• Data available:

    : The data from the following runs is available:

    Run# Start time N events t (hit) max jitter shaping time noise occ run key
    1 05-Jan-99 18:36 105 17-19 31 100 ns 0.7% $9996
    2 05-Jan-99 18:36 130 8-11 31 100 ns 0.7% $9996
    3 06-Jan-99 10:10 106 2-3 7 100 ns 0.7% $9996
    4 06-Jan-99 10:25 309 2-3 7 100 ns 0.7% $9996
    5 06-Jan-99 10:51 313 2-3 7 100 ns 0.7% $9996
    6 06-Jan-99 11:13 315 2-3 7 100 ns 0.7% $9996
    7 06-Jan-99 11:43 304 2-3 7 200 ns $9996
    8 06-Jan-99 12:07 311 2-3 7 200 ns $9996
    9 06-Jan-99 13:45 383 2-3 7 200 ns $9996
    10 06-Jan-99 14:16 80 2-3 7 400 ns $9996
    11 06-Jan-99 14:26 295 2-3 9 400 ns $9996
    12 06-Jan-99 14:53 309 2-3 9 400 ns $9996
    13 06-Jan-99 15:19 261 2-3 9 400 ns $9996
    13 06-Jan-99 15:42 350 2-3 7 100 ns $9997
    14 06-Jan-99 16:11 383 2-3 7 100 ns $9997


    making it a total of 3954 events.

  • Tools, parameters, etc.:

    : There were two different approaches used in this analysis, both involving some modification of the current track finder, SvtTrackFinder. In one of them, used by Eric Charles, the track finder was modified in such a way as to do a least square fit to a line. Details on the parameters used, as well as pictures of some of the reconstructed tracks, are available here. In the other approach, accepted by Gerry Lynch and myself, the tracks were fitted to helices.

    In my tests, I've also set noise levels and thresholds to the average values measured in run 20204 (100 ns, Qn = 2). Here are the values (courtesy Owen Long):

    Noise, fC Threshold counts
    Layer 1, phi 0.20 3.6
    Layer 1, z 0.17 3.0
    Layer 2, phi 0.20 3.6
    Layer 2, z 0.19 3.4


    These numbers were extracted from this plot.


  • Resolution:

    : Gerry Lynch reported the resolution of 26 microns in z and 35 microns in phi in run 6 (see this plot of the residuals in this run). Track parameters for the tracks found in this run are available here.

    Gerry has also provided the first estimate of the resolutions if alignment is taken into account (translations only for the moment). The description of the method he used is available here, and the resulting resolution estimates are 22 microns in phi and 18 microns in z. We are still missing a reliable accounting for the Coulumb scattering in the resolution estimates.


    • Efficiency:

    : There were two approaches to estimating the clustering efficiencies for these runs. The description of the method used by Gerry is available here. In particular, his estimate for run 6 is that the single hit efficiency is 97%, and that of the space point is 93%.

    I used a slightly different approach to this described here. For run 6, this method gives me 98% efficiency for phi hits, 96% for z hits, and 94% for space points.

    It is interesting to notice that most of the missing hits actually occur right on the edges of the active volumes. See, for example, Eric's plots for run 1 and run 6 showing the locations of the missing hits, or the following print out I get from run 1:

    Missing strip Layer Module Half Wafer
    Phi strip 799 1 2 forward 2
    Phi strip 874 2 2 backward 1
    Phi strip 874 2 6 forward 1
    Z strip 411 1 2 forward 3
    Z strip 441 2 2 backward 1
    Z strip 1 2 2 forward 2


  • Conclusion and suggestions for the next run:

    : It must be said that so far our analysis has not shown any outstanding problems with the reconstruction software. Track parameters, resolutions, and clustering efficiencies look quite good,

    For the next run, it might make sense to use a broader trigger window. These runs have shown an encouraging performance; but one must remember that most of the runs were quite clean. It would be good to see how we are doing when things aren't quite so clear-cut.

    Concerning David Kirkby's suggestion that the thresholds could be made low enough to mimic the expected background occupancies, one must remember that high occupancies from noise and equally high occupancies from background are not the same thing. The priority should probably remain taking data at realistic thresholds.