Calibration tool for Headphone + Hearing capability

[Dreyfus]

Thanks for mentioning your Earful tool, @pkane. That's a really neat piece of software!

I still do not fully agree that sine and noise signals are completely interchangeable without perceptional differences, though. I just did a quick run through Earful and ended up with the following (minimum hearing level) curves:

I have to redo and verify that with a better isolating headphone and less tiredness any time soon. But so far I can conclude that the white noise adds a distinct downward slope to the curve. That said, the overall trend looks very similar. Also, the slope could probably be countered by simply using pink instead of white noise.

At any rate, if you aim for the comparison to dynamic music, I would say that gated noise is still closer than narrow sine peaks due to its wider bandwidth and random character.

 

[Peter Verbeek]

Here's a thread dedicated to the Earful app itself, including comparisons to F-M curve and ISO 226-2003

Thanks for the link. Great to know we're looking for the same thing.

I'm guessing that phon 0 line is probably not the right place to correct for ear response

I'm just wondering. The shape of the middle section of the contour is almost the same at different Phon's. For equalizing this middle section perhaps any Phon contour can be chosen. But in general it seems that 0 Phon isn't good for getting a good bass and high treble equalization, probably due to the fact you also pointing at. At least that sounds logical to me. I wish I know the science behind it (if there is any).

 

[Peter Verbeek]

I still do not fully agree that sine and noise signals ....
... gated noise is still closer than narrow sine peaks due to its wider bandwidth and random character.

There is a difference for sure. I agree that noise maybe closer to music then a sine. On the other hand, if you look at your test profiles the 2 curves are very similar in the main frequency band of our hearing, 1 kHz to 5 kHz. This band is also (obviously) music is focused on. So making an equalization for this main band either curve can be chosen, although this could only be true in your recent test case.

Btw. thanks again for testing. I appreciate your effort

adds a distinct downward slope

Yes, that's interesting. When including the frequencies below 1 kHz in your case for the sine curve it would result into a "bassy" equalization compared to the noise curve. Fortunately it's only a slope related matter. When the resulting equalization has too much or little bass this is quite easy to correct by an opposite going bass (slope).

using pink instead of white noise.

I'm not sure if that will be the case. If the taken noise band is very narrow, does this make much difference? Anyway, it would be nice to test that too.

 

[pkane]

Thanks for the link. Great to know we're looking for the same thing.

I'm just wondering. The shape of the middle section of the contour is almost the same at different Phon's. For equalizing this middle section perhaps any Phon contour can be chosen. But in general it seems that 0 Phon isn't good for getting a good bass and high treble equalization, probably due to the fact you also pointing at. At least that sounds logical to me. I wish I know the science behind it (if there is any).

Yes, around 60-80dB is the normal listening range, so I'd pick an equal loudness curve around there somewhere to start with. I plan to add an equal loudness measurement to Earful, but it may take a while. Maybe something you could add to your calibration tool in the meantime?

 

[Dreyfus]

Just found a plausible explination for the differences between the two (noise vs sine) curves:

Although the A-weighting curve, in widespread use for noise measurement, is said to have been based on the 40-phon Fletcher-Munson curve, it should be noted that determinations of equal loudness made using pure tones are not directly relevant to our perception of noise. This is because the cochlea in our inner ear analyses sounds in terms of spectral content, each 'hair-cell' responding to a narrow band of frequencies. The high-frequency bands are wider in absolute terms than the low frequency bands, and therefore 'collect' proportionately more power from a noise source. Equal loudness curves derived using noise bands therefore show an upwards tilt above 1kHz and a downward tilt below 1kHz.

Source: Lindos Electronics - Equal-Loudness Contours

This can also explain why a broad noise signal is better at simulating dynamic music than a narrow sine signal.

This band is also (obviously) music is focused on. So making an equalization for this main band either curve can be chosen, although this could only be true in your recent test case

There is a lot of guesswork without having a complete baseline.

And I still don't know how stable a compensation based on the averaged F-M curves mixed together with individual uncertainty in a different test environment with variable ambient noise, headphone designs and ear resonances can be.
But I think we do all agree that we need more empirical data to prove and improve the procedure.

If the taken noise band is very narrow, does this make much difference?

Sure! It changes the curve that the reference signal follows. If the reference signal follows a downward tilt (as pink noise does), you would add less bass and more treble to compensate.

 

[Peter Verbeek]

add to your calibration tool in the meantime?

A 0 Phon and a 0 Phon (quite arbitrary) corrected curve are already available in the first version of the test tool. I'm developing a dropdown selection box for the Phon's from 0 to 100. By matching the closes one of these curves to one's test curve one can visually see where the "weaker" frequencies of one's hearing/headphones are located frequency-wise speaking. And Peace (the test tool is part of this app) can create a decent equalization which can be a starting point for one's own equalization.

I plan to add an equal loudness measurement to Earful

That would be nice.

 

[Peter Verbeek]

The explanation you've found is quite nice. It gives the scientific base (or part of) we're looking for.

But I think we do all agree that we need more empirical data to prove and improve the procedure.

Yep. Your empirical data helps a lot. But you're right. There are so many uncertainties (different test environments, improper test procedures, difference in headphones placement, using noise/sine, using background noise, fatique, using equal-loudness contour based on sine, etc.).

If the reference signal follows a downward tilt (as pink noise does), you would add less bass and more treble to compensate.

Doesn't the effect of this downward tilt depend on the width of the narrow band noise? I mean, when using a narrow band the tilt of pink noise might by too tiny for having an effect.

 

[g0ne]

@Peter Verbeek

Hi! I can also confirm I'm getting a way to bright sound or when inversed a complete can-boomy sound. If i'm understanding things correctly implementation is meant to balance everything out; you need higher dB to hear bass clear which should mean equalizer will add some bass as you're not hearing it as good at similar loudness for example?
This is a test I did, having noise on as high but still comfortable level as possible (Sennheiser HD6xx)

I previous test with left and then right ear with quite low noise loudness level overall got me

which seemed less reliable.

Is there something I could do to get a better result?

I like the idea of getting what would essentially be flat monitors in your brain

Also, are we supposed to run multiple tests at multiple noise levels and then average it out in relative %? or is it set up in such a way that we should have an absolute loudness value as in reference listening volume (as loud as possible yet comfortable)?

After some tweeking I figured out the mid section are quite spot on from this test. The low bass and high treble though needs to be done with a different Phon?

 

[Silvian]

@g0ne

As one of the participants in developing this tool I'll try to explain my point of view, from how we implemented it to how I use it. By the way, in case you didn't read my write up on how to best use it, here it is. The last point, 11, is the answer to your question, but I think it pays to read it all, I also talk about what I do with the very high frequencies, the ones I hardly hear anymore, if at all.

Basically, the focus of the tool should be the midrange area. And you agree that that part sounds "right". It is based on how we, as a species, are sensitive to sound. This is what our ears have been "trained" for during the milenia. The bass and treble, the extremes of our hearing abilities, are a lot less important for us and especially in recorded music they are very recent "inventions", made possible by the technological developments of the last 100 years. We do not have the same sensitivities to perceive them plus there isn't as much happening there in comparison with the midrange.
As you can see if you check the equal-loudness contour curves in Wikipedia, it really is only at the extremes of the audio spectrum where they vary. That's because we lose sensitivity as the volume goes down - hence the "Loudness" button which was ubiquitous on the amplifiers from a few years ago.

We were tempted to not include bass and treble at all in the tool. But we did and the way it worked was that it simply was our, Peter's and mine preference at the time. But they should be treated separately by each individual based on their own preferences, listening volume, time of the listening session, etc. These are the variables, for each individual, which make impossible to have just one level for bass and treble to satisfy everyone.
As you may have noticed the majority of the discussions in this thread revolve around these extremes - and that should not have been the focus.

So basically just adjust them as you like. Other than that, as far as I am concerned, there is nothing to worry about.
I am not familiar, on a practical level, with any of the other potential variables mentioned already such as the type of test signal. I am sure there will be ways to improve the tool, but, to me, it is quite apparent that it solves a very high percentage of the imperfections of one's hearing + the particular headphones used.

I am not sure if the drop-down list Peter mentioned in earlier entries in this forum is available in the latest official release - I have access to it. Here, one can choose from a number of equal-loudness contours curves. I tested it and it is another way to introduce varied levels of bass and treble. See if you prefer this approach.

 

[Dreyfus]

Doesn't the effect of this downward tilt depend on the width of the narrow band noise? I mean, when using a narrow band the tilt of pink noise might by too tiny for having an effect.

That might well be. The wider the bandwith the more energy in the specific band.

David Griesinger recommends 1/3 octave pink noise for loudness experiments. I think that is specified in ISO 532-2:2017.

 

[Peter Verbeek]

David Griesinger recommends 1/3 octave pink noise for loudness experiments. I think that is specified in ISO 532-2:2017.

Thanks, That's really helpful. I hope I can implement it in a future version.

 

[Peter Verbeek]

After some tweeking I figured out the mid section are quite spot on from this test. The low bass and high treble though needs to be done with a different Phon?

It's like Silvian explained. And you've noticed it yourself, primarily the test tool is for the region from, say, 1 kHz to 5 kHz. Treble and especially bass is usually a matter of preference. So this can be added/subtracted yourself. The default frequency region for creating an equalization based on the test result is set from 200 Hz to 12 kHz. I think this could be narrowed down.
The selection for matching the test result to a equal-loudness contour at a Phon, will be in the upcoming version which isn't released it yet. Basically by matching the bass (slope) of the test result with an equal-loudness contour from this selection list (dropdown box) will result into a flat bass equalization so it can be added to taste.

 

[Dreyfus]

Some supplementation regarding the type of the test signal:
@pkane added a variable Q to his noise function. I just repeated the procedure with my R70X, starting with sine, then white noise with a width of 1/3 octave. The result is very similar to the first round commented in post #41:

Then I switched to pink noise:

The plot shows exactly what I expected. The downward slope of the pink noise compensates for the increase in treble sensitivity when switching from the sine to the noise signal. Now the curves are very close to each other. So close, that I would even say that it probably doesn't matter whether you use the sine or pink noise generator. However, one thing I found is that after switching to pink noise I had less trouble to concentrate on the test tone. It is more pleasent to listen to in longer sessions (adds less fatigue). Another advantage is that it probably causes less errors with headphones that suffer from a lot of unpredictable resonances since the wider bandwith of the 1/3 octave noise does also excite the neighbouring frequencies. This should smoothen the response and make it less susceptible to abrupt peaks and dips with the limited number of data points.

I hope that is some useful information for further developments.

Regards,
Dreyfus

 

[Peter Verbeek]

I had less trouble to concentrate on the test tone. It is more pleasent to listen to in longer sessions (adds less fatigue). Another advantage is that it probably causes less errors with headphones that suffer from a lot of unpredictable resonances since the wider bandwith of the 1/3 octave noise does also excite the neighbouring frequencies.

Those are quite good advantages. Thanks for that

 

[g0ne]

@Silvian @Peter Verbeek If we know the entire frequency response curve of a pair of headphones we own and then have the loudness correction based on our created spectrum how do we go about deducting what would essentially be level?

Where would the 0 or middle from which we subtract or add dB be in the loudness graph we created when placed over the frequency response curve of our headphones; how does one arrive at which frequencies needs to be adjusted in what amount?

How would one compensate for how music generally is mastered; the relative loudness of different instruments and vocals? Can we assume that the source is "flat", that is.

If the bass and treble are where we get weird results is there a method theorized where we could deduce what actually is flat there or is it just pure taste between roll off/diffuse/harman bump?

What I'm trying to ask is if your test could be improved upon when taking into consideration the particular frequency response curve in relation to the user's headphone which already is a sort of rough general pinpoint. It doesn't make sense to me why this shouldn't matter at all so I'm putting it out there as measuring the start frequency response curve for your headphones seems to be very important in relation to how much in % you should add or reduce based upon the loudness test. We lose resolution/clarity if we apply a lot of EQ generally

 

[Silvian]

@g0ne

Let me start with a disclaimer: I quite often have problems understanding what it is being said on subjects which either do not have a well enough developed language (describing sound is a good example), or where people talking are not at the same level with that language. I am way behind some of the contributors to this thread. I promised myself that I'll try to bring myself up to speed during this holiday.
So I am not sure if I completely understand the question. But having read it a few times, I'll try to contribute.

The tool measures for both the headphones' frequency response curve AND your hearing curve. When a sine signal at a certain frequency is being played, the level you tell the system your hearing threshold is, is a result of both curves. If, for example, you have 2 pairs of headphones, one say mids focused and another with a V kind of response (scooped in the mids region), then you'd be hearing a signal in the mids region earlier from the headphone which is mids focused.
Once all the measurements are done, the resulting curve is a "composite" curve of both what that particular headphone was able to reproduce and what you were able to hear. Then Peter "amplifies" - I don't know how he does it, the various points on this curve to match the target curve, which is the 0 phone equal-loudness curve, compensated for bass and treble to match what Peter and I happened to like at the time.

That target curve is not "flat", it is what 100 years of experiments showed what most of us humans hear like. I proposed to Peter that perhaps he should also include a flat curve as a target so that people would understand, objectively, what the result of that would be like. I'm sure it is not something anyone would appreciate - but as of now no one can really test, but it would certainly be useful to demonstrate why aiming for "flat" (the same dB values throughout the spectrum) with headphones is not correct.

The other question might be what is it that the music makers aim for? But with this tool, it becomes irrelevant. As long as the listener's perception is corrected for both the headphones and his hearing capabilities, the correction being targeted to what we are "designed" to hear like, then it follows that we'll be hearing what the producer intended. Of course, the assumption is that they know what they are doing, and I read that they are trained to differentiate signals 0.5dBs appart. They also have their share of challenges: what they produce is supposed to sound good on a boom-box, in a car as well as on a hi-Fi system/headphones, with no EQ, because most people do not care for one.

As for bass and treble, I do not consider anything "weird", it is just a result of the many variables I mentioned before. Once Peter releases that drop-down list of equal-loudness curves at various phon, you'll see very easy how the bass and treble decrease or increase relative to the mids, depending on the assumed listening level. This is a very good summary of the terms and the theory https://www.physicsclassroom.com/getattachment/actprep/act9ag.pdf

Again, I am not sure if I answered to much of what you asked Peter will probably intervene at some point too.

 

[g0ne]

@g0ne

Let me start with a disclaimer: I quite often have problems understanding what it is being said on subjects which either do not have a well enough developed language (describing sound is a good example), or where people talking are not at the same level with that language. I am way behind some of the contributors to this thread. I promised myself that I'll try to bring myself up to speed during this holiday.
So I am not sure if I completely understand the question. But having read it a few times, I'll try to contribute.

The tool measures for both the headphones' frequency response curve AND your hearing curve. When a sine signal at a certain frequency is being played, the level you tell the system your hearing threshold is, is a result of both curves. If, for example, you have 2 pairs of headphones, one say mids focused and another with a V kind of response (scooped in the mids region), then you'd be hearing a signal in the mids region earlier from the headphone which is mids focused.
Once all the measurements are done, the resulting curve is a "composite" curve of both what that particular headphone was able to reproduce and what you were able to hear. Then Peter "amplifies" - I don't know how he does it, the various points on this curve to match the target curve, which is the 0 phone equal-loudness curve, compensated for bass and treble to match what Peter and I happened to like at the time.

That target curve is not "flat", it is what 100 years of experiments showed what most of us humans hear like. I proposed to Peter that perhaps he should also include a flat curve as a target so that people would understand, objectively, what the result of that would be like. I'm sure it is not something anyone would appreciate - but as of now no one can really test, but it would certainly be useful to demonstrate why aiming for "flat" (the same dB values throughout the spectrum) with headphones is not correct.

The other question might be what is it that the music makers aim for? But with this tool, it becomes irrelevant. As long as the listener's perception is corrected for both the headphones and his hearing capabilities, the correction being targeted to what we are "designed" to hear like, then it follows that we'll be hearing what the producer intended. Of course, the assumption is that they know what they are doing, and I read that they are trained to differentiate signals 0.5dBs appart. They also have their share of challenges: what they produce is supposed to sound good on a boom-box, in a car as well as on a hi-Fi system/headphones, with no EQ, because most people do not care for one.

As for bass and treble, I do not consider anything "weird", it is just a result of the many variables I mentioned before. Once Peter releases that drop-down list of equal-loudness curves at various phon, you'll see very easy how the bass and treble decrease or increase relative to the mids, depending on the assumed listening level. This is a very good summary of the terms and the theory https://www.physicsclassroom.com/getattachment/actprep/act9ag.pdf

Again, I am not sure if I answered to much of what you asked Peter will probably intervene at some point too.

Thanks you so much for taking the time to go through all of this!
I must say this feature is one that really excites me. I look forward to also see it improved in the future as well.

The biggest constraint I can see now is mainly like you explained how the bass and treble differs based on loudness and that one needs to both know about this and how much of an offset one wants to apply (normal, weak, weaker, weakest etc) in order to get the right results (right being what we mean when we say flat which is ofc not flat in the sense that it is for speakers). The solution to understanding how one would tinker with bass and treble you proposed with different Phon levels is interesting.
From strongest-weakest offset inputs when loading a test how do one know which is more correct? Is normal based on your own preference or what happens behind the curtains in this decision making last step so to say? Can you make an algorithm that makes an offset based on some standard?
How do I know that I'm listening to the right loudness level when testing (perhaps some relation to the standards set forth when measuring headphone frequency response curves)?
If the process is dynamic it make sense to me to use the volume corresponding with the loudest yet comfortable listening level, what's your take on this?
I see great potential in this and thanks for the wonderful feature!

 

[Peter Verbeek]

As Silvian explained when you do a test it's by using headphones which have a "built-in" curve (and your hearing which has a certain curve too). Each headphones manufacturer tries to build a pair of headphones according to some curve. By nature this is like the equal-loudness contour which isn't flat to begin with. A manufacturer decides which frequencies will be emphasized in the headphones. Of course the used materials (cheap or expensive) do "select" a curve as well. In other words, there isn't an ideal "flat" response of a pair of headphones (or speakers for that matter). There's only a rough "flat" response which works for most listeners. One could wonder why so many people are searching the internet for the frequency response of their headphones. Isn't an expensive pair of headphones not good enough on its own? Why are people not satisfied with their headphones? I once flatten my Beyerdynamic DT770 32 Ohm headphones using its measured frequency response (AutoEQ). It took all the juice out of my headphones. I wonder what to do next. Should I create my own equalization? And how to do that? Instead of doing so, I realized that using a measured frequency response got me nowhere. To begin with it doesn't account for my hearing, my audio taste and my music taste. Moreover, the flattened headphones took the manufacturers curve out of it and left me basically a set of very cheap unuseful headphones. That can't be the goal of finding the frequency response of one's headphones.

The biggest constraint I can see now is mainly like you explained how the bass and treble differs based on loudness and that one needs to both know about this and how much of an offset one wants to apply (normal, weak, weaker, weakest etc) in order to get the right results (right being what we mean when we say flat which is ofc not flat in the sense that it is for speakers). The solution to understanding how one would tinker with bass and treble you proposed with different Phon levels is interesting.

As described above one's test result should show one's weak and strong points of one's hearing and headphones. If there are some spikes upwards or downwards then these need to be balanced with strong equalization. But even this is a matter of taste therefore the normal, weak, weaker, weakest options. Remember that some one has these spikes for a long time so he/she have gotten used to it.

From strongest-weakest offset inputs when loading a test how do one know which is more correct? Is normal based on your own preference or what happens behind the curtains in this decision making last step so to say? Can you make an algorithm that makes an offset based on some standard?

I'm experimenting with a curve which basically shows the weak points or better said the needed equalization. This curve is the difference between the ideal and the tested curves. The ideal being the selected equal-loudness contour. One issues has arised: Adding another curve in the graph has made things look more complex. Anyway, there isn't an algorithm and/or theory I know of. There could be of course. To create an equalization the test tool calculates the difference curve which describes the weak and strong points. Then creates an equalization based on the settings of the user of the "Use test for equalization" window.

How do I know that I'm listening to the right loudness level when testing (perhaps some relation to the standards set forth when measuring headphone frequency response curves)?

Basically knowing the above you can't. The test is done with your headphones which have a predefined curve. So what is the right loudness level? To only test one's hearing a pair of ideal headphones should exist. But headphones always will have a curve. Besides the results of that test can't be used for equalization for all headphones. Headphones just differ too much for. This again leaves us to do the test with our target headphones. Of course one could test different pairs of headphones to match them to eachother by using resulting equalizations. I did this matching with my cheap Philips and Beyerdynamics pairs of headphones. Although this can be seen as flatten the curve it was just matching using my hearing ability. I did learn something: Somehow even cheap headphones can be largely improved through the test tool. In my case I did need to dial back the pre amp to compensate for the resulting clipping after equalization. But that's just a result of the "cheapness" of the Philips headphones not able to produce some frequencies well enough.

I'm not sure if the above helps answering your questions. Let me know what you think.

 

[g0ne]

As Silvian explained when you do a test it's by using headphones which have a "built-in" curve (and your hearing which has a certain curve too). Each headphones manufacturer tries to build a pair of headphones according to some curve. By nature this is like the equal-loudness contour which isn't flat to begin with. A manufacturer decides which frequencies will be emphasized in the headphones. Of course the used materials (cheap or expensive) do "select" a curve as well. In other words, there isn't an ideal "flat" response of a pair of headphones (or speakers for that matter). There's only a rough "flat" response which works for most listeners. One could wonder why so many people are searching the internet for the frequency response of their headphones. Isn't an expensive pair of headphones not good enough on its own? Why are people not satisfied with their headphones? I once flatten my Beyerdynamic DT770 32 Ohm headphones using its measured frequency response (AutoEQ). It took all the juice out of my headphones. I wonder what to do next. Should I create my own equalization? And how to do that? Instead of doing so, I realized that using a measured frequency response got me nowhere. To begin with it doesn't account for my hearing, my audio taste and my music taste. Moreover, the flattened headphones took the manufacturers curve out of it and left me basically a set of very cheap unuseful headphones. That can't be the goal of finding the frequency response of one's headphones.

As described above one's test result should show one's weak and strong points of one's hearing and headphones. If there are some spikes upwards or downwards then these need to be balanced with strong equalization. But even this is a matter of taste therefore the normal, weak, weaker, weakest options. Remember that some one has these spikes for a long time so he/she have gotten used to it.

I'm experimenting with a curve which basically shows the weak points or better said the needed equalization. This curve is the difference between the ideal and the tested curves. The ideal being the selected equal-loudness contour. One issues has arised: Adding another curve in the graph has made things look more complex. Anyway, there isn't an algorithm and/or theory I know of. There could be of course. To create an equalization the test tool calculates the difference curve which describes the weak and strong points. Then creates an equalization based on the settings of the user of the "Use test for equalization" window.

Basically knowing the above you can't. The test is done with your headphones which have a predefined curve. So what is the right loudness level? To only test one's hearing a pair of ideal headphones should exist. But headphones always will have a curve. Besides the results of that test can't be used for equalization for all headphones. Headphones just differ too much for. This again leaves us to do the test with our target headphones. Of course one could test different pairs of headphones to match them to eachother by using resulting equalizations. I did this matching with my cheap Philips and Beyerdynamics pairs of headphones. Although this can be seen as flatten the curve it was just matching using my hearing ability. I did learn something: Somehow even cheap headphones can be largely improved through the test tool. In my case I did need to dial back the pre amp to compensate for the resulting clipping after equalization. But that's just a result of the "cheapness" of the Philips headphones not able to produce some frequencies well enough.

I'm not sure if the above helps answering your questions. Let me know what you think.

Reading your own personal journey and experience with trying to tinker with headphones to sound, well "flat" was really interesting. The experienced curve with weakpoints and suggestions sounds like an awesome idea! Choosing between normal,weak,weaker etc requires some common sense as well but quantifying this process or making it tangible could lead to an huge improvement to the implementation. To make the headphones sound too little like how they should take account for ones hearing and freqcurve is not as bad as overdoing it but my gut is telling me there is more here to actually choosing the magnitude of the offsets, that must be in relation to something quantafiable as the hearing test in itself is done based of different variables that seems to be important as this loudness curve illustrates

One thing I really would like is even more sensitive test dials; more nuanced than 1dB steps.

I've gotten the best results when listening to the tones with noise matching my environment along with the volume nob of my amp at the place of loudness I usually listen to. Under 500hz and over 9000hz it's gets a bit too unreliable but I find the current setting I've got enjoyable that is based of a combination of preference at bass region and the region between 250hz-7000hz mainly, perhaps on par/better than an autoeq offset to the Harman curve. [EDIT: I'm starting to like the EQ derived from your test more than the AutoEQ offset to Harman]

I can't help but want to be able to dial in everything more precise in the future though ^^

The sound I get seems to be less dry and airy, instead it's more sweep-like and harmonized, definitely more balanced.

It's a big, great step in the right direction the implementation of yours!

Would it be possible to implement something in particular for setting the bass under 200hz and the treble over 5000hz that would produce reliable adjustments or do you see we'll always have to adjust these regions to our own tastes by hand compared to diffuse or harman for example?

Wouldn't we ideally want to match the right Phon measurement (with corresponding noise and signal measurements) to the exact dB listening level we listen to music? Like a dynamic situation where you would have let's say three different presets for different dB listening? After all we perceive sound different at different loudness?

 

[Peter Verbeek]

One thing I really would like is even more sensitive test dials; more nuanced than 1dB steps.

I'll try to make the step 0.5dB.

my gut is telling me there is more here to actually choosing the magnitude of the offsets

I'm not sure what you're talking about. Perhaps it's this. Looking at the contours there's an obvious progression from 0 to 100 Phon: less bass slope as we hear this much better when the sound becomes louder, lesser low point at 3 kHz and 16 kHz and the frequency high point shifts from 10.5 kHz to 9 kHz. This seems to be our hearing "algorithm". In that sense there's more to it. Is this your gut feeling?

preference at bass region and the region between 250hz-7000hz mainly

Yep, that's what our experience also telling us. This mid region is the main focus, or better said the region where the most can be gained by testing as your next sentence shows

It's a big, great step in the right direction the implementation of yours!

Yes, we're getting there

Would it be possible to implement something in particular for setting the bass under 200hz and the treble over 5000hz that would produce reliable adjustments or do you see we'll always have to adjust these regions to our own tastes by hand compared to diffuse or harman for example?

I'm playing with the thought of an automatic match based on the test on the bass section so below 1 kHz (at least). Of course, if implemented this has to be tested. By doing things automatic (so taking out the user interaction) a user might expect a perfect match. He may be disappointed with the resulting equalization by this automatic selection of an equal-loudness contour. But it could be very nice if it does work.

Wouldn't we ideally want to match the right Phon measurement (with corresponding noise and signal measurements) to the exact dB listening level we listen to music?

Very nice you're having this thought. Basically having a single one equalization for all kinds of music, songs and volumes is pretty ridiculous if I may say so. So of course you're right. Depending on the used volume the equalization needs to change according to some equal-loudness contour which is created for a specific genre (remember that the equal-loudness contour is "measured" with sines). In other words, an proper equalization has a static part (say sliders) and a dynamic part which continuously measures the played song and adjusts accordingly to genre, volume, equalization and set taste. Anyway, all the test tool can give is a selection dropdown box for creating the equalization. Equalizer APO (thus Peace) doesn't have any dynamic features for sound processing. It could have but it doesn't.