I build the proto type, this high voltage supply can do more than 2 kV output. It runs from 5 V DC, and uses about 11 mA with 1250 V output. It is controlled by a DC voltage of 0 to +5V, the output voltage is always 546 x this input voltage. So for a control voltage of 0V, the output is 0 V, and for a control voltage of 2 V the output is - (minus) 1092V. The relation is linear, but the sign is reversed.
Upon multiple requests and mainly because I could not do this in ASCII art, and all I have is pencil sketches, I did draw the diagram in a more Christmas like color (coming up soon):
formatting link
The interesting part is the control circuit. I forgot to order CMOS opamps, and anyways those often have no offset compensation. But I found some BF245 FETs while looking for OA91 diodes? Or did I find the diodes looking for the FETs? anyways build my own opamp. After several sketches (no spice I do not use it), and doing calculations for a LM324 like, but an LM324 has some input current, and this is all very high impedance, that would create > 20 % error. And FETs have large tolerances and not much gain in circuits like this. Also the common mode range was problematic for all those kind of solutions. So, here it is, zero common mode range, and oh yes, the output goes negative but there is no negative opamp supply. So this circuit solves this. The gate of Q7 is always at zero volts +- a few millivolt. There is quite a bit of gain, as Q5 drives the top series regulator (beta). A PIC can generate a 0 to 5 V voltage using 255 PMW steps (or more, it can do 10 bits too at a lower frequency), this is filtered and then used as control voltage, The impedance of the control system is 2M2, and should be driven from a low impedance source (relative to 2M2). The factor 546 I have done in the software, so you can enter the PMT HV voltage directly. The high voltage divider is simply[] 10 120 MOhm SMD resistors is series:
formatting link
on the left side, vertical. That capacitor should go to the other side later:
formatting link
Oops, 7 transistors! And no I did not have a smaller trimpot of the correct value. All the way on the right the 2M2. The capacitors are 400V DC types, because, after all.. 5 x 546 = 2730 V. Have to tried that high, and will not try that high. Some other features: There is a build in current limiting for the PMT, so it does not fry if you accidently keep it in the light, This current limit is set by the beta of the series regulator (measured 250), and the current mirror's resistor to +5V. And other interesting aspect is that this circuit has different voltages for the different dynodes, for example the voltage between the cathode and A1 is 2 x the voltage between A1 and A2, and the voltage between A8 and A7 is supplied by a separate rectifier and 1.5 x the voltage between A1 and A2. The whole thing uses about 11 mA at 5 V for 1250 V output.