So, 6-10mA or so? A bit much for a CRT flyback, but something in that range, certainly. ~mV filtering will need a lot of cleanup, including consideration of common mode coupling (for example, the input likely needs to be filtered as well). You may consider a linear pass reg, which is feasible with MOSFETs these days, or there's the old (also a Jim Williams special) trick using a vacuum tube (Fig. 11A):
https://www.analog.com/media/en/technical-documentation/application-notes/an2f.pdfNote that LC filters aren't good for much more than cleaning up HF trash; design filters for this band, more-or-less at impedances near free space (ballpark 100 ohms?), since that's basically what matters up there (i.e., filtering against radiated or conducted emissions). Some damping will be needed, so for example, a CLC(LC..) filter with a shunt R+C at the end is a good way to do it.
For bulk filtering (switching ripple, subharmonics / modulator noise, something or other), RC filtering tends to be the better way to go: there's no hope of matching a filter to RL (almost a megohm!), you can't get inductors that big and even if you could they'd have hopelessly massive stray/self capacitance. The R gives more impedance, at almost all frequencies, so is the better deal. It also serves to limit short-circuit current, improving lifetime of transformer, rectifiers, capacitors, etc. (Use high voltage, pulse rated resistors.)
I am not very good at web forums, so please let me know if this needs a separate (new) post.
I am interested in building an HV supply too. Specifically something that doesn't use a flyback. I have miles and miles of magnet wire down to #32 AWG and I am wondering how to wind my own transformer. I understand the basic idea where if I have 1000 turns of primary and 100,000 turns of secondary, I should get 100 times the voltage.
What I do not know how to do is build the transformer for a specific resonant frequency. How is this done? I know a standard power transformer has a resonant frequency between 50 and 60 Hz (or does it?). How would I make one that has about 1:1000 ratio and operates at >100 KHz? Is there a commercial off-the-shelf option for this? Something with some real power where I could get more than a few mA output at say 5KV?
Eh, close enough; a transformer topic would be closer I guess?
The turns ratio is set by the required voltage ratio; the number of turns, alone (or in terms of volts/turn, say), is set by driven frequency, voltage, and core size:
N = V / (4(.44..) F Bmax Ae)
N = number of turns
V = applied voltage
F = applied frequency
Bmax = peak flux density (depends on core material and acceptable losses at frequency)
Ae = effective area (cross section of the core)
Typical values for ferrite are F in the 10kHz to 1MHz+ range, Bmax under 300 or 400mT (ferrites can't handle much more than this, in general, with lower values being typical at higher frequencies, like say 100mT at 500kHz, etc.), and a typical EE33 core has Ae ~ 100 mm^2. (That being a core of two 'E' shaped pieces facing, 33mm being the long dimension, the 'spine' of the 'E', with the other dimensions being typical for the core set, whatever that is -- see the datasheet.)
The constant is 4 for square waves, or approx. 4.44 for sine waves (V in RMS).
So like, the uh lemme see here, I made this thing a while ago,
It's an adjustable regulated supply, though because of how the oscillator works, it doesn't go all the way down to zero, and tends to idle high at light loads. It's also not very efficient, it uses a linear pre-reg to control it, but it's fine, only like a 10W capacity anyway.
So the oscillator runs at, oh I didn't write it down did I, dang, want to say it was around 60kHz? And it's getting around 12V (sine, RMS; though the exact figure may be higher, because of how the push-pull circuit works), into 6 turns (each side, alternately), and Ae is about 100 mm^2 for this (EE33 core set), so it would seem,
Bmax = (12V) / (4.44 (6t) (60kHz) (100 mm^2)) = 75mT
which is well under the 0.4T this would saturate at, and also quite comfy at this low frequency (I never noticed the transformer getting warm). Evidently I could reduce the primary turns count a bit (2 turns might be pushing it, but 3 is certainly fine) and get even more voltage from this... heh, well, if I had varnished the thing maybe, as-is it's just loose layers of tape so I don't trust it to very much voltage.
As for resonant frequency, that's a complicated matter, for example: in this very circuit, operating frequency varies widely with load condition, typically being around 60kHz (or whatever it is), but jumping up to something like 200kHz when heavily loaded (or shorted). The reason is, the primary has inductance against the secondary: leakage inductance. This acts in series, between the two windings. Leakage acts to limit the secondary current, or worsen output regulation (when primary voltage is fixed), or reduce signal bandwidth, etc. Magnetizing inductance acts in parallel (as the winding is drawn, as you likely expect it to), and dominates when the output is unloaded. Hence why the frequency changes with load: the oscillator sees a different inductance.
Leakage doesn't have to act to limit current: as it's an inductance, it can be resonated out just like anything else. You can even make a dual resonant transformer, where both primary and secondary are tuned; this isn't a good configuration (the E core with stacked layers of windings) to do that*, but there are others that are better. (CCFL (old LCD display backlights?) transformers are a good example, having a long aspect with many sections of secondary winding, and one thick primary at the end. They also typically use the "ZVS" oscillator (called "Royer" by others above, but it's actually properly a different name, but anyway..) design, like I did!
*Complicated reasons (namely, the coupling factor is too high), and kinda out of scope to explain why, but there is a reason.
With enough tuning, you basically end up with a Tesla coil, in a compact package (with a ferrite core stuck in it). I don't know that I'd recommend this, but it can be done, anyway, and helps when you need really high ratios.
As for the usual kind of resonance, where the secondary magnetizing inductance resonates with its self-capacitance, and ignoring how to drive it for a moment -- that's roughly the 1/4 wave frequency of the wire itself. It actually doesn't matter much that it's balled up into a winding, the wire length still holds great significance. Now, the core being there, the frequency is lowered proportionally, so it's not actually 1/4 wave, but some factor less. So, bigger transformers, with more turns, always have a lower self-resonant frequency (SRF).
SRF is most important for switching converters: SRF must be a fair factor about Fsw, so that neither the switching itself, nor any significant harmonics, excite the resonance (thus drawing excess reactive power, increasing losses in general, or causing more switching losses, etc.). When this isn't feasible, resonant converters are the way to go -- this way it can be driven at/near resonance, and if anything there's most likely gain rather than loss as a result!
Tim