One of my workstations refuses to stay asleep. Often,
> just after being COMMANDED to sleep, it would reawaken.
> Sometimes, a short while later (i.e., just long enough
> for me to leave the room and return to find it running,
> again).
>
> Windows logs suck. But, I was able to trace the cause
> of the wake-up to "USB root hub" (I think there are
> six things with this name in the device tree!).
>
> I narrowed it down to the keyboard; unplug keyboard,
> command it to sleep and it stays asleep.
>
> [N.B. The keyboard works completely as expected when powered on!] >
> Fine. Replace keyboard, problem solved.
>
> *BUT*, I can't EASILY get any more of this particular model
> of keyboard (eBay-roulette?) so I would like to salvage it
> given that the "problem" is likely a hardware issue.
>
> First question is what *might* (likely) Windows be seeing that
> makes it think that a valid USB event has been detected?
> I.e., is it a "smart" pattern recognizer (like a WoL packet)?
> Or, just "any activity" on the data lines to the i/f?
>
> It seems that power conditioning on the keyboard should be a good
> starting place (it's ALWAYS power-related!).
>
> It also seems like Windows only leaves USB ports powered up
> during sleep *if* they were active while the machine was awake (?). >
It appears to me that the device drives a 'K' state onto the data lines for between 1 and 15ms. The 'K' state is the opposite of the idle state.
It's not necessarily the keyboard's fault - unplugging it prevents it from waking the system up, if that's what it's doing, but it also means that it's not plugged in. It could be the latter fact that makes the spurious wake-ups go away.
You could try plugging something else into that port instead.
Also plugging the keyboard in via a separate usb hub.
The results may hint at what to look at next.
Sylvia.