How to freq resp calibrate ones own hearing?

[John Dyson]

The primary goal is 'calibrating' ones hearing for freq response, I have an almost immediate need and it must be reliable.

This is bordering on 'emergency last gasp'

Also, this request is as much as about  'anyone have any other ideas' as 'does this idea seem okay'.

 

The final question below is the most important to me:  "ANY IDEAS FOR SUCCESS?"

 

Here is the problem:

I have a project where there are numerous A/B response balance comparisons needed.  A big part of the project is to reliably compare response balance of recordings.

This is not about 'tuning' freq response as there are 'steps' or 'building blocks' that need to be assembled, eventually in multiple dimensions (the goal reminds somewhat of Tetris.)

It is also 'reverse engineering' and not 'design'.   There is little freedom away from the ultimate goal, and the steps ALL need to be correct.

 

Unfortunately, my hearing changes by quite a bit even through the day.

Also, this is NOT just about 'HF response', but also includes my variable hearing sensitivity at both LF AND HF.

Does anyone have experience in trying to 'calibrate' their hearing, and how much success with this experience?

 

Preface this idea with the following:

 

I just tried a 'seat of the pants' 'first measurement', and the initial first data points seem interesting.

Has anyone  had sucess making a spot measurement of hearing with this method below?

 

note: I am sensitive to the smallest audible nuances, but because of unreliable&variable  frequency sensitivity, almost blind to frequency response balance.

 

Periodically during the day, run response curves by using a subjective 'normal loudness' level vs freq.

1)  At the beginning of each period, start with a 'nominal loudness' at perhaps 1kHz

2)  Match the subjective loudness at a group of standard frequencies... e.g. 50Hz, 100Hz, 200Hz, 500Hz, 1kHz, 2kHz, 3kHz, 6kHz, 9kHz, 12k .

 

When matching the loudness, log the needed gain to match levels, and remember the +-dB setting.

I am thinking that this log of loudness vs. freq can be converted into an approx hearing response curve.

 

HAS ANYONE TRIED TO CALIBRATE THEIR HEARING?

HAVE THEY USED THIS METHOD, or SOME OTHER METHOD?

HAS THIS PRODUCED USABLE RESULTS?

ANY IDEAS FOR SUCCESS?

 

Normally, I would feel comfortable  'inventing' the method all by myself, but this is a time constrained situation.

I am grasping at straws, and considering the nearly immediate need, might be the only way to proceed.

 

One might ask about the headphones used:  Beyerdynamic DT990 (a special ediiton).  The DT990 is more of a 'semi accurate measurement' type device instead of being the best for entertainment listening.  The DT990 seems to have good, extended response, and subjectively 'flat enough' for what I do.  (I have two  DT770s that are totally useless, yet I was misled by them for several years.)

 

FRUSTRATED...

 

 

[Jud]

2 hours ago, John Dyson said:

The primary goal is 'calibrating' ones hearing for freq response, I have an almost immediate need and it must be reliable.

This is bordering on 'emergency last gasp'

Also, this request is as much as about  'anyone have any other ideas' as 'does this idea seem okay'.

 

The final question below is the most important to me:  "ANY IDEAS FOR SUCCESS?"

 

Here is the problem:

I have a project where there are numerous A/B response balance comparisons needed.  A big part of the project is to reliably compare response balance of recordings.

This is not about 'tuning' freq response as there are 'steps' or 'building blocks' that need to be assembled, eventually in multiple dimensions (the goal reminds somewhat of Tetris.)

It is also 'reverse engineering' and not 'design'.   There is little freedom away from the ultimate goal, and the steps ALL need to be correct.

 

Unfortunately, my hearing changes by quite a bit even through the day.

Also, this is NOT just about 'HF response', but also includes my variable hearing sensitivity at both LF AND HF.

Does anyone have experience in trying to 'calibrate' their hearing, and how much success with this experience?

 

Preface this idea with the following:

 

I just tried a 'seat of the pants' 'first measurement', and the initial first data points seem interesting.

Has anyone  had sucess making a spot measurement of hearing with this method below?

 

note: I am sensitive to the smallest audible nuances, but because of unreliable&variable  frequency sensitivity, almost blind to frequency response balance.

 

Periodically during the day, run response curves by using a subjective 'normal loudness' level vs freq.

1)  At the beginning of each period, start with a 'nominal loudness' at perhaps 1kHz

2)  Match the subjective loudness at a group of standard frequencies... e.g. 50Hz, 100Hz, 200Hz, 500Hz, 1kHz, 2kHz, 3kHz, 6kHz, 9kHz, 12k .

 

When matching the loudness, log the needed gain to match levels, and remember the +-dB setting.

I am thinking that this log of loudness vs. freq can be converted into an approx hearing response curve.

 

HAS ANYONE TRIED TO CALIBRATE THEIR HEARING?

HAVE THEY USED THIS METHOD, or SOME OTHER METHOD?

HAS THIS PRODUCED USABLE RESULTS?

ANY IDEAS FOR SUCCESS?

 

Normally, I would feel comfortable  'inventing' the method all by myself, but this is a time constrained situation.

I am grasping at straws, and considering the nearly immediate need, might be the only way to proceed.

 

One might ask about the headphones used:  Beyerdynamic DT990 (a special ediiton).  The DT990 is more of a 'semi accurate measurement' type device instead of being the best for entertainment listening.  The DT990 seems to have good, extended response, and subjectively 'flat enough' for what I do.  (I have two  DT770s that are totally useless, yet I was misled by them for several years.)

 

FRUSTRATED...

 

 

 

John - 

 

Yes, no wonder you're frustrated.  A then B is a method that is bound to lead to frustration, as echoic memory lasts about 4 seconds, and after that you're working from a recollection of how a sense impression made you feel, not a memory of the sense impression itself.

 

Suggestion: There are two potential courses, but since one of them would take weeks of intensive listening, I think there is only one realistic choice, and that is to remove memory from the equation. However, this will only work if your hearing is close to equivalent in both ears.

 

The two items of software I'd suggest are @pkane2001's DeltaWave Null Comparator ( https://deltaw.org ) and Audacity ( https://www.audacityteam.org ). The procedure is to use Audacity to combine a left and right input (for example, 50Hz signal as left input, 100Hz signal as right input) into one stereo track.  Then use DeltaWave to compare left and right.  You have to "switch sides" with the stereo track (that is, 50 Hz left channel first time, right channel second time) to account for hearing differences between your left and right ears.  And all you have to do is decide which side (if either) sounds louder in simultaneous listening to two tones, no feats of audio memory involved.  You can increase the loudness of the frequency that sounds softer until you reach equality, and you'll have a fairly precise measurement of how much you've had to increase one frequency until both sound equally loud.  Proceed in this way using frequency pairs until you know the ratios among all of them.

 

If there's anything in this description that's confusing, please let me know and I'll try to be more clear.

 

[pkane2001]

1 hour ago, Jud said:

 

John - 

 

Yes, no wonder you're frustrated.  A then B is a method that is bound to lead to frustration, as echoic memory lasts about 4 seconds, and after that you're working from a recollection of how a sense impression made you feel, not a memory of the sense impression itself.

 

Suggestion: There are two potential courses, but since one of them would take weeks of intensive listening, I think there is only one realistic choice, and that is to remove memory from the equation. However, this will only work if your hearing is close to equivalent in both ears.

 

The two items of software I'd suggest are @pkane2001's DeltaWave Null Comparator ( https://deltaw.org ) and Audacity ( https://www.audacityteam.org ). The procedure is to use Audacity to combine a left and right input (for example, 50Hz signal as left input, 100Hz signal as right input) into one stereo track.  Then use DeltaWave to compare left and right.  You have to "switch sides" with the stereo track (that is, 50 Hz left channel first time, right channel second time) to account for hearing differences between your left and right ears.  And all you have to do is decide which side (if either) sounds louder in simultaneous listening to two tones, no feats of audio memory involved.  You can increase the loudness of the frequency that sounds softer until you reach equality, and you'll have a fairly precise measurement of how much you've had to increase one frequency until both sound equally loud.  Proceed in this way using frequency pairs until you know the ratios among all of them.

 

If there's anything in this description that's confusing, please let me know and I'll try to be more clear.

 

 

In addition to DeltaWave, I wrote a simplified "audiology" exam software that lets you measure the frequency response of the combined system (headphones + your hearing). It's called Earful, and you can try it here: https://distortaudio.org/earful.html

 

At the very least it can help you compare your frequency response changes between sessions, but it can also give you an idea of how close or far you are from some industry standard averages.

 

 

 

[John Dyson]

Both of you so far have given me a good pointer to something that might definitely help.

As some might know, I have been having real troubles with finishing my project, and only recently it is working as expected mostly by brute force.   The project itself is definitely working very well and does what it was intended to do.  The core part of the project is not the only goal.

 

There are still demos, some needing limited mastering.  I have been screwing up the demos that require the limited mastering.   The screwup invariably happens when a raw recording might come with a different EQ than the normally available commercial versions of a recording.   Trying to  match the sound of the new version to the old recording has been hit or miss.   Of course, the resulting new version has a possibility of being profoundly better to those with hifi audiophile hearing.   The fly in the ointment is that my botched EQ destroys the usefulness of the new recording to both audiophiles and casual listeners.

 

It would have been a good to objectively address the issue a few years ago.  Instead, over the last couple of years, I ended up beating my head against the wall, test people's patience and cause a large number to lose interest to the 'crackpot'.   Now, I forced the completion of the project by persistence instead of wisdom.

 

Thanks again for the info, and will look at 'earful'.  It seems to be close to what might be useful.   I know that I cannot ask for perfection in matching my environment, but in desperation I actually tempted to run applications software on WIndows 🙂.

 

[Jud]

34 minutes ago, John Dyson said:

Both of you so far have given me a good pointer to something that might definitely help.

As some might know, I have been having real troubles with finishing my project, and only recently it is working as expected mostly by brute force.   The project itself is definitely working very well and does what it was intended to do.  The core part of the project is not the only goal.

 

There are still demos, some needing limited mastering.  I have been screwing up the demos that require the limited mastering.   The screwup invariably happens when a raw recording might come with a different EQ than the normally available commercial versions of a recording.   Trying to  match the sound of the new version to the old recording has been hit or miss.   Of course, the resulting new version has a possibility of being profoundly better to those with hifi audiophile hearing.   The fly in the ointment is that my botched EQ destroys the usefulness of the new recording to both audiophiles and casual listeners.

 

It would have been a good to objectively address the issue a few years ago.  Instead, over the last couple of years, I ended up beating my head against the wall, test people's patience and cause a large number to lose interest to the 'crackpot'.   Now, I forced the completion of the project by persistence instead of wisdom.

 

Thanks again for the info, and will look at 'earful'.  It seems to be close to what might be useful.   I know that I cannot ask for perfection in matching my environment, but in desperation I actually tempted to run applications software on WIndows 🙂.

 

I guess there’s always WINE, I hear it’s better these days. 😉

[pkane2001]

8 hours ago, John Dyson said:

Thanks again for the info, and will look at 'earful'.  It seems to be close to what might be useful.   I know that I cannot ask for perfection in matching my environment, but in desperation I actually tempted to run applications software on WIndows 🙂.

 

I've been able to run my other software under WINE on Linux, and it works, like Jud suggests. I've had some problems with accessing audio devices for playback, though, but it could be just a problem with the VMWare driver implementation (running on a Mac) or maybe the WINE version or configuration -- I'm neither a Linux nor WINE expert. In any case, you could try WINE, but the surest way to get these working would be to use Windows. Audacity, I believe, runs on multiple platforms, so you don't need emulation to use that piece of software.

[John Dyson]

57 minutes ago, pkane2001 said:

 

I've been able to run my other software under WINE on Linux, and it works, like Jud suggests. I've had some problems with accessing audio devices for playback, though, but it could be just a problem with the VMWare driver implementation (running on a Mac) or maybe the WINE version or configuration -- I'm neither a Linux nor WINE expert. In any case, you could try WINE, but the surest way to get these working would be to use Windows. Audacity, I believe, runs on multiple platforms, so you don't need emulation to use that piece of software.

I do have Audacity, and use it from time to time.   Using it a few years ago, ran into a minor nit.  Maybe will also reinstall Audacity because it has/had a very strange bug.  That bug represented normal enough results that it misdirected me on the project for a few days.   I still avoid that section of Audacity.  Might as well address the issue now and just reinstall it also.

 

After I get rid of the strange, not often occuring HF matter in the V7.0B of the decoder,  I'll start the experiment to calibrate my hearing.   I already delayed the user by a day or so, need to make sure that the decoder bug/problem stays in front of the queue.

 

THANKS AGAIN!!!
 

 

 

 

[PeterG]

10 hours ago, John Dyson said:

Both of you so far have given me a good pointer to something that might definitely help.

This is surprising to me..  As someone with a small amount of hearing loss at the high end, I was certain you would note Jud's caveat on hearing in both ears needing to be equal.  Have you had your hearing measured by an audiologist?  They provide a detailed report of loss at each frequency in each ear.  You could imagine this report plus an equalizer (or two?) adjusting for loss.  This will not help with the intraday issues you report, but might at least provide a baseline

[Jud]

2 hours ago, PeterG said:

This is surprising to me..  As someone with a small amount of hearing loss at the high end, I was certain you would note Jud's caveat on hearing in both ears needing to be equal.  Have you had your hearing measured by an audiologist?  They provide a detailed report of loss at each frequency in each ear.  You could imagine this report plus an equalizer (or two?) adjusting for loss.  This will not help with the intraday issues you report, but might at least provide a baseline

 

He would soon get a good rough idea from the procedure if the results differ upon switching frequencies between ears.  Of course then the question arises which if either of the two is closer to being useful, and if the answer is neither, then measurements are the only thing.  Except for what John is doing, as I understand it, standard measurements won't work very well.

[John Dyson]

59 minutes ago, Jud said:

 

He would soon get a good rough idea from the procedure if the results differ upon switching frequencies between ears.  Of course then the question arises which if either of the two is closer to being useful, and if the answer is neither, then measurements are the only thing.  Except for what John is doing, as I understand it, standard measurements won't work very well.

About standard techniques not working.  One time, on anofher forum, I was given a tutorial on matching freq response, and out of frustration, got very angry about the presentation about the "obvious."  Thinking "already tried that a long time ago",  I was very desperate for actual help.    The project results in  something that does NOT have a fixed freq response, not a fixed gain, yet subjectively needs to appear to be flat.  It really doesn't seem that there is any easy or even reliable comparative measurment possible without a specification and a model to compare against. :-).  Unfortunately, on the project, there is neither a specificiation nor model.   Luckily for raw DolbyA, I did have a model, 2 different kinds of schematics, and measurements.   So, emulating a DolbyA was mostly a matter of determining what was important, what was not important and learning more about the area of nonlinear control systems where I had little practical experience.   Perhaps the trickiest thing was avoiding patent infringement, but that wasn't really all that difficult.   The biggest real problem for the project was an engineer with lots of past success, who felt that he knew more than he really did (me.)

 

The larger, more complex project, that uses DolbyA as a major component is a whole other story, including NOT knowing the technical goals, what it should sound like, desired behavior, at the project started.   DIdn't even know the technical goals until far into the project.   The variability of my hearing had further confused the situation.   Simply, if one looks at the whole picture, I chose the wrong project for a reasonable chance to succeed.

 

The hearing matter had become more important for sections/modules of the project don't have  100% internal connections, so the external connections (EQ) have another confused set of variables on the input and output nodes.    The fully internal nodes are a little easier to deal with because of things equivalent to 'impedance match' in the EE realm, but internally still more variables than just the one analogous  analytic variable called 'impedance.'   For the internal nodes there are more constraints, paradoxically making their interfaces easier to develop.    Thank goodnes for being able to hear 'tells' even though my hearing sucks in some very important ways.

 

Again, thank you all for the ideas/pointers.  Sometimes, I just cannot invent everything myself -- I think that I am in invention/design burnout.

Anytime that someone really intends to help, I appreciate and respect it so very much.