I had pretty much dropped out of this thread because it looked as if you and Ian were working it out.
I think we need take a step back. The original problem was that the LM2596 is getting hot. The LM2596 regulates the 5V VCC rail but the HV generator does not run off that rail - it runs off the VIN (presumable 12v) rail. So why are we looking so hard at the HV generator?
Something is loading down the VCC rail. but as far as I can tell the HV generator is incapable of doing that. But that current has to be going somewhere. So the best I can think of is to try running it with all chips removed from their sockets except for the Atmega. Please try it and report back...
Look into it later when the dust is clearing off the crater.
At the moment only the Atmel is installed on the board.
The 5V rail runs completely normal without the Atmel. I would also say that the noise comes from the HV rail. Unfortunately I don't know how the HV regulation must look like. Therefore I find it difficult to interpret the oscilloscope images.
I will interrupt the HV-rail again today to make sure that the HV-rail is not the problem.
I still suspect that the Atmel is doing something stupid
I'm using a new Atmel and I programmed it directly. So it always runs on the internal crystal. That's why the wrong capacitors had no effect. The internal crystal should run with 10Mhz by default (the program was compiled for 16Mhz), I'll have to find out the fuse settings for the arduino - with a little luck that's the problem.
I suggest to mention the fact with the fuses in the manual (when programming with another programmer).
Unfortunately I could not reach the Atmel via the STK500. So I uploaded the hex-file via the Atmelstudio (you can find it under:
or search for the file FirmwareV2.ino.hex
the current consumption is about 0,05A and after a while it drops to 0,01A (surely after the HV string was switched off)