How? Variable high-voltage PSU

[matseng]

When I saw another thread there about a high voltage psu I realized that a HV PSU is something missing in my lab.

Can anyone give me some ideas and pointers in the right direction for designing a PSU with variable output 100 up to 1000 volts AC 50/60 Hz with, say, 100 watts peak power or so?

I guess it would end up involve re-winding the secondary of a decently large toroid transformer to withstand 1KV and having a centertapped primary on it connected to two banks of PWM driven power fets. All of that powered by a reasonably high (20-40 volts?) supply.

Or am I totally off here?

[Rerouter]

Audio amp with step-up transformer?

[richard.cs]

I'm in the middle of making a 0-500V, 0-100 mA (200mA below 250V) bench supply, regulated output voltage/current. That's a fairly complex beast.

For what you want perhaps just a variac followed by a step-up transformer and do it all at 50 Hz, no power fets required. If you are in a 120V country you could use a 240V-primary microwave oven transformer with the shunts knocked out - it'll run nice and cool on 0-150V from the variac and give you around 0-1500V ac output.

[Psi]

Audio amp with step-up transformer?

+1 for this

Rewind the primary on a microwave oven transformer so it can take voltage from a 100W car amp. Then feed that with a adjustable amplitude 50hz sinewave.

Note: high power 1000V isn't the safest thing to have around your lab, put a keyswitch on it!  (not on the output  )

[Someone]

Beyond 300-400 Volts it all gets "unusual" to say the least, check out the old HP 65XX series of high voltage supplies (either for inspiration or just buy one). I wouldn't recommend designing anything over 500V just for fun or because you think you might need it.

[retrolefty]

In the good old days ham radio vacuum tube days used a lot of 120/800vac 'plate transformers' for a typical 100-200 watt output. They may still be available surplus/used. Of course that leaves an adjustable regulator that would have to be fairly efficient over such a large output range.

[dom0]

For a DC supply it really isn't any different from any other DC lab supply, you just need some high-voltage high-power pass transistor and have a voltage sense divider that doesn't have issues with 1 kV (i.e. 4+ resistors in series).

[richard.cs]

For a DC supply it really isn't any different from any other DC lab supply, you just need some high-voltage high-power pass transistor and have a voltage sense divider that doesn't have issues with 1 kV (i.e. 4+ resistors in series).

In principle yes but as someone who is partway through building one I can say that suitable finding pass elements is tricky. Both bipolars and modern switching mosfets have severe second breakdown limits at these voltages, even huge transistors may only be good for a few tens of milliamps at dc. I spent a lot of time toying with series stacked pass transistors and in the end gave up. I've ended up with a valve as the pass element  in an otherwise solid state design.

[T3sl4co1l]

No big deal, just wire it up like any other.  You'll have to use resistors instead of current sources as pull-ups, and parallel a few "linear" type MOSFETs for the pass devices (or pick up a few power tubes and sockets and heater transformers, but that's a lot of bother to go to, unless you already have them on hand).  The feedback divider, voltage reference and error amplifier still work the same, the error amp's output is merely scaled to a much higher voltage range.

You could also make a switching supply that's primary side controlled, so you don't need to worry about dropping high voltages directly.  This one does that, on a somewhat higher voltage range and less power (< 20W):
http://seventransistorlabs.com/Images/HVPower1.png
http://seventransistorlabs.com/Images/HVPower2.jpg

Tim

[dom0]

For a DC supply it really isn't any different from any other DC lab supply, you just need some high-voltage high-power pass transistor and have a voltage sense divider that doesn't have issues with 1 kV (i.e. 4+ resistors in series).

In principle yes but as someone who is partway through building one I can say that suitable finding pass elements is tricky. Both bipolars and modern switching mosfets have severe second breakdown limits at these voltages, even huge transistors may only be good for a few tens of milliamps at dc. I spent a lot of time toying with series stacked pass transistors and in the end gave up. I've ended up with a valve as the pass element  in an otherwise solid state design.

I'd suspect that if you go the stacked transistors route in a HV design you'll need additional clamping for every CE (e.g. Zener). But yeah, high voltage, linear and power doesn't really mix well with anything.

[calexanian]

I would use a Variac. there are plenty of transformers out there for use in tube circuits that will give you an appropriate voltage range. Series up and balance some filter caps and you are in business.

[dom0]

For a DC supply it really isn't any different from any other DC lab supply, you just need some high-voltage high-power pass transistor and have a voltage sense divider that doesn't have issues with 1 kV (i.e. 4+ resistors in series).

In principle yes but as someone who is partway through building one I can say that suitable finding pass elements is tricky. Both bipolars and modern switching mosfets have severe second breakdown limits at these voltages, even huge transistors may only be good for a few tens of milliamps at dc. I spent a lot of time toying with series stacked pass transistors and in the end gave up. I've ended up with a valve as the pass element  in an otherwise solid state design.

I'd suspect that if you go the stacked transistors route in a HV design you'll need additional clamping for every CE (e.g. Zener). But yeah, high voltage, linear and power doesn't really mix well with anything.

Hmm, interesting, there are some IGBTs that are qualified for linear (DC) operation. For example, IKW 25N120 can take 300 mA at 1.2 kV (of course only at Tc = 25 °C, which is, as usual, kinda hard to maintain :-)

[bktemp]

Hmm, interesting, there are some IGBTs that are qualified for linear (DC) operation. For example, IKW 25N120 can take 300 mA at 1.2 kV (of course only at Tc = 25 °C, which is, as usual, kinda hard to maintain :-)

Almost all Infineon IGBTs have DC specs in the datasheets.
I have spent many hours searching for information about IGBT linear operation, because IGBTs are cheap and readily available for >1kV working voltages. There are only a few appnotes going into detail. Most manufacturers (like IR) simply say IGBTs are not the intented operation mode, therefore they do not test it.
There is one interessting appnote with actual results of IGBTs tested in linear operation. They show a derating curve for higher voltages similar to the second breakdown derating of bipolar transistors. The DC curves in Infineon datasheets are a straight line giving 350W independent of the voltage. This looks wrong. I doubt it can handle anything near 300W at 1.2kV even at cold temperatures.

Edit: Found the appnote from Microsemi: How to Make Linear Mode Work
http://www.digikey.com/Web%20Export/Supplier%20Content/microsemi_278/pdf/microsemi-power-an-make-linear-mode-work.pdf?redirected=1

A 600V, 283A, 682W IGBT fails at only 114W at 500V.

[T3sl4co1l]

IGBTs have higher current density than BJTs.  They are absolutely optimized for switching.  They're already near ratings in saturation at rated current.  Drawing even a small fraction of that current, at a high voltage drop, results in extreme heating.  Which causes spot heating, which enhances the hFE of the PNP structure, and reduces the Vgs(th) of the FET structure, making it doubly worse than spot heating of BJT or MOSFET structures alone.

MOSFETs exhibit 2nd breakdown as well, because they have been optimized to the point where power density is into that critical range.

As far as I know, so-called linear MOSFETs are simply constructed with lower cell density (and probably some source resistance or something), so the die area and Rds(on) is higher (and cost, too) given the ratings, but the power density is below the critical point so their SOA is useful.

Tim