linac errors

[astecpete]

Hi Michael,

We're doing a simple error analysis and have a little problem. Attched are a couple of files. We run twice, the first is a fiducial run with one particle, the second we apply an error to the voltage of cavities and track 1000 particles. The trouble comes when we make this error small. In the file i have made the error zero. When we do this and plot out Cdelta, we see that it still diverges from the centroid and ends up at ~6e-4 fractional difference. When we have large errors, we see that in Cdelta but as we turn it down, Cdelta asymptotes to this 6e-4 value. Looking in the .error file, the values for VOLT are correct (i.e. the values from the second run are the same as the first). This effect seems to be undiminished when i increase particle number so doesn't seem to be statistical.
Can you help!

Pete (and Sara from MaxLab)

[michael_borland]

Pete and Sara,

This happens because of the first_is_fiducial=1 flag in &bunched_beam. This causes elegant to fiducialize with a single-particle bunch. That bunch apparently sees very different wakes than a realistic bunch, so it isn't a good fiducial. I should probably put a warning about this in the manual.

I recommend setting first_is_fiducial=0 in &bunched_beam. Also, you might want to consider setting one_random_bunch=0, so that the same particles are used for all runs. That way the effects you'll see will only be due to random errors assigned via &error_element.

--Michael

[astecpete]

Ah, i didn't think of turning the wakes off.
Thanks very much Michael