I'm not sure I follow the theory on current density. Yes, the resistance in series with the load could be less, but this is always small compared to plate resistance. The resistance in series with the supply is going to be about the same. The DC heat will be about the same. The AC heat is a non-issue, even considering the way the two currents sum, because we flow maximum AC signal much less than 1% of the time. The PSRR for a
pentode (does anybody still use those?) is actually better if the winding returns to B+ instead of to cathode. A triode will gain 6dB-12dB if returned toward the cathode, true. I see the point about getting the power supply out of the audio business, but I'd view that added choke as "power supply", just a hi-Z supply instead of a low-Z supply, and possibly with more audio "character" than good caps. Efficiency difference will be very small, and if efficiency were any sort of goal, we'd be running class-B transistors. I don't disagree with much said there, just that I see the balance of factors differently.
>
try an active CCS instead of the choke.
A choke lets the tube tell it how much current to pass.
A CSS tells the tube how much current it may pass.
So tube bias becomes critical. In a triode, the plate voltage change will tend to find a "working" bias current. This may not be the operating current and voltage you would pick, and will generally be lower voltage than you'd get with a choke. In a pentode, you have two high-impedance current sources that have to pass the same current: it won't be easy to make them play together.
Nevertheless, it is often done. Someone sells a board.
Another approach is a gyrator to emulate a many-many-henry choke with low DC loss and eating any DC current it is fed. A gyrator can be made with a unity-gain amplifier, so implementation may be simple. If unidirectional DC current is drawn, it can be a simple cathode/emitter-follower. The nice thing is that it will self-adjust to any current the tube wants to draw (within reason).
In
Rod's Fig 7, R1 is similar to a choke's DC resistance. R2 is similar to shunt resistance across the coil. For high Q (in this case, for high gain without large DC drop) we want small R1 and large R2.
I forget what the hell we wanted a super-choke for. 12AY7 output stage with 30K load? Right away we have to face one problem with a CCS or a gyrator: they don't store or kick-back energy. With a 300V B+, a choke or transformer can swing up to 600V peak-peak, a resistor, CCS, or active gyrator only 300V peak-peak (and in reality much less). For the CSS or gyrator to "replace" the choke with the same max output we need more like a 600V B+ and double the power consumption. Sometimes we down't need max output, but in this case with a 7:1 output ratio we might.
Say the 12AY7 runs at 300V and 10mA. Use a BJT (if you can find one with the ratings). Gm of a BJT running 10mA is like 1/(3Ω), so try a 100Ω emitter resistor to the 12AY7 Plate to swamp BJT nonlinearity. Your load is 30K, so there aint much point in raising shunt resistance above 300K. Tie 300K from Base to +300V. Since this is a large-signal stage, you need to take the Collector up to +500V, +600V or so to cover the swing (no energy stored or kicked).
Using L=R1*R2*C1 {or more exactly" L=C1*R1*(R2-R1)} we need about 3uFd for 100H. This cap has about 1.6V across it, and "should" never go over 2.6V or under 0.6V, so it could be quite cheap. Leakage current will cause a voltage offset (drop) which I think is insignificant for good Japanese caps and these conditions.