Beta Test: Earful -- a hearing test App

[pkane]

So you didn't mean to use the RAW frequency responses of the database? Or yes and it, like you said, flatten the response?
Because if so it is kinda different than use a compensation curve based on the Harman target ones and indeed the results; from your flattening method to using a impulse filter (targeted to have a Harman curve), will be different for obvious reasons.

I was aware of that I just wanted to try the new function

The raw measurement of the headphone frequency response (measured with no simulated ear, pinna, HRTF, and no Harman curves, etc.) will create a flat frequency response, compensating for any frequency errors of the headphones. That's what you would want to use when measuring threshold of hearing: flatten out the headphone frequency response.

 

[zermak]

The raw measurement of the headphone frequency response (measured with no simulated ear, pinna, HRTF, and no Harman curves, etc.) will create a flat frequency response, compensating for any frequency errors of the headphones. That's what you would want to use when measuring threshold of hearing: flatten out the headphone frequency response.

I see but the RAW measurements available on the database are mostly (at least the ones from oratory1990 and the Reference Audio Analizer) with the dummy head which simulates ear/pinna also so they are not the ones you are looking for.
For example my hand/home made masurements on the KZ10 Pro I showed here in my old post (UMINK-1 straing on the memory foam tips) is the one you want I guess.

 

[pkane]

I see but the RAW measurements available on the database are mostly (at least the ones from oratory1990 and the Reference Audio Analizer) with the dummy head which simulates ear/pinna also so they are not the ones you are looking for.
For example my hand/home made masurements on the KZ10 Pro I showed here in my old post (UMINK-1 straing on the memory foam tips) is the one you want I guess.

I could be wrong, but at least the InnerFidelity headphone measurements show the raw frequency response, uncorrected for HRTF or simulated pinna. Here's a description: https://www.innerfidelity.com/content/headphone-measurement-proceedures-frequency-response (the curve at the bottom of the chart is what we're interested in):

I was recently using Beats Solo 3 wireless headphones for testing. Here's the curve from AutoEQ website:

 

[zermak]

Oratory's RAWs are different then:

You can see the gears he uses to get the RAW measurements.

 

[pkane]

Oratory's RAWs are different then:

You can see the gears he uses to get the RAW measurements.

Sorry, but where does he show how the raw measurements were acquired? They look similar to InnerFidelity, but I'm not really familiar with his process.

 

[zermak]

It is written on the bottom of the image. That particular test was made on a GRAS 43AC/43AG coupler with KB5000/5001 anthropometric pinna. It meas the RAW data is with the simulated ear and so on (if you search for the parts numbers you will see what I am talking about).

 

[pkane]

It is written on the bottom of the image. That particular test was made on a GRAS 43AC/43AG coupler with KB5000/5001 anthropometric pinna. It meas the RAW data is with the simulated ear and so on (if you search for the parts numbers you will see what I am talking about).

I see. So yes, a direct microphone to an acoustic coupler to headphones measurement of frequency response is what's needed for Earful.

 

[zermak]

Ok, so what I thought since the begin and I suggest you to point it out then necasue many AutoEQ compensation curves are based on RAW measurements on simulated head/ears (even Inner Fidelity uses it indeed but maybe they share the RAW without the simulated head/ears; and well I have just seen the video in your link and indeed he said so but it's the program that compute the flat response curve from the RAW with the simulated head/ears).

 

[pkane]

Ok, so what I thought since the begin and I suggest you to point it out then necasue many AutoEQ compensation curves are based on RAW measurements on simulated head/ears (even Inner Fidelity uses it indeed but maybe they share the RAW without the simulated head/ears; and well I have just seen the video in your link and indeed he said so but it's the program that compute the flat response curve from the RAW with the simulated head/ears).

By the way, here's a comparison I did a few years ago of HD650 headphones measured on my head with in-ear microphones, and with a simple acoustic coupler. Green is the coupler, red is my ear response to the same headphones:

 

[zermak]

Nice and indeed it kinda reflects the measurements and what we know. The lower frequencies differences are probably due to the different sealing (headphones properly in your head vs coupler on some disc with hole I suppose) and the peaks at 2kHz it's becasue of the ear shape while the coupler and what you used misses that kind of shape.

 

[jae]

Are you able to add an option for individual curves for each ear? Or at least an option to mute left or right channel completely. Would be a godsend

 

[pkane]

Are you able to add an option for individual curves for each ear? Or at least an option to mute left or right channel completely. Would be a godsend

Sure! Sounds easy enough, just need to find the time

 

[m_g_s_g]

Are you able to add an option for individual curves for each ear? Or at least an option to mute left or right channel completely. Would be a godsend

Under Windows 10, you can just use the balance and move the sliders for your left or right channel to test a specific ear. [Control Panel->Sound->Playback Tab ->Your playback headphones/speakers->Properties button->Balance]. Anyway, thanks for bumping this thread; I'd completely missed it!

@pkane: Thanks for this! One question: how exactly should I do to calibrate the SPL for my Drop HD6XX headphones? I've used the HD650 AutoEQ mic calibration file you linked in a previous post. I have an UMIK-1 available. Should I just "touch" the inner side of my headphones cups to the mic and just open REW's SPL meter to read the SPL level? dB (A/C/Z) (F/S)?

 

[Dreyfus]

Hi @pkane,

can you tell me how the noise signal is compiled? I noticed that it is different from standard broadband white noise.
Just to make sure everything is set up correctly on my side, here is a record of the generated signal (sine, warble and noise):
https://drive.google.com/file/d/1zWimNWkue2t76Kz0kK_2C47P74zLueEa/view?usp=sharing

I actually expected something like 1/3 octave noise pinned to the selected target frequency.
Any comments on this subject?

By the way, thanks for sharing this wonderful tool with the community!

Regards,
Dreyfus

 

[pkane]

@pkane: Thanks for this! One question: how exactly should I do to calibrate the SPL for my Drop HD6XX headphones? I've used the HD650 AutoEQ mic calibration file you linked in a previous post. I have an UMIK-1 available. Should I just "touch" the inner side of my headphones cups to the mic and just open REW's SPL meter to read the SPL level? dB (A/C/Z) (F/S)?

Not sure why I'm not getting notifications, sorry didn't see this question.

Yeah, you can use your UMIK-1 and REW, as long as these are calibrated by some other SPL meter. If you have an actual SPL meter, then you could just use that directly. Don't touch the mic to the cup, but rather position it where you'd imagine your eardrum to be. To do this reasonably well, you'll need to create some artificial enclosure around the cup and the microphone. I made this fixture that worked reasonably well for me out of a small plastic bottle:

https://www.audiosciencereview.com/...headphone-software-correction.3453/post-85252

The photo shows the mic pushed all the way into the "ear", but for this measurement, you'd want to pull it back a bit to create some distance from the headphone.

A or C weighting is fine, if you use Earful SPL calibration, it'll be a tone around 1kHz. The weighting doesn't matter much there. Use the S setting for longer averaging.

 

[pkane]

Hi @pkane,

can you tell me how the noise signal is compiled? I noticed that it is different from standard broadband white noise.
Just to make sure everything is set up correctly on my side, here is a record of the generated signal (sine, warble and noise):
https://drive.google.com/file/d/1zWimNWkue2t76Kz0kK_2C47P74zLueEa/view?usp=sharing

I actually expected something like 1/3 octave noise pinned to the selected target frequency.
Any comments on this subject?

By the way, thanks for sharing this wonderful tool with the community!

Regards,
Dreyfus

It's using generated wide-band white noise (up to Nyquist) filtered to 1/3 octave around the target frequency using a sharp biquad (notch) filter.

 

[Dreyfus]

Thanks for the info!

The signal doesn't sound right to me. Sounds like a single sine tone which is simply being modulated in its level. You can hear it in the file linked above.

I could confirm that with another soundcard and another Windows PC.

 

[pkane]

Thanks for the info!

The signal doesn't sound right to me. Sounds like a single sine tone which is simply being modulated in its level. You can hear it in the file linked above.

I could confirm that with another soundcard and another Windows PC.

That's just a question of how wide the band-pass filter really is. I can make it settable using a Q value. As you make it wider, it'll sound more like what you might expect white noise to sound, but it'll also include many more frequencies than the one you're trying to measure.

 

[pkane]

Updated version of Earful, v1.0.9 is now available.

Changes as discussed recently:

• Added Left, Right, or Both ears selector (@jae )
• Added Pink noise option (@Dreyfus )
• Added band-pass Q setting for pink and white noise (@Dreyfus )
• Added Clipping indicator

Please let me know of any issues or any additions/suggestions.

 

[Dreyfus]

Wonderful, thank you for the extremely quick update!
I will try the latest version and let you know what I think asap.

but it'll also include many more frequencies than the one you're trying to measure.

This is the actual purpose I strive for. Since I evaluate the whole spectrum with a limited number of bands, there has to be some kind of "smoothing" that addresses the neighbouring frequencies lying between the data points. A wider Q of say 1/3 the octave does allow me to reduce the errors induced by abrupt resonances in the treble region.