I'll try
them (my "wooden" bus allows me to snip the diodes off and replace
them).
If all seems to be as before, or I get any extra benefit, then schottky
may be the way to go.
Does anyone see any potential harm in me leaving the mobos at "OK to
remove power"??
Ta,
S.
Mike Tomlinson
skotl <scott....@btclick.com> writes
>Does anyone see any potential harm in me leaving the mobos at "OK to
>remove power"??
No.