Do you have a question? Post it now! No Registration Necessary. Now with pictures!
- Posted on
- Testing USB ports
- Robert Miles
June 13, 2014, 4:25 am
rate this thread
Do any of you know a suitable way to test which of the USB ports have this problem, so I can stop using those ports and add a USB hub instead?
Re: Testing USB ports
Plug in a USB key and see if it is detected ?
USB ports don't seem to be like Parallel ports or Serial ports,
where some limited GPIO capabilities exist. On serial ports, you
can assert some flow control pins. And there are plenty of home
projects that use Parallel ports as programmable GPIO pins. But USB
doesn't seem to work that way. In general chip design terms,
I/O ports can be checked with boundary scan at the factory,
but that's about the only thing I can think of that might work.
There is a chip with a known problem in that regard. The Intel
ICH5/ICH5R Southbridge, we had a fair number of reports of failures
in the news group. When the first poster showed up with a report,
there were quite a few "me too" postings confirming a problem.
Reports after that have been more sporadic. The ICH5 was used
in Pentium 4 designs of the S478 era.
When an ICH5 fails in a major way, like in this picture, the
motherboard no longer boots. Since this chip never has a heatsink
on top, in designs, you can easily visually inspect for the mark.
The chip will not say "ICH5" on top - it will likely have an
Intel 82 type part number of some sort, which is obscured here.
Compare to this one. The "secret" printed on this one, means
the chip is from an Engineering Sample lot, and the motherboard
would be a relatively early prototype. Here, the part number is
If the bond wires inside the ICH5 "pop" like fuses rather quickly,
then all that happens is the USB I/O ports lose power. The logic
blocks are intact. The drivers install just fine. But the USB
ports end up "deaf", because the thing that drives out logic 1's
and 0's, no longer has power. So no logic state changes make their
way inside the chip.
| X <--- pops open here
Logic ----------- I/O ---> part of four pin connector
Ground Ground I/O pad goes into latchup due to ESD,
forms a PNPN junction (SCR) between +5V
and ground. The dead short, burns the
bond wire to +5V.
The above diagram and interpretation of it, is pure *speculation*
on my part. There is no public information from Intel, indicating
there is a problem with the part (excessive ESD sensitivity).
The Intel Journal years ago, boasted that Intel USB ports
have a 5 or 6kV or so rating on ESD sensitivity, which is
relatively good. The only thing better than that, is a MAX232
style Serial port, which is rated for 15kV. And such a high rating
was developed for those, after the MC1488/MC1489 era, where
our serial ports used to "burn out like light bulbs". I had to
replace a number of those in the lab. Eventually,
the industry figured it out, and the serial ports became
relatively bullet proof. The ESD rating uses HBM (human body model),
a high voltage probe with capacitive storage, delivers a jolt
to I/Os, as a means of proving they won't be damaged when an
end-user scuffs over the carpet and touches the pins of the port.
(Example reference to HBM...)
I have an ICH5R here, and mine is still alive :-) I always
get a chuckle out of that, when I write one of these answers.
Why am I so lucky ?
Another chip with some indication of troubles, is NEC USB2
chips. On those, the ports pop independently, so you can
have a quad port card, with one or two ports burned out.
And the test method there is the same - walk a USB key from
port to port, find the ports that don't seem to respond,
and mark them with a pen. Those failures don't have
a visible manifestation - there's no evidence of latchup
when those fail. And not really enough good quality field
reports, to say much more about them. For example, on the
ICH5, at least one failure occurred during POST and boot,
almost suggesting the user didn't actually shock it, and
it just spontaneously popped. Which is strange for latchup.
Re: Testing USB ports
What's a USB key?
This motherboard is too new for that to be likely. It uses an Intel
Today, enough computer programs reached a reasonable stopping point
that I tried a test for that. I plugged an external USB hub into each
of the external USB ports. Its indicator lights indicated that each
of the ports, including those with problems, had power. I then left
it plugged in, and plugged all the external USB cables into it. They
all worked. So now I have a workaround, with some delay for the
external USB hub.
That's more like my problem, except that the USB ports are on the
motherboard. I was unable to identify the chip supplying these