battery display broken

[Ian Smith]

On Tue, 20 Dec 2011 11:10:34 +0100, Dominic Fandrey wrote:
 > On 16/12/2011 16:28, Ian Smith wrote:
 > > On Fri, 16 Dec 2011, Dominic Fandrey wrote:
 > >   >  It seems something broke with the battery display. Last night it
 > >   >  showed 94% remaining capacity for more than 2 hours.
 > >   >
 > >   >  Afterwards I docked the machine (HP6510b) and rebooted it. Since then
 > >   >  more than 8 hours have passed, but it still shows 16% (the LED
 > > indicators
 > >   >  state that the battery is full and no longer charging).
 > >   >  ...
 > > 
 > > At least four things can go wrong.  The battery charging circuit might
 > > be broken (my first T23 failed that way after 5 years); the battery
 > > might just need 'conditioning' (discharged to exhaustion, beyond normal
 > > low-battery shutdown, then fully charged - perhaps twice), to reset its
 > > internal Coulomb Counter; the CC chip may be faulty; or the battery
 > > itself may be failing / have failed, usually one cell first.
 > 
 > The battery still does the expected 3 hours. That's about as much as I
 > can expect.

Good!  I suspected hardware because I've seen all of the above, and I'd
seen no software problems with charging on FreeBSD {6,7,8}.x, so for the 
sake of your hardware I'm glad all that was irrelevant :)

 > > Is the battery hot at this stage?  If that 'Present rate' is correct, a
 > > 3.35A/39.6W charge should tend to overheat the battery over time, if
 > > charging continues beyond full capacity, which may indicate a bad cell.
 > 
 > Are you certain about that? Decent lithium cells can be charged with 6A
 > without becoming perceivably warmer. I'd expect that at least 39W of
 > the 39.6W get stored in the battery.

Certain.  High charge rates are fine for bulk capacity, but once full, 
continued charging will (must!) overheat the battery, and with LIion 
batteries, failure modes may include explosion and/or fierce fire.

One recent local fire that completely consumed a house and sheds began 
just that way, unattended overcharging of large LIion batteries on an 
electric bicycle; I've yet to ask the poor guy what sort of charger it 
was, but any decent charge circuits monitor voltage, temperature and 
likely dV/dt at minimum - so please excuse my unwarranted concern :)

Even with shorted cell/s, laptops monitor temperature too .. "Brand X 
laptop brings down jetliner" must be a manufacturer's worst nightmare.

 > > 11.8V seems too low for a fully-charged 10.8V nominal LIon battery.  I
 > > have several 4.0 and 4.4Ah like the below, which charge to ~12.4V, and
 > > only get down to 11.8V while discharging, at around 85% nom. capacity.
 > 
 > It states 16% capacity. Everything fits that. It simply doesn't update
 > the readings. When I boot with a charged battery:
 > Present voltage:	12540 mV

Just so.  This bug isn't hurting hardware, just messing with minds in 
this manifestation, while we wonder what other effect it may have?

 > > Interesting that your LEDs display a different view; perhaps BIOS + EC
 > > just monitors voltage and cuts charge, but then what's reporting that
 > > fairly high charge rate?  I'd expect the Embedded Controller to be doing
 > > that .. any dmesg indications of ACPI problems talking to the EC?
 > 
 > Well, that's interesting, it reports 2 batteries:
 > battery0: <ACPI Control Method Battery> on acpi0
 > battery1: <ACPI Control Method Battery> on acpi0
 > 
 > I think there is an option for a second battery, but I'm certainly
 > not using it.

acpiconf -i1 should report it as unavailable.

 > > The 16% is likely from the battery's onboard coulomb counter, but then
 > > so might be the (bogus?) charge rate report.  All speculative, I know ..
 > 
 > The data is all perfectly fine, it was simply from a moment in time
 > a couple hours in the past. It seems the reading is cached and never
 > updated.

Obvious once known :)  Curious to check acpiconf.c diffs 8.2 vs 9, I see 
only this latest change, which if I read it right, was in 8.2 anyway:

http://www.freebsd.org/cgi/cvsweb.cgi/src/usr.sbin/acpi/acpiconf/acpiconf.c.diff?r1=1.27;r2=1.28 

and that shouldn't be pulling cached data, so maybe this issue is deeper 
in the ACPI/EC code?  Well out of my depth, and no time to go diving 
now; anyway, another data point for kern/162859.

 > > All that said, I don't know specifically how HP do things, or what
 > > normal full charge voltage is expected.  Tried another battery?
 > 
 > Tried another OS. Windows shows the battery status just fine.
 > So did RELENG_8. This is a regression in RELENG_9 there isn't a
 > problem with the battery. Or if there is RELENG_8 was able to
 > cope.

So it seems.  When I get full backups done I'll brave bsdinstall again, 
but last time it wasn't happy to reuse my 7.4 slice's partitions without 
external newfs'ing, and I'm still hoping not to nuke its existing /home.

cheers, Ian