Some info I collect from listening to people presenting trigger work.
Main Flow Line:
L1
Collision rate is 40MHz(25ns bunch distance). After L1 trigger, rate is reduced to 100kHz, which is maximum L1 rate.
It's the first fast algorithms with lower precision to reduce rate
HLT
1KHz is the limitation of HLT output. HLT uses result with L1, and make uses of full topology of detector, to give better precision.
It's high-precision reconstruction with near-offline performance
TDaq
After HLT trigger passes, TDaq will assemble all information from sub detector/trigger system to form an event. See this for detail:
DAQ
Terminologies
Zero bias
trigger on random (filled) bunches
Minimum bias
trigger on minimum detector activity (e.g. hits in specialized small scintillation counters)
Prescale
Not all triggers can or need to run at full rate.
- Rate might be just too high.
- A subsample of data is enough.
- Adding triggers when luminosity drops makes optimal use of resources. Unbiased reduction of the rate of a trigger by randomly rejecting it
- Controlled by prescale value:
- E.g. PS(L1_2EM3) = 1000 means 1 in 1000 events with at least 2 EM3 objects this L1_2EM3 is causing an L1-accept.
Muon
BC0 Fractions
BC0 means Bunch Crossing Zero. It is the bunch crossing that we are studying. There's also bc 1/-1, etc, meaning next/previous bc.
The bc0 fraction of hits are the fraction of hits that is ascribed to bc0 when it's actually coming from bc0. If muon RPC timing has some issue, larger fraction of hits might be ascribed to bc 1/-1. So it's a quality metric for trigger timing. The numerator is easy to get, just by counting hits that are triggered. Denominator has to come from other parts, like calorimeter timing.
--
RongkunWang - 2017-09-05