Minutes of the Testbeam Meeting, held Tuesday, October 22, 2002
 =================================================================

 (Present:  ML, NK, TI, TS, MF)


 -  Naoko is back from CERN. She reported that she met with Rob McPherson,
    and they discussed the pulse shape for digital filtering.
    There is a note on the EM barrel pulse shape (ATL-LARG-2001-008)
    that discusses a FFT technique to reconstruct the pulse. Naoko
    is looking into this.
    She now also has access to the calibration pulse data for the HEC.


 -  Michel showed us a very comprehensive overview over his timing studies.
    He used the stability runs to calibrate the timing.
    Pavol has created a number of weights files and Michel is in charge of
    evaluating their distribution.
    Pavol had used cell 45 and the results of Michel's program to generate
    the distributions.

    Predictions are that there will be a large number of wac's for
    the combined EM/HEC testbeam period.

    In detail, Michel looked at run 12745 (150GeV pions into cell 45).

    He showed us plots of:
    - signal distributions for cubic fit and for digital filtering.
    - raw tdc  vs. time at the peak from cubic fit
    - distributions of wac-corrected peak position
      for cubic fit and digital filtering
    - normalized adc values   vs.  time dig corrected
      This particular plot shows the range over which the weights
      were generated. There are 5 points for each event in the plot:
      one point for each of the time slices that was used in the
      digital filtering.
    Essentially this part of looking at the timing makes
    use of the tdc information for digital filtering.
    The challenging problem for this particular run is that there
    are two distinct lines/peaks in all of the above distributions.
    It looks like there could have been two triggers contributing
    to the e/pi trigger that have somewhat different timing.

    In order see whether the tdc information could possibly be
    circumnavigated, Michel looked what happens when using the cubic fit
    information for the digital filtering.
    For that:
    - use only cells with a deposited energy of more than 50 adc counts.
    - Now the double line goes away in the digital filtering, but
      it is still there in the cubic fit (suggesting that there
      are 2 triggers contributing to e/pi).

    Michel now has an automated way to find the timing for every cell
    that has data (>50adc) in it: This code stores timing information
    for EVERY event .and. EVERY cell above threshold in a multimap
    during the athena "execute". In "finalize" he does his own
    digital filtering by looping over all the events in the multimap
    and iterates t0 for each cell until the cell has the correct t0.
    This can be done for either cubic fit or tdc.

    Michel showed us plots for about 5 different cells before and after
    this analysis for cubic fit and digital filtering (based on cell 45).
    It looks very good, but it remains to be seen what biases are
    introduced e.g. by setting thresholds and deliberately selecting
    cells according to the deposited energy in them.