DØ Level 3
Algorithms Meeting: 1st May 2002 in the Farside
Talk
-
Ntuple-based monitoring for L3
online: Michiel Sanders.
Michiel discussed the work he has been doing to enable reco_analyze to
be run online as an "examine" and how this might be applied to running
l3fanalyze in a similar way.
The basic idea would be for the roottuple produced by l3fanalyze to be
read in a continuously updating fashion by a root macro that fills histograms
defined by the L3 group.
Currently xgooey would be used to view these histograms.
Michiel has agreed to make the necessary modifications to l3fanalyze
and each tool author should provide definitions for a few histograms
to go into the root macro.
It seems like a good idea for us to run d0_analyze with reco_analyze
and the L1/L2 analyze packages switched off initially.
(Once we get something running then the L1/L2 analyze packages
could be switched on and L1/L2 monitoring
could be easily available via this route.)
Note added Thursday 2nd May: Yann Coadou and Michiel sat down in
front of a terminal and
figured out exactly what changes Michiel needs to make and where
to get this all
to work.
Discussion Items
-
Getting L3 tracking into the trigger list.
Daniel Whiteson and Terry Wyatt sat down with Elizabeth today and made
global_CalMuon6.10. On each of the four single muon triggers at L1/L2 this
has a script with 100% mark&pass fraction and
a standalone track filter requiring
at least one L3 global track with p_t > 3.0 GeV.
N.B. Each single muon triggers at L1/L2 also has a script
with 100% mark&pass fraction and a loose muon filter.
Daniel and Robert will run some tests and then let everyone where to
get hold of the appropriate level3.sim files.
The hope is to get L3 muon and tracking running online early next week.
-
Need for greater rejection at L3.
With luminosity now at around 1.6 * 10**31 at the beginning of the
fill, several triggers have acquired L1 prescales because the rate to
tape is too high. This is clearly inefficient and undesirable.
In particular, some of our favorite electron triggers CEM(1,10) and
CEM(2,5) have prescales of 2 at 1.6 * 10**31.
We should definitely tighten the shower shape cuts to those we
proposed originally.
Volker and Ulla will measure the additional rejection that could be
obtained by increasing the p_t cuts for the two electron filters
(with and without shower shape cuts) that we currently have hanging
off of CEM(1,10) and CEM(2,5).
Another possibility to sharpen the turn on curve is to make use of
primary vertex information....
-
.... Next steps for track-based primary vertex tool.
It would be good to see the results of the primary vertex tool if the
cut on the number of stereo hits were reduced from 8 to, say, 3, in
order to pick up the tracks that have stereo information only in SMT.
In addition, results on the most recent possible data, which has fully
instrumented CFT stereo, would be interesting.
-
Inconsistencies between on and offline results for tau tool: Yann Coadou.
Yann finds slight differences in the properties of taus found online
and when running the simulator offline. In rare cases this causes the
number of candidate taus to be different on and offline.
Offline, he gets exactly identical results when he runs on d0mino and
clued0.
Next step seems to be to use Gustaaf's technique of running in
"online" mode on an offline machine.
-
Problems with event_info: Yann Coadou.
Yann has found that when running on real data event_info delivers the
online trigger results even when the simulator is run!
Solution 1: event_info needs a switch to tell it whether to deliver
the online or simulator results.
Prefered Solution: Both event_info and l3fanalyze should provide both
online and simulator results to enable event-by-event comparisons
using one roottuple.
-
We urgently need someone to make the information on tool timing
collected by the L3 monitor program visible in the control room.
Scribe: Terry Wyatt.