Introduction to FSW Event Filtering

Introduction to FSW Event Filtering

Overview

The LAT's downlink data rate allows only a small fraction (2-10 percent) of detected events to be reported. Consequently, the vast majority of events must be discarded. Fortunately, this can be done without damage to the scientific objectives of the instrument.

The LAT will "trigger" (i.e., detect events) on a variety of phenomena, including both high-energy photons (i.e., Gamma rays) and cosmic background radiation (e.g., protons, electrons, and positrons). In fact, cosmic rays and other "noise" account for 99.8% of the detected events. The event filtering software attempts to discard this noise, using a series of tests. As discussed below, the software's goal is to preserve analyzable gamma-ray data by identifying and discarding cosmic rays (a relatively easy and fast process), then identifying and retaining analyzable gamma rays (harder).

In order to clarify the physical basis for the tests, we will use a simplified diagram of the LAT instrument. The following diagram depicts a (somewhat idealized) event, where a Gamma ray is entering the top of the instrument. It passes through the Anti-coincidence detector (ACD) tiles without interaction, but is detected by the Tracker (TKR) and Calorimeter (CAL).

G G ---------------------- | ==================== | T | ==================== | K | ======EP============ | R | ======E=P=========== | | ======E==P========== | | ======E===P========= | | ======E====P======== | ......*.....*....... C .....***....***..... A ....*****..*****.... L .....***.....***....

The ACD is a set of plastic scintillator "tiles" ('-' and '|' in diagram) which surround the TKR. The ACD reports which tile(s) have responded to a cosmic ray.

The TKR is a set of 16 towers, arranged in a 4x4 grid. A tower contains 18 detection layers ('=' in diagram), each consisting of a Tungsten "radiator" sheet and an (X, Y) pair of "Silicon-strip Detectors" (SSDs).
When a Gamma ray (G) passes through a radiator, it may produce an electron-positron pair (E, P). Each SSD reports the (X or Y) locations at which it detects charged particles. By analyzing the data from multiple SSDs, the software can "project" an (X,Y,Z) path for each charged particle and thereby, the Gamma ray.

The CAL is a set of four detection layers, each consisting of an (X, Y) pair of scintillator arrays. Each array ('.' in diagram) is a set of instrumented Cesium-Iodide crystals, which report the event's energy level. By analyzing the data from multiple crystals, the software can "project" an (X,Y,Z) energy distribution.

Tests

At the highest expected data rate (10K events per second), the LAT computers have only 100 microseconds (on average) to handle each event. Consequently, "cheaper" tests are made before "expensive" ones.

Trigger discriminator tests

These tests, which check whether the data is consistent with the presence of cosmic rays, are very fast and quite effective (eliminating 75% of the candidate events). They use data from the GEM (Global Trigger), including anti-coincidence hit maps, which tracker towers had "3-in-a-row" triggers, and trigger information from the CAL (low- and high-energy) discriminators.

Veto tile response; no CAL LO

P P ---------------*------ | ===============P==== | T | ================P=== | K | =================P== | R | ==================P= | | ===================P | | ==================== | | ==================== | .................... C .................... A .................... L ....................

If any ACD tile near the top of the instrument ("veto tile") responds to the event and the CAL low-energy (CAL LO) trigger discriminators all read zero, the event is rejected (64.4%).
This test targets charged track events (e.g., incoming electrons or positrons) which miss the CAL, but produce a tower tracker "3-in-a-row" trigger.
Note: At energies which are too low to activate the CAL LO discriminators, Gamma-ray interactions in the CAL will not produce backsplash in the ACD veto tiles. Thus, this must be an incoming cosmic ray.

ACD Splash Veto (pass 0)

-------*-------*------ * ======E======P====== | T | P=====E=====P======= | K | =P====E====P======== | R | ==P===E===P========= | | ===P==E==P========== | | ====P=E=P=========== | | =====PEP============ | ......*............. C .................... A .................... L .................... G G

If the ACD gets too many hits and the CAL high-energy (CAL HI) trigger discriminators all read zero, the event is rejected (10.2%).
This test targets events which enter from the bottom of the detector and interact in the CAL.

Energy-level tests

These tests requires significant compute time (~10x that of the previous set). They use energy information from the CAL, which must be unpacked.

Veto tile response; CAL energy < 350 MeV

Any ACD tile response; CAL energy < 10 MeV

P P ----*----------------- | ==================== | T | ==================== | K | ==================== | R | ==================== | | ==================== | | ==================== | | ==================== | .................... C .................... A .................... L ....................

If any ACD tile responds to the event and the CAL detects energy of less than 10 MeV, the event is rejected (3.0%).
This test targets cosmic-ray events and events where there is no potential for backsplash from the CAL to hit an ACD tile.
Note: This is a much more inclusive version of the initial test ("Veto tile response; no CAL LO"). At 10 MeV, no backslash can reach even the lowest ACD tiles, so all tiles can be included in the test.

ACD Splash Veto (pass 1)

-------*-------*------ * ======E======P====== | T | P=====E=====P======= | K | =P====E====P======== | R | ==P===E===P========= | | ===P==E==P========== | | ====P=E=P=========== | | =====PEP============ | ......*............. C .................... A .................... L .................... G G

If the CAL detects energy of less than 40 GeV (this is almost always the case) and the ACD gets too many hits, the event is rejected (0.4%).
This test targets events which enter from the bottom of the detector and interact in the CAL.
Note: Because we now know the energy of the event, we can apply the "ACD Splash Veto" test to more events.

TKR tower match with ACD top tile

P P ------*--------------- | =====P============== | T | =====P============== | K | =====P============== | R | =====P============== | | =====P============== | | =====P============== | | =====P============== | .................... C .................... A .................... L ....................

If a responding ACD top tile shadows a tower with the "3-in-a-row" tracker trigger, the event is rejected (2.7%).
This test targets charged particles (e.g., protons) which enter through the ACD. This is a slightly more sophisticated version of the hardware throttle.
Note: This test, coupled with the following test, refines the geometric resolution to a single tracker tower. A follow-on test ("TKR/ACD matching") will refine this to a single ACD tile.

TKR tower match with ACD side tile

P ---------------------- * ==================== | T | P=================== | K | ==P================= | R | ====P=============== | | ======P============= | | ========P=========== | | ==========P========= | .................... C .................... A .................... L ....................

If a responding ACD side tile shadows a tower with the "3-in-a-row" tracker trigger, the event is rejected (2.0%).
This test targets charged particles (e.g., protons) which enter through the ACD. This is a slightly more sophisticated version of the hardware throttle.

No connection between CAL energy and TKR (pass 0)

---------------------- | ===PP=============== | T | ==PP================ | K | =PP================= | R | PP================== | | P=================== | ? | ==================== | ?? | ==================== | ? .................... C .................... A ???? ..*.***.**..*.*.*.** L ???? *..*.*..**.*.**.**..

If there is any CAL energy and no tracker tower has hits in more than four of its bottom six planes, the event is rejected (7.7%).
This is an inexpensive test for bottom- and side-entering showering tracks (e.g., caused by a particle interaction with the spacecraft).

CAL layer 0 energy < 1 percent of total

---------------------- | ==================== | T | ==================== | K | ==================== | R | ==================== | | ==================== | | ==================== | | ==================== | .*.............*.... C ...*......*..*..*... A .*....*..*....*...*. L ...*.*..*..**.*..*..

If the energy detected by layer 0 (the top later) of the CAL is less than 1% of the total energy, the event is rejected (1.0%).
This test looks for behavior that is inconsistent with electromagnetic showering.
Note: Although the diagram only shows a lower number of hits in layer 0, the total energy level is also affected (and is measured by this test).

CAL layer 0 energy > 90 percent of total

---------------------- | ==================== | T | ==================== | K | ==================== | R | ==================== | | ==================== | | ==================== | | ==================== | *..*.*.*..*..*..*..* C .*..*..........*.... A .*.......*........*. L ...........*.....*..

If the energy detected by layer 0 of the CAL is more than 90% of the total energy, the event is rejected (0.6%).
This test looks for behavior that is inconsistent with electromagnetic showering.
Note: Although the diagram only shows a higher number of hits in layer 0, the total energy level is also affected (and is measured by this test).

CAL crystal ratio

---------------------- | ==================== | T | ==================== | K | ==================== | R | ==================== | | ==================== | | ==================== | | ==================== | *....*....*.....*..* C .*..*...*......*.... A .*..*....*....*...*. L .....*.*...*.....*..

If the number of CAL crystals with less than 1% of the energy exceeds 20% of the total number of hit crystals, the event is rejected (0.25%, from recent work).
This test looks for behavior (e.g., a diffuse scattering) that is inconsistent with electromagnetic showering. (This is more characteristic of hadronic showering.)

Track-projection tests

These tests depend on finding X-Z and Y-Z projections, using the fully-unpacked tracker data. They are very compute-intensive, requiring perhaps 50x the processing time of the initial tests. Fortunately, we're down to a small fraction of the initial events, so the aggregate compute time isn't a big issue.

No projections

---------------------- | ==================== | T | ================P=== | K | ==============P===== | R | ================P=== | | ==================== | | ==================== | | ==================== | .................... C .................... A .................... L ....................

If no projections are found, the event is rejected (TBD%).
This test targets two classes of events which are barren of tracks: false "3-in-a-row" triggers (generally caused by noise) and non-tracker (e.g., CAL LO) triggers.

TKR/ACD matching

P P -------------*-------- | ===========/======== | T | ==========/========= | K | =========/========== | R | ========/=========== | | =======/============ | | ======/============= | | =====/============== | .................... C .................... A .................... L ....................

If a track projection points at a responding ACD tile, the event is rejected (1.7%).
This test targets charged particles which enter through the ACD. This is a very sophisticated version of the hardware throttle.
Note: Although the projection has much greater resolution, we can only compare against a single ACD tile. Because of the increased geometric correlation, noise is not a factor (so we can target more events).

Track(s) project into skirt region

---------------------- | ======/============= | T | =====/============== | K | ====/=============== | R | ===/================ | | ==/================= | | =/================== | | /=================== | .................... C P .................... A P .................... L ....................

The "skirt region" lies between the CAL and the lower row of ACD tiles. Thus, it is outside of the area covered by the CAL, but inside the active area of the instrument. If a track projection points into this region, the event is rejected (0.5%).
This test targets charged-particle tracks that escape detection by the ACD and CAL, as well as Gamma rays which miss the the CAL. (The rejected Gamma rays are nearly useless for analysis, because they lack energy information.)

CAL energy < 350 MeV; fewer than two tracks

---------------------- | =\================== | T | ==\================= | K | ===\================ | R | ====\=============== | | =====\============== | | ======\============= | | =======\============ | ........*........... C ........**.......... A ........***......... L .........**.........

If the energy level detected by the CAL is less than 350 MeV and there is no evidence of two or more tracks, the event is rejected (3.0%).
This test targets events which are inconsistent with the topology expected from relatively low-energy Gamma rays. At this energy, one expects to see two separated tracks (from the initial electron-positron pair). At higher energies, the separation between the two tracks is insufficient for the tracker to resolve.

No connection between CAL energy and TKR (pass 1)

---------------------- | ===PP=============== | T | ==PP================ | K | =PP================= | R | PP================== | | P=================== | ? | ==================== | ?? | ==================== | ? .................... C .................... A ???? ..*.***.**..*.*.*.** L ???? *..*.*..**.*.**.**..

If there is any CAL energy and no projection points at that energy, the event is rejected (TBD%).
This is an expensive (but more precise) test for bottom- and side-entering showering tracks (e.g., caused by a particle interaction with the spacecraft).

Albedo Gamma-ray rejection

G ---------------------- G| ==================== | T | ==================== | K | ==================== | R | \=================== | | =\================== | | ==\================= | | ===\================ | .................... C .................... A .................... L ....................

If the track of the candidate event (now assumed to be a Gamma ray) is consistent with pointing at the Earth's limb, the event is rejected (TBD%, out of an absolute rate of 250-750 Hz).
Gamma rays coming from charged-particle interactions in the Earth's atmosphere are irrelevant to LAT's mission. We may allow a certain percentage of these through, for calibration and monitoring purposes.

Note: Except where otherwise indicated, the rejection percentages given above are based on the total count of input events. The values given are based on simulation results, reported at the DOE/NASA Peer Review, March 19-20, 2003.