Re: [PATCH 2/2] dt-bindings: iio: adc: document TS voltage in AXP PMICs

[Evgeny Boger <boger@wirenboard.com>]

22.11.2021 14:35, Samuel Holland пишет:
> On 11/22/21 5:17 AM, Evgeny Boger wrote:
>> 22.11.2021 13:49, Maxime Ripard пишет:
>>> On Thu, Nov 18, 2021 at 05:12:33PM +0300, Evgeny Boger wrote:
>>>> Most AXPxxx-based reference designs place a 10k NTC thermistor on a
>>>> TS pin. axp20x IIO driver now report the voltage of this pin via
>>>> additional IIO channel. Add new "ts_v" channel to the channel
>>>> description.
>>>>
>>>> Signed-off-by: Evgeny Boger <boger@wirenboard.com>
>>> Would it make sense to put the resistance in the DT as well or is it
>>> made mandatory by Allwinner?
>>>
>>> Maxime
>> Well, I don't think so. Basically, by default AXP20x injects 80uA
>> current into the TS pin and measure the voltage. Then, there are
>> voltage thresholds to stop charging if the battery is too hot or too
>> cold. The default thresholds were calculated by the manufacturer for
>> default 10k resistance and 80uA current. Finally, if TS pin is
>> shorted to GND, the AXP2xx will detect it and won't shut down
>> charging. Note that AXP2xx doesn't convert the measured voltage to
>> temperature.
> Agreed, since the ADC driver only works with voltages, the resistance is
> not relevant to it, so a resistance property does not belong here.
>
>> So while it's possible to use AXP2xx with resistance other than 10k,
>> it will require us to override these protection thresholds.
>> Moreover, if one want to put the actual resistance in DT, then the
>> driver would need to calculate these protection thresholds based on
>> NTC parameters and injection current.
> That means we do need a resistance property for the battery charger
> driver, because it does need to calculate temperature.
Right now the charger driver just doesn't touch the default voltage 
thresholds at all.
> Regardless of the reference design, the resistance is variable in
> practice. At least some early v1.0 PinePhones shipped with batteries
> containing a 3 kOhm NTC. And the battery is removable, with an
> off-the-shelf form factor, so users could install aftermarket batteries
> with any NTC resistance.
I think we can easily expose *voltage* thresholds as DT properties.

>
> Right now, people with these batteries are disabling the TS; otherwise
> the PMIC refuses to charge them. It would be good to re-enable the TS by
> coming up with the proper voltages for the min/max thresholds. And there
> are power supply properties we can use to expose the current temperature
> and those thresholds to userspace (at least as read-only).
So my idea was to convert voltage to temperature elsewhere. Again, I 
personally
think that it makes sense, because the hardware itself (including 
charger) doesn't
deal with temperature at all, only with threshold voltages and injection 
current.

For instance, in  our board we now use hwmon NTC driver, like this:

     /* Huge pullup voltage is here to emulate constant current */
     bat-temp {
         compatible = "murata,ncp15xh103";
         pullup-uv = <1000000000>; // 1E9 uV
         pullup-ohm = <12500000>; // pullup_uv/80
         pulldown-ohm = <0>;
         io-channels = <&axp_adc 4>;
     };

As far as I know, there are IIO AFE patch set under review, which also add
voltage to temperature conversion.

If we indeed want the axp20x charger driver to convert voltage to 
temperature
(and vice versa, for computing charging thresholds), then all that logic 
will need to
be reimplemented in driver. Note that there wouldn't be a single 
"resistance" property,
instead there are huge resistance vs voltage tables which are slightly 
different for
different vendors.

In principle, having temperature reported by power supply driver would 
be great. But
I don't see how it can be implemented without a massive effort.


>
> Regards,
> Samuel


-- 
С уважением,
     Евгений Богер / Evgeny Boger
     CTO, Wiren Board Team
     https://wirenboard.com/ru
     +7 495 150 66 19 (# 33)