Here is a bit more explanation about the graphs and measurement artifacts. Above 2kHz it is very difficult to measure headphones with dummy heads.
When transducers are measured, typically an IEC-standard baffle is used, where we see very tight and consistent response. Audeze’s transdcuers are matched in pairs using IEC Baffle.
Judging headphones’ performances above 2kHz just by using graphs is quite difficult. The dummy heads that are used can also be very different. There are several different models, and each comes with several sets of ears (small/med/big, male/female, etc). On top of all these variations, the microphone placement is also quite different. Some have the microphone near the entrance of the ear canal, some have it near the ear drum. If you measure one with a blocked ear canal then you don’t see the ear canal resonances, which are very important. We use 4 different types of testing rigs to measure headphones during manufacturing.
The ones we publish use a Neumann KU100. In this model, the microphone in the test rig picks up sound at the entrance of the ear and the positioning of the headphone makes a big difference in measurements above 2kHz. Slight changes in position results in significant changes in the measurements. When testing, we usually take 5 different position measurements and average it (Similar to what Tyll at Innerfidelity does). When using other rigs (eg - IEC baffle), this is not necessary and one measurement is sufficient.
Here is a measurement of a non-audeze headphone taken from innerfidelity.
Here is the same audeze headphone measured with slightly different positions
Conclusion:
When measuring headphones on dummy heads, graphs will have measurement artifacts. This is true for not just Audeze headphones, but most other headphones as well. These are measurement artifacts, so you shouldn’t read too much into them and/or treat them literally. If you measure the same unit 5 times in an IEC baffle, the lines will be virtually indistinguishable.