Re: libbattery was Re: [RFC PATCH 5/5] power: generic-adc-battery: Add capacity handling

[H. Nikolaus Schaller]

Hi Tony,

> Am 19.10.2017 um 18:24 schrieb Tony Lindgren <tony@xxxxxxxxxxx>:
> 
> * H. Nikolaus Schaller <hns@xxxxxxxxxxxxx> [171018 08:49]:
>> 
>>> Am 18.10.2017 um 15:22 schrieb Tony Lindgren <tony@xxxxxxxxxxx>:
>>> 
>>> * H. Nikolaus Schaller <hns@xxxxxxxxxxxxx> [171018 05:49]:
>>>>> Am 18.10.2017 um 14:28 schrieb Pavel Machek <pavel@xxxxxx>:
>>>>> 
>>>>> So I started something, it is at.
>>>>> 
>>>>> https://github.com/pavelmachek/libbattery
>>>>> 
>>>>> My battery on n900 is currently uncalibrated (and charging), still it
>>>>> gets some kind of estimation:
>>>>> 
>>>>> Battery -1 %
>>>>> Seconds -1
>>>>> State 1
>>>>> Voltage 3.88 V
>>>>> Battery 63 %
>>>>> 
>>>>> Of course, there's a lot more work to be done.
>>>> 
>>>> Nice start but not a solution to our problem.
>>>> 
>>>> Our problem is that people simply expect that for example https://packages.debian.org/wheezy/xfce/xfce4-battery-plugin
>>>> displays the battery percentage.
>>> 
>>> I think we could make things compatible with various battery apps by
>>> having libbattery write back the capacity percentage and time remaining
>>> to the kernel driver via sysfs or a dev entry. Then the kernel interface
>>> can just display the data to whatever apps.
>> 
>> Hm. That would be quite difficult to understand and maintain code.
> 
> How so? The libbattery can do it all, then the kernel drivers needing
> that will just display the most recent values to maintain compability
> with battery apps.

Well, it looks as if you are thinking about a much more complex solution
than I am...

The proposal we are discussing as

[RFC PATCH 5/5] power: generic-adc-battery: Add capacity handling

just reads the current voltage, charging current and uses a (nonlinear)
mapping to 0% .. 100%.

For that we don't need a library, and we don't need a mechanism to

"write back the capacity percentage and time remaining
to the kernel driver via sysfs or a dev entry.

> 
>> Why not have the kernel driver do the simple calculations (they do
>> not need float) and provide the standard /sys/class/power attribute?
> 
> Because the current remaining capacity and battery empty state depend
> on maintaining a database of previous history for battery wear.

For remaining capacity I agree that you need that. But this RFC doesn't
even want to provide remaining capacity. To avoid such issues. Just
provide POWER_SUPPLY_PROP_CAPACITY:

https://lkml.org/lkml/2017/8/1/910

So we are discussing to replace these 10 lines with a library, a
daemon and a mechanism that the driver can read that from user space?

IMHO this can be added in a further step, but shouldn't be backpacked
onto this here.

> This
> data needs to be preserved across reboots, so most likely on a file
> on a disk is the way to go.

It could also be stored in some i2c eeprom (DT could provide a reference).

> 
> There's a nice summary what all is involved here:
> 
> http://www.mpoweruk.com/soc.htm

Oh yes, that is really nice!

Using their terms, we just want to add "Voltage Based SOC Estimation"
for typical "Lithium chemistry".

BTW: this already exists in the twl4030-madc-battery driver and we use it
for the GTA04 derivatives without fuel gauge:

http://elixir.free-electrons.com/linux/v3.17/source/drivers/power/twl4030_madc_battery.c

It uses a table driven approach which was rejected to be converted to DT
some time ago. DT maintainers asked to use an estimation formula instead
of a table and interpolation.

And this patch set does not want to solve all problems in one step :)

But it wants to:
* make it useable for any battery voltage providing iio adc and not only twl4030
  by modifying the generic-adc-battery driver
* make the driver device tree compatible (twl4030-madc-battery still isn't)
* replace table by fixed point estimation formula fuel_level_LiIon(() as requested

Not more.

Of course a libbattery can do a lot of additional calculations to predict
e.g. time to emtpy or full charge, estimate battery health etc. and provide
more precision by filters.

We would be very happy to have a simple estimation because it already
covers most use cases (some early smartphones had only 4 levels for SOC).
And we want to replace the pdata based driver with a DT driver asap.

BR and thanks,
Nikolaus