Results for preco03.02 with overlap, smt extension and forward fixes.

Results for preco03.02 with overlap, smt extension and forward fixes

D. Adams
19jan99 1350

Previous results were for preco03.02 with overlap and smt extension fixes.

Release preco03.03 should be equivalent. Here we start with that release but use a newer version of gtr_find (v00-19-54), smttrack (v00-04-22), smt_path (v00-03-18) and smt_detector (v00-04-13) which speed up forward tracking. I have requested this fix be included in preco03.03.

Here are results for single muons and Pythia Z -> mu mu with zero, one and two minbias background events. As usual these are for pT above 0.5 GeV/c.

 run bg  chk   CPU  CLK mem swp tm  pe  pTt # mc   eff  misr  fake  fcsq  mcsq  nclu
---- -- ----  ---- ---- --- --- -- ---  --- ----  ----  ----  ----  ----  ----  ----

m50      401          1  57  79  0   1  0.5 1629  .936  .003  .008  18.4   5.4  17.0
m03      401          1  56  78  0   2  0.5 1609  .931  .023  .006  21.6  10.9  16.6
m01      401          0  56  78  0   0  0.5 1642  .607  .072  .003  24.9  16.7  14.4

pzmm  0  401     9   10  68  91  0   0  0.5  683  .739  .139  .028  19.1  13.3  13.3
pzmm  1  401    58  101 121 145  0   8  0.5 1892  .816  .326  .077  19.0  16.5  11.2
pzmm  2  401   357  663 204 230  0  33  0.5 2938  .901  .451  .135  18.7  17.4  10.2
pzmm  3  401   321 3362 612 637  2 115  0.5

For the Pythia events, here are results for tracks with pT above 10 GeV/c.

                                       10.0  154  .955  .014  .020  19.1   5.8  17.6
                                       10.0  160  .944  .079  .073  19.1   6.4  16.1
                                       10.0  186  .957  .185  .112  18.3   5.4  15.8
                                       10.0

The columns in these tables are defined here.

Here are plots showing the found (top) and lost (bottom) tracks in the z0-tan(lambda) plane.

The figures include curves showing the subdetector boundaries. The central horizontal band is the CFT acceptance. The thin bands above and below are the overlap region. The regions above the top V and below the bottom V are the SMT forward acceptance. There are also regions in the upper left and lower right between the CFT overlap and the SMT forward which are not covered. These are the "gaps".

m50
m03
m01
pzmm + 0 bg  low pT  high pT
pzmm + 1 bg  low pT  high pT
pzmm + 2 bg  low pT  high pT

Here is the clock time (sec/event) broken down by track-finding step as reported by the framework (I do not believe the CPU times):

 run bg   cc   co    c   cs    f  csf
 --- --  ---  ---  ---  ---  ---  ---
pzmm  0    1    4    1    1    2    0
pzmm  1    5   66    3   11   12    3
pzmm  2   14  522   13   48   44   11
pzmm  3   28 2909   60  207  116   44

There is a problem here that the times for the last entry do not add up to the total observed time. I suspect a problem in the framework.

Comments

Changes since the last posted results:

There has been a small decrease in efficiency.
There are small decreases in the chi-squares.
The forward tracking CPU time is much less.

The most serious remaining problems:

The misreconstructed fraction and the fakes increase significantly as backgound events are added. (This may be because the central CFT efficiency drops and the tracks are found in the overlap CFT).
The CPU time for overlap tracking is very large for complex events.
The fit quality (match chi-square) is not good at low pT.
There is a gap between the overlap and forward regions. Presumably this is responsible for the 5% inefficiency obsrved for the high pT tracks in the pzmm events.
The CPU times returned by the framework are not correct.

Removing the overlap

The overlap region covers a small region of phase space but uses most of the CPU time. We used a new feature (released in preco03.03) to remove this region and repeated the above studies.

 run bg  chk   CPU  CLK mem swp tm  pe  pTt # mc   eff  misr  fake  fcsq  mcsq  nclu
---- -- ----  ---- ---- --- --- -- ---  --- ----  ----  ----  ----  ----  ----  ----
pzmm  0  401     6    6  68  90         0.5  683  .627  .082  .016  18.3  12.0  13.7
pzmm  1  401    26   26 109 133  0   4  0.5 1892  .580  .141  .035  18.5  15.2  12.5
pzmm  2  401    73   73 154 178  0   2  0.5 2938  .557  .162  .034  19.8  16.5  12.2
pzmm  3  401   164  165 252 277  0   5  0.5 4554  .563  .172  .038  18.3  16.9  11.2
pzmm  4  401   343  344 419 442  1  12  0.5 5663  .569  .192  .041  19.4  16.9  11.2

Again we show results for tracks with pT > 10 GeV/c:

                                       10.0  154  .890  .007  .015  19.3   5.6  17.5
                                       10.0  160  .831  .030  .023  19.9   6.2  17.0
                                       10.0  186  .785  .041  .000  19.2   5.2  17.5
                                       10.0  171  .766  .069  .022  21.4   5.8  17.2
                                       10.0  169  .799  .104  .022  22.9   5.8  17.7

Here is the breakdown of the clock time:

 run bg   cc   co    c   cs    f  csf
 --- --  ---  ---  ---  ---  ---  ---
pzmm  0    1    0    1    1    2    0
pzmm  1    5    2    2    2   11    2
pzmm  2   14    3    5    4   42    3
pzmm  3   28    5    7    6  110    6
pzmm  4   61    7   10    9  247    8

Here are the plosts showing the tan(lambda)-z positions of the found (top) and lost (bottom) tracks:

pzmm + 0 bg  low pT  high pT
pzmm + 1 bg  low pT  high pT
pzmm + 2 bg  low pT  high pT
pzmm + 3 bg  low pT  high pT
pzmm + 4 bg  low pT  high pT

As expected, the processing was much faster. The best efficiency is now 89% instead of 95% because tracks are not found in the overlap region. This efficiency decreases as backgound events are added because the central CFT efficiency decreases. The misreconstructed fraction and the fakes are much lower.

dladams@uta.edu