The main mb trigger used for the HFT physics program will be vpdmb-5-p-nobsmd.

The pileup protection for contaminated events are defined as (March 22, 2014):
ZDC-Coin & ZDC-TAC & VPD-Coin & TofMult>350
Protection window is 400 time bins for both past and post protection.

The following figure shows the observed rate after implementing the pileup protection compared to the expectation.

Solid and open circles:   observed tape rates (built events) (physics runs from 15081034-15081043)
Black solid line:  Expected maximum tape rate  ( VpdXsec = 6barn, VPDMB-5/All = 0.065, TofMult>350 => top 50% central collisions).
Red solid line:  Expected maximum gating-grid firing rate (after the past protection at L0).
Blue dashed line: targeted tape rate goal ~ 500Hz to get "good" event rate at around 300 Hz, so that we can reach 1B "good" event goal with 72 DAQ hours per week for 14 weeks.

Conclusions:

1) The observed rates are very close to the expectations.

2) To reach the mb physics goal, we need to boost up the tape recording rate to the black solid line (maximum expected), which means generally no prescale on the vpdmb-5-p-nonsmd.
     - From Run11 experience, with the stochastic cooling in place, one would expect the typical stores with ZDC rate from ~45-20 kHz (green dotted lines). This means the GG rate  need to fire at around 600-900 Hz.
     - The longitudinal cooling may improve the VPDMB-5/All ratio, thus all the rates in the plot. We may apply some prescale, but the targeted tape recording rate is about still close to the solid black line.

===================
Update on March 24
===================
So to reach the goal, the proposed plan is shown in the following figure.

The solid red and black lines are the targeted L0 (including prescales in possible) and tape recording rates in two different scenarios (different VPDMB-5/All ratios). 
- With the current beam condition, VPDMB-5/All is about 0.065, to reach the goal, we will not have the room to prescale at L0.
- If VPDMB-5/All is improved, we should have more available collisions so we can apply some prescales particularly in the beginning of the fill to reduce the GG rate. The more this gets improved, we can have more collisions available in the end of the store (lower luminosity) where the rejection factor is small, then we can put more prescales in the beginning of the store to balance the total TPC GG rate.

The open circles are the estimated L0 rate from the runs listed above. So generatelly, we need to boost that up to about 700-800 L0 Hz to reach the goal. 

===================
Update on March 26
===================
CAD has pretty much the full stachastic cooling on since March 25. This gives us a better estimate of the collision rate we going to experience in the run. Taking one long store starting from roughly midnight to early morning March 26, here are the observations and the tentative plan for the VPDMB rate we are going to run.

Black dots are the recording rate for the main vpdmb physics trigger we are currently running in the store (solid - the real recording rate including ps, open - no prescales)
Red dots are the estimated L0 rate based on the recording rate and the expected future protection rejection factor. This is just a quick estimation assuming a simple exponential rejection. Similarly, the solid ones are those after prescales, and open ones are those without prescales. The estimated L0 rate without prescale is toughly consistent with the expectated L0 rate, considering the VPDMB-5/all ratio changes from 0.08 in the beginning of store to 0.06 at the end of store.
http://www.star.bnl.gov/protected/common/common2014/trigger2014/plots_au200gev/VPDMB-5-nobsmd_450014.pdf

We are already now running at 1.4 (1.5 at the very beginning) KHz TPX rate at the beginning of the store (TPC limit is now set to 1.8 KHz). So saying the L0 rate for VPDMB trigger fires at 800Hz or below (shown as the solid blue line), the total TPC rate will be ~1.7 KHz. The recording rate for VPDMB trigger will be around 400-500Hz shown as the black line. In this scenario, we will be likely 10-20% short from our goal.

More L0 rate at the beginning of the store is not so effective to improve the recording rate due to the high rejection rate from future protection. To meet our goal, there are two possible directions

1) Tighten the cuts used to define the contaminated events, perferrably increase the TofMult threshold cut used to , i.e. protect against more central collisions, this will end up with a smaller rejection factor from L0 -> tape rate. Currently, the TofMult cut is set to be 350, roughly correspond to 50% central events. If we increase this cut to be roughly 40% central, that will meet our targeted goal on the recording rate.

2) The longitudinal coolling seems to improve the VPDMB-5/All ratio, but it drops very quickly with the store (faster than the real luminosity). If CAD can improve further the longitudinal cooling, particularly in the second half of the store, that will be very helpful. Currently the available collisions is already below 800 Hz at roughly 40KHz zdc rate, 

===================
Update on March 27
===================
Jeff raised the concern of the total DAQ bandwidth in the beginning of the store, there is little room to increase as we are almost at the limit. So if we still don't try to sacrifice other triggers, the plan after the discussion at a special trigger board meeting is
  - increase the TofMult cut for the contaminated event definition from 350 -> 500, this corresponds to roughly 10% increase in the selected centrality.
  - increase the protection time window from 400 time bins to 425 time bins.
  - only increase the desired L0 rate at the beginning of this store a bit, and slow increase the rate when the luminosity drops. (here my proposal is a linear change for the L0 rate from 650 at 50KHz and 800 at 40KHz)

When I include all these, the expected L0 and tape rate are shown in the following. The two dotted lines in each group are the expected L0/tape rates in the current and proposed protection configurations.

The tape rate goal is based on 72 DAQ hours per week and 60% of recorded events are "good" (|Vz|<5cm).

Jeff made the modifications to the protection setup according to the above (TofMult>500, time bins 425), prescales adjusted a bit too. Here is the observed rates from the store on March 27 from 1:30pm to 11:30pm (10 hours!!!).

Near the end of the store, the recording rate is close to the target goal - 500 Hz. On average, we are probably 10-20% short from the goal assuming the same uptime!

===================
Update on March 28
===================
Observed rates from the two stores on March 28.

Fill number 18133	Fill number 18135

===================
Update on March 29/30, April 3
===================
Observed rates from the stores on March 29/30. In the second store, we changed the strategy. In view of the very high instantaneous luminosity in the beginning of the store, HFT detectors and triggers will be only turned on when the ZDCx < 50kHz.

Fill number 18136	Fill number 18142

Fill number 18145	Fill number 18151

Fill number 18159	Fill number 18161

Fill number 18166	Fill number 18168

Fill number 18169	Fill number 18170
Fill number 18171	Fill number 18172
Fill number 18174	Fill number 18177

Fill number 18178	Fill number 18180

Updated on April 22

Fill number 18231	Fill number 18232