Re: Reading ADC that comes from a multiplexer

[Jonathan Cameron <Jonathan.Cameron@Huawei.com>]

On Wed, 22 Sep 2021 11:50:14 -0300
Fabio Estevam <festevam@gmail.com> wrote:

> Hi Peter,
> 
> On Wed, Sep 22, 2021 at 11:28 AM Peter Rosin <peda@axentia.se> wrote:
> 
> > Nice!
> >
> > While I don't completely understand that iio-device node in the beaglebone
> > dts that didn't work for you, it looks like it's just a renumbering thing?  
> 
> The beaglebone dts uses some undocumented properties such as:
> iio-channels and iio-channel-names.

Some of this comes from the dts-schema repo.  We haven't been strict in
adding the entries to individual ADCs until they actually use them - which
has the advantage it gives us a window to think about the of_xlate (see below)

https://github.com/devicetree-org/dt-schema/blob/main/meta-schemas/iio.yaml
Not that it helps much as little in the way of docs in the dt-schema repo.

> 
> > However, your version only remapped 4 channels, and in that case your new
> > iio-device only had those, i.e. 0-3. But the iio-mux was looking for the
> > missing channel 4. Maybe that was why that variant didn't work?  
> 
> Yes, this is where I got confused.
> 
> The stmpe811 has 8 channels. On the apalis board, the first four channels
> (0 to 3) are used for touchscreen. The other 4 channels are for general purpose.
> 
> The ADC that is connected to the MUX is channel 4 (which is the first
> one that is
> free for general usage), so I had to pass:
> 
> io-channels = <&adc0 0>;
> 
> in the mux, instead of  io-channels = <&adc0 4> that I was originally trying.
> 
> and now the mapping is correct and I can read proper voltages when I
> switch the mux.

It's possible to add a translation routine to a given driver to deal with this
sort of case. I guess no one needed on the that driver before + all this
infrastructure post dates that driver.

See the of_xlate callbacks that let you map more obvious numbering to a particular
channel.

We are in an unfortunate mess here, but I'd argue the lack of io-channels entry
in the dt binding should in theory mean no one is using this property (as they
should be verifying against that).  The problem will occur if we have a pre
yaml conversion binding out in the wild with a mux or other consumer.  We could
cross our fingers and fix this now...

Jonathan


> 
> Thanks!