[VNandor]

The difference would need to be more than 3dB to be at all audible, and anything under 5 would be pretty small. I don’t see any difference between the responses you’re citing there in the last chart.

Differences under 3dB in commercially recorded music are too small to matter. 3dB with test tones is different than 3dB in a narrow band of frequencies with recorded music. Numbers are meaningless without context. It’s a good idea to research thresholds of human perception and experiment with sound editing programs so you know what the numbers actually sound like.

Although I agree a high-Q notch or peak would be imperceptible with most music, it's a moot point because a lot of music fills up the spectrum fairly uniformly and the curves @ADUHF is talking about are showing a wide band attenuation of frequencies. The differences between these targets are defintely enough to cause a change in sound that could be easily heard. Maybe you should take your own advice an try to dial in some EQ. Just by eyeballing the chart, the high frequencies start to gently roll off at about ~1kHz and goes to ~-2.5dB at ~10kHz. The bass seems to be differing below ~200Hz and the difference eventually maxes out around ~2.5dB at ~60Hz. I think that would be a fair representation of the differences between the two responses. There's no way you wouldn't hear any differences if the music you use contains enough high and low frequencies to begin with. I know I do.

 

[bigshot]

3dB is the threshold where I generally can hear clear differences with EQ under music. Anything under that, I'm just splitting differences. Granted, 3dB at 2kHz is different than 3dB at 15kHz or 40Hz.

Maybe we are talking about different charts again. I'm referring to the one with the two very close curves in the middle, and the wide variances above and below it... preferred in-room black headphone cyan. If the overall levels were matched, with headphones you would get a gentle boost of what looks to be just under 3dB between 200Hz and 600Hz. That wouldn't be "night and day". It would be just on the edge of being audible.

 

[ADUHF]

I think the sensitivities can be smaller than 3 dB with broader ranges or bandwidths. I haven't looked at the research on it though (as bigshot suggested). So this is just based on my own impressions. I think it probably is more difficult to hear the changes over very small bandwidths though. So I would tend to agree on that. Using a bracketed, over-under type of approach like I described earlier can probably help to improve that though imo. At least somewhat.

The degree of sensitivity and its significance in terms of sound quality improvement may depend on the specific application though. And whether you're using the headphones more for entertainment/enjoyment, or more for critical listening and evaluation purposes, for example. And also on an individual's general pickiness about such things, I suppose. So it could be somewhat debatable in that sense.

I use my headphones for both entertainment and also critical listening. And I like to hear as many of the small nuances and details in the music as possible, for both these purposes. So therefore I feel it is a worthwhile expenditure of my time and effort to attempt to iron out as much of the unwanted unevenness and imbalances in my transducer's response as my hearing and other facilities will allow, using various means and approaches. Including physical modifications in some cases.

For others this may be a complete waste of time and energy though. And/or simply a distraction from the enjoyment of their music.

 

[VNandor]

3dB is the threshold where I generally can hear clear differences with EQ under music. Anything under that, I'm just splitting differences. Granted, 3dB at 2kHz is different than 3dB at 15kHz or 40Hz.

Maybe we are talking about different charts again. I'm referring to the one with the two very close curves in the middle, and the wide variances above and below it... preferred in-room black headphone cyan. If the overall levels were matched, with headphones you would get a gentle boost of what looks to be just under 3dB between 200Hz and 600Hz. That wouldn't be "night and day". It would be just on the edge of being audible.

We are talking about the same charts. You are just off about the difference. It's not night and day but it's not "too small to matter" either and most certainly not too small to be heard. Here's the same chart with the cyan curve shifted up so it matches the black one at the low and high frequencies. Man, I'm way too good at paint.

 

[bigshot]

Shift the speaker line up a little more to split the difference. That would be closer to equal volume. I doubt that would be a clearly audible difference with most music. I'm not arguing it would be imperceptible. I'm saying it wouldn't matter much at all. Create the curves, apply them to a music track, normalize and listen to them side by side. It isn't going to be much of a difference. In fact, after twenty or thirty seconds, your ears would easily acclimate to a difference that small and that gradual.

In general 1dB is the JDD with tones and 3dB is the JDD with music. And that is for overall volume level. It would probably be higher for just a section of the frequency spectrum, with 2-5kHz being the range where ears are most sensitive.

I'd suggest that you try to create those curves and listen to them. I think you'll find that the difference is just barely detectable if it is detectable at all.

 

[VNandor]

I'd suggest that you try to create those curves and listen to them. I think you'll find that the difference is just barely detectable if it is detectable at all.

I already did in the way I described it in my post before which is why I insist the difference is big enough to care about. I created one curve which is meant to be the difference between these two. I did the level matching separately from the EQ because the amount of gain needed to do a proper comparison depends on the music as well, not only on the EQ curve. The difference was not only enough to be heard, in my opinion it was enough to change the music in a meaningful way.

 

[redrol]

IMO within the Pinna gain area, differences less than 3db can be heard. Thats the most sensitive part of human hearing. Im massively picky with mids.

 

[bigshot]

Yes but if you look at the chart, the difference is centered between 200Hz and 1kHz. From there up, it’s a slow roll off to 3. Also I need to remind everyone again. We are talking about music, not tones.

 

[ADUHF]

We are talking about the same charts. You are just off about the difference. It's not night and day but it's not "too small to matter" either and most certainly not too small to be heard. Here's the same chart with the cyan curve shifted up so it matches the black one at the low and high frequencies. Man, I'm way too good at paint.

This is just my personal opinion based on my own EQ tests. But if you had the ability to easily flip back and forth between the two preferred response curves shown in black and cyan above, I think you could fairly easily hear some difference between the two. Would it make some difference in your enjoyment of the music?... Based on my own EQ tests, I think it would. But can't really speak for anyone else.

When I'm creating a target EQ response curve for my headphones, I will often start with just a flat/constant diffuse field slope. Usually somewhere in the neighborhood of -1.2 dB per octave. And then begin to make adjustments from there.

And one of the tweaks to that slope will usually be the addition of a depression, or Fletcher-Munson type effect in the midrange, generally somewhere on the order of -2 to -4 dB at the approximate center of the frequency response range. Which looks something like this...

This brings out more detail in the lower and higher frequencies, giving the sound a great sense of depth. Which can be especially helpful when listening at lower volumes.

The difference between the cyan and black curves on the above plot is probably no more than about 2 dB at the center of the frequency range btw. Which is not a huge amount imo. But enough to be audible. And to make some difference in the sound quality imo.

 

[ADUHF]

Perhaps a more salient question to this topic though is whether either the cyan or black curves is a good model for a room response. And the answer to that imho is not very.

I've spent the last couple weeks looking at the sound power plots of different loudspeakers here...

https://pierreaubert.github.io/spinorama/

...In an attempt to try to better understand and identify what a neutral response should look like, for both a good loudspeaker, and also a potentially good pair of headphones. And most of the curves look a bit different than the cyan and black curves in the aforementioned graph.

Loosely speaking, they seem to break down more or less into three different categories or shapes: linear, V-shaped, and L-shaped.

 

[ADUHF]

The loudspeakers with linear sound power curves look something like this, where their SP curves roughly approximate a constant slope, between about -1.0 to -1.5 dB per octave...

https://pierreaubert.github.io/spinorama/KEF Reference 5/KEF/vendor/CEA2034.html
https://pierreaubert.github.io/spinorama/Genelec 8341A/ASR/asr-vertical/CEA2034.html

This is the type of shape that most speaker manufacturers would probably like to achieve, but can't. Because it's usually quite difficult to reliably achieve both a flat direct response, and also such a flat/uniform indirect response, without some fairly sophisticated engineering, electronics and calibration. So calibrated loudspeakers like the KEF Reference 5 and Genelec 8341A with a more linear sound power/indirect response are generally more the exception than the rule.

There are other speakers which get pretty close to this type of indirect response though. Including the Revel F328be, F208 and F206...

https://pierreaubert.github.io/spinorama/Revel F328be/Revel/vendor/CEA2034.html
https://pierreaubert.github.io/spinorama/Revel F328be/ASR/asr-v2-20210415/CEA2034.html
https://pierreaubert.github.io/spinorama/Revel F208/Revel/vendor/CEA2034.html
https://pierreaubert.github.io/spinorama/Revel F208/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Revel F206/Revel/vendor/CEA2034.html

And a few others...

https://pierreaubert.github.io/spinorama/Ascend Acoustics Horizon Center/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Dynaudio LYD 5/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Kali IN-8/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/JBL 308P Mark ii/ASR/asr/CEA2034.html

Some compromises in the direct sound are usually necessary though to achieve such a linear sound power/indirect response though. As you can probably see from a couple of the examples above.

 

[ADUHF]

Loudspeakers with a "V-shaped" indirect/sound power response will usually look something like this, with a somewhat more pronounced dip, usually at around 2k (or somewhere in that general frequency range), at the cross-over of the tweeter and midrange drivers...

https://pierreaubert.github.io/spinorama/Adam T8V/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Buchardt Audio S400/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Genelec 8050B/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Revel F35/ASR/asr-v2-20210415/CEA2034.html
https://pierreaubert.github.io/spinorama/Infinity Entra One/Infinity/vendor/CEA2034.html
https://pierreaubert.github.io/spinorama/Revel Ultima Salon2/Revel/harman-v1-2007/CEA2034.html
https://pierreaubert.github.io/spinorama/Yamaha HS7/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Definitive Technology Demand D11/ASR/asr/CEA2034.html

 

[ADUHF]

The "L-shaped" speakers also have a dip at around 2k, at the crossover of the midrange and tweeter drivers. But instead of rising back up, they will tend to stay a bit flatter in the upper mids and treble. The Neumanns (which are generally highly regarded btw), tend to fall more into this L-shaped category...

https://pierreaubert.github.io/spinorama/Neumann KH 80/ASR/asr-v3-20200711/CEA2034.html
https://pierreaubert.github.io/spinorama/Neumann KH 80/Neumann/vendor/CEA2034.html
https://pierreaubert.github.io/spinorama/Neumann KH 310A/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Neumann KH 120A/Neumann/vendor/CEA2034.html

Several of the Genelecs also follow this L-shaped pattern...

https://pierreaubert.github.io/spinorama/Genelec 8010A/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Genelec G2/Misc/misc/CEA2034.html
https://pierreaubert.github.io/spinorama/Genelec 8030C/ASR/asr/CEA2034.html

And a few of the KEFs...

https://pierreaubert.github.io/spinorama/KEF Q350/ASR/asr-horizontal/CEA2034.html
https://pierreaubert.github.io/spinorama/KEF R3/ASR/asr/CEA2034.html

And some of the JBLs and Revels...

https://pierreaubert.github.io/spinorama/JBL 708P/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/JBL 705i/JBL/vendor/CEA2034.html
https://pierreaubert.github.io/spinorama/Revel F226be/Revel/vendor/CEA2034.html
https://pierreaubert.github.io/spinorama/JBL M2/JBL/vendor/CEA2034.html


And several others, including the Infinity IL10, ELAC DBR-62, and the Focal Aria 906, among others...

https://pierreaubert.github.io/spinorama/Infinity Interlude IL10/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Infinity Interlude IL10/Infinity/vendor/CEA2034.html
https://pierreaubert.github.io/spinorama/Elac Debut Reference DBR-62/ASR/asr/CEA2034.html
https://pierreaubert.github.io/spinorama/Focal Aria 906/ASR/asr/CEA2034.html

So this is also a very common type of response on many of the better loudspeakers. (Possibly even the most common.)

 

[ADUHF]

There are also speakers which fall sort of in-between the above 3 shapes, which will have more of what I would describe as a soft V-shape.

 

[bigshot]

Listen. Don't talk for a minute. I'm getting exhausted with this. I'll repeat myself one more time.

I did not say that it was not a discernible difference. I said it is perceptible but IT DOESN'T MATTER because when listening to music your ears would easily accommodate to it in a very short space of time. Ears can adjust to response deviations. Small imbalances, especially long gradual ones just don't matter. The differences between one headphone copy and another of the exact same make and model are probably as great or greater than this.

You're arguing some sort of absolutist straw man theory that I'm not claiming. You can go ahead and say a tiny molehill is a mountain to you. That is fine. I will understand your obsession and leave you alone. But at least understand what I am saying.

THERE IS NO PERFECT RESPONSE CURVE. THE HARMAN CURVE IS AVERAGED FROM A RANGE OF DIFFERENT PEOPLE'S PERFECT RESPONSE CURVES. THERE'S NO NEED TO DEMAND ACCURACY TO A FRACTION OF A PERCENT BECAUSE THE CURVE ITSELF ISN'T ABSOLUTE FOR EVERYONE'S EARS.