triggeralgs

A place to put the trigger algorithms for DUNE. This code can then be linked inside ArtDAQ and LArSoft, so that we can run the exact same code in simulation and on real data.

TOC

• Future
• Organisation
• Compile
• Contribute
• Data structures

Future interesting things to do

• Improve documentation!!
• Make the whole setup more realistic
  • TPs from PDFs of what we expect from LArSoft/simulations
  • TAs algorithm better
• Find a way to estimate the efficiency for TAs and TCs.
• Implementation of "TP window" and "TP zipper":
  • TP window = something that ensures that the TPs are all in a time window,
  • TP zipper = something that merges source of TPs.

Organisation

The code is quite simple. Essentially, the trigger algorithm rely on 3 structs:

• TriggerPrimitive (also known as hits)
• TriggerActivity (clusters)
• TriggerCandidate (trigger)
• TriggerDecision (readout request ?)

Their format will be discussed at a DAQ meeting and fixed at some point.

The trigger primitives are created within ArtDAQ or LArSoft. These are passed to one or more TriggerActivityMaker classes which create the TriggerActivity. The TriggerActivity are then passed on to one or more TriggerCandidateMaker classes, which create the candidates, which in turn will be fed to TriggerDecisionMaker, etc. Each rely on their pure virtual operator:

• void TriggerActivityMaker ::operator()(const TriggerPrimitive& input_tp, std::vector<TriggerActivity >& output_ta)
• void TriggerCandidateMaker::operator()(const TriggerActivity& input_ta , std::vector<TriggerCandidate>& output_tc)
• void TriggerDecisionMaker ::operator()(const TriggerCandidate& input_tc, std::vector<TriggerDecision >& output_td)

that any TriggerActivityMaker, TriggerCandidateMaker and TriggerDecisionMaker need to implement, respectively.

Note the TPs can also be created here, but given how this happens now in real life, it doesn't look like these libraries will be used for creating TPs (the data structures for the raw data are very complicated and closely tied to ArtDAQ, that is not so much the case in LArSoft), hence the TriggerPrimitiveMaker class exists with its corresponding operator:

• void TriggerPrimitiveMaker::operator()(const void* data, std::vector<TriggerPrimitive>& tps)

but that may not necessarily be used.

These operators do all the work, they should be reasonably fast to handle the rate at which their input arrive in the real system. The "makers" get input data, rearrange it, and then push_back to the output vector. An example is available in src/trivial/TriggerCandidateMaker_trivial.hh. It can be tested with a exec/run_trivial_candidate.cxx, which creates fake TPs, feeds them into it and then dumps them in a csv.

Note none the granularity of the TriggerActivityMaker, TriggerCandidateMaker and TriggerDecisionMaker isn't be decided here. It is the ArtDAQ's developers' job to decided how many TriggerCandidateMakers there should be in one APA for example. The focus of this library is the algorithm, ArtDAQ should create and handle the DuneTriggers objects and the calls to them (this is realised in triggermodules)

To compile it

You need to have the boost libraries for testing (this dependency will be dropped soon, we don't need it)

git clone https://github.com/DUNE-DAQ/triggeralgs.git
cd triggeralgs
mkdir build
cd build
export TRIGGERALGS_DIR=$(pwd)/../install # this is for triggermodules to know where this install is.
cmake -DCMAKE_INSTALL_PREFIX=${TRIGGERALGS_DIR} ../
make -j`nproc`
make install # to move all the insteresting stuff in ${TRIGGERALGS_DIR}
ctest # if you want to run the tests, right now it doesn't do anything interesting

To contribute

Fork this repo.

More details on the structs

Note all the times here refer to 50 MHz clock (since epoch), as is now standard in DUNE.

TriggerPrimitive

This struct represent hits, it contains:

Variable	Type	Comment
time_start	int64_t	Start time
time_peak	int64_t	Time of the peak ADC
time_over_threshold	int32_t	Time over threshold
channel	uint32_t	Channel number
adc_integral	uint16_t	ADC integral
adc_peak	uint16_t	ADC peak (could be 12 bits, strictly)
detid	uint32_t	detid (a flag that represents the detector which created the TP)
type	uint32_t	some flag that says what it think it is (PDS/TPC for example)
algorithm	uint32_t	some flag that says which algorithm created it (CPU/Firmware for example)
version	uint16_t	version of above
flag	uint32_t	extra flags.

TriggerActivity

Represents a cluster of hits, it contains:

Variable	Type	Comment
time_start	int64_t	Start time
time_end	int64_t	End time
time_peak	int64_t	Time where ADC the highest
time_formed	int64_t	Time at which the object was created (maybe we don't need that?)
channel_start	uint32_t	Start channel
channel_end	uint32_t	End channel
channel_peak	uint32_t	Channel of the highest ADC
adc_integral	uint16_t	ADC integral
adc_peak	uint16_t	ADC peak (could be 12 bits, strictly)
detid	uint32_t	detid (a flag that represents the detector in which it ocurred)
type	uint32_t	some flag that says what it think it is (Argon39/Muon for example)
algorithm	uint32_t	some flag that says which algorithm created it (PDS/TPS/SN/high energy/solar for example)
version	uint16_t	version of above
tp_list	std::vector<TriggerPrimitive>	the list of TPs that was used to create it

TriggerCandidate

contains a trigger "request" (i.e. please trigger). In this case, it's based on activity in the detector, but it could be external trigger as well, if one removes the ta_list:

Variable	Type	Comment
time_start	int64_t	Time at which to start the readout
time_end	int64_t	Time at which to end the readout
time_decided	int64_t	Time at which this decision was taken
detid	uint32_t	some flag that represents which part of the detector to readout
type	uint32_t	some flag that says what it think it is (SN/Muon/Beam for example)
algorithm	uint32_t	some flag that says which algorithm created it (SN/high energy/solar for example)
version	uint16_t	version of above
ta_list	std::vector<TriggerActivity>	the list of TAs that was used to create it

TriggerDecision

contains a readout request (i.e. you MUST trigger), based on all the candidates in the detector:

Variable	Type	Comment
time_start	int64_t	Time at which to start the readout
time_end	int64_t	Time at which to end the readout
time_triggered	int64_t	Time at which this decision was taken
detid	uint32_t	some flag that represents which part of the detector to readout
type	uint32_t	some flag that says what it think it is (SN/Muon/Beam for example)
algorithm	uint32_t	some flag that says which algorithm created it (although I think there will only ever be one of this)
version	uint16_t	version of above
tc_list	std::vector<TriggerCandidate>	the list of TCs that was used to create it