Archived from groups: alt.comp.periphs.mainboard.asus (
More info?)
In article <P%pde.3484$V01.1747@newsread1.news.atl.earthlink.net>, " -"
<xvvvz@hotmail.com> wrote:
> >>In your description, you noted "Today, while doing other things,
> I noticed the screen go black...". Did you, perhaps, plug in a
> USB device to the computer, just before this chain of events ?
> There is an issue with all motherboards based on ICH4/ICH5/ICH5R
> Southbridge chips, where electrostatic discharge into a USB
> port, causes the Southbridge to go into latchup. That is where
> a phantom SCR forms inside the chip, when ESD exceeds the max
> current allowed on any input or output. In full latchup, the chip
> can get hot enough to burn, and you might notice visible damage
> to the Southbridge chip (and no, adding a heat sink to the
> Southbridge will not stop the phenomenon from happening,
> and neither would it have prevented device destruction). <<
>
> Paul -
>
> Do you have any more information you can share/point us to in regard to how
> to prevent this problem or know when it is occurring so that you can
> minimize damage? I am unfamiliar with the term "latchup." Is this
> something that would be reset if you were lucky enough to coincidentally
> power your PC off (hopefully before things burned up) or is it a "once the
> damage is done, the damage is done" type of thing?
>
> Thanks in advance,
> Doug
http://tw.giga-byte.com/Motherboard/Support/FAQ/FAQ_456.htm
There are different levels of latchup. "Micro-latchup" involves
a tiny portion of a circuit. The kind of latchup in the ICH5
is the major kind, and device destruction would happen so quickly,
I doubt you could reach the power switch in time to stop it.
(Under a second ?) Only an electronic circuit designed to detect
overloads would have a chance to limit damage, and that is not
a normal design requirement for digital circuit design. (This is,
after all, a chip fabrication error, and is not a normal behavior
that can be planned for.)
My recommended workaround goes further than what Gigabyte is
recommending. One poster here claims he lost a motherboard
by plugging a USB device into the rear motherboard USB connectors,
and those connectors do have working ground shields. That
tells me the Intel problem is much more severe than the
Gigabyte web page above is suggesting.
As a result, my current recommendation is to use a USB 2.0 PCI
card. Do not connect to the Asus motherboard USB headers at all,
and do not use the USB motherboard connectors on the back of
the computer. That will reduce the exposure of those ports
to static electricity to the maximal extent possible.
Static discharge can work directly, or a static discharge in
one conductor can induce energy into an adjacent conductor.
That means, for these sensitive USB ports, we don't want any
"antennas" on them, so the less wire connected to the USB
headers/ports the better. The USB 2.0 PCI card should give you
a good USB solution, without the same risk of a meltdown. One
would hope the PCI card faceplate and its connection to the
chassis, would form an effective path for dissipating a
static discharge.
ESD protection of ICs has come a long way while I've worked
in the industry. Many devices get a 1001V or a 2001V ESD
rating now, and in the beginning, you would not get nearly that
much protection. In the case of the Intel Southbridge,
some mistake must have been made in the protection
structures for this to happen, either that or a layout mistake
was made during the design. While the core of chips is laid
out by automated tools, there are still design issues around
the pad area of chips that are handled manually. The people
who do that work are called "polygon pushers", as they can
make manual changes to device structures. Some mistake
or bad assumption must have been made in the pad area of
the chip, in particular the D+ and D- signals on each of
the eight USB ports.
The worst ICs I've ever worked with, were destroyed while
sliding down the inside of an ESD protection plastic tube!
50% of the devices were destroyed before even making it
to my ESD protected benchtop. Adding external clamp diodes
to those chips, in fact, made the chips bulletproof, and
the chips could then be safely handled and passed from person to
person. So, there have been chips which are much worse
than this Intel problem. The manufacturer of those
devices I was using, admitted there was absolutely no
protection from static on the devices.
The worst transistors for static are insulated gate MOSFETs.
Those ship with a steel spring wound around their legs, to keep
all the signal leads at a common potential. You solder the
transistor into the circuit board, then pull on the coil spring
to remove it from the legs of the transistor. A transistor
like that can have an ESD rating of only 30V, which is no
protection at all. The best RS-232 transceiver chips,
have a 15000V ESD rating. So, there is quite a range of
static proofing in semiconductor devices.
One thing that surprises me, is I don't actually see a stated
ESD rating on Intel datasheets. Makes me wonder...
You can see some mention of testing for static electricity
protection here (near the bottom of the page). The
"human body model" involves modelling a human as a capacitor
charged to a high voltage. There are some cute HV probes you
can get, that will deliver a zap to a piece of electronics
hardware, and a company that does a thorough job of designing
their electronics, would use this kind of testing to prevent
sensitive designs from leaving the factory. I suppose
most motherboard manufacturers would assume the chip maker
had taken care of this detail.
http://www.maxim-ic.com/appnotes.cfm/appnote_number/654
HTH,
Paul