Electronics > Projects, Designs, and Technical Stuff

High Voltage Bench Power Supply Design

<< < (11/17) > >>

Wolfgang:
Never seen an IGBT with DC soar. Interesting ! Any types recommendable ?

duak:
H713, the pass transistor stage may be oscillating by itself.  Bipolar transistor emitter follower circuits can and do oscillate with capacitive loads.  The solution there is to add damping resistance in series with the base.  I expect that FET source followers could do the same.  You already have series gate resistors here of a reasonable value so I would try increasing the values of the source resistors R7 - R9 to a few ohms and see what happens.  As I mentioned before, larger resistors here will help balance the currents in the pass transistors.  You can also try a series RC damping network between the output terminals - try 10R and 100n and see what happens.  You've already found that electrolytic capacitors with their higher losses do not cause the trouble that the film caps do.

You've got a lot of voltage gain in this circuit, and it's distributed among a number of stages each of which introduces a delay - this is a recipe for an oscillator.  I would analyze the circuit to determine its frequency and phase response and do what I could to improve the phase margin.  Do you know about Bode plots or Root Locus analysis as applied to feedback systems?  (I used to know more at one time - if you don't use it, you lose it)

H713:
A little further testing revealed that smaller values of electrolytics do cause a small amount of oscillation, something like 30-40 millivolts. A series RC network did not cure the problem, nor did a gate resistor on Q1. I currently don't have a source resistor in the test sample, as there is only one pass element being used. I can try that, but I doubt it will cure the issue.

What did cure the issue is the brute-force solution of putting a 100uF electrolytic on the output. This sort of defeats the current limit protecting the DUT, though it still protects the pass transistors. While I personally would not have any problem using a supply like this (it's still less scary than a lot of the power supplies I work with), It's really a brute-force solution to the problem and makes this design considerably less useful for others.

Duak,

Bode plots and Root Locus analysis is a bit over my head, or at least the terminology is. With audio amplifiers, I occasionally do test the frequency response and phase shift to get an idea of where instabilities are coming from, though normally with well-designed tube amps it comes back to the output transformer, which there isn't a ton you can do about. That said, I'm really not sure of the best way to test a circuit like this. It's not like I can feed an AC signal to the input and look at it on the output.

duak:
H713, Many linear power supplies do indeed use an output capacitor for stability and it sure does render the current limiting moot for transient loads.  A sure fire way to get around the current transient is to design a supply that inherently current limits, but the voltage is varied by clamping it to the desired value.  The hp 6186C supply works this way.

Back to your circuit, consider it to be the positive going side of a Class-A audio amplifier.  By injecting an AC coupled square wave, say 100 Hz, into U1 pin 2 and then following the signal path you can see how the various stages respond and try various things.  Because the feedback loop is delaying the signal, you can try paralleling R13 with some small capacitance, say 1n0 with some series resistance.  This is called lead compensation.   Also, because you have gain to burn and the problem is that there is too much at high frequencies, reducing gain without adding delay can be helpful.  Try increasing the value of R3.

H713:
I didn't have a chance to do the frequency/phase analysis today, however I did experiment with a few of your suggestions. I was able to increase the value of R3 to about 2.5K if it is connected to the -18V supply rather than GND. This makes things better but doesn't kill the issue cold. Unsurprisingly, the power supply is less stable at lower output voltages. With this configuration, it is good down to about 75 volts with a 33uF output capacitor, but below that threshold and the 33uF capacitor causes an oscillation in it of itself. With a 100uF capacitor it is fine. If I have time tomorrow I'll try the 1nF cap across R13 and see what that brings about. As I understand it, this is essentially to roll off the frequency response (and lower the overshoot on square waves).

Quite honestly, I don't have a big issue with a big output capacitor in the 100uF range. It will take up considerable board space and add cost, but this isn't a production unit so that is not a big deal, especially when I have the parts in stock. I also don't have an issue with its impact on current limiting. The point of the current limit in this design is not to make this thing idiot-proof (no 500V power supply will ever be), but rather to avoid literally melting the DUT and/or blowing the pass elements. The main thing I don't like about it is the fact that I feel like it is a band-aid fix to avoid proper debugging and circuit modifications. With that said, it seems to be a fairly common solution to this problem. I'd like to get this circuit as stable as possible before adding the output capacitor, however.

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod