[BiggerHead]

I'm not the most experienced headfier here but I've got a long history on more technical matters than this and I've paid quite a bit of attention to this so I'm going to butt in here and give my take on some of this, similar to another thread about measurements elsewhere yesterday.
 
If you stand at Niagra falls, and let's just pretend that niagra falls is white noise, lets' imagine that a perfect microphone placed at some spot not inside a head would measure a flat spectrum.  Ok this is a real thing, now you go and stand where the mic was and  you hear it.  However you hear it, that's how it really should sound.*  (every time I make this star it will mean give or take the bass boost room correction, see my comment below)  It doesn't sound like a flat spectrum.  It sounds like the inverse of an equal loudness curve, which incorporates some product of curves, basically HTRC*: "transfer function" from diffuse field(or free or whatever) to ear canal and BRC: Brain Response Curve (I just made that up).
 
Now if you have a perfect speaker system (easier than headphones so lets start there),  and play back your Niagra recording it should not  sound flat to you because the real Niagra didn't sound flat to you.  So if you do a frequency sweep with your speakers you should not eq the result to a flat sounding curve.  With reference speakers of course (give or take room effects) you won't have to eq it at all, but if you eq it, it should* be eq'd so a mic in the listening position would pick up a flat spectrum.  Now if you play back niagra, the mic will pick up a flat spectrum just like there was in front of Niagra.  If you put your head there, you'll have the same input to your head as you had at Niagra, a flat spectrum, and you'll hear the same thing, not flat.  If you played an equal loudness curve, that by definition would sound flat*.  I think maybe this is what Joe Bloggs does, but not what he said to do.  (I'm intentionally oversimplifying diffuse field, free field and directional stuff, fine, but if you would have been facing Niagra, maybe that all works out reasonably ok).
 
Ok so you put on a set of headphones and you want to hear that same thing.   Tyll measures what sound gets inside his dummy ear.  Someone said he has to correct that for human hearing and someone else objected.  Well he has to correct it for how sound gets in your ear, if he wants  flat on his graph to mean Niagra sounds real.  Of course that's not correcting the headphone response for human hearing for a couple of reasons.  1) It's not measuring "the headphone response".  That's actually impossible.  There's no such thing.  There is only the response produced by the headphones, measured somewhere, in some setting, but anyway, this is far from "the headphone response".  It's measuring "the headphone response" modified already by part of the human hearing system, the ear canal.  2) It's only modified by part of the human hearing system, not the BRC (neural bits).  So it's correcting the headphone response modified by part of the human hearing system... for that part of the human hearing system that already modified the actual measurement.  That really isn't "correcting the headphone response for human hearing".  This really is an important point I think**. 
 
 
So why not just measure the headphone response with a mic outside the ear?  Because that won't/shouldn't be the same as a diffuse field niagra recording either and while it doesn't depend as strongly on ear shapes, the relationship between that and the sound at your eardrum probably does depend significantly on headphone shapes.  Now you'd need an HTRF for every headphone. This is a problem speakers didn't have and is the reason speakers can be measured outside of the dummy head.  I think this is an important detail often missed.
 
So.. he measures inside the ear and wants to have a flat input to sound to the mic like the HTRC.  If that happens and he subtracts (divides... but subtract if we're working in db ie log scale) the HTRC he should see a straight line.  Of course that doesn't happen because headphones aren't perfect.  He still subtracts though and we see how close it gets.
 
So back to eq'ing...  I certainly also didn't understand (and still don't) in the original post how the HTRC was meant to be used, step by step, in the process, but I just can't see how it could be if you're using hearing as the judge.  The HTRC is the sound in your inner ear from a flat spectrum at Niagra*.  It's not what you hear.  What you hear at Niagra is the inverse equal loudness curve, probably because of nerve and brain issues beyond that.  So how can you use a human hearing based eq system and also use HTRC as any useful reference?  This doesn't make sense to me.  What would make sense to me is an equal loudness curve played on speakers should sound equally loud in all frequencies (that's what it means).  And that same curve played in headphones should sound the same, ie equally loud in all frequencies, in headphones.  We should NOT be correcting measured heaphone spectra with equal loudness curves, that's a different issue and common false claim, but we aren't talking about measurements here.  So this actually seems very simple to me.  Just play equal loudness curves and make them sound like equal loudness regardless of what equipment they are played on.  After all, speakers should sound the same as headphones right?  Which equal loudness curve should you play?  That's the hard part.  It should be the one corresponding to the volume you are hearing when you do the eq.  Your head will modify appropriately for other volumes.  So you need a way to sort that out. 
 
 
* About that room bass.  I think it's neat that the industry has finally realized that people listen to music in their living room and the brain expects that, but the brain hears many directional cues of where the walls are with speakers.  With headphones that doesn't happen and I'm somewhere else anyway.  I don't want it then, and I think it's wrong and is just justification to play into people's apparent enjoyment of bass (which is fine, in the recording).
 
** I hope/think Tyll is still using his own measurements with his own dummy for his own version of an HTRC.  This means that he's correcting the mic in his dummy head with the HTRC for his dummy head.  There actually shouldn't be much caveat left about head to head variation.  He makes it sound right on his dummy head.  Of course it will sound different in your ear canal, but everything sounds different in your ear canal.  It's supposed to.  Niagra does too. There's no problem here unless the resonances set up between your ear canal and a particular headphone create a strong effect that doesn't exist normally without headphones and also that doesn't exist in the other head.

 

[BiggerHead]

.. and white noise is probably a bad example because it brings in other technical points but please ignore that and take it for how it's meant in the context of the argument... a real source that when measured by the mentioned mic, produces tones at all frequencies (one at a time if you like), in equal volume.

 

[Joe Bloggs]

There's no problem here unless the resonances set up between your ear canal and a particular headphone create a strong effect that doesn't exist normally without headphones and also that doesn't exist in the other head.

The rest of your post went a bit over my head :blink: , but this last part, I think it does happen a lot of the time.

 

[spruce music]

I think the short version of the Niagra post is this.
 
If you EQ your speakers for a log sweep equal loudness result using your ears, take that EQ and modify it until you get log sweep equal loudness in your phones. You will now have an EQ that makes your phones have the same response as your speakers. 
 
Is that the essence BiggerHead?
 
If you don't like your speakers or don't use speakers then you have a bootstrap problem.

 

[AudioBear]

This has been a very interesting discussion in which I have now become totally confused. Too many good ideas and alternative approaches. 
 
Why wouldn't it work to simply put on your phones and use individual tones and log sweeps to create an equalization curve that to us (subjectively) sounds flat?  This eliminates Niagra Falls, speakers, rooms, transfer factors, diffuse fields, Ty's measurements and everything else.  It combines all of those into a curve that our brain interprets as flat.  Lunatique has already provided us the files to download. 
 
EDIT:  what I guess I am asking is whether the idea is to create physically flat response curves or perceptually flat response curves.   When I read other threads I chuckle at how many in this hobby are going the opposite direction by looking for headphones and IEMs that are fun, warm, analytical, bright, etc but never for those that are accurate.

 

[spruce music]

  This has been a very interesting discussion in which I have now become totally confused. Too many good ideas and alternative approaches. 
 
Why wouldn't it work to simply put on your phones and use individual tones and log sweeps to create an equalization curve that to us (subjectively) sounds flat?  This eliminates Niagra Falls, speakers, rooms, transfer factors, diffuse fields, Ty's measurements and everything else.  It combines all of those into a curve that our brain interprets as flat.  Lunatique has already provided us the files to download. 
 
EDIT:  what I guess I am asking is whether the idea is to create physically flat response curves or perceptually flat response curves.   When I read other threads I chuckle at how many in this hobby are going the opposite direction by looking for headphones and IEMs that are fun, warm, analytical, bright, etc but never for those that are accurate.

A perceived flat response absolutely guarantees a non-flat measured response.  Which guarantees a lower subjective fidelity playback. 
 
A measured flat response guarantees a playback with subjective fidelity.  Maybe not subjective preference in sound, but subjective fidelity to the music. 

 

[AudioBear]

 
A perceived flat response absolutely guarantees a non-flat measured response.  Which guarantees a lower subjective fidelity playback. 
 
A measured flat response guarantees a playback with subjective fidelity.  Maybe not subjective preference in sound, but subjective fidelity to the music. 

I get that but I'm not convinced perceptually flat isn't what we want with phones.  I see a fundamental difference between speakers, rooms and correction than with phones and IEMs.  We can measure the sound that is incident on our ears from speakers and correct that to flat.

 

[BiggerHead]

  I think the short version of the Niagra post is this.
 
If you EQ your speakers for a log sweep equal loudness result using your ears, take that EQ and modify it until you get log sweep equal loudness in your phones. You will now have an EQ that makes your phones have the same response as your speakers. 
 
Is that the essence BiggerHead?
 
If you don't like your speakers or don't use speakers then you have a bootstrap problem.

 
Yes.  I took aim at a few other misconceptions as I see it, but this is exactly right*.  This is simple.  Of course you can cheat, and just prepare a source sweep made with a pre-programeed equal loudness curve if you don't have the speakers to bootstrap through, and then eq that to flat in your headphones.  But the point is the right thing when using ears is equal loudess, not HTRC.  HTRC is not what a human perceives.  It's what a mic in your ear would perceive.  It's the wrong curve unless you're using measurements.
 
*Edit: (one important caveat in the technique as you described it ... the correct eq for your headphones now, maybe it was obvious, is the difference eq applied in the second step, not the total eq.  You don't want to leave the equal loudness eq in your already flat response speakers, and you don't want to leave it in the headphones either.  We are just using it as a tool to get a benchmark.)

 

[BiggerHead]

The rest of your post went a bit over my head

, but this last part, I think it does happen a lot of the time.

 
Maybe, but I think it's a smaller effect than either by itself.  Headphones do setup resonances, and different ears do propogate the sound differently, but Tyll's technique is fine with that.  The only correction it misses is the extext to which the particular ears and particular headphones combine to make the sound say very near the driver, very different in one ear and another ear.  The propagation inward (edit: to the extent it can be thought of just as a propagation inward) is properly accounted for, even considering ear to ear variation. 

 

[spruce music]

 
I get that but I'm not convinced perceptually flat isn't what we want with phones.  I see a fundamental difference between speakers, rooms and correction than with phones and IEMs.  We can measure the sound that is incident on our ears from speakers and correct that to flat.

You may be conflating two different ideas here. 
 
With speakers truly flat response is perceived as way too bright and thin sounding.  A number or reasons one being close miking of recordings is unnatural.  Another being in large concerts you are far enough away the highs are absorbed by passage thru the air.  So a slight gentle downward tilt typically will sound like a balance more like what we hear in real life.
 
Same is true in headphones.
 
The difference in perceptually flattening response with EQ is we aren't talking about that.  We are talking about a roller coaster response varying over 10 db if we do it that way.  Do that and play music and it won't sound perceptually flat in the end.

 

[BiggerHead]

Audio bear spruce music is dead on right.  aA flat sweep doesn't sound flat to a human and thus shouldn't, period, with any playback device.  You an actually measure speakers and headphones, the same, way.. in the ear, and use the HTRC for both.  Again, though don't confuse measurement and listening based correction.  Listening is not HTRC.  It's equal loudness.

 

[BiggerHead]

So including my edit in response above I guess you could say I proposed two main techniques.
 
Bootstrap technique
 
version a)
 
1) Get reference class speakers (in a reference class room).
2) Eq them so a sweep sounds flat to you.
3) Record the eq values.
4) Play this eq on your headphones.
5) Eq it further to sound flat
6) Record the HP eq values.
7) Subtract 3 from 6.  This is your headphone eq. 
The problem here is this works for graphic eq.  For parametric, you'll have to subtract the eq curves and then try to parameterize that subtracted curve, which is maybe not so easy.  Maybe good software can do this?
 
so version b)
 
1) Get reference class speakers.
2) Prepare tones at many frequencies so that they sound flat to you on the speakers
3) Play the tones on your headphones.
4) Eq the headphones so they sound flat.
5) Caveat, you probably have to take out small spikes before or after with sweeps like in the O.P.
 
 
Equal loudness curve technique
 
1) Play on your headphones a sweep or tones already biased for a generic equal loudness sweep at the right volume.
2) Eq that to sound flat on your headphones.
This assumes your personal equal loudness curve is reasonably generic, but it doesn't require reference speakers in a reference room.

 

[Joe Bloggs]

BiggerHead, version (a) and (b) are superceded by the two-EQ technique:
1. Get reference class speakers in a reference room.
2. EQ them so a sweep sounds flat to you (with a parametric EQ or otherwise). Call this your personal equal-loudness EQ.
3. Play this EQed sweep through your headphones.
4. Adjust a second EQ until it sounds flat (with a paremetric EQ or otherwise). This already is your headphone EQ.

I illustrated this process in this pic:

From *right* to left: tone generator, personal equal-loudness EQ, headphone EQ.

Added bonus, the EQed sweep of (3) can be exported for others to play with. I made my EQ guide with a video that plays an EQed tone sweep of (3) in time with the video footage that shows the exact frequency being played at any instant. All you have to do is to find a video player that supports EQ and adjust the EQ until the sweep sounds flat. Doesn't get much simpler than that.
http://www.head-fi.org/t/794467/how-to-equalize-your-headphones-2016-update

This would be the "Equal loudness curve technique" but with my personally-tuned equal loudness curves (modified in the hopes of working for a larger audience)

 

[BiggerHead]

Sorry, maybe I misunderstood.  I think you're saying your software allows you to pile eq's on top of each other.  If so, then yes this works.  If you're just saying start with the first eq and modify it some more and that's it, then I would disagree strongly.

 

[BiggerHead]

deleted.