I think that this study's methodology makes a bit immaterial to know which pair of headphones was which. It was discussed in this thread (that post and subsequent posts) :
https://audiosciencereview.com/foru...ng-society-2023-conference.48497/post-1741065
The issues are :
- they made a very poor choice for the reproduction headphones : Beyer DT 770 Pro. This pair of headphones will have more coupling issues and inter-individual variation than a pair of headphones like the HD800, for example. Besides, if it is like the DT 700/900 Pro X, the dip at around 4kHz might be of the kind that's resistant to EQ and makes comparisons in that band a bit pointless.
- they used a 30 band GEQ to EQ the reproduction headphones to the different FR curves under test (whether the approximated targets, the headphones, or the modifications to these curves). This prevented the reproduction of sharp peaks and dips, and as some headphones' individual features have been shifted on the X axis, the GEQ could have had a role to play in that regard.
- even still, it seems that the approximated targets, which were created by Mad_Economist
in that thread, were not perfectly captured. They're also shifted on the X axis a little bit, and I am not certain that the use of a GEQ alone could have caused this. Don't hesitate to ask him about this, he's probably looked into it more than I did. The "v2" versions of the two approximated targets are a modification of them, with additional energy around 3-9kHz.
So it's really difficult to know exactly what the listeners actually experienced and compared, and in my view it is inappropriate to interpret that article with excessive exactitude. In my opinion it's best to keep it at a very general picture of the kind "people want their headphones' response to be sensibly shaped", "sensibly" meaning "very roughly in the spirit of the general shape of Harman's target", ie avoid zany responses, and by "very roughly" I mean don't get caught up in any nitpicking within a 3-5dB window at the very least (and don't necessarily think that APHarm2018v2 is the right target to use for the 5128).
The detailed preference data is freely available here :
https://zenodo.org/records/8388242
ThomasXia provided some useful graphs to analyse it in the thread linked above.
Also provided as well are the simulated FR curves using the 30 band GEQ (so not exactly the FR curves of the actual headphones).