I've found that with 2A3s, and an AC hum balance pot (usually 50 ohms, WW, or use 2 20-22 ohm 2W resistors and a 10 ohm WW pot in the middle) will get the 60Hz hum out. As the filament voltage rises, this approach can be somewhat less effective, and for 300Bs, 6B4Gs, 6A3s, etc DC is better, overall.
Though there some traditionalists who claim that the audio quality diminishes with DC filament power, I never found it to be so, as far as I could tell. Then my ears are merely copper, not golden Yellow_Light_Colorz_PDT_08 Yellow_Light_Colorz_PDT_09
The older circuits with DHTs simply used center tapped 2.5, 5 or 6.3VAC filament trannies, the cathode resistor connected to the CT, and no hum balance pot was used. This works if the trannie is well-balanced and the tube filament has evenly distributed resistance along its total length.
The balance pot works for 60 Hz (or 50 Hz for overseas) AC only, the 120Hz (or 100 Hz) PSU ripple is a whole different situation.
Any of the JE Labs circuits are fine, though in one of Jose's "Simple 2A3" circuits he used a shared cathode resistor and filament trannie for both 2A3s, that works, but the tubes have to be well matched, and this reduces channel separation (and imaging) somewhat. I haven't trried the "Aikido" 300B approach, it looks like it should work well, and John Broskie really knows his stuff, so I see no problem other than increased circuit complexity.
FWIW, Jeff Larson's Aikido 2A3 link is here:
http://home.att.net/~abraxasaudio/proje ... 2a3-2.html
HTH
/ed B in NH