Thanks for the reply. So only way to do is measure a known headphone and compensate the difference.
So it can be done without that artificial ear also.
yes, it can be done without an artificial ear BUT mounting a mic on a hard (echoing) surface will create other problems that need to be compensated as well. Care must be taken to mimic thin skin on bone.
Also the space the ear takes up as well as the 'shielding' of soundwaves from the rear also has some influence on the plots you get that way.
To make things worse... for instance when you measure 2 different headphones on the same 'flatbed' rig and add a Pinna (so without the ear canal) the plots you get may differ yet again.
A speaker sound comes from your in front of you and the ears are 'trained' to interpret that.
A headphone sound comes from the side and the pinna works VERY different from that angle as from the front.
When you listen to your speakers from a few meters in front of you and then move them much closer and on the far right and left the same speakers will sound VERY different.
This is why it is hard to measure headphones 'correctly' and the 'needed' compensation is debated quite often.
Who is right, who is wrong or perhaps who is more right or wrong ?
If the plots 'show' what you are hearing you will be close.
I would dare to say that trained ears are almost a must if you want to 'calibrate' that way (by ear).
Now I have lot of iem for which graphs are available.
But for headphones I have just vintage orthodynamic headphones which I need to modify.
For in ears you can get away with a specified length of tube but personally think the artificial ear is more suited.
As Marvin (Purrin, Marvey) also clearly states measurements done above 9kHz are not very reliable as the compensation you need will differ yet again and the insertion depth as well as getting a good seal may not be that easy to achieve.
Here is example of zebrano wood T50rp I built.
http://cdn.head-fi.org/9/99/99f28a47_t50rp-wood.jpeg
http://cdn.head-fi.org/4/4f/4f320076_t50rp-wood-wf.jpeg
It is hard to evaluate the plot if it isn't clear how it was created.