Re: [PATCH 6/8] dt-bindings: iio: adc: ad7476: Add ROHM bd79105

[Matti Vaittinen <mazziesaccount@gmail.com>]

On 06/08/2025 18:15, David Lechner wrote:
> On 8/6/25 2:04 AM, Matti Vaittinen wrote:
>> The ROHM BD79105 is a simple, 16-bit, 1-channel ADC with a 'CONVSTART'
>> pin used to start the ADC conversion. Other than the 'CONVSTART', there
>> are 3 supply pins (one used as a reference), analog inputs, ground and
>> communication pins. It's worth noting that the pin somewhat confusingly
>> labeled as 'DIN', is a pin which should be used as a chip-select. The IC
>> does not have any writable registers.
>>
>> The device is designed so that the output pin can, in addition to
>> outputting the data, be used as a 'data-ready'-IRQ. This, however, would
>> require the IRQ to be masked from host side for the duration of the data
>> reads - and it wouldn't also work when the SPI is shared. (As access to
>> the other SPI devices would cause data line changes to be detected as
>> IRQs - and the BD79105 provides no means to detect if it has generated
>> an IRQ).
>>
>> Hence the device-tree does not contain any IRQ properties.
> 
> There are lots of other ADC chips that have a ready signal like this
> and we've made them work.

Ah. I had no idea. Thanks for the insight!

> Since devicetree bindings should be as
> complete as possible even if the driver doesn't use all of the
> features, I think we should be including the interrupt in the binding.

After what you wrote above, I do agree. There may be systems where the 
IRQ is usable, so dt should have it even if the Linux driver never used it.

> We have also found that some interrupt controllers won't work
> as you have suggested and in that case we also needed a ready-gpios
> to be able to read the state of the pin.

Oh. My thinking was just hard-coding the conversion-time delay, but this 
can indeed make sense - especially if there are other examples :)

Thanks a lot for the insight!

Yours,
	-- Matti