Why evaluating the sound of a single speaker is essential

[Floyd Toole]

So I tend to agree with your assessment that reviewers saying 2-3kHz is a bad range, were probably due to bad implementations.

My perspective is that the historic problem in this range is the mismatched loudspeaker directivities at crossover, especially in popular 6" and 8" two-way systems, combined with tweeters that often generated distortion immediately above crossover, especially before properly designed waveguides were discovered.

 

[Floyd Toole]

I'd posit that how the contours change with level may perhaps be more relevant than the shape at any single leve

Agreed. If you get into the ear's sensitivity as a function of frequency, you might as well get into sensitivity as a function of direction as well, because we know that the direct sound is a major factor in sound quality and it arrives along a specific HRTF - which will be different for each of us. Why not admit that these are life-long characteristics we live with and adapt to.

 

[MarkS]

Pink noise is the most revealing signal for detecting the presence of resonances in multiple-loudspeaker comparison tests. That's all.

Thank you for the reply. But I stand by my contention that out-of-phase pink noise (that is, mono pink noise with one channel with reversed polarity) on a stereo pair of speakers is more revealing of speaker directivity than any single-speaker listening test. I haver never seen any published results of such a test, if you know of any I would appreciate the reference. Thank you.

 

[gnarly]

My perspective is that the historic problem in this range is the mismatched loudspeaker directivities at crossover, especially in popular 6" and 8" two-way systems, combined with tweeters that often generated distortion immediately above crossover, especially before properly designed waveguides were discovered.

Thank you ! That sounds like messed up acoustic designs from the gitgo.
And maybe, (just speculating now) wherever the crossover frequency happened to fall, reviewers might have blamed that crossover frequency, not knowing the real underlying issues.

For me, a good acoustic design would not have directivity mismatches, or a tweeter driven into undesirable excursion.
But I guess I've been into DIY long enough, that I'm out of touch with consumer audio. Do hope those historic problems have faded away.

 

[Duke]

Back in post 599 the curve in (b) by Stevens was done using bands of sound in a normally reflective room and arguably is the most relevant to this discussion.

Thank you! It looks like the ups-and-downs of the equal-loudness contours are considerably less pronounced in the Stevens curves.

Do we know what the directivity of the source was? I'm assuming that would matter.

 

[Heinrich]

... I'm assuming that would matter.

In post #597 you introduced a speculation on why an x-over at 1kHz would be superior, supporting an anecdotal observation, that in itself wasn't conclusive. There is no causation chain from A to Z, actually. And now you argue about differences between equal loudness curves. As if they were objective data, rather than an aggregate of many individuals' subjective experience. And as if there was a relation between A and Z to begin with.

... not knowing the real underlying issues.

Exactly, speculation over speculation, dismissing the very basics of scientific investigation deliberately, or by just not knowing. Sometimes it reads like throwing a wild assumption at a problem, and see what might stick. You might do that, but it shouldn't be the only method. At least if it peels off like water from teflon, one should acknowledge the fact.

 

[Floyd Toole]

Thank you for the reply. But I stand by my contention that out-of-phase pink noise (that is, mono pink noise with one channel with reversed polarity) on a stereo pair of speakers is more revealing of speaker directivity than any single-speaker listening test. I haver never seen any published results of such a test, if you know of any I would appreciate the reference. Thank you.

I did an elaborate stereo vs mono evaluation, described in Section 7.2 of the 3rd edition, and in my 1985 JAES paper. A review of research on the topic in 2009 concluded that I still stood as the "most relevant to date". Listeners heard spatial/directivity effects in mono listening.

Nowadays a spinorama yields data on directivity of the loudspeaker, but not the interaction with your listening space.

 

[Duke]

In post #597 you introduced a speculation on why an x-over at 1kHz would be superior, supporting an anecdotal observation, that in itself wasn't conclusive.

I see a possible correlation between perception and the equal loudness curves, and you do not. Is that a fair observation?

And now you argue about differences between equal loudness curves...

You can look at post #599, illustration (b), and see for yourself the differences between different sets of equal loudness curves. I'm not sure what there is to "argue" about... ?

 

[Heinrich]

I see a possible correlation between perception and the equal loudness curves, and you do not. Is that a fair observation?

Of course, there's nothing unfair with taking different perspectives. May I ask what the correlation could be? Let's take the perception part as a given, fair? Now, what do the curves bring to the table? The curves as such are not an auditory, but more a visual feat. From looking at these I may conclude something. What is it?

I suggest that a human, when exposed to sound may perceive it differently, depending on the physical intensity. If my recollection doesn't trick me, to put that very fact at display is the intention of the curves. The correlation is there, in a mental model of "hearing", it refines that model in one direction, nice (so far we perfectly agree). But what does it tell in regard to speaker design, especially when considering--as you did, the crossover in non-coaxial speakers?

You say there's a particularly sensitive range. Sensitive to what parameter of the soundfield, and how does that correlate to engineering? I'm still not convinced that there is a clear path from the shape of the curves to auditory sensations related to a soundfield's properties originating in the crossover design.

You can look at post #599, illustration (b), and see for yourself the differences between different sets of equal loudness curves. I'm not sure what there is to "argue" about... ?

None, only that as long as the avove is not clarified (at all, me thinks), there's no reason to argue about minor differences in distinct representations of a basic observation.

nb: when playing with LLMs, you'll find that these give an answer like your's like "Because the threshold of hearing is minimal, the range 2..5k is most sensitive, hence is easily disturbed by anomalies in speakers." One additional question ("On (3), that is my point. How could those curves support a statement like (3)? ") is enough so that the LLM turns 180° like "Nope, correction, different psychoacoustic domain". Even the LLM finds, that it is a misuse of terms; "sensitive"--to what? Of course, sensitive to the presence of sound at all, the threshold of hearing is the lowest. But what else, and it there was anything, how to derive from the curves? I'm desperate in evaluating the logic here.

See also this thread, https://www.audiosciencereview.com/...way-point-source-qualities.63707/post-2336923

Someone's argueing, that a KEF R3 is inferior to an LS50 because of a crossover in the 350ies range.

 

[Duke]

You say there's a particularly sensitive range. Sensitive to what parameter of the soundfield, and how does that correlate to engineering?

According to the ISO 226-2003 equal loudness contours, at 80 dB, the ear is about 4dB less sensitive to sound pressure level at 1.5 kHz, and about 3 dB more sensitive to SPL at 3 kHz, relative to its sensitivity to SPL at 1 kHz. (This is normal and should not be "corrected for" by equalization.)

I think this is an indication that the ear would be less sensitive to SPL-based anomalies (in particular peaks, and spectral discrepancies between the direct sound and the reflection field) in the 1.5 kHz region, and more sensitive to such in the 3 kHz region.

I'm still not convinced that there is a clear path from the shape of the curves to auditory sensations related to a soundfield's properties originating in the crossover design.

A spectral discrepancy (SPL-curve discrepancy) between the direct sound and the reflection field is the type of soundfield anomaly we are most likely to encounter in the crossover region, due to differing driver directivities.

 

[Heinrich]

I think this is an indication that the ear would

You repeat yourself, me thinks. I won't do the same. Where is the step by step connection between the Fletcher/Munson curve and being, as you put it ...

... sensitive to SPL-based anomalies (in particular peaks, and spectral discrepancies between the direct sound and the reflection field) in the 1.5 kHz region, and more sensitive to such in the 3 kHz region.

We could also wrap up the conversation before we get into a second cycle ;-)

 

[bmc0]

I think this is an indication that the ear would be less sensitive to SPL-based anomalies (in particular peaks, and spectral discrepancies between the direct sound and the reflection field) in the 1.5 kHz region, and more sensitive to such in the 3 kHz region.

I'm not sure there's good reason to draw this conclusion. As I pointed out in post #620, looking at the same data (ISO 226:2003) from a different angle leads to a different conclusion: the region from 1.25kHz to 6.3kHz requires a larger change in SPL for the same change in loudness compared to frequencies outside this range. Does this mean anything for crossovers? Again, I don't know.

It may be important to consider as well that ISO 226:2003 describes the steady-state behavior of the auditory system to pure tones. Music is not pure tones, is not spectrally flat, and is full of transients. Cochlear compression is not instantaneous. What might be the implications for crossover audibility? Yet again, I don't know, but the point I'm trying to make is that given the complexities of hearing, I don't think one can reasonably assume that because equal loudness contours indicate that a pure tone (or noise band, for that matter) at frequency X is perceived as more/less loud than Y at the same SPL, it follows that X is a worse/better place for a crossover.

 

[Duke]

I'm not sure there's good reason to draw this conclusion. As I pointed out in post #620, looking at the same data (ISO 226:2003) from a different angle leads to a different conclusion: the region from 1.25kHz to 6.3kHz requires a larger change in SPL for the same change in loudness compared to frequencies outside this range. Does this mean anything for crossovers? Again, I don't know.

It may be important to consider as well that ISO 226:2003 describes the steady-state behavior of the auditory system to pure tones. Music is not pure tones, is not spectrally flat, and is full of transients. Cochlear compression is not instantaneous. What might be the implications for crossover audibility? Yet again, I don't know, but the point I'm trying to make is that given the complexities of hearing, I don't think one can reasonably assume that because equal loudness contours indicate that a pure tone (or noise band, for that matter) at frequency X is perceived as more/less loud than Y at the same SPL, it follows that X is a worse/better place for a crossover.

You may be right. I've been wrong before and will probably be wrong again.

 

[Salt]

Well played...
Double - means +,.
Where I would agree: the most sensitive area is 1 to 4 kHz, and that is where FR, phase, diffraction etc. should behave as good as possible.

 

[Thomas_A]

My take (as I've probably mentioned many times). the 1-5 kHz region; never ever have a dip 1-2 kHz and/or a peak 3-4 kHz. This applies on-axis, lateral dispersion, power response. I prefer the opposite within +/- 1.5 dB since it subjectively sounds better to me. In stereo.

 

[Duke]

Where I would agree: the most sensitive area is 1 to 4 kHz, and that is where FR, phase, diffraction etc. should behave as good as possible.

My take (as I've probably mentioned many times). the 1-5 kHz region; never ever have a dip 1-2 kHz and/or a peak 3-4 kHz. This applies on-axis, lateral dispersion, power response. I prefer the opposite within +/- 1.5 dB since it subjectively sounds better to me. In stereo.

Maybe I'm mistaken to see a correlation between such observations and equal loudness curves; that certainly seems to be a disputed correlation to make. My observations are in the same general ballpark as both of yours.

David Griesinger finds that time coherence (phase coherence + arrival time coherence) is desirable north of 1 kHz; his field is concert hall acoustics and psychoacoustics but I think the same psychoacoustic principles apply. I recall Earl Geddes saying that multiple sensitivities peak around 4 kHz, including sensitivity to diffraction.

So perhaps there is room for some consensus on the frequency regions that are most critical, even if the reasons why are disputed.

 

[bmc0]

I've been wrong before and will probably be wrong again.

Ditto. Anyone who has spent any time on a complex subject and claims otherwise is surely lying.

Edit: This paper is not very recent (1966), but might be useful: Review of Research and Methods for Measuring the Loudness and Noisiness of Complex Sounds. Figs. 13 and 14 are "equal annoyance" contours for octave and 1/3rd octave bands and they certainly have a different shape than equal loudness contours.

 

[Thomas_A]

Maybe I'm mistaken to see a correlation between such observations and equal loudness curves; that certainly seems to be a disputed correlation to make. My observations are in the same general ballpark as both of yours.

David Griesinger finds that time coherence (phase coherence + arrival time coherence) is desirable north of 1 kHz; his field is concert hall acoustics and psychoacoustics but I think the same psychoacoustic principles apply. I recall Earl Geddes saying that multiple sensitivities peak around 4 kHz, including sensitivity to diffraction.

So perhaps there is room for some consensus on the frequency regions that are most critical, even if the reasons why are disputed.

Yes loudness curve match, but is it just a coincidence? So do the comb filtering dips and peaks in stereo triangle setups. And also in speakers where the crossover is around 2.5-3 kHz, when the tweeter energy takes over from a more beaming woofer (solvable though). Psychoacoustic; changes in this region can affect "image depth" ("laid-back or forward sound"). Peaks in frequency response, especially tweeter range, makes me reveal the speaker position more easily, decreasing the illusion. It is both general tweeter level but specifically 1-5 kHz.

 

[sceptical1]

I doubt, very much, that a circle of confusion or equal loudness curve is something I need to be concerned with since neither have been a problem when listening for the last 55 yrs. Either I haven't been paying attention or everyone else thinks confirmation bias does not include what you read.

 

[kyuu]

I doubt, very much, that a circle of confusion or equal loudness curve is something I need to be concerned with since neither have been a problem when listening for the last 55 yrs.

That's great. I'll make sure to file away your 55 years of listening experience where it belongs relative to actual research.