Well actually not quite. What I am suggesting is two independent windings (or transformers) with 6.3V 1.2A (or greater) windings.
Let me explain. In a DHT the noise induced on the signal by the heater voltage is equal in magnitude to that due to B+ supply ripple. This is why DHTs are specifically designed for DC heaters. It is possible to use a balanced AC heater but the noise immunity for such will always be suspect. As such, given the requirements for high fidelity, a set of DC heaters are a must.
The requirements for the heaters are being driven by ripple (noise) and channel separation. The target ripple requirement for the B+ supply is .003% or -90dBv. Since the heater ripple is induced directly on top of the bias voltage this means that the requirement for heater ripple is...
Vr = 71v * (10 ^ (-90/20) ) = 2.25mV
Due to the incredibly low effective load presented by the heaters, 4.17Ω, acquiring this level of filtering on a high current circuit with just passive filtering will be virtually impossible. As such I am strongly
recommending the use of LT1085 low dropout regulators to supply the heater circuits. These regulators give almost 70dB of ripple rejection by themselves and I believe that by using them the 2.25mV total ripple will be achievable.
This is the circuit I am proposing for each 300B heater. Please note that the heater circuit is not grounded in any way. Each heater supply must float approximately 71v above ground and will be ac coupled to ground via the driver stage cathode coupling capacitor.
I know that this is a violation of your normal approach to building amps (Retro-Thermonic) but I believe that in this case the use of these regulators is warranted. The LT1085 is a low dropout regulator with excellent startup characteristics that will handle the difficult job of initial heating of the cold filament. In addition, the snubbing of the heaters with a 0.1µf tantalum might also be a good idea.
The other requirement driving the heater supply design is the channel separation requirement. In order to preserve the soundstage of the amp, a channel separation of at least 70dB is required. One of the reasons that 300B amps do so well as mono-blocks is that the channel separation is guaranteed because the only connection between the amps is via the very low impedance mains. If both heater circuits are in the same box then either two magnetically isolated transformers are required or the DC circuits needs significant buffering to prevent crosstalk. The reason I recommended the dual secondary Hammond is because with each regulator offering 70dB of ripple rejection, the channel separation via conduction is guaranteed to be 140dB plus another 10dB or so for the coupling loss in the transformer. Thus preserving a deep and wide soundstage.
I hope this sheds some more light on the heater supply design. It's not final as there are a couple of prototyping experiments I would like to try, but this should give you some insight into my thinking on the matter.