Performance of a Silicon Drift Detector
count rates (<70-percent dead time). For the 0.5- and 0.8-ms time constants, the energy shiſt is ≤10 eV if the dead time is kept below ~54 percent (input count rates of 425 and 310 kcps, respectively), which would keep the input count rate at or below that for maximum throughout. Te soſtware offers a post-acquisition routine for removal
Figure 5: Energy resolution (FWHM) of the Mn Ka peak measured at 20 kV as a function of time constant and input count rate. It should be remembered that the 40-mm2 chip is an older Ketek chip that has been replaced by newer chips in currently available SDDs.
collected at significantly higher dead time (DT) values than that for maximum throughput. Such non-optimized counting conditions normally would not be employed for actual X-ray microanalysis. Similar behavior is observed for the FWHM of the Mn L line, with slightly wider peaks resulting from the presence of multiple, unresolved L lines. Te displayed energy of X-ray peaks generally shiſts
to a slightly higher energy as a function of input count rate (Figure 6). Te dominant origin of this shiſt is the assymetric broadening or high-energy tailing of the peaks at higher count rates (for example, see Figure 2). Te slight variation in indicated peak energy at a low count rate for the different time constants simply reflects small differences in energy calibration. Te apparent increase in peak energy for the four slower time constants (that is, 1.6–12.8 ms) is ≤10 eV as a function of input count rate over the entire range of usable
of pulse pile-up effects. Figure 7 shows the result for a spectrum collected with the 3.2-ms time constant at 40-percent dead time before and aſter running the correction routine. Tough the correction partially removes some pile-up effects (sum peaks, higher background between and above the Mn Ka and Kb peaks), it does not completely remove the sum peaks. Te best correction removes ~60 percent (Figure 7) at 3.2-ms TC, 40-percent DT. For many other SEM and SDD operating conditions, the correction routine removes 40–50 percent of sum peaks. Te correction routine removes more of the Ka peak than the Kb peak and does so asymmetrically. Te correction routine [5] is based on the earlier work of Wielopolski & Gardner [6], Statham [7], and Bristow and Harrison [8]. In his paper, Elam indicates that the spectrum correction requires “a realistic model of the hardware response,” which apparently is somewhat lacking for the current SDD. In addition, the correction sub-routine deletes the counts in the sum peaks and other pulse pile-up effects but does not return them to their correct (original) energies, as would be desired. In order to understand the apparent non-linearity
between input count rate and specimen current, as indicated by the saturation of the indicated input rate (Figure 3), the count rate was measured as a function of probe current, which was measured using a Faraday cup and picoammeter (the results for the 0.5-ms time constant are shown in Figure 8). The software package offers two measures of the input count rate. The first is displayed during spectrum acquisition, and the second is calculated after spectrum acquisition using a ratemeter function for the entire spectrum (0–40 keV for spectra collected at 20 keV in order to capture the counts at >20 keV resulting from pulse pile-up effects). Both of these rates are non-linear as a function of probe current, indicating a problem with the “fast channel” of the detector/pulse processing electronics that are used to determine the input count rate. If the true input count rate is estimated by linear extrapolation of the low count rate data as a function of probe current, the result is the straight line shown in Figure 8. The non-linearity probably starts below 200 kcps and becomes more pronounced at higher count rates. (The actual count rate is underestimated by a factor of ~3 at 100 nA (~500 kcps, ~75 percent dead time)). The results from the other time constants basically overlay the 0.5-ms data. The manufacturer agrees that this effect is present in the current generation of detector/processing electronics and processing software and will be addressed in the next generation of electronics hardware that should be available in the near future. Because the indicated count rate saturates at 500 kcps for
Figure 6: Apparent energy of Mn Ka peak as a function of time constant and input count rate.
44
pure manganese at 20 keV, the true count rate is drastically underestimated as the saturation point is reached or exceeded. When conducting performance tests on other materials and other accelerating voltages, it became clear that the saturation
www.microscopy-today.com • 2011 May
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76