Human hearing and measurement

Audio Gear Talk HQ Player
81

I dispute that, it is one of the original claims by JWM that have been debunked. Neither do any ´system-related distortion´ of cone drivers exist and no-one could so far explain to me what kind of inherent distortion that should be. Nor is the MST anyhow different from a conventional electrodynamic transducer in terms of voicemail mass, movement or diaphragm behavior.

In contrary, there are indications that the particular geometry of the MST´s voicecoil and diaphragm induce certain phenomena like distortion, resonance, out-of-phase movement and directivity errors originating from the diameter of the excited diaphragm area and the acoustic behavior of the material.

The two main factors that set the MST apart from other dynamic drivers are the voicecoil size (3" if I am not mistaken) and the fact that it is a suspension-less, surround-less bending wave diaphragm with high inner damping. This has implications particularly for its dynamic behavior (distortion, compression, resonances) as well as the directivity. These factors seemingly explain why the MST sounds very different.

That is nothing unusual and explainable by the fact that a large voicemail is combined with the aforementioned bending-wave diaphragm which is also very light.

There are other drivers covering a similar or larger frequency range (fullrange drivers as well as 1.5-way or coaxial designs), there are other bending-wave concepts and there are suspension-less or surround-less drivers on the market. They all sound very very different from the Manger hinting that this is not the main reason for its particular sound but rather other factors such as the directivity.

Concepts with AMT midranges exist and they sound very very different from the Manger.

What is the inherent problem of the cone drivers that the Manger does not bring?

I agree that it sounds different, but not necessarily more ´real´ or ´natural´. My theory is that the different impression largely originates from the directivity as the MST is seemingly a resonance-laden, bending-wave design in the middle of its frequency range (1-4K) combined with conventional pistonic behavior below and a very large ring-radiator towards higher frequencies. It is pretty visible in Stereophile´s polar plot:

1119ManP1fig5

I suggest an interesting experiment: Listen to the Manger in a near-field environment in a very damped studio control room (or an anechoic chamber) and subsequently in a normal living room. They sound like they are two completely different types of loudspeakers supporting my theory that it is the directivity which is defining the particular sound.

It is interesting you are mentioning the Kii 3 as an example of utter discrepancy from the MST’s sound as that would mean debunking two essential claims made my Manger: That time-coherence matters and that the moving mass of pistonic drivers sounds unnatural. The concept of Kii is perfectioning the former via DSP to at least the level of the MST and the latter thanks to active cancellation similar to open-baffle or cardioid concepts. If Manger´s claims were true the Kii had to sound very similar in regard of timing, impulse response and dynamics.

But it does not, you rather describe it as an antithesis ´natural´ vs. ´soulless´ (I would not subscribe to the descriptions, but we can agree on the existince of huge differences). Maybe your impression also has to do with directivity as the Kii 3 represents an antithesis to the Manger par excellence: constant medium-high directivity in the midrange and slightly decreasing DI with the tweeter kicking in (Kii) vs. omnidirectional/diffuse soundfield in the midrange plus a steep step towards very high DI in the treble area (Manger).

I do not see any hint this was the case with the success of Kii. I guess people were rather easily convinced that the concept of actively suppressing reflections and booming in a compact speaker design was what they really needed in their rooms and I fully support this idea.

82

You haven’t ‘explained’ anything actually. You’ve only made some claims, maybe parroting what the manufacturer claimed in the Stereophile measurement response.

I’ve seen this with some companies and Stereophile measurements - they respond by saying JA is measuring it wrong, without any hint of advice of how JA should have measured it. Or explaining how they measure.

I’m actually open minded and want to genuinely hear an explanation - but nobody ever explains it !

3 Likes
83

As a side note, Vandersteen has been making time coherent analog loudspeakers for quite a while. There are of course number of other time coherent approaches such as electrostatics (with high vertical and horizontal directivity).

1 Like
84

You‘re quoting me out of context, so I‘ll rephrase:
Triggered by the link to step responses by another member in the context of the thread topic, I‘d actually been trying to convey, that, although there is plenty anecdotal testament to it, it’s very unlikely for anyone to be able to hear electronics’ time domain minutiae, when a speakers‘ gross misbehavior in this regard seems obviously inaudible.

Secondly, a step response graph is merely one of several possible representations, like frequency, phase, impulse, energy time, group delay, etc responses of mathematical transforms of one measurement with a maximum length sequence stimulus, and not a measurement in itself.

Or, not being native English speakers, are we just getting lost in semantics here?

1 Like
85

As a native English speaker, I think all of you are doing just fine. We talk past each other constantly and have to restate and clarify. Even if I put on my editor’s hat (never mind the semantics), the only nit would be “a speaker’s” should be possessive singular while your just previous “electronics’” being possessive plural is both correct and more sophisticated than most native writers can typically summon. And all of us would dash off a note and not worry over such trivialities.

Break over: carry on! This is fascinating.

86

Fully agree, and I did not mean to imply that you had been making any claims on the importance of step response. It was rather a contribution to the Manger-related discussion.

From mathematical point that is true. In the real world, step response graph is showing merely limitations of the complex response of a loudspeaker like the limited time window of the phase response measurement (which should include at least a 20ms window or more for usual speakers). Plus underrepresenting interesting aspects and overrepresenting inaudible ones.

In the time domain, a non-capped group delay graph over frequency and a waterfall plot without time/frequency windowing would be interesting.

87

So, which part of my explanation on (partly) bending wave transducers, diffuse soundfields and parts of the diaphragm being out-of-phase with each other did you fail to understand?

It is actually a bit funny that you speak of ´parroting´, as I am seemingly the only one here having made measurements plus several listening tests with MST equipped speakers myself. And the result was that the frequency response at a single position did not show any correspondence with a result of a listening test which is explainable.

That does not mean I am preferring these speakers or defending the manufacturer. I did not refer to the Stereophile dispute and I am well aware of manufacturers or distributors claiming their devices were ´wrongly treated in the lab´ which I usually do not take seriously. In this case the manufacturer has a point which I can confirm whith having made any claims that the loudspeakers show perfect measurements or are sounding perfect.

I am neither associated with the manufacturer nor the reviewer but from my own experience I can say that any loudspeaker showing tendencies towards bending wave behavior or parts of the diaphragm being out of phase with each other should not be measured at a single position but rather in a 2-dimensional window. Sometimes a binaural dummy head measurement in an anechoic chamber makes sense for that but poses the problem of not being comparable with other measurements.

That is the case with an astonishingly wide range of speakers, from true bending-wave concepts (Goebel, Manger, German Physiks, Walsh) to suspension-less planar or ribbon speakers (Magnepan, early ML), drivers with mechanical crossovers (some Thiel, Pfleid, Naim Ovator) or huge lightweight, flexible cone fullrange drivers (e.g. Lowther, Voxativ and many others).

Will try again. The MST is partly a bending wave transducer. That means in certain frequency bands parts of the diaphragm are out of phase with each other leading to a cancellation under specific angles, most noticeable 0deg on axis. It is easy to understand if you look at the polar plot at 1.7K and 3.5K as these are the two narrow bands showing an almost circular or concentric out-of-phase behavior. At 1.7K you have maximum cancellation on axis and addition due to in-phase behavior around 30-50deg, at 3.3K you have the exact opposite, maximum cancellation under 30+deg angles and addition hence maximum level on axis.

Any single-point measurement at 0deg would not be even a far representation of what one hears in the listening room. This is true not only to the aforementioned narrow bands but to the whole area of bending-wave behavior which gives way to a more ring-radiator-like directivity the higher you go in frequency (probably due to diaphragm damping or inertia thereof.

88

Can you Please show us the data - your measurements?

And explain how you measure and how it is different … that is the explanation I was after :wink: So far in all that text above we have this below. And it is interesting! Any photos of your measurement setup? So you have 2 mics and how do you combine or splice the data?

89

Sorry Arindal, you’re wrong here.
I had problems with some room modes in my living room system with my Manger P1 at the time, and passive bass traps were out of the question due to the living room situation. So I spent a lot of time measuring the room and the speakers using REW and various measurement methods. My hopes of achieving an improvement with the convolution filters generated for different house curves in Roon were not fulfilled. Unfortunately, I didn’t achieve a result where the overall sound was better with the digital filters than without any filters.
I’ve had the P2 with the new W06 MST for a year now which is a noticeable improvement. Since then I’ve been trying to persuade my wife to install PSI Audio AVAA C214, unfortunately so far without success.

90

I am sorry, did not mean to ignore your efforts, I was referring to measurements in anechoic conditions as a reference point for judging the speakers, not the situation in your room.

So were basically coming to the same conclusion that standard measurements at a single point, no matter under which condition, will not accurately represent the tonal behavior of the speaker nor can be used to calculate working correction curves.

Am I getting it right your problems are mainly in the lower frequency regions? Did you try applying only calculated correction filters under 250Hz and leave the rest uncorrected?

That is a solid method of ignoring miscalculated filters. The only thing you might want to do as a second step is setting a filter for smooth transition from the region of calculated filters to the bypassed bands. As this frequency band of lower midrange is also the source of typical audible direcitivity errors, you might anyways want to experiment with such a PEQ, I suggest:

f=350Hz, Q=0.7, level range=-2…-0.75dB

That is in such cases smoothening the audible step in directivity as well.

91

I will check what was published and is still publicly available. That I used to do such measurements does not mean they are my intellectual property.

If traces of narrow-banded cancellation, diaphragm-induced comb-filter effects or anything like that occurs, a binaural measurement with a dummy head in an anechoic chamber under nearfield conditions is usually the way to verify the audibility of the dips. Some labs might also increase the smoothening tolerance but this is what I found not to be useful.

Dummy heads usually have 2 mics. Please note that the result will not be comparable with any other FR under room or anechoic conditions. It is solely for comparison with other speakers and not to judge the candidate.

Manufacturers, reviewers and scientists alike seemingly are pretty reserved when it comes to publishing such data as it might be misinterpreted easily. That is particularly true since amateur measurements are publicly discussed and lots of people obviously are more hunting for even lines on a graph than interpretable measurements.

To give an idea about the importance of narrow-banded, narrow-angle interference and how to measure/interprete them, there were some things publicly discussed concerning edge diffraction. It is a different topic but there are similarities to out-of-phase behavior or parts of the diaphragm as these phenomena also spoil the on-axis FR but most of developers agree this should not be corrected. You might want to translate and read this including some measurements (it is in German so Google translator should do the job):

If you want to dive more into the topic of non-pistonic diaphragms, BMR are a good example, as measurements on ASR have shown (same manufacturer, different geometry and drivers):

Recommend to take a close look at FR under multiple angles and the polar diagram. You see a pretty similar behavior of the BMR around 900 Hz and 1.7K compared to the Manger MST which indicates that inner and outer diaphragm rings are out-of-phase in very narrow bands in both cases.

Despite from on-axis FR graph getting ugly I have not noticed any downsides in the listening test, either.

92

Other than measurements of FR, there’s nothing scientific here by the 2 Engineers.

Just that they did listening tests and things sounded more natural with the very slight scoop at ~3kHz, with longer listening sessions.

I’d be interested how they performed these longer listening sessions in a scientific way.

I’m not disputing by the way. I just did this A/B testing with headphones, to generate a custom PEQ for my headphones: https://peqdb.com

The resultant PEQ definitely sounds more exciting and lively and ‘better’ but only for 30 seconds, similar to the length of the test music samples.

I get listening fatigue with this ‘winning’ PEQ after a minute.

Listening fatigue is definitely a thing, even if the EQ obtained is the one you preferred (in short music sample A/B testing!)

93

That is completely congruent with my vast experience.

I did not show this example to deliver a scientific proof but to show there are problems in loudspeaker design which you cannot judge from a single point or angle nor draw any conclusions on sound quality out of a few measurements. So having a kinked on-axis FR might be a legitimate outcome of tuning a loudspeaker to desired natural tonal balance in the listening room.

In any case I think on-axis FR measurements are largely overestimated, they do not tell much about the speakers performance in a room and if ever the two are matching, you can easily apply DSP correction methods.

I found it more useful to look at a loudspeaker´s behavior in a larger two-dimensional window and the overall directivity behavior. These do not allow precise predictions on how it will sound, either, but you get a pretty accurate understanding about room compatibility and if any problems might occur that can not be corrected by DSP.

Surprisingly the latter is pretty common in many cases the on-axis FR has been optimized to the max for optical reasons or you find such speakers being praised by self-declared ´audio objectivists´ trying to draw straight lines into the graph instead of doing a listening test.

What is the main frequency band making the difference you are describing? Do you experience listening fatigue with these headphones with some tracks when not enabling any EQ?

Just checked the page and found their method to be utterly pointless as they seemingly rely on music samples with which you cannot judge anything nor optimize a DSP.

That is exactly what I referred to when the discussion came to A/B testing with slight differences in level. If listeners are prone to experience fatigue from a given setup/music they might prefer the lower level or an EQ setting which is slightly reducing the level of frequency bands contributing to that fatigue. Being exposed to short-term samples without high risk of fatigue they usually prefer the louder variant.

I am pretty sensitive to listening fatigue as I was listening up to 8h a day while not very much liking the solution of other recording engineers to put the level down in order to reduce fatigue. So I anyways choose loudspeakers and headphones by the subjective absence of fatigue giving a vast range of applying EQ filters without ever reaching this threshold of fatigue kicking in.

94

You can upload your own track also.

Well it does quite a bit of EQ in bass, mids and treble regions. All based on my preferences in the A/B tests.

It sounds quite impressive overall listening to my music - except only for a short time.

If I spend more time I could trace which area/s are causing fatigue but I just go back to my preferred EQ and am much happier

Nothing scientific about my method of course

So its an interesting discussion relevant to the thread - has anyone done credible studies on listening fatigue . We’ve all experienced it but it might be difficult to do scientific listening tests? Or measure

95

This thread is literally about human hearing and measurement and you apparently care about unsubstantiated opinion instead of actual science. That seems to sum up quite a lot of the posts on this thread.

I mean we literally have people quoting the sales pitch of DAC company owners here… :joy:

As I said in my first reply to this thread, it’s clear that most people do not understand science. Unfortunately, it also appears that most people don’t want to either as that would conflict with their existing views.

1 Like
96

Nice feature, but what is the point if you yourself do not know which of your favorite tracks lead to fatigue in which narrow frequency band? I could upload a hundred tracks and none would get fatiguing with randomly set EQ filters while the 101st hits my ear with a laser beam in a narrow frequency band the software did not foresee nor offer during comparison tests.

That was my point. The software is randomly changing pretty broadband filters in level unrelated to the actual headphone models and test track, asking for the listener´s preference in a direct or indirect comparison.

If you would use a 3-band graphic EQ embedded in the app of pretty affordable in-ear monitors, you would with a bit of training most probably come to the same result much quicker.

I am not really surprised by this result. If you want to improve the SQ of headphones or adopt them to your individual perception, it takes much more than just some random broadband (Q<=2) PEQ filters.

You might have to select half a dozen of tracks out of hundreds which are most prone to give you the impression of fatigue and subsequently precisely nail down the narrow bands that cause this fatigue. We are talking about notch filters (Q>10) and a very precise tuning of the center frequency. I personally like to adjust PEQ by reserving the assumed level change and try to nail down the annoying or boosted frequencies in a first step, eventually setting the filter level back to the desired attenuation.

This is particularly critical in the frequency region 2…6K in which the ear is most sensitive and many narrow-band acoustic phenomena are in play.

Once you have identified few narrow bands which are causing fatigue or harshness you have to bring the neighboring bands of harmonic series into balance (f0.5, f0.666, f1.5, f2 and alike) which is a delicate but necessary process in order not to kill overall tonal balance and dynamics with your main PEQ filters.

Such a process requires hundreds of single filter constellations which have to be applied on dozens of tracks so several 1,000 quick listening tests. A filter set which takes away harshness from one track might darken or dull the tonal balance of another one or take away its dynamics. Trained recording engineers might tell you the result of a single test after a few seconds and several bypass A/B comparisons, but if you are not trained it takes a long time and might result in frustrating circular errors once you reduce harshness of a system, then you notice it has lost treble resolution and dynamics, and reverse the EQ filters…

It is already difficult and time-consuming to do such things from a personal perspective and without scientific methods. If you expand this to several individuals it actually gets very complicated as each and everyone of them might be having completely different perception.

Measuring is very useful for nailing down frequency range and cause of particular problems which you have encountered during listening tests. For headphones it is pretty useless when it comes to predicting how something will actually sound.

97

I am also pretty critical when it comes to claims made by manufacturers or reports on audible differences which have no substantial explanation or technical base.

Nevertheless we should admit that the scientific base for subjective listening tests is actually very thin. A lot of claims have never been remotely verified or fall under scientific principles, so from scientific point of view the fair answer would be in most cases ´we do not know it and no-one has every tried to verify it in a scientific process´.

Nevertheless there is a lot of knowledge based on solely experience combined with applying what we actually know for sure about acoustics. I respect people whose claims have been verified in practically applying them and are congruent with both the scientific base and common sense alike. Some people might call it anecdotical but that applies to many claims, even those coming in the form of AES papers of study-like experiements.

I do not see any scientific findings colliding with what I consider to be solid practical experience, but nevertheless I would be interested to know which examples you mean by that. And I am not talking about obviously esoteric claims.

98

The „science argument“ shouldn’t be always used to only diss or look down on people.
If your are so familiar with science then educated people with it and explain the relevant science to them.

99

I don’t see it that way. When it comes to differences, double-blind ABX tests are pretty good at removing bias and determining the physical limits of our hearing. That is the ultimate practical test and does not require any particular understanding of the hearing process. If such tests don’t produce a statistically significant result, then the burden of proof lies with the side that claims such differences are audible.

1 Like
100

There are many efforts in this forum to do just that. Assuming that all concepts - some of which are quite complex - could be explained in intuitive terms (Einstein believed they could), the majority of these attempts end up in ridiculous claims like:

  • You’re just jealous because you can’t afford my system/don’t have my ears.
  • You can’t possibly know unless you try it yourself.
  • Science is some kind of elitist activity.
  • Science is purely theoretical and doesn’t apply to real world.
  • It’s a hobby, so I can’t be wrong.

Nobody is looking down on anyone, and the science argument (unquoted) should always be considered.

3 Likes