EQ-ing headphones by measurements

I'm experimenting with methods for EQ-ing headphones based on available measurements. I'll share my subjective evaluation in the end, using measurements of headphones I own. This may or may not yield satisfactory results. I'm keeping an open mind while I'm working out the methods.

Steps to take:
Choose a target - This is a no-brainier, you can't work without a goal. The most recent extensive work has gone in to the Harman target, last updated in 2015 (blue curve).

Find a graph created using a HATS - Measurements on HATS, rather than simple rigs, is basically a prerequisite for the Harman target, but differences between HATS systems will cause unexpected results. For now I'll leave it as-is. TODO: Work out how to factor in differences between HATS systems.



Extracting the raw responce curves - Innerfidelity provides raw responce curves, rendered as overlayed discrete curves for each position. The compensated curves have been been averaged and are easier to work with but must be de-compensated with the following curve:
http://www.innerfidelity.com/images/110401_measure_graph_headacousticshrtf.jpg
TODO: Find the compensation curve used in older Golden Ears graphs.

Creating a preliminary curve - I convert the graph curves into .csv files using WebPlotDigitizer and load the .csv's into a chain of three instances of CurveEQ.

The first has the reverses the responce curve, effectively flat on the measurement rig. The second has the compensation curve inverted, effectively electronically flat. The last one applies the Harman target. The CAL! are now EQ'd to match the Harman target. In theory the sound should be excellent, but on my CAL! the sound is unlistenable. Let's see what you can do about that.

Tweaking the final curve - For different reasons, such as differences between HATS systems mentioned earlier, you have to modify the curve before it actually improves the listening experience. A simple way is to scale the gain of the curve. I carried this out by processing white noise though my EQ chain and outputting as .wav files. This allows me to capture the curve with FabFilter Pro-Q 2, which applies the curve as a minumum-phase IIR filter. This type of filter has a faster transient responce than FIR or linear-phase. After critically listening to various recordings I settled on a gain scaling value of 18.4%.


Another method involves thresholding the Y-axis, similar to Mathaudio Room EQ. This avoids overcorrecting nulls. It can also be combined/blended with the gain scaling. TODO: Evaluate Y-axis thresholding.


Results - I won't use audiophile terms, I'm primarily a musician. Tonally on my CAL! the sound strikes me as having less of a murky and stuffed quality. Impact and intelligibility has improved, especially on bass-heavy recordings. Aspects of ambiance come through that went unheard before. Overall more neutral but the treble has a slightly tiring quality.

TODO: Other measurement sources/sites. IEMs and different types of occluded ear simulators.

 

hint:

Accudio
Sonarworks

They work on your principle.
That principle has inherent faults when you start using rather complex EQ settings based on measurements.

I suggest to use simple/gradual EQ instead or ... buy a better headphone.

 

It's may be early to say it's a flawed principle. First you must identify what the flaws are. I took a shortcut and applied a partial EQ, and this shortcut isn't in its final stage of sophistication yet but the results are promising.

 

I'll withhold judgement on what's autistic or not. *shrugs* EQ-ing by ear, I've done that, using the same EQs I use for audio production. Honestly my current method I concocted is much faster. The issue is how your brain adapts and "fills in the blanks", especially on your favorite reference recordings, making you have to revisit your work, which takes time.

 

@markanini, why don't you head over to the New Member thread and let us know what you are about.

 

Consider this: you're choosing the Harman curve as your personal target... however, this curve is actually just an average of individial study participants. Now let's look at one example of measured individual responses in this study, and let's for the sake of the argument imagine that you are one of the participants:



So, with these measured individual responses varying up to 12dB between 2 - 4kHz, yours might be the one that's 6dB below average. Or possibly the one that's 6db above average. See where I'm getting at? The Harman target (as well as other averaged targets) is certainly useful as a general reference. But imho not a sound basis for individual EQing.

No harm using it as a starting point. But, as others have said, rather learn to use your ears for personal fine-tuning.

 

You can use your ears if you know how neutral sounds. Best bet would be using a rig which allows for fast switching between the headphone in question and a calibrated full range speaker system in a well treated room.