80 likes | 175 Views
Rawhits Status A short update on only pattgen “sector mode”. Kohei Yorita University of Chicago March 13 th 2007 @ FTK Meeting. Some Conventions. > Physical Layers are different from logical layers : So in order to avoid confusions, followings are defined :
E N D
Rawhits StatusA short update ononly pattgen “sector mode” Kohei Yorita University of Chicago March 13th 2007 @ FTK Meeting
Some Conventions > Physical Layers are different from logical layers : So in order to avoid confusions, followings are defined : PIX(1,1,1) + SCT(11,11,11,11) -> “,” separates physics layers. -> Each SCT layer has two sides (r-phi z) e.g. PIX(1,1,1) + SCT (00,10,10,10) corresponds to 3 Pixel layers + 3 SCT r-phi layers (in 2nd, 3rd, 4th). > In total, there are 7 physical layers & 11 logical layers. -> Maximum 14 coordinates are available. > With SP, configuration being used is PIX(1,1,1) + SCT(0,1,1,1) -> 6 physical layers with 12 coordinates. -> This corresponds roughly to PIX(1,1,1) + SCT(00,11,11,11) in rawhits format.
Reminder : Rawhit quality In Pixel : ~40% of tracks have two hits in one layer. PIX(1,1,1) kills too many tracks! -> ~100% in the same module and next to each other in pads -> So just use #module for sector finding. (ignore hits if it is near enough) In SCT : ~88% have two hits in one “physical layer” -> Require one for stereo, one for r-phi and no more hits.
Modifications(to make sure if my understanding is correct) For now only pattgen with sector mode is considered : struct atlhit{ isPixel barrel_ec layer_disk phi_module eta_module pi_side ei_strip x,y,z } -> hit by hit struct atltrack{ atlhit(nhits) + MC param + nhits } -> ev-by-ev struct hit{ sector plane type data e.t.c. } -> ev-by-ev • *Added “RAW_MODE 1” in config file so that we can change • SP and rawhits very easily (though switch not yet fully completed..) • Modified codes (for rawhits handling) : • > th_rdStep/readHits : return atltrack/atlhit. • -> Then the atltrack goes to “find_patt_atltrack” routine. > In find_patt_atltrack in pattgen.c file, -> atltrackToHitsM(atltrack,&nhits) : • -> make hitarrayList using “atlhitToHitM”. • * remove another hit in a pixel layer here. • -> each atlhit converted to “hit” structure. • -> filterHitsSec : Remove noisy tracks. (unchanged) > Now we have “sector”. • Along these changes, lots of other files were modified. • e.g. rmap, pmap, pzoffmap files and their input routines. e.t.c.
all_layers_12.pmap file(specifies which layers to be used) SP version (old) Rawhit version (new !) Ex) NUMBER_OF_PLANES = 6 PIX(1,1,1) + SCT (0,1,1,1) mode Ex) NUMBER_OF_PLANES = 9 PIX(1,1,1) + SCT (00,11,11,11) mode 3 pixel 8 SCT Pixel 0 plane 0 phi 22 eta 13 Pixel 1 plane 1 phi 38 eta 13 Pixel 2 plane 2 phi 52 eta 13 SCT 0 stereo 0 plane -1 phi 32 eta 13 SCT 1 stereo 1 plane -1 phi 32 eta 13 SCT 2 sterep 0 plane 3 phi 40 eta 13 SCT 3 sterep 1 plane 4 phi 40 eta 13 SCT 4 stereo 0 plane 5 phi 48 eta 13 SCT 5 stereo 1 plane 6 phi 48 eta 13 SCT 6 stereo 0 plane 7 phi 56 eta 13 SCT 7 stereo 1 plane 8 phi 56 eta 13 3 pixel 4 SCT Pixel 0 plane 0 phi 22 eta 13 Pixel 1 plane 1 phi 38 eta 13 Pixel 2 plane 2 phi 52 eta 13 SCT 0 plane -1 phi 32 eta 13 SCT 1 plane 3 phi 40 eta 13 SCT 2 plane 4 phi 48 eta 13 SCT 3 plane 5 phi 56 eta 13 So it is very easy to turn on/off specific logical layer. e.g. drop all stereo layers e.t.c. * plane –1 : not active.
Comparisons > Ran modified pattgen with “sector mode” to compare the following configurations. > SP format (old style, SP input as a reference) : 1. PIX(1,1,1) + SCT(0,1,1,1) : current configuration -> 6 physical planes 2. PIX(1,1,1) + SCT(1,1,1,1) : Max info. with SP -> 7 physical planes > Rawhit format (new pattgen, Rawhit input) : 3. PIX(1,1,1) + SCT(00,10,10,10) : Minimum option (only phi) -> 6 logical planes 4. PIX(1,1,1) + SCT(00,11,11,11) : More less the same as SP1 -> 9 logical planes 5. PIX(1,1,1) + SCT(11,11,11,11) : Maximum option -> 11 logical planes Any others options to be investigated ?
Finding sectors Looks like efficiency was recovered in rawhits. Mainly due to - Allow two hits in a pixel layer. if they are near. - Treat side0/side1 as a different logical layer. Comparison would be nice to be done - 1 vs 4 (current info.) - 2 vs 5 (full info.)
Summary & Plan > Modified pattgen for rawhits handling. > Compared sector finding with some possible configurations. > Observed High efficiency for sector with rawhits. ~ 50% increase with the same information. > Have to make sure there is no bugs in it. It always takes long(er) time (than expected) to understand someone’s code especially written by > 1 people …. But now I have a feeling that I almost-fully understood ! > After making sure everything is OK, I will move to corrgen. -> generate constants with sectors generated with rawhits. > Superbin definition for patterns e.t.c. > Will make it more speedy.