That it's wrong, sorry
.
But it's a long, long explanation.
I too share the idea that we should be hearing what the recording / mixing / mastering engineers intended us to hear. Even if we're talking about Brokencyde for which I'd rather listen to with an all spectrum -100dB filter.
But that seems quite a bit difficult to achieve.
So for a start, Adaptive EQ (which is something that has already existed in various form with ANC headphones for a while, albeit perhaps not quite as sophisticated as Apple's implementation) doesn't work above 1khz :
https://twitter.com/oratory1990/status/1343323936801644544?s=20
See the faint grey traces ? That's the multiple measurements with slight positional variations on the dummy head that are then averaged. They coalesce into one trace where Adaptive EQ can work its magic. Above, it's a mess, and beyond the audibility threshold.
The same sort of variation will happen on your own head.
So take away n°1 :
headphones' FR curve vary with position over your head, not just seal quality. Adaptive EQ is a great solution to sealing issues, but not to positional variation.
Also, since we all have different ears, heck since even our left from right ears aren't the same, frequency response above 1khz will vary between us, again above the threshold for audibility :
From that study I believe :
https://www.aes.org/e-lib/browse.cfm?elib=16877
See the thick grey trace behind the solid black curve ? That's how FR curve near the eardrum of the test subjects vary (measured by probe microphones). You can notice that it can become quite thick (ie high variation) at lower frequencies with some headphones (that's a sealing issue), and above 1khz
with all of them, and even more so above 5khz (that's variations related to the shape of your ears / canal).
Now it's possible that some variation in that range is desirable, as we all hear differently the same sound source in a natural environment (that article :
https://en.wikipedia.org/wiki/Head-related_transfer_function).
But the problem is that headphones seem to interact with our ears in a way that is quite a bit different than natural sound sources or a pair of speakers. As a result we see high variation between listeners, and a variation that is undesirable.
There aren't a lot of good solutions to that problem at the moment. EQing this range by ear is a PITA.
But theoretically the headphones of the future will be able to gather enough anatomical data in real time of your ear shape that they'll be able to
a) tailor the FR curve to your own anatomy so that the FR curve delivered at your eardrum is similar to what a natural sound source would have produced,
b) modulate it in real time to compensate for positional variation.
So take away n°2 :
headphones' FR curve increasingly varies above 1Khz across several listeners because of anatomical variations, in a way that could be at least partly unnatural, and remains quite difficult to EQ well, particularly, IMO, above 5khz.
But we'll still need to EQ it at some point
.
Also, Harman's own research fully acknowledges that personal preferences variation come into play, particularly in terms of bass response below a certain frequency - independent of seal, positional, or anatomical variation :
https://www.headphonesty.com/2020/04/harman-target-curves-part-3/
So take away n°3 :
People's preferences, particularly in the bass, may differ quite a bit.
And finally hearing damage or age can modulate what you hear and perhaps require some form of compensation over time.
So in regards to the AirPods Max FR curve and its capacity to deliver recordings as intended by the engineers / artists :
- Below 1khz it measures very rationally (ie corresponds fairly well to Harman's research, that is to say to a "decent pair of speakers in a decent listening room", with a sub bass bump according to the majority of preferences, and is very smooth with no weird bumps or wiggles,), and theoretically Adaptive EQ can ensure that it's quite constant across listeners and immune to sealing variation. But it doesn't feature a bass shelf EQ adjustment which we know should be there according to Harman's research.
- Between 1khz and 4khz the general shape of it corresponds very well to an average of how humans' anatomy naturally modulate the FR curve, with the ear canal gain peak at around 3000hz for example. Top marks for that. But the whole region is quite depressed compared to what Harman's research would suggest corresponds better to "decent speakers in a decent listening room". Difficult to know exactly by how much and as it's in the range where we start to see anatomical variation across humans it's difficult to say "2.7khz is 4.75dB below where it should be". But given how many people have commented on how they prefer it with the "balanced" setting in headphones accommodation it's quite likely that this region is perceived as depressed by quite a few people.
- Above 4khz measurements show a very high degree of variability, in fact even more so than quite a few headphones,
even on the same type of rig (GRAS). Difficult to extract any information from them.
- The capacity to apply an audiogram to change its FR curve might be an interesting solution to the latter point.
BTW I just had a bit of fun using in ear mics to measure some of my headphones to see whether third party measurements can be relied on to EQ headphones to below audibility differences. Don't take it too seriously, I don't have the means or the knowledge to do it to a particularly rigorous degree, but the answer is most likely not :
https://www.head-fi.org/threads/the...-at-a-breakthrough-value.943107/post-16300055