[Community-news] OLPC News 2007-07-28

[Walter Bender]

1. FCC certification: Quanta's Kenny Chung reported that XO has been
granted full FCC certification. This includes SAR (specific absorb
rate) at less than 20cm (the distance between the WiFI antennas and a
child can be less than 20cm). Other tests passed: Conducted Emissions;
Radiated Emissions; Power Harmonics; Electro-static Discharge; Fast
Transients and Burst Immunity; Surge Immunity; Induced Radio-Frequency
Immunity; Power Frequency Magnetic Field Immunity; and Voltage Dips.
Further UL and CE certifications are in process; these will require
C-build machines for final approvals.

2. Free Drop:  Mary Lou Jepsen and Quanta reviewed all the 10-point
free-drop data that we have been collecting over since December 2006.
The units are dropped on all corners, all side bumpers, and front and
back. Initially, we had dropped onto plywood, but this spring we made
the test tougher: we have been dropping on a hard steel plate, with
and without a carpet. B4 units pass a 150cm 10-point drops onto a
carpet-covered steel plate; a 105cm simulated slanted-desk "slide"
onto a steel plate; and a 80cm 10-point free drop onto a steel plate.
The laptop,  when dropped on the antennas, withstands a 150cm drop. To
put these data into perspective: a standard laptop only passes a 45cm
10-point drop on plywood (a much softer material than steel).

3. Trial-2 testing: Sometimes we easily miss the forest for all the
trees, including the hundreds of bugs we deal with. This week we have
a working mesh that does not require access points and operates across
the full internet as well, the ability to share most of our basic
activities including Chat, Write, Etoys, and Record. Kim Quirk and Jim
Gettys conducted a "train" test (they commute by train without
Internet access), which saw successful peer-to-peer picture sharing in
the Record Activity and text, audio, and project sharing in the Etoys
Activity. Sharing through the mesh is fun; it got the attention of a
number of people on the train.

The functionality of Trial-2 code is at a very good state; it is now
time to close it down and release it. To do that we are asking that no
one checks in bug fixes or any code without an express OK from Jim,
Kim or Dan Williams. At this point, we are only going to allow fixes
associated with critical suspend/resume issues and critical
regressions.

4. Trial-3 planning: Thanks to Simon McVittie and Dafydd Harries for
coming to the OLPC office in Cambridge to work through some specific
mesh, tubes, Salut, and Gabble issues and to help plan the Trial-3
feature set. Next week a team from OLPC and RedHat will be working
with Pentagram to finalize the mesh, UI, and Journal features for
Trial-3.

5. Etoys: The current focus of the Etoys team is stabilizing the Sugar
integration. Bert Freudenberg led the effort and now an Etoy project
can be saved into Journal. In the meantime, the other members of the
team continue to improve the system. Takashi Yamamiya and Korakurider
are working on the interface with gettext; now, the phrases used in
the Etoys system can be translated by standard external tools.
Yoshiki Ohshima and Bert made the system locale conscious; the system
switches language upon starting. Scott Wallace has made improvements
to some UI elements, including simplifying the sound recorder. Scott
also helped Kim Rose, Alan Kay, and Rebecca Cannara with the
documentation effort. Ted Kaehler worked on improving some example
Etoys that will be bundled with the distribution.

6. Temperature testing: Joel Stanly and Arjun Sarual set up a food
warming oven in the OLPC office. The oven is large enough to house
eight fully opened XOs and allows us to examine the behavior of the
laptops under temperatures ranging from a warm 40°C, up to a toasty
60°C and above. Some preliminary tests were conducted, examining the
operation of the battery charging systems under the extreme heat that
may be encountered by, say, a laptop sitting in full sunlight. One
motivation for this testing is that the NiMH batteries that are used
in some of the XOs lose the ability to be charged above 55°C. (The
newer LiFePO4 technology allows charging above these temperatures, for
when the need arises.) We are pleased to report the XOs ran flawlessly
in the extreme heat, even when the oven's unpredictable thermostat
inadvertently allowed the temperature to reach 68°C. Further testing
will take place over the coming weeks.

7. Touchpad: Joel also tested the new touchpad/keyboard subsystem from
ALPS, verifying the new power-saving mode it allows. We can
successfully power it down, saving almost 12mW, and bring it back out
of power saving mode when the user requires it again.

8. Backup: Chris Ball worked with Scott Ananian on the USB
backup/restore script for Trial-2. By putting restore into the
activation ramdisk, we have backup/restore happening with one command
(./usbupgrade) in the developer console. We would like to get that
down to zero commands, though, by having the autoreinstallation image
make the backup automatically too.

9. Performance: Marco Gritti, Tomeu Vizoso, Dan Williams, and Chris
Ball explored reducing our memory usage to make the current images
usable on B2-1 laptops (with only 128M of DRAM). All in all, there
should be 20M of (resident) memory that we can reclaim so far for
Trial-2; 10M due to Sugar importing libraries and icons that it
doesn't need, and 12M for the heavyweight DHCP server that we run when
associated with an access point. We'll keep working on this, as it has
a positive impact on performance on all versions of the XO.

10. School server: Scott Ananian, Dan Margo, and John Watlington
pushed forward on School server software builds. A stumbling block is
automating the install from a live CD. Dan ended his summer internship
this week. His efforts in this area have been appreciated, and we hope
that he continues to help remotely in the future.

11. Embedded controller: Richard Smith discovered that the SCI wakeups
were not handled properly and the power button was being asserted on
resume (unnecessarily), causing spurious power button events.  Richard
also discovered a race condition that was the reason for some EC
timeouts; Andres Salomon added a 2ms delay in the kernel as a
workaround, while Richard is working on a proper fix for the EC. After
three weeks of chasing a bug, the cause of a NiMH battery-charging
problem has been found. Q2C19 will fix this bug. The source of the
mistaken "power button" on resume was located and fixed. With these EC
bugs fixed, suspend/resume code should begin to function as expected.

12. Donations to the OLPC Foundation can now be made by credit card
through Google Checkout or with a PayPal account. The link to the
giving site is http://laptopfoundation.org/en/participate/. OLPCF
continues to accept checks as well.

Upcoming Highlights:

Jul 31 – Aug 6	Wikimania, Taipei
Aug 1–3		Squeakfest07, Chicago, IL

-walter
-- 
Walter Bender
One Laptop per Child
http://laptop.org