Harman preference curve for headphones - am I the only one that doesn't like this curve?

[andreasmaaan]

That's all good and understandable, and it would be informative if you provided what type of tilt you believe is neutral, so we can get one more opinion on this. Thanks!

Good question, but I'm not sure I can answer it, at least not on theoretical grounds. In fact, it's a bit like questions about what the most "neutral" loudspeaker directivity is, or what the most neutral listening environment is. This information is just not contained in any recording, which makes it impossible to say that a particular loudspeaker directivity or room acoustic is the most faithful.

It's similar with headphones: I think there's good justification for a downward tilt, but I don't think there is any specific degree of tilt that is the most correct: this information is just not contained in the recording.

In practical terms, if you have the measurements available, I would start with a flat diffuse field response and then adjust bass and treble until it sounds subjectively neutral to you (which is not so different to what Harman did with their research subjects).

But I'm afraid I don't have an answer to the general question :-/

 

[bluefuzz]

I do agree with you that some degree of downward slope in a headphone target response is likely necessary for subjectively neutral sound. But I would disagree with the idea that the Harman target is the most neutral.

As others have suggested, the Harman target should be considered a somewhat fuzzy range rather than a sharply defined line. It will obviously never be the 100% most neutral response for every headphone on every head for every personal preference. But broadly speaking it is more right than wrong and, until better science is available, still the best target to aim for a neutral(ish) and accurate(ish) headphone response.

 

[andreasmaaan]

As others have suggested, the Harman target should be considered a somewhat fuzzy range rather than a sharply defined line. It will obviously never be the 100% most neutral response for every headphone on every head for every personal preference. But broadly speaking it is more right than wrong and, until better science is available, still the best target to aim for a neutral(ish) and accurate(ish) headphone response.

Back in circles here, but the Harman target has nothing to do with, and was not intended to have anything to do with, neutrality. It is an average preference target. That doesn't mean it isn't reasonably neutral, but it doesn't mean it is reasonably neutral, either.

 

[Feelas]

Good question, but I'm not sure I can answer it, at least not on theoretical grounds. In fact, it's a bit like questions about what the most "neutral" loudspeaker directivity is, or what the most neutral listening environment is. This information is just not contained in any recording, which makes it impossible to say that a particular loudspeaker directivity or room acoustic is the most faithful.

It's similar with headphones: I think there's good justification for a downward tilt, but I don't think there is any specific degree of tilt that is the most correct: this information is just not contained in the recording.

We could say that the recording process itself & splitting into stereo is a form of lossy compression Yes, I agree, the information is not there so we
are left guessing. All in all, you could say that there is just no "best", merely lots of reference points so you won't get lost in an infinite option space. Discretizing is almost always a good idea to solve something.

After all, the problem would be solved if recordings came with recording room PRIR & mastering engineer's HRTF, so we could correct from there. This is of course impossible, but it might solve the dilemma. Yet, to add up to the problem, you could consider mastering room's impulse response a company secret - funny how the listeners' and studios' interests don't align.

In practical terms, if you have the measurements available, I would start with a flat diffuse field response and then adjust bass and treble until it sounds subjectively neutral to you (which is not so different to what Harman did with their research subjects).

I'd stray from saying that diffused field has any practical matter rather than merely being a reasoning device; even unfurnished rooms in living spaces aren't diffused fields, since walls aren't infinitely smooth, the reasoning goes that way, I think. EQ-ing doesn't help with uneven reverb time, which might heavily ruin the sound. That is said, of course, in regards to far-field.

 

[andreasmaaan]

I'd stray from saying that diffused field has any practical matter rather than merely being a reasoning device; even unfurnished rooms in living spaces aren't diffused fields, since walls aren't infinitely smooth, the reasoning goes that way, I think. EQ-ing doesn't help with uneven reverb time, which might heavily ruin the sound. That is said, of course, in regards to far-field.

Well, that's why I said "as a practical matter". Since headphones don't resemble listening rooms in any way whatsoever really, you have to start from somewhere. What starting point would you suggest, and why?

 

[raistlin65]

As others have suggested, the Harman target should be considered a somewhat fuzzy range rather than a sharply defined line. It will obviously never be the 100% most neutral response for every headphone on every head for every personal preference. But broadly speaking it is more right than wrong and, until better science is available, still the best target to aim for a neutral(ish) and accurate(ish) headphone response.

Why is there a need to push HTR as defining "neutral" when Harman's research describes it as preference?

 

[bluefuzz]

It is an average preference target.

Yes, we're probably going a bit round in circles.

I feel the word preference here is a bit of a stumbling block. It reminds me of when creationists call evolution 'just' a theory, meaning 'anyone's guess' rather than a testable scientific hypothesis. The same goes for preference in the context of audio testing. I feel it has more rigor behind it than simply 'what a bunch of dudes kinda thought sounded cool that day' (not that I think that is what you personally are saying). My understanding of preference with regard to audio testing is that for most people it strongly correlates with neutrality and uncoloured response.

 

[Feelas]

Well, that's why I said "as a practical matter". Since headphones don't resemble listening rooms in any way whatsoever really, you have to start from somewhere. What starting point would you suggest, and why?

I guess it'd be most informative to ask a lot of mastering studios what they're exactly targetting; I think that most engineers have some concept of how records translate onto gear - in fact, I'm yet to hear (and I've heard quite a bunch of stuff already) recordings that would sound bad on anything. Not fasctinating anywhere - yes, but for broken ones I'd pick maybe a dozen. There IS some commonality. We should do what was done by Harman and target *some* speaker reference - at least stop targetting some intangible headphone target, since humans aren't accustomed to acoustic coupling of headphones in nature. Going that way - since the HTC was derived form Harman's reference room w/ JBL speakers, why not take that as a signpost and why are we trying to avoid that? It's said out loud how the preference diverged between target groups, so I don't think there's anything missing.

I think it would make no sense to consider omnis & various electrostatic or reflected-sound speakers a reference. Why? Simply because I don't think any studios take these into account, so in the end it'll be artifical in regards to what the engineer heard, and the engineer is the reference, the only signpost.

Why is there a need to push HTR as defining "neutral" when Harman's research describes it as preference?

Perhaps since it was derived scientifically, and the preference range is highlighted, not hidden from ones interested in it for what it is. The group might be small, but in the end I believe we have some semblance of what sounds correct, since we're living in it (and not in headphones) since we're born and since people exist, in general. That's why both free field & diffuse field make no sense - nobody lives either in anechoic nor in extremely live space.

Behind "merely the preference" lives a bigger, more objective matter of what humans in general consider right - and that's what we should be looking for. I think that the difference between trained and untrained listeners stems only from the lack of experience, and we'd converge on some specific curve, given enough time for untrained becoming trained.

 

[andreasmaaan]

Going that way - since the HTC was derived form Harman's reference room w/ JBL speakers, why not take that as a signpost and why are we trying to avoid that?

But this is the common misunderstanding re: the HTC. It was not derived this way at all.

This is essentially the process that was followed:

1. EQ a pair of loudspeakers in the Harman listening room so that the in-room steady state response is flat.
2. Measure this response using a HATS (NB: the obtained response is almost identical to a flat diffuse field response - as you would expect given the measurements were taken beyond the room's critical distance: see graphs below).
3. EQ headphones to match this measured HATS response.
4. Give subjects these EQ'd headphones and a pair of shelving filters and ask them to EQ the sound until they arrive at their preferred bass/treble balance.

The only two steps in this process that I would change if I had the resources design a pair of headphones for myself in this way would be (1) and (4).

In the case of (1), I would use a true flat diffuse field response instead of a flat in-room steady-state response, because - why not? (Although, as I mentioned, I don't think much at all turns on this given the HATS response Harman measured; again, see graphs below).

And in the case of (4), I would use a bass filter with a slightly different Q, since the filter used in the Harman study was underdamped, leading to a slight reduction in lower-midrange as bass level increased (although this was done intentionally and arguably for good reason, it is not how I would do it).

As to the significance of using a diffuse field starting condition vs the starting condition chosen by Harman, as I said, I don't think much turns on it.

The Harman starting response is shown below (green trace on the left graph), along with a number of typical diffuse field (flat) responses. You'll notice that the Harman response and the diffuse field responses are essentially the same, tracking essentially perfectly right through to the upper treble (NB: the diffuse field graph is smaller so that the scales of the two graphs match):

In other words, the Harman starting condition essentially was a flat diffuse field response.

 

[Robin L]

Well, that's why I said "as a practical matter". Since headphones don't resemble listening rooms in any way whatsoever really, you have to start from somewhere. What starting point would you suggest, and why?

I know ya didn't ask me, excuse me for jumping in. I'm on the other side of this question, I really like the AKG K371 headphones, they are now my default. Are they perfect? No, but the ways they are good are more important to me than the ways they are not so good. First off, they go all the way down, without eq. It's boosted at the very bottom, other than that, having full response into deep bass makes everything a little more "real". While the fact that the AKG 'phones are super efficient isn't directly related to the adherence to the Harman curve, that feature does make these headphones easier to drive hooked up to just about anything. The two areas when I'd most desire improvements are associated with the two major deviations from the Harman curve, the slight excess of bass [and associated distortion] and the "sparkly" end of the treble, where there's a lack of energy and an excess of harmonic distortion. Then again, they're going for $120 at B + H, so ya pays yer money and makes your choice. But seriously, I can't see how the Harman curve is a bad thing. Anyway, it's a moving target and the complaints I'm seeing are mostly carping.

 

[markanini]

I find EQ above 2Khz unreliable on headphones so wouldn't necessarily use that to form any conclusions on personal preference.

 

[Feelas]

But this is the common misunderstanding re: the HTC. It was not derived this way at all.

This is essentially the process that was followed:

1. EQ a pair of loudspeakers in the Harman listening room so that the in-room steady state response is flat.
2. Measure this response using a HATS (NB: the obtained response is almost identical to a flat diffuse field response - as you would expect given the measurements were taken beyond the room's critical distance: see graphs below).
3. EQ headphones to match this measured HATS response.
4. Give subjects these EQ'd headphones and a pair of shelving filters and ask them to EQ the sound until they arrive at their preferred bass/treble balance.

The only two steps in this process that I would change if I had the resources design a pair of headphones for myself in this way would be (1) and (4).

In the case of (1), I would use a true flat diffuse field response instead of a flat in-room steady-state response, because - why not? (Although, as I mentioned, I don't think much at all turns on this given the HATS response Harman measured; again, see graphs below).

And in the case of (4), I would use a bass filter with a slightly different Q, since the filter used in the Harman study was underdamped, leading to a slight reduction in lower-midrange as bass level increased (although this was done intentionally and arguably for good reason, it is not how I would do it).

As to the significance of using a diffuse field starting condition vs the starting condition chosen by Harman, as I said, I don't think much turns on it.

The Harman starting response is shown below (green trace on the left graph), along with a number of typical diffuse field (flat) responses. You'll notice that the Harman response and the diffuse field responses are essentially the same, tracking essentially perfectly right through to the upper treble (NB: the diffuse field graph is smaller so that the scales of the two graphs match):

View attachment 110901

In other words, the Harman starting condition essentially was a flat diffuse field response.

It was in essense the flat diffuse field FR, but the conditions weren't of a perfectly reverberant room, but a treated one, if I get what DF is supposed to approximate correctly. If not, sorry! The DF target was probably taken because of being the previous point of reference, or a known anchor, that's all, I think.

Still, I think you're right and we would agree that it doesn't really matter where we start, since knowing the specifics of the living space we EQ, it's possible to estimate what'll the end result be... but since the "flat in-room equalized" wasn't the preferred response (for nobody!), it'd be pointless to take this curve as a starting point, don't You think? If it doesn't

And after all, the curve derived for headphones was then experimentally confirmed for being the "best one" for a group of listeners in "A Statistical Model that Predicts Listeners’ Preference Ratings of Around-Ear and On-Ear Headphones" study. No more bass/treble mumbo jumbo, just pure DBT.

The curve itself won the preference as the hidden high-anchor, when K712 were EQ'd that way - I think that it does have some semblance of sense, doesn't it?

 

[andreasmaaan]

Yes, we're probably going a bit round in circles.

I feel the word preference here is a bit of a stumbling block. It reminds me of when creationists call evolution 'just' a theory, meaning 'anyone's guess' rather than a testable scientific hypothesis. The same goes for preference in the context of audio testing. I feel it has more rigor behind it than simply 'what a bunch of dudes kinda thought sounded cool that day' (not that I think that is what you personally are saying). My understanding of preference with regard to audio testing is that for most people it strongly correlates with neutrality and uncoloured response.

Ok, sure, I don't mean to suggest that "preferred" implies that it is not neutral, and I certainly don't mean to imply that it is not good. I just am hesitant to make a generalisation about the relationship between preference and neutrality that comes from one set of studies involving speakers (and with which I already have some reservations) and apply to another set of studies involving headphones. I can understand why you're keen to do that, though

But seriously, I can't see how the Harman curve is a bad thing. Anyway, it's a moving target and the complaints I'm seeing are mostly carping.

Absolutely, and I'm not trying to argue that HTR is a bad thing at all.

I'm just trying to be clear about what it is (an average preference curve) and what it isn't (an attempt at neutrality)

 

[andreasmaaan]

It was in essense the flat diffuse field FR, but the conditions weren't of a perfectly reverberant room, but a treated one, if I get what DF is supposed to approximate correctly. If not, sorry!

You're basically correct there. A DF response is a response in a highly reverberant room. The thing is, since we are dealing with headphones here, it doesn't really matter whether the response was obtained in a diffuse field or in a just-normally reflective room - if the two responses are essentially the same, they are essentially the same.

The DF target was probably taken because of being the previous point of reference, or a known anchor, that's all, I think.

The reasoning behind using this response as the starting condition was not explained in the study, but I have a theory (see next para).

Still, I think you're right and we would agree that it doesn't really matter where we start, since knowing the specifics of the living space we EQ, it's possible to estimate what'll the end result be... but since the "flat in-room equalized" wasn't the preferred response (for nobody!), it'd be pointless to take this curve as a starting point, don't You think?

The reason I believe a flat DF response makes the best starting point is that it ensures as much as possible that the FR is smooth. If we don't EQ to some kind of flat starting response, we end up with irregularities in the response that will not be removed by wide-band bass and treble shelf filters. A flat DF response (or indeed the response that Harman used) ensures that there are no narrow-band resonances that the subjects can't correct (that is, until you take the headphones off the dummy head and put them on the head of a subject - but that's a whole separate can of worms!).

And after all, the curve derived for headphones was then experimentally confirmed for being the "best one" for a group of listeners in "A Statistical Model that Predicts Listeners’ Preference Ratings of Around-Ear and On-Ear Headphones" study. No more bass/treble mumbo jumbo, just pure DBT.

The curve itself won the preference as the hidden high-anchor, when K712 were EQ'd that way - I think that it does have some semblance of sense, doesn't it?

Sure, I am not trying to be critical of the HTC here. I think there is more than a semblance of sense to it All I'm trying to do is to caution against extending it beyond what it was designed to do...

 

[Feelas]

You're basically correct there. A DF response is a response in a highly reverberant room. The thing is, since we are dealing with headphones here, it doesn't really matter whether the response was obtained in a diffuse field or in a just-normally reflective room - if the two responses are essentially the same, they are essentially the same.

Well, of course - if we would somehow get the same FR, it must sound the same (given other factors don't change).

I just think that they won't be the same, if you take a room with a different reflectivity, since the resultant FR w/ the same direct sound would be different, just as you mentioned - remember that angle of attack towards either a HATS or a person is bound with a different frequency response - alas a different perception. People hear 0° and 90° differently. So, if we do (and in fact I believe that is exactly what happens in all these DF-tuned headphones) apply DF curve into headphone design process, we'll end up with an unrealistically "live room" perception.

Yet, what's more important - I don't believe that the tonality will be the only off thing! Since there's a connection between angle of attack and perceived response (not merely the time cues), I think that the distance & imaging perception must also be off, then. Having an identical mix - i.e. 50% and 50% - of indirect and direct sound will result in a different FR in perfectly reverberant spaces and in partly reverberant ones. Harman research being transparent on approximating a 30°/30° (somehow treated) environment is, I think, a good thing. We know what they've been trying to approximate, exactly. Does DF target give any directionality that it was taken for? The exact direct vs indirect balance MIGHT (but I don't have anything to show for it, sadly, theorizing here) be crucial to proper soundstaging & perceived width, also the perceived space, in itself. Get the indirect vs direct wrong and nothing will sound "3D".

"Holography" is essentialy an audiophile buzzword, but it makes sense to use that word; it simply means that layering is correct and what should sound distant, does. I think that DF target makes the reflections "seem" absurdly close and in fact that's all the fuss about "wide" imaging in headphones (which don't have angled drivers at the same time) - the reflections are simply made closer, thus illusion of a wider, but less in-depth, soundfield.

As a sidenote - I've read a study on some small group of people (which I'd have to dig for on the SSD) which correlated typical front vs backwards confusion in headphones with FR dips and shown that correcting for the dips alleviates the problem. Interesting, honestly.

The reason I believe a flat DF response makes the best starting point is that it ensures as much as possible that the FR is smooth. If we don't EQ to some kind of flat starting response, we end up with irregularities in the response that will not be removed by wide-band bass and treble shelf filters. A flat DF response (or indeed the response that Harman used) ensures that there are no narrow-band resonances that the subjects can't correct (that is, until you take the headphones off the dummy head and put them on the head of a subject - but that's a whole separate can of worms!).

So in essence any smooth target would work - I think that theory makes sense. Perhaps that's partly a polemic on whether DF is fine or not, and the answer is: not.

Sure, I am not trying to be critical of the HTC here. I think there is more than a semblance of sense to it All I'm trying to do is to caution against extending it beyond what it was designed to do...

Aight, never thought you are. Still, I don't think that we should put much more into "neutral" than what we might expect; the reference must be shared and known between everyone to make sense.

 

[andreasmaaan]

I just think that they won't be the same, if you take a room with a different reflectivity, since the resultant FR w/ the same direct sound would be different, just as you mentioned - remember that angle of attack towards either a HATS or a person is bound with a different frequency response - alas a different perception. People hear 0° and 90° differently. So, if we do (and in fact I believe that is exactly what happens in all these DF-tuned headphones) apply DF curve into headphone design process, we'll end up with an unrealistically "live room" perception.

I don't see how a headphone's frequency response can give a "live" or "dead" perception per se.

And I think you're glossing over an important fact regarding the Harman starting response. Although it was taken with two speakers in a room, it was taken beyond that room's critical distance, such that the majority of the sound measured was indeed reflected sound.

My opinion is, the more reflected sound, the better, since a 360° sound field reduces the FR irregularities caused by the HRTF specific to any particular azimuth angle of the source. In other words, the more reflective the sound field, the smoother the response.

Yet, what's more important - I don't believe that the tonality will be the only off thing! Since there's a connection between angle of attack and perceived response (not merely the time cues), I think that the distance & imaging perception must also be off, then. Having an identical mix - i.e. 50% and 50% - of indirect and direct sound will result in a different FR in perfectly reverberant spaces and in partly reverberant ones. Harman research being transparent on approximating a 30°/30° environment is, I think, a good thing. We know what they've been trying to approximate, exactly. Does DF target give any directionality that it was taken for?

I tend not to agree here. We are talking only about the starting condition. The main consideration IMHO is therefore not tonality, but smoothness of response. DF is the surest means to achieve that. Tonality is then determined in the experiment itself.

I'm not sure though I understood your question, "Does DF target give any directionality that it was taken for?"

So in essence any smooth target would work - I think that theory makes sense. Perhaps that's partly a polemic on whether DF is fine or not, and the answer is: not.

I'm not sure I understood this either

 

[Feelas]

I don't see how a headphone's frequency response can give a "live" or "dead" perception per se.

FR different than perceptibly flat will also bring out any reverb that has been taken on the recording, but the level will be increased - keep in mind, that not only the direct sound is captured when recording. As such, the recording will seem to be recorded in a space livelier than it was. If you took -20dB components into -15dB range (or -25dB range, whichever way shows better), that'd be a sure perceptible difference on how live the recording room is perceived. And I think that it's not only the problem of "sharp razors in your ears" with overestimated treble - in fact, overthrown balance will bring out components which the engineer didn't intend highlighted, and this might end up... well, I believe in lots of different ways.

The tonality is one thing, the size of point sources connected with correct (smooth) reverb structure & perceived distances is something entirely different, yet both result from proper or improper FR, nothing more, nothing less, apart from THD&N and other issues. Playing with speaker EQ on drum records shows how the size of instruments might be exaggareted or made smaller, by bringing out different sound components. That all is a prime matter, and I think completely lost in the FR dilemma, where people only regard boomy bass or shouty treble. The FR that gives proper size & width of sources should be regarded as correct, since it must align with natural hearing & perception of space.

And I think you're glossing over an important fact regarding the Harman starting response. Although it was taken with two speakers in a room, it was taken beyond that room's critical distance, such that the majority of the sound measured was indeed reflected sound.

My opinion is, the more reflected sound, the better, since a 360° sound field reduces the FR irregularities caused by the HRTF specific to any particular azimuth angle of the source. In other words, the more reflective the sound field, the smoother the response.

Well, that's an important distinction to make, and I didn't actually know that. Taking that into account, I still think that the actual mix of direct & indirect is important. The HRTF might be getting smoother, but I don't actually think that makes a lot of difference - there's some dominant sound coming from some direction and it'll affect what is heard the most. Assuming a target that doesn't contain direct sound would make no sense, would it?

I don't tend to agree here. We are talking only about the starting condition. The main consideration IMHO is therefore not tonality, but sm
oothness of response. DF is the surest means to achieve that. Tonality is then determined in the experiment itself.

I'm not sure though I understood your question, "Does DF target give any directionality that it was taken for?"

Yes, that was the question. Is DF is an average over all angles that sound comes from when perceived in-ear or is it taken at some specific azimuth in-ear?

I'm not sure I understood this either

Oops, just me being cryptic again. I meant that Harman might've just chosen to show that DF target is not proper, albeit phrased it... strangely. Made sense to me, hah!

 

[andreasmaaan]

FR different than perceptibly flat will also bring out any reverb that has been taken on the recording, but the level will be increased - keep in mind, that not only the direct sound is captured when recording. As such, the recording will seem to be recorded in a space livelier than it was. If you took -20dB components into -15dB range (or -25dB range, whichever way shows better), that'd be a sure perceptible difference on how live the recording room is perceived. And I think that it's not only the problem of "sharp razors in your ears" with overestimated treble - in fact, overthrown balance will bring out components which the engineer didn't intend highlighted, and this might end up... well, I believe in lots of different ways.

That will be true in some cases, but completely recording-dependent. It all depends on the relative spectra of direct sound and reflected sound in the recoridng, and the nature of the deviation in FR upon reproduction.

A deviation in the reproduced FR may just as well make a recording sound less live, depending on these factors.

Well, that's an important distinction to make, and I didn't actually know that. Taking that into account, I still think that the actual mix of direct & indirect is important. The HRTF might be getting smoother, but I don't actually think that makes a lot of difference - there's some dominant sound coming from some direction and it'll affect what is heard the most. Assuming a target that doesn't contain direct sound would make no sense, would it?

As I argued before, the starting condition is not a target. It is just a starting condition, and the primary goal should therefore be for it to be as smooth as possible. If you wanted to truly randomise the starting conditions, you might decide, for example, to introduce random tilts to the flat DF response, and I think this would actually be a way to improve the experiment (since starting from a flat DF or steady-state response is arguably confounding). But in all cases, I would prefer a smooth DF starting condition.

I do see where you're coming from though. You're saying that, when we listen to speakers in a room, the first-arriving sound we hear has (ideally) a +/-30° azimuth, and you think that this should be reflected in the starting condition. I can see the logic to that, but I think that a typical HRTF at +/-30° is going to be way too erratic to be of any use. Also, I don't think it's necessarily clear that headphones are supposed to emulate the response of loudspeakers at +/-30°, let alone the question whether they even can, given all the fundamentally different conditions under which they operate (different azimuths, absence of delayed reflections, absence of crosstalk, etc etc).

Finally, you must keep in mind that the HRTF at +30° is radically different from the HRTF at -30°, and when we listen to speakers in a room, we receive both in both ears at the same time (and then, milliseconds later, a host of additional strong early reflections from a variety of other angles).

Yes, that was the question. Is DF is an average over all angles that sound comes from when perceived in-ear or is it taken at some specific azimuth from ear?

Exactly, it essentially gives the FR at the eardrum independent of azimuth.

Oops, just me being cryptic again. I meant that Harman might've just chosen to show that DF target is not proper, albeit phrased it... strangely. Made sense to me, hah!

Got it I think Harman may have been thinking along your lines, but that in the end, the result they obtained from the in-room steady-state measurement was barely different from a DF response, anyway.

 

[Feelas]

That will be true in some cases, but completely recording-dependent. It all depends on the relative spectra of direct sound and reflected sound in the recoridng, and the nature of the deviation in FR upon reproduction.

A deviation in the reproduced FR may just as well make a recording sound less live, depending on these factors.

Yes, of course, that'll depend heavily on what type of tilt we're seeing - albeit the distortion of overall imaging & depth factors will surely be a fact.

As I argued before, the starting condition is not a target. It is just a starting condition, and the primary goal should therefore be for it to be as smooth as possible. If you wanted to truly randomise the starting conditions, you might decide, for example, to introduce random tilts to the flat DF response, and I think this would actually be a way to improve the experiment (since starting from a flat DF or steady-state response is arguably confounding). But in all cases, I would prefer a smooth DF starting condition.

And yet I'm not arguing against smoothness; it's perhaps one of the most important factors. I believe I've seen @Sean Olive mentioning that the starting shelves were indeed randomized at each trial, maybe even at this specific topic.

I do see where you're coming from though. You're saying that, when we listen to speakers in a room, the first-arriving sound we hear has (ideally) a +/-30° azimuth, and you think that this should be reflected in the starting condition. I can see the logic to that, but I think that a typical HRTF at +/-30° is going to be way too erratic to be of any use. Also, I don't think it's necessarily clear that headphones are supposed to emulate the response of loudspeakers at +/-30°, let alone the question whether they even can, given all the fundamentally different conditions under which they operate (different azimuths, absence of delayed reflections, absence of crosstalk, etc etc).

Well, I think that it's important to note that merely changing the angle of attack in the same room will change the direct to indirect ratio and result in a different FR, since direct sound will be subject to HRTF variation w/ angle and indirect (being heavily variable) might be subject to DF-like averaging. That'd show one more glaring point against an universal curve, since azimutal differences will affect what preferences one has.
Yet, if we're really dealing w/ DF-like (far-field) scenario, I am sorry to hear that probably most I've written doesn't apply Oh well!

Finally, you must keep in mind that the HRTF at +30° is radically different from the HRTF at -30°, and when we listen to speakers in a room, we receive both in both ears at the same time (and then, milliseconds later, a host of additional stron early reflections from a huge variety of other angles).

... and here I see a strong argument to consider the situation more of a diffuse field. Well, somehow now that makes sense!

Got it I think Harman may have been thinking along your lines, but that in the end, the result they obtained from the in-room steady-state measurement was barely different from a DF response, anyway.

Guess it's just me getting trying to get into their state of mind, not the other way around, haha! After all that, I can see Your point.

 

[raistlin65]

Perhaps since it was derived scientifically, and the preference range is highlighted, not hidden from ones interested in it for what it is. The group might be small, but in the end I believe we have some semblance of what sounds correct, since we're living in it (and not in headphones) since we're born and since people exist, in general. That's why both free field & diffuse field make no sense - nobody lives either in anechoic nor in extremely live space.

Behind "merely the preference" lives a bigger, more objective matter of what humans in general consider right - and that's what we should be looking for. I think that the difference between trained and untrained listeners stems only from the lack of experience, and we'd converge on some specific curve, given enough time for untrained becoming trained.

You misunderstood the question.

You have explained why you want to equate need and preference. But you haven't explained why you need to. Just stick to using the term preference since there is no disagreement that's what HTR describes. There is no need to use the term neutral when there is obvious dissension with your attempt to do so.