filtering a 3000V power supply output.

[(*steve*)]

I currently have a power supply capable of 250W output, with the output voltage adjustable from 5V to 1000V (yep, that would be 250mA at 1000V). I'm soon to get a similar power supply capable of 600W and 3000V.

These are deadly (I know that).

They are electrophoresis power supplies, and I picked them up for next to nothing.

A problem they have is that that are not well regulated, varying a bit with load, noisy, and with an output voltage that tends to wander. They have a few other issues, but filtering is the one I'm tackling at the moment.

3000V capacitors are not really an option. However I have plenty of 400V 100uF caps that I will place in series (10 in series).

Balancing the voltage across each of them is obviously important, and using resistors would be very wasteful of power.

My thought is to place 350V transient voltage suppressors (1.5ke350a) in parallel with the caps. At 300V they leak about 1uA, and they clamp at a voltage safe for the caps. Because they are basically a zener, they will also limit three reverse voltage across the caps when they discharge.

The problem is pretty obvious though -- whilst they're rated for 1500W, that's only for something like a millisecond. (Constant dissipation is something like 3W) The current through these devices will depend significantly on the variation between the capacitors, and I don't expect it to be very high.

Do you think it will work? Any alternate suggestions?

I plan on trying this with a lower voltage array first (for around 1kV) before scaling it up to the full 3kV.

10uF isn't a lot of filtering, but I plan to make a linear regulator to follow it.

 

[Harald Kapp]

What about using TVs in series with resistors? The tvs will keep away most of the voltage from the resistors and the resistors will limit the current through the tvs.
Once the string of capacitors is charged and balanced, the current through the tvs/resistors will be rather small as the voltages across the capacitors are then balanced and only the ripple current needs to be taken into account which should be much lower than the initial inrush current.

 

[(*steve*)]

The tvs diodes do not seem to have as sharp a knee as a regular zener diode. So I'm not sure exactly how much current will flow. I might do some tests on some 400V tvs diodes I do have on hand. The datasheets are a little hard to interpret.

The idea of this circuit is not to balance the voltage across the caps per se, but to limit the maximum voltage across each capacitor. This will act to balance the voltage once a capacitor nears the limit, but not before.

 

[hevans1944]

I think transient surge suppressors will not work in your application. It is my understanding that once their threshold voltage is "triggered" they become low-impedance "shorts" to absorb the energy surge that led to the over-voltage condition. How often, and how long, this short-circuit can occur without permanent damage to the component depends on how much energy is dissipated while they are in a conducting state.

In any event, triggering of one device will likely cause an over-voltage condition across other devices in series with it... a domino effect... that will lead to the entire output of the power supply being short-circuited within a few microseconds. It is easy enough to connect three or four in series to test this hypothesis.

Perhaps a better approach would be to find some high-voltage caps on the surplus market. I have several 10 μF, 4000 V, plastic dielectric, capacitors, each packaged in an hermetically sealed metal can with ceramic-insulated terminal posts, just waiting for a project to come along. Unfortunately I don't have enough of 'em to do anything really interesting, and the capacitance isn't really large enough for a vacuum tube RF linear amplifier, which is pretty much obsolete now... still, it might be fun to build one and play with it on the air for awhile.

I need to find a suitable power transformer, i.e., cheap or free before beginning such a project, and also see what I can afford to use for a vacuum tube... This would have to be a "single amateur band" amplifier because I have neither the time nor the patience to design and build a band-switched design. All the latest solid-state linears are "no tune" types covering from 80m through 6m at a cost of a few hundred bux for a few hundred watts. I will try out my still-unused KXPA100 solid-state 100 watt linear first before deciding if more power will deliver my transmitted signal as far as I can hear any replies.

But good luck, Steve, for voltage sharing among your 400 volt series-connected capacitors. I would go with the "tried and true" resistor method first before discarding the idea. Hams have done this since forever with good results. A side benefit: the resistors "bleed" off the charged capacitors when power is removed, not an inconsequential thing if you have ever been "bitten" by a charged capacitor bank. Don't ask me how I personally know about this.

 

[(*steve*)]

The tvs works like a zener diode, I've just run some tests using my 1000V regulated power supply.

I'm testing a 1.5KE400A, and on my multimeter they don't show any leakage up to about 395V, where it rapidly increases to 1mA at 407V.

Something that is really interesting is that they clearly have a significant positive temperature coefficient. Seting the voltage to 410V have me an initial 1.6mA, but that dropped to 0.25mA over about 30 seconds. That's with the device dissipating 0.1W, well within the 3W continuous dissipation. (Actually, it's still falling now down to 0.12mA).

A test at 460V started with the power supply limited to 20mA, but a gradual fall off to 460V and 5.5mA. it gets pretty warm, but doesn't fail. I'm thinking if using the tvs diode to drive the base of a high voltage transistor.

I also have some 55uF 450VAC motor start capacitors, maybe these would be better than the 100uF 400VDC electrolytics? I'll have to check their specs for DC voltage...

 

[(*steve*)]

I've just connected up the 450V caps in parallel with these TVRs, and they seem to work well.

Using 450V 270uF caps, all the in series charge up within about 10V of each other with the power supply delivering around 1080V. Disconnecting the power supply and discharging one cap, then reconnecting the power supply, the caps balance out to 370, 386, and 328 volts (the last was the one discharged). Decreasing the voltage, until the caps are 32, 53, and 3.7 v, and then further... The voltage on the caps falls to a minimum of -1.1V, which is almost certainly safe for these caps. Seems to work well.

 

[(*steve*)]

Here's the 1000V test using the electrophoresis power supply.

The whole setup (note the special high voltage cabling!)


And closeup of the caps


I'm still seeing a lot of ripple...



That's 500mV per division, but with a 10x probe, so anything up to 20V of ripple.

Hmmmm...