# 8.2 Black Spot module testing (2D testing)

113

# Module 4

Mean std. dev. of centroid Y position (pixels)

# 0.025

0.02

0.015

0.01

0.005

0

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

Time (samples)

Figure 8.10 The standard deviations of the centroid’s Y position over 2 minute periods throughout a day. The spikes are outliers as discussed in Appendix G.

graph. These spikes are treated as outliers and their cause is investigated in

ppendix G.

These outliers should not have a large effect on results due to the duration of the spikes with the exception of the worst case results shown in Table 8.3.

The two minute intervals in this data set can be analysed to produce a best case, worst case, and mean case scenario. gain, the marker closest to the centre of each module’s FOV was chosen for this analysis. The best case scenario was calculated by looking at the smallest standard deviation across all 2 minute windows. The mean case was calculated by taking the mean of the standard deviations over time and the worst case was calculated by finding the largest standard deviation. The results are shown in Table 8.3. The figures in the rows named ‘Best std. dev. of centre marker’ are the best noise performance that one could hope to expect over a 2 minute period. Module 4 has the best performance in the X direction with a standard deviation of 0.0047 pixels and 0.0045 in the Y direction. These numbers correspond to a standard deviation (at the measurement distance of approximately 2 m)

of 9.3 µm and 9.0 µm respectively. The worst case is the largest standard deviation of the centre marker over all 2 minute windows. This is shown in the Table 8.3 in the rows named ‘Worst std. dev. of centre marker’. These values are distorted by the outliers as shown in Figure 8.10 and, until further tests can be completed, these results have little significance. The mean standard deviation is more informative and has values of 0.0061 pixels in X and 0.0074 pixels in Y. This corresponds to 12.2 µm and 14.8 µm respectively.

In summary, the system is stable over the long term. Without continuous calibration the best result is a standard deviation of 1.4% of a pixel. Using a threshold of 2 standard