To demonstrate the influence of integration time on the NRMSE, the same water specimen as used above was scanned vertically at a high intensity (35 kV, 700 µA) and at a low intensity (30 kV, 200 µA) for integration times of (1, 5, 10, 20, and 30) s. After five seconds of integration, x-ray counts were ≈4300 and ≈59 600 for the low and high intensities, respectively. The 30 s scan at the high intensity was assumed to be the true value for the computation of NRMSEs. As the integration time increased, the NRMSEs decreased (Fig. 3). For example, at the high intensity, the NRMSE decreased from 1.3 % to 0.5 % when the integration time was increased from 1 s to 20 s. At the high intensity, integration times of 5 s and higher result in practically the same NRMSE. The NRMSEs are again slightly higher than the uncertainty estimated from the square root of the number of counts. In general, integration times of 5 s or longer produced NRMSEs that are less than 2 % for both intensities tested. Note that the high intensity produced NRMSEs that are two to three times lower than those produced by the low intensity. Higher intensities and longer integration periods, then, decrease the uncertainty in the measurements, as expected for a Poisson process.
Fig. 3. Normalized Root Mean Square Errors (NRMSE) as a function of integration time for experiment WVP at low (30 kV; 200 µA; ≈850 counts per second) and high (35 kV; 700 µA; ≈12 000 counts per second) x-ray source intensities. The data from the 30 s integration time at the high intensity was used as the true profile for the calculation of the NRMSE. See Table 1 for more information on experiment WVP.