Griesinger's Headphone EQ for Frontal Localization

[pozz]

Has anyone tried this? http://www.dgsonicfocus.com/

It sort of worked for me on the AKG K371. With any headphone or IEM I have a persistent sense of elevation (vertical soundstage) as the frequency goes up. The best I could get with the DG method was a narrower range of elevation. That said, listening became fatiguing and the overall tone was ruined.

It's possible of course that I'm doing something wrong. I used a KH80 a mono speaker reference and listened from about a 40cm away. I didn't try loudness equalization below 500Hz to avoid room effects.

When I was doing it I thought the third octave bands were too wide (DG uses a Q of 5). I would set gains different depending if I was going in ascending or descending order (I'm pretty sure I'm locking onto different partials intraband). Instead of just using the 500Hz reference band to match, I also quickly flipped between adjacent bands until they were equal. Perhaps this was an error.

I also didn't try any binaural music (I have some on a hard drive somewhere) beyond what Griesinger embedded into a related presentation. From his papers and talks it doesn't seem like this method is exclusive to binaural recordings alone, and should work for stereo.

 

[David Griesinger]

To answer your questions:

Please read the "About" file in the app carefully, and watch the how-to video on YouTube. My website is also useful - but I update
it frequently. Chrome insists on using older versions unless you delete history. So do that periodically.

Your major error is not using alternating noise bands to find your equal loudness data. There is an extensive literature on equal
loudness tests which states unequivocally that to make an accurate map of the base sensitivity of human ears a test tone or noise
band must alternate with a reference tone or noise at about a one second rate. The reason is that human hearing continuously
adjusts the sensitivity of the basilar membrane to optimize the signal to noise ratio of the ear as a function of frequency. So if
there is a loud midrange signal the mid frequencies are turned down and if higher frequencies are soft, the higher frequency
sensitivity is turned up. The ear/brain system keeps track of the changes in gain, so spectrum you perceive is still useful for detecting
the identity of the source of sound and for determining its azimuth and elevation. Comparing the loudness of a test band to the
loudness of a reference band with a rapid alternation defeats the auto gain control, and an accurate value can be found.

So you MUST use the app as it is designed. You must adjust the loudness of each band under test to match the loudness of the
reference band as best you can. Doing this can seem difficult at first - but considerable experience with equal loudness tests finds that
almost everyone can do the test reliably. It does not matter if the answer you get is "correct". It only matters if you get the
same answer consistently. In the literature almost everyone is able after a few tries you get the same answer for each band
with a consistency of +-1dB. The app uses this consistency to find the difference between the equal loudness from the speaker
and the equal loudness loudness from the headphone. The proper eq for the headphone can then be found by subtracting
the speaker data in dB from the headphone data in dB. With practice you should get the same answer going up and going
down. I prefer to go up first, and then go back down making changes. When I go back up I seldom need to change anything.

I chose to use 1/3rd octave frequencies for each band because they are an international standard, and there are many 1/3 octave
equalizers on the market that have been used successfully professionally for many years. Once the data is found with our app you
can use it in many ways. For example there is a Windows app that allows you to set up a 1/3 octave equalizer of Q3 or Q5 for
everything you play through the headphone jack. I chose Q=5 for our app to get a little better accuracy. The professional
standard is Q=3 - but the Q=5 data works fine in a professional eq.

My website now includes a link to a folder with the older Windows apps. There is a binaural music file in that folder. Also in my powerpoints
on proximity there are several examples of binaural recordings from other halls. All these files are equalized to be frequency flat from
the front - which means a frontal source would be recorded with the same spectrum as a studio microphone. Some commercial dummy heads claim to have a "free field" setting which is supposed to make them frequency flat from the front - but in my measurements very few meet
this criterion. "Diffuse field" eq is definitely wrong. So it is useful to check the spectrum of a commercial binaural recording by comparing it to
an equivalent recording made for stereo reproduction. The spectra from a well recorded orchestra are pretty consistent.

David Griesinger

 

[JohnPM]

Chrome insists on using older versions unless you delete history. So do that periodically.

Putting this in the head section of your pages might help with that:

<meta HTTP-EQUIV="CACHE-CONTROL" CONTENT="NO-CACHE">

 

[Hugo9000]

@David Griesinger
Welcome, Dr. Griesinger! It's good to see you here!

 

[pozz]

To answer your questions:

Please read the "About" file in the app carefully, and watch the how-to video on YouTube. My website is also useful - but I update
it frequently. Chrome insists on using older versions unless you delete history. So do that periodically.

Your major error is not using alternating noise bands to find your equal loudness data. There is an extensive literature on equal
loudness tests which states unequivocally that to make an accurate map of the base sensitivity of human ears a test tone or noise
band must alternate with a reference tone or noise at about a one second rate. The reason is that human hearing continuously
adjusts the sensitivity of the basilar membrane to optimize the signal to noise ratio of the ear as a function of frequency. So if
there is a loud midrange signal the mid frequencies are turned down and if higher frequencies are soft, the higher frequency
sensitivity is turned up. The ear/brain system keeps track of the changes in gain, so spectrum you perceive is still useful for detecting
the identity of the source of sound and for determining its azimuth and elevation. Comparing the loudness of a test band to the
loudness of a reference band with a rapid alternation defeats the auto gain control, and an accurate value can be found.

So you MUST use the app as it is designed. You must adjust the loudness of each band under test to match the loudness of the
reference band as best you can. Doing this can seem difficult at first - but considerable experience with equal loudness tests finds that
almost everyone can do the test reliably. It does not matter if the answer you get is "correct". It only matters if you get the
same answer consistently. In the literature almost everyone is able after a few tries you get the same answer for each band
with a consistency of +-1dB. The app uses this consistency to find the difference between the equal loudness from the speaker
and the equal loudness loudness from the headphone. The proper eq for the headphone can then be found by subtracting
the speaker data in dB from the headphone data in dB. With practice you should get the same answer going up and going
down. I prefer to go up first, and then go back down making changes. When I go back up I seldom need to change anything.

I chose to use 1/3rd octave frequencies for each band because they are an international standard, and there are many 1/3 octave
equalizers on the market that have been used successfully professionally for many years. Once the data is found with our app you
can use it in many ways. For example there is a Windows app that allows you to set up a 1/3 octave equalizer of Q3 or Q5 for
everything you play through the headphone jack. I chose Q=5 for our app to get a little better accuracy. The professional
standard is Q=3 - but the Q=5 data works fine in a professional eq.

My website now includes a link to a folder with the older Windows apps. There is a binaural music file in that folder. Also in my powerpoints
on proximity there are several examples of binaural recordings from other halls. All these files are equalized to be frequency flat from
the front - which means a frontal source would be recorded with the same spectrum as a studio microphone. Some commercial dummy heads claim to have a "free field" setting which is supposed to make them frequency flat from the front - but in my measurements very few meet
this criterion. "Diffuse field" eq is definitely wrong. So it is useful to check the spectrum of a commercial binaural recording by comparing it to
an equivalent recording made for stereo reproduction. The spectra from a well recorded orchestra are pretty consistent.

David Griesinger

Thanks David. I'll try again at some point and report my impressions here. My trouble was that partway through I always noticed the difference in relative loudness between distant bands and tried to account for that.

Does the current Windows version include a player like the Android app? The one I downloaded did not and I imported the results into MathEQ for foobar using the generated text file. Also, the Android app hasn't been released yet, correct?

 

[amirm]

To answer your questions:

Warm welcome to the forum Dr. Griesinger! It is a privilege to have you here.

 

[amirm]

Putting this in the head section of your pages might help with that:

<meta HTTP-EQUIV="CACHE-CONTROL" CONTENT="NO-CACHE">

Does this still work? Chrome caching is very pesky and ignores many directives. When I used to manage our company website manually, I put in every trick in the book and it would still take it a while to fetch the new page.

 

[JohnPM]

Does this still work?

I think so. Could use "must-revalidate" rather than "no-cache" to tell Chrome to ask the server if there is a more recent version than the page it has cached.

 

[dasdoing]

To answer your questions:

Please read the "About" file in the app carefully, and watch the how-to video on YouTube. My website is also useful - but I update
it frequently. Chrome insists on using older versions unless you delete history. So do that periodically.

Your major error is not using alternating noise bands to find your equal loudness data. There is an extensive literature on equal
loudness tests which states unequivocally that to make an accurate map of the base sensitivity of human ears a test tone or noise
band must alternate with a reference tone or noise at about a one second rate. The reason is that human hearing continuously
adjusts the sensitivity of the basilar membrane to optimize the signal to noise ratio of the ear as a function of frequency. So if
there is a loud midrange signal the mid frequencies are turned down and if higher frequencies are soft, the higher frequency
sensitivity is turned up. The ear/brain system keeps track of the changes in gain, so spectrum you perceive is still useful for detecting
the identity of the source of sound and for determining its azimuth and elevation. Comparing the loudness of a test band to the
loudness of a reference band with a rapid alternation defeats the auto gain control, and an accurate value can be found.

So you MUST use the app as it is designed. You must adjust the loudness of each band under test to match the loudness of the
reference band as best you can. Doing this can seem difficult at first - but considerable experience with equal loudness tests finds that
almost everyone can do the test reliably. It does not matter if the answer you get is "correct". It only matters if you get the
same answer consistently. In the literature almost everyone is able after a few tries you get the same answer for each band
with a consistency of +-1dB. The app uses this consistency to find the difference between the equal loudness from the speaker
and the equal loudness loudness from the headphone. The proper eq for the headphone can then be found by subtracting
the speaker data in dB from the headphone data in dB. With practice you should get the same answer going up and going
down. I prefer to go up first, and then go back down making changes. When I go back up I seldom need to change anything.

I chose to use 1/3rd octave frequencies for each band because they are an international standard, and there are many 1/3 octave
equalizers on the market that have been used successfully professionally for many years. Once the data is found with our app you
can use it in many ways. For example there is a Windows app that allows you to set up a 1/3 octave equalizer of Q3 or Q5 for
everything you play through the headphone jack. I chose Q=5 for our app to get a little better accuracy. The professional
standard is Q=3 - but the Q=5 data works fine in a professional eq.

My website now includes a link to a folder with the older Windows apps. There is a binaural music file in that folder. Also in my powerpoints
on proximity there are several examples of binaural recordings from other halls. All these files are equalized to be frequency flat from
the front - which means a frontal source would be recorded with the same spectrum as a studio microphone. Some commercial dummy heads claim to have a "free field" setting which is supposed to make them frequency flat from the front - but in my measurements very few meet
this criterion. "Diffuse field" eq is definitely wrong. So it is useful to check the spectrum of a commercial binaural recording by comparing it to
an equivalent recording made for stereo reproduction. The spectra from a well recorded orchestra are pretty consistent.

David Griesinger

welcome.
could you please explain the concept of frontal localization?

 

[David Griesinger]

Putting this in the head section of your pages might help with that:

<meta HTTP-EQUIV="CACHE-CONTROL" CONTENT="NO-CACHE">

Thanks - I did a major re-write of the site and added this header. Hopefully it will work.
David

 

[Twelvetone]

Greetings to Dr. G.! Any notion of when the Mac OSX version of your SW might appear?

 

[Ron Texas]

@David Griesinger and @JohnPM thank you for dropping in.

 

[scott wurcer]

Hi Dave, good to see you here. Were we neighbors on Raymond St. in Cambridge, I remember someone trying to help with our dreadful acoustics in the planned community project down the street? If I'm mistaking you for someone else my apologies. IIRC Doug Preis also lived a few blocks away.

 

[David Griesinger]

Greetings to Dr. G.! Any notion of when the Mac OSX version of your SW might appear?

A Mac OSX version of the headphone app is now in the dropbox link below or the link in www.dgsonicfocus.com. There is also a VST3 version of our app with low latency. The VST3 works with modern devices, such as the Blue Cat Audio VST jukebox. It may not work with older sound programs.

Our new apps also include a low latency ASIO connection to an external (analog) interface.

The dropbox link is in dgsonicfocus.com, as well as here:
https://www.dropbox.com/sh/l7oxszh0dwvwjr6/AADsbLTomcogGzl6oALOcDOUa?dl=0

If the new VST3 does not work, try the older VST3 in the dropbox folder with the ASIO only apps. The apps use the older user interface, but work the same as the new ones. See www.dgsonicfocus.com or:
https://www.dropbox.com/sh/uvgel6xkbaq61k6/AAB1guqPFm_QGnsyU835t3iea?dl=0

You do not need a built-in player for these apps. There are Windows and Mac apps that provide a virtual cable that can connect any audio stream on your computer to our apps. Download 'Virtual Cable" for Windows or "Soundflower" for Mac. Set the audio output for your computer to "Virtual Cable" for Windows, or "Soundflower" for Mac. Then use the "Options" tab on our app to set the input to "Virtual Cable" or "Soundflower" and the output to your computer sound card.

 

[ObjectiveSubjectivist]

So as far as I know I cannot use laptop and an ordinary microphone?
The microphone has to be calibrated?