[k00zk0]

When EQing a system, what is the ideal response? Such that every frequency of a sine or pink sweep sounds the same? Or such that every band of ____ curve sounds the same? There are tons of weighting curves. Some users say to make pink noise sound equal loudness. Some post complexly weighted sine sweeps which are generally louder at the mids than at the bass and treble ends. This looks like the equal-loudness curve. What is the ideal here? If you recall seeing where a user posted a weighted sweep, please link as a white or pink weighted noise/sweep that some are using is definitely not the ideal EQ reference.
 
The EQ guide says to eliminate peaks due to resonances of the ear canal and headphone, but definitely the drop in lows and highs generally seen with single-driver headphones, compared to full-range speaker systems, are to be corrected as well. All of these weighing and equal-loudness curves follow a similar trend: the ends are to be raised to correct for the fact that the single dynamic headphone driver is generally more efficient in the middle of the audible range.

 

[MalVeauX]

Heya,
 
Depends on your goal.
 
If you're equalizing to achieve neutrality, or a flat line, you want the perceived decibel to be the same across the board. Note, I said perceived, because when you get to higher frequencies, it takes spikes and troughs to actually achieve the same perceived decibel volume due to the shape of the ears and how sound is refracted/absorbed/collides.
 
Very best,

 

[k00zk0]

Sounds fair, but the perceived decibel of uniformly powered white noise? Because when you do this, response lacks tremendously on the highs. The HF in recordings is at a much lower signal level than than LF. This seems like what pink noise is for, right? Much lower hf so that in equalizing, it is leveled to the way you hear it.
 
It still seems like something smiley-face shaped is more of the ideal. Every weighting curve has raised bass as well as raised highs, reflecting the quiet highs in recordings, as well as the lack of subwoofers in headphones. Is it the equal-loudness contour you shape a sweep with, which seems to follow pink noise distribution, until some more complex areas, or what?
 
I actually very accurately modeled my headphone response in a parametric. It took 29 filters. A sine sweep spread over anywhere from 10 seconds to 10 minutes will not have changes in loudness, and picking random frequencies will not yield any either.
 
Unfortunately it is fatiguing and I found myself not automatically going for music any more. When I disabled it and listened raw, the music sucks me back in. But, it sounds incomplete somehow and not like speakers do, flat across the board.

 

[stv014]

In theory, you want to equalize according to these curves (click to zoom):

For this to be accurate, you need to calibrate the SPL level at 1 kHz (if not having a better way to do it, then by measuring the voltage on the headphone, and calculating the SPL from its impedance and sensitivity), and preferably do the equalization with the minimum amount of outside noise.
However, it is not easy to do in practice, as comparing the loudness of tones at different frequencies is difficult, and it can also get fatiguing after a while. I experimented with this and created a FIR filter for the presumably "equalized" response, but found the actual results rather awful when listening to the filtered music.