I need a small, low-current, very low noise high voltage supply to bias an exotic and expensive photodiode, measuring the energy of a
13.5 nm light source. I have +12 or maybe +15 to start with, and I want around +100 at less than a mA average. I looked at all the conventional approaches, flyback inverters and such, and didn't like them much: too big, too noisy, custom magnetics, snubbers, loop stability issues, simulation, too much hard thinking.I was reminded of my previous concept, making a C-W multiplier chain using switched SSRs instead of AC drive and diodes. Phil Hobbs, certainly in a spirit of friendliness and good cheer, called this idea the Groucho Marx Generator.
OK, how about this:
ftp://jjlarkin.lmi.net/Harpo_Marx.JPG
This is the Harpo Marx Generator. A C-W/Groucho increases voltage linearly with the number of stages. This one doubles voltage every stage, sort of like the octaves on a harp. The dual diodes would, ideally, be more SSRs, but the diodes are small and cheap and I can tolerate the drop here.
Both Marx circuits can use big caps and switch at low frequencies, 500 Hz maybe, so they will be very quiet. The caps can be biggish ceramics or aluminums.
Each SPDT switch is a dual SSR. If I do three stages and drive each SSR control input from an FPGA pin - I have tons available - I can gate the clocks and select +24, +48, or +96 volts out, minus some diode drops. Good enough. Or build a hybrid, a true Marx Brothers Circuit, that does 24, 48, 72, or 96.
John