[charleski]

I'm experimenting with EQing my iems to meet approach different target curves (within a few dB) and wondered if there's any official data out there that defines them.

Right now I've been estimating the curves from looking at pictures of them: The Olive & Welti 2014 curve (for over-ear headphones) shown here and the 2017 revision of the Harman target for in-ear 'phones shown here. Both sources use 5dB divisions and my estimates are a bit rough. They're probably good enough (especially considering how primitive my measuring setup is and the fact that I'm not aiming to produce a perfect match), but it would be nice to have a precisely-defined target.

 

[bigshot]

Are you using published measurements of your headphones, or are you measuring them yourself? If you're using published measurements, your own copy may vary as much as 3 to 5 dB due to manufacturing tolerances, so ball parking it within a couple of dB of the Harman curve would be fine I'd say.

Also remember that the Harman Curve is based on people's preferences and there was a range of preferences that were averaged to reach it. Don't feel afraid to tweak it yourself to suit your own ears. Just do it in a systematic fashion so you don't make random changes.

 

[charleski]

Measuring them myself (though as I said, my equipment is very primitive). You definitely can't rely on published responses. While my setup isn't properly calibrated (because I'm not spending hundreds of pounds on it) it does produce repeatable results (i.e. it may not be accurate, but it is precise) and this allows me to verify the effect of changes in the EQ curve.

You're absolutely right in saying these curves are just guidelines based on studies of overall preference, but it would just be nice to have a proper definition of them, if only for the sake of consistency.

 

[castleofargh]

strictly speaking, something like diffuse field or free field, doesn't define a compensation but a measurement method. so when you see compensations offered, they usually involve a very specific dummy head where the different measurements were done, so they have an effective relation they can apply to predict what the other measurement would have shown on that rig. so in a sense, the dummy head is the reference and the various compensations are made for it specifically. you can always take the graphs from people who own such systems and apply those compensations to your rig to "look right", which is what most people end up doing, but I would argue the merits of doing that at all, because it will give a false sense of objectivity. your coupler or tube may not have the same resonances as a professional setup, meaning that compensating for a given IEM might then let you get reasonably close measurements for a few IEMs, and also some pretty noticeable differences for other IEMs just because they are vented or have a different insertion depth, etc.
the other obvious issue with our cheapo setups is that our raw measurement often don't correspond with any sort of standard(although some fake Chinese couplers do come pretty close). so we tend to lack even the initial reference(let's call it calibration or an arbitrary reference of neutral). how could we possibly know how much compensation to apply for the various target responses if we are unsure as to what we're initially measuring?

now that I've talked like a guy who cares about proper stuff done properly, you might be interested in jumping on the little group experiment we have going on there https://www.head-fi.org/threads/ety...phone-for-your-ears-and-your-couplers.908512/ (fast if you're in the US)
it probably won't give you exactly what you were initially looking for, but it will give you some hand on references for what your initial RAW measurement is showing compared to more or less fancy measurement rigs. and then based on that you might decide to apply some fake DF or Harman compensation like many people do.

 

[charleski]

I suppose there might be something buried behind the AES pay wall (it would sure be nice if audio engineers followed the lead of others and dumped their papers on arXiv or engrXiv). But it just surprised me that people spend so much time talking about these targets without really having a solid definition of what they are.

Obviously the target curves shouldn't be regarded as a gold standard, even for those with an accurate setup. And even if the measurements are precise there are other confounding factors when comparing different 'phones, as you may end up using different insertion depths for purely mechanical reasons. But how much variability is acceptable? (And what does the term 'acceptable' mean here anyway?) There are lots of other questions concerning different types of variability, joint variability across the spectrum, etc.. But it's hard to consider these matters when you're not really sure what the underlying basis is.

I suppose it's tempting to fall back on the 'if it sounds good then it is good' adage. But my ears are poor discriminators, and I've found earphones with which I was quite happy could be transformed into something much better with the right EQ as long as I measured things along the way. (EQing blind just tended to end up with a mess.)

 

[bigshot]

I find that targets are a really good place to start. They get you 90% of the way there. But you often need to adjust a hair one way or the other to get it just the way you want it. Cans differ, ears differ, tastes differ.

I'm sure Olive's original paper at the AES has the exact measurements. Be aware that they revised the curve later. You'll have to choose which one you like better.

 

[EZAndante]

Precise values for the various permutations of the Harman target can be found here:

https://github.com/jaakkopasanen/AutoEq/tree/master/compensation

These data are part of Jaakko Pasanen's AutoEQ project. The were made by using some software tool to "parse" available graphs. (Read the README.md text file on that site for more information. The concept of the site is to combine the high quality measurements for various headphones available on-line on sites like InnerFidelity, Rtings.com and Oratory1990's Reddit page with a target curve then make the results available for use in various EQ tools like EqualizerAPO.)

You can read the data on-line by clicking on any .csv file, or you can import them into a spreadsheet.

Since I don't have a measurement rig, I work with the available measurements. But I combine a headphone measurement set with the Harman 2013 measurements in a spreadsheet to generate a compensation curve, then I manually transfer that into the 31-band interface of EQMac2 on my MacBook. Then I adjust by ear.

Not sure how you would work with these data in your situation, but I'd be more than happy to explain what I do on the spreadsheet side if that would help.

 

[SilverEars]

Over years of looking at measurements and hearing iems, I'm getting a sense of my preference target, but for headphones, I haven't seen ones that is really tuned to hit a target or looks that close.

Seems like iems have a better standard in rig consistancy and accuracy compared to full-sized.

I'm uneasy with the accuracy of full-sized measurements and therefore hard to know what rig really is most precise. Rtings seems to put out the most consistant and looks most accurate, but at the same time it seems quite a feat to tune a full-sized to hit a target curve. Planars seems to be more difficult if one has seen enough measurements, and dynamic drivers have most potential in hitting the target in the high frequencies.

I would like to hear something that used a known precise measurement rig to tune and hits a target curve to make an assessment. In order to assess, I would need variations of responses.

Iems are easier to assess if Harman is near my target since I have more confidence in the graphs.

 

[dazzerfong]

Over years of looking at measurements and hearing iems, I'm getting a sense of my preference target, but for headphones, I haven't seen ones that is really tuned to hit a target or looks that close.

Seems like iems have a better standard in rig consistancy and accuracy compared to full-sized.

I'm uneasy with the accuracy of full-sized measurements and therefore hard to know what rig really is most precise. Rtings seems to put out the most consistant and looks most accurate, but at the same time it seems quite a feat to tune a full-sized to hit a target curve. Planars seems to be more difficult if one has seen enough measurements, and dynamic drivers have most potential in hitting the target in the high frequencies.

I would like to hear something that used a known precise measurement rig to tune and hits a target curve to make an assessment. In order to assess, I would need variations of responses.

Iems are easier to assess if Harman is near my target since I have more confidence in the graphs.

Isn't it the other way around? IEMs have different insertion depths, and that plays a huge part on whether you'd like the treble or not, and how it measures too.

On the other hand, headphone measurements, especially closed, are seen as pretty much universal.

 

[charleski]

Precise values for the various permutations of the Harman target can be found here:

https://github.com/jaakkopasanen/AutoEq/tree/master/compensation

These data are part of Jaakko Pasanen's AutoEQ project. The were made by using some software tool to "parse" available graphs. (Read the README.md text file on that site for more information. The concept of the site is to combine the high quality measurements for various headphones available on-line on sites like InnerFidelity, Rtings.com and Oratory1990's Reddit page with a target curve then make the results available for use in various EQ tools like EqualizerAPO.)

You can read the data on-line by clicking on any .csv file, or you can import them into a spreadsheet.

Since I don't have a measurement rig, I work with the available measurements. But I combine a headphone measurement set with the Harman 2013 measurements in a spreadsheet to generate a compensation curve, then I manually transfer that into the 31-band interface of EQMac2 on my MacBook. Then I adjust by ear.

Not sure how you would work with these data in your situation, but I'd be more than happy to explain what I do on the spreadsheet side if that would help.

Sorry for not replying earlier. Thanks for these, they're a bit more precise than my own efforts to trace the various graphs. Probably the best we're going to get. I just import this as a 'house curve' in REW and then use the built-in EQ modeler to work out where to start.

The principal departure between the over-ear and in-ear curves is clearly the extended shoulder from 4-8kHz, which suggests an exaggeration of the main difference between free-field and diffuse-field measurements. I've spent the past couple of weeks playing around with EQing a set of MH755s to match the 2017 IEM curve and find it far too bright and shouty. After repeatedly refining my EQ settings I'm finding I prefer something which looks a lot more like the free-field response, at least as far as the high frequencies are concerned, with a deep cut from 6-13kHz and a notch to take out the ~8kHz resonance (and foam tips to dissipate it further). Since the target curves are based on aggregated preferences, I'm starting to wonder if my ears are just a departure from the norm or if any others have found something similar. Going by the explanation in innerfidelity, this suggests my conchae are a bit ... different .

Once I've got the high frequencies sorted out I'll play around with the mid-range. It looks like they just applied a low-shelf at around 150Hz (and have been upping the gain on that in successive years). It'll be interesting to see if varying that frequency has an impact, and how far I have to go before I notice the difference.