Note to Jeff: befw11s4 hangs cured by cryogenics

Repost: corrected unintended long lines.
=================

In a previous thread, I mentioned that my befw11s4-based outdoor access
point was experiencing the well-known hangs during high throughput and
that I had used an SNMP power controller driven by monitor scripts to
power cycle the box as a solution. I also mentioned that after a
prolonged deep freeze (several weeks at minus 20 degrees F or colder),
the unit 'healed' and had not experienced further hangs.

Now that the weather is warmer, I can report that it continues to
operate without hanging during high throughput. The nature of the
'repair' is however unknown, and speculation is welcome. The unit
remained under power during the severe cold period, and connectivity
on the LAN port was uninterrupted, however the radio drifted off channel
and clients could not connect until the temps rose to above zero degrees
F. for some days. One could imagine some trivial dimensional changes
at the pcb level, or perhaps something more fundamental involving
changes in capacitor chemistry, IC bonding wire integrity, discrete
components changed in value or who knows what in the silicon is
at work? In any case the results seem to be permanent and decidedly
for the better.

Michael

On Mon, 28 Apr 2008 01:37:17 -0500, msg <msg@_cybertheque.org_> wrote:

>Repost: corrected unintended long lines.
>=================
>
>In a previous thread, I mentioned that my befw11s4-based outdoor access
>point was experiencing the well-known hangs during high throughput and
>that I had used an SNMP power controller driven by monitor scripts to
>power cycle the box as a solution. I also mentioned that after a
>prolonged deep freeze (several weeks at minus 20 degrees F or colder),
>the unit 'healed' and had not experienced further hangs.

Amazing. My only guess is that the wire bonds inside the chips are
moving around. When high thruput causes the chips to warm up, they
move again.

Did you ever try the 1.52.06 firmware that was leaked out of Linksys?
<http://802.11junk.com/jeffl/crud/BEFW11S4v4/>
<http://groups.google.com/group/alt.internet.wireless/msg/f618788a5e1739cc>
Note that's it's for the v4 version only and has to be installed using
TFTP as the build didn't include proper checksums on each line. I
have it installed on 2 units, but have not had the time or inspiration
to try it.

>Now that the weather is warmer, I can report that it continues to
>operate without hanging during high throughput. The nature of the
>'repair' is however unknown, and speculation is welcome.

Magic?

>The unit
>remained under power during the severe cold period, and connectivity
>on the LAN port was uninterrupted, however the radio drifted off channel
>and clients could not connect until the temps rose to above zero degrees
>F. for some days.

It was probably slightly above freezing inside the box due to self
heating. That implies that condensation was involved, where the radio
was basically under water. It's fairly difficult for the radio to
"drift" off frequency. However, a water soaked board is going to
cause problems. I ran into this problem trying to test electronics
over a wide range of temperatures. The worst temperature was the
transition zone around the dew point, where water condenses on
everything. A good quick test is to freeze a board (not running),
bring it out of the freezer to room temperatures, and quickly apply
power. The results convinced me that conformal coatings (of various
types) and good board cleaning (to prevent ionic contamination) is
required to survive such environments.

>One could imagine some trivial dimensional changes
>at the pcb level, or perhaps something more fundamental involving
>changes in capacitor chemistry, IC bonding wire integrity, discrete
>components changed in value or who knows what in the silicon is
>at work?

Yep. It might broken PCB traces inside the multi-layer board. I
would tend to favor the wire bond theory. Electrolytics have almost
no capacitance left at low temperature, but are usually oversized to
compensate. A change in their value might induce some hash on the
power supply leads, but I don't think that would cause hangs.

>In any case the results seem to be permanent and decidedly
>for the better.

Recycled wisdom: If it works, don't touch it.

>Michael
--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558