Bug #1807
Tracker reconstruction efficiency and resolution poor
Added by Rogers, Chris over 7 years ago.
Updated about 5 years ago.
Start date:
05 January 2016
Description
General issues with tracker reconstruction efficiency and resolution.
Files
|
bfield_vs_z.png (7 KB)
bfield_vs_z.png |
|
Rogers, Chris, 05 January 2016 17:12
|
|
|
transmission_vs_z.png (6.96 KB)
transmission_vs_z.png |
|
Rogers, Chris, 05 January 2016 17:12
|
|
|
emit_vs_z.png (8.04 KB)
emit_vs_z.png |
|
Rogers, Chris, 05 January 2016 17:12
|
|
|
baseline_recon.tar.gz (54.8 MB)
baseline_recon.tar.gz |
|
Rogers, Chris, 05 January 2016 17:12
|
|
|
transmission_vs_z.png (6.85 KB)
transmission_vs_z.png |
|
Rogers, Chris, 06 January 2016 12:14
|
|
attachment:bfield_vz_z.png!
Bfield vs z - this is a solenoid mode
attachment:transmission_vs_z.png!
Blue line shows the number of tracks at each virtual plane vs z; circles shows the number of reconstructed tracks at each tracker station vs z.
attachment:emit_vs_z.png!
Circles show reconstructed emittance; line shows MC emittance. This is for the same set of particles (i.e. only particles that were reconstructed).
I attach analysis scripts, geometry, etc in the tarball so you can see the details of what I did (and berate me if I made a mistake).
Thanks Chris, I have had a first look. TkUS seems reasonable, but TkDS is dropping spacepoints and tracks far too regularly. I wonder if this might not stem from the recent modifications to fix real data - we know this introduced an issue into the MC which we are still working to fix.
Thanks for looking. I realise TKD looks much worse, but TKU is way out of spec also.
Should I try running with an older geometry/MAUS version? It would be great to get this into the optics review paper (or more like, it would be crap if we can't get it in).
Running with 1.3.0 (the current release branch) should determined whether it is the new tracker modifications or not, so if you could re-run with that, that would be great...
I ran over 100 events, the efficiency looks worse: 50 % u/s and 15 % d/s. I will run with 1000 events and see what happens.
Indeed, running over 1000 events in 1.3.0 also yields terrible efficiency, if anything worse than 1.3.2. The upstream resolution looks good, but the downstream resolution is still poor...
Interesting, Chris Hunt, is this consistent or at odds with your MC studies?
Just ran my efficiency script over the output, got the following:
Total recon events: 1000
10 spacepoint events: 883
TkUS 5 spoint events: 928
TkDS 5 spoint events: 902
10 spoint PR tracks: 224
TkUS 5 spoint PR tracks: 850
TkDS 5 spoint PR tracks: 243
TkUS 3 to 5 spoint PR tracks: 921
TkDS 3 to 5 spoint PR tracks: 809
Verfies that TkDS is going awry. Will investigate further...
An idea, the geometry has been updated to fix the tracker flips - that could be causing the problem even with an old version of maus. Chris R, could you try it with the previous set of geometries please, id 72 I think? Still with 1.3.0. Thanks...
The fiber positions have changed. I switched the fiber x positions, also the radius of the tracker planes was increased slightly which had the effect of lengthening the fibers, finally, the x position of the fibers were shifted a little to make them symmetric about the center fiber. These changes haven't been extensively tested, obviously, but they have pasted all my tests so far.
The position of the planes (in Z) needed to be flipped in upstream (?) tracker as well, but this impact should be minimal.
I ran ID 72 and it is still not happy... is this expected? Maybe you should run an independent simulation to check that I didn't make a typo somewhere?
- Status changed from Open to Closed
- Assignee changed from Dobbs, Adam to Franchini, Paolo
- % Done changed from 0 to 100
I think this can be closed.
Also available in: Atom
PDF
