Update on March 17, 2016

On March 16, we have deployed the trigger plan B upgrade to improve the online VpdVz resolution. Here are the findings from run 17076043 (this run with only vpdmb-5-p-sst trigger) from a fast offline production

The average purity is 84.4% when using the first vertex that is correlated with VzVpd.

================

Fast offline production data for run 17038023. TPC S&C and Gridleak calibration is not in place yet.
This run was taken with only vpdmb-5 trigger after the slewing correction table uploaded (no trigger plan B upgrade yet). Trigger rate (corrected ZDCx rate) is 39kHz.

We know the vertex ranking algorithm in recent year Au+Au data is not optimial. In the following study, I used two algorithm to select the primary vertex

1. - default first vertex from MuDst
2. - loop over all vertices (with ranking ordering) and find the first vertex that satisfies |Vz-VpdVz|<3cm.

The following figure shows the correlation between Vz from the TPC and Vz from the VPD.

I looked at the Vz distributions from two different algorithms vs. refmult. See the following plot. Solid squires - means from Gaussian fits. Solid circles - widths from Gaussian fits.

The right plot shows a comparison of the refmult distributions from two algorithms (red- default, blue - with VzVpd constrain).
It is clear that the default vertex ranking often pick vertices with very few tracks as the highest rank. With the VzVpd constrain, we can re-gain many of the central and mid-central events.

The vertex purity (ratio of events within |Vz|<6cm over total) are shown in the following plot. The solid lines show the averaged ratio over all events which is

48.5% - default
72.8% - with VzVpd constraints. 

I think we can use the number 72.8% to estimate our number of good events taken with this trigger. A few possible more studies
1) Check the real physics trigger (vpdmb-5-p-sst)
2) luminosity dependence or in-store time dependence
3) Check the online HLT labeling efficiency and purity
4) With TPC calibration updated, may check these again