Terje Mathisen <terje.mathisen@tmsw.no> writes:

> Thomas Koenig wrote:
>
>> Terje Mathisen <terje.mathisen@tmsw.no> schrieb:
>>
>>> This is where my signal processing background steps in and says: Start
>>> with the 24 bits/sample, 48 KHz master, and compress that with aac/ogg
>>> down to something a bit less than CD bitrates, maybe ~50% which is what
>>> you typically get from the lossless CD compression formats.(FLAC?)
>>
>> What about the properties of the low-pass filter you will need before
>> digitizing? Aliasing higher frequencies into the audible range can
>> really spoil hearing experience...
>
> By definition, the master digital recording have already applied that
> lowpass filter, right?
>
> Back in the analog days we had the same absolute limits on what the
> recording microphones could physically react to, as long as you sample
> that signal at more than 2X the highest frequency you will be OK.
>
> If your master is using 24 bits/sample @ 96 kHz then you will surely
> be well above any need for phase-changing lowpass filters, [...]

Not so. Human ears are sensitive to phase information content
up at least 200 kHz or so, as anyone with a background in
high school physics, geometry and trig can easily verify for
themselves with a simple thought experiment. Imagine yourself
in a completely dark environment, mostly quiet. Suddenly there
is a sharp sound. Point to where you think the sound is. What
is the angular resolution of where the sound is coming from?
Translate that angular resolution into a delay in microseconds.

Incidentally the time-delay resolution of ears has been directly
measured at somewhere near 10 microseconds.

Tim Rentsch wrote:
> Terje Mathisen <terje.mathisen@tmsw.no> writes:
>
>> Thomas Koenig wrote:
>>
>>> Terje Mathisen <terje.mathisen@tmsw.no> schrieb:
>>>
>>>> This is where my signal processing background steps in and says: Start
>>>> with the 24 bits/sample, 48 KHz master, and compress that with aac/ogg
>>>> down to something a bit less than CD bitrates, maybe ~50% which is what
>>>> you typically get from the lossless CD compression formats.(FLAC?)
>>>
>>> What about the properties of the low-pass filter you will need before
>>> digitizing? Aliasing higher frequencies into the audible range can
>>> really spoil hearing experience...
>>
>> By definition, the master digital recording have already applied that
>> lowpass filter, right?
>>
>> Back in the analog days we had the same absolute limits on what the
>> recording microphones could physically react to, as long as you sample
>> that signal at more than 2X the highest frequency you will be OK.
>>
>> If your master is using 24 bits/sample @ 96 kHz then you will surely
>> be well above any need for phase-changing lowpass filters, [...]
>
> Not so. Human ears are sensitive to phase information content
> up at least 200 kHz or so, as anyone with a background in
> high school physics, geometry and trig can easily verify for
> themselves with a simple thought experiment. Imagine yourself
> in a completely dark environment, mostly quiet. Suddenly there
> is a sharp sound. Point to where you think the sound is. What
> is the angular resolution of where the sound is coming from?
> Translate that angular resolution into a delay in microseconds.
>
> Incidentally the time-delay resolution of ears has been directly
> measured at somewhere near 10 microseconds.
>
This is the part I really do believe: There should be no way for human
ears to determine front/back direction, but we are obviously able to do so!

It has been shown that if you artificially smoothen the ear canal
(making it a cylinder), that capability goes away, so something in our
ears are able to detect really tiny phase modifications.

Anyway, what's really amazing is that our brains have more volume
dedicated to sound than to vision!

Terje

--
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"

On 21/07/2023 13:25, Terje Mathisen wrote:
> Tim Rentsch wrote:

>> Not so.  Human ears are sensitive to phase information content
>> up at least 200 kHz or so, as anyone with a background in
>> high school physics, geometry and trig can easily verify for
>> themselves with a simple thought experiment.  Imagine yourself
>> in a completely dark environment, mostly quiet.  Suddenly there
>> is a sharp sound.  Point to where you think the sound is.  What
>> is the angular resolution of where the sound is coming from?
>> Translate that angular resolution into a delay in microseconds.
>>
>> Incidentally the time-delay resolution of ears has been directly
>> measured at somewhere near 10 microseconds.
>>
> This is the part I really do believe: There should be no way for human
> ears to determine front/back direction, but we are obviously able to do so!
>
> It has been shown that if you artificially smoothen the ear canal
> (making it a cylinder), that capability goes away, so something in our
> ears are able to detect really tiny phase modifications.

The human hearing system (and that of other animals, especially mammals)
is extremely complicated. It is not just two microphones, and position
determination is done based on a large number of factors.

The phase information is not really that a big influence, and the
signals are /not/ timed anything like that accurately. (Phase
information /is/ used, but it is not a big part of the process.) There
are audio illusions where pairs of sounds are recognized by the brain in
different orders from how they are played - timing resolution is of the
order of tens of milliseconds at best. Your brain will fill in missing
sound of up to 50 milliseconds without noticing - what you think you
hear is reconstructed by your brain, just as is the case for your
vision. Phase /difference/ between your ears is much more accurate, but
that is very different from phase /resolution/.

You can test the influence of phase by inverting the loudspeaker wires
on one of your speakers. You might be able to notice a difference, but
not much, and you won't notice a difference in position of the sounds
despite a half wave-length change in the phase difference.

You can also see that Tim's numbers are completely unrealistic. In 10
microseconds, sound travels about 3.5 mm. So moving your head two
millimetres while listening to a sound would have the same effect, yet
it will do absolutely nothing to your perception of the source of the sound.

Or listen to a sound from far away - you can still determine the source
(roughly) even though the phase difference is utterly negligible in
comparison to the distance, and varying air movements, echoes, and other
effects totally swamp any kind of phase information.

Two bigger influences in determining position are relative volume and
phase difference, and the way the sound is affected by your head and
outer ears. Sound passes through bone faster than through air, but is
distorted in odd ways by the shape of your skull. Your ears and your
brain go through years of fine-tuning themselves to use this kind of
information for positioning. Just as changes to your ear canal mess
with your ability to locate sound sources accurately, so do things like
losing some or all of an outer ear, or part of your skull (such as from
injury or brain surgery). You can re-learn afterwards.

The biggest effect in real life (as distinct from very artificial
experiments) is post-deduction rationalisation - you are convinced you
know the direction by other factors (something you see, or otherwise
know), and you then know that's the direction you are hearing it from.

It's worth noting that people who have only one working ear still have
some directional hearing. It's not nearly as good as if you have two
ears, but it still works. (Compare this to 3D vision - with only one
eye, you have virtually as good 3D perception as you had with two eyes.)

>
> Anyway, what's really amazing is that our brains have more volume
> dedicated to sound than to vision!
>

That /would/ be amazing if it were true. About 25-30% of the brain is
for visual processing, and about 8% of it is for audio processing. The
figures are inevitably quite rough, because there's a lot of other
things highly interconnected, such as speech (and therefore language)
and your brain's models of the world around us.

What /I/ think is amazing is that there are more nerves that carry
information from the brain to the ears, than there are carrying
information from the ears to the brain. These are especially important
in the first few years of life (particularly the first six months), as
the brain and ears work together to tune your ears to detecting and
interpreting speech. This process lets you understand people talking
about the chatter in a crowded room, or other noise that is higher
volume than the person you are listening to. It also explains why adult
language learners have trouble learning new phonemes that are not in the
language(s) they heard as a small child - their ears are not trained to
recognize them.

On 7/21/2023 6:44 AM, David Brown wrote:

snip

> It's worth noting that people who have only one working ear still have
> some directional hearing.  It's not nearly as good as if you have two
> ears, but it still works.  (Compare this to 3D vision - with only one
> eye, you have virtually as good 3D perception as you had with two eyes.)

With vision, your brain has learned other clues, such as closer objects
partially occlude your view of farther away objects, and objects farther
away are generally smaller than closer objects. Of course, these can be
fooled, For example, that is how painters can show the illusion of 3D
scenes on a 2D medium, or you can perceive depth in a 2D photograph or
when viewing a movie.

--
- Stephen Fuld
(e-mail address disguised to prevent spam)

On Friday, July 21, 2023 at 8:44:13 AM UTC-5, David Brown wrote:
> On 21/07/2023 13:25, Terje Mathisen wrote:
> > Tim Rentsch wrote:
>
> >> Not so. Human ears are sensitive to phase information content
> >> up at least 200 kHz or so, as anyone with a background in
> >> high school physics, geometry and trig can easily verify for
> >> themselves with a simple thought experiment. Imagine yourself
> >> in a completely dark environment, mostly quiet. Suddenly there
> >> is a sharp sound. Point to where you think the sound is. What
> >> is the angular resolution of where the sound is coming from?
> >> Translate that angular resolution into a delay in microseconds.
> >>
> >> Incidentally the time-delay resolution of ears has been directly
> >> measured at somewhere near 10 microseconds.
> >>
> > This is the part I really do believe: There should be no way for human
> > ears to determine front/back direction, but we are obviously able to do so!
> >
> > It has been shown that if you artificially smoothen the ear canal
> > (making it a cylinder), that capability goes away, so something in our
> > ears are able to detect really tiny phase modifications.
> The human hearing system (and that of other animals, especially mammals)
> is extremely complicated. It is not just two microphones, and position
> determination is done based on a large number of factors.
>
> The phase information is not really that a big influence, and the
> signals are /not/ timed anything like that accurately. (Phase
> information /is/ used, but it is not a big part of the process.) There
> are audio illusions where pairs of sounds are recognized by the brain in
> different orders from how they are played - timing resolution is of the
> order of tens of milliseconds at best. Your brain will fill in missing
> sound of up to 50 milliseconds without noticing - what you think you
> hear is reconstructed by your brain, just as is the case for your
> vision. Phase /difference/ between your ears is much more accurate, but
> that is very different from phase /resolution/.
>
> You can test the influence of phase by inverting the loudspeaker wires
> on one of your speakers. You might be able to notice a difference, but
> not much, and you won't notice a difference in position of the sounds
> despite a half wave-length change in the phase difference.
>
> You can also see that Tim's numbers are completely unrealistic. In 10
> microseconds, sound travels about 3.5 mm. So moving your head two
> millimetres while listening to a sound would have the same effect, yet
> it will do absolutely nothing to your perception of the source of the sound.
<
In a stereo with good sound stage presentation, moving your head absolutely
does change the direction from which you percieve sound coming. That is,
you rotate your head and the sound stage stays where it is/was while the
phase at your ears changes.
>
> Or listen to a sound from far away - you can still determine the source
> (roughly) even though the phase difference is utterly negligible in
> comparison to the distance, and varying air movements, echoes, and other
> effects totally swamp any kind of phase information.
>
>
> Two bigger influences in determining position are relative volume and
> phase difference, and the way the sound is affected by your head and
> outer ears. Sound passes through bone faster than through air, but is
> distorted in odd ways by the shape of your skull. Your ears and your
> brain go through years of fine-tuning themselves to use this kind of
> information for positioning. Just as changes to your ear canal mess
> with your ability to locate sound sources accurately, so do things like
> losing some or all of an outer ear, or part of your skull (such as from
> injury or brain surgery). You can re-learn afterwards.
>
> The biggest effect in real life (as distinct from very artificial
> experiments) is post-deduction rationalisation - you are convinced you
> know the direction by other factors (something you see, or otherwise
> know), and you then know that's the direction you are hearing it from.
>
> It's worth noting that people who have only one working ear still have
> some directional hearing. It's not nearly as good as if you have two
> ears, but it still works. (Compare this to 3D vision - with only one
> eye, you have virtually as good 3D perception as you had with two eyes.)
> >
> > Anyway, what's really amazing is that our brains have more volume
> > dedicated to sound than to vision!
> >
> That /would/ be amazing if it were true. About 25-30% of the brain is
> for visual processing, and about 8% of it is for audio processing. The
> figures are inevitably quite rough, because there's a lot of other
> things highly interconnected, such as speech (and therefore language)
> and your brain's models of the world around us.
>
> What /I/ think is amazing is that there are more nerves that carry
> information from the brain to the ears, than there are carrying
> information from the ears to the brain. These are especially important
> in the first few years of life (particularly the first six months), as
> the brain and ears work together to tune your ears to detecting and
> interpreting speech. This process lets you understand people talking
> about the chatter in a crowded room, or other noise that is higher
> volume than the person you are listening to. It also explains why adult
> language learners have trouble learning new phonemes that are not in the
> language(s) they heard as a small child - their ears are not trained to
> recognize them.

Terje Mathisen <terje.mathisen@tmsw.no> writes:

> Tim Rentsch wrote:
>
>> Terje Mathisen <terje.mathisen@tmsw.no> writes:
>>
>>> Thomas Koenig wrote:
>>>
>>>> Terje Mathisen <terje.mathisen@tmsw.no> schrieb:
>>>>
>>>>> This is where my signal processing background steps in and says: Start
>>>>> with the 24 bits/sample, 48 KHz master, and compress that with aac/ogg
>>>>> down to something a bit less than CD bitrates, maybe ~50% which is what
>>>>> you typically get from the lossless CD compression formats.(FLAC?)
>>>>
>>>> What about the properties of the low-pass filter you will need before
>>>> digitizing? Aliasing higher frequencies into the audible range can
>>>> really spoil hearing experience...
>>>
>>> By definition, the master digital recording have already applied that
>>> lowpass filter, right?
>>>
>>> Back in the analog days we had the same absolute limits on what the
>>> recording microphones could physically react to, as long as you sample
>>> that signal at more than 2X the highest frequency you will be OK.
>>>
>>> If your master is using 24 bits/sample @ 96 kHz then you will surely
>>> be well above any need for phase-changing lowpass filters, [...]
>>
>> Not so. Human ears are sensitive to phase information content
>> up at least 200 kHz or so, as anyone with a background in
>> high school physics, geometry and trig can easily verify for
>> themselves with a simple thought experiment. Imagine yourself
>> in a completely dark environment, mostly quiet. Suddenly there
>> is a sharp sound. Point to where you think the sound is. What
>> is the angular resolution of where the sound is coming from?
>> Translate that angular resolution into a delay in microseconds.
>>
>> Incidentally the time-delay resolution of ears has been directly
>> measured at somewhere near 10 microseconds.
>
> This is the part I really do believe: There should be no way for human
> ears to determine front/back direction, but we are obviously able to
> do so!

My understanding is the shape of the outer ear also plays a role
in direction location. Also, some animals, I believe owls are
an example, have ears that are slightly offset relative to one
another, and that property helps in direction locating.

> It has been shown that if you artificially smoothen the ear canal
> (making it a cylinder), that capability goes away, so something in our
> ears are able to detect really tiny phase modifications.

Making it a cylinder also has the property that the radius
stays constant. Is that what you meant? It's easy to
believe a changing radius could be a factor here (ie, so
the ear canal would be more like part of a cone than a
cylinder).

> Anyway, what's really amazing is that our brains have more volume
> dedicated to sound than to vision!

The ears listen well in all directions.

On 21/07/2023 16:19, Stephen Fuld wrote:
> On 7/21/2023 6:44 AM, David Brown wrote:
>
> snip
>
>> It's worth noting that people who have only one working ear still have
>> some directional hearing.  It's not nearly as good as if you have two
>> ears, but it still works.  (Compare this to 3D vision - with only one
>> eye, you have virtually as good 3D perception as you had with two eyes.)
>
> With vision, your brain has learned other clues, such as closer objects
> partially occlude your view of farther away objects, and objects farther
> away are generally smaller than closer objects.

Yes. There are /many/ ways your brain identifies 3D positioning when
you look at something - in the list of the top ten biggest factors,
stereoscopic vision would barely make it onto the list.

Some of these factors are also used in audio positioning. Absolute
volume of a sound you recognise will influence your idea of how far away
it is, for example, just like size of familiar objects for your vision.

>  Of course, these can be
> fooled, For example, that is how painters can show the illusion of 3D
> scenes on a 2D medium, or you can perceive depth in a 2D photograph or
> when viewing a movie.
>

Indeed. And there are auditory illusions too, though they are not as
common.

On Sat, 22 Jul 2023 07:04:04 -0700, Tim Rentsch
<tr.17687@z991.linuxsc.com> wrote:

>Terje Mathisen <terje.mathisen@tmsw.no> writes:
>
>> Tim Rentsch wrote:
>>
>>> Terje Mathisen <terje.mathisen@tmsw.no> writes:
>>>
>>>> Thomas Koenig wrote:
>>>>
>>>>> Terje Mathisen <terje.mathisen@tmsw.no> schrieb:
>>>>>
>>>>>> This is where my signal processing background steps in and says: Start
>>>>>> with the 24 bits/sample, 48 KHz master, and compress that with aac/ogg
>>>>>> down to something a bit less than CD bitrates, maybe ~50% which is what
>>>>>> you typically get from the lossless CD compression formats.(FLAC?)
>>>>>
>>>>> What about the properties of the low-pass filter you will need before
>>>>> digitizing? Aliasing higher frequencies into the audible range can
>>>>> really spoil hearing experience...
>>>>
>>>> By definition, the master digital recording have already applied that
>>>> lowpass filter, right?
>>>>
>>>> Back in the analog days we had the same absolute limits on what the
>>>> recording microphones could physically react to, as long as you sample
>>>> that signal at more than 2X the highest frequency you will be OK.
>>>>
>>>> If your master is using 24 bits/sample @ 96 kHz then you will surely
>>>> be well above any need for phase-changing lowpass filters, [...]
>>>
>>> Not so. Human ears are sensitive to phase information content
>>> up at least 200 kHz or so, as anyone with a background in
>>> high school physics, geometry and trig can easily verify for
>>> themselves with a simple thought experiment. Imagine yourself
>>> in a completely dark environment, mostly quiet. Suddenly there
>>> is a sharp sound. Point to where you think the sound is. What
>>> is the angular resolution of where the sound is coming from?
>>> Translate that angular resolution into a delay in microseconds.
>>>
>>> Incidentally the time-delay resolution of ears has been directly
>>> measured at somewhere near 10 microseconds.
>>
>> This is the part I really do believe: There should be no way for human
>> ears to determine front/back direction, but we are obviously able to
>> do so!
>
>My understanding is the shape of the outer ear also plays a role
>in direction location. Also, some animals, I believe owls are
>an example, have ears that are slightly offset relative to one
>another, and that property helps in direction locating.

That is true of all animals that have ears - humans included. If you
actually measure (map the head), you'll find the ears always are
asymmetrically placed - even if only slightly.

Owls and bats are among the animals that have the greatest differences
in their ear positions ... on some species you can easily see the
difference just by looking at them.

>> It has been shown that if you artificially smoothen the ear canal
>> (making it a cylinder), that capability goes away, so something in our
>> ears are able to detect really tiny phase modifications.
>
>Making it a cylinder also has the property that the radius
>stays constant. Is that what you meant? It's easy to
>believe a changing radius could be a factor here (ie, so
>the ear canal would be more like part of a cone than a
>cylinder).
>
>> Anyway, what's really amazing is that our brains have more volume
>> dedicated to sound than to vision!
>
>The ears listen well in all directions.