To gather the linearity test-data, The external DAC was stepped over five settings, beginning at 8.88 mV and increasing in steps of 39.11 mV (nominally corresponding to energies: 5, 27, 49, 71, and 93 keV). The data were captured in compressed event format into a file. Data analysis steps: ------------------- Removed "noise events", i.e.events not created by the test pulsers, and events with pathological data (e.g. zero or negative "time-of-rise" denominator). Then the data were sorted by pixel and an "energy" estimator E derived for each event prior to entering the event into a histogram. Once all the events for a given pixel are sorted into the histogram, the five peaks were located and the mean and FWHM of E for each peak was calculated by fitting with a Gaussian. A least squares best line was fitted to the five means and the deviations of the means from the line reported as a fraction of the mean value for the full scale (93 kev) peak. For each pixel, the transfer function gain and offset is proportional to the slope (m) and offset (b) of this fit line, respectively. The energy estimator was derived only from the central pixel info in the compressed event by the following floating point calculation: E = ( postsum - presum - offset(starting cap#)) * (1. + (numr/dnom)*2.15e-3) where postsum and presum are the post and pre event sample sums for the 5th (central) pixel, numr and dnom are the "time of rise" numerator and denominator, and offset(starting cap#) was derived from the data separately for each pixel as follows: For each pixel sort the events by starting cap# and create 16 separate histograms of (postsum - presum). For each starting cap#, calculate the mean of the five peaks and average the five means. This yields a set of 16 numbers. Find the average of the 16 numbers and subtract it from each of the 16 to yield a final set of "offset(starting cap#)" values. Note: a more-involved method was also investigated, where offset[starting cap#][peak#] was calculated for each of the 5 peaks separately, and then the overall offset[starting cap#] was calculated by averaging the 5 offsets calculated for each peak. This method did not produce offsets that were significantly different from the simpler method. Update: However, this method is more robust in the case of low statistics, where one or two of the 5 peaks cannot be reliably determined for a particular cap/pixel combination. So, this method is now used. For each ASIC the linearity analysis report currently consists of a PDF file containing the plots listed below, plus a space-delimited text file containing data that could be used for further analysis (see bottom of this document for file format), plus a text file listing the pixels that could not be analyzed Page 1: Histograms of gain (m) in Channels/KeV, and offset (b/) in KeV. Page 2: Plots of (m/) and (b/) as function of pixel#. Page 3: Histos of the calculated FWHMs (Channels), for each of the 5 peaks Pages 4 and 5: Plots of FWHM as function of pixel#, for each of the 5 peaks Page 6: Histo of the largest deviation of the means from the best-fit line, for all pixels Page 7: Histos of the Starting-cap offsets, one histo for each cap Pages 8-11: Plots of Starting-cap offsets vs pixel#, one plot for each cap The 32 following pages: The deviations of the means from the best-fit line are calculated as a fraction of the mean value for the full scale (93 kev) peak, and plotted, for each pixel. Format of data text file ------------------------ Each line in the file is a data-record for a single pixel. Fields in each record are space-delimited (NOT tab-delimited). Field Description 1 X pixel coord 2 Y pixel coord 3-7 Centroids of the 5 peaks, from gaussian fit to Energy-estimates 8-12 FWHMs of the 5 peaks 13 Slope of best-fit line to the peaks (Channels/KeV) 14 Offset (Channels) 15 Correlation coefficient of best-fit line 16-20 Deviations of the means from the line as a fraction of the mean value for the full scale (93 kev) peak 21-36 16 starting-cap offsets