Dutch & Dutch 8Cs

[Cosmik]

Hahahahaa

I'm still not sure I understand what you're trying to say in concrete terms though. I don't understand what your "real explanation" is.

It seems to me that you agree with the conventional wisdom that the direct sound and the sound field should not be conflated above Schroeder. Correct?

But it seems that you disagree with the conventional wisdom that the two should be conflated below Schroeder? I.e. you disagree with the idea that EQ should be used to smooth the steady-state response below the Schroeder frequency. Correct?

And the other thing I don't understand about your position: Do you believe that both the direct sound and the sound field should measure flat? Or that only the direct sound should be taken into consideration, i.e. the sound field is not of any interest?

Or are you saying something else?

I do try my best to make my position clear. I can't believe it's not obvious!

Let's try a different angle for anyone who is baffled by what is being said.

Let's be very precise. Basic experiments provide an observation that listeners to speakers in a room prefer a sound that, when measured with a microphone and FFT with a fixed-duration time window, yields a drooping frequency response. That is it. That is all that has been observed, and it is explanation-, theory- and understanding-free.

As has been seen, the author of a book on DSP, no less, extrapolates from this that people's hearing and brains prefer a drooping frequency response. From this can be extrapolated that EQ can/should be used to modify the sound from speakers in a room (any speaker, any room) to give this drooping frequency response. It is a 'target'. If the speaker happens to give this in-room curve without adjustment, this must mean that the speaker is a good one - although this view may be tarnished if it turns out that it doesn't do the same thing in different room...

Other people, upon intense discussion, make a distinction between the near field and the far field - but they provide no explanation for it; it is merely a (correct) observation. They do not register the idea of the listener hearing the speaker and room separately; they are still fixated on a single stream of 'audio soup' with a frequency response. Now, such people may not be quite so ready to accept that the drooping response should be a target, but they don't do anything to contradict it.

They don't have any method for saying what the response should be at 1', 2', 5', 10' from the speaker or at different angles. Their only method would be to make laborious observations with listening tests and multiple listeners at multiple distances and angles from the speaker. After many tests they might be in a position to create a set of tables of typical preferred frequency responses at different distances. These observations could even be correct for that room and speaker. Interpolation might be used to infer the preferred responses in between.

Now for the technology! Logically, a speaker system could be devised that used a clever range-finding system to determine the location of the listener in the room and adjusted the EQ to preset settings in order to always provide the scientifically-preferred EQ setting. Mysteriously, it might be found that the speaker doesn't yield the same curves in every room, but short of performing listening tests on the client, the speaker would be calibrated by carrying a microphone around the room and setting it up to provide the right measurements at all locations in that room. Result! Science!

What is wrong here? Pretty much everything, even though it could be argued it is a logical conclusion to draw from the original scientific observation.

What is the right answer? Well, an explanation or idea is like an engine: it has power, promotes progress, economy, elegance, performance, correctness. In this case, the explanation is: the speaker should have a flat response anechoically and have uniform dispersion because the listener perceives the direct sound of the speaker separate from the room. That is it. (That is my position @andreasmaaan!)

From that flows the simple design and performance of the Kii Three/D&D 8C, the great reviews, the observations of their neutral sound regardless of location, the lack of need for EQ settings. And also flows the lack of necessity for a listener tracker and dynamic EQ system, the lack of the need for listening tests, the lack of the need for in-room measurements. And possibly why hi-fi in the 1970s sounded better than typical audiophile systems now - wider baffles, sealed drivers, speakers set up in anechoic chambers independent of room, etc.

Is there something wrong with science then? No. Good science starts from a hypothesis, which should eventually turn into the explanation described above. But audiophile 'science' doesn't try to make the hypothesis an explanation. It just makes it a true/false criterion for the observation, and is only interested in differences or preferences while listening to music. It cannot possibly lead to understanding or a productive explanation, and results in dead ends and rabbit holes such as 'room correction' and 'target curves'.

 

[andreasmaaan]

What is the right answer? Well, an explanation or idea is like an engine: it generates progress, economy, elegance, performance, correctness. In this case, the explanation is: the speaker should have a flat response anechoically and have uniform dispersion because the listener perceives the direct sound of the speaker separate from the room. That is it. (That is my position @andreasmaaan!)

Ok, but I still feel that what you're saying is unclear (sorry!).

It seems to me there's a contradiction between your view that the speaker should have uniform dispersion on one hand, and your view that "the listener perceives the direct sound of the speaker separate from the room. That is it."

If that really is it, then it follows that the dispersion characteristic of the speaker is irrelevant. But in the same breath you are saying that the dispersion is fundamentally important and that there is one particular way that it should be (namely uniform).

This is the aspect of what you're saying that I'm still very confused about.

And finally, I'm still unclear as to whether you believe in corrective EQ below the Schroeder frequency, which is what I'd consider to be the "conventional wisdom" as it is the position taken by the top authority in these matters, Floyd Toole (whether you agree with him or not).

So I'm interested in what your view is on that?

PS I'm asking all these questions because I find what you're saying interesting and I want to get to the bottom of it. Not trying to be argumentative here, which I hope is clear.

 

[Cosmik]

Ok, but I still feel that what you're saying is unclear (sorry!).

It seems to me there's a contradiction between your view that the speaker should have uniform dispersion on one hand, and your view that "the listener perceives the direct sound of the speaker separate from the room. That is it."

If that really is it, then it follows that the dispersion characteristic of the speaker is irrelevant. But in the same breath you are saying that the dispersion is fundamentally important and that there is one particular way that it should be (namely uniform).

This is the aspect of what you're saying that I'm still very confused about.

And finally, I'm still unclear as to whether you believe in corrective EQ below the Schroeder frequency, which is what I'd consider to be the "conventional wisdom" as it is the position taken by the top authority in these matters, Floyd Toole (whether you agree with him or not).

So I'm interested in what your view is on that?

Ah, I see. Your misunderstanding comes from assuming that when I say "separate from the room", I am meaning "discarding the room" - I do not mean this. Again, this stems from the 'audio soup' notion and the implicit idea that the ideal speaker 'projects' dry sound directly to the listener.

To quote Dr. Toole, the listener is hearing separate 'streams' for the direct sound and the room. But 'the room' is not an independent source: it is derived from the speaker's sound. Even if the direct sound is correct, 'the room' will be 'incorrect' if it is not derived from a speaker with uniform dispersion. Both have to be consistent for it to sound right.

As regards Schroeder, obviously I am saying that I want no 'correction' above Schroeder. Unlike some other people, however, I am fairly sure I don't want to mess with the bass either! I, personally, keep the anechoically-measured bass flat except for a (not particularly carefully) chosen roll off frequency and -12dB per octave. For those with bass reflex, timing misalignments, phase errors and so on, I wish you the best of luck because maybe you need to do more messing about with it. Personally, I have the notion that I want the room's effects in the bass to be consistent with its effects higher up.

 

[andreasmaaan]

To quote Dr. Toole, the listener is hearing separate 'streams' for the direct sound and the room. But 'the room' is not an independent source: it is derived from the speaker's sound. Even if the direct sound is correct, 'the room' will be 'incorrect' if it is not derived from a speaker with uniform dispersion. Both have to be consistent for it to sound right.

This is actually not quite what Dr Toole says. His view is that the off-axis performance of the speaker should be smooth, which includes but is not limited to speakers with uniform dispersion.

And as you know I mostly disagree with you about bass reflex vs sealed, although I certainly agree that if you are not using subwoofers, a 12dB roll off is preferable to a 24dB roll-off.

But thanks for clearing that up

 

[svart-hvitt]

I do try my best to make my position clear. I can't believe it's not obvious!

Let's try a different angle for anyone who is baffled by what is being said.

Let's be very precise. Basic experiments provide an observation that listeners to speakers in a room prefer a sound that, when measured with a microphone and FFT with a fixed-duration time window, yields a drooping frequency response. That is it. That is all that has been observed, and it is explanation-, theory- and understanding-free.

As has been seen, the author of a book on DSP, no less, extrapolates from this that people's hearing and brains prefer a drooping frequency response. From this can be extrapolated that EQ can/should be used to modify the sound from speakers in a room (any speaker, any room) to give this drooping frequency response. It is a 'target'. If the speaker happens to give this in-room curve without adjustment, this must mean that the speaker is a good one - although this view may be tarnished if it turns out that it doesn't do the same thing in different room...

Other people, upon intense discussion, make a distinction between the near field and the far field - but they provide no explanation for it; it is merely a (correct) observation. They do not register the idea of the listener hearing the speaker and room separately; they are still fixated on a single stream of 'audio soup' with a frequency response. Now, such people may not be quite so ready to accept that the drooping response should be a target, but they don't do anything to contradict it.

They don't have any method for saying what the response should be at 1', 2', 5', 10' from the speaker or at different angles. Their only method would be to make laborious observations with listening tests and multiple listeners at multiple distances and angles from the speaker. After many tests they might be in a position to create a set of tables of typical preferred frequency responses at different distances. These observations could even be correct for that room and speaker. Interpolation might be used to infer the preferred responses in between.

Now for the technology! Logically, a speaker system could be devised that used a clever range-finding system to determine the location of the listener in the room and adjusted the EQ to preset settings in order to always provide the scientifically-preferred EQ setting. Mysteriously, it might be found that the speaker doesn't yield the same curves in every room, but short of performing listening tests on the client, the speaker would be calibrated by carrying a microphone around the room and setting it up to provide the right measurements at all locations in that room. Result! Science!

What is wrong here? Pretty much everything, even though it could be argued it is a logical conclusion to draw from the original scientific observation.

What is the right answer? Well, an explanation or idea is like an engine: it has power, promotes progress, economy, elegance, performance, correctness. In this case, the explanation is: the speaker should have a flat response anechoically and have uniform dispersion because the listener perceives the direct sound of the speaker separate from the room. That is it. (That is my position @andreasmaaan!)

From that flows the simple design and performance of the Kii Three/D&D 8C, the great reviews, the observations of their neutral sound regardless of location, the lack of need for EQ settings. And also flows the lack of necessity for a listener tracker and dynamic EQ system, the lack of the need for listening tests, the lack of the need for in-room measurements. And possibly why hi-fi in the 1970s sounded better than typical audiophile systems now - wider baffles, sealed drivers, speakers set up in anechoic chambers independent of room, etc.

Is there something wrong with science then? No. Good science starts from a hypothesis, which should eventually turn into the explanation described above. But audiophile 'science' doesn't try to make the hypothesis an explanation. It just makes it a true/false criterion for the observation, and is only interested in differences or preferences while listening to music. It cannot possibly lead to understanding or a productive explanation, and results in dead ends and rabbit holes such as 'room correction' and 'target curves'.

This discussion on FR curves may clutter the DD8c theme, but the curve discussion is interesting nonetheless.

The distinction between direct and sum(direct+reverberation) is interesting; and I can’t reconcile my thoughts on this even after reading Toole.

So I thought there may be a way around this theme, to converge thoughts and camps on the FR curve issue. Say you have a measurement in-room that understands what is the correct curve slope in that specific room. It would basically be a smart algorithm that looks at the in-room response. From the in-room response the algo understands what’s the natural correct slope of the curve. Because every room is different, the slope of this curve will always be s bit different. But the algo identifies deviations from this natural, correct curve slope that will be an impediment to the chain of information that is a sound reproduction system. So it will take away excess energy caused by the room in order to reduce the tracking error versus a smooth curve. What the algo will give you, is an FR curve that has the natural correct slope for that particular room; however, the curve will have been smoothed a bit by taking away excess values of the tracking error (negative error, nulls, will be left untouched).

@Cosmik ’s position mirrors that of John Watkinson. But John Watkinson’s position is based on his ideal of a speaker, which is different from @Cosmik ’s speaker (ideal).

The problem is, I have never seen if this theoretical idea is valid given today’s speakers. In the meantime, it seems like DSP has objective value at least up to Schroeder (given simple tests like recording a speaker with and without room DSP).

To sum up: I think it boils down to @Cosmik believing a smoothing (which is not the same as making a set sloping!) of the curve in-room by DSP is flawed, while others believe it has merit; some, or most (?), in the pro DSP camp believe smoothing is flawed above Schroeder.

Trying do decipher what @mitchco really means on this issue, I think it seems like he is gravitating towards @Cosmik ’s position that a good speaker should be left untouched by DSP, wouldn’t you all agree?

 

[andreasmaaan]

So I thought there may be a way around this theme, to converge thoughts and camps on the FR curve issue. Say you have a measurement in-room that understands what is the correct curve slope in that specific room. It would basically be a smart algorithm that looks at the in-room response. From the in-room response the algo understands what’s the natural correct slope of the curve. Because every room is different, the slope of this curve will always be s bit different. But the algo identifies deviations from this natural, correct curve slope that will be an impediment to the chain of information that is a sound reproduction system. So it will take away excess energy caused by the room in order to reduce the tracking error versus a smooth curve. What the algo will give you, is an FR curve that has the natural correct slope for that particular room; however, the curve will have been smoothed a bit by taking away excess values of the tracking error (negative error, nulls, will be left untouched).

I'm sorry @svart-hvitt, I just need to clarify a few things to understand what you're saying.

What do you mean by "a measurement in-room that understands what is the correct curve slope in that specific room." How does it come to determine what this is?

And what do you mean by "tracking error"?

Trying do decipher what @mitchco really means on this issue, I think it seems like he is gravitating towards @Cosmik ’s position that a good speaker should be left untouched by DSP, wouldn’t you all agree?

I guess it's up to Mitch to clarify here, but my understanding from reading his most recent two reviews is that his standard practice is to use room-corrective DSP, although I'm not sure exactly for what.

 

[svart-hvitt]

I'm sorry @svart-hvitt, I just need to clarify a few things to understand what you're saying.

What do you mean by "a measurement in-room that understands what is the correct curve slope in that specific room." How does it come to determine what this is?

And what do you mean by "tracking error"?



I guess it's up to Mitch to clarify here, but my understanding from reading his most recent two reviews is that his standard practice is to use room-corrective DSP, although I'm not sure exactly for what.

Ok, @mitchco and many DSP enthusiasts talk about the correct curve as if it were a fixed target with a certain slope. The thing is, every room has its own, natural slope, as pointed out by @Cosmik . So my suggestion is to discard the idea of a universal curve target.

An intelligent algo will figure out what is the natural sloping of the in-room FR curve. This natural sloping is not primarily based on a theory; it’s based on observation of the specific in-room FR response. However, this natural in-room curve will be used to make a reference, a benchmark against which the actual FR curve can be compared.

However, I am not ready to buy everything @Cosmik says. I am suggesting that the measured in-room FR response can be improved by smoothing out the natural curve by taking away positive errors, i.e. excess energy. The algo will minimize tracking error by correcting the positive excesses. So we will have an algo that is used to create an in-room reference which will be used to minimize tracking error. So primarily a smoothing algo.

This method is interesting, isn’t it, as both @Cosmik and the average DSP enthusiast will disagree. Yet, it represents a middle-way. Truth is often somewhere in the middle.

 

[andreasmaaan]

An intelligent algo will figure out what is the natural sloping of the in-room FR curve. This natural sloping is not primarily based on a theory; it’s based on observation of the specific in-room FR response. However, this natural in-room curve will be used to make a reference, a benchmark against which the actual FR curve can be compared.

By "natural slope", do you mean something like the steady-state response that would be produced if an omnidirectional point source located at the position of the speaker were measured at the listening position?

I am suggesting that the measured in-room FR response can be improved by smoothing out the natural curve by taking away positive errors, i.e. excess energy. The algo will minimize tracking error by correcting the positive excesses. So we will have an algo that is used to create an in-room reference which will be used to minimize tracking error. So primarily a smoothing algo.

The problem is the one that @Cosmik mentioned: although sound field plays a role above Schroeder at least, we seem to be keenly attuned to the on-axis response of the speaker when judging tonal balance. Smoothing out the in-room response at the expense of flat on-axis response, even if done with reference to a curve that is somehow tailored to the "natural slope" of the room, will in my estimation tend to result in a perception that the tonal balance of the speaker is incorrect.

 

[svart-hvitt]

By "natural slope", do you mean something like the steady-state response that would be produced if an omnidirectional point source located at the position of the speaker were measured at the listening position?



The problem is the one that @Cosmik mentioned: although sound field plays a role above Schroeder at least, we seem to be keenly attuned to the on-axis response of the speaker when judging tonal balance. Smoothing out the in-room response at the expense of flat on-axis response, even if done with reference to a curve that is somehow tailored to the "natural slope" of the room, will in my estimation tend to result in a perception that the tonal balance of the speaker is incorrect.

By «natural slope» I mean an R^2 optimized reference based on actual in-room measurements.

A SOTA speaker has (about) the same on as off axis response.

 

[andreasmaaan]

By «natural slope» I mean an R^2 optimized reference based on actual in-room measurements.

A SOTA speaker has (about) the same on as off axis response.

Could you be more specific? I'm not familiar with the term "R^2 optimized reference". Just a link would be fine

And what's the relevance of the speaker having the same on as off axis response?

 

[mitchco]

Hey guys, interesting discussion. I like @andreasmaaan quote here:

"Also, I think you're mistaken about room correction and the target room curve, or at least you've been reading different studies to what I have. The downward slope in the steady-state response is not supposed to be generated by EQ, it's supposed to follow from an anechoicically flat speaker with downward sloping power response being optimally placed in a typical listening room. Only below the Schroeder frequency is room EQ to be used to tailor the response to the target. If the response doesn't meet the target in the mid and high frequencies, EQ is not the answer."

Yes, this is the idea, but there is an issue as to why so many speakers don't have this downward tilt. This is because the "single" measurement on axis in the anechoic chamber approach is a flawed approach. This is why Harman uses the spinorama approach which takes 72 anechoic measurements on two axis and uses a number of computational algorithms to produce the spinorama chart. As Kevin Voecks says, "As our research has long indicated, the listening window is a far better indicator of direct sound quality than is any on-axis curve. Harman researcher's work on the importance of first reflections--especially the side walls--proved that the speaker's output that contributes to these critical first reflections is as important as the direct sound."

If you look at the research https://www.computeraudiophile.com/...-with-roomfeel-headphone-review-r720/#science and scroll to the links at the end - btw, Sean Olive peer reviewed what I wrote there, but see all the reference links. If you pile through them, you will see many studies where participants sat in Harman's typical listening room and with a speaker eq'd flat, the participants adjusted bass and treble controls, adjusted until it sounded accurate or neutral to their ears, as to what they preferred. Guess what, statistically, it was 20 Hz to -10 dB at 20 kHz measured in-room steady state response. You can also see it in Figure 14 in Dr. Toole's http://www.aes.org/e-lib/browse.cfm?elib=17839

Harman, did the same thing with headphones too. And cross correlated back, as it was the same as the speakers preference. Over and over again. It got to the point that based on the spinorama's and how it measured in the room and how people perceived it, were able to make it a repeatable process to the point where there is an 0.86 correlation coefficient that based on the spnioramas can predict the tilted in room response that sounds accurate, neutral, preferred, or whatever you want to call it. As said before, Dr Toole indicates that accurate and preferred are synonymous.

And now it is an industry standard: https://standards.cta.tech/kwspub/published_docs/ANSI-CTA-2034-A-Preview.pdf

So a well designed loudspeaker, like the Salon2, will have a that -10 dB tilt if measured in the room based on it's spinorama. I know as somone sent me a measurement You can also see it in the listening window and even JA's measures in Stereophile. Paul Barton's Imagine T3 in the link I shared above, same response, in-room, no eq. But as you know, Paul, Floyd and Sean have been through the NRC in Canada...

To be totally clear, I did not apply an external eq to the 8c's or Kii Three's at all. In fact, I don't want to apply any eq or tone control to anything above 500 Hz if possible. But until loudspeaker manufacturers up their ante to using an industry standard (i.e. spinorama) linked above, we will continue to get speakers that, certainly to my ears, sound way too bright, no matter what room they are in...

 

[Blumlein 88]

Peter Walker used to suggest a 3 db per decade slope as a good target. He got that idea from those old B&K research papers in either the late 60's or early 1970's. 3 db per decade is a 9 db slope from 20 to 20 khz. They actually also suggested 10 db the 3 db per decade was a close approximation. So the idea has been around if people had paid attention.

 

[andreasmaaan]

Yes, this is the idea, but there is an issue as to why so many speakers don't have this downward tilt. This is because the "single" measurement on axis in the anechoic chamber approach is a flawed approach. This is why Harman uses the spinorama approach which takes 72 anechoic measurements on two axis and uses a number of computational algorithms to produce the spinorama chart. As Kevin Voecks says, "As our research has long indicated, the listening window is a far better indicator of direct sound quality than is any on-axis curve. Harman researcher's work on the importance of first reflections--especially the side walls--proved that the speaker's output that contributes to these critical first reflections is as important as the direct sound."

Thank Mitch

I agree with Harman's use of "listening window" measurements rather than a single on-axis measurement when designing the speaker crossover, and my "anechoically flat" statement was a shorthand for basically this.

But I don't agree that it's this approach that is the reason that Harman speakers tend to exhibit the downward tilt, while others don't. In fact, the two are not really related.

The "listening window" approach is about ensuring the flattest possible direct sound by averaging out a number of inevitably slightly different measurements in the "window" within which the listener's ears are likely to be located. It is also a way of ensuring that little FR response anomalies that happen to occur at precisely 0° aren't compensated for in the design, as this would skew the response at all locations not exactly 0° in front (and with two ears, we're never at precisely 0° in front).

The downward tilt in the sound field is actually unrelated to all this, it comes about as a result of two things primarily: (1) most speakers tend to become more directional as frequency increases, so the sound going out into the room and being reflected back is weighted towards the lower frequencies, and (2) most rooms become increasingly absorptive as frequency rises.

It's therefore not just Harman speakers that tend to exhibit this downward tilt, it's most dynamic speakers in most rooms. Harman however goes to significant lengths to ensure that this downward tilt is smooth and controlled (more so than some other manufacturers, who pay less attention to off-axis performance / power response).

 

[Blumlein 88]

I'm simply using logic:
A neutral speaker (flat response) in the near field (effectively anechoic or in an anechoic chamber) sounds neutral.
A neutral speaker in the far field sounds neutral.
Logical conclusion: humans find a flat response to sound neutral.

But the 'conventional' explanation/conclusion i.e. espoused by all room correction enthusiasts, is that something happened when the listener moved from near to the speaker to far away. Their hearing changed so that they now prefer a drooping frequency response - because an indiscriminate microphone/FFT measurement shows that if you accumulate a timed window of in-room sound and then examine the contents of the FT bins, there is a downwards slope. If the duration of the window is changed, the slope changes, too. Messy.

The other explanation (did someone mention Occam's Razor?) is that the human hears the direct, neutral sound from the speakers wherever they are in the room relative to the speakers!

It's simple logic, could have been (was?) predicted before these speakers were developed, fits perfectly with everyday experience and what evolution would have provided if it wanted an animal's hearing to give it a survival advantage. It also fits perfectly with the 'hollow recording' experiment where an in-room microphone picks up something different to what a human hears in the live situation. But I could be wrong...

No this is not it. An anechoicly flat speaker won't measure flat in room. It isn't that you slope the response when moving far away. If the response were flat nearfield, and you do nothing except measure in the far field you'll end up with some sloping response result when nothing has changed.

Now our hearing does stuff to ignore the reflections that cause that sloping measured result. We as you say hear the flat direct response in room. But our measurements show a slope. So you get unknown speakers in a room and measure farfield. If they don't conform to the expected measurement slope you EQ them to fit that slope. The assumption, which might be invalid, probably is partly invalid due to unknown directional effects of the speaker, is that you EQ your measured result to fit the same slope a flat speaker would give you. If you do this correctly, then presumably a nearfield or anechoic measure of this 'corrected' speaker response would be flat or at least closer to flat. In room with our brain processing early reflections we also would hear a flat neutral result.

There are plenty of reasons why that simplified assumption working backward from in room measurements might be in error. The basic idea isn't so crazy. We are using a slope to compensate for measurement artifacts. Sure build the speaker right in the first place and you don't need this.

 

[mitchco]

Thank Mitch

I agree with Harman's use of "listening window" measurements rather than a single on-axis measurement when designing the speaker crossover, and my "anechoically flat" statement was a shorthand for basically this.

But I don't agree that it's this approach that is the reason that Harman speakers tend to exhibit the downward tilt, while others don't. In fact, the two are not really related.

The "listening window" approach is about ensuring the flattest possible direct sound by averaging out a number of inevitably slightly different measurements in the "window" within which the listener's ears are likely to be located. It is also a way of ensuring that little FR response anomalies that happen to occur at precisely 0° aren't compensated for in the design, as this would skew the response at all locations not exactly 0° in front (and with two ears, we're never at precisely 0° in front).

The downward tilt in the sound field is actually unrelated to all this, it comes about as a result of two things primarily: (1) most speakers tend to become more directional as frequency increases, so the sound going out into the room and being reflected back is weighted towards the lower frequencies, and (2) most rooms become increasingly absorptive as frequency rises.

It's therefore not just Harman speakers that tend to exhibit this downward tilt, it's most dynamic speakers in most rooms. Harman however goes to significant lengths to ensure that this downward tilt is smooth and controlled (more so than some other manufacturers, who pay less attention to off-axis performance / power response).

Thanks. May I refer you to Dr Toole's presentation from 2002... https://www.harman.com/sites/default/files/LoudspeakersandRoomsPt2_0.pdf See page 6:

"Using a loudspeaker that we know, in retrospect, had a design flaw, let us see what happens in a room. This loudspeaker was designed using the philosophy that the direct sound, the on-axis sound, is the most important... If we are to try to anticipate how a loudspeaker will sound in a room, it is necessary to measure everything, and not just a few curves around the on-axis measurement." I think you may know where this is headed...

Same presentation... Listening window = "The second curve is a spatial average over +/- 30° horizontal, and +/- 10° vertical, representing the direct sound for listeners seated in a row of chairs or a large sofa, and possibly standing and sitting".

Re: "most speakers tend to be directional as frequency increases..." Both the Kii THREE and the D&D 8c are constant directivity from 100 Hz to 20 Khz... There is absolutely no increase in directivity as frequency increases for these speakers. But the Kii THREE needs -5 dB attenuation from 3 kHz to 20 kHz, whereas the 8c's do not...

PS. My room's measured RT60 or REW's optimum room decay shows that my room is not more absorbent as frequency increases:

Based on the research that Toole and Olive have performed and certainly based on the measurements that I have made on several loudspeakers that have come from my room, very few speakers actually exhibit a downward tilt in frequency response. In fact, it is way fewer than people would even believe. It is simply that the (vast) majority of speakers are based on a flawed approach of using a single on axis measurement in an anechoic chamber to determine in room frequency response. Toole's paper points this out in 2002 and not much has changed since, even when the right way to do it is available for $100.

OK, I am not being fair, to properly do it requires a substantial investment in being able to take 72 measurements in an anechoic chamber. But so far, no-one has been able to prove Toole's and Olive's research incorrect. Conversely, we have an excellent example of a very neutral loudspeaker based on this research: Revel Salon2, which still is the leader since 2009 for a neutral, downward tilt in-room frequency response. And now the 8c's...

 

[Hrodulf]

I was always under the impression that downward sloping in home systems occured from sitting off-axis. Not many people toe in their speakers.

 

[Cosmik]

No this is not it. An anechoicly flat speaker won't measure flat in room. It isn't that you slope the response when moving far away. If the response were flat nearfield, and you do nothing except measure in the far field you'll end up with some sloping response result when nothing has changed.

Now our hearing does stuff to ignore the reflections that cause that sloping measured result. We as you say hear the flat direct response in room. But our measurements show a slope. So you get unknown speakers in a room and measure farfield. If they don't conform to the expected measurement slope you EQ them to fit that slope. The assumption, which might be invalid, probably is partly invalid due to unknown directional effects of the speaker, is that you EQ your measured result to fit the same slope a flat speaker would give you. If you do this correctly, then presumably a nearfield or anechoic measure of this 'corrected' speaker response would be flat or at least closer to flat. In room with our brain processing early reflections we also would hear a flat neutral result.

There are plenty of reasons why that simplified assumption working backward from in room measurements might be in error. The basic idea isn't so crazy. We are using a slope to compensate for measurement artifacts. Sure build the speaker right in the first place and you don't need this.

I think you are actually agreeing with me despite your first sentence!

The idea of the frequency response target is backwards. The observation that a neutral speaker gives the sloping response in a dumb (in the British sense of the word) FFT in a typical room derives from simple, understandable physics. But by understanding where the curve comes from, rather than just making the observation, it can be predicted that the neutral speaker would produce a different curve in a different room. Also, a more, or less, directional speaker (which still counts as neutral) would produce a different curve.

So, even if we don't stray into non-neutral speaker territory, it would be a mistake to use any curve whatsoever as a target for adjusting the EQ of a neutral speaker. It shouldn't be adjusted at all, whatever the in-room measurements.

In this particular case it is observed that "I didn't have to make any adjustments to meet the target curve", but this is just an indication that the room was not far off 'typical', and the speaker is neutral and happens to have similar directivity to the one used when measuring the special curve. It is a backwards way of establishing a speaker's neutrality - which would be better established without reference to any in-room curve at all.

 

[Cosmik]

I was always under the impression that downward sloping in home systems occured from sitting off-axis. Not many people toe in their speakers.

I think there's a very logical reason for not toe-ing in ordinary speakers. Because ordinary speakers' directivity generally narrows as frequency increases, this creates an inconsistency between the direct and reverberant frequency responses. It therefore makes perfect sense to listen to them off axis, the listener getting less of the 'top end' that is playing into the room. The direct sound may then correspond better subjectively with the reverberant sound. Tweaking the top end to be brighter when measured on-axis may also be a perfectly sound strategy.

At the same time, the baffle step is also a factor i.e. where the speaker sounds deficient in the bottom end and this requires compensation.

Making non-neutral speakers sound better must be an art, but can be aided with a little science. Knowing the frequency of the baffle step through calculation would remove one variable from the permutations you would need to cover, for example.

Before anyone says it, I don't think the in-room target curve works here either. Each speaker will have its own foibles to be dealt with that *should* show up to some extent in the in-room curve (if you have time on your hands and measure it). The lumpy dispersion of one speaker cannot be magicked away completely so that it sounds the same as a neutral speaker when their in-room curves match. If the curves match, the direct sound of the lumpy speaker is now way off neutral and the listener will hear that. The compensation should only be partial, not total.

 

[Soniclife]

Re: "most speakers tend to be directional as frequency increases..." Both the Kii THREE and the D&D 8c are constant directivity from 100 Hz to 20 Khz... There is absolutely no increase in directivity as frequency increases for these speakers. But the Kii THREE needs -5 dB attenuation from 3 kHz to 20 kHz, whereas the 8c's do not...

Do you have any idea why these two, that I would have guessed from measurements would produce the same in room profile differ?

 

[svart-hvitt]

Do you have any idea why these two, that I would have guessed from measurements would produce the same in room profile differ?

If you measure the sum of direct and reflected sound in-room, two anechoically equal speakers of different design will not measure the same in-room, wouldn’t you agree?