http://gilmore2.chem.northwestern.edu/library.htm could keep you in reading material for a while
I cite this one a lot http://gilmore2.chem.northwestern.edu/articles/hearing_art.htm - read twice - for protecting yourself and then for Dynamic Peak SPL estimates - which lets you size a amp for a given headphone's impedance, sensitivity
for the HD-650 getting to 120 dB SPL peak requires ~ 12 Vpk, 40 mApk - this is a pretty high amount of dynamic headroom you may never have a clipping event depending on music genre and whether you really ever try to literally reproduce live event SPL peaks
yet with a op amp + buffer it is a pretty easy goal to meet for the HD-650
another overkill approach common in headphone amps is to run the output in Class A bias - avoids a few error mechanisms and at headphone power levels it doesn't cost too much for desktop amps
the current and power to do this with the HD650 are within the limits of some op amps but the easiest way is to use a separate buffer amp giving greater freedom for choosing the input op amp
I strongly advocate putting the buffer inside the op amp Global feedback loop - Walt Jung has articles, explanations, check out his Op Amp Applications book (free online) - particularly the audio line diver circuits - section 6 http://www.analog.com/library/analogDialogue/archives/39-05/op_amp_applications_handbook.html
like Jung I prefer current feedback op amps for the output "buffer" - the TPA6120 gives an impressive spec as a headphone amp on its own - but I still wrapped a OPA627 around it in my amp (avatar)