Hi Radu,

I played around with the R1 and R2 values (by soldering in potentiometers) and did not find that they made much of a difference with higher values– in terms of voltage stabilization or reducing the duty cycle.

However, I did find that lowering the inductor frequency would drop the duty cycle 10% or more with each 1khz reduction.

At 7-8khz the duty cycle is solid at 50%. This seems really low– do you think this would affect accuracy or is the khz primarily a concern of audible noise?

It seems I have two separate problems, the high duty cycle and the wild voltage fluctuations.

The duty cycle was fixed by lowering the freq. Then I found that the wild voltage fluctuations went away by attaching the ENC28J60. Without the ethernet module attached, there are 40-50 voltage swings. With the module attached it settles to deviations of 10 volts but with most of the time resting at a steady 380v

I have watched the 3.3v rail and it stays clean, no noise. I tested powering from batteries, DC jack, rigol bench power supply, and USBasp programmer set to 3v. I believe that the voltage regulator is fine as this was tested with the DC jack.

When it was running steady at 380V / 50% duty, I logged data for about an hour and noticed waves of brief high CPM readings (low to normal most of the time, then 5-10 seconds of high readings). I question their accuracy however, there is nothing around that should be producing high radiation. Do you know if the tubes are susceptible to RF interference?

I will continue to investigate. It may just be this first build is haunted. I will begin constructing another one for comparison