The schematics of HV supplies I've seen in the literature use diode strings with each diode loaded by a cap and/or a resistor to keep the reverse-bias equal. Without that one diode will "steal" all the voltage, die short, followed by the others in the string.
If it's an experiment, you might - no kidding - consider using an old color TV rectifier tube, a 1B3 sort of thing. You can run the filament off a C-cell battery. You can even adjust the filament voltage to tweak the HV.
Or, as suggested, rip most of the HV section out of an old teevee.
ebay has surplus HV supplies, out of copiers and such, but probably not 20KV.
John Larkin Highland Technology, Inc
picosecond timing laser drivers and controllers
Modern HV rectifiers are designed to partially avalanche when the reverse voltage begins to exceed their rated PIV. A group of identical HV controlled avalanche rectifiers connected in series will effectively "self-equalize" themselves without the need for external resistors or capacitors. Internally, your existing microwave oven rectifiers actually consist of a number of lower-voltage controlled avalanche rectifiers connected in series. Connecting two (or more) of these rectifiers in series can be used to handle higher voltages.
However, you may need to add a small (500 pF 30 kV ceramic "doorknob" capacitor across your two-plate load to reduce HF ripple.
On Sat, 06 Jun 2015 22:00:45 +1000, firstname.lastname@example.org Gave us:
An HV diode IS a stack of P/N junctions in itself. Really high voltage examples are upwards of 60 elements seriesed together in one package, and that is why the conduction voltage threshold is so high.
Here is one example:
Note the 16kV version takes 35 volts to begin conduction. There are literally about 20 or 30 or even up to 60 elements in each one package.
You could use a silica powder filled epoxy to act as a thermally conductive bond between the diode and a plate,. yet remain insulated and try to insure that they both keep the same temperature.
We always used Philips, but they stopped making them, IIRC.
The web search appears to show a few companies offering replacements.
If you want to put a diode at 20kV, you need a 30 or 40kV diode as well, BTW.
At those voltages, you want the soldered nodes to be round balls as well. No sharp protrusions like the lead at the end of a fillet type solder joint. "big blobby balls" is best. Absolutely NOT inspectable by a picky QA person.
"philips HV diodes" was the simple search criterion I gave.
You are an absolute idiot yourself.. If you can't read between the lines of what someone is attempting then maybe you should shut your trap. Most people in their position may simply assume they need raw HV of desired level to start with and want to use some elementary method of rectifying it. Does it really hurt to suggest a better way peckerhead?
You must be some throwback from the neanderthal days.
Resistors. That's how I always did it, a lot. Decades ago and the highest voltage diode I could afford as a student was 1kV so that sometimes meant a lot of soldering.
Nowadays one can easily buy diodes for several kV, like these:
The GP02-40 can be had in singles for 50c or less. Trr looks somewhat ok for a 20kHz job. Depending on how square the square wave really is.
If there is a substantial risk of spikes coming along there may also be a need for a small capacitor across each diode but I've not needed that very often.
It depends on the max spread in reverse current and thus also on the max tempearure they are exposed to. Better to err on the conservative side rather than risk a white-knuckle ride followed by tsssk ... *KAPOW*
Horizontal flyback diodes like Tauno suggested are also an option if t's a low power application. However, due to the advent of flat-screen TVs a long time ago they have become nearly extinct and may no longer be suitable for new designs. For hobby they can be scrapped out of an old set or maybe Ebay.