GEM Cluster Finder
( FermiLab test beam data taken in May of 2007 , based on May7_run34 )
Updated, September 26,2007

  1. Data from U & V-planes of detector B from a single run have been processed.

  2. Pedestal correction
    Details of this procedure are given at here , (see 200 events PDF )

    Example of one event: raw ADC (top) , after ped correction (bottom). Plane U (V) spans X-axis 0-191 (192-383).
    Vertical green lines mark chip boundaries. Pedestal are marked be the ~horizontal green line. For this event pedestal fit failed QA for the 3rd chip ( magenta) and all ADC's are cleared to zero.

  3. Cluster finder algorithm

  4. QA of found clusters & events is illustrated on frequency plot below. Bins count events passing subsequent criteria, separately for U (left) and V-plane (right)
    1. all input events
    2. events with non-zero cluster per
    3. events with exactly one cluster
    4. all strips in this cluster have STAT=0 (it is 'good' clusters)

  5. Characteristics of 'good' clusters is shown on subsequent plots, separately for U (left) and V-plane (right)

    1. Seed cluster distribution vs. strip index, last chip in V-plane is dead

    2. Seed cluster amplitude. Peak at 580 in the U-plane comes from pegged raw ADC. V-plane is positioned after U-plane and sees less energy.

    3. Seed cluster amplitude vs. strip index. Now one can identify which chip saturated most often

    4. Cluster energy (in ADC units).

    5. Ratio of seed ADC to Cluster energy. V-plane clusters are wider (seed caries less energy), despite saturation of the U-plane.

    6. Ratio of ADC of 2nd highest strip/seed strip. The peak at ~1.0 in U-plane is probably due to both strips saturate.

    7. Averaged cluster shape. Events aligned based on position of max strip. Re-normalized to one event. Fit("gaus","I").

  6. 2D cluster reconstruction
    * Event QA, condition defined as follows
    1. input eve
    2. all 5 chips w/ good peds
    3. any chip w/ good pedestal
    4. both plane any cluster
    5. both plane exactly one cluster
    6. both plane exactly one and good cluster ('good' is defined in item 4.4 )

    * LEFT: 2D cluster position (condition 5) - looks like a beam spot to me.
    * RIGHT: cluster energy correlation (condition 6). Clusters w/ saturated strips has NOT been removed, most likely cause of nonlinearity in reco energy.

Conclusions:
* later plane sees less energy and wider profile
* Based on my M-C cluster profile study , (table 2) the width of U-plane cluster is slightly above 0.4*strip width, V-cluster is wider: ~0.6*strip width - This is not good!
* cluster energy in both planes is proportional to ~20% -30%

  • Used code is here

  • For reference: Frank's analysis page