Re: Charge complete detection?

Hello Kevin,

Thursday, December 28, 2006, 2:05:11 AM, you wrote:

> Hi,

> I've recently updated my phone to support battery monitoring. The
> only thing left for basic PDA functionality on HTC Apache is battery
> charge support.

> I know how to charge the battery -- I've identified the AC detect
> GPIOs and the charge enable GPIOs. However, I'm not sure when
> charging should be considered complete. I want to avoid over-charging
> batteries, so until I've figured this out I'm planning on disabling
> the charger.

> I'm writing this email to solicit feedback from other phone developers
> on how they handle the charge complete sequence.

> The HTC Apache phone does not appear to have a w1 bus. Instead it
> looks like it uses a simple ADC chip (the ad7877) to monitor battery
> voltage and current.

  h1910, h4000 have the same (except it is ads7846). I also not sure about
current, but I guess it must be sampled somehow too.

[]

> When charging, the normal battery voltage appears to be always around
> 4.2 volts. However, I've seen WinCE do a "trick" where it disables
> the charger for 10ms and then senses the battery voltage again. It
> seems to do this more frequently as it "completes" the charge
> sequence.

  I wonder, if that charge enable/disable GPIO's purpose is *solely*
to be able to sample actual battery voltage. My personal opinion of
necessity to control LiON battery charging in such way is urban
legend. It's well known that LiON cells need sophisticated charging
control, and that every LiON battery has builtin voltage/temperature
sensors and charge controller. And that would be true unless you want
to believe vendors put little bombs in each PDA/phone.

  Anyway, what I can say is that on h4000, there's currently no
"charge control" at all. After booting with HaRET, battery can be
charged while in Linux, but during the charging, voltage is skewed as
you describe. In WinCE, standard applet in control panel neither shows
capacity during charging (it shows running progress bar). However, 3rd
party utils show it, so I guess, there's some way to sample battery
voltage during charge too.

> I've also noticed that even after WinCE reports that the battery is
> fully charged, if the AC is left in, it will occasionally re-enable
> the charger for brief periods (without reporting that it is). As a
> guess, the phone may run off of the battery even when AC is present,
> and the software periodically recharges to account for recent losses.

> Any thoughts? Any references I can research?

  I'd say, if you have reference implementation to look at, just do as
it does. This way, we at least will be correct and won't be worse than
it. When we'll have few correct implementation around, we'll be able
to compare them, and try to optimize/elaborate somehow.

> -Kevin

-- 
Best regards,
 Paul                            mailto:pmiscml_at_gmail.com