Personalized Equalization with in-ear mics

And this is where I couldn't disagree more. In my opinion the in-ear measurements are the only headphone measurements that make sense without compensation. Hands doesn't compensate at all for the fact that the mic is in his ear. Neither do I. Go look at our measurements. Do they look that off to you? To me they look more accurate to what I hear than the coupler measurements.
I know it seems a bit counter intuitive, but the ear-canal opening measurements look very different than the ear drum measurements Tyll does.

Measurement repeatability is indeed shit compared to a real coupler. With my coupler I can get measurements within 0.01db if I measure them within a few minutes and disassemble and reassemble my coupler. With the in-ear mic the measurement error in the mid-treble is closer to 0.5db and it's even worse in the upper treble. Resonances shift slightly from various tries. Tiny positioning differences with the mic and the headphone have a much bigger effect than with the coupler. Sometimes your measurements are off by a couple dbs for no apparent reason, before going back to normal the next day. Sometimes you even break your mic.
(You can get them reasonably close, though. Like within 0.1db or less if you try hard.)

Essentially this. I also prefer modding to EQing. You can get the HD800 to sound neutral, believe me.

I've not seen one measurement that wasn't too noisy to show how much good headphones like the SR009 and the HD800 really distort at realistic levels, except for the low bass where distortion does hit measurable levels. The innerfidelity ones don't get close, either. That's the reason why the 100db curve is below the 90db curve. The rest is a function of SNR varying with frequency due to its FR.
In theory the SR009 should probably destroy the HD800 in distortion, but the SR-Omega measurements I've seen that show 0.03% in the midrange at 98db do show more distortion than what I've measured for my HD800 at 95db. It's probably close.
For accurate distortion measurements you'd probably need a 1" microphone. Those tend to be very bandlimited though. Forget measuring ultrasonics with one of those.

The real great thing about the HD800 is its driver. That driver has no resonances until 38kHz. Very good FR. 5db more efficient than HD6X0 drivers (its housing wastes more energy, so in headphone use it's closer to 2.7db). I'm not even sure if the SR009 driver has this impressive CSDs tbh. The stock HD800 doesn't have impressive CSDs, though.

 

Lol, I've told you before. I know you can't believe it and, to be honest, I also expected the measurements to look more like my speaker in-ear measurements before I first tried it. The thing is, part of the direct sound of the driver almost completely bypasses your outer ear, so you don't get any of the ear-gain around 3-5kHz. The only compensation I use is for my microphone.

In the end I think it only comes down to which measurements we find more accurate. With every headphone I've measured the ear-canal opening measurements were more accurate in a FR sense compared to what I hear than the coupler measurements, but that doesn't have to be true for you. Again, I don't compensate for the ear-gain, it's part of the equation. There's no 3-4kHz dip in my compensation curves.

This is another reason why the coupler measurements don't really work. They don't include those effects even if it's a big factor of how we hear. What I mean is that if two different headphones have almost the same FR on the coupler, they don't have to have a similar FR on the head.

(I'm not saying that coupler measurements are useless, but IMO they not very good at accurate FR measurements. They are very good for checking if your headphone has resonances or not, where the in-ear measurements fail in the treble.)

This is also what I'm wondering. I bet the measured difference between 0.001% and 0.005% or 0.05% with sine tones may be deceiving because with music the distortion ends up being much higher than that.

The FR of the driver is much better than the HD600 driver's. Efficiency matters if you build speakers. Even the HD800s sensitivity is very low. I think about 82db/1m/2.83Vrms.

(I'm not saying that I want to use the HD800 driver for speakers, but I do think it correlates a bit with some aspects of sound, better efficiency generally being better as long as you don't introduce tons of resonances.)

 

Depends on the angle and various other factors. That one was head-on (nose facing speaker) at 6m distance in-room where things will look different from a 30° angle. You will get much more of that bump if the ear is facing the speaker. I feel a stereo reproduction should mimic a mono source FR. Or what others call the "BBC dip", but it's a little more complicated than that. The speaker target FR IMO depends on the room, angle, directivity of the speakers, ...

No, it doesn't. Here's my HD600 ear-canal measurement without any compensation for ear-gain. If you don't believe me that's fine. @Hands also does it like this.

You don't get the same bump at 3-4kHz. I also mentioned that the ear drum FR target will be different for speakers than for headphones. A few months later Tyll also said something like that in one of his articles. I don't think measuring speakers with a dummy head is the right approach to making a new headphone target FR, but it may work somewhat for measuring speakers, but measuring speakers with a dummy head is overcomplicating the issue, I think.

That may be true, I was talking about the drivers. The HD800s housing is fucked up while the HD600s housing may be the best thing Senn ever did in terms of FR.

True.

Yeah, I was talking about an equivalent HD800 driver 2pi speaker measurement at 1m.

 

Attached my compensation files. Here's measurement file (HD600+HD800X+Utopia):
https://www.dropbox.com/s/fuqnoledgqlhh2z/HD800 vs HD600 vs Utopia.mdat?dl=0
The measurements look a little worse without compensation because my mic is half-broken by now. The "In-Ear" compensation file compensated for a 5kHz resonances that I don't hear as measured and the huge upper-treble rolloff that also isn't there. No compensation for ear-gain in the upper mids. The Utopia needs its own compensation file because I think something went wrong with the measurement, which should explain the dip. Atleast that's what I think.
Use both the In-Ear and the mic compensation, but feel free to look at them without the in-ear compensation.

 

Those were different studies, with the mic either at the ear-drum level (for headphones) or when measuring speakers. Ear-canal opening measurements for headphones do look like the ones Hands and I post. I've found other measurements that show the same thing (like the Audeze ones, or ones on HF).
Let me show you another thing: Here's a measurement of my Riva S showing the differences between "headphone mode" and when used as a speaker:
https://www.dropbox.com/s/sk2uzhdolvih7er/Ultrabike.mdat?dl=0
Two measurements at a slight distance from the ear (about 1/2 ft) and two measurements right next to the ear. When it's close to the ear it bypasses most of the outer ear gain, like headphones do.

The distance makes a big difference, even though the Riva S is essentially a single-driver (the front facing speaker). It doesn't have as much bass up-close because of the passive radiators.

 

Well just to throw it out there.. here's my HD600 with Realiser in-ear mics vs what Serious posted above..

 

I'm not sure. You can try it with the dummy ears, but you have to get it really close to the driver. Those are 2-way speakers, right? That'll get you some issues. I would try to get it as close to the tweeter as possible. Like so that it's almost touching the tweeter.

That looks weird. Might be the coupling (mic to ear canal) or maybe the insertion depth. The measurements me and Hands post are with the mic flush at the ear canal opening and with the ear canal completely blocked off (eartip with glue).
The realiser uses some sort of foam, doesn't it? That will change the results completely.

 

Fixed the image, now yours is without compensation...

Yes Realiser has the mics in foam. Used them as manual instructed, fully inside the canal, flush with opening. Always averaged measurements 3+ reinsertions etc, not much difference there.

 

Oh, that means the microphone is facing the eardrum? That puts the diaphragm much closer to the eardrum. Yeah, not surprised that it looks more like that. I'm kind of surprised that you get so much less bass from your HD600, but mine seems to sound almost as bassy as the HD650. I removed the rear-foam.
I think that measuring like this most likely works well with the Realiser. They probably tried many different things for their measurements. I'm sure they know what they're doing, but I don't think it gives the most accurate FR (and I don't think that's the point). I don't have any experience with the Realiser.

(BTW: I found your HD800 measurements on HF from over a year ago:
http://cdn.head-fi.org/0/00/00bfbf00_hd800_hekeli2.png
I think they're actually reasonably close to what I get with my in-ear measurements with the stock HD800. Stock HD800 is not even close to neutral.)

 

I think it's just that with the Realiser the microphone is close to the eardrum with the ear canal opening mostly open, with only the foam blocking it slightly. In my case I completely ignore the ear canal. The microphone is where the hole for the ear canal is, facing outwards.

 

I think you got it backwards.. Realiser foamtip size is chosen so it blocks the canal snugly. Obviously the microphone is "listening" outwards, you can see the pinsized hole in pic above. If it was inside the canal, how could it hear anything..