accuracy is subjective

[castleofargh]

Does not exist, but can we listen to something that was recorded with cues close enough to achieve a fair horizontal spatial effect?

if all the recording and playback process was focused on this, then sure enough we could achieve amazing results. my statement was based on how almost all albums ever made were recorded.
and of course as you very well know, for headphones it doesn't help that most albums were mastered on speakers. but again it's only a matter of recording with a particular use in mind.

 

[aukhawk]

Originally Posted by johncarm /img/forum/go_quote.gif
 
... So, she found that vinyl was much more accurate than a typical CD player. In vinyl playback, the instruments sound a lot like themselves. For instance, a trombone has the right quality of overtones. The E string of a violin is bright and powerful without being harsh, just as in real life. 
As well, in vinyl playback, the instruments blend properly. That is important if you want recordings to be useful for studying conductor technique.
I AM AWARE THAT SOME AUDIOPHILES OR RECORDING ENGINEERS DISAGREE. Some find CD to be more accurate, generally.
Therefore, accuracy is subjective.

 
New poster here.

It seems to me that it is fairly simple to make an objective assessment of vinyl vs digital - simply needledrop a vinyl record and digitally record the result.  Then just compare the two preferably blind (admittedly that's not very easy - synchronisation and level-matching).  If digital is accurate it will faithfully reproduce the vinyl sound.
 
I have done quite a lot of needledrops of my old vinyl collection and I actually think the digital versions sound better - this is mainly down to two things, (1) I record with the sound turned down low, thus eliminating any acoustic feedback in the vinyl playback, and (2) I can de-click the recorded result.
 
ff

 

[jgazal]

if all the recording and playback process was focused on this, then sure enough we could achieve amazing results. my statement was based on how almost all albums ever made were recorded.
and of course as you very well know, for headphones it doesn't help that most albums were mastered on speakers. but again it's only a matter of recording with a particular use in mind.

Well, to be sincere, I don't really know.

I have never listened to a Bacch-SP processor or a Realiser set to avoid crosstalk.

So I just cannot tell what percentage of an unbiased (representative) sample the universe of recorded music available has been deliberately mixed with ILD (panpot) and ITD so that, albeit artificial, would render an natural stereo soundstage, as Dr. Chouriri is stating.

Even if I had such processors readily available, my sample would not be large enough to guarantee a good level of confidence. I only have few CDs...

That's why I am asking everybody.

P.s: Perhaps I should have asked them not only to make available and optional function to not add crosstalk, but also an option to change a. the crosstalk level and b. the time the signal from the left virtual speakers arrives to the right headphone driver (and vice-versa). Then such option could be used with recordings in which the mixing relied only in level differences between L and R channels. So we could tweak the artificial panpoted soundstage, But that wouldn't be accurate and the mixing engineer may get angry with us.

 

[gregorio]

So I just cannot tell what percentage of an unbiased (representative) sample the universe of recorded music available has been deliberately mixed with ILD (panpot) and ITD so that, albeit artificial, would render an natural stereo soundstage, as Dr. Chouriri is stating.

 
I'm not really sure of the context of the statements you've quoted. But on the face of it, some/many appear to be nonsense. The percentage of popular music recordings deliberately mixed with both ILD and ITD is tiny. At a guess, less than 1% and probably a lot less! Popular music is always recorded as a collection of mono sound sources or of one or two stereo sources mixed with mono sound sources. The stereo-image is therefore constructed artificially and even some of those stereo sources are commonly artificial (stereo synth pads for example). So, virtually without exception, popular music has a stereo image which is an artificial construct and then the question becomes, how do we construct it? Well, it's a combination of tools, one of which is reverb. Some reverbs are mono in, stereo out, others are stereo in, stereo out. The latter potentially providing a reasonably natural/realistic ITD relative to the mixed L/R position of the of the source channel/s feeding the reverb, the former cannot. If we're talking about the L/R position of the individual source channels themselves though (rather than the reverb applied to those channels), then almost without exception that is accomplished purely with ILD (panning) and in fact, ITD is usually deliberately avoided, let alone realistic values calculated and applied! The deliberate use of ITD for panning (more commonly called "psycho-acoustic panning" by the audio engineering community) is generally avoided for a few reasons: 1. It's far more time and resource consuming to set-up initially and adjust later. 2. The mix is unlikely to have decent mono compatibility and 3. The resultant L/R position achieved by psycho-acoustic panning on a channel is very fragile/unreliable: A. Any subsequent application of any delay based effects (chourusing, doubling, DD or reverb for example) to that channel will almost certainly change or completely destroy the L/R position. B. It's far more sensitive (than ILD panning) to small changes/inaccuracies in speaker positioning, room acoustics and listener position. C. I can't even imagine trying to create a mix where all the L/R positioning is achieved by psycho-acoustically panning individual channels, I don't know how you'd avoid a complete mess.
 
The only exception I'm aware of is an old, rather obscure trick on those rare occasions where it's desired that the kick and/or bass guitar be positioned some place other than the centre or near centre and psycho-acoustic panning maybe employed to more evenly distribute the high energy levels between channels/speakers.
 
All the above relates to popular music recordings, as quoted, it's not necessarily true of classical recordings.
 
G

 

[jgazal]

(...) The deliberate use of ITD for panning (more commonly called "psycho-acoustic panning" by the audio engineering community) is generally avoided for a few reasons:

1. It's far more time and resource consuming to set-up initially and adjust later.

2. The mix is unlikely to have decent mono compatibility and

3. The resultant L/R position achieved by psycho-acoustic panning on a channel is very fragile/unreliable:

A. Any subsequent application of any delay based effects (chourusing, doubling, DD or reverb for example) to that channel will almost certainly change or completely destroy the L/R position.

B. It's far more sensitive (than ILD panning) to small changes/inaccuracies in speaker positioning, room acoustics and listener position.

C. I can't even imagine trying to create a mix where all the L/R positioning is achieved by psycho-acoustically panning individual channels, I don't know how you'd avoid a complete mess.

(...)

All the above relates to popular music recordings, as quoted, it's not necessarily true of classical recordings.

G

I would like to thank again for this post.

I wonder the percentage of the subgroup classical recordings that were made using the ORTF or Jecklin Disk microphone arrangements directly to tape without mixing/mastering.

What do think about those methods and what is your guess about the incidence of such recording in the classical subgroup?

Do you believe a mixing software that relies not only in ITD/ILD, but also in spectral cues, such as the theoretica.us - bacch-dsp, would render poor compatibility with mono and standard loudspeaker playback (in other word, without a crosstalk cancellation DSP)?

Do you believe that ambisonics and vector panning is a better option than "Binaural Audio through Two Loudspeakers (BA2L)" regarding such aspects?

 

[gregorio]

[1] I wonder the percentage of the subgroup classical recordings that were made using the ORTF or Jecklin Disk microphone arrangements directly to tape without mixing/mastering.

[2] What do think about those methods and what is your guess about the incidence of such recording in the classical subgroup?

[3] Do you believe a mixing software that relies not only in ITD/ILD, but also in spectral cues, such as the theoretica.us - bacch-dsp, would render poor compatibility with mono and standard loudspeaker playback (in other word, without a crosstalk cancellation DSP)? Do you believe that ambisonics and vector panning is a better option than "Binaural Audio through Two Loudspeakers (BA2L)" regarding such aspects?

 
1. Classical isn't really a sub-group as far as recording is concerned, it's more like a group of different sub-groups. Recording a large symphony orch is quite a different challenge to recording say a single unaccompanied soloist for example. Answering your question in terms of say orchestral recordings: Again, I should imagine very few or virtually none! From the digital age, it's been more usual to employ a Decca Tree or similar mic array plus some spot mics (on individual sections/sub-sections of the orch). Prior to this it was more common to just use a Decca Tree or even just a stereo pair but then we're talking about vinyl distribution where mastering was essential to counter some of vinyl's inherent weaknesses (applying the RIAA curve for example).
 
2. Both have their place, as do the other stereo mic techniques such as near co-incident, spaced and one of my favourites, an M/S pair. The best stereo mic technique theoretically is a Blumlein pair but it's very rarely used because it provides little control and has some practical use issues. That's the problem with all stereo mic techniques, each has it's own advantages and disadvantages and depending on exactly what one is recording, where it's being recorded and what the desired result is, the recording engineer will pick the one best suited for that particular circumstance. Again though, it's not typical to only use a stereo pair, even when we're talking about a small ensemble or soloist.
 
3. I can't answer these questions to be honest. I know nothing about the Bacch-dsp system and my experience of ambisonics is limited to occasionally receiving B-format files which I've decoded for use in a 5.1 theatrical mixes.
 
G

 

[Lex2]

It does require a lot of restraint when mixing with psycho-acoustic tools employing delay effects coupled with L/R phase inversions. I find that delays of less than 7 ms would usually be more than sufficient. Anything beyond that, well God help you. It does work but takes more time and practice. Gregorio is right in that most stereo images are artificial anyways achieved via the good old pan pot from mono sources. Some can be as inelegant as drums at 12 o'clock; guitars (rhythm and lead) at 9am and 3pm; and backing singers at 10am. Or go old school and pan vocals and guitars all the way to the right, with drums and bass on the left. No rules here. Or there are some rules but they're all meant to be broken.

 

[pinnahertz]

I wonder the percentage of the subgroup classical recordings that were made using the ORTF or Jecklin Disk microphone arrangements directly to tape without mixing/mastering.

That would be very hard to say. Certainly many use ORTF, but unlikely only that, there's usually a few spot mics at very least. There are many classical recordings mixed directly to stereo from quite a number of mics and arrays, though far fewer today with multitrack recording on site.  All the big classical labels have lots of recordings mixed on site at a performance venue, and in the early days of digital when there was no practical on-site multitrack, it all got mixed to stereo.  Telarc did this first, but so did all the big guys.  

What do think about those methods and what is your guess about the incidence of such recording in the classical subgroup?

I've always rather liked ORTF because it can provide at least some inherent ITD. But in terms of universal compatibility with all play situations, it kind of fails on its own (as do most stereo-pair-only techniques), but probably worse than coincident pairs. Blumlein works quite well for compatibility, though on its own the image ends up a bit flat. Same with M/S. It's all a compromise, and as I said above, hardly any orchestral recordings have been made with only one stereo pair of any kind. They are all supported by spots and even other pairs, like widely spaced omnis even.

Do you believe a mixing software that relies not only in ITD/ILD, but also in spectral cues, such as the theoretica.us - bacch-dsp, would render poor compatibility with mono and standard loudspeaker playback (in other word, without a crosstalk cancellation DSP)?

Unless the software takes the playback configuration into account, it will not be very compatible. And there's the problem. How do you produce a mix that is specifically optimized for a unique playback configuration but remains universally compatible? You can't. That's why there are post-processes that take the intended playback configuration into account along with the actual playback configuration. But those will only work well if both conditions are very fully and precisely characterized.  Have fun with that one.

Do you believe that ambisonics and vector panning is a better option than "Binaural Audio through Two Loudspeakers (BA2L)" regarding such aspects?

No experience with the BACCH system, can't compare. But logically, to properly and fully decode either one the playback environment and conditions must be accurately known and compensated for. The problem with these systems is they encompass at least two translations: 3d space encoded, and encoded signals decoded to 3d space. The decoded space will never match the encoded one physically or from the standpoint of transducer location, and there will always likely be a lack of full characterization of the playback system. The complete transcoding must therefore include compromises, and while that's always true in any stereo recording, at least with common stereo mixing techniques the mix monitoring environment, though controlled and idealized, is not a bad match to a well designed stereo playback system in a good listening room. In that case the "encoding", or mix is created on a system that has reasonable "match" to the playback system. This is not the case with Ambisonics, BACCH, binaural, or anything like that, at least not without full playback system analysis. Even though it is possible, that last bit, the full characterization of the play system, is what's missing from nearly every play situation. And the kicker is, what are you going to analyze? One seat with a head clamp? For the ultimate precision, you'd have to. But that's impractical, so it ends up being a compromise too.
 
If I may...from Toole...kind of sums it all up:
 

 

[jgazal]

(...) the playback environment and conditions must be accurately known and compensated for. (...) though controlled and idealized, is not a bad match to a well designed stereo playback system in a good listening room. In that case the "encoding", or mix is created on a system that has reasonable "match" to the playback system. This is not the case with Ambisonics, BACCH, binaural, or anything like that, at least not without full playback system analysis. Even though it is possible, that last bit, the full characterization of the play system, is what's missing from nearly every play situation. And the kicker is, what are you going to analyze? One seat with a head clamp? For the ultimate precision, you'd have to. But that's impractical, so it ends up being a compromise too.

 
Interesting that you say it is impractical to accurately know the playback enviroment.

I think we start with somewhat different premises:
 

(...)

So for me at least, even though the standards are such that you should have some reverb / avoid direct field from the surround channels in regular setup, I actually seem to prefer PRIRs taken in rather dry room (so I need the personalized xfeed but not so much the acoustic imprint of the room).

So to my idea: am wondering if it would make sense to get my own "PRIRs" (more like hrtfs) with the speaker rather close and stuffing the room walls to attenuate reflections as much as possible so that direct field dominates.

(...)

Then, I'd "simply" rotate the head to get the various headings, including elevation channels. The angle might not be accurate but, as long as the head is steady during the recording, it would be fine perhaps.

 


 
Now that you mentioned the circle of confusion, I would like to know your opinion with such somewhat "bypassed" playback environment as mentioned below:
 

Do you believe that such dsp engine (with convolution of a PRIR measured in a more and less dead room, no added acoustic - or virtual - crosstalk, decay room equalisation - in time domain -, causal headphone equalisation and tactile transducers, as described in the quotes above) can minimise the filtering effects that we are used to hear from the rest of the chain (mainly from playback room, playback amplifiers and playback transducers) compared to the current standard?

Please consider the spatial accuracy only in the horizontal plane (disregard then spectral cues). I did asked about the comparison between binaural and a third order ambisonics using such DSP engine, but apparently no one in head-fi except the creator has listened both.

 

I would also like to know what is your theoretic opinion about this:
 

Suppose you have your own HRTF measured with two stereo speakers in a low reverberation room (anechoic) and you set the Realiser to not add cross-talk at convolved output.
Do you think the elevation cues - filtered by the binaural head and torso microphone transfer function - only change the listener perception of elevation of a recorded point source (in other words, the listener understand that the source is above or under 0 degree, but the listener doesn't realize the true/original elevation of the recorded point source) or completely ruin the elevation perception (the listener do not hear the source as it were above or under 0 degrees elevation)?
Now suppose you have a HRTF measured with an sphere arrangement of sixteen speakers (eight at 0 elevation, 4 at +45 degrees and four at -45 degrees) in a low reverberation room (anechoic) and set the Realiser A16 to decode ambisonics b-format to a third order convolved output. Does this second arrangement improve the listener elevation perception compared to the first arrangement?
If you think the second arrangement is worst than the first arrangement, how many channels the second arrangement would need in order to achieve the perception performance of the first arrangement?
In other words, do you believe the playback of binaural stereo recordings with head tracking and personalized dynamic convolution without the addition of crosstalk has the same performance than playback of 16 channel ambisonics output with the same head tracking and personalized dynamic convolution playback? 
My criterion would be the number of errors a listener has comparing the the elevation he believes a point-source (a person speech for instance) is and the true/original n elevation positions the source was recorded.

 

 

[pinnahertz]

   
Interesting that you say it is impractical to accurately know the playback enviroment.

I think we start with somewhat different premises:
Originally Posted by arnaud View Post (...) So for me at least, even though the standards are such that you should have some reverb / avoid direct field from the surround channels in regular setup, I actually seem to prefer PRIRs taken in rather dry room (so I need the personalized xfeed but not so much the acoustic imprint of the room). So to my idea: am wondering if it would make sense to get my own "PRIRs" (more like hrtfs) with the speaker rather close and stuffing the room walls to attenuate reflections as much as possible so that direct field dominates. (...) Then, I'd "simply" rotate the head to get the various headings, including elevation channels. The angle might not be accurate but, as long as the head is steady during the recording, it would be fine perhaps.

Before we go on, I would like to know your definition of the word "practical". Thanks.

 
Now that you mentioned the circle of confusion, I would like to know your opinion with such somewhat "bypassed" playback environment as mentioned below:

Before we go on, please let me know how I (or anyone), producing and mixing a recording, has any way to know if the playback environment has indeed been bypassed, and by how many of my customers. Thanks again.

I would also like to know what is your theoretic opinion about this:
 
Suppose you have your own HRTF measured with two stereo speakers in a low reverberation room (anechoic) and you set the Realiser to not add cross-talk at convolved output. Do you think the elevation cues - filtered by the binaural head and torso microphone transfer function - only change the listener perception of elevation of a recorded point source (in other words, the listener understand that the source is above or under 0 degree, but the listener doesn't realize the true/original elevation of the recorded point source) or completely ruin the elevation perception (the listener do not hear the source as it were above or under 0 degrees elevation)?
 
Now suppose you have a HRTF measured with an sphere arrangement of sixteen speakers (eight at 0 elevation, 4 at +45 degrees and four at -45 degrees) in a low reverberation room (anechoic) and set the Realiser A16 to decode ambisonics b-format to a third order convolved output. Does this second arrangement improve the listener elevation perception compared to the first arrangement? If you think the second arrangement is worst than the first arrangement, how many channels the second arrangement would need in order to achieve the perception performance of the first arrangement?
 
In other words, do you believe the playback of binaural stereo recordings with head tracking and personalized dynamic convolution without the addition of crosstalk has the same performance than playback of 16 channel ambisonics output with the same head tracking and personalized dynamic convolution playback? My criterion would be the number of errors a listener has comparing the the elevation he believes a point-source (a person speech for instance) is and the true/original n elevation positions the source was recorded.
 

There's a very old principle applied to asking questions in a broadcast interview: If you ask multiple questions, you'll only get an answer to the last one.  
 
So, no.

 

[drtechno]

  If we define accuracy in audio playback as "how close something is to the original," then we have to ask "what is the original?" and "what are we comparing it to?" The original is a sound field experienced by a human being. The reproduction is a sound field experienced by a human being.
 
These two sound fields are never the same, as current audio technology does not reproduce 3-D sound fields. So suppose we have an original sound field A, and two different reproductions B & C. How do we decide which is closer to A? 
 
I claim that there is no way to do that without involving an individual human decision. For some people B is closer. For some people C is closer.
 
Therefore accuracy is subjective.

Accuracy isn't really subjected. Its the "sounds good" opinion the master engineer had to deliver for the record company. Because "correct" and "sounds good" are two different things they actually have to balance. Because they have to make it sound good across all devices, acurracy goes out the window the minute the signal sees a limiter or a compressor.
 
But as far as capturing, yes a lot of harmonics are getting lost. But on another note, they do tune out a lot of other frequencies so it will sound better in cheap consumer devices, but its because they have to lower the dynamic range as low as 4db to give it the "louder" illusion that the record company calls "good". I think this is one of the bad things that caused CDs to sound bad after 1996.
 
Yes digital can not capture all the things the analog recording devices did, but hopefully that might be no more, because at the end of the year, the 32bit converters will hit the pro market.

 

[jgazal]

Before we go on, I would like to know your definition of the word "practical". Thanks.

(...)

Before we go on, please let me know how I (or anyone), producing and mixing a recording, has any way to know if the playback environment has indeed been bypassed, and by how many of my customers. Thanks again.

You are right. Practical is an ambiguous word.

I have to define my premises better.
But before let me state this clearly: I am a consumer and I would never say to any engineer that he is currently doing something right or wrong.

I am not asking questions about what is currently correct, but about what would be viable in the near future, since my questions are all related to playback devices that are just hitting the market.

I am sure you all are fulfilling the needs of your clients, otherwise you wouldn't get paid.

Now to my understanding of practical.

Measurement of HRTF in anechoic chambers is definitely not practical.

Living in a room with very low reverberation is definitely not practical.

Measuring a binaural impulse response only once with only one professional monitor (that is specified to have high directivity and flat frequency response on axis in an anechoic chamber) in the near field while stuffing fist reflections on ceiling, floor and walls seems to me more practical than the first two options. Currently possible with Smyth Research Realiser.

Measuring a BRIR in your acoustically controlled mixing and mastering room seems as practical as the prior option.

How many of your costumers are going to have such controlled environment? I wouldn't mind at least listening in the same way you are.

So I agree with you, very few. You are right, you made your point.

But then how many of your customers have a listening room with the acoustical treatment your mixing and mastering room has?

Buying beamforming transducers arrays have just been demonstrated by Comhear (beamforming or yarra 3dx). I would expect the percentage of your costumers with access to that kind of playback environment to rise.

If that occurs, then ITD starts to become important.

Buying crosstalk cancellation DSP is currently expensive, but the playback environment is similar to the prior option.

Having head and torso scanned or photographed and a HRTF derived from such data is not currently possible. If it becomes available, then virtually all headphone listeners will have access to controlled externalized playback environment.

There's a very old principle applied to asking questions in a broadcast interview: If you ask multiple questions, you'll only get an answer to the last one.  

So, no.

If your answer to my last written question is no, then you believe or that third order ambisonics render elevation in a better way than binaural encoded files OR the other way around.

I appreciate your sincere anwee, but I still just do not know if a third order is enough.

With such uncertainty, ambisonics choice with convolved HRTF would at least work around the potential problem of matching the HRTF of the dummy head (or the standard HRTF of the mixing code) with the listener HRTF.

I would like to know if such matching is so critical or it just introduces a negligible tonal inaccuracy.

Cheers.

 

[pinnahertz]

You are right. Practical is an ambiguous word.

I have to define my premises better.
But before let me state this clearly: I am a consumer and I would never say to any engineer that he is currently doing something right or wrong.

Thanks, but you are nothing like the typical consumer.

I am not asking questions about what is currently correct, but about what would be viable in the near future, since my questions are all related to playback devices that are just hitting the market.

Unfortunately, we've had systems that have been able to measure and apply correction to consumer sound systems for almost 15 years. While millions of units have been sold the actual proper application of the auto-cal technology has been abysmal. And those are low, or percieved "no-cost" features included with units that would have been purchased without them. I use auto-cal as an example of something that offers significant improvement to the user experience but has not been widely accepted.  
 
Expecting the general application of something as advanced as the Realizer is just fantasy.

Now to my understanding of practical.

Measurement of HRTF in anechoic chambers is definitely not practical.

Living in a room with very low reverberation is definitely not practical.

Measuring a binaural impulse response only once with only one professional monitor (that is specified to have high directivity and flat frequency response on axis in an anechoic chamber) in the near field while stuffing fist reflections on ceiling, floor and walls seems to me more practical than the first two options. Currently possible with Smyth Research Realiser.

Measuring a BRIR in your acoustically controlled mixing and mastering room seems as practical as the prior option.

How many of your costumers are going to have such controlled environment? I wouldn't mind at least listening in the same way you are.

So I agree with you, very few. You are right, you made your point.

But then how many of your customers have a listening room with the acoustical treatment your mixing and mastering room has?

Translation of a mix done in a well designed studio to playback in a typical home is actually pretty good. Mostly what happens at home is listening outside the sweet spot (something that can be checked in the studio), and poor frequency response (again, easily simulated). There are compromises, but the correlation is actually fairly good, and at least known.

Buying beamforming transducers arrays have just been demonstrated by Comhear (beamforming or yarra 3dx). I would expect the percentage of your costumers with access to that kind of playback environment to rise.

Well, it's at zero now, so a 100% increase would be fairly simple to accomplish. But do you have any idea what "typical consumer" means? There will never be that kind of technology in the home of a "typical consumer".  

If that occurs, then ITD starts to become important.

It's important now, just not in the way you seem to think.  ITD can add, if done right, a sense of depth and space to any stereo mix, and the expense of mono compatibility. 

Buying crosstalk cancellation DSP is currently expensive, but the playback environment is similar to the prior option.

Crosstalk cancellation has been around for 37 years in various forms, nothing as involved as DSP, but some actually pretty astounding. It's never penetrated the market beyond the boutique level for the same reasons and problems we still have. We've also had two significant attempts at getting more than two-channel stereo into the home. First with Quad, which we thought failed because of format confusion, a tiny sweet spot, and the expense and inconvenience of 4 speakers.  But later, 5.1 music taught us that none of that was the problem. With 5.1 music we had standardized format, standardized speaker plans, a huge listening window, clear improvement over stereo, a huge pre-installed base of 5.1 systems (the extra speakers were already there for movie surround sound) and it still can't penetrate the market beyond the boutique.
 
Sorry if you don't see it this way, but DSP crosstalk cancellation systems will never be more in the market than at the boutique level, and as such cannot be considered in music production beyond the specialized demo material, etc.  An obvious parallel is binaural, again, around for 50 years.  Today there are billions of headphone listeners, more than ever.  But can you buy your favorite music in binaural today?  Not even a little.  It's still at the curiosity, and boutique stage, and not likely to change.

Having head and torso scanned or photographed and a HRTF derived from such data is not currently possible. If it becomes available, then virtually all headphone listeners will have access to controlled externalized playback environment.

I think your view of "virtually all headphone listeners" is way, way out of whack. You know what "every headphone listener" has access to now? Precision headphone-specific EQ developed from high resolution measurements. It's available on the most common headphone listening platform in the world, and is so cheap as to be a non issue.  Ever heard of that? Didn't think so.
 
"Have access to" and actually having something proliferate are two very different things.

If your answer to my last written question is no, then you believe or that third order ambisonics render elevation in a better way than binaural encoded files OR the other way around.

I didn't say that at all, did I? I believe they are two different things and don't present the same way.

With such uncertainty, ambisonics choice with convolved HRTF would at least work around the potential problem of matching the HRTF of the dummy head (or the standard HRTF of the mixing code) with the listener HRTF.

That's not a small problem, though, it's actually a deal-breaker.  And because it's hard to get it right without customizing, it severely limits the commercial application of the technology.  And yet, without a good listener HRTF in there somewhere, the whole thing just doesn't work very well.

I would like to know if such matching is so critical or it just introduces a negligible tonal inaccuracy.

I have not played much with Ambisonics, but from my binaural work, matching the HRTF to the listener is critical to the palpability of the resulting image. It's not so much a tonal issue as an imaging accuracy issue.  There's also an artistic problem too.  It isn't always desirable to have that full 3D sound field. 

 

[gregorio]

  [1] Because they have to make it sound good across all devices, acurracy goes out the window the minute the signal sees a limiter or a compressor.
 
[2] But as far as capturing, yes a lot of harmonics are getting lost. [2a] But on another note, they do tune out a lot of other frequencies so it will sound better in cheap consumer devices, but its because they have to lower the dynamic range as low as 4db to give it the "louder" illusion that the record company calls "good".
 
[3] Yes digital can not capture all the things the analog recording devices did, [3a] but hopefully that might be no more, because at the end of the year, the 32bit converters will hit the pro market.

 
1. Accuracy goes out the window long before the signal sees a limiter or compressor. In the case of pop music, accuracy is the last thing we want! We don't want the listener to be aware that certain parts of the lead vocals were overdubbed days later, that the drums were recorded at a different time/place and what they really sounded like, plus a whole bunch of other things. With classical/acoustic music genres we usually don't want exact accuracy either, we don't want them to be aware of any edits/overdubs, of all the audience noise and usually of the actual sound waves rather than the "experience". So accuracy goes out of the window from the point of mic choice/position and the moment the "rec" button is pressed.
 
2. Not sure what you mean by "a lot of harmonics are getting lost"? Compared to what, to what the instruments are actually producing or compared to what the signal an audience member would hear? If it's the former, then some rather than "a lot" might be lost and if it's the latter then very little would generally be lost.
2a. With popular music genres this is true although even the most crushed ones have more than 4dB dynamic range. Far less compensation is applied for cheap consumer devices as far as the vast majority of classical/acoustic music is concerned and the dynamic range is never reduced anywhere near the amount of popular music.
 
3. No! Digital can absolutely capture "all the things the analogue recording devices did", far more in fact! Analogue recording devices typically add to what was captured (noise, tape saturation, etc.) and also typically distort what they did capture, particularly in the HF range. Even the very best analogue recording devices were not able to capture "all the things" as accurately a digital.
3a. Won't make any difference whatsoever. 32bit converters cannot capture any more than 24bit converters, no commercial converters capture more than about 20-21 bits at best anyway and even 16bit converters capture far more than any analogue recording device ever could!
 
G

 

[pinnahertz]

 
But as far as capturing, yes a lot of harmonics are getting lost.

Clearly you've never actually seen the frequency response of a tape recorder, or the real spectrum of the original audio. No, digital doesn't miss capturing harmonics. Analog actually did a far worse job.

But on another note, they do tune out a lot of other frequencies so it will sound better in cheap consumer devices, but its because they have to lower the dynamic range as low as 4db to give it the "louder" illusion that the record company calls "good".

You've confused two entirely different processes. No, "they" don't "tune out" anything for cheap consumer devices, yes, it is processed to sound louder...but that's not why.

I think this is one of the bad things that caused CDs to sound bad after 1996.  

I own a lot of post 1996 CDs that sound fantastic. You seem to think they all sound bad...wrong-0.

Yes digital can not capture all the things the analog recording devices did, but hopefully that might be no more, because at the end of the year, the 32bit converters will hit the pro market.

This is almost comic. I spent many decades with analog recorders, trying to squeeze the last dB of quality out of them. Compared with any professional (and some semi-pro) digital recorder, every single analog recorder was inaccurate, of limited bandwidth, highly distorted, noisy, had an unstable time-base, and on and on.
 
32 bit converters? Funny, since we don't even have 24 bit converters with actual 24 bit performance. Only one in the world comes even close. I believe you've been seriously misinformed...about a great many things.