High-voltage bench power supply

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I prefer whole range adjustment, since all my other PSUs cannot supply over 6-70VDC.
Smaller range is easier and can be done more efficiency with SS. IMHO with whole range like 0-450V tube based circuits would be safer although efficiency is a big trouble. Also, old-school tube based PSUs like UIP-1 or Fluke 407 naturally come with auxiliary bias supply as an added value.
 
I've been working on a lab supply, based on Pete's design and made a few mods. Attached is a spice file showing the complete supply.

First thing I noticed when you simulate Pete's regulator was that if you use a linear potentiometer in the voltage adjustment branch, the output voltage vs. rotation wasn't linear. I changed it to an adjustable current sink and fixed resistor. This linearizes the voltage adjustment nicely, and also increases the loop gain in the regulator at low voltage settings. The current sink uses a voltage reference chip, potential divider, and op amp/transistor voltage to current converter.

I simulated Pete's cheap and dirty current limiter and found it a bit imprecise, as expected. I devised a closed loop current limiter that works roughly as follows: output current flows through a 0.5 ohm current sensing resistor. An op amp amplifies the voltage across the resistor and incidentally is used to drive a current meter. This voltage is converted to a current and sent to an optoisolator, whose output is in series with the cathode of a triode connected in parallel with the pentode used as the error amplifier. When the output current exceeds a certain amount, the optoisolator/triode shunts the control voltage to the negative rail, reducing the output voltage and limiting the output current. The current limit threshold is adjustable with a potentiometer between the current sense amplifier and the optoisolator.

Minor changes: I'm considering using a 6AW8A Pentode/Triode in place of the 12BY7A pentode as the error amplifier, and I'm using 6550s instead of 807s as the pass tubes.

I've changed the voltage reference gas regulator tubes to temperature compensated zener diode strings, just because I don't have any regulator tubes. Most of the other differences in our circuits have to do with the power transformers I have available in my junk box, so you can neglect those differences.

I'm in the middle of building it up and will report back later once I have test data to compare with the simulation.

PS. has anyone noticed that the error amplifier tubes used in Pete's, Heathkit's and even HP supplies are used in a way that the maximum plate voltage spec is exceeded? Is this a concern, or can I assume that the data sheet design maximum specification are conservative?
 

Attachments

  • Lab Power Supply.asc
    37.8 KB · Views: 79
4 independent outputs with different ranges. One isolation transformer with 2 secondaries, 2 doubling rectifiers. One Zener string with potentiometers across Zeners, 4 MOSFETs. One high voltage BJT that shunts Zeners when current goes out of range (3 separate ranges). White backlight goes red when the BJT that shuntes Zeners start conducting.
 

Attachments

  • IMG_0916.jpg
    IMG_0916.jpg
    712.5 KB · Views: 307
  • IMG_0917.jpg
    IMG_0917.jpg
    507.8 KB · Views: 294
  • IMG_0918.jpg
    IMG_0918.jpg
    631.5 KB · Views: 287
Anatoliy, your post got me thinking... about current limiting by adjusting the reference to the servo versus brute force shunting the control signal inside the feedback loop, so I'm redesigning the current limiting function in that vein, and when I'm done I will post back. I realize that I'm just creating another loop, but hopefully this loop will be more amenable to analysis.

Glad I have not gone too far in construction yet. Posted via my tablet so please excuse the spelling and punctuation, such a pain typing on a touch screen. Later.
 
I'm sensing current at the positive output, but I could shift that down to "ground" if necessary, but seeing that my reference is sitting at -100 volts negative I would still need some sort of level shifting apparatus...right now I'm using a pair of optocouplers arranged such that the transfer function is more or less linear, so I don' t see the need to modify my setup at present, but that might change later on.

I love ltspice, because you can breadboard to your heart's content and not burn anything up in the process!
 
I'm sensing current at the positive output

Here is the LTspice file for a HV power supply that I started designing 5 or 6 years ago. It has a novel current sense / translation circuit.

Life, loss of job, and two moves totaling 1200 miles prevented it from ever being built. I will get back to it some day, but for now it only exists in the mind of the computer. Copy and proceed at your own risk......parts blow up in violent and nasty manners at these power levels!!!!!

I have hit the limit on the 1 KW HP supply seen in post #15, which was about the same as the limit in the electrical service in the lab in my old house. My new lab (under construction) has no practical limit on power (240 volt 200 amp breaker box next to the bench), so I am looking at building a supply for 0 to 750 volts and 0 to maybe 3 amps. The circuit shown will be my starting design, unless I come up with a better idea before I start building. I have already collected most of the parts.

The raw DC supply (V2 in the spice file) is made of several 160 volt supplies connected in series. Each is made from a 650 VA isolation transformer with a FWB and big cap on its secondary. There are 6 such supplies wired in series for -160V, 0, +160V, +320V,+480V, +640V, and +800V.

I have several "medical grade" toroidal transformers with enough primary to secondary isolation to allow stacking to these voltage levels. Some cheap stuff may break down when used in this manner. Another alternative would be a 480 volt to 120 volt industrial control transformer wired backwards. These are cheap on ebay, and will provide a reliable 600 volt DC source.

There will be several separate regulated outputs:

One (maybe two) negative regulators fed by the -160V source to provide negative 0 to 150 volts or so. The regulator has not been designed yet. The "bias" supplies in most tube based power supplies is not designed to support any current draw, and therefore useless for mosfet grid drivers.

There will be at least two positive regulators with electronic means to switch their inputs to any of the positive raw voltage sources depending on the output voltage desired. The positive regulator is shown in the LTspice file.

U2, M6 and M2 comprise the high side current sensor / translator. the current through R13 is translated to a ground based voltage across R16. I believe it was 1 volt per amp, but I don't remember for sure. The level translator came from a app note, LT, or maybe TI.

M8 and M9 form the current error amp / comparator The voltage across R16 is compared to the voltage from source V2 and if the current through R13 creates a voltage across R16 that is above V2, bias from I1 is stolen by M4 reducing the output voltage to make the currents equal.

M5 and M7 form a similar error amp / comparator for the output voltage. R1 and R2 divide the output voltage by 1000 and it is compared to V3.

D4 and D5 decide which comparator will steal the pass device's bias, allowing constant voltage or constant current operation, and M4 translates the "steal" current to the raw B+ voltage level.

The voltage sources for the current (V5) and voltage (V3) control could be generated by a pot off the 15 volt supply. I intend to use DAC outputs. A 12 bit DAC will provide 1/2 volt steps which is plenty good enough for a HV power supply.

The pass device will in reality be several devices in parallel, each with their own source resistor and zener diode. The diode (D1) / resistor (R3) combination should be chosen such that the maximum short circuit current does not blow the fet in case of a dead short on the output and a brief moment of instability in the current limit circuit. A 10 volt zener and a 1 ohm resistor with a Fuji 2SK3675 mosfet shown should self limit to about 2 amps which is far to much dissipation for a single fet, even the Fuji 200 watt 40 amp (peak) fets. Fuji does not provide curved below 5 volts Vgs, so some testing will be needed.

A thermal sensor, and speed controlled fan will be needed

R8 and all the other parts downstream of Vout were used for transient testing and are not part of the design.
 

Attachments

  • PS_5a.asc
    7.6 KB · Views: 80
I have several "medical grade" toroidal transformers with enough primary to secondary isolation to allow stacking to these voltage levels. Some cheap stuff may break down when used in this manner. Another alternative would be a 480 volt to 120 volt industrial control transformer wired backwards. These are cheap on ebay, and will provide a reliable 600 volt DC source.

I want something like this to pop up in my area cheap. It has all sorts of primaries that when flipped around would give you multiple secondary voltages to choose from. 5kVA should be sufficient, I wonder how much it will go for.

http://www.ebay.com/itm/NEW-PBF-Gro...259493?hash=item2ef0ab6325:g:kagAAOSwM4xXZLh0
 
5kVA should be sufficient

I guess your dreams of building the "big one" are bigger than mine. As I get older I have learned not to collect too many things that weigh nearly as much as I do. I have some rather steep and narrow basement stairs to contend with. I passed on some rather nice equipment racks with glass doors because I couldn't possibly get them down here, and they didn't come apart.

I had a similar transformer, in 1KW size that I ripped out of a Japanese industrial printer in Florida. It had taps on one winding from 100 volts to 480 volts. The other winding had taps for 100 and 120 volts. I think it weighed about 40 pounds. I gave it and a bunch of similar transformers away when I had to move.

The common industrial control transformer has two 120 volt windings on one side, and two 240 volt windings on the other side. They come in sizes from 50 VA up. I used a 500 VA model in my SE 845 amp, feeding a voltage doubler made with 5AR4's to make 1050 volts at 200 mA. They often go unbid on Ebay so a minimum bid will take them. Just pick the one with the lowest shipping cost. An honest and motivated seller will stick one in a USPS flat rate box. A 750 VA model can fit into a medium box with some padding. This may really irritate your letter carrier!
 
PS. has anyone noticed that the error amplifier tubes used in Pete's, Heathkit's and even HP supplies are used in a way that the maximum plate voltage spec is exceeded? Is this a concern, or can I assume that the data sheet design maximum specification are conservative?

Yup, that was intentional. The 12BY7A has a "design center" plate voltage rating of 300V. Usually, one can assume that it could operate to 2x that, at least in a transient condition. (Consider driving an output transformer at B+=300V).

If you look at the internal construction, it appears to have plenty of "room". So I figured it was a low risk to run at ~400-450V.

Pete
 
Thanks, Pete for the info. I guess these tubes were conservatively rated to allow for higher voltages as long as screen and plate ratings weren't exceeded beyond a short duration. I'll make sure that the plate connection is doubly insulated to preclude flash-over at the socket pins, but beyond that I'll just hold my breath and pray that the tube can stand the stress. Since the external plate resistances are fairly high, at the very least I know that any fault currents, should they occur, will be limited to milliampere levels if any arcing occurs.

I'm just about finished my simulations of a low side current limiter, and cleaning up the spice file for perusal, so soon, very soon. Nice clean current limit, infinitely adjustable and fairly stiff. Just need to simulate some extremes of the operating envelope to make sure nothing goes boom.
 
A more simple and elegant way for current limiting (or rather current limit HV cut-off) is the one shown in Bruce Rozenblit's bench power supply. The idea is very simple: a resistor in series with the HV supply line before the pass element senses current and triggers the LED of an optocoupler, which, in turn, triggers the coil of a relay which interrupts the mains supply of the B+ transformer. This, of course, assumes that there is a single transformer dedicated solely for supplying the B+ line. With several resistors mounted on a rotary switch one can choose different levels of current limit.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.