"Jim Gibson" said:
Interesting..
I've just RMA'd a P4800 SE that had all 8 USB ports die.
Everything else was fine. It would boot, all drives worked, video never
changed, my parallel printer kept on printing, PS/2 ports worked well.
But trying to get a USB port to work was useless.
I spent 4 days trying everything under the sun - but not a thing worked. I
tried everything in the message below + half a dozen otherthings.
All USB ports died while plugging in a Minolta camera into one of the front
USB ports of my Antec case.
The system kinda gave a quick power down - kinda like a brief 'moaning'
sound, then - no USB.
I had connected the same camera many times before.
I had read about USB 'latch-up' and static discharge and attempted to ground
everything before I used the camera.
In desperation I fired off a message to Antec asking for advice and a
schematic of their USB board at the front of my SX635B case.
I receveived a generic reply part of which read:
"The Antec USB connector follows the Intel standard layout. Majority of the
motherboards follow the same standard as well. Intel Standard 10pin Layout
Pin 1 = Power Pin 2 = Power<BR>
Pin 3 = Data - Pin 4 = Data -<BR>
Pin 5 = Data + Pin 6 = Data +<BR>
Pin 7 = Ground Pin 8 = Ground<BR>
Pin 9 = Key or Blocked pin Pin 10 = Ground <BR><BR>
Warning: If your motherboard contains a USB pin labeled OC or NC, make sure
that no wire goes to it. You can lift the black tab and pull the wire out or
simply clip the wire going to it. The pin for it is usually in pin 10. When
a wire is connected to OC or NC, all the USB ports can stop working, the
system can automatically shut down, or the secondary IDE is disabled.
<BR><BR>"
On checking my USB cable from the case to the board header, there was indeed
a ground wire leading to pin 10 (NC).
Was this the cause of the failure??
JG
Sorry I didn't look back at this thread sooner.
So, now we have two reports of a slightly new twist on the
problem. And the "moaning" sound tells us it is hardware :-(
In terms of the USB interface, there are four pins. Two data
pins D+ and D-, the VCC and GND. According to an Intel reference
schematic (available for free download from Intel), there is a
switching device in the VCC path to the USB port. Looking at
my P4C800-E Deluxe motherboard here, there are two Polyfuses
(marked with "160" on them) near the USB headers. That tells
me the power to the USB ports is limited. It also tells me
that if the VCC pin on the USB header is shorted to GND, not
enough current can flow to cause the power supply to "moan".
(These Polyfuses likely have a rating in the 1 amp vicinity.)
Now, with respect to the Antec information. Some USB headers
have a signal called OC. It stands for OverCurrent, and is
an active low input on the motherboard. The idea is, a device
plugging into the header, is allowed to measure the USB port
current. If the current exceeds half an amp, the measurement
circuit alerts the motherboard via the OC signal. If you
accidently connected a GND wire to the OC signal on some brands
of motherboards, about the worst that can happen, is the
motherboard will be concluding the port is in overcurrent.
Thus, the port would be out of service, but I don't see any
way for permanent damage to result. Once the cable is removed,
the logic signal will say there is no overcurrent.
I just got out my multimeter, set it on its highest range,
and touched the NC pin on my P4C800-E Deluxe USB header, while
grounding the other lead. The meter reads infinity. We cannot
definitely conclude anything from the reading, except to say
the pin is not grounded.
I would say, if a USB front cable happens to ground the
Asus NC pin there is no damage. If a USB front cable grounds
an OC pin placed in the same location (pin 10), the port won't
work right, but there shouldn't be any permanent damage.
Removing the ground wire from pin 10 of the cable assembly,
should restore the port to its function again.
According to the Intel schematic, there are a couple of options
for OC. If a motherboard is equipped with fuses, then there is
no need for the OC signal. If a motherboard is unfused, then
the USB interface board is supposed to measure the current and
report to the processor in the event that more than 500mA is
drawn. The motherboard would then use its series FET to switch
off the USB header power. So, perhaps that is why I cannot
find a switching device on my motherboard. It could be, with
the Polyfuses in place, that no switch is needed, and neither
is an OC pin needed on the 2x5 header.
Now, to the analysis of the latest evidence. The moaning tells
us that a slug of current was drawn from one of the power
supply rails. The fact that the moaning only lasted a few
seconds, tells us that the short, wherever it was, is now
burnt out :-( The fact that the USB mouse still lights up,
means the fault was not in the USB power path after all.
Unfortunately, that leaves with only one other path for a
problem. It is the latchup problem. Something happens via
D+ and D- data leads. We know, based on a report of a failed
motherboard via the rear USB ports, that this problem is
not likely to be purely via ESD. That is because the design
of the USB connector and cable, guarantees the shield is
grounded as the connector is inserted. There should be much
less (but not zero) chance for ESD under those circumstances.
I still don't feel comfortable blaming this problem only on
ESD, at least for a rear port failure.
Based on people observing the Southbridge getting really
really hot, that tells me most people suffer major latchup
that causes complete destruction of the Southbridge.
As I've explained in a previous post, there are several
levels of latchup, of which the only two I've experienced
in the lab, are micro latchup (where a small number of logic
gates refuse to function, and powering down the device
leads to complete recovery of function - in this case, not
enough current can flow through the tiny gate to burn
anything), and major latchup (device gets red hot). There
is another kind of latchup, but I don't remember the
description right off hand.
It could be that the PNPN phantom SCR circuit, that forms
between the power rails of a portion of circuitry about
to go into latchup, is drawing its current mainly from a
few bonding wires near the USB pads. That burns out the
bonding wires that power D+ and D- pads, leading to the
rest of the logic inside the chip being in perfectly good
order, but the interface is toast.
That, of course, raises the question of how the problem
knows to "just burn out the USB pad power" versus
"frying the whole damn chip". I don't have an explanation
for that selectivity, so the theory isn't perfect.
While this is an interesting twist on the "completely
toasted" southbridge problem, it won't change my
recommendation:
1) Don't use any of the USB ports on a motherboard that
uses ICH4, ICH5, or ICH5R Southbridge. There are no
reports yet of the ICH6 based P5xxx boards being
affected (and strangely, there are no reference
schematics available for the 915/925 chipsets either,
so we won't be able to observe any external components
added to the motherboards to solve the problem).
2) Do not connect front panel wiring to the USB headers.
Doing so is just creating an "antenna" for ESD. Since
we don't know what the lower limit is on ESD causing
latchup, there is no reason to be taking chances.
3) Install, use, and enjoy, a separate USB 2.0 PCI card
with your Intel motherboard. If you buy a USB card
that has a 2x5 header or a 1x5 header on it, you might
even manage to connect your front panel wiring to it,
and get at least one port on the front of the computer.
HTH,
Paul
