For each channel its offset or `pedestal'-value was determined continuously during the beam period by applying a random trigger. The pedestal value had to be subtracted for each channel separately. The width of the pedestal peak is determined by the amount of electronic noise present. Most of the noise was coherent regarding several channels. The subtraction of coherent noise is referred to as secondary pedestal correction. After applying this correction a threshold value was obtained to separate electronic noise from valuable energy information.
As mentioned above, intercalibration effects contribute to the constant term of the energy resolution. Since an electromagnetic shower develops over several crystals, the crystals have to be cross-calibrated against each other. The so-called gain matching was applied in a two step approach. In the first place the sum of both calorimeters defined the invariant mass of the decaying p 0. For each crystal this energy information was adapted such, that the peak position for each channel was equalized. In a second step, shower leakage also was taken into account. To this end the individual spectra for each crystal were obtained by simulation and compared with the measured spectrum.
Shower leakage also causes the presence of a tail to the left side of a peak. In order to reduce the extension of that tail, that is to minimize the contribution of the tail of the 129 MeV peak under the SCX distribution, a cut limiting the deposited energy in the outermost crystals was implemented.
Further background reduction had to be achieved off-line. Due to a high rate some 8 MeV neutrons from radiative caption fell into the time window of the photons. Furthermore scattered beam pions reached the CsI-calorimeter. A cut on the timing spectrum of the beam counter reduced these sources of background notably.
The background and the implications of the cuts were studied using a GEANT simulation of the LH2 target. Furthermore the detector acceptances for photons of different energies were determined using the same simulation.
Finally, a clean photon spectrum with two well-separated photon distributions was obtained. Applying fit functions to these distributions the integrals were calculated and such the Panofsky ratio was obtained.