Fast offline has run, and we have triggers set up and elevated to physics.  This is a first look at multiplicity distributions.

Checking two triggers: 260001, which is a vpd-mb, and 260101, which is a central trigger.

Looked at runs

11003001
11003002
11003004
11003005
11003006
11003010
11003011
11003012
11003013
11003014
11003015
11003016
11003017
11003029
11003030
11003031
11003062
11003063
11003064
11003065
11003066
11003067
11003068
11003075
11003079
11003082
11003101
11003102
 

Looking at refmult distributions, actually better refmult^1/4 distributions.  For reference, earlier runs had these distributions:

First, the vertex seems to have good resolution for everything but the most peripheral events or so,

and is cut at ~35 cm.

One can make a profile histogram out of this to check where the bias sets in.  Looks like 40 cm or so.

The endpoint for the vpd-mb is a bit shifted down relative to Run 4:  4.7 (Run 10) vs 5 (Run 4), which indicates a 20% or so loss in efficiency.

This should cleanly go down to 1.2 or so in refmult^1/4 (pp refmult is ~2-2.5, 2.5^1/4 =1.25)

Appears to be the start of bias ~2 in refmult^1/4, which is ~(4.7-2)/(4.7-1.2) =~70-80% centrality

Looking at the central trigger 260101, it appears that we are not fully dealing correctly with the double-valued, since there is significant population at very low refmult.

If I were to take the old 5% central cut in run 4, of 520, and scale it down by 4.73/5^4, get 420, which is about where the central trigger 260101 appears to fall off from 260001.