Author Topic: High Voltage Bench Power Supply Design  (Read 15837 times)

0 Members and 1 Guest are viewing this topic.

Offline duak

  • Super Contributor
  • ***
  • Posts: 1048
  • Country: ca
Re: High Voltage Bench Power Supply Design
« Reply #75 on: July 14, 2019, 05:11:27 pm »
Thx kindly for the schematic.

You can likely improve the ripple rejection by adding a 1N4007 and another RC filter stage in series with R6.  This will provide a cleaner and slightly higher DC voltage to the driver stage.

For the current limit, how about using a 22R resistor and a pot connected as a voltage divider it to vary the voltage applied to Q5?  If the voltage drop across the 22R resistor is too high, then how about two ranges? eg. 500 mA and 50 mA.  If you have a built in meter for current, it could switch the meter's range as well.
 

Offline H713Topic starter

  • Regular Contributor
  • *
  • Posts: 180
  • Country: us
Re: High Voltage Bench Power Supply Design
« Reply #76 on: July 15, 2019, 05:18:05 am »

You can likely improve the ripple rejection by adding a 1N4007 and another RC filter stage in series with R6.  This will provide a cleaner and slightly higher DC voltage to the driver stage.


That's kind of what I was thinking, the only thing I don't like about it is that it does require another capacitor on the board that has to deal with the full DC bus voltage. To those unaware, once you exceed 450V, electrolytic capacitor prices skyrocket and availability plummets. This means capacitors in series, with equalizing resistors, and this quickly starts to eat up board space. I managed to squeeze it in though, since the improvement was dramatic. I tested with just a 25uF filter capacitor on the DC bus (just a 660V oil-filled cap), with a 150mA load, and adding the cap right after D6 was about the same as adding 900uF of capacitance on the DC bus. Schematic is attached. I haven't had time to play with the new current limit idea, but if it behaves well then it too may be worth the complexity. IMO, adjustable current limit on a HV bench supply really doesn't need to be that accurate, but Rheostats are not the easiest thing to come by, and while I probably have something in stock, most people don't and I like to minimize the usage of needlessly expensive parts when I can.

Speaking of expensive parts:The OPA604 is an rather pricey op-amp targeted at the audio market, and at $3.30, ridiculous for a power supply. The circuit works just as well with a TL071, the only reason I used an OPA604 is that 1) I have lots of them, and 2), they can handle very high supply rail voltages- convenient for testing. Most generic FET-input op-amps should work fine.
 

Offline ArthurDent

  • Super Contributor
  • ***
  • Posts: 1193
  • Country: us
Re: High Voltage Bench Power Supply Design
« Reply #77 on: July 16, 2019, 10:34:23 pm »
I have a module from an industrial electronic load that is rated 600 VDC and fused for 20A. This shouldn’t be that much different than a pass module of a high voltage power supply and might give you some ideas on MOSFETs to use. There are 8 identical sections in parallel on this module, each using two 900 volt N-channel MOSFETs in TO-247AC high voltage insulated packages. Each of these sections uses a 0.82 ohm balancing resistor in series with the two MOSFETs, which would have to handle about 2.5 amps. Here are a couple of photos. The overall view only shows one side of the heatsink assembly so there is another identical half with more MOSFETs on the other side. The entire electronic load had several of these modules for very high current and heat was removed by three fast 5” square fans that sounded like a cyclone.

You had mentioned the IRFPF50 earlier and while the specs don’t say specifically it is for linear use, that is what this electronic load module uses.
 

Offline duak

  • Super Contributor
  • ***
  • Posts: 1048
  • Country: ca
Re: High Voltage Bench Power Supply Design
« Reply #78 on: July 17, 2019, 04:52:03 pm »
AD,  I see in the first picture that one device is an IRFPC40 and the other is an IRFPC50.  Are they in series or in parallel?   Would you have any schematic diagrams or other info to give us an idea about the design and maybe a way to better use power FETs in linear mode?

As H713 is showing with connecting the FETs in series, it looks like using a high voltage device at lower voltages and higher currents puts the device in a more reliable combination in the SOA.  It makes sense as electro-thermal instability occurs with high drain to source voltages.
 

Offline ArthurDent

  • Super Contributor
  • ***
  • Posts: 1193
  • Country: us
Re: High Voltage Bench Power Supply Design
« Reply #79 on: July 17, 2019, 10:15:38 pm »
Like most unique industrial equipment there is no schematic and I haven't tried to totally figure it out because it was designed to work and I'll believe it does and I'm too lazy to go any further. Each of the eight sections on this module has two MOSFETs and a .82 ohm power resistor plus two 1/8 W 15 ohm resistors. Each of the two MOSFETs seems to be driven separately by one section of a dual JFET opamp. At first glance I thought the two MOSFETs were in parallel but that isn't the case, if I traced that part of the circuit out correctly. If you look at the schematic I have below, the IRFPF40 (rated 4.7A) has the two small 15 ohm resistors in between its source and the source of the IRFPF50 (rated 6.7A)., and the IRFPF50 has the .82 ohm power resistor basically in series with both MOSFETs. As a guess I think the lower rated MOSFET handles the load when the current is lower, perhaps up to 1.5A, with the IRFPF50 off and the higher rated MOSFET, with the power resistor, is used in addition to the lower rated MOSFET when the load increases from 1.5A to 20A.  Maybe they get better control at the low end that way-I don't know.

If you were only looking for 300ma and each of the IRFPF50 MOSFETs can handle 2.5A, I'd assume you might be able to just use one of these or a similar part on a proper heatsink with your own feedback and control circuit.   
 

Offline duak

  • Super Contributor
  • ***
  • Posts: 1048
  • Country: ca
Re: High Voltage Bench Power Supply Design
« Reply #80 on: July 17, 2019, 10:44:37 pm »
Fair enough AD.  No reason comes to mind other than different current ballasting resistors for different ranges.

I remembered a design idea in EDN for a wide range load where paralleled FETs were used and the gate drive switched in and out for high/low ranges.


 

Offline H713Topic starter

  • Regular Contributor
  • *
  • Posts: 180
  • Country: us
Re: High Voltage Bench Power Supply Design
« Reply #81 on: July 18, 2019, 06:19:02 am »
I also found that there is a "sweat spot" range for FET voltage rating. Too low, (100V like the IRF510), and if obviously has minimal SOA, but too high (like the FQA8N90C and the 2SK3675), and they seem to perform rather poorly at 120V in the linear region compared to the IRFP450s. Some of this may be due to the age and product family of the transistors, but I think that 900V parts are not designed to handle as much current as parts in the 500V range, and I suspect that this is hurting their performance here.

Really, I think parts like the IRFP460 and probably IRFP360 like Duak mentioned are probably the best option. A lower power version could be built using IRF640s, they performed reasonably well (they don't have DC specs, but as we have seen that means little), but the TO-220 package is a limitation. They were used by BSS Audio in the EPC 760 and EPC 780 power amplifiers from the early 90s as output devices, and they are a good part. For a 0-500V 0-100mA supply, they would be pretty comfortable with about 6 in parallel. The FQA8N90C performed considerably worse in this application, blowing up at or before the IRF640s did, and costing over twice as much.

IMO, range switching drastically improves this situation, as a 200V @500mA output (perfectly reasonable for many applications) would result in a total dissipation of 200 watts. That's a lot of heat and comes out to about 32W per transistor. That doesn't seem bad, but that makes for a big heatsink, noisy fan and and efficiency of just 33%. Using something with a pair of 120V primaries would solve this issue, at least in North America. In the low range, the primaries would be wired in series (as if wired for 240 operation), while in the high range they would be wired in parallel. This transformer doesn't have to cost of a fortune as a result. In civilized parts of the world, you would want something with a pair of 240V windings on each side.

While we're on the subject of the power transformer, I know there may be people who worry about the use of a control transformer wired backwards to get a pair of 120 windings on one side and a pair of 240 windings on the other. It is fine in most cases to use a transformer backwards (secondary as primary) so long as the voltage is correct. That is, applying 240V to a 120V secondary will saturate the core. When this happens, the inductance drops like a rock and the winding will pull heavy current and burn up in short order. By contrast, applying 120 to a 240 winding is fine, though you need to keep your copper losses in mind.

The really nice way of doing this is to use a variac driven by a servo motor for this. I've seen this done on some really big magnet power supplies from the 70s. With a smaller variac (rather than about 500 pounds worth of them), it could be pretty fast-tracking. Something for someone a little more ambitious than me. That could, in theory, drastically increase the possible output current and make this quite an efficient power supply.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf