It's just that what's measured on an ear simulator may not be what happens on your own head below threshold of audibility, that's all
.
To the point where the question of "follows the Harman curve" or even "what does the Harman target actually sounds like ?" is not a straightforward subject to answer IMO.
I think that most of us who have EQed headphones with presets based on ear simulator measurements would have noticed that headphones still sound different, occasionally quite jarringly so. There’s nothing new here, and it’s something that’s fully acknowledged by, for example, Oratory :
https://www.reddit.com/r/oratory1990/comments/gbdi7v/_/fpay3b5
It’s something that I tried to illustrate, as far as my own experience is concerned, in this post from a few months ago, with on head, in situ measurements :
https://www.audiosciencereview.com/...snt-like-this-curve.19668/page-22#post-844726
While I rarely measure headphones with EQ presets as, these days, I’m starting to get a fairly decent idea of where I prefer the FR to land anyway, I recently measured 8 headphones with Oratory’s EQ presets and without, with three different types of mics (DIY probe, in-concha microphones, blocked ear canal entrance microphones), in an attempt to get an idea of what the Harman target exactly sounds like.
Perhaps I'll share my experience in a more complete post with more headphones involved, maybe a new thread, after sending my own samples to Oratory if he's interested, and re-measure them with his generic profiles and profiles tailored to my own samples, at some point in time in 2022.
But for now that's how these eight headphones measure, relatively speaking, without EQ, with my DIY probe mics (averages of five individual traces, right channel only, normalised across one octave centred at 500hz - remember that
the absolute values are inaccurate, and that
these results are not valid for you, using the exact same tools and methodology on your own head would yield different results):
View attachment 169341
And how they measure with Oratory's latest profiles applied :
View attachment 169342
How confident I can be in the
relative results from that DIY probe is a long boring ass subject that would probably be best left to a new thread but involves criss-crossing the results from different types of mics and compare the "different difference" between their results in the range where they logically are the most relevant (ie, there are good reasons why my DIY probe and my in-concha mics with open ear canals are agreeing somewhat well in the 2-3kHz region, but the blocked ear canal mics don't). The TLDR is : not confident enough that it’s below threshold of audibility, but quite enough that the degree of inaccuracy is a good deal lower than the residual differences observed. See the spread at 2kHz after EQ ? Well at that frequency I'm
very confident that what I'm actually hearing is similarly spread.
I'm also tempted to suggest, given the magnitude of some of the residual differences, that listening tests alone should be enough to attest to some these differences.
So Oratory’s profiles were quite successful to bring the headphones quite a bit closer to each others. That’s particularly the case for headphones with a significant departure from the target, even more so the case when it happens in the range where ear simulator are the most accurate to everyone’s experience (such as the headphones with a strong dip at around 1500Hz).
On the other hand the residual differences are still quite important and audible, whether it's because of sample variation, profiles incompletely correcting the FR, HPTF concerns, pads breaking in / wearing in, etc.
Some of the residual difference is “on purpose”, in the sense that Oratory’s profiles don’t correct for it. This is quite possibly because
some of these nulls are of the “difficult to EQ” kind and Oratory preferred to leave them alone (the 3.6kHz or so null for the red trace for example).
While the question of sample variation is tempting to raise (and one of the reasons it could be interesting to have my own samples measured by Oratory - I have reasons to suspect that some of the HPs involved deviate from each other after correction indeed because of sample variation), I'm already pretty certain that it is
far from being the sole valid explanation. For example, in the lot above there actually are two different samples of the same headphones, and both produce the same sort of peak at around 6kHz that the profiles don't account for.
And in the case of the APM sample variation is stupidly low. That's already something that I tried to illustrate in this thread.
But I realised that if I used the exact same pads at the same state of wear it's even better :
View attachment 169349
Two APMs manufactured two months appart, average of five individual traces, same pads for both APM, traces not normalised (same volume set on the source device), blocked ear canal measurements, same channel.
It's not that the pads have poor tolerances, far from it. In fact I'm utterly impressed by their tolerances compared to other headphones. But I have some reasons to suspect that at least some ANC headphones are quite sensitive to some variables in the 1-5kHz region (here's a rather jarring example with the QC45 :
https://www.audiosciencereview.com/...d-fi-and-sean-olive.27017/page-12#post-935561)
The pads breaking in over a few weeks do have an influence in the 1-3kHz region BTW.
So that's why above 800Hz, I just don't know what the Harman target is supposed to sound like, or measure like (on my head). It's quite tempting to average all of the traces after EQ, but I have a lot of concerns about doing this as a way to extract what the Harman target is supposed to sound like. Let's just say that I am far more confident in the fixture to fixture and fixture to head translation of the Sennheiser HD650 than some other headphones.
And, interestingly, it also would mean that in the ear canal gain region, it would disagree quite audibly with the two Harman products I already have (K371 and 710BT), or with the headphones that I find somewhat acceptable as is in that region.
So it's a bit of a puzzle.
Below 800Hz however, the AirPods Max allow me to know quite well what the target is supposed to sound like as the fairly robust feedback mechanism combined with the lack of FR variation with volume means that they can deliver an exact dB value at my eardrum for a specific digital value, and that my on-head measurements can be "matched" with ear simulators, as long as the sample variation is low (which it is, particularly in that range), and the ear simulator measurements don't exhibit methodological issues (which can happen with the APM but less so when ANC is enabled).
In no way is this a criticism of ear simulator measurements BTW, I'd like to make this very clear. Nor is it a criticism of Oratory's profiles, far from it (in general they're the ones I tend to personally prefer and I quite enjoy how at least for the bass shelf they're designed to help with tuning for individual preferences). It's just a way to illustrate in a visual way that I think is reasonably representative of my own experience something that I would have otherwise been limited to simply say "still sounds different after EQ to the same target", and perhaps a way to illustrate that the question of following the Harman target is not quite clear-cut once you're experiencing your own samples on your own head.