Re: [PATCH 6/7] dt-bindings: iio: adc: add adi,ad7606c-{16,18} compatible strings

[Krzysztof Kozlowski <krzk@kernel.org>]

On 21/08/2024 22:26, Jonathan Cameron wrote:
> 
>>>> +    type: object
>>>> +    $ref: adc.yaml
>>>> +    unevaluatedProperties: false
>>>> +
>>>> +    properties:
>>>> +      reg:
>>>> +        description: The channel number.
>>>> +        minimum: 0
>>>> +        maximum: 7
>>>> +
>>>> +      diff-channel:
>>>> +        description: Channel is bipolar differential.  
>>>
>>> There is diff-channels property, why do we need one more?  
>>
>> Yeah, I wanted to use that.
>> Maybe I will try another spin at that.
>> The thing with "diff-channels" is that it requires 2 ints.
>> So,  diff-channels = <0 0>.
>> To use it here, a rule would need to be put in place where  "reg ==
>> diff-channels[0] && reg == diff-channels[1]".
>> That also works from my side.
>> Part of the reason for this patchset, was to also get some feedback
>> (if this is the correct direction).
>>
> So I 'think' this is a datasheet matching thing.
> In many cases, even for strictly differential devices, the pin
> naming allows for a clear A - B channel description. Here
> in the non differential modes, the negative pins are effectively
> not used (from a really quick look at the datasheet)
> 
> So we 'could' introduce magic channels (give them high numbers) for
> the negative ends. I think we may want to do that for the
> userspace ABI (0-0 on the few times it has come up has been a
> calibration / self check mode not what you have here - it
> wires the actual inputs together).  Alternative is just present
> them as a simple voltage and don't worry about the differential aspect
> as it's not hugely different to bipolar (where the zero level is
> effectively the negative input of a differential ADC.
> 
> For the binding I'm fine with the binding using A, A as you suggest
> with an update to adc.yaml to cover this corner.

Yep, let's add it to adc.yaml.

> 
> We never (I think) have bindings for the self check case where the input
> is wired to both sides. It's just a mode that is applied to
> any inputs that are wired. 
> 
Best regards,
Krzysztof