USB-OVP

[Werner Almesberger]

Rod Whitby wrote:
> I have two GTA01's that are dead due to 12V being applied to the power
> supply of an upstream USB hub instead of 5V ...

That's an interesting scenario because it shows the effect of abuse
of an inherently less reliable interface being propagated to a
supposedly more reliable interface.

E.g., this discussion is about whether it's okay to feed USB's VBUS
to an input that has an "absolute maximum rating" of 5.5V without
any further protection.

The USB 2.0 spec says that the maximum VBUS is 5.25V, so the question
was "how likely is it that someone is violating the spec ?"

My gut feeling would be about the same as hopping on the swivel chair
in my office to change a light bulb. Probably nothing bad will happen,
but it's not safe. (So yeah, I do it anyway.)

Then Joerg found that the USB charger spec allows transients of up to
6V. I didn't expect that. So then the question became "how serious
are NXP about this 5.5V limit ?"

So that's more like running a red light with your car. Chances are
that you still get away with it, particularly if you're "careful",
i.e., you watch for conflicting traffic, cameras, or the police, but
you're probably aware that you're doing something very wrong.

Now the hub scenario extends this: anything connecting to USB is
almost certainly another USB device, so we can be reasonably
confident that this device will make a honest effort of playing by
the rules.

However, whether the power supply that you plug into that barrel
connector of your hub outputs the 5V the hub expects, 12V, -12V,
or something meandering around the general vicinity of 5V, the
plug will fit just as well.

So now the safety of the Neo depends on either nothing ever going
wrong at that fragile interface, or the hub's own protection
circuits preventing further disaster. Now, if we're too cheap to
care about adequate protection for our USD 399 device, how likely
is it that the makers of as-cheap-as-possible no-name generic USB
hubs aren't ? (And your example shows that the honest effort we
assumed above doesn't seem to go that far.)

So this is more like building your chalet in a known avalanche
zone and counting on everybody else being prudent enough to not set
off an avalanche.

I can see a trend in the risks we discover in this process, and it
seems that the initial gut feeling that this isn't right was quite
correct.

- Werner