An Enticing Marketing Story, Theory Without Measurement?

[andreasmaaan]

But the simulated 'reflection' is coming from a point-like source with its own directivity characteristics..? The discrepancies between what reaches the two ears will be different from a real reflection, and as the listener moves their head, more information about the source of the sound will become available.

I still can’t quite understand this one. Unless the source is extremely directional, there won’t be a significant difference between the spectrum at each of the listener’s ears (except that caused by the listener’s head, which is a natural and constant part of the listener’s HRTF and which would be present whether the reflection were real or simulated).

I'm saying that because the listener is most likely not fooled by the simulated reflection, they're going to register it as a source of sound. And then the experiment is really about what locations for sound sources humans have evolved to find the least/most threatening. Sounds from above might be especially threatening, so people registering ceiling 'reflections' as detrimental to the sound are really just indicating some evolutionary mechanism. If it was a genuine reflection they wouldn't register it in the same way.

You’re starting from the assumption that the listener isn’t “fooled” by the simulated reflection, but otherwise I think that’s an interesting idea.

 

[Cosmik]

I still can’t quite understand this one. Unless the source is extremely directional, there won’t be a significant difference between the spectrum at each of the listener’s ears (except that caused by the listener’s head, which is a natural and constant part of the listener’s HRTF and which would be present whether the reflection were real or simulated).

That's pure frequency domain thinking

I'm thinking of it like this: imagine the source being point-like, and a pulse of some kind spreading out from it as a sphere with increasing diameter. If it hits a continuous, large reflective surface, it bounces off and continues to spread out down towards you. You have two ears at fixed spacing, and the direct wavefront moves past them as a spreading circle of a certain radius proportional to the distance from the source, and the reflection has a larger diameter, as it has been spreading for longer.

Imagine an animation of spreading concentric circles and two points (your ears) 'seeing' the circles moving past.

A simulation of a reflection based on a speaker effectively re-starts the reflection as a small sphere, so that when it reaches your ears it may only have the same radius as the original source. This changes the 'animation' of relative phase and/or delay moving past the ears. (It would be good to create an animation to demonstrate that it would, indeed, be different).

Maybe there are a few ambiguous positions, so if you are oriented perfectly perpendicular to the source and reflection the simulated reflection is indistinguishable from a real one but I'll bet that if you shifted position slightly, all would be revealed.

 

[andreasmaaan]

This changes the 'animation' of relative phase and/or delay moving past the ears. (It would be good to create an animation to demonstrate that it would, indeed, be different).

Interesting thought experiment. But why would changing the radius of the hypothetical sphere result in the relative phase being different?

Keep in mind that the simulated reflection is timed to emanate from the secondary speaker after the simulated direct sound from the primary speaker, with the delay corresponding to a simulated distance between the direct sound and the reflective surface.

EDIT: I think your argument has clicked now. You're saying that the radius of the wavefront would affect the inter-aural time difference - is that correct? Very interesting idea.

If humans were sensitive to this phenomenon, this would be a problem for stereo/multichannel sound in general, because the ITD of two separate spherical wavefronts arriving from two different directions would be different than for a single spherical wavefront arriving from a point between the two (or more) sources. Yet the phantom image still appears. If this is so, surely so must the phantom reflection?

A secondary point would be that the radius of the wavefront (assuming it's spherical, which is not necessarily the case anyway) could be altered by moving the secondary source. There's no rule stating that the primary and secondary sources have to be equidistant from the listener in the anechoic chamber.

Your other point about small head movements not affecting the relative directions of the direct and "reflected" sound in the same way as they would in a real room is more concerning though IMO.

 

[Cosmik]

If humans were sensitive to this phenomenon, this would be a problem for stereo/multichannel sound in general, because the ITD of two separate spherical wavefronts arriving from two different directions would be different than for a single spherical wavefront arriving from a point between the two (or more) sources. Yet the phantom image still appears. If this is so, surely so must the phantom reflection?

I am thinking of these as two separate phenomena.

One of them is: how do we read our natural acoustic environment? I think it is akin to 'focusing', just as we do visually. There is a 'ray tracing' element to modelling it. We learn to interpret reflections and 'focus' on the source.

When we look at reflections in a mirror we are not focussing on the mirror, but are able to 'focus back' to the source. If we see a mirror it can be confusing, but with two eyes and physical movement we quickly establish what is the mirror and what is the source. Hearing is the same, I think: a reflection from a large surface still leads us back to the source (through the 'radius' phenomenon). Think of the listener's hearing as being a lens gathering 'rays' and looking to where they point back to. Like an array of correlated radio telescopes, the distance between our ears effectively sets the diameter of the notional 'lens', and physical head movement makes it bigger, increasing its resolving power.

Certainly, you could imagine the two-mic-with-movement system being used with a neural network to learn to navigate by sound alone - I believe our brains and ears have evolved/learned to do this in order to locate sounds sources in reflective environments. Underneath, it is creating possible models of the static environment based on phase, timing and volume differences and the more information it has, the less ambiguous the model. It may be expressed mathematically, but with a neural network you don't need to know any maths explicitly; you just 'train' the system.

A secondary point would be that the radius of the wavefront (assuming it's spherical, which is not necessarily the case anyway) could be altered by moving the secondary source. There's no rule stating that the primary and secondary sources have to be equidistant from the listener in the anechoic chamber.

Yes, I wasn't assuming any particular relationship between them nor that it would be a perfect sphere (I did mention earlier that the speaker will have directivity at different frequencies and that it is not a point, but is relatively 'point-like'). It's just that each configuration will differ between what it is trying to simulate and the reality that our brains have evolved to interpret.

The visual analogy for the second phenomenon, stereo, is , I think, the putting on of 3D goggles. It hijacks the human hearing system by doing something that cannot happen in nature: producing identical sounds from two separate sources - and Blumlein stereo really does this: it creates inter-speaker volume differences which, thanks to the crossfeed from each speaker to both ears, creates inter-aural time differences. To our hearing, I think it is more about time domain correlation than explicit phase relationships - I don't think the brain is using those as its primary cues with stereo.

The truly amazing thing is that if you model speaker-based stereo using simple correlation between what the ears would be picking up as the model of the human's hearing system, it produces a perfect 'auditory scene' that even stays stable with head movement. But this scene is static, like a 3D television image with special glasses: if you move around you are not going to gain extra information from 'parallax'.

The great gift for audiophiles is that we are able to fuse the static artificial scene with the natural acoustics of our environment.

Both artificial scene and room remain stable as we turn our heads and - to a limited extent - move around. It really is an amazing system; the fantastically compelling, detailed imaging from a good stereo system it is not just a wishful illusion. It may be an illusion in one sense, but it is 'real' in that it can be conjured up deliberately and repeatedly.

 

[Juhazi]

Cosmic, no need to develop new theorieswww.davidgriesinger.com

Vision and hearing are also quite different and vision occupies much larger part of brains, much more calculating power! Still, both systems can be fooled easily, this is why we can enjoy audio and video reproduction.

 

[Bjorn]

Dealing with the floor reflection is easily heard with most speakers IMO. Clarity and intelligibility is better and the sound is less blurry. Personally I like that, but I can't debate with anyone who says he doesn't. Using a carpet is not a good way to treat it though. It's both very band limited functioning as a filter/EQ and seldom attenuates well even for higher frequencies.

Back to topic. A common problem arising in the debates about room correction seems to be a lack of understanding of time and magnitude and how they correlate. The person behind the Audiolense software said at a norwegian forum that he disagreed with those who said the time domain was more important and thought that the frequency response more important. This shows a serious lack of understanding. The frequency response follows the time domain, not the other way around!

 

[Cosmik]

Cosmic, no need to develop new theorieswww.davidgriesinger.com

LookVision and hearing are also quite different and vision occupies much larger part of brains, much more calculating power! Still, both systems can be fooled easily, this is why we can enjoy audio and video reproduction.

Which parts of the link should I be looking at? (There's quite a lot there!)

Am I not allowed to develop any new theories of my own?

We all have our ways of looking at things: mine come from being a 'systems' engineering person who blends hardware with software more seamlessly than most people in my profession - who tend to be experts in hardware OR software. I use audio DSP, image processing, pattern recognition, neural networks for practical purposes and I think I do have an insight into how a 'dumb' neural system can take raw information from transducers and be trained to, effectively, perform complex mathematics on it - without really doing anything but summing with weights.

Some people will look at the problem of hearing as a transducer hardware problem, because that is what they are experts in. They will attempt to find physical hardware in the ears that leads to hearing phenomena like localisation etc. And I am sure that such mechanisms exist in the hardware as pre-processors. But they are relatively crude in comparison to what could be done afterwards in the brain using 'software' - or at least 'firmware'.

In your own comment you betray something of the hardware-centric mentality in that you state that vision and hearing are quite different on the basis that they occupy different regions of the brain. This does not follow. Where the processing power resides and how big it is physically has no bearing on what it is doing in principle. For sure, vision requires more hardware, but it does not follow that evolution hasn't given a human the ability to 'see' with sound and vision in closely-related ways.

 

[andreasmaaan]

One of them is: how do we read our natural acoustic environment? I think it is akin to 'focusing', just as we do visually. There is a 'ray tracing' element to modelling it. We learn to interpret reflections and 'focus' on the source.

I agree with this in principle. I’m suggesting though that the stereo illusion is a good example of how a non-natural acoustic environment can be used to trick the brain into perceiving a particular natural environment.

The other more specific point is that a sound source 30 degrees to the side will create a different ITD to a sound source directly in front. Yet - despite false ITDs in comparison to a real acoustic event emanating from the location of a phantom image - the brain perceived two equal off-axis sound sources as producing a “natural” sound source directly between the two sources. Arguably the same mechanism is at work if a simulated reflection creating a different ITD to a particular real reflection will be perceived as a “natural” reflection. Indeed, the stereo phantom image can be considered to be a particular instance of the precedence effect, just as can the effect of adding a simulated reflection (which has the effect of widening or shifting the apparent image).

It's just that each configuration will differ between what it is trying to simulate and the reality that our brains have evolved to interpret.

True, but there is no single reality here. Each particular real reflection will have its own particular phase characteristics determined by the wavefront of the direct sound and the acoustic properties of the reflective surface. There is no single real reality that the experiment seeks to replicate, but rather a range of possible realities that it approximates.

Both artificial scene and room remain stable as we turn our heads and - to a limited extent - move around. It really is an amazing system; the fantastically compelling, detailed imaging from a good stereo system it is not just a wishful illusion. It may be an illusion in one sense, but it is 'real' in that it can be conjured up deliberately and repeatedly.

I agree

 

[Juhazi]

cosmic, have you noticed prof. Tapio Lokki's studies? Pretty close to your interests, combining perception of sound and computers, analyzing and replicating/synthesizing virtual acoustics. These go well over my head... I have biological/medical background only. Griesinger has nice papers and videos where he explains acoustics and perception of sound, which is a very complex entity.

https://people.aalto.fi/tapio.lokki#publications

 

[Guermantes]

Let us pause then and reflect on the wonder that is man. From the blink of our existence to the past 100 years we have "evolved" into an animal that creates "music" and has also created the means of reproducing it with our brilliant mechanical tools so that we may debate objectively on how our early ancestors might respond to such unnaturalness using science.

Aside from our natural tendency to exploit and destroy most of what we can we also come up with the good stuff. We may have evolved too fast for our own britches but ahhhhh, the ecstasy of sound done well.... Which makes me wonder how and with what tools we will be listening in another 50 to 100 years.

Have you read this book?
https://mitpress.mit.edu/books/million-years-music

It's about the evolution of capacities in hominins that gave rise to "musicking" and, more broadly, "culture". I'm half-way through at the roughly 100,000 years ago mark. Fascinating stuff.

 

[Guermantes]

The truly amazing thing is that if you model speaker-based stereo using simple correlation between what the ears would be picking up as the model of the human's hearing system, it produces a perfect 'auditory scene' that even stays stable with head movement. But this scene is static, like a 3D television image with special glasses: if you move around you are not going to gain extra information from 'parallax'.

The great gift for audiophiles is that we are able to fuse the static artificial scene with the natural acoustics of our environment.

Both artificial scene and room remain stable as we turn our heads and - to a limited extent - move around. It really is an amazing system; the fantastically compelling, detailed imaging from a good stereo system it is not just a wishful illusion. It may be an illusion in one sense, but it is 'real' in that it can be conjured up deliberately and repeatedly.

I think this is where there is some dissonance between talking about the speakers as sound sources in environments which we react to in an evolutionary sense (i.e. sound sources which we can locate in space via reflections in the environment and cues between our ears, like the crack of a twig behind us in some primeval forest), and as sound sources which we are using to create an illusion of an acoustic space or soundstage.

In the former, we are arguing that room modes, floor bounce, side reflections, etc., are natural aspects of our listening experience in relation to speakers as sound-sources-in-themselves. But in the latter, we want the speakers to disappear as sound-sources-in-themselves behind the stereo illusion of a recorded acoustic. Do we want to hear speakers in a room? Or do we want to hear the auditory illusion?

 

[Dimifoot]

McEachern [17] argues that signal detection, identification, and location (localization) were the most critical signal processing tasks for the eyes and ears of early humans. (I recall an unknown reference that observed that an object must emit or disturb energy to be detected by a biological or man-made sensor.) To hunt and avoid being eaten, our ancestors had to detect nearby movement, identify it as prey, predator, or human, and determine its location so they could run towards or away from it as necessary.

Didn’t we hunt (and got hunted) for hundred thousand years on soil/grass/turf? The hunting grounds weren't made of concrete, wood or marble all these years, like our home floors are today.
So our evolution of ground reflection wasn’t a result of survival process on highly reflective ground surfaces.

 

[Cosmik]

Didn’t we hunt (and got hunted) for hundred thousand years on soil/grass/turf? The hunting grounds weren't made of concrete, wood or marble all these years, like our home floors are today.
So our evolution of ground reflection wasn’t a result of survival process on highly reflective ground surfaces.

Low frequencies would still be reflected. But that is besides the point: I think that if we have the ability to identify sound sources in reflective environments, that is a general ability regardless of whether the reflections come from trees, cave walls or the ground. Like your vision works equally well wherever you are looking.

If you were to create an artificial listening system, I think you would be failing if you built into it a 'ground reflection factor'. The general solution shouldn't care about that. Another part of the system might be concerned with clues to orientation and so on, but this would not be part of the low level 'hearing and source identification function'.

 

[DWPress]

Have you read this book?
https://mitpress.mit.edu/books/million-years-music

Thanks for the suggestion Guermantes, looks like a good read.

 

[Dimifoot]

But that is besides the point: I think that if we have the ability to identify sound sources in reflective environments, that is a general ability regardless of whether the reflections come from trees, cave walls or the ground.

My post was a suggestion against the argument that “floors need no treatment, because it sounds unnatural, since we are used to floor reflections for evolutionary reasons”.

Yes, we are used to the floor being there, but it hasn’t been highly reflective for hundred thousand of years.

 

[svart-hvitt]

On floor (sic!) or lack thereof: It may have been addressed already, but what about the floor bounce effect on the recorded material?

Twice the reflections is double good?

 

[Dimifoot]

Furthermore I find the discussion on the (lack of) scientific evidence on the benefits of Room Eq (above the transition frequency) misleading.

Shouldn’t we examine each software separately? They are so remarkably different, Ypao vs Dirac for example, or Audyssey vs Trinnov...we shouldn’t be condemning (or accepting) the use of Room Eq in general.

I think that there are huge differences between them, and also in each one’s interaction with different speakers/rooms.
It’s obvious that research is needed, and I agree that it should be based on listeners preference and not measured “perfections”.

 

[Cosmik]

On floor (sic!) or lack thereof: It may have been addressed already, but what about the floor bounce effect on the recorded material?

Twice the reflections is double good?

Floor bounce on the recorded material sounds like colouration because the listener can't separate it from the sources (not enough sound field information). When 'live', floor bounce has no audible effect, except as has been mentioned, as part of a 'spaciousness' effect.

To the frequency domain-centric person this does not compute, because in both cases, a measurement would reveal a distinct effect on the spectrum of continuous sine waves. How can it not sound the same? The answer is most likely that humans are not just frequency response analysers: in a live sound situation they are 'sound field analysers' and 'focus' on the source; they literally do not hear the tonal changes that dumb measurements would suggest.

A recording doesn't contain the information to allow the human to perform this focussing, so any effect on the spectrum due to recorded reflections becomes part of the (composite) source, and is heard as a tonal change of the source.

Does this not tally with your experience? (e.g. recording a concert from in the audience and then being shocked at how coloured it sounds)

All relevant to the idea of so-called room correction...

 

[svart-hvitt]

Floor bounce on the recorded material sounds like colouration because the listener can't separate it from the sources (not enough sound field information). When 'live', floor bounce has no audible effect, except as has been mentioned, as part of a 'spaciousness' effect.

To the frequency domain-centric person this does not compute, because in both cases, a measurement would reveal a distinct effect on the spectrum of continuous sine waves. How can it not sound the same? The answer is most likely that humans are not just frequency response analysers: in a live sound situation they are 'sound field analysers' and 'focus' on the source; they literally do not hear the tonal changes that dumb measurements would suggest.

A recording doesn't contain the information to allow the human to perform this focussing, so any effect on the spectrum due to recorded reflections becomes part of the (composite) source, and is heard as a tonal change of the source.

Does this not tally with your experience? (e.g. recording a concert from in the audience and then being shocked at how coloured it sounds)

All relevant to the idea of so-called room correction...

I think I am of the opinion (note: OPINION as opposed to a scientifically based statement) that much music would sound really great in an anechoic chamber. Having heard VERY big speakers, near field in a dead room, I wonder if reflections in room are sometimes a compromise. Having heard a 7.4.1 setup, which I believe worked as a bombardment of direct sound, it was as if I was transported to the church and its acoustics in that room.

So I cannot, based on anecdotal experience, forget the idea that reflections in room are a compromise, colouration. Some like colouration. I may not. Most don’t know how much room colouration is just right...

I would very much have liked to experience 2.1 and multichannel in a big anechoic chamber....

 

[JohnPM]

Anechoic chambers are unpleasant places to spend time in, can be quite disconcerting.