I have been looking into why I can't get my wireless Microsoft keyboard and mouse to work properly. That is missed keys, sticky keys and fairly broken network connectivity at the same time.
A wired HP keyboard and mouse through a PS/2 adaptor are fine.
People have written loads about power supplies, checking 5v, doing things with USB cables and that the device is unfortunately not compatible. I don't buy it.
"The NCP1117 is a low dropout regulator" only when "low" has the same meaning as in the sentence "bankers generally have low renumeration."
The design doesn't leave a lot of headroom so I have replaced my 3v3 regulator with another device that drops 250mV to be sure. I've also used different power supplies, both a powered and unpowered hub and scoped the 5v and 3v3 rails.
Also with the same wireless receiver dongle plugged into my Marvell Sheevaplug also running a recentish Debian I can dump the appropriate /dev/input device and it's fine in that though I'm not overly familiar with input event data I can see events that correspond to what my paws are doing on the keyboard and I can see what is convincingly the "keyup" event as there is a zero in a particular byte.
My other observation is the vast quantity of kernel output from the smsc driver which, if you didn't have performance problems before, you will now. Many "kevent 4 may have been dropped" per second but I have also one or two others. Thanks to the configuration many log files get this and for some reason not instead the "last messages repeated a squillion times" traditional sysloggers did by default.
Given that the smsc device runs off 3v3 now via my better regulator and that the network breaks at the same time I suspect some or all of
The smsc device
some electrical problem
poor device driver
Any thoughts welcome.