By "new battery" colin was talking about turning N900 off, replacing almost empty battery with freshly-charged one (charged externally, not in N900). He wasn't talking about new or old battery (age) AFAIK.

I understand that it's upstream module Pali and that You can't do much about it. Again, AFAIK as i know it work pretty well in notebooks etc, but for some reason doesn't work on our N900 as it should - maybe joerg_rw got more info why.

I reported it, cause - as documentation that you quoted said - charge_now and charge_full DOESN'T use charge_design to calculate value, but pull it from last know full and empty battery states, calculating uAh charged/discharged by monitoring current in given time. Technically, it works like advanced RC battery charger - with difference that RC battery charger charge/discharge with stable, given current (can change it when charging/discharging is almost done, to lower value) and measure time, with adjusting calculation when processor lower changing current for safe charge-finish.

Our N900 use different current all the time, depending on usage of device - chip act (via bq module) as super digital charger, adjusting calculations all the time, to get real uAh value. After few full charge/discharge cycles done internally on N900 (via USB charging) these values (charge_full and charge_now) should be quite correct - little lower than real battery capabilities, cause N900 discharge it only to 3,3 V.

Ho ever, in current approach, bq module reports much higher values than battery really posses (even after many cycles of charging/discharging), and i have no idea why.

Also, current_now value seems quite higher than it's really. If I got battery that i know have 1100 mAh, active usage of 200 000 uA - 300 000 uA (200-300 mAh) should drain battery in 4-5 hours. Ho ever, battery last much longer in tests. Also, as i mentioned reported values for "charge_full" and "charge_now" are higher that it should be. Mixing it with wrong current_now values, internal calculations are correct, except that higher than it should be. Higher current_now readings result in higher charge_full and charge_now values.

Maybe that 22/3570 uA value (instead of 1 uA) in kp46 was there on purpose? I wonder if current, patched bq module, just adjusted to operate at "wrong" 22/3570 uA values could provide real charge_full and charge_now data... Maybe data pulled out from chip is wrong, and that "incorrect" value put in bq was there in fact to correct it, to adjust calculations by error factor?

I would LOVE to have working charge_full and charge_now values (of course, also charge_empty), because that would allow everyone with kp to quite precisely measure real battery capacity in mAh (after calculating it from uAh), so everyone could contribute to threads like "battery test", and check if super-uber advertised battery really have it advertised capacity.