Hewlett-Packard (HP) 9831 calculator for initial field processing. The results were printed out on an HP printer, and the data then stored on cassette tape. A Pb-Cd collimator was used to limit the detector field-of-view for 60 keV gamma rays to a finite area on the ground (see Section 3.2.8). The collimator consisted of 1.6 mm (1/16") thick soft lead backed by 0.8 mm (1/32") thick cadmium. Both the lead and cadmium were supported on a 1.6 mm thick aluminum cone. The collimator slipped around the detector housing cap and then extended down 12 em at an angle of 50° from the vertical. A 1.27 em thick soft lead collar, 2.54 em long, was placed around the detector housing cap to further reduce background counts in the 24lam photopeak window due to air scatter. In order to adequately support the Enewetak cleanup project, it was necessary to fabricate three complete in situ systems, i.e., three IMPs. All three systems were identical. Two systems were routinely deployed in the field while the third system provided a complete backup. 3.2.3 Data Reduction Procedures Field Processing. The initial stage of the data reduction was performed in the field immediately following each measurement. The main advantage of this procedure was that the operator could perform quality control checks on the system after each measurement, which shortened the data turnaround time. In addition, the program allowed the operator to input certain bookkeeping information through the HP 983] calculator; usually, this consisted of island name, stake number, percent of brush cover, date, time, weather conditions, and the detector serial number. This information and the spectral data were then stored on magnetic tape. The field program was restricted to analyzing five specific narrow regions of the spectrum to yield data for 241Am, 155£u, 137Cs, and 69Co (69Co in two regions). This restriction, and the technique used to extract the photopeak data, enabled the field processing to be completed during the time it took to move between locations. Photopeak shapes for the four isotopes (five photopeaks) were determined empirically on Janet for the first two HPGe detectors to arrive at Enewetak. Resolution of both units was 1 keV to 1.2 keV full width at half maximum (FWHM) at 59.5 keV under normal field operation conditions. Detailed manual (graphical) analysis was performed on each of the five photopeaks for count rates ranging from background to those of the calibration sourees--tens to hundreds of times background. Peak shapes were constant over the count rate ranges within the limits of recognition imposed by statisities at lower count rates. Careful measurements were then made, using the high count rate data, to determine the points at which the peak rises out of the background. delineated was used in the program to determine the centroid and net Each region so photopeak counts. Symmetrical windows adjacent to the peak region were used to determine (by straight line interpolation) the background under the peak. To find a peak, a narrow predetermined segment of the spectrum was examined. This method, which contributed greatly to the quickness of the program, was viable because each measurement was analyzed immediately, so the IMP operator could adjust the gain and zero of the analyzer system, when necessary, to keep the peaks where they belonged. For peak finding, the raw data were first smoothed bya sliding interval filter of near-optimum width. The filtered data were searched for the channel with the most counts. This channel was the "peak" channel. No further use was made of smoothed data. The central peak region and background windows were positioned with respect to the peak channel as described in the previous paragraph. Then the peak centroid, background counts, and net peak counts were determined. The one sigma standard deviation was calculated from the total counts (peak plus background) and a statistical counting error was assigned (sigma/net counts). The centroid (in channel number) was converted to energy. Net counts were converted to equivalent soil concentration using a conversion coefficient stored in the library array and the live time measured by the analyzer during spectrum acquisition. The coefficient stored in the library had units of (pCi/g)/eps. Determination of that number is described in Section 3.2.5. The error assigned to the soil concentration result was the statistical counting error, plus a 10 percent error to account for uncertainties in the conversion coefficient (see 86 Section 3.2.6). It should be pointed