Summary of the Analysis of IES Units F5 - F9
Joshua M. Glasel for Prof. T. A. Fritz
Boston University Center for Space Physics
The same procedure was used for all of the units in carrying out the analysis. First, the raw data was obtained from wherever the testing for a given unit took place. In the cases of Los Alamos National Laboratory (LANL) and Goddard Space Flight Center (GSFC), the data was obtained from the BCE program and saved as excel files in order for a spectrum to be plotted from the data in excel spreadsheets where the analysis took place. All Barium testing occurred at LANL and all electron beam testing occurred at GSFC. The F7 and F8 units were the only two tested with a beam. Gadolinium testing occurred at Lindau Germany and the data for those tests had to be obtained via ftp.
Once the data were in useful excel files they were plotted on a semi log graphs, displaying the counts accumulated in each of 256 (energy) channels. This is known as the histogram mode in the integrated RAPID unit electronics. The spectra were then fitted with Gaussian curves. The Gaussians were used to fit the pedestal and signal peaks within a single spectrum. The pedestal is a result of the electronics of the IES and is prominent in all plots of any data, distinguished by its height, its position in the spectrum, (the left most peak), and its similarity to an actual Gaussian curve. The pedestal is also assumed to be the "zero" of energy. Results of the analysis of multiple pedestal spectra of the X-ray source data have confirmed that this assumption is correct. Any additional counts or peaks in the spectrum are a result of external phenomena and are considered signal data. Only data with signals were used here. Fits to the peaks were obtained using a chi-squared method. As a result of this process, beam data was much easier to fit than source data. This is because the electron beam used was at a given energy so the signal counts were concentrated around a centroid and resulted in a nice peak, which was easy to fit with a Gaussian curve. Sometimes there would be additional peaks beside the one predicted by the energy of the beam which were due to pile up, but these peaks were of no significance in the analysis done. The spectra of radioactive X-ray sources were sometimes very difficult to fit and at times had to be manipulated by eye. This was because the radioactive sources emanate X-rays at many different energies. This introduces a lot of noisy interference throughout the spectra. In the case of Barium 133 the two most prominent signal peaks occur at 30.97 and 80.99 KeV. In the case of Gadolinium 153, the prominent peaks occur at 41.54 and 97.43 KeV. These peaks at times, (especially the higher energy peaks), were engulfed by the counts from other energies since they did not emit as many X-rays as the lower energy peak did. In this type of situation, a satisfactory fit could not be obtained by the computer and had to be fitted by eye. Knowing which peak to fit, a width was assumed not too dissimilar from that of the pedestal, which gives a reference as to how wide a Gaussian fit should be for a given run. This, though, was not as accurate.
Once the plots were fitted with Gaussians, certain values could be read off of the spreadsheet and used further. After the fitting process, plots had to be made to determine the gain and channel offset of a given unit, I.D. and time constant. These plots were labeled Energy vs. Channel #. The data that was needed for these plots were the energies of a given run and the channel # that a given peak was centered about with respect to the pedestal, i.e. the peak positions channel number minus the pedestals peak position. The beam data, which was recorded at energies of 55, 110, 350 and 400 KeV, and the source data with their two peak energies each, provided the values of energies that were used on the plots. The channel # was read off of the spreadsheets. When this data was plotted linearly, the gain, (in KeV/Channel), and the offset in KeV were read off of the graph. The gain must be obtained to associate a channel in an I.D. with a specific energy range. The offset predicts whether our assumption of the pedestal being at the "zero" of energy is correct. In the case of source X-rays, the offset is usually no more than 2 KeV which tells us that our assumption is probably correct. For electrons however, the offsets are larger ranging anywhere from 8 to 15 KeV. This tells us that there is an energy loss for the electrons as they penetrate the contacts of a given I.D. detector for a given unit. While the X-rays energy is completely absorbed when the X-ray is counted, the electrons' energy that is absorbed is reduced somewhat from that of the incident electron energy. The gain like the offset, fluctuates between I.D.s and time constant. For source data though, the gains average is 2.25 KeV/Channel ranging from 2.2 to 2.4. The beams gain seems a little more consistent. It fluctuates between 2.1 and 2.2 averaging at 2.15 KeV/Channel. In addition to an electron beam calibration for the F7 and F8 units, there was also a test with an ion beam and an angular response test. The proton tests were only done at 50 us. Because of the large energies and long time constant, the peaks in these runs are extremely broad and extremely easy to fit. Also, energies lower than 500 KeV were engulfed in the pedestal and energies greater than 850 KeV did not register on the instrument. For F7, the gain is between 1.1 and 1.3 KeV/Channel and for F8 between 1.3 and 1.45 KeV/Channel. The proton threshold for F7 ranges from 500 KeV to 580 KeV and for F8 from 475 KeV to 500 KeV. Clearly a significant amount of energy is lost from the ions due to their energy and size when they hit the contact. For the angular response, each head consisting of 3 I.D.s was rotated, at different time constants, about a vertical axis with the beam held fixed. >From the spectrum of an I.D., a cosine effect is seen from the increase and decrease in counts as the head is rotated. This effect is plotted using the area of the Gaussian fitted to the beam signal, which changes in proportion with the number of counts. The graphs of these effects were normalized to the three highest data points. In every case, the cosine effect was exhibited in every angular run, some more clearly than others. These plots are graphed semi-logarithmically to accentuate this effect. For the rest of this analysis, even though the instruments were tested at different temperatures the only temperature that was used was room temperature, roughly 20 C.
The next step was to obtain an average gain and pedestal position of an I.D. for a given unit. Tables were produced in the section of the notebook labeled "averages", with these values. The bold numbers in the row labeled "gain" are the desired gain numbers for an I.D. The bold numbers in the row labeled "Ped. Pos." under the part that reads "Channel # of Pedestal" are the desired pedestal averages. The gain values are just the averages of all the gains for the individual runs for an I.D. The Ped. Pos. values however follow a different procedure. The section of the page labeled Ped. Pos. contains the actual channel numbers that the Pedestals in each run are centered on. This number is then added to the offset number of the corresponding run divided by the gain of that run. This is because the offset values are in KeVs and must be converted to channels in order to be added to the pedestal position. Then after this calculation in the section of the page labeled "Channel # of Pedestal", in bold, are the resulting pedestal averages for each I.D. The final values under "Channel # of Pedestal" do not vary that much beyond the final values under "Ped. Pos.". This is because a larger gain number is dividing the offset, which is a smaller number. This produces an even smaller number than the offset. So when this number is added to the Ped. Pos of a specific run, a big difference is not expected.
Although F7 and F8 also had this work done for the beam data, these values are not used in the final Ped. Pos. calculation, but are submitted just for comparison to the source data. This is to be more consistent with our calculations from unit to unit because the gains for the beam runs average about 0.1 Kev/Channel less than the source runs. The offsets are different because of the electron energy loss, in the contact of the detector. The inclusion of these numbers would introduce a great discrepancy.
As you can see from the source data, the gains fluctuate from I.D. to I.D. as well as the Ped. Pos. The variation of Ped. Pos. has to do with different foil thicknesses of the different contacts in the different I.D.s. The average gains for the source data vary from unit to unit but seem to stay around 2.27. In the case of F8 though the gains are significantly larger averaging about 2.35, nearly a tenth higher. The beam gains, conversely, average a tenth lower, about 2.15. All gain values have units of Kev/Channel.
The final part of the analysis consisted of calculating a source count rate for each unit, I.D. and time constant for the source data. This is done to find out whether or not the count rate is basically constant within an I.D. as it should be. This was not done for the beam data because the count rate was known from a monitor readout of the beam statistics. First, if it exists, an accumulation time, (Tacc), was found. If not, it was calculated, (Tcalc), from the total counts of a run multiplied by a number dependent on time constant. If this time agreed with the logged time, then the logged time was used. If not, or if there was no logged Tacc, the Tcalc was used. It appears that for some tests, a consistent disagreement of times occurs where Tcalc is less than Tacc. It may be due to the fact that whatever clock was used to log the Tacc, was running slower than what is predicted by the data. This answer though is not known and could probably be resolved by a communication. The total counts, (Total Cnts.), were also needed and readily available from the excel spreadsheets when all of the counts of a run were summed up. Then the source count rate could be calculated by dividing the number of signal counts, (excluding pedestal counts), by another number dependent on time constant. One thing that can be said after the analysis is that the hopes of consistent source count rates were not fulfilled. Another observation would be that for a given I.D., the Cnt. Rate. decreases with increasing time constant constantly. Upon further analysis of the plots containing the spectra, one direct cause for random disagreements in the Tacc and Tcalc are what seem to be features resembling pitchforks. These pitchfork features, which occur only in the pedestal, significantly reduce the total number of counts. This reduces the Tcalc, which in turn reduces the Cnt. Rate. It may be that this is a result of some sort of overflow in specific channels when they are saturated. Regardless, there still is an incredible inconsistency in all of the data.
The F5 unit (or Phoenix unit) was tested once at LANL with an Americium and Barium 133 source and twice with a Gadolinium 153 source, in April 98 and in Aug 98 at Lindau. The I.D.s with an Americium source were not used in any analysis due to the fact that the spectrum exhibited by this source only gives one peak energy signal with which to work. This was useful to see the response from the I.D. but not useful for any calibration efforts. I.D.s 1,2 and 3 were tested with Barium and consequently used in the analysis. This test at LANL also consisted of 2 runs, one at 0 C and one at 20 C, (roughly room temperature). The first Gadolinium test, on Apr. 98 used all I.D.s on the detector and was tested at room temp. The second test on Aug 98 used I.D.s 1-3 and 7-9, but not 4-6. There were also 3 runs at -10, 20 and 40 C. The F6 unit was tested at LANL with Barium on I.D.s 1-3 at temperatures of 0 C and 20 C. At Lindau, Gadolinium was tested on all heads during Dec. 98 at room temperature.
The F7 unit was tested at GSFC with an electron beam and an ion beam. These tests covered all heads at room temperature. The ion tests only occurred at 50 us. Also, angular response tests were run. The data for these runs are sometimes sparse due to time constraints. The Barium tests were conducted at LANL on I.D.s 4-6 at 0 C and 20 C. The Gadolinium tests were at Lindau on Mar. 99, covered all heads and were tested at room temperature.
The testing for F8 at GSFC were conducted in the same manner as F7. The Barium test at LANL covered I.D.s 7-9 at 0 C and 20 C. The Gadolinium tests at Lindau covered all heads at room temperature and data was taken on Aug. 99.
The F9 unit was only tested once at LANL with Barium at 0 C and 20 C on heads 4-6. As a result, I would consider the analysis of F9 somewhat inconclusive.