It uses a MOSFET (sort found in PC switching power supplies, I used a 2SK2098) a zener diode (to set the voltage for the output, and a LM393 comparitor chip. It also requires that the powerline/mains voltage be a true sine wave, not a "modified" sine wave (which is really a square wave with some dead time). The comparitor chip determines when the rectified but unfiltered supply voltage is, further from zero, or closer to zero, than the zener reference voltage. (think of absolute value here). This in turn switches on or off the MOSFET via a buffer transistor, (any PNP like a 2N2907). When the MOSFET is switched on, the supply voltage is closer to zero than the zener reference voltage is. The output filter cap C2, having discharged some since last time (a cycle of AC waveform ago), will then be recharged when the rectifer diode AND the MOSFET conduct. The MOSFET is switched off when the rectified but unfiltered voltage exceeds the zener reference voltage. Thus the output cap C2 is charged up only to the same voltage as the zener reference. The timing (over the course of a cycle of AC sine waveform) of the cap C2 recharge current pulses is different than the usual B+ filtered rectifier recharge current pulses (that recharges its filter cap at the peaks of the high voltage secondary AC sine waveform). This timing mismatch should reduce stress on the power transformer. Note that we need the rectifier diode to keep the input 12VAC waveform source from reverse draining the C2 cap voltage charge. And there's a separate rectifier circuit to get voltage to run the comparitor circuit.

Note that the circuit is "upside-down" from the usual viewpoint, that is, it produces a regulated (albeit with ripple) negative voltage with a positive ground. This is because most MOSFETs I had on hand were N channel, and I'd need a P channel to create a regulated (albeit with ripple) positive voltage. But a tube heater isn't going to care.

Here is the full wave version. As long as the AC source is free floating, you can ground one side of the output. If one side of the AC source is grounded, the DC output would appear to hop up and down at 60Hz rate, which would negate the point of using DC on a heater.

*Pro: little waste heat from this circuit vs a linear voltage regulator.

*Con: there is ripple, just like you have with a simple filtered rectifier. But the overall output voltage won't vary if the input powerline voltage varies, and you can select this voltage with the selection of the zener diode. Also no short circuit protection, though the MOSFET I used could pass 20A. I used a PTC (positive temperature coefficient) device as a fuse.