wad at media.mit.edu
Sat Dec 19 02:50:51 EST 2009
Turns out my mail to on Wednesday night was a little optimistic.
On Thursday, we dug deeper into a self-test failure
that we had all noticed but it hadn't been a high priority.
It turns out that a series of errors were causing the lid
switch in most XO-1.5 (B2-C1) laptops to "flap". We had
left old debug code left in the EC and zero ohm resistors
populated so that both the EC and the VX855 Southbridge
were connected to the lid switch. While a debouncer in the Linux
lid switch detection was preventing this from being noticed,
units do see an abnormally high rate of lid switch interrupts!
OFW, on the other hand, frequently though the lid had been
open and closed. This will be fixed by the next firmware release.
The build of the remaining units started late on Thurs.,
and didn't finish until Friday morning. When Quanta
tried to download the Linux image, about two-thirds of
the laptops couldn't talk to their WLAN card. (Which OFW
reported as a MAC address of all FFs.)
We are still verifying, but what I believe happened is
that the modifications Gary and I made Wed. (and he applied
to all machines) to fix a WLAN 3.3V leakage problem exposed
a power sequencing issue with the WLAN card. This card
was specially designed to use multiple supply voltages from
the MB. This saves power, but we are the first to do this with
the 88W8686. Gary has tested an ECO on three boards and it
worked, but conclusive test are still needed. If the WLAN card is at
all warm, the existing power sequencing works so this problem
is only seen on first boot when the unit is cold.
The C1 laptops will hopefully be shipped out to 1CC on Monday
or Tuesday of this week.
More information about the Devel