[Soekris] PCI card + 5V kills board

Bernd Walter soekris-tech at cicely.de
Mon Jul 7 15:25:33 UTC 2003

On Mon, Jul 07, 2003 at 04:56:47PM +0200, Poul-Henning Kamp wrote:
> In message <20030707142617.GA31285 at cicely12.cicely.de>, Bernd Walter writes:
> >I tried a Longshine USB2.0 card in a 4501 at 12V which seemed to work.
> >Next I tried JP6 supply with GND at 1+2 and 5V at 3+4.

I meant JP9...

> >The board died with a hearable tick :(
> >After replacing the Q3 (the driver bridge for 3.3V regulator) the
> >board worked again at 12V without additional cards.
> >I retried 5V with the USB card which killed Q3 again.
> >
> >I have no clue if my 5V supplying was wrong or the USB2.0 card isn't
> >really 3.3V aware - altough it has a 3.3V notch and claims to be
> >PCI2.2 which implies 3.3V operating.
> >My impression is that the card is just crap but I not happy to try
> >out cards until it runs.
> >Has anyone any expiriences to share?
> There are two issues:  PCI signalling voltage (3.3 and 5.0) and 
> PCI card power drain (3.3 and 5.0).

Thats's clear.

> The PCI signalling can only be 3.3V, end of story.  That is what
> the notches control.

Asuming the notch won't lie :(

> The power drain is an entirely different story: both 3.3V and 5.0V
> power is routed to the PCI card, but the 5.0V supply in the Soekris
> has very little spare capacity so if the card drains too much power,
> you have probably seen the result.
> If you feed your Soekris with +5V, this is not an issue.

It also wouldn't explain why the 3.3V died.
The 5V regulator as well as the switching regulator for 3.3V have
overload protection - at least in specs...
But my situation showed that the problem is reproduceable.

I've just found the reason.
PCI specs defines multiple powerlines: 5V, 3.3V, UIO, ...
UIO is either 5V or 3.3V - on the soekris it's 3,3V.
Now this idiotic Longshine board connects UIO with 5V - arglll.
So the fact that the card has a 3.3V notch at claims to be PCI2.2 is
just bogus.
The driver just broken by trying to be a hero and pulling down the
5V source to 3,3V.
I asume it worked with the 12V supply just because the 5V regulator
did not drive to much power for the 3.3V driver to break.
Sigh - but now I know what I have to lock for when I try another
board - so the next cards are not russion roulete.

As a warning to other users to watch out before trying:
The broken card connects at least 8A and 10A.
They are on the bottom side counted from the slot blend.

B.Walter                   BWCT                http://www.bwct.de
ticso at bwct.de                                  info at bwct.de

More information about the Soekris-tech mailing list