Re: generic-adc-battery volatge-to-percent formula

From: "H. Nikolaus Schaller" <hns@goldelico.com>
To: Pavel Machek <pavel@ucw.cz>
Cc: Sebastian Reichel <sre@kernel.org>,
	pali.rohar@gmail.com, kernel list <linux-kernel@vger.kernel.org>,
	ivo.g.dimitrov.75@gmail.com, patrikbachan@gmail.com,
	serge@hallyn.com, abcloriens@gmail.com,
	Belisko Marek <marek.belisko@gmail.com>,
	Linux PM mailing list <linux-pm@vger.kernel.org>
Subject: Re: generic-adc-battery volatge-to-percent formula
Date: Mon, 24 Oct 2016 15:09:26 +0200	[thread overview]
Message-ID: <3F9199B6-233E-4260-9362-498FB1AA3FB7@goldelico.com> (raw)
In-Reply-To: <20161024123404.GA24185@amd>

[-- Attachment #1: Type: text/plain, Size: 1941 bytes --]

Hi Pavel,

> Am 24.10.2016 um 14:34 schrieb Pavel Machek <pavel@ucw.cz>:
> 
> Hi!
> 
>> On Mon, Feb 22, 2016 at 06:58:13AM +0100, H. Nikolaus Schaller wrote:
>>>>> static inline int fuel_level_LiIon(int mV, int mA, int mOhm) {
>>>>> ...
>>>>> }
>>> 
>>> To which header file should this go?
>> 
>> I think it should get its own header file in include/linux/power/.
>> Maybe something like "generic-fuel-gauge.h". I'm open to other
>> solutions, though.
> 
> I'd like to use use this formula in my own code. Is it somewhere in
> the kernel already?

Not that I am aware of. Marek is still working on the generic-adc-battery
driver augmented by DT + iio-ADC + formula before publication.

Here is the latest draft so you can cherry-pick it into your work:

	http://git.goldelico.com/?p=gta04-kernel.git;a=patch;h=22ab047ae296e998379c1aa29fe1210043cfa040

But beware: I think it is quite wrong using the sqrt() function above 19.66%
and doing linear interpolation below.

By using the sqrt() function it has a steepness that goes to infinity when reaching
19.66% from above. This makes a quite non-realistic curve with a sharp bend at 19.66%
(which is equivalent to 3.756 V).

No real battery I have seen and measured with a coulomb counter has such a strange
bend at 3.756 V...

IMHO it would be a better approximation to adjust the factors so that e.g. a realistic
voltage for "empty" (e.g. 3.3 V) is taken as the 0% point where steepness goes through
the roof.

It should then be something like:

	SOC = sqrt((Volt - 3.3 V) / (4.2 V - 3.3 V))

This goes more smoothly between 100% and 0%.

Or we could even use an exponential function with an exponent different from 1/2
(1 being linear interpolation):

	SOC = pow((Volt - 3.3 V) / (4.2 V - 3.3 V), factor)

But this would probably need floating point arithmetic in the kernel or some numeric
approximation algorithm.

BR,
Nikolaus

[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 801 bytes --]