An Enticing Marketing Story, Theory Without Measurement?

[andreasmaaan]

It also results in flawed scientific experiments such as the one that purports to simulate a large reflective surface with a small speaker, on the basis that it kind of produces similar looking signals as would be captured by a single mic from a real surface if viewed in terms of frequency response and phase. But in terms of simulating 'the lake' (i.e. comparing the outputs of two mics simultaneously) it fails dismally.

We don’t know whether it fails. We’d need to test listeners to see whether they can discern a difference between identical soundwaves of differing radii (I can imagine how one might go about this).

Moreover, the fact that stereo creates a phantom image despite similar theoretical shortcomings suggests that such a simulated reflection is likely to be processed by the auditory system in much the same way a real reflection would.

 

[andreasmaaan]

Frequency and phase are 'abstractions' which may have some direct relevance to the hardware in human ears but not all of human hearing and acoustics.

All knowledge is arguably abstraction. Is this what you mean by this statement or are you making a more practical point about the way soundwaves travel through air?

 

[Cosmik]

All knowledge is arguably abstraction. Is this what you mean by this statement or are you making a more practical point about the way soundwaves travel through air?

Well sound can be thought of as vibrations, or three-dimensional sound fields, or time domain waveforms, or frequencies, etc. People go from a true three dimensional model at a fundamental level to a mathematical formula that describes a small portion of the system's output captured at a point in space - and which they have to assume repeats forever - and then they pronounce that it represents the entire system.

 

[Cosmik]

We don’t know whether it fails. We’d need to test listeners to see whether they can discern a difference between identical soundwaves of differing radii (I can imagine how one might go about this).

You could assert that about every conceivable experiment: make random, ad hoc substitutions for things that are difficult to get hold of or arrange, based on another experiment to see whether subjects can tell the difference.

So if I wanted to know whether a particular type of aircraft control algorithm was going to cause motion sickness I could think to myself "It's going to be tricky to test this in the real aircraft, but I can use one of those things you see at fairgrounds where we substitute tilting for acceleration; it'll probably be near enough. If we find that people can't describe the difference if we ask them to, that will be incontrovertible proof a demonstration of high probability that it's a reasonable substitution."

At every stage you build in another level of error that undermines the very thing you're doing the tests for. Why bother doing the tests if you already know what is an allowable substitution?

 

[andreasmaaan]

You could assert that about every conceivable experiment: make random, ad hoc substitutions for things that are difficult to get hold of or arrange, based on another experiment to see whether subjects can tell the difference.

They are neither random nor ad hoc. The fact that listeners perceive a phantom image indicates that the particular phenomenon which you consider to be a fatal flaw is of either marginal or zero psychoacoustic importance. If it were otherwise, we would perceive two separate sound sources and not a phantom centre.

But putting that to one side, the experiment seeks to answer the question: What types of early reflections (if any) do listeners prefer?

Reasoning through all the possible ways to answer that question, it becomes apparent there is no perfect way to answer it. So we need to choose the least worse way (or abandon trying to answer the question).

Personally, I want to find the best answer to the question I possibly can, because I want to know what best decision to make when designing or setting up listening rooms. So I’ll take the best answer rather than no answer at all.

In any event, your contention that the experiment is flawed on this basis would be trivial to put to the test: you’d just need to conduct an experiment to determine whether listeners can discern between a real and simulated reflection, which would be easy enough to set up.

 

[Cosmik]

They are neither random nor ad hoc. The fact that listeners perceive a phantom image indicates that the particular phenomenon which you consider to be a fatal flaw is of either marginal or zero psychoacoustic importance. If it were otherwise, we would perceive two separate sound sources and not a phantom centre.

As I was saying in another thread, stereo is a very, very special case that doesn't happen in nature: two identical sounds coming from different locations, and it is this that fools our hearing - thank goodness .

What our hearing receives is ambiguous - if we were trying to build an artificial system that could find the locations of sources in arbitrary reverberant environments, it would, at least in part, have to attempt to match plausible models of the acoustic scene to what it was picking up. 'Transients' are essential for the system to work - as discussed elsewhere, delays cannot be determined from steady tones. (Even so, the human can make their own transient from a steady tone: if there's something beeping from inside or outside our house but I don't know where, I find that I start by turning my head, and if that doesn't help me, I'll get up and walk around with more 'dynamic' head turning, and I suspect that something like creating my own Doppler shift helps me to locate the source).

From a single stationary mic, there might be any number of possible models of the acoustic scene that would produce the same result. Two mics would cut the ambiguity down. More mics, or moving the mics around dynamically would reduce ambiguity even further. It would make no sense, however, to train the system on anything that did not occur in nature; 'odd' acoustics are naturally going to fool us e.g. the domed ceiling that plays tricks with our hearing.

But putting that to one side, the experiment seeks to answer the question: What types of early reflections (if any) do listeners prefer?

But in doing so, it replaces a reflection with a fake reflection based on someone's idea of what the essence of a reflection is. It's not unlike deciding to run an experiment on whether subjects prefer butter or margarine, but deciding to use a butter substitute instead of butter - because the scientist already has an idea of what it is about butter that humans like.

If the scientist in charge of the reflection experiment has already stated something similar to what you say above ("The fact that listeners perceive a phantom image indicates that the particular phenomenon which you consider to be a fatal flaw is of either marginal or zero psychoacoustic importance") then they are already showing that they have drawn a conclusion based on a 'reduction' of acoustics. In this case they have concluded that because 'sound' emitted from two separate locations is perceived as emerging from a phantom point, sound emerging from a point (the speaker) is equivalent to sound coming from multiple locations. This does not follow!

I’ll take the best answer rather than no answer at all.

I can see flaws in that approach to life in general! There might be a better way. For a start, the definition of 'best' is usually just as much in question as everything else.

In any event, your contention that the experiment is flawed on this basis would be trivial to put to the test: you’d just need to conduct an experiment to determine whether listeners can discern between a real and simulated reflection, which would be easy enough to set up.

So why not use real reflections anyway, and stop people like me from causing trouble by questioning the results of the original experiment?

But I also see a problem with that 'trivial' approach, anyway. As I said above, if I create my own Doppler shift by moving around, for example, I may not know why I am moving around in the way I do, and if asked to describe what I am hearing, I may have no conscious idea of hearing any difference at all. If played a recording of the sound of distant beeping from a stationary mic, versus the same thing from a moving mic, I may hear no difference whatsoever in the context of sitting still and listening to the recording. Conclusion: humans gain nothing from dynamic head movements. But it wouldn't be true.

If, in your experiment to validate an experiment, you prevent me from getting up and walking around, you are creating an artificial situation (from an artificial situation). If I can walk around, I will spot the speaker (even in my blindfold) without any trouble. But even that's a weird situation I'm not used to. If you prevent me from walking around, that's a very artificial restriction, too. You can wave these anomalies away with words and a section on 'assumptions', but it really does look like an experiment designed to demonstrate an already-desired result.

 

[andreasmaaan]

In this case they have concluded that because 'sound' emitted from two separate locations is perceived as emerging from a phantom point, sound emerging from a point (the speaker) is equivalent to sound coming from multiple locations. This does not follow!

I think you’re misunderstanding my point here

Firstly, the sound coming from the speaker is not presumed to be equivalent to sound coming from multiple points. It’s presumed to be equivalent to a single reflection, that is, a sound coming from a single point.

Secondly, the fact that two sound sources create phantom images in stereo indicates not what you have written, but this: The two sources in stereo emit soundwaves that arrive at the ears from different angles (than would a single sound source in the location of the phantom image). Therefore, it can be inferred that any resultant ITD error is of lesser importance (or no importance) with respect to localisation than other factors.

But I also see a problem with that 'trivial' approach, anyway. As I said above, if I create my own Doppler shift by moving around, for example, I may not know why I am moving around in the way I do, and if asked to describe what I am hearing, I may have no conscious idea of hearing any difference at all. If played a recording of the sound of distant beeping from a stationary mic, versus the same thing from a moving mic, I may hear no difference whatsoever in the context of sitting still and listening to the recording. Conclusion: humans gain nothing from dynamic head movements. But it wouldn't be true.

That’s a huge number of steps in your hypothetical, all completely speculative. Could you point to an actual experiment concerning this point that has resulted in a similar series of missteps and misunderstandings on the basis of the reasons you speculate about?

I can see flaws in that approach to life in general! For a start, the definition of 'best' is usually just as much in question as everything else.

Everything is always in question. I’m using a combination of experimentation and reasoning to reach my conclusions, which is the only valid approach to this type of problem IMHO

 

[Cosmik]

Firstly, the sound coming from the speaker is not presumed to be equivalent to sound coming from multiple points. It’s presumed to be equivalent to a single reflection, that is, a sound coming from a single point.

Hmm. It sounds as though the experimenter just needs to suspend a reflective 'point' somewhere in the chamber rather than bother with a speaker then. A coin? I don't think that a reflection from a single coin would be audible or behave as intended at all frequencies. Doesn't this illustrate why the approach is not really a simulation of a reflective surface? But for sure, in the context of an anechoic chamber with a single reflective surface or simulation thereof, and a listener who cannot move around, maybe you could get some statistics to show that the speaker simulation is 90% effective or something. But the missing 10% might be the crucial difference that falsely gives the listener their preference for side reflections rather than ceiling. Why not just use a real reflective surface if it is trivial?

Secondly, the fact that two sound sources create phantom images in stereo indicates not what you have written, but this: The two sources in stereo emit soundwaves that arrive at the ears from different angles (than would a single sound source in the location of the phantom image). Therefore, it can be inferred that any resultant ITD error is of lesser importance (or no importance) with respect to localisation than other factors.

When we hear a phantom image in stereo, our normal hearing is being fooled with a static image - if we get up and walk around we gain no new information, and the image falls apart as we move. There is no information we can glean from 'radii' and the like, because there isn't any. It is just a beautiful 'stereoscopic scene' that hangs in front of us courtesy of the hijacking of one of our hearing abilities. Like the domed ceiling, we cannot work out the true situation - which I am glad about. But we cannot then extrapolate from that that any arbitrary acoustic situation can be simulated by a number of discrete point sources (or close to).

That’s a huge number of steps in your hypothetical, all completely speculative. Could you point to an actual experiment concerning this point that has resulted in a similar series of missteps and misunderstandings on the basis of the reasons you speculate about?

This is backwards! . My approach is "Don't just do something; stand there!".

If an experimenter wants to show something about perception of reflections, it is up to them to justify the short cuts they have taken. *They* are the ones creating the huge number of *active* hypothetical steps. I may be speculating, but it is passive speculation because I don't propose to take any short cuts or to reduce the dimensions of audio such as claiming that 'phase doesn't matter' (for example). If I don't like the sound of my room, I am not proposing to take the short cut of merely altering the signal, but would actually have to lay a real, physical carpet or whatever.

Because I can see that every artificial 'dimensionality reduction' of nature carries with it errors, ambiguities and the possibility of unconscious responses to stimuli that cannot be isolated experimentally, the best approach is to 'do nothing' if possible. So, I am going to attempt to make my speakers sufficiently neutral (as linear phase and time aligned as I can make them, too) without tricks like bass reflex or arbitrary peculiarities like spraying anti-phase out of the back, and I am going to assume that my hearing will 'hear through the room' (notwithstanding that it can also be hijacked with stereo).

The irony is that although this is the 'no short cuts' approach, it works out much less troublesome than the alternative .

 

[andreasmaaan]

So, I am going to attempt to make my speakers sufficiently neutral (as linear phase and time aligned as I can make them, too) without tricks like bass reflex or arbitrary peculiarities like spraying anti-phase out of the back, and I am going to assume that my hearing will 'hear through the room' (notwithstanding that it can also be hijacked with stereo).

Now the topic is shifting, but this is interesting. You say your ideal speaker has constant directivity, but how would you achieve that without either impracticably large horns and/or arrays, or some kind of cardioid arrangement (i.e. "spraying anti-phase out the back")?

 

[Cosmik]

Now the topic is shifting, but this is interesting. You say your ideal speaker has constant directivity, but how would you achieve that without either impracticably large horns and/or arrays, or some kind of cardioid arrangement (i.e. "spraying anti-phase out the back")?

I was careful to use the word "sufficiently" .

As I said in another thread somewhere, I would only trust the Kii approach if the anti-phase was all cancelled in the air i.e. not sprayed around the room. I am no longer as convinced that this is being achieved as I assumed before.

In another thread last night, I was pointing out that my sentiments are now more Grimm LS1. Maybe it has slightly less spectacular figures on paper (in themselves 'dimensional reductions'), but it is less radical, less 'Frankenstein'.

 

[andreasmaaan]

As I said in another thread somewhere, I would only trust the Kii approach if the anti-phase was all cancelled in the air i.e. not sprayed around the room. I am no longer as convinced that this is being achieved as I assumed before.

I'm hoping I can find a pair of these to measure a some point...

 

[svart-hvitt]

This thread is already 12 pages long, this is comment #232...

After so many words I’d like to quote Freeman Dyson again, as I did in the OP comment #1:

«Generally speaking there are two kinds of information: Observations and there are theories. So generally speaking you can believe the observations and you don’t need to believe the theories».

With observations in audio I mean listening tests (needless to say, of the competently managed kind).

AFAIK there has been only one such test, and even this test was of the rather shallow kind (8 trained listeners, all researchers connected to one company):

http://www.aes.org/e-lib/browse.cfm?elib=15154

The Olive, Jackson, Devantier, Hunt and Hess (all Harman employees) AES convention paper is already 10 years old. They found that some digital room compensation (DRC) packages were rated as positive by some of the trained listeners. In other words, this «observation» doesn’t rule out DRC as a valuable tool. But one observation doesn’t constitute evidence. It’s just an anecdote in scientific terms. In other words, we cannot say whether DRC is a positive or a negative thing as a matter of fact based on a wide range of observations.

I think it’s a mess - and maybe a reflection of the lack of resources in audio research - that we decades after DRC was introduced, still have no canonical papers on the issue or a wide range of observations to support such canonical papers.

 

[Duke]

Hi Dr. Toole,

Thanks for posting here and for mentioning that the third edition of your book is out. Cha-ching. That was the cash register on Amazon dot com - I just bought one.

In post number 174 you talk about listening fatigue and mention several things that are possible causes. I'd like to ask your opinion about another possible cause.

First, let me quote from post number 63:

... loudspeakers delivering flat and smooth direct sound, followed by reflected sounds having similarly smooth, but not necessarily flat responses, are consistently highly rated. Part of that may relate to the fact that the precedence effect works best when the direct and delayed sounds are similar. [emphasis Duke's]

So, how does the ear/brain system process a reflection that is spectrally somewhat dissimilar to the first-arrival sound, such as one might get with speakers that have good on-axis response but poor off-axis response (typical 6.5" two-ways come to mind)? For one thing, I would think the dissimilar spectral balance of the reflection accordingly skews perceived timbre a bit.

But how does the ear/brain system ultimately decide that it's a reflection and not a new sound event? My guess is that it's not as easy for the ear/brain system to correctly classify a reflection which is not a clear-cut "match" for a sound already in the precedence effect's short-term memory.

Let me digress for a moment then I'll come back:

David Griesinger points out why it is so fatiguing and inefficient to listen to a lecture given in an overly-reverberant hall: It literally takes the ear/brain system longer to understand each word, by which time the speaker has moved on to the next word, and the brain doesn't have sufficient time to link together the meanings of the different words into comprehension of complex ideas and thence conversion into long-term memory. Students who sit in the back of an overly-reverberant lecture hall soon become mentally exhausted by the sheer effort required to understand the words and don't remember the complex concepts being taught very well, if at all. Instead of leaving the lecture hall with a head full of new knowledge, he or she leaves with a headache and exhaustion. Anyway the most relevant part of this is, the listening fatigue induced by the ear/brain system's "extra CPU power" and extra time needed simply to understand each word.

Okay, back to the listening room, and our hypothetical speakers that have poor off-axis response, resulting in spectrally somewhat dissimilar reflections. Might it require more "CPU power", and/or more time, for the ear/brain system to comprehend and correctly classify these reflections? And might this process result in a form of "listening fatigue", and perhaps even a headache given sufficient time?

(The above first occurred to me when I noticed a correlation between listening fatigue and speaker configurations that looked to me like they'd have poor off-axis response. Of course correlation is not necessarily causation).

Anyway, any thoughts you might have on the subject would be greatly appreciated.

 

[svart-hvitt]

CONCERT HALLS AS A SOURCE OF CONFUSION

Many experienced listeners and engineers recommend younger people to learn how an unamplified performance sounds. Late Peter Aczel recommended just this recipe: Unamplified concert and music. @Floyd Toole has reported on his many visits of concert halls too.

I am not a regular goer to concerts. During the ones I have been to, I sat 15 meters or more from the performers. Lucky me, I have attended private concerts too; distance from performer(s) was 1 meter and up to 10 meters.

I have hifi friends that go to symphony orchestra concerts more regularly. So I remember one of my friends telling me a symphony orchestra should sound dark on a high resolution, well calibrated hifi system. I never thought of this remark in a critical fashion before I met this rather obvious illustration by Lokki et al.:

Source: http://www.caa-aca.ca/conferences/isra2013/proceedings/Papers/P040.pdf

Take a look at the frequency response of the virtual loudspeaker orchestra in the top left FR chart. Higher frequencies are attenuated significantly due to sound power dispersion as a function of distance in air. There is nothing new in this chart; but it made me look at an old statement on «dark symphony sound» from this very perspective.

My friend uses digital room compensation to calibrate his system. He got help setting this drc system up from a performer of classical music.

Then it strikes me; if you go regularly to classical concerts and are used to the air related attenuation of higher frequencies you would anchor your evaluation of reproduced classical music as if you had a seat 10-15 meters or more from the performers. If you bring this experience - which may have become habit and a preference - you would expect the same attenuation of higher frequencies from your speakers. However, in case the recording you listen to is made closer to the microphone than 10-15 meters, the recording may sound too bright. Right?

My point is: Coming home from a symphony concert may tilt your preferences of such sound in a way that is not easily compatible with recordings of symphony orchestras.

The air attenuation problem; how does one deal with this in a recording and unknown reproduction scenario?

Should recording of symphony orchestras sound dark (i.e. attenuated in low-frequency terms) or as bright as the recording dictates?

Should a fan of classical music EQ his reproduction system to attenuate higher frequencies through DSP?

I am confused

ADDENDUM: In private concert where I was much closer to performers, I often found the sound shrill, too bright. This adds to the suggestion that distance related air attenuation of higher frequencies is of essence in both recording and playback, right?

 

[Floyd Toole]

@ Duke. Struggling to understand speech in rooms having excessive reverberation is a well known issue, and it undoubtedly contributes to what one could call "fatigue". Such rooms tend to be above 0.5s RT, which is unlikely to occur in furnished domestic rooms. There is an extensive literature on it in the context of classrooms in JASA.

The precedence effect really boils down to the direct sound dominating localization. Perception of secondary images associated with delayed sounds (reflections in our context) defines the breakdown of the precedence effect. When the precedence effect is functioning, we are aware only of the direct sound as far as localization is concerned, but later sounds can affect loudness and timbre. This is discussed in some detail in my book, starting in Section 7.6.4. It is very signal dependent.

One of the factors known to cause delayed sounds to be more apparent is a difference in spectrum. So the perceptual inhibition (NOT masking) of delayed sounds as just described is related to the spectral similarity of the direct and reflected sounds. Delayed sounds that are spectrally similar to the direct sound are most successfully inhibited, as far as localization is concerned. Nobody I am aware of has chased down the quantitative differences necessary to be useful in the context of sound reproduction. All that I can say is that loudspeakers exhibiting "similar" direct and off-axis performance receive elevated sound quality ratings, while those misbehaving off axis do not. Those are also the speakers the have the fewest audible resonances, and that tend to "disappear" behind the screen in double-blind tests.

Direct sound dominates in terms of sound quality as was shown in my early papers in 1985-86, and later, and it is also dominant in terms of sound localization if it is not corrupted by spectrally dissimilar reflected sounds. Does it require more "CPU power" when off axis misbehavior exists? I can only speculate that if sound quality is degraded - which is evident in every program played - I can imagine that it would eventually become an annoyance, even if it is a subconscious one. This is where the science that I am aware of fades.

I just spent several hours listening to new music on Tidal (I am getting tired of my old repertoire and things are still extremely variable in the context of spatial presentation. With subtle Auro3D upmixing, some cuts put me in the middle of an enormous space extending beyond the walls of my room. It was exhilerating. Others reverted to the old fashioned L, C, R dominated format - a flat specially deprived soundstage, even with the upmixing engaged. Same speakers, same room.

 

[Floyd Toole]

CONCERT HALLS AS A SOURCE OF CONFUSION

Many experienced listeners and engineers recommend younger people to learn how an unamplified performance sounds. Late Peter Aczel recommended just this recipe: Unamplified concert and music. @Floyd Toole has reported on his many visits of concert halls too.

I am not a regular goer to concerts. During the ones I have been to, I sat 15 meters or more from the performers. Lucky me, I have attended private concerts too; distance from performer(s) was 1 meter and up to 10 meters.

I have hifi friends that go to symphony orchestra concerts more regularly. So I remember one of my friends telling me a symphony orchestra should sound dark on a high resolution, well calibrated hifi system. I never thought of this remark in a critical fashion before I met this rather obvious illustration by Lokki et al.:
View attachment 24590
Take a look at the frequency response of the virtual loudspeaker orchestra in the top left FR chart. Higher frequencies are attenuated significantly due to sound power dispersion as a function of distance in air. There is nothing new in this chart; but it made me look at an old statement on «dark symphony sound» from this very perspective.

My friend uses digital room compensation to calibrate his system. He got help setting this drc system up from a performer of classical music.

Then it strikes me; if you go regularly to classical concerts and are used to the air related attenuation of higher frequencies you would anchor your evaluation of reproduced classical music as if you had a seat 10-15 meters or more from the performers. If you bring this experience - which may have become habit and a preference - you would expect the same attenuation of higher frequencies from your speakers. However, in case the recording you listen to is made closer to the microphone than 10-15 meters, the recording may sound too bright. Right?

My point is: Coming home from a symphony concert may tilt your preferences of such sound in a way that is not easily compatible with recordings of symphony orchestras.

The air attenuation problem; how does one deal with this in a recording and unknown reproduction scenario?

Should recording of symphony orchestras sound dark (i.e. attenuated in low-frequency terms) or as bright as the recording dictates?

Should a fan of classical music EQ his reproduction system to attenuate higher frequencies through DSP?

I am confused

ADDENDUM: In private concert where I was much closer to performers, I often found the sound shrill, too bright. This adds to the suggestion that distance related air attenuation of higher frequencies is of essence in both recording and playback, right?

The downward tilt of the curves (If they are steady-state curves, most likely) is primarily caused by bass rise - all musical instruments, and loudspeakers, tend towards omnidirectional radiation at low frequencies (Figure 10.5 in the 3rd edition). Air attenuation begins around 1-2 kHz. All of this is discussed in the context of cinema sound in Chapter 11 in the 3rd edition.

Many people have argued that the air attenuation effect is automatically accommodated by humans who associate it with the listening location in the specific venue - and treat it as normal. This is one of the discussion points with respect to cinema sound. Do we employ array speakers that can deliver "flattish" sound to all parts of the audience, or simply let the HF roll off with increasing distance. The jury is still out.

Orchestras in concert halls do exhibit a bass-heavy energy response, which I mention briefly in Chapter 18 in connection with the Acoustic Research AR3a where they justified the dramatic high frequency rolloff on the basis of it being similar to the Boston Symphony orchestra spectrum in Symphony Hall. Wrong . . . This is what mixing and mastering engineers are supposed to take into account as they combine the sounds picked up by mics that are close to and above the orchestra - therefore "hearing" something very different from a member of the audience, especially in the main floor "orchestra" seats.

In my experience of listening to several versions of the same classical movement in different recordings I find significant variations (this is easy to do on Tidal). Some are almost mono and "distant". Others are clearly stereo, but with a flat soundstage. Still others, the one's I prefer, deliver some depth and space. All with the same speakers in the same room. It is all microphone and mixing technique. Recording and mastering engineers contribute greatly to what we hear.

 

[Duke]

Thank you very much for that in-depth reply.

When the precedence effect is functioning, we are aware only of the direct sound as far as localization is concerned, but later sounds can affect loudness and timbre.

Ime a significant benefit of speakers like the Mirage M1 is the rich timbre from having a relative abundance of spectrally-correct reflections. Of course "spectrally-correct reflections" is a cat which can in more ways than one be skinned, but the bipolar configuration of the M-1 also offers the prospect of introducing a fair amount of path-length-induced time delay before the "backwave" energy reaches the listening position, and I think the time delay is beneficial... and perhaps even moreso if the speakers are toed-in such that their backwave energy actually arrives after bouncing from backwall to sidewall, then bouncing towards the listener from the theoretically ideal (IIRC) 10:00 and 2:00 angles.

Did your team at Harman evaluate any speakers conceptually similar to the M1, aside from that one infamous hybrid electrostat? I would not expect a bipolar speaker to fare well on the Spin-O-Rama because its spectrally-correct backwave energy would have the shorter wavelengths significantly attenuated by the curtained area behind it.

In retrospect that may be an invalid assumption on my part, as your multi-room paper indicates that listeners are able to hear through the room and reliably evaluate speakers on their own merits.

So the perceptual inhibition (NOT masking) of delayed sounds as just described is related to the spectral similarity of the direct and reflected sounds. Delayed sounds that are spectrally similar to the direct sound are most successfully inhibited, as far as localization is concerned.

That makes sense, and imo is an argument for very well-behaved directivity, like in the JBL M2 and Revel Ultima.

I suppose a question which naturally arises is, how wide should that well-behaved pattern be? Which in turn implies, what sort of direct-to-reverberant ratio is preferable, and what are the implications of having correspondingly more or less energy in early reflections? My guess is the answer is some variation of "it depends", but I'm not sure of WHAT it depends on.

Also, my recollection is that the M2's curve is very "flat", while the Ultima has a gently downward-sloping on-axis curve, perhaps in approximation of the air attenuation effect? (My memory may be faulty). Does the radiation pattern width inform what the on-axis curve should look like?

 

[Dimifoot]

Such rooms tend to be above 0.5s RT, which is unlikely to occur in furnished domestic rooms.

Seems to be more common amongst my friends, and quite common in Greek forums.
Could it be a different approach to architecture/interior design in recent years, or a local thing in Greece or the Mediterranean, comparing to US/UK?

Take a look:

I am not an acoustician or a calibrator/installator (actually an Orthopaedic Surgeon, MD), and I only do measurements for myself and close friends. So its not that I have a huge database of measurements to be able to pick only the +0.5s ones to show.

These were done for friends who found their systems tiring.

Measurement House n3 was done in my house , some years ago, when my growing interest in improving the sound of my system turned my attention to REW measurements. That led me to treat the room extensively (diffusion on first reflection and back, lots of bass traps/resonators, treatment of a stairway), and the result is a hugely improved feeling, in home cinema, multichannel and stereo music but also conversation.
(It might be just a touch less "live" from what I would like, and I plan to get it closer to 0.35s in the nest weeks)

Was I able to "hear my music through the room" before treatment?
Of course, as I am able to recognise my wife's voice through the 0,2$ loudspeaker of my mobile phone.
So yes, but I didn't enjoy it that much (the music, not the wife)

 

[DDF]

Seems to be more common amongst my friends, and quite common in Greek forums.
Could it be a different approach to architecture/interior design in recent years, or a local thing in Greece or the Mediterranean, comparing to US/UK?

In interview, KEF's Jack Oclee-Brown (head of acoustics) described how, in his experience, European living acoustics are typically more lively in the upper octaves than in North America, requiring some adjustment to the design targets, on average (unlike DIY, KEF are targeting product for the broadest possible applications). Its not too difficult to infer what those might be, such as tighter and better controlled dispersion and/or slightly more aggressive off axis higher frequency roll off. I think this is reflected in their design choices, such as uni-q.

 

[Guermantes]

It also results in flawed scientific experiments such as the one that purports to simulate a large reflective surface with a small speaker, on the basis that it kind of produces similar looking signals as would be captured by a single mic from a real surface if viewed in terms of frequency response and phase. But in terms of simulating 'the lake' (i.e. comparing the outputs of two mics simultaneously) it fails dismally.

@Cosmik, are your referring to the studies by Lokki, et al? I see that there are others following this line of producing virtual acoustics:
http://www.aes.org/e-lib/browse.cfm?elib=19713

This last one has just come across my desk today. It includes a substantial references list, so I look forward to having a read.