[Gta04-owner] Qt upgrade eating battery?

[Neil Jerram]

NeilBrown <neilb at suse.de> writes:

> Those numbers look like numbers for when the CPU is awake, but the kernel
> logs clearly show the the device is going to sleep. e.g.
>
> [12233.411468] PM: late suspend of devices complete after 1.983 msecs
> [12981.035766] Successfully put all powerdomains to target state
>
> is good to see. 748 seconds of real suspend.
>
> So either 
>  1/ the numbers are misleading
>  2/ something other than the CPU is sucking power.

I've since discovered a bit more, I think.

Forgetting, for now, the current in suspend, and concentrating on
current_now when the phone is connected by USB to my laptop, and so
charging...

- Originally (i.e. in a similar state as when I reported the suspend
  current numbers above) I was seeing -120mA.

- Then I decided to fix my bluetooth UART speed, as QtMoko v46 seems to
  assume that.  After that (and rebooting), I was seeing -170mA.

- Then I went and played with all the QtMoko scripts for powering things
  off.  Unfortunately in a rather unscientific way - but after that I
  was seeing between -350mA and -390mA.

Since, I believe, my USB supplies about 500mA, those numbers indicate
consumption respectively of 380mA, 330mA and 150-110mA.

Then, if I let the phone suspend when in the last state, leave it for a
while, and then wake, I see a suspend current around 30-40mA as others
have reported

Then, leaving the phone plugged into my laptop, the steady state
current_now is about -300mA, => 200mA consumption.

In summary, it looks like I can now get down to a reasonable suspend
current, but I need to experiment a bit more carefully to discover
exactly what the relevant steps are.

> 1:  I'm guessing that the number here is an 'instantaneous current' read from
> the battery and it is hoped that it is an old enough value to represent
> current usage during suspend.  Maybe it is but I don't trust it.
> If you really want to know how much current was used during suspend, you need
> to sample the 'charge_now' value and the time before entering suspend, sample
> them, again when waking up, then calculate
>   (first_number - last_number) * 3600 / (last_timestamp - first_timestamp)

Thanks.  I think perhaps that the suspend current number _is_ mostly
trustworthy, but I'm also interested in recording the charge history
against time, and plotting that.

> 2: there isn't much that can draw 200 mA.  Maybe the wifi, but I'm pretty
> sure that gets turned off at suspend.  Does the device feel warm at all
> before you wake up.  I'm pretty sure that if it is really drawing  200mA
> you'll notice the difference in temperature - particularly if it is in a
> pocket where the heat cannot escape easily.

Yes, it definitely did always feel warmer than it should do when
suspended.

Regards,
        Neil