[HELP] High Voltage Power Supply problem

[valvedoctor]

Hello all,

I'm looking at this circuit (attached PDF) from the Art of Electronics 3rd Edition.

I'm wondering if anyone can help me solve exercise 9.15. I'd like to use a similar thing in a design I'm doing. Any help at all would be greatly appreciated.

Best regards,
Doc

[rs20]

As with any book-or-homework question like this, we have to ask what progress you've made -- have you got any design in mind? Have you figured out why an LM385-2.5 might be useful? You surely can't be entirely 100% stuck?

[valvedoctor]

Hi rs20,

I have limited knowledge with solid state devices. I'm more or less a tube guy. I need this to supply about 350 volts at 200ma maximum. I was hoping somebody could educate me on this and how I can determine the exercise.

Thanks,
Doc

[David Hess]

Remove Q2 and its resistor, remove D1, and short D2.  Place the LM385-2.5 in series with the gate of the changed Q1 to raise its Vgs.  Add a resistor from the drain to source of Q1 to bias the LM385-2.5 correctly or use a current source like the original Q2 circuit.

Carefully check for catastrophic failure modes which may require adding some diodes.

This is a pretty standard circuit for adding a cascode for high voltage operation to a voltage regulator.  There is no reason to use a depletion mode device for Q1.  The original circuit configuration is probably more economical.

[valvedoctor]

Hi David,

Here's a LM317 variant based off other designs I saw, which would you suggest?

EDIT: UPDATED SCHEMATIC

[David Hess]

Check out my revision. Is this correct? I may have a mistake made...

Also here's a LM317 variant based off other designs I saw, which would you suggest?

The first one needs an R2 between the drain and gate to supply bias to the gate and LM385-2.5.  Also Q2 is a depletion mode MOSFET in this one.

The second one is what I would recommend and I have used it before.  I did not have an resistor in series with D7 and I did not bother with D5 and R4 and R3.  The voltage of D7 is adjusted to get the best voltage across the regulator.  There are refinements which can be made but they depend on performance requirements.  R2 needs to be sized for an appropriate current through D7; it will be necessary to replace R2 with a high voltage current source for wide output voltage range.

[valvedoctor]

Hi David,

I kept R3 and R4 just to prevent oscillation. What purpose does the D5 serve?

I only need around 325 - 375 vDC.

Otherwise, the schematic looks okay?

[David Hess]

D5 combined with R4 operate as a current limit which is not a bad idea.

The voltage across R4 is limited to the breakdown voltage of D5 minus the Vgs of the MOSFET.  Since it relies on Vgs it is not well controlled but it does not have to be; it is just for protection against faults.

[David Hess]

Take a look at figures 11 and 12 on pages 6 and 7 of Linear Technology Application Note 2:

http://www.linear.com/docs/4099

They show and discuss the same cascode configuration with a floating linear regulator you are considering.  I have used this same design several times including in a military project which went onto the F/A-18.  When my boss asked if it could handle the specified over voltage condition of like 90 volts, I just laughed and plugged it into the wall socket.

Over on funwithtubes@yahoogroups.com, there has been a recent discussion about building tube based bench power supply which led me to digging up the above application note by Jim Williams and passing it along to you.  The August 2014 edition of QST also has a design for a tube based power supply which you may find interesting:

http://www.arrl.org/files/file/QST/This%20Month%20in%20QST/August%202014/JULSTROM.pdf

The changes I would make to the cascode design include foldback current limiting which lowers the output current as the output voltage drops to protect the cascode transistor and input supply and a current limit indicator.  The foldback current limiting should be used in your case because with say a 500 volt input supply and 200 milliamp current limit, the short circuit power dissappation will be about 100 watts which while possible to handle is inconvenient and will require a much larger cascode transistor and heat sink.

Another feature which might be worthwhile is SCR crowbar protection on the output.  If it detects an over voltage condition which could occur if the pass element shorted out or if the voltage control potentiometer went open, it shorts out the output to protect the circuit being powered and blows the input fuse.

The only other features that immediate occur to me which you might want are a precision adjustable current limit and improved output regulation.  Both would add considerable complexity however.  Improved output regulation is probably a waste of time unless you want to use the supply as a high voltage calibration reference; integrated voltage regulators already have good enough performance for typical power supply applications.

Something to watch out for in this design is the lower resistor of the output voltage divider.  There is an assumption about the minimum output current that LM317 style regulators require which is normally provided by the upper resistor but it also flows through the lower resistor.  If this current is say 10 milliamps and the output is 500 volts, then the lower resistor has to be able to dissipate 5 watts which will require a pretty beefy potentiometer.  This requirement is lowered of course if a series resistor is used to set a minimum output voltage.  There are also other ways to lower the current through the lower divider resistor if necessary.