[alsa-devel] [RFC PATCH] ucm: Add support for device positions

[Liam Girdwood]

On Tue, 2017-01-03 at 08:50 -0600, Pierre-Louis Bossart wrote:
> 
> On 01/03/2017 04:09 AM, Mengdong Lin wrote:
> >
> >
> > On 01/03/2017 04:21 PM, Vinod Koul wrote:
> >> On Tue, Jan 03, 2017 at 02:09:18PM +0800, 
> >> mengdong.lin at linux.intel.com wrote:
> >>> From: Mengdong Lin <mengdong.lin at linux.intel.com>
> >>>
> >>> Users can provide prosition info of audio devices as a device value. 
> >>> This
> >>> will help the sound server to choose an audio devices from some 
> >>> candidates
> >>> based on the their locations and the status of the machine.
> >>
> >> This is good thing to have but some bits are bit unclear to me atm. 
> >> When you
> >> say device, do you mean sound card?
> >
> > No, the device here does not mean a sound card, but mean an I/O device 
> > like speaker or dmic, defined by 'SectionDevice'.
> >
> >> For me, position makes sense for dmics and speakers. So can you help 
> >> me out
> >> by understanding how we can describe these...
> >>
> >
> > For example, if there are two DMICs, one is in the front and one on 
> > the back, we can define their positions as below:
> >
> > SectionDevice."MainMIC" {
> >
> >        ...
> >
> >         EnableSequence [
> >                ...
> >         ]
> >
> >         DisableSequence [
> >                 ...
> >         ]
> >
> >         Value {
> >                 Position "Front"
> >         }
> > }
> >
> > SectionDevice."SecondaryMIC" {
> >
> >     ...
> >
> >     EnableSequence [
> >     ...
> >     ]
> >
> >     DisableSequence [
> >     ...
> >     ]
> >
> >     Value {
> >         Position "Back"
> >     }
> > }
> 
> What is the entity that will make use of the position information? For 
> what purpose? This type of information needs to be known at a lower 
> level (typically a DSP), I wonder what UCM would do with it?
> Also it can be tricky to define left and right for a tablet which works 
> in all possible orientations, you'd need sensor information to refine 
> the position. And last if you have a mic or speaker array you'd need to 
> provide geometrical information?

This should actually be split into "location" and "position" (maybe
better descriptors are available).

"location", mainly used by sound server to represent the device to the
user (e.g. on a GUI) and handle changes to laptop physical configuration
(device tilted, screen folded, etc)

"position", used by DSP for beamforming etc. Coords can be processed in
userspace for any physical changes prior to being fed into DSP.

Liam