more HV supply stuff

Still working on the HV supply, did some test measurements, and designing a pcb in eagle (cadsoft).
The circuit is really working well!
This is the eagle schematic, pcb will follow soon when I have al the parts.

It is working much better than a normal voltage multiplier, and I was wondering how much stages could be added, I tried 11 stages, the circuit still works fine in the simulation, with about 3.6W of output:

The next image shows the "why", the driving voltage peaks are very short spikes, the lower line of capacitors have a very low impedance for these spikes (or in other words, the capacitors form a high-pass filter, but only a very small part of the energy of a delta function is in the lower frequencies)

It should be possible to add a lot more stages without loosing much energy.

HV supply for neutron counter or geiger müller tubes

I tried to make an HV supply for counter tubes before, but making the transformer took some time, and it didn't work well.
Based on the ideas in my previous post I tried to combine a switched DC-DC converter with a voltage multiplier, this is the good-old-pen-and-paper result:

You can click on the image to see it enlarged.
I'm using the MAX1771E, and the circuit is actually quite simple.
The only non-trivial component is the 1M resistor between the 1.5V REF en the FB input, it is needed to lift the FB input above 50mV on startup, so it forces the MAX1771E into non-bootstrapped mode, even if the output voltage is 0.
Without it the MAX1771E won't start unless the supply voltage is below 12V.
When the output voltage is stable, the voltage on the REF and FB are both 1.5V, so the 1M resistor does not introduce any error.
The off-switching of the MOSFET creates a high dU/dt peak at the drain, the lower line of capacitors have a very low impedance for this, because of the high dU/dt.
The upper line of capacitors are charged by the peaks, but they don't have any AC feed, so the output voltage at the end of the upper line is stable.
The peak drain voltage is limited by the energy in L and the combined Cdg and Cds of the MOSFET (about 1n for the NDF10N60ZG, dI/dt is not the limiting factor), peak can be calculated with LI^2=Cmosfet * U^2, where I is the max current set by the current sense resistor.
I started with 30mOhm current sense, giving an Imax of 3.33A, and a peak of about 500V (calculated) or 670V (measured).
Output current was 0.5mA at 1600V.
It worked great, but the MOSFET was getting a bit hot, maybe be because of breakdown, but probably because the on resistance is about 0.7 ohm.
Also, the circuit keeps pumping in power (about 1.25A at 12V), even without load.
With a sense resistor of 55mOhm everything worked fine, and the input current without load dropped to 250mA.
So the next step is using a MOSFET with a lower on resistance, bigger HV capacitors, and better diodes.
I've ordered the SiHG47N60E MOSFET's (64 mOhm), AU2PK diodes and got some 100nF/1kV caps from ebay.
This is a simulation of the maximum output power at >1600V (1637V) into 200K.
Seems to be ok, 13.4W out at 12W in, a little bit impossible but at least the simulation doesn't show any big losses.
I used 7us on, 2.3us off and 100ns rise and fall time to simulate the max power output signal of the MAX1771, this gives a peak current of 4A in the inductor.

Adding a stage and simulate with a 2M load gives an output voltage of 3800V at 1.9mA, but the really great thing is that the voltage step per stage doesn't get less at higher voltages, this is because of the high dU/dt of the spike at the drain of the mosfet, the impedance of the capacitors of very low for this spike. I wonder how many stages can be added before each stage starts to step up less than the previous one.