Re: V4L control units

[Hans Verkuil <hverkuil@xs4all.nl>]

On 05/09/2014 03:46 PM, Sakari Ailus wrote:
> Hi Hans,
> 
> On Fri, May 09, 2014 at 02:57:13PM +0200, Hans Verkuil wrote:
>> On 05/08/2014 11:04 AM, Sakari Ailus wrote:
>>> Heippa!
>>>
>>> On Wed, May 07, 2014 at 03:57:11PM +0300, Antti Palosaari wrote:
>>>> What is preferred way implement controls that could have some known
>>>> unit or unknown unit? For example for gain controls, I would like to
>>>> offer gain in unit of dB (decibel) and also some unknown driver
>>>> specific unit. Should I two controls, one for each unit?
>>>>
>>>> Like that
>>>>
>>>> V4L2_CID_RF_TUNER_LNA_GAIN_AUTO
>>>> V4L2_CID_RF_TUNER_LNA_GAIN
>>>> V4L2_CID_RF_TUNER_LNA_GAIN_dB
>>>
>>> I suppose that on any single device there would be a single unit to control
>>> a given... control. Some existing controls do document the unit as well but
>>> I don't think that's scalable nor preferrable. This way we'd have many
>>> different controls to control the same thing but just using a different
>>> unit. The auto control is naturally different. Hans did have a patch to add
>>> the unit to queryctrl (in the form of QUERY_EXT_CTRL).
>>
>> Well, that's going to be dropped again. There were too many comments about
>> that during the mini-summit and it was not critical for me.
> 
> Ok. Thanks for the information.
> 
>>> <URL:http://www.spinics.net/lists/linux-media/msg73136.html>
>>>
>>> I wish we can get these in relatively soon.
>>
>> Sakari, I think you will have to push this if you want this done.
> 
> Ack. I think I proposed something like this already a few years ago so I'm
> fine picking it up. :-) Now it's a good time to add the required space in
> the struct as we're going to have a new IOCTL anyway.
> 
>> One interesting thing to look at: the AVB IEEE 1722.1 standard has extensive
>> support for all sorts of units. I don't know if you have access to the standard
>> document, but it might be interesting to look at what they do there.
> 
> I have access to it but I don't see this would be that interesting in
> regards to what we're doing. In any case, we should document the units so
> that different drivers end up using exactly the same string to signal a
> particular unit.

Actually, it is interesting. AVB has controls as well, and those map pretty
much exactly to the sort of controls we have, and they support a wide range
of units. Look at section 7.3.3 ("Control Value Units") of IEEE 1722.1.

If nothing else, it gives a good insight into how others handle this.

Regards,

	Hans

> I prefer to have a prefix as well: a lot of hardware devices use binary
> fractions so that even if we provide an integer control to the user the
> actual control value may well be divided by e.g. 256. That is a somewhat
> separate topic still.
>