Whoa! You don't have a gate-source zener protection diode. Anything from 8V to 15V should work fine -- that'll probably solve your blown FET problem.
Another point, you've paralleled two MOSFTEs, but in reality one will dominate in a linear circuit like this. If you need the extra power capability of a second FET you can use a volt or so of source-degeneration, or you can use the active MOSFET current-divider circuit I posted here last year or so. Or you can use a series connection for the MOSFETs, dividing the extra unused input voltage between them, that's a common approach.
I don't see any loop compensation, so I'd expect the circuit to be unstable, or marginally stable. We show some typical values in our circuit figure 6.47 on page 369 of AoE. There's also a lengthy discussion there about your circuit's approach.
Hi I have been experimenting with some ideas and designs for a HV labsupply and I have come up with a design I would like to have some response on.
Look here:
formatting link
I will off cause have a negative supply combined with this one, and a current limiter - but that will come later.
The circuit is working very nicely and is very stable, the only problem is that the HV mosfets is blowing once in a while if I connect or disconnect anything then power is on and I have a high voltage on the output. I would like very much to have some ideas on how to optimize the circuits to protect from the sparks and similar things that properly is destroying the fets.
What could be the best way to do it? and what kind of capacitors should I use to get the best protection?
Another thing to consider would be a bleeder resistors across each output capacitor. This will help to ensure that the DC voltage across the output capacitors remain balanced. Without these resistors, a small amount of leakage current can cause a large DC imbalance in series connected capacitors.
Switched MOSFETs (with high gate voltages, each device fully turned on) may be paralleled and will share current per each transistor's conductance, I ~ 1/Ron. Since Ron goes up with temperature this is a safe scheme. Modern power MOSFETs are not so often used as linear elements, and some mis-information circulates about this mode of operation. For example, at any reasonable power level, for a fixed gate voltage, the current increases with temperature, so direct paralleling is not safe.
I bought a few of these Xantrex 0-600V 1.7A switchers on eBay.
formatting link
The seller might have a few more of them.
OK, that was an aside. Have you set a reasonable lower voltage at which you'd want a full 0.5A, such as 300V? That's still a high 175W of hot-air wasted heat, if your DC input is 550V. If you want a 0.5A rating down to zero volts (e.g. a short circuit) you'll need to dissipate 350 watts. Ouch! Space-heater-city.
A second little-known issue if that it's not the FET's Ron rating you should be looking at, but the Pmax rating, or more accurately its thermal resistance. Another factoid, you'll find that some of the lowest thermal resistances available are for high-voltage power MOSFETs. Ahem, another eBay opportunity is the STW8NB100, rated at 190W with an impressive thermal resistance of 0.66 C/W,
formatting link
and eBay link:
formatting link
This guy only wants $2.99 for these great MOSFETs. Note their large surface area, which improves heat-sink conduction. Nice!
You can ignore the 1kV rating, it's as if the voltage rating was infinite. Given typical heatsink issues and saftey margins, you might end up dissipating only 40W or so max in each MOSFET, which means you'd want at least five or six as regulators. You could wire them in series, with gate dividers as we show in AoE figure 6.52, page 372. Or, if you want to use them in parallel, you'll have to provide a current-sharing scheme to deal with the positive drain-current temperature coefficient. A simple way would be to use 22-ohm 1/2-watt degenerating source resistors, these will drop 2V at 100mA, and effectively prevent any one FET from taking more than its share of the current. You could use a little servo for each FET to lower the voltage drop, but 2V is not a problem.
This design uses six MOSFETs (only 83mA max) and drops 1.8V max across the source current-sharing resistors.
. HV regulator, paralleled MOSFETs, source R . . HV in STW8NB100 22 500mA max . ----+---- D S ----/\\/\\---+----->
. | G | 175-550mA CL . | | 22 | . +---- D S ----/\\/\\---+ . | G | . | | 22 | . +---- D S ----/\\/\\---+ . | G | . | | 22 | . +---- D S ----/\\/\\---+ . | G | . | | 22 | . +---- D S ----/\\/\\---+ . | G | . | | 22 | . +---- D S -+--/\\/\\---+ . | G | | . | | | \\ 12V | . '-/\\/\\--+-+--|--| You could use a little servo for each FET to lower the voltage
The servo isn't too complicated, here's how Agilent does it.
. HV regulator, paralleled MOSFETs, servo . . HV in 22 500mA max . ----+--------- D S --+--/\\/\\--+----- . | G | | . | | b | 0.2mA HV . | ,--/\\/\\--+---c e--+---|--- current . | +--||----' | | sink . | | 22 | | . +---|----- D S --+--/\\/\\--+ . | | G | | | . | | | b | | . | +--/\\/\\--+---c e--+ | . | +--||----' | | . | | 22 | | . +---|----- D S --+--/\\/\\--+ . | | G | | | . | | | b | | . | +--/\\/\\--+---c e--+ | . | +--||----' | | . : | : : . control
Basically a set of differential-amplifier transistors act to keep each paralleled MOSFET running at the same current as the rest. You design for 4 to 8 volts across the gate resistors to allow for differing FET gate voltages without dramatically changing each BJT's current. This scheme allow you to use low-value current- sensing resistors, e.g., dropping 100 to 200MV at full current.
Which part was that? I don't see it on our drawing.
I usually try to design with no capacitor at all across the HV feedback resistor. If you do have one, isolate the opamp's input with a series resistor and diodes to the supply rails, like we show in figure 3.75 on page 169. If one is used, it should be a small value to make a high-frequency zero in the control loop. Also, this way it won't contain much energy in the event of a sudden output short, creating a high dV/dt on the output. Remember, the cap's current is I = C dV/dt.
I do use zener diodes.
My usual practice is to use more control current (the book's circuit shows about 450uA max). Anyway, adding a degeneration resistor gives you a chance to add a zero in the control loop, which is a good idea. Note: We didn't say 10pF.
Also, I'd not drop more than a few volts across the resistor so I'd say 15k is too high a value to use.
Actually, my practice is not to use a power MOSFET for Q1, because they have so much gate capacitance. I usually use mpsA42 transistors in cascode, keeping the voltage across each one to under 250V. But using a MOSFET is simpler.
Ok, I have added that one now. New pdf is oploaded.
Well, I have read some discussions about this subject and there seems to be a some disagreement on how to parallel MOSFETS. But I would like to but do both things: series and parallel so I can use som less expensive mosfets and stil have a high current out (max 500mA). As for now I have removed one of them, and then I think I will try to do experiment with it...
Yes, I know the problem about the missing loop compension, but I dont know the right way to do it in this circuts. Is it about a resister from the voltage divider to non-invert, and fron non-invert to ground?
(i don't have AoE but I will try to get it as soon as possible!)
I just got a scan of a few pages of AoE (i got to have that book!) and I have places a 10n and 10p similar to that design. New pdf is uploaded.
Comments? I have not tried it out yet.
About the parallel/series FET: if I should try to place them in series, should I place any extra components also, or can they just have the same gat and then D from the first to S on the second?
That was then, this is now. I use zener diodes. One might argue a silicon diode's low capacitance is preferable, but the zener diode's higher capacitance is still swamped by the MOSFET's very high gate capacitance.
Thanx, now I have something to get me a lot further with this project. I have just tried with the 10nF on the opamp from out to invert, and it is working fine. Without it I had a lot of oscillating today on the circuit. But I had to remove to 100nF from out to GND.
I tried with 10nF instead of 10pF, it gave a much more stable output. 100nF blew the opamp :-( what would be right here?
I just have a few questions to the circuit on page 369:
What kind of "tube experimenting"? What voltage and current rating? Since you are interested in tube experimenting, have you considered using one or two 6AS7G's for the pass element. Back in 19959 or 1960 I built myself a regulated 0 to 300 volt, 300mA supply using them. It worked well but I didn't use it much because by 1962 all my designs were solid state. (I gave the supply away in 1963.)
Anyway, I always add a 10 to 15V zener across the MOSFET gate-source, and in the next edition we'll show that. Try it and let me know. One other thing, you don't have a DC POWER switch on the FET's input do you (this would create a very high dV/dt at the circuit's input).
I have a small question I hope you will answer. It is about the HV project I am experimenting with these days. On page 369 in AoE you are using enhancement mosfets right? but the symbol is depletion. I have the problem that I blow the MOSFETs when I put fx. a 60uF cap on the output.
It must have something to do with the reverse voltage. I have the diode on the mosfet, and everything is ok if I dont have any caps on output but only a pure resistorload. I have the hypothesis that it must have something to do with the fact that I am using enhancement MOSFETS so I can't get the reverse voltage away from the output caps when the circut is regulating before the MOSFETS is blowing.
How can I decide the value of those resisters and caps? I tried to put them in parallel and had to put in a 10R gate resistor to avoid ringing, but the current is not very uniform at the mosfets so the best way would absolutely be the series connection.
What is the optimal size of the resistor in series with the output cap? I found out that this resistor is absolutely necessary otherwise I will have a lot of low feq. charge/discharge curves at the output depending on the load. How can that be? in low. voltage PS I have never seen this problem and that resistor..
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.