Re: [LAD] Floating point processing and high dynamic range audio

Excerpts from lieven moors's message of 2010-07-22 15:20:48 +0200:
> On 07/22/2010 05:31 AM, Philipp Überbacher wrote:
> > Excerpts from Philipp Überbacher's message of 2010-07-22 03:16:00 +0200:
> >
> > > Excerpts from fons's message of 2010-07-22 02:24:04 +0200:
> > > > On Thu, Jul 22, 2010 at 01:05:01AM +0200, Philipp Überbacher wrote:
> > > >
> > > > > I think the word loudness is a problem here. Afaik it usually
> > refers to
> > > > > how it is perceived, and twice the amplitude doesn't mean twice the
> > > > > perceived loudness. It may mean twice the sound pressure level,
> > energy,
> > > > > or intensity (if we ignore analogue anomalies, as you wrote in
> > some other
> > > > > answer).
> > > >
> > > > Subjective loudness is a very complex thing, depending on the
> > > > spectrum, duration, and other aspects of the sound, and also
> > > > on circumstances not related to the sound itself.
> > > >
> > > > For mid frequencies and a duraion of one second, the average
> > > > subjective impression of 'twice as loud' seems to correspond
> > > > to an SPL difference of around +10 dB.
> > >
> > > I had a brief look at the section about loudness in musimathics and it
> > > mentions 10 dB based on the work of Stevens, S.S. 1956,
> > > "Calculation of the Loudness of Complex Noise" and 6 dB based on
> > > Warren, R. M. 1970,
> > > "Elimination of Biases in Loudness Judgments for Tones.".
> > > I think I've encountered the 6 dB more often in texts, which doesn't
> > > mean it's closer to the truth, if that's possible at all.
> > > Knowing a 'correct' number would be nice for artists and sound
> > > engineers, but if it varies wildly from person to person, as Gareth Loy
> > > suggests (no idea where he bases this on) then this simply isn't
> > > possible. Picking any number within or around this range is probably as
> > > good as any other.
> > >
> > > > I often wondered what criterion we use to determine which
> > > > objective SPL difference sounds as 'twice as loud'. We don't
> > > > have any conscious numerical value (there may be unconscious
> > > > ones such as the amount of auditory nerve pulses, or the amount
> > > > of neural activity), so what it this impression based on ?
> > > >
> > > > The only thing I could imagine is some link with the subjective
> > > > impression of a variable number of identical sources. For example
> > > > two people talking could be considered to be 'twice as loud' as
> > > > one. But that is not the case, the results don't fit at all (it
> > > > would mean 3 dB instead of 10).
> > >
> > > I never thought about that to be honest. It's immensely complex. It
> > > might have to do with each persons hearing capabilities, for example
> > the
> > > bandwidth of loudness perception or the smallest discernible loudness
> > > difference. If it really is very different from person to person, then
> > > an explanation that takes the different hearing capabilities into
> > > account could be sensible, don't you think?
> >
> > I did find some more approaches to the problem, but those are just
> > ideas. From my personal experience I have to say that I have a very hard
> > time saying when something is twice as loud. A musically well trained
> > person might have an easier time, I wouldn't know, but for me twice as
> > loud is something that is very vague. This might already explain the
> > large deviation between subjects as described in musimathics. It lead me
> > to another idea though, the evolutionary perspective. Evolutionary it
> > likely never was important whether a sound is twice as loud. The only
> > situation I can imagine where judging loudness probably was important
> > is judging distances. How far is the animal I can't see, be it prey or
> > predator, away from me? We know that this takes more than the SPL into
> > account, and 'twice as loud' doesn't have relevance in this context. So
> > maybe the loudness perception is linked with spatialization.
>
> I think this is a very interesting idea. Could this be linked to some
> kind of
> avarage SPL of all the sounds human beings are exposed to (and this variable
> changes throughout history). Because when we try to judge the distance of
> a barking dog, our brain would use the knowledge of all other dogs we heard
> barking before, to estimate the distance of that dog. If we never heard
> a dog
> before, maybe we would use the sounds of other animals as a reference,
> and so on...
>
> greetings,
>
> Lieven

Hi Lieven,
again, I don't know, I can only deduct from experience and reasoning,
but I think it could work in a similar way. We can discern lots of
different sounds, even sounds that are very similar. For that we need
some experience with that particular sound or source, but I think we can
learn to recognise sounds very fast. Sounds have multiple properties
that allow us to keep them apart, a nice example might be again my
lack of experience. In many cases I might not be able to keep a violin
and a cello apart. I might be able to tell you that it's a stringed bow
instrument, but I lack the experience to tell you whether it's the one
or the other. Multiple properties might allow this though. Knowledge of
the pitch range of each instrument for instance might allow to tell that
the low pitched sound is a cello. There are other, finer differences,
probably called timbre, that might allow this at any pitch. In my case
it's even worse with brass instruments, I could maybe tell you it's
brass, maybe tell you when it's a sax, and that's it. We know similarity
but need experience.
An example for the learning ability of fine detail are people who work
with perceptual coding and frequently do listening tests between
different codecs like mp3, vorbis and others. With careful listening and
time you learn to hear typical artifacts of those codecs that are
specifically designed to fool the ear. Someone well trained might even
be able to tell which codec was used by the artifacts alone.
Of course this only works as long as the ears are better than the codec.

Anyway, among the many properties of sounds that help us identify their
sources might be typical loudness at a distance. It might well be that
we try to identify a sound source before we try to obtain its position,
because the distance information might be more accurate this way, and
because the sound might not be relevant to us at all.

To get closer to the original question again, maybe all this
identification and spatial mechanism plays a role when we try to judge
loudness and loudness relationships. If so, then it's likely very hard
to come up with a formula that describes it.

-- 
Regards,
Philipp
--
"Wir stehen selbst enttäuscht und sehn betroffen / Den Vorhang zu und alle Fragen offen." Bertolt Brecht, Der gute Mensch von Sezuan
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev@lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-dev