So here are the results of an interesting test prompted by a paper by Benchmark Media:
https://benchmarkmedia.com/blogs/application_notes/12838141-headphone-amplifiers-part-1
In a nutshell, it says that real headphone loads substantially increase the distortion of the amplifier versus dummy/resistive loads. Here is the example they give of their headphone amp and that of two other unknown units:
With dummy load:
With Sony MDR-V6 headphone:
We see that their line in navy blue at the bottom does not change. The distortion profile of the other two amplifiers change a lot.
Here is my version of that test, with a 150 ohm dummy load, volume set to -10 dB on front panel (to keep the headphones from being damaged) and four different headphones including the MDR-V6:
Starting with 150 ohm dummy load in cyan, we see that the distortion profile is pretty flat now unlike the max volume that showed a rise in low frequencies.
When we change the load to headphones, the distortion profile wildly changes unfortunately. So there is some secret sauce in Benchmark headphone amps that the RME does not have.
Based on running this test on other headphone amps, the HifiMan HE400i seems to cause least amount of variations. It is essentially the same as dummy load in mid-frequencies where our hearing is most sensitive.
The grado is brutal, causing such large swings. Looking at its impedance curve does give much clues to this variation:
https://www.innerfidelity.com/images/GradoSR60i.pdf
Comments?