CST - Completely unable to get sensible results from a basic PIC simulation.

PEC tube with vacuum, beam energy 35 MeV, with a 800 fs longitudinal sigma, cutoff length 3*sigma, 0 offset, 6*sigma distance. It's being emitted from a PEC surface 225 um radius. The tube itself is PEC with a width of 3.33 mm, height of 1.4 mm and a length of 10 mm. PEC walls are 100um thick, the boundaries are open, and the background is normal.

The bunch generates significant noise, with rippling on the self field, and huge wakes (in a PEC tube). I assumed it was numerical Cherenkov so I've tuned the "special dispersion" properties of the vacuum to the frequencies of both what the calculator macro calculates, and what the FFT of a point probe outputs with various resonance widths. I have varied the mesh between 5-50 mesh/lambda at 1 THz and the effect is still present. I've used different units ensuring the absolute value is still the same. I've turned down the stability factor per timestep incrementally from 1.0 to 0.6. Automatic time sampling is off, as is local time stepping in the integrator setting.

I'm completely at a loss as to why I cannot get sensible results from such a simple calculation. The results are obviously wrong, and yet nothing I seem to do can fix it. Attached is the E-field monitor showing E_abs after 35ps propagation.

CST Studio Suite ​​​​​​​