How do we know if something is actually "neutral"?

[vid]

If you play a sound (from a flat-response source) next to the ear and record it with a (flat-response) microphone at the eardrum, the frequency distribution of the recording will be different from that of the original sound. The difference is the head-related transfer function, i.e. HRTF: how the shape of the ear modifies the sound passing by it.
 
The raw response (measured at the eardrum) attempts to show the original frequency distribution + the effect of HRTF; while the compensated response attempts to portray how you actually perceive that sound, i.e. how your brain interprets it, by subtracting the HRTF from the raw response. Since the shape of your ear is modifying every sound that comes in, your idea of a neutral sound includes that effect, meaning that a sound with its energy evenly distributed (flat) will only be perceived as flat if it comes to your eardrum as non-flat (flat + your HRTF). A flat sound at the eardrum, on the other hand, would not sound flat to you unless you have no ear at all and just an exposed eardrum.
 
You've got it right that the raw graph shows what the ear 'hears', but of course you don't actually hear with your ear (you feel with it) - the brain does the hearing. And the compensated graph is what tries to display what the brain perceives, i.e. what the listener hears.
 
Then again, if I've got it wrong, corrections are welcome.

 

[TMRaven]

I could be wrong too, who knows.  I'm just trying to understand it.
How come speakers aren't measured with hrtf then?  It's still sound entering the ear.  
 
 
For reference, Tyll considers the LCD-3 fr nearest to the ideal for flat sound headphone:
 
http://www.innerfidelity.com/images/AudezeLCD3Rev2sn2613375circa2012.pdf

 

[LFF]

Here you go....
 
THE ANSWERS TO ALL YOUR QUESTIONS!!!!

 

[vid]

Quote:

I could be wrong too, who knows.  I'm just trying to understand it.
How come speakers aren't measured with hrtf then?  It's still sound entering the ear.  
 
For reference, Tyll considers the LCD-3 fr nearest to the ideal for flat sound headphone:
 
http://www.innerfidelity.com/images/AudezeLCD3Rev2sn2613375circa2012.pdf

 
I've no idea about speakers in general, but consider that unlike speakers, headphones are attached to your ears. You'll get a mess of sound waves bouncing between the ear and headphone baffle - and ideally you'd like that mess to resemble a natural curve at the eardrum. Which curve you get depends on who's making the headphones; they'll equalize their phones with a specific HRTF curve in mind. When you're measuring headphones, you also need to decide which HRTF curve you want to use, since there's apparently no standard and also no universal agreement on which curve is the best or most natural. In any case, you don't want headphones that sound like the sound is coming from within an inch of your ear and encased in a small chamber, thus you (the manufacturers and measurers) muck about with HRTF.
 
Tyll himself has expressed some doubts about his compensation curve (most recently in that Zoro review). Indeed, if the compensated graph doesn't look flat for headphones that sound flat, you're possibly using a misleading compensation curve.