An Enticing Marketing Story, Theory Without Measurement?

[Juhazi]

^Link to McGill Uni http://sites.music.mcgill.ca/vat/virtualacoustics/

 

[j_j]

It is a triumph of measurement over logic. It is based on a vague, unstated assumption: that humans are purely frequency content analysers and that they can't distinguish between direct sound, reflections and reverberation; so as long as the contents of an FFT window more-or-less give the right proportions of 'frequency response stuff', it will sound perfect.

Uh, no. Not even the room correction buried deep inside your windows 10 box is that simplistic. Furthermore, the goal of room correction has several components.

1) reduce level in places where the room stores too much energy at low frequencies.
2) make the DIRECT signal accurate at medium/high frequencies (let's not talk "schroeder frequency" here, that's a different issue).

And that is accommodated even in the W10 method.

Logic and common experience tells us that this cannot be true, otherwise we would not be able to hear the directions of sound sources in reverberant environments, and people's voices would sound different as we walked from outside to inside.

Again, multiply wrong. First, even if one did correct the "whole signal" the time domain cues would remain in the reverberent environment, and localization would proceed apace. You're ignoring the very first basis of the ear's function, which is as a time/frequency analyzer. No room correction algorithm I've studied even comes remotely close to the kind of behavior you seem to think room correction is designed to do.

As to people's voices sounding different, actually, they do, as they naturally should, as we move and they move, and that part of adaptation is learned at a young age. You don't notice the difference because it's expected. You only notice when it doesn't happen. Wandering around an anechoic room with some big flat panels blocking direct sound will make this clear apace.

We have evolved, then learned all our lives, to distinguish between a sound source and reflections (a.k.a. 'the room'); we do it expertly and without conscious thought. In our hearing we are not eliminating the room, but we are distinguishing between the source and the room. If we change the sound source to fit some notional 'target curve' it will sound 'changed' but the room will sound just the same! - the opposite of what is intended.

Again you reach this conclusion because you fail to consider the time domain issues in hearing. It's just wrong, and proper "room correction" does what's intended, and not the opposite. Your conclusion is counter to fact, and is based on many faulty assumptions and a huge leap of faith rather than proper logic.

Confusion arises because real speakers are often not very neutral sources, with odd directivity at different frequencies. This cannot be corrected, but it can be compensated for to make the speaker sound subjectively better. The room cannot be corrected, but the speaker in that room in that particular placement may be improved subjectively if there is some understanding of what is going on. Calling it 'room correction' is an indication that that understanding is absent...

And, to some extent, true. Perhaps "source correction" would be better, because in position, speaker level, time delay can be corrected to a very, very high-accuracy extent. Correcting for speakers with moderately different amplitude and phase response, likewise. So, no, let's not quite say that the "room can not be corrected", some low-frequency reflections can be mitigated, after all, and that is often a thing of great value. Speaker distance can be fixed to an astonishing degree, as can differences between speakers, including those due to acoustics of reflective objects, walls, etc, near the speaker.

"Room Correction" is not completely right, but a good room, corrected, will always be better than a good room, left alone.

 

[svart-hvitt]

Uh, no. Not even the room correction buried deep inside your windows 10 box is that simplistic. Furthermore, the goal of room correction has several components.

1) reduce level in places where the room stores too much energy at low frequencies.
2) make the DIRECT signal accurate at medium/high frequencies (let's not talk "schroeder frequency" here, that's a different issue).

And that is accommodated even in the W10 method.


Again, multiply wrong. First, even if one did correct the "whole signal" the time domain cues would remain in the reverberent environment, and localization would proceed apace. You're ignoring the very first basis of the ear's function, which is as a time/frequency analyzer. No room correction algorithm I've studied even comes remotely close to the kind of behavior you seem to think room correction is designed to do.

As to people's voices sounding different, actually, they do, as they naturally should, as we move and they move, and that part of adaptation is learned at a young age. You don't notice the difference because it's expected. You only notice when it doesn't happen. Wandering around an anechoic room with some big flat panels blocking direct sound will make this clear apace.


Again you reach this conclusion because you fail to consider the time domain issues in hearing. It's just wrong, and proper "room correction" does what's intended, and not the opposite. Your conclusion is counter to fact, and is based on many faulty assumptions and a huge leap of faith rather than proper logic.


And, to some extent, true. Perhaps "source correction" would be better, because in position, speaker level, time delay can be corrected to a very, very high-accuracy extent. Correcting for speakers with moderately different amplitude and phase response, likewise. So, no, let's not quite say that the "room can not be corrected", some low-frequency reflections can be mitigated, after all, and that is often a thing of great value. Speaker distance can be fixed to an astonishing degree, as can differences between speakers, including those due to acoustics of reflective objects, walls, etc, near the speaker.

"Room Correction" is not completely right, but a good room, corrected, will always be better than a good room, left alone.

@j_j , thanks for your comment to @Cosmik ’s ideas.

You wrote:

«Room Correction" is not completely right, but a good room, corrected, will always be better than a good room, left alone».

That made me curious. Do you speak of your own experience or a broader set of research efforts?

And am I wrong if I interpret you as having a more open stance on digital room compensation than @Floyd Toole ?

 

[j_j]

@j_j , thanks for your comment to @Cosmik ’s ideas.

You wrote:

«Room Correction" is not completely right, but a good room, corrected, will always be better than a good room, left alone».

That made me curious. Do you speak of your own experience or a broader set of research efforts?

And am I wrong if I interpret you as having a more open stance on digital room compensation than @Floyd Toole ?

It's complicated. We don't use room correction. We use digitally corrected speakers, in a near-anechoic space. But that's what we need for what we are working on.

In an average room, "room correction" can mitigate some of the worst problems.

In a good room, it can fix floor and wall problems that remain, and do some improvement. No, I have not written a paper on this, if you have the funding, I could propose such a test.

I'm not sure if I disagree with Floyd or not. He's right that you can't fix egregiously bad stuff. He's also right that you want to fix the direct sound, and that speakers should be both power-flat and direct-flat if you can manage it, which is an interesting problem at low frequencies that requires that the room match the speakers and their effective directivity at low frequencies.

What I think you CAN do is fix mismatches between speakers and rooms, if the speaker radiates energy properly (with no directivity at low frequencies, of course) and the room isn't egregious.

 

[DDF]

He's also right that you want to fix the direct sound, and that speakers should be both power-flat and direct-flat if you can manage it, which is an interesting problem at low frequencies that requires that the room match the speakers and their effective directivity at low frequencies.

Small correction not detracting from jj's coherent argument:
Harman doesn't advocate constant directivity, but rather a flat on axis and listening window with a DI showing a smooth positive tilt with no obvious shenanigans off axis. Pg 13:
https://www.harman.com/sites/default/files/AudioScience_0.pdf

 

[Duke]

...which is an interesting problem at low frequencies that requires that the room match the speakers and their effective directivity at low frequencies.

Can you explain what you mean by "the room match[ing] the speakers and their effective directivity at low frequencies"?

Thanks!

 

[Cosmik]

You're ignoring the very first basis of the ear's function, which is as a time/frequency analyzer.
...
Again you reach this conclusion because you fail to consider the time domain issues in hearing.

Of all the things to accuse me of!

I'm practically the only person here who talks about human hearing as a time/frequency analyser and who is not completely taken in by the frequency domain-only way of thinking. Just check back on my posts - I drive everyone mad by continuously pointing out the time domain as well as the frequency domain.

For example, I have only just been discussing the issue of the inverted backwave from open baffle speakers. I am the only person who points out that it is not a good idea for time domain reasons.

(Something tells me you'll find a way to disagree with me on that point, though...)

 

[Floyd Toole]

Hi JJ. Just an addition.

You said: "the ear's function, which is as a time/frequency analyzer". That is a good simple description, but humans have two ears - i.e. binaural hearing - which are active full time thereby adding : direction, spaciousness, envelopment, binaural discrimination, precedence effect, and more to the list of auditory capabilities. Thus my simple description of one-mic room EQ (with or without spatial averaging) is that a mic and an analyzer are not the functional equivalents of two ears and a brain. A dummy head with HRTFs is also insufficient because it is a dummy - no brain . My book points to evidence that perception of spatial elements seems to rank with perception of sound quality, timbral accuracy. This complicates things if one is seeking the ultimate listening experience because most of the easily perceived spatial information is in the recordings and they can be very different. Naturally, this varies with playback format: spatially-and-directionally-deprived stereo vs. have-what-you-want multichannel.

One can of course imagine elaborate mic arrays and signal processing that could isolate direct and reflected sounds, including their angles of incidence, but the brain part, and its adaptive and cognitive contributions are far from being simulated. The challenge is to embed within the EQ system the intelligence to ignore measured aberrations that are not audible problems and focus on those that are, and that includes the ability to address or ignore the effects of reflected energy arriving from different directions, at different times and amplitudes. The fact that the audibility of delayed sounds varies enormously with the temporal structure of the sound is a separate problem overlaying it all (Figure 7.29), and music and movies are ever changing.

As for what I think room EQ can and cannot address, some of it is in: Toole, F. E. (2015). “The Measurement and Calibration of Sound Reproducing Systems”, J. Audio Eng. Soc., vol. 63, pp.512-541. This is an open-access paper available to non-members at www.aes.org http://www.aes.org/e-lib/browse.cfm?elib=17839 And much more of it is in my book. The reality I see is one in which only high resolution (e.g.1/20-oct) anechoic measurements (the spinorama consists of a database of 70 curves) can adequately reveal the timbral accuracy, lack of resonances, and directional properties of the sound source. If one starts with a timbrally neutral speaker with well-behaved directional properties the rest is made much, much easier. Listeners can adapt to, and to a significant extent "listen through" live performance venues and playback venues. There are limits, of course, and it is up to research to reveal the issues and the magnitudes above which intervention is required. We have some of the answers. In many cases only physical acoustic treatments will suffice.

If one has no expectations about the loudspeaker - potentially anything from junk to jewels - it is asking a lot of an in-situ system to find the problems with the source, fix them and go on to make the rest of the setup adjustments for the loudspeakers in the playback arrangement: signal level, time of arrival, direction of arrival, etc. - the easy parts. It is the audible medium- and high-Q resonant misbehavior that in-room measurements have difficulty finding and measuring with sufficient accuracy to correct. A windowed on-axis curve is not enough. Strong frequency-dependent directivity is something that EQ may be able to somewhat address depending on frequency, but cannot correct. One needs a better loudspeaker.

With bass performance accounting for about 30% of our overall factor weighting in sound quality assessments there is work for EQ at low frequencies - at least for the prime listening location. The fundamental problem is that all bass sounds are propagated through a three-dimensional acoustically resonant chamber - the room. There is no dominant "direct" sound in the normal sense because at all resonance frequencies the energy builds at a rate determined by the Q, and correspondingly decays. This behavior is different at every location in the room, meaning that multiple listeners do not share the same bass experience. To address the needs of multiple listeners multiple subs are powerful assets in attenuating room resonances and thereby reducing seat-to-seat variations. With signal processing in the signal paths to each of the multiple subs room modes can be made to almost disappear, certainly pushed well below thresholds of detection (e.g. Harman's Sound Field Management). Section 8.3 describes elaborate research on this topic, one finding of which was that active multiple sub solutions were better than necessary at attenuating room resonances - a nice result. Because humans tend to ignore ringing - now there was a surprise - even relatively crude frequency response smoothing at bass frequencies can be greatly beneficial.

All said, discussions of this kind are good - we need to combat the truly uninformed stream of verbiage floating through audio forums. Cheers.

 

[RayDunzl]

Ear/Brain etc...

I was surprised when I created a file with 400Hz in the left channel and 405Hz in the right channel, put on headphones, and heard the 5Hz beat, which didn't exist in either ear, nor in the air.

 

[Floyd Toole]

Ear/Brain etc...

I was surprised when I created a file with 400Hz in the left channel and 405Hz in the right channel, put on headphones, and heard the 5Hz beat, which didn't exist in either ear, nor in the air.

Yes, this is the binaural hearing system in action - they are called binaural beats, essentially the 'phantom image' being swept rapidly between the ears.

 

[j_j]

Hi JJ. Just an addition.

You said: "the ear's function, which is as a time/frequency analyzer". That is a good simple description, but humans have two ears - i.e. binaural hearing - which are active full time thereby adding : direction, spaciousness, envelopment, binaural discrimination, precedence effect, and more to the list of auditory capabilities.

My goodness, that falls directly out of the time-delay in the compression in the ear. Precedence effect to a first order approximation is a direct result.

This also then leads directly to BMLD and all of the effects that arise from the emphases of signal-lock at low frequencies and leading-edge lock on the envelope at high frequencies.

So, I'm not sure quite why we're having this discussion, it's kind of obvious, isn't it? From BMLD, time arrival, and level, the rest of the properties come about very nicely by looking at properly derived partial loudnesses as a function of time. This also exposes a serious problem with spatial issues in block-oriented coding systems, but that's not the subject here.

 

[j_j]

Can you explain what you mean by "the room match[ing] the speakers and their effective directivity at low frequencies"?

Thanks!

The interaction between the DI of the speaker and the resonances of the room can be "interesting". Having the room have a peak where the power response of the speaker rises vigorously as a result of the DI changing can be "interesting". Not necessarily in a good way, mind you.

 

[Guermantes]

I understand that part of @Cosmik's criticism of room correction EQ is that he views it as a case of theory-following-practice and that this has resulted in proponents looking for justifications for what may be misconceptions.

What is its history? I tended to believe that it was a domestic application of the use of graphic equalisation in live sound reinforcement -- the tuning of a system by ear to mitigate problems such as resonances and microphone feedback, etc. Thus we saw the fashion of home hi-fi graphic equalisers with so few bands that really they were little more than somewhat technical-looking tone controls. Was that based on theory or the idea that what the professionals use must be good for the home, too? Some trickle-down from professional audio practice to the dosmetic listening enviroment is natural but some is misguided (e.g. "they use an external master clock so I need one, too").

Of course, DRC is now much more sophisticated and much theory and measurement has gone towards explaining its benefits. I like to believe that as the theory develops, the misconceptions will atrophy and best practice will be validated.

 

[j_j]

The challenge is to embed within the EQ system the intelligence to ignore measured aberrations that are not audible problems and focus on those that are, and that includes the ability to address or ignore the effects of reflected energy arriving from different directions, at different times and amplitudes. The fact that the audibility of delayed sounds varies enormously with the temporal structure of the sound is a separate problem overlaying it all (Figure 7.29), and music and movies are ever changing.

Pretty much exactly. There is another gotcha, in that corrections higher in frequency limit the space the correction applies to, if they are too detailed.

But that is the dual of "not audible problems", of course. This is where looking at the time/frequency response of the ear is absolutely critical.

Fortunately the two are not terribly contradictory. They can be in the 300-800 Hz range. We luck out on the time/envelope confusion around 1kHz to 2kHz, and above that, early arrival saves us.

And much more of it is in my book. The reality I see is one in which only high resolution (e.g.1/20-oct) anechoic measurements (the spinorama consists of a database of 70 curves) can adequately reveal the timbral accuracy, lack of resonances, and directional properties of the sound source.

Personally, I would suggest 1/3 ERB spacing. This results in 90 or so bands.

If one starts with a timbrally neutral speaker with well-behaved directional properties the rest is made much, much easier.

I would say "the rest is made possible". If a speaker has no near-zeros, and has a smooth DI, yes. This is where high-power crossovers with phase shift (yes, corrected, etc, assume "done right" here) can be problematical due to the phase response around the crossover point. Still, it is possible to not do this too poorly, I agree.

Listeners can adapt to, and to a significant extent "listen through" live performance venues and playback venues. There are limits, of course, and it is up to research to reveal the issues and the magnitudes above which intervention is required. We have some of the answers. In many cases only physical acoustic treatments will suffice.

Which is why I say "room correction" can be beneficial to decent or good rooms. If a room stores too much energy, it stores too much energy. Adding more doesn't help (to correct dips at the listening position), and remove energy (again to fix peaks at the listening position) does (*&(& all for the time hangover.

There is no dominant "direct" sound in the normal sense because at all resonance frequencies the energy builds at a rate determined by the Q, and correspondingly decays.

One can separate out such analytically, with special stimulii and a fair amount of work, but as far as the ear is concerned, that's about it.

But again, analyzing with the ear's time/frequency response in mind is a great start.[/quote][/quote]

 

[j_j]

Of all the things to accuse me of!

I'm practically the only person here who talks about human hearing as a time/frequency analyser and who is not completely taken in by the frequency domain-only way of thinking. Just check back on my posts - I drive everyone mad by continuously pointing out the time domain as well as the frequency domain.

For example, I have only just been discussing the issue of the inverted backwave from open baffle speakers. I am the only person who points out that it is not a good idea for time domain reasons.

(Something tells me you'll find a way to disagree with me on that point, though...)

Well, your assertion of "FFT" without having varying window with frequency, etc, made it seem that way.

But I'm not looking for an argument, and there can be value to "room correction" systems, but you can't fix a turd. I'll agree with that.

 

[j_j]

I understand that part of @Cosmik's criticism of room correction EQ is that he views it as a case of theory-following-practice and that this has resulted in proponents looking for justifications for what may be misconceptions.

What is its history? I tended to believe that it was a domestic application of the use of graphic equalisation in live sound reinforcement -- the tuning of a system by ear to mitigate problems such as resonances and microphone feedback, etc. Thus we saw the fashion of home hi-fi graphic equalisers with so few bands that really they were little more than somewhat technical-looking tone controls. Was that based on theory or the idea that what the professionals use must be good for the home, too? Some trickle-down from professional audio practice to the dosmetic listening enviroment is natural but some is misguided (e.g. "they use an external master clock so I need one, too").

Of course, DRC is now much more sophisticated and much theory and measurement has gone towards explaining its benefits. I like to believe that as the theory develops, the misconceptions will atrophy and best practice will be validated.

Well, for sure, the use of single-length FFT frequency response calculation, and correction based on that, breaks a whole bunch of things, from the actual capture of the signal (confusing direct with reflected, and all the resulting HF disasters), to the need for "accurate" correction at high frequencies, well,l the list is endless.

All I can say to some of the attempts is "it's not that simple".

 

[j_j]

Yes, this is the binaural hearing system in action - they are called binaural beats, essentially the 'phantom image' being swept rapidly between the ears.

Status Quo, 1960's? But yes, this is a classic example.

Another one is playing a constant, reasonable level 50Hz tone in both ears, then slowly vary the phase (I mean SLOWLY) on one ear. No, no directionality, but surprise, that's a spatial effect for sure.

 

[Floyd Toole]

The interaction between the DI of the speaker and the resonances of the room can be "interesting". Having the room have a peak where the power response of the speaker rises vigorously as a result of the DI changing can be "interesting". Not necessarily in a good way, mind you.

Fortunately the world has decided that omni woofers are to be the norm. Even most "full bandwidth" dipole speakers transition to monopole bass drivers. I think dipole subwoofers have joined the dodo bird, as they should. There are a couple of cardioid woofers out there somewhere I think.

It is a real world problem. Around 1990 I had my first interaction with noted subjectivist Harry Pearson (RIP) when I was to be called to help him with a problem. He had heard a speaker he thought he liked and wanted to review it. When he got the speakers he placed them in his carefully selected "reference" locations and seated in his "reference" seat something was wrong with the bass. The designer of the speaker was called to find out what was wrong. I was invited to join him in what was likely to be an interesting venture. We arrived at Harry's house in the AM, and without listening to a note of music, we got out a tape measure and a calculator. He looked on in stunned amazement. Then we asked if the problem was a deficiency of mid-bass. He said "yes, how did you know?". It turned out that his "reference" location was arrived at while listening to full bandwidth dipoles, and the speaker he was auditioning was a monopole. We moved both the speaker and his chair and all was well. Dipoles couple maximally at a pressure minimum/velocity maximum and monopoles couple maximally at pressure maxima/velocity minima. Dipoles have the additional complication of being vectored sound sources, meaning that the orientation relative to a room-mode null matters. Monopoles are not. There are good reasons to use monopole woofers and subs. He remained in a kind of trance, exclaiming that he had never experienced anything like that. For pure subjectivists science, even really basic science, is a great mystery and/or threat.

Any woofer or subwoofer I have ever encountered does not change its power response "vigorously" - they are minimum-phase systems that are quite well behaved. However, room modes/standing waves do change dramatically with location of the ears or mic. That is the problem to be addressed. Mode cancelling/attenuation using multiple subs greatly simplifies the situation, but only when the budget allows. Good news is that with multiple subs the total system efficiency rises, so they can be smaller subs.

 

[Floyd Toole]

Pretty much exactly. There is another gotcha, in that corrections higher in frequency limit the space the correction applies to, if they are too detailed.

But that is the dual of "not audible problems", of course. This is where looking at the time/frequency response of the ear is absolutely critical.

Fortunately the two are not terribly contradictory. They can be in the 300-800 Hz range. We luck out on the time/envelope confusion around 1kHz to 2kHz, and above that, early arrival saves us.



Personally, I would suggest 1/3 ERB spacing. This results in 90 or so bands.


I would say "the rest is made possible". If a speaker has no near-zeros, and has a smooth DI, yes. This is where high-power crossovers with phase shift (yes, corrected, etc, assume "done right" here) can be problematical due to the phase response around the crossover point. Still, it is possible to not do this too poorly, I agree.


Which is why I say "room correction" can be beneficial to decent or good rooms. If a room stores too much energy, it stores too much energy. Adding more doesn't help (to correct dips at the listening position), and remove energy (again to fix peaks at the listening position) does (*&(& all for the time hangover.



One can separate out such analytically, with special stimulii and a fair amount of work, but as far as the ear is concerned, that's about it.

But again, analyzing with the ear's time/frequency response in mind is a great start.

[/quote][/QUOTE]

Toole: And much more of it is in my book. The reality I see is one in which only high resolution (e.g.1/20-oct) anechoic measurements (the spinorama consists of a database of 70 curves) can adequately reveal the timbral accuracy, lack of resonances, and directional properties of the sound source.
JJ: Personally, I would suggest 1/3 ERB spacing. This results in 90 or so bands.

The ERB/critical band is not the resolution of the hearing system as it relates to all things. It is better described as being related to the discrimination of tones and how multiple tonal components interact with each other, thereby entering the domain of timbre. Roederer says: "It [the critical band] plays an important role in the perception of tone quality and provides the basis for a theory of consonance and dissonance of musical intervals." In short, if there are frequency response variations within a critical band/ERB there will be differences in musical timbre, which is exactly what Sean Olive and I found in our investigation of the audibility of resonances, some of which were much narrower than the critical band/ERB. In order to see with ones eyes the measured evidence of clearly audible and identifiable narrow-band/high-Q resonances one needs much higher resolution in the frequency domain than your suggestion allows. One might debate the 1/3 ERB spacing, but in this technical age why do we invoke such an imprecise, subjectively determined, metric. These days it cost nothing to have whatever resolution one wants.

Brian Moore, the author of ERBs, said to me: "The auditory filter bandwidths (ERBn values) are about 1/6 to 1/4 octave at medium to high frequencies. And . . . within band irregularities in response can have perceptual effects". It is quoted in my book.

All of this is discussed in Section 4.6.5 in the 3rd edition.

 

[j_j]

Any woofer or subwoofer I have ever encountered does not change its power response "vigorously" - they are minimum-phase systems that are quite well behaved. However, room modes/standing waves do change dramatically with location of the ears or mic. That is the problem to be addressed. Mode cancelling/attenuation using multiple subs greatly simplifies the situation, but only when the budget allows. Good news is that with multiple subs the total system efficiency rises, so they can be smaller subs.

Actually, I have to disagree. Even though they are monopoles, effectively, there can be some very weird things I've encountered in highly tuned (passive radiator, usually) systems. Actually, I doubt they are minimum phase, although they probably should be inside their actual bandwidth.

And a system that goes to 120 degrees between front and back radiation is not exactly a monopole, either. This can put an interesting shape into the power response (i.e. it does not rise monotonically as one goes down in frequency (of course, staying above cutoff).