Hi All,
It seems that, no matter what the project, there's ultimately discussion bandied about regarding running the tubes' heaters on DC. I thought maybe it was time to dedicate a thread to this subject.
The first thing I want to emphasize is that DC heating is not a panacea, nor is it a substitute for good "sanitary" wiring techniques. Think of it as a sort of four wheel drive: you want to go as far as possible in two wheel drive, but then, if you reeeaaly need it, the 4WD's there.
Output tubes, including directly heated triodes, are designed to operate with AC on their heaters or filaments unless explicitly specified otherwise. They have sufficient thermal inertia to overcome the on-off cycle of AC, provided they are operated at their rated voltage. Operating a heater at subspec voltage invites cathode stripping, while operating a filament at subspec creates a mutant of crossover/intermodulation distortion.
Sustained operation of a relatively high current heater or filament on DC creates a problem that's a side effect of the "Edison Effect". The side of the heater or (especially) filament that's connected to tne negative side of the supply will deteriorate, causing premature failure. Small signal tubes are less sensitive to this phenomenon, but it's a good idea to run them at slightly lower (5% or so) voltage to minimize the effect, just the same.
High value/low voltage electrolytic caps have become both a boon and a curse in DC heater supply design. Boon, because we now have enormous capacity in small packages available at low cost. Curse, because the "More is better" myth kicks in.
I don't know if the "ELI the ICE man" brain tickler is known by anyone except old farts like Ed Brown and me; for those uninitiated, voltage (E) in an inductive circuit (L) leads current (I), and current (I) in a capacitive circuit (C) leads voltage (E). Handy for those of us who survived the '60s. In any event, until an ultra-high value cap is fully charged, it is drawing current, and thus must be considered as part of the load on the supply. If the current capacity of the transformer is marginal, the current requirement(s) of the tube(s) plus the charge current of the cap could prevent the supply from reaching its intended voltage for several seconds to indefinitely.
This problem is exacerbated when one tries to design a pi network filter using the megacaps. The resistor between the first and second caps further impedes the charging process, as it slows the charging of the second cap (which is in parallel with the heater load) even more.
One solution is to forego the use of the megacaps in favor of more reasonable values--say 470~680 µF. Another possibility is to use a choke input L filter, but that would take a fairly hefty choke to overcome the critical current threshold. Ed's approach--brute force into a FWB with 10,000 µF hung across it--is simple and effective (but I'll bet the lights dim a little when he turns the thing on).
Tomorrow I thought we could discuss the pros and cons of regulation--assuming anyone's around for the weekend.