How to equalize your headphones: advanced tutorial (in progress)

[AstralStorm]

This is HRTF, it's personalized, the equalization is tuned to sound equally loud at all frequencies. There's nothing closer to linear to my ears - this might not hold true for average ears (DF is derived from those) or your ears.
 
I've tuned the eq since to be a bit more accurate to flat sounding in the highest end and painted it better, and it looks like:

 
Now you know why I found VSonic GR07 sibilant. (also, that has some extra resonances)

 

[xnor]

But equal loudness at all frequencies is not what calibrated speakers sound like. For example your pinna boosts frequencies around 3 kHz. You don't equalize speakers for that, instead they produce a flat FR and what arrives your ear is a peak at 3 kHz. Especially in-ears don't care about your pinna at all so you have to have that peak in the headphones. That's why we have FF/DF-equalization in the first place.
 
Just take a look at the ER-4S here (DF-equalized):

 
Without DF-equalization:

 

[AstralStorm]

High fidelity speaker response is nearly linear, as in equally loud at all frequencies as heard by myself (or anyone else really including measurement microphones) - that's the ideal often not possible to achieve.
Obviously, you need a very well matched/prepared room to come close to that as well as excellent speakers.
Live performance (or a very good recording of such) is the actual reference, not any brand of speakers.
 
Headphone equalization with sine tones is more or less equivalent to applying your own 0 degree HRTF measured in an anechoic chamber. (Independent-of-Direction or ID)
Note that I'm talking about headphones/IEMs equalized to sound linear (sound equally loud at all frequencies) to your own ears. Not be linear - not reproduce all frequencies equally loud or sound linear to a measurement head, which likely doesn't match yours in one way or another.
 
Adding a properly set up crossfeed on top of the above simulates the speaker-related inter-aural time difference, most relevant for low and mid frequencies.
You can also add any kind of stereo convolution reverb to simulate a room response in high frequencies.
 
--
This approach is better in my opinion than attempting to encode a room-related response convolved by average HRTF (the definition of diffuse field). When the headphone is DF-equalized, adding a convolution reverb will cause the response to be doubly equalized. (since the microphone records a diffuse field)
 
In simpler terms:
1) equalize the IEM/headphone for binaural playback (ID equalization) = equally loud sine tones
2) apply a convolution reverb to simulate spectral cues
3) apply crossfeed to simulate ITD cues
 
I'm also not sold on whether ID or DF is more accurate in case of IEMs. DF has been shown to be more accurate with headphones.
Furthermore, the differences are restricted to the 500 Hz - 2.5 kHz range - ID equalized headphone/IEM has bit less loudness in that range.
Source: http://www.head-acoustics.de/downloads/eng/application_notes/BinauralMeasurement_06_11e.pdf

 

[xnor]

Quote:

High fidelity speaker response is nearly linear, as in equally loud at all frequencies as heard by myself (or anyone else really including measurement microphones) - that's the ideal often not possible to achieve.

Measurement microphones measure the frequency response in dB SPL or dBr referenced to a certain SPL and not loudness. Equal loudness is what a listener perceives and is measured in phons. Think about it this way: no matter if you measure speakers at 60, 80 or 100 dB SPL the frequency response ideally stays exactly the same. Equal loudness has nothing to do with it.
 
 
Quote:

Headphone equalization with sine tones is more or less equivalent to applying your own 0 degree HRTF measured in an anechoic chamber. (Independent-of-Direction or ID)

A measurement with an azimuth of 0° (speaker in front of the mic/head) in an anechoic chamber is a free sound field (FF) measurement.
 
I don't see how this could be equivalent to headphone equalization. Speakers are usually set up in an equilateral triangle at +- 30°. Therefore you need a mix of FF+DF, which ID comes close I guess.
 
 
 
Quote:

Note that I'm talking about headphones/IEMs equalized to sound linear (sound equally loud at all frequencies) to your own ears. Not be linear - not reproduce all frequencies equally loud or sound linear to a measurement head, which likely doesn't match yours in one way or another.
[...]
In simpler terms:
1) equalize the IEM/headphone for binaural playback (ID equalization) = equally loud sine tones

Again, equal loudness has nothing to do with these equalizations, not even if you had your personal EQ curves. Phons just describe how you percieve the loudness of pure tones.

 

[Kaffeemann]

After installing VAC I don't get any sound at all.
 
I set the output to Line 1(Virtual Audio Cable) in SineGen and foobar but ist doesn't work.
 
Oh, one more question: is there a stand alone version of electri-q?
 

 

[AstralStorm]

Indeed, it attempts to provide equal loudness referenced in my case to loudness at 500 Hz (not 1 kHz as per definition of phon unit)
Suprisingly though, the equalization produces very similar ATH as to the one measured by my audiologist some time back, so it doesn't compensate for equal loudness curve. Any reason why? My tone generator does not adjust for it.
 
The equalization is made at a specific level (semi-calibrated using speech samples) and is usually reasonably stable with volume. The equalization software uses dBFS apparently which is directly related to dBV, but not necessarily to dB SPL. (especially in case of multi-driver designs)
 
The result matches my old reference speakers reasonably well with exception of frequency response edges. (IEMs provide better subbass handling, but are rolled off at the highest end.)
(Reference setup: JBL Control 5, +/- 3 dB 75 Hz - 20 kHz, in a relatively small reflection-dampened studio room, 2.5m distance, in sweet spot, 60 degree spread; room is 4 m x 3 m x 2.25 m)
 
Is it possible that in a closed ear, equal loudness curves are different? This warrants some further study and equipment I no longer have access to.
(specifically, calibrated in-ear microphones to check SPL)
 
The equalization definitely doesn't audibly match any "loudness" preset and as you can see above in the B2 diagram, the result is quite close to flat when DF-equalized. The difference is probably due to personal HRTF and/or different playback levels.
 
--
Thanks for catching my mistake, anechoic chamber is FF indeed. ID is the plugged ear response, so that's what you will get instead with this equalization method and IEMs. With headphones it's likely that the resulting equalization is a compromise between DF and ID.

 

[AstralStorm]

Quote:

After installing VAC I don't get any sound at all.
 
I set the output to Line 1(Virtual Audio Cable) in SineGen and foobar but ist doesn't work.
 
Oh, one more question: is there a stand alone version of electri-q?
 

You need to route audio between VAC and the chosen soundcard, either using some VST host which allows separate input and output (I'd like one), ASIO4ALL with more than one device or the included Audio Repeater.
 
Currently my setup uses 2 VAC cables set to the same clock, linked together with ASIO4ALL through LiveProfessor. The output from the second one is then routed to the card using the Audio Repeater.
 
I'd love some alternative VST host which can use WASAPI exclusive or Kernel Streaming instead to get rid of the relatively lousy ASIO4ALL, which is not immune at all to even minor amounts of clock drift and thus requires the additional cable.

 

[xnor]

Quote:

Indeed, it attempts to provide equal loudness referenced in my case to loudness at 500 Hz (not 1 kHz as per definition of phon unit)
Suprisingly though, the equalization produces very similar ATH as to the one measured by my audiologist some time back, so it doesn't compensate for equal loudness curve. Any reason why? My tone generator does not adjust for it.

I have no idea what you're saying but if you're talking about the threshold of hearing - that's still psychoacoustics and has nothing to do with objective SPL measurements.
 
 
Quote:

The equalization definitely doesn't audibly match any "loudness" preset and as you can see above in the B2 diagram, the result is quite close to flat when DF-equalized. The difference is probably due to personal HRTF and/or different playback levels.

Ahm between 3 and 4 kHz you have a more than 15 dB dip, that's not flat.

 

[Kaffeemann]

@AstralStorm
 
Thank you I think I got it right now. I am using Audio Repeater. Wave in is VAC and Wave out is my audio card.
 
But here's the next problem: Although I integrated electri-q in VSThost it doesn't affect SineGen at all. It works perfectly for foobar though.

 

[AstralStorm]

Quote:

I have no idea what you're saying but if you're talking about the threshold of hearing - that's still psychoacoustics and has nothing to do with objective SPL measurements.
 
Ahm between 3 and 4 kHz you have a more than 15 dB dip, that's not flat.

 
While 3-4k -15 dB is not flat, the resulting curve shape doesn't approximate an equal loudness contour.
Note that I've cheated when drawing the graph and assumed linear dBFS <-> dB SPL relation. This should hold true for a wide SPL range.
(Otherwise large amounts of harmonic distortion would be noted.)
 
About the ATH, I meant the quick check using an ATH/hearing test tool mentioned somewhere earlier in the thread. This test for ATH resulted in a similar shape to ISO ATH curve.
 
I'd require a calibrated in-ear probe microphone to attempt a direct mapping... unfortunately, I don't have access to this kind of hardware.
 
Quote:

@AstralStorm
 
Thank you I think I got it right now. I am using Audio Repeater. Wave in is VAC and Wave out is my audio card.
 
But here's the next problem: Although I integrated electri-q in VSThost it doesn't affect SineGen at all. It works perfectly for foobar though.

 
In VSTHost, the DirectSound and MME options work pretty badly if at all, but if you want to try them out, you should pick VAC as input and soundcard as output device. Make sure to select a large enough buffer size - here, 768 samples worked with those outputs.
 
Alternatively, install ASIO4ALL, select that as the IO device in VSTHost, then enable VAC In / Soundcard Out in its configuration.

 

[Joe Bloggs]

Quote:

Originally Posted by AstralStorm /img/forum/go_quote.gif
 
Quote:

@AstralStorm
 
Thank you I think I got it right now. I am using Audio Repeater. Wave in is VAC and Wave out is my audio card.
 
But here's the next problem: Although I integrated electri-q in VSThost it doesn't affect SineGen at all. It works perfectly for foobar though.

 

In VSTHost, the DirectSound and MME options work pretty badly if at all, but if you want to try them out, you should pick VAC as input and soundcard as output device. Make sure to select a large enough buffer size - here, 768 samples worked with those outputs.
 
Alternatively, install ASIO4ALL, select that as the IO device in VSTHost, then enable VAC In / Soundcard Out in its configuration.

 
If using VSTHost you just need one VAC cable.  Go to VSTHost->Devices->Wave... select MME: Line 1 (virtual audio cable) as input device and MME: loudspeaker (or something like that) as the output device.  Then put Electri-Q between the input and output in the VSTHost main window and link them up

 

[keyro]

hi guys very interesting thread, i have a pair  Beyerdynamic DT 770 Pro-80 but i dont quite understand all that has been explained in the guide (yup i´m a newbi) so i´d like to know if anyone has done this to this pair of headphones could provide me with the final EQ so i can test the sound of it since iam really interested in listening how it sounds with the final result, thanks in advance guys

 

[hollowkenshin]

Okay, so if none of these methods work, how would we equalize the right way?
 
Lets say we:
1. Purchase a flat response in-ear microphone (e.g. http://www.soundprofessionals.com/cgi-bin/gold/item/SP-TFB-2)
1. Insert them in your ear and then place your headphones over the top.
3. Run pink noise through the headphones.
4. Take the resulting eq-curve captured by the microphone and invert it.
5. If desired, one could get some stereo impulses, room simulators or a crossfeed like xnor's to externalise the sound.
(Now just if we could homebrew a headtracker...)
 
Would this equalisation method be accurate?

 

[xnor]

You cannot just invert the FR captured by the mic. The response of a flat speaker is not flat at the eardrum. There are resonances like about +15 dB around 3 kHz etc.. So first you need this/your personal target curve.
My crossfeed tries to keep the FR as flat as possible because most headphones are already DF-equalized, at least to some degree.

 

[k00zk0]

Just wanted to note that sinegen 2.5 has both of its mirrors 404'ing and the sinegen 2.1 from tucows has unavoidable malware attached. I have uploaded a legitimate sinegen 2.1 here:
 
https://docs.google.com/open?id=0B6IxqGzcxDWTeGhycEdFeEdhZzg
 
Unfortunately the first "how to equalize your headphones" thread still has not had the OP link changed from the tucows one. Shame on head-fi; sorry but I don't approve of you pointing thousands of people hitting that thread from Google to download malware. I noted it two weeks ago in that thread.