How to automatically EQ your headphones to online measurements: a step by step guide
Jan 8, 2019 at 4:33 PM
Alright I definitely understand what you meant now at least, even if not why haha. Thanks

When you say it decreases in volume smoothly, I assume you mean that, a linear line on the graph corresponds to a linear decrease in volume?

What frequency would you say in the treble is equal to half the loudness of 3kHz? Looking at a Harman curve just now, I would say 9-10kHz. Is that right?

Last edited:
Jan 8, 2019 at 8:11 PM
What frequency would you say in the treble is equal to half the loudness of 3kHz? Looking at a Harman curve just now, I would say 9-10kHz. Is that right?

You don't need to worry about this if you're following the guide. Just make sure that it smoothly decreases in volume without any obvious peaks (except the three quarter wavelength resonance peak) or dips.

Jan 8, 2019 at 8:19 PM
Is there no relation between the perceived loudness of a neutral headphone (i.e. if it is calibrated to the diffuse or harman target) and the equal loudness contour then? In other words, if you get a 'neutral' headphone and do a frequency sweep over it, you're saying it won't correspond in loudness with an equal loudness contour? Surely, just like the equal loudness contour, at least the sub bass and high treble loudness would be attenuated, and you mentioned that 3kHz is the loudest frequency as well.
the issue here is that the exact equal loudness contour is specific to the listener. the curve we see everywhere is an average from many curves done by many people.
then the notion of flat for a headphone is also listener specific(because of the missing HRTF cues). so depending on how much accuracy you're looking for and which references you're starting with, you will be right or completely wrong ^_^. but in general, you wouldn't even be able to find a neutral headphone, and if you somehow did you would create your own ELC from it, that may or may not align exactly with the latest version published. only the general idea will hold hold, for example you will most certainly find 2khz louder than 8khz. if you didn't, then for sure your headphone wasn't remotely flat.
so I'd answer you with a firm yes and no.

Jan 8, 2019 at 8:54 PM
Castleofargh, Griesinger published a file named Binaural_hearing_and_headphones.ppt (you should be able to find with Google and/or Internet Archive) which contains instructions for building affordable in-ear microphones that rest at the eardrum. These would be absolutely ideal for EQing headphones. The only problem is they use a Radio Shack part which is obviously now unavailable. If anyone can suggest a similar affordable mic capsule for that purpose, please do so. Also, for those with 4 digits to drop comfortably, Etymotic makes an inner ear probe mic system.
sorry I can't help, I imagine that Taobao would be the place to look for stuff like that, but I go mad trying to navigate Chinese websites on my own.

Jan 27, 2021 at 10:41 PM
Jan 30, 2021 at 7:00 AM
How does one find the actual curves - correctional, etc, on that page?
At the top you can see various folders and files, if for example you click on "compensation" you get into that folder and see a bunch of them. The .png are pictures, so beside giving you a look at the shape of the curve, you won't care for that. And the .csv have the curves as so many pairs of frequency+gain(if there is more, they might not work in REW depending on what the extra stuff is). If you click on one like say "diffuse_field.csv" you get to see those values and can confirm that we have indeed just one freq and one gain value per line(could find phase too if they came from actual measurements). To try it the other day before posting, I then clicked on "raw"(on the right) so the page only had the FR and gain data, and did ctrl+s(or right click "save page as") to save the .csv. In this example, that would be "diffuse_field.csv" that I would save on my computer.

I tried 2 randomly and I could use them as calibration curve in REW without problem. I can't say if those are relevant/accurate for a given series of measurements(because different measurement rigs would need their own curves), but they seem usable in REW.

Jan 30, 2021 at 4:32 PM
At the top you can see various folders and files, if for example you click
on "compensation" you get into that folder and see a bunch of them.
The .png are pictures, so beside giving you a look at the shape of the
curve, you won't care for that. And the .csv have the curves as so many
pairs of frequency+gain(if there is more, they might not work in REW
depending on what the extra stuff is). If you click on one like say
"diffuse_field.csv" you get to see those values and can confirm that we
have indeed just one freq and one gain value per line(could find phase
too if they came from actual measurements). To try it the other day before
posting, I then clicked on "raw"(on the right) so the page only had the
FR and gain data, and did ctrl+s(or right click "save page as") to save
the .csv. In this example, that would be "diffuse_field.csv" that I would
save on my computer.

I tried 2 randomly and I could use them as calibration curve in REW
without problem. I can't say if those are relevant/accurate for a given
series of measurements(because different measurement rigs
would need their own curves), but they seem usable in REW.

All I see are Harman target curves and coordinates. No actual headphone compensations.

Jan 30, 2021 at 6:19 PM
I gave an example for a compensation(free field, diffuse field, harman compensated to alter how your measurements will look). If you seek some measurements of a given IEM or headphone, maybe try the folder called "measurements"?
And if you want to see the EQ to apply to get something alledgedly close to Harman target(different measurement sources, different rigs, so the accuracy is what it is), or see what the result looks like in a graph, browse "results".

Jan 31, 2021 at 7:17 AM
I gave an example for a compensation(free field, diffuse field, harman compensated to alter how
your measurements will look). If you seek some measurements of a given IEM or headphone, maybe
try the folder called "measurements"?
And if you want to see the EQ to apply to get something alledgedly close to Harman target(different
measurement sources, different rigs, so the accuracy is what it is), or see what the result looks like
in a graph, browse "results".

EUREKA!!

I found curves plus plots for two of my sets(img 0162 and 0163 are for my Sony). It just took a bit of poking around in those folders.

My question now is, do these curves aim for Harman or for flat? My goal is flat - as per the attached Sonarworks.

#### Attachments

• IMG_0162.PNG
207 KB · Views: 0
• IMG_0163.PNG
42.5 KB · Views: 0
• IMG_0223.JPG
385.5 KB · Views: 0
Last edited:
Jan 31, 2021 at 2:21 PM
Harman is an attempt to achieve audibly flat, as close as possible to calibrated speakers in a room. True flat isn't audibly flat because headphones don't reproduce the same way as speakers.

Jan 31, 2021 at 3:11 PM
True flat isn't audibly flat because headphones don't reproduce the same way as speakers.

I'm sure that has a lot to do with HRTF(head-related transfer function).

So, doe that mean the curves in the link provided by Castle are aiming for the Harman curve?

Jan 31, 2021 at 4:26 PM
I believe so. He says “Harman compensated”.

Jan 31, 2021 at 4:32 PM
@jaakkopasanen mentions flat sometimes but the EQs offered point toward Harman curves for reference(or options to match one headphone with another one). So, it would probably be more accurate to say that the aim is to approach the generally preferred response(which was the aim of Harman's research). How similar that is to the idea of flat is a long conversation about statistics, subjectivity, and choosing a reference that probably doesn't apply to typical music listening anyway. I'm of the not so unique opinion that there is no universal flat for headphone and IEMs. Because everybody's HRTF is different in some ways, and the headphone/IEM will alter or bypass some of the unique filters giving what we call normal hearing(normal hearing being the assumed flat reference for a given listener). Meaning the correction may need to be custom made for most people to really achieve something they perceive as flat. but there can be such a thing as close enough, and between the huge variations of FR we find in headphones and IEMs, and the Harman targets, chances are that those targets get most people using most headphones, closer to their perceived flat, or at least something many listeners would prefer. So I think that approach makes a lot of sense overall(until the audio market moves on to a more HRTF compliant ecosystem).
Even for me, who would want to custom EQ anything and everything, starting from a response close to Harman's target can save me some time.