I am building a HV (5kV) supply to bias an ionization chamber.
The supply needs to be clean, both from the converter ripple and the external electrostatic influences.
The logical solution would be a large bypass cap, but at 5kV, it is unpractical.
I plan to use 1.5nF caps, and multiply their effect using an active circuit.
This circuit works:
But I wonder if simple improvements are possible, without going overboard, compromising the stability or using insane components values.
Any ideas?
The supply needs to be clean, both from the converter ripple and the external electrostatic influences.
The logical solution would be a large bypass cap, but at 5kV, it is unpractical.
I plan to use 1.5nF caps, and multiply their effect using an active circuit.
This circuit works:
But I wonder if simple improvements are possible, without going overboard, compromising the stability or using insane components values.
Any ideas?
Member
Joined 2009
Paid Member
You're on the wrong forum for that.All xerox machines have 4...10kv bias circuits for that.
A high frequency simple resonant circuit with a voltage multiplier will do the job very well.
I can heartly recommend this circuit from Dave:
https://highvoltageforum.net/index.php?topic=1439.msg11002#msg11002
Then you add a voltage multiplier with resistor bleeders of 1...50megohms in parallel with each capacitor .
There are people there that did everything you can imagine in the high voltage domain.
A high frequency simple resonant circuit with a voltage multiplier will do the job very well.
I can heartly recommend this circuit from Dave:
https://highvoltageforum.net/index.php?topic=1439.msg11002#msg11002
Then you add a voltage multiplier with resistor bleeders of 1...50megohms in parallel with each capacitor .
There are people there that did everything you can imagine in the high voltage domain.
No ideas from Linear Tech/Analog Devices application note 118? https://www.analog.com/media/en/technical-documentation/application-notes/AN118fb.pdf
Jim Williams did a few app notes on the subject, but out of convenience he placed the inductor in the wrong place.The inductor should be from positive rail to mid tap of the transformer as Baxandall placed it innitially to be more efficient.
The top resistor of the feedback divider is 1gigaohm; it is very high, but it consumes 5µA, and the flyback converter operates at a very low frequency, low duty cycle: a few tens of Hz, a hundred max., and the conduction time is a few µs. This is sufficient to cause a significant ripple on the final 10nF filter cap.
This very discontinuous mode of operation is required to achieve a low current consumption, a few mA, because the equipment is portable and operates from a 9V battery.
Adding a number of >5kV capacitors is certainly possible, but unpractical for a small, hand-held device.
The resistors would consume much more power than the 1G feedback resistor, which is already borderline.
In addition, the converters operate in a ~continuous mode, which is incompatible with a low power consumption
Same remarks, and the voltages illustrated are an order of magnitude smaller than 5kV. The filter caps are also proportionally much larger.
Basically, I don't want to dramatically alter the topology/configuration: the current circuit provides a well-regulated, 5kV supply whilst consuming under 5mA, which is quite nice considering the simplicity of the circuit: one BJT, one MOSfet and a standard CMOS IC.
It is very compact too. I want to keep all of those features, but at the same time reduce the noise to insignificant levels. I know that I can use brute force: increase the cap values, or push R2 to hundreds of megohms, or use a multi-opamp corrector, but if a simple and clever fix exists for the current filter circuit I would like to know about it; I have already searched the net, but I have found no satisfactory solution.
As the current consumption is insignificant, I could increase R1, but it would make the circuit more sensitive to external electrostatic perturbations, which is undesirable
Member
Joined 2009
Paid Member
5mA at 5kV is not a small current.
What about a shunt regulator, it's current hungry compared with a series regulator but....
What about a shunt regulator, it's current hungry compared with a series regulator but....
He means 5mA at 5...15v maybe for the op amp operation which makes the current at 5KV so small that even the 1.5nF caps will take a while to charge up...
Yes, the output current for the 5kV is ~5µA, due to the feedback divider. The ~5mA is the input, drawn from the 9V battery
I have explored this site already, but the ion chambers there are operated in the plateau region, ie true ion chamber mode. I have experimented with it, but now I want to upgrade to the proportional mode, requiring much higher, and preferably stable and clean voltages. The advantage is a greatly increased sensitivity, intermediate between a plain ion chamber and a GM tube, but with a discrimination ability
Capacitors at high voltage exhibit leakage currents that can scale up in the microampere range depending on voltage and capacitance value also on cap type so avoiding the bleeding resistors is not a guarantee for lower power .Usual v multipliers don't necessarily have bleeding resistors.
Last edited:
No, not at all: it is just good clean fun.
The last time I underwent a scintigraphy, I remained detectable from everywhere in the house for a few hours 😵💫 ☢️
It gives you a new perspective on radiological safety....
Hi Elvee, not sure if that topology will work for your project; but I want to mention if it does, the output voltage will be 110kV.
The current through R1 =1M is the same across R2=22M so Vout=(5kV/1M) 22M = 110kV
😵
The current through R1 =1M is the same across R2=22M so Vout=(5kV/1M) 22M = 110kV
😵
You have a nice FDNR (Y = K s2) with damping capacitance now. Replacing R2 with a T network, with the branch to ground being an RC series network, could maybe result in an K s3 term and damping terms in the admittance from out to ground and make it a third-order filter. The keyword is maybe, I haven't calculated the transfer yet.
Not really relevant, but thanks anyway for your tentative
Thanks for that analysis: what I tried to achieve when I arrived at that particular configuration was a low impedance as seen from the junction of C1, C2, R1, whilst using only ~low value caps (because they are HV).
I didn't aim at or see a FDNR in it before you mentioned it.
I will try to simulate the circuit using a T-network (thus far, I didn't simulate anything: it was just mental, pen and paper, and physical.)
What I dreamt of was a capacitor-multiplying circuit using a single capacitor, C1 or C2 as a reference, something like this:
It would kill many birds with a single stone, but thus far I didn't manage to find a solution (but I have a deep, strong feeling that it is possible).
The best I could find was the circuit of the first post.
Using a synthetic, negative capacitor having a value close to 1.5nF could create a positive combination much larger than 1.5nF, but the problem is that it uses a second capacitor, and it would need to track closely the 1.5nF, which is a HV ceramic, X7R at best, not particularly stable.
The schematic shown is just the barebone concept; the real circuit includes many additional resistors, protection diodes etc. to deal with mishaps inevitably inherent to HV circuits. I don't want to confuse the situation by adding components having no function during normal operation.
I have tested Marcel's idea: it works very nicely, and allows asymmetrical values, which is convenient, as one of the HV caps can be reduced to 330pF:
The reduction in performance is minimal, and the circuit allows for more than 40dB reduction at 50Hz. The settling time should remain tolerable.
I have tested Marcel's idea: it works very nicely, and allows asymmetrical values, which is convenient, as one of the HV caps can be reduced to 330pF:
The reduction in performance is minimal, and the circuit allows for more than 40dB reduction at 50Hz. The settling time should remain tolerable.