I played with that sort of, a few years ago. Specifically with 6HV5, also comparing with 6LB6.
Standard setup, bias tees for input and output, fast pulse in, measure the output. So the output will be negative-going, and also represents plate current since the load is largely resistive.
Input comes from a BJT avalanche pulser, which is much faster than the tube (<1ns), and more than adequate to deliver required grid drive (~100Vpp for the 2N3904, up to almost 300V with a higher voltage type).
Setup:
Power supplies on the left (I think either screen (for 6LB6) and/or grid (C-) on left, plate in middle), tube in the cage to keep the measurement clear of induced fields.
And here's the, best result I think it was, for 6HV5. Note grid voltage is 10x and plate voltage is -100x. So that's +60Vpk grid, enough to get the plate going -- these "beam triode" pulse regulator types are basically sweeps minus the screen grid, so have very low perveance (not even class B at zero bias), but very high gm and mu. 2A peak is pretty impressive, but clearly it's lacking in speed, and clearly it's only a tiny fraction of the available 2kV supply.
I just wanted to see if these tubes are any better in pulsed operation -- they're already very awful for any other kind of operation, so it's a low bar; but alas, they aren't. I forget what grid current was under this condition; "hFE" in the 2-5 range maybe?
6LB6 did better (unsurprisingly),
This is x100 on the plate again, but with a 500 ohm load resistor (450 in series to 50 ohm output, into a 10x attenuator). The grid waveform is double-peaked by a PFN (pulse forming network), and as you can see, much of the plate current is spent slewing its capacitance rather than actually delivering output. So, it's not going to cut it for sub-10ns range waveforms.
This was in part motivated by seeing what else could be used to generate EFT waveforms (~5ns risetime, 50ns exponential decay), other than a hydrogen thyratron or SiC MOSFETs (Si should also be
just suitable, but needing more parts). My survey found something like:
- BJTs could be used (avalanche mode), but you need about a hundred acting perfectly in unison, with a tonne of 0° power combiners to merge them all together;
- Si MOS, you need one or two dozen, something like that; maybe affordable, competitive anyway vs. SiC;
- SiC, you can basically make a stack of, 6 or 9, I forget how many exactly it was -- and drive the output directly;
- Hydrogen thyratron, you can go direct drive of course, and to quite high voltages (or, preferably with a transformer at the highest voltages), but, well, you need one of these fairly rare components;
- Or a spark gap, but it's not a very controllable and consistent device so I wouldn't recommend it.
As you can see, even with best case grid drive, these tubes just aren't going to cut it. Maybe a vacuum modulator (as found in RADAR equipment -- often alongside a hydrogen thyratron, incidentally) but you'll need much more drive (typically rated closer to 1kVpp grid drive), besides the screen and heater power of course.
Better suited vacuum tubes are the radial-beam and planar types, usually of metal-ceramic construction, which have been used in physics apparatus -- pulse amplifiers for beam kickers and such (where a pulse of some kV is required, at just the right couple of nanoseconds, to electrostatically deflect a bunch of particles in a beam, say to transfer between cyclo/synchrotron loops). In addition to high current density and reasonable impedances (still higher than you'd like for transmission line connection), they're quite reasonable in distributed amplifiers (in which case, the parallel connection can match to 50 ohms reasonably well for say 4+ in parallel).
I hope you are not planning on making a DC-DC converter with tubes.
Really not a big deal; a typical sweep tube is comparable to (but, not
better than, these days!*) a 4kV 1A depletion-mode MOSFET with low gm. They were indeed quite capable of switching, as was the intended application -- horizontal deflection in television service. "Rak(on)" of 30-100 ohms is typical; and of the companion damper diodes, a voltage drop of 10-30V (incremental resistance 20-50 ohms) was typical. They're actually quite capable devices: a plate efficiency over 80% is achievable. Mind, an overall efficiency of more than 60% is quite challenging due to heater power and control circuitry!
Empirical evidence:
https://www.seventransistorlabs.com/tmoranwms/Elec_Compound2.html (now there's an old page of mine, hah)
*As of a decade or two ago, it used to be, sweeps were unsurpassed, at least among commercial offerings, in the very narrow class of high voltage parts. For a while there, I believe there was a 2500V 150 ohm MOSFET by IXYS, which clearly doesn't outperform when compared purely with plate characteristics of a 6LB6, 6LQ6, etc. (Obviously you'd still prefer it for ease of gate drive, size, and lack of heater power, but, practical matters aside, eh?
) 4kV+ parts have since arrived, not to mention SiC (of which, HV parts aren't widely available, but do apparently exist, including MOSFETs and IGBTs up to 10kV or maybe a bit more; I think they're commercially restricted, partly by lack of need or maybe safety, but also ITAR I think??!).
Anyway, as for the likes of 6BK4C -- I think you will pretty clearly see the problem: a pittance of current. I doubt you can get 100mA peak through their tiny cathodes, and I don't know how much of that would even reach the plate in that case -- hFE < 1 may be the limiting factor. Meanwhile, the plate is this huge hunk of metal, you simply can't avoid the stray capacitance. The effective bandwidth will be quite low I suspect.
Oh, peaking coils! The 6LB6 test above had a peaking coil; since only* series peaking is possible into a resistive load such as this, only a modest (20%?) bandwidth extension is possible. This goes up particularly slowly with order, so, you have almost nothing to gain by going to 2nd or 3rd order peaking as well. So, as dearly needed as peaking would be for a 6BK4, it won't help nearly as much as you want.
*Citation needed
No please, I need a citation. I would love to know if there is a better network topology!
Tim