Useful Things To Know about The Data
- After integration with
the space craft:
-
Starting from December 8, 2006 (EM p29em1)
there is a problem with the time tones. They do not seem to be properly
stored. This affects the event time stamps. More details in JIRA
ROOT-26. This seems to only have affected this specific release.
- Starting from
when we integrated the LAT with the space craft there is a large
increase in the number of events with a TKR trigger and <6 TKR hits.
This behavior is seen when we are not staying powered on for a long time
(< 1 shift at a time). After integration we have only been on for one
shift per day. When running continously for longer times the efficiency
goes back to normal.
- More
continuous running started with the CPT on March 24 (around run number
077014293).
- There are
still problems with rollovers in the time tones. One consequence is that
time sometimes goes backwards. This is supposed to be fixed in FSW
B0-10-0 which as of June 12, 2007 still hasn't been used on the LAT.
-
During EMI there are numerous GPS
dropout ('lock loss'). Note that the meaning of the GPS lock loss bit in
the data is currently flipped.
- Spectrum Astro (SASS):
- Gamma filter CAL pedestals and
gains: Before run 077010608 it used ideal calibration constants.
Starting from this run it uses realistic ones. The consequence is that
the rate drops from 49 Hz to 25 Hz. Most of the removed events seem to
have been low energy events (see for example CalRecon energy
distribution in the SVAC Recon reports).
-
The MIP filter runs are currently
configured to preferentially select off-axis events (see track direction
distribution in the SVAC Recon reports).
-
TKR FIFO full: Starting from around run
077011026 we changed the TKR FIFO settings to allow for a maximum of 14
hits per layer side. We also reserve space for 126 hits in the GTCC
buffer so that we will now never see any TKR FIFO full errors. One
consequence of this is that we now have a handfull of phasing errors per
30 minute run.
- TKR Bay 10, layer 0, left side
problem: Runs between 07011690 and 077012433 suffer from errors and
stale data from the left side of layer 0 in TKR bay 10. Then, by
mistake, we reverted back to the original split starting with run
077012749. As of 11.21.2006 we are still using this split and get stale
data. One exception: as of 11.17.2006 the following run types have
the correct (all right) split: 22x and 71x muon runs and the 20x MIP
passthru filter runs. Gammafilter runs (both passthru and not passthru)
and the MIP filter runs still have the original split.
- LAT LICOS/FSW
70x and 71x Muon runs NRL/TVAC:
- Bug!
All runs except the 71x runs have a wrong high energy diode CAL
calibration. This means they have no xtals with energy >1 GeV (the exact
threshold varies somewhat). This is fixed starting with run 077008316.
-
LAT LICOS/FSW 70x and
71x Muon runs April 2006:
- Bug!
All runs except the 71x runs have a wrong high energy diode CAL
calibration. This mans they have no xtals with energy >1 GeV (the exact
threshold varies somewhat). All LAT 70x runs in the April runs list have been
reprocessed with correct calibrations. Unfortunately these new
calibrations had two problems. Seven offline(!) LAC thresholds were put
to zero. I have verified that the effect of this is negligeable. The
offline CAL LO and CAL HI thresholds used by TriggerAlg (we run
TriggerAlg as a monitoring tool for real data) were too low. This means
that the two Cal entries in the GltWord are wrong for data. In any case,
you should use the GEM information for real data.
- Periodic trigger: No periodic
trigger before run 77003051, 10 Hz for runs 77003051-77003064, 2 Hz
for runs 77003165-77003571, 10 Hz again for runs 77003573-present.
-
ACD Veto threshold: 0.1 MIPs before run
77003572, 0.3 MIPs after that. At the same time we also changed from
GSFC pedestals to pedestals found with the ACD on the LAT.
-
Both ACD VETO LVDS drivers enabled until
April 20.
- CNO threshold: 20 MIPs
-
LAT 71x and 4-range readout: Because of a
configuration bug, the runs 77002457-77002465 and 77002497-77002502 do
not have 4-range readout of the TKR&&ROI trigger engine. The bug
was fixed for runs after 77003572.
- LAT 71x runs: After run 77003572
there is a factor three more CAL HI triggers in bay 2 than in the other
bays. Known effect.
- Lights on/off: ON in runs 77003656 and
77003657. OFF in runs 77003658-77003665. The runs are normal LAT 70x
runs.
- Light on/off: ON in
runs 77004796-77004798. OFF in runs 77004799-77004801. The runs have a
special configuration with a 100 Hz periodic trigger with unsuppressed,
1-range readout. Even though the other triggers are enabled, only the
periodic trigger is allowed to open the trigger window.
-
Gem sequence counter problem: In several
runs there seems to be a block of events where the GEM sequence counter
(the 'Event ID') is wrong. Looking at the elapsed time (and also the GEM
discarded counter) it looks like the events should have been earlier in
the sequence than they actually are. This will affect people doing
deadtime studies as it will appear that time goes backwards for some
events. More details in JIRA COM-54.
- Other interesting JIRAs under
investigation: COM-9, COM-51, COM-56, COM-57 and COM-58
-
-
- LAT SVAC muon runs
January 13-16, 2006:
- B30 runs: Same as B10 except CAL LO
and CAL HI are not allowed to open the trigger window.
-
B30 runs: Residual retriggering from Bay 2.
To remove them, require no CAL HI trigger in Bay 2. Use for example SVAC
ntuple variable: GemCalHeVector[2]==0. This removed about 0.08% of the
total event sample.
- Six
xtals are always read out because of too low LAC (zero-suppression)
thresholds. They are (Bay,Layer,Column): (0,1,8), (2,0,3), (3,3,5),
(8,0,5), (10,1,5) and (11,5,10). They at least sometimes make it through
the energy cut in CalRecon. These xtals do not cause extra triggers!
-
CNO is allowed to open the trigger
window. The threshold was set to
7.8 20
30 MIPs. - ROI was not allowed to open the trigger
window. It had one Region Of Interest which included all the ACD tiles.
It worked as a trigger i.e. if any tile had a signal above the Veto
threshold the ROI bit in the GEM word was set. The Veto threshold was
set to 0.2 MIP, but was actually more like 0.1 MIP.
-
There was a 10 Hz periodic trigger that was
allowed to open the trigger window.
-
The Merit variable 'EvtRun' is not correct
for runs in the range we have in the data and should not be used. It is
correct for the DC2 runs though.
-
The ACD tile in row 3, column 4 (Tile ID 34) has the wrong thickness: 12
mm instead of 10 mm.
-
Six towers:
- We have now enabled a 10 Hz
periodic trigger that is allowed to open the trigger window.
-
Bug in a CAL calibration files
concerning all 4 and 6 tower data: The high energy range is not correct.
Both for runs with flight and the high energy muon gain this corresponds
to above 1 GeV in a single xtal.
-
TKR TOT was not processed with all the
latest fixes (charge scale is a little off).
- Four towers:
- Bug in a CAL calibration files
concerning all 4 and 6 tower data: Only the lowest energy range is
correct.
- We still have the Event ID
initialization bug.
- SVAC
ntuple 'TkrNumClusters' has a bug and should not be used until further
notice!
- The
Cal Low trigger thresholds for the CALs in bay positions 1 and 5 were
set to one fixed value - a little on the high side.
-
The LAC thresholds for the CALs in bay
positions 1 and 5 were not correct (too low) so we have lots of noise
hits - about 67 and 44 xtals read out per event respectively. Very few
of these make it into recon which has an energy cut - only 1 per event
for all the towers combined.
-
We are now using the
1-shot TKR trigger.
-
Most of the TKR trigger primitives are missing - only about 10% are there.
This is probably due to timing changes introduced by the 1-shot TKR
trigger. Note that the CAL trigger primitives should not be affected by
this, but of course: most of the triggers are TKR triggers.
-
The TKR Hot
strips calibration file had a bug. All trackers except FM A and FM B
were marked as dead. This did not affect the data digis. However, any
missing hit in TkrRecon was flagged as a hot strip.
-
TKR TOT was not processed with all the
latest fixes (charge scale is a little off).
- Two towers:
- Event ID is now 32 bits i.e. no more
roll overs (multiple events with the same Event ID).
-
Initialization bug in online means the Event ID
sometimes starts at 131073.
- Trigger window is now 12 ticks (1 tick is 50 ns) instead of 5 ticks which we
used for the 1 tower runs.
- New
SVAC variable available: EvtTick. Can be used to find the time between any
two events. Does not saturate like the GemDeltaEventTime.
-
SVAC ntuple 'TkrNumClusters' has a bug and
should not be used until further notice!
-
To find out which tower triggered, use
GemTkrVector[iTower], GemCalLeVector[iTower] etc.
-
TKR TOT was not processed with all the
latest fixes (charge scale is a little off).