Difference between revisions of "January 19, 2011 Tracking CDC/FDC"
(→Tracking Issues) |
(→Tracking Issues) |
||
| Line 51: | Line 51: | ||
Dave Lawrence spoke about some issues in the tracking software... | Dave Lawrence spoke about some issues in the tracking software... | ||
# In the reconstruction software certain events would hang or segfault, but not reproducibly, on the IU 32-bit Linux farm. This might be associated with the Dec 7 transition from the LSG fitter to a Kalman fitter, maybe mis-alignment of 32bit machines with simd, the Kalman simd has been disabled... this is ongoing. | # In the reconstruction software certain events would hang or segfault, but not reproducibly, on the IU 32-bit Linux farm. This might be associated with the Dec 7 transition from the LSG fitter to a Kalman fitter, maybe mis-alignment of 32bit machines with simd, the Kalman simd has been disabled... this is ongoing. | ||
| − | |||
# Different mass hypotheses for the same track (eg fit pi+ to both p and pi+) have been found to give a large difference in mass, possibly due to a different hit selection. Simon is looking at activating the Kalman filter to reject further away hits and Dave is looking at changing the CDC hit selection, the sigma calculation looked wrong... this will be ongoing for the next few days. | # Different mass hypotheses for the same track (eg fit pi+ to both p and pi+) have been found to give a large difference in mass, possibly due to a different hit selection. Simon is looking at activating the Kalman filter to reject further away hits and Dave is looking at changing the CDC hit selection, the sigma calculation looked wrong... this will be ongoing for the next few days. | ||
| − | + | # Historically -999 in code has been used to indicate not a valid number, some of these were propagating by division into a sensible number. Dave is now changing these to NAN. So... in future we will need to check if something is NAN before operating on it. | |
| − | # Historically -999 in code has been used to indicate not a valid number, some of these were propagating by division into a sensible number. Dave is now changing these to NAN. | + | |
| − | So... in future we will need to check if something is NAN before operating on it. | + | |
Revision as of 11:44, 20 January 2011
Contents
Meeting Time and Place
Wednesday January 19, 2011 at 11:00am At Jefferson Lab, the meeting will be held in F326
Connections
To connect from the outside, please use ESNET
1.) ESNET: 8542553
2.) Phone:
+1-800-377-8846 : US
+1-888-276-7715 : Canada
+1-302-709-8424 : International
then enter participant code: 77438230# (remember the "#")
Agenda
- Announcements
- The Collaboration Meeting
- FDC Update logbook
- CDC Update [1]
- Tracking Issues
Minutes
Participants: Curtis, Naomi, Beni, Simon, Lubomir, Dave and Mark
Collaboration Meeting
Curtis has scheduled one talk each for the FDC and CDC groups. The next section is Offline Reconstruction which will follow on nicely, there will probably be an overview by Mark and reports on Dmitri's calibration database and tracking status. Beni and Lubomir will let Curtis know in the next few days if they need to make another talk on the FDC.
FDC Update
CDC Update
Curtis reported on the CDC construction status: the remaining straws in row 10 will be put into place today and glued tomorrow. We are completing about one layer each week. An undergrad student is working on cutting more straws to size and others are checking the parts - there are still a few 1000 pinholders to be checked and cleaned.
Installing the straws takes 3 people, usually these are the 2 techs and Gary; Curtis and Naomi have been trained for use if needed.
We have not found any more bad resistance straws - these all came from one section of the box. Initially found 2 straws where where the straw-donut glue-joint made no electrical connection, this has not been seen since then. The resistance of each straw and straw-donut is being checked.
Gary has made some modifications to Slava's test electrical hookup. He found that it can be sealed with less torque than initially specified. He also found that the wires leaked gas (along the mesh) but can sealed with cyanoacrylic over the rubber boot at the pin end.
We plan to train students to prepare the wires - cut, strip, solder ball, seal, add the rubber boot and make sure that the finished wires have low resistance.
Cleanroom... our particulate meter was returned from servicing and indicates that we still have a class 2000 cleanroom. The low humidity has been causing static electricity problems (eg giving people a small shock when they touch the table) so we have been trying to raise humidity and also grounded the cleanroom and the table to the building ground. After installing the straws in the CDC, the ends are sealed with a plastic cap to prevent dust from collecting inside them.
Naomi has been working installing Xilinx software so that she can download software from Gerard into the new fADC to increase the pre-trigger buffer time. She is preparing to send some of our MFCs back to MKS for recalibration.
Tracking Issues
Dave Lawrence spoke about some issues in the tracking software...
- In the reconstruction software certain events would hang or segfault, but not reproducibly, on the IU 32-bit Linux farm. This might be associated with the Dec 7 transition from the LSG fitter to a Kalman fitter, maybe mis-alignment of 32bit machines with simd, the Kalman simd has been disabled... this is ongoing.
- Different mass hypotheses for the same track (eg fit pi+ to both p and pi+) have been found to give a large difference in mass, possibly due to a different hit selection. Simon is looking at activating the Kalman filter to reject further away hits and Dave is looking at changing the CDC hit selection, the sigma calculation looked wrong... this will be ongoing for the next few days.
- Historically -999 in code has been used to indicate not a valid number, some of these were propagating by division into a sensible number. Dave is now changing these to NAN. So... in future we will need to check if something is NAN before operating on it.
The software workshop had a visitor Dmitri from the ATLAS tracking group at RAL. The ATLAS tracking process takes 1ms whereas the estimates for GlueX are around 20ms. Simon made some changes based on this. He found that the stepsize is too small and there are too many iterations/path so he doubled the stepsize and decreased iterations, now down to 10ms with no noticeable effect on resolution. Dave has added hi-res timers. Checking material maps for boundaries adds time. Other ATLAS characteristics are very similar, a 2T field track, lower measurement range of 100MeV, momentum resolution of a few %.
Dave and Simon are going to look at probabilistic fitting, including L and R tracks.
Curtis said that Will found that some errors in the kinematic fitter were possibly too small, he will report on this at the collaboration meeting.
Next meeting
will be in 4 weeks.
