[pulseaudio-discuss] Port icons, +11 dB, ABI stability

[David Henningsson]

2012-05-09 07:45, Colin Guthrie skrev:
> 'Twas brillig, and David Henningsson at 09/05/12 02:36 did gyre and gimble:
>
>> 2) +11dB
>> The people within the discussion - including myself - seem to mostly
>> agree that for the sound settings UI, it makes most sense to remove the
>> possibilty to set more than 0dB output for the device volume, but
>> instead allow it for stream volumes. The argument is that the most
>> common use case for this is that you're listening to DVD/flash movies,
>> where the particular application would be what's bothering you.
>> Therefore it would make more sense to boost the application, rather than
>> the device.
> Interesting idea. However, as you guys do not use flat volumes, how can
> a volume>0dB for a stream ever make sense if the system volume is at
> -30dB?  As it doesn't make sense to push a stream volume>0dB if the
> system volume is anything less than 0dB itself, would you be adopting
> flat volumes to implement this feature?

No, we're not planning to switch to flat volumes. The use case is that 
somebody listens to a video on youtube, and that the audio is not using 
the full signal range (e g, for 16 bit audio, maximum value for a sample 
is 32767, but for this particular sound it never exceeds say 8000). In 
this case it does not hurt to just software amplify the signal.

And tbh, as for digital clipping, the holy grail for many of us is to 
avoid it, but being pragmatic, for a lot of source material (e g 
speech), a little bit of digital clip here and there is not noticeable 
for most people.

> If not then a stream volume of
> +11dB with a sink volume of -30dB would result in -19dB overall for that
> stream which doesn't, to me, seem like the clearest method of presenting
> this feature. Internally I would hope that this could be implemented
> sensible (i.e. setting the hardware to -19dB even if we present the sink
> volume as -30dB and attenuate other streams accordingly etc. etc.), but
> this may make the the number of volume objects we track even more
> confusing (although it may fit the current module anyway?)

It is correct that we could set -19 dB on the hardware in the above 
case, and that would be an improvement, but I don't plan on doing that 
right now. We would software amplify +11 dB on the source stream and 
keep the hardware at -30 dB.

> And if you do adopt flat volumes, as the sink volume is always no less
> than maximum of the stream volumes, how would you present that in the
> GUI? Would it just limit the slider but still allow it to be technically
>> 0dB (much like the media keys on keyboards are handled just now)?
> I've probably complicated this idea for you now :p

I agree that flat volumes *are* complicated and raises mind-boggling 
questions like the above one. Hmm, if upstream standard (in practice, 
Fedora) is to use flat volumes, and we want to upstream the sound 
settings UI (which we do), upstream might decide differently than Ubuntu 
at this specific point.

>
> On a related note, it's maybe also worth thinking about how "0dB" is
> handled on different h/w. e.g. how do we deal with "Base Volume" on
> hardware where "all ALSA sliders at max" != 0dB.
>
> IMO, we should have a per-sink/source persistent setting that allows us
> to opt to use the ALSA 0dB point as our 0dB point (thus effectively
> opting to NOT have a base volume on this device (as if turned on, base
> volume == 0dB == 100%).
>
> Of course in this setup, we'd still want volumes>0dB (boosted) to use
> the alsa controls to do that rather than software amp while not making
> full use of the kcontrols.

What would be the practical benefit of doing this?

// David