battery voltage scale is not correct

[laforge@…]

Since we don't have a coulomb-counting battery gas gauge, we cannot give proper
battery metering.

Our current battery percentage reading is based on battery voltage, and uses a
linear scale between 2.8 and 4.2V (see drivers/i2c/pcf50606.c:battvolt_scale()).

This scaling is obviously wrong and just a 'proof of concept' scale.

Somebody needs to do proper charge/discharge measurements and come up with a
non-linear algorithm expressing estimated battery life based on present battery
terminal voltage.

This needs a lot of experimentation/measurement, as well as a couple of
calculations. Please provide me either with a patch against battvolt_scale(),
or give me a mathematical formula expressing battery lifetime percentage as
function of terminal voltage.

[philipp.zabel@…]

battemp/battvolt measurement during discharge

[philipp.zabel@…]

charge1.txt.gz, charge log

[tony_tu@…]

I will transfer to Tim later about this bug, and see if we could change
battery on next version and give right scale

[alphaone@…]

Have you been able to obtain the correct scale yet?

[tony_tu@…]

We have change the resister value on the PMU battery circuit, should give
better scale compare to Bv3 version, but we did not put all rest Bv4 SMT due
to holiday/quene issue. We will verify the boards rest SMT finished.

[speedevil@…]

I'm doing a simple patch for this, based on a simple curve-fit. Just polishing ATM.

[willie_chen@…]

Did this patch publish in kernel?So we can close this bug.

[andy]

We took a patch for charging curve some months ago, on GTA02 we do have coulomb counter, so closing.