Multiple strips -> tracking resolution.
Having a cluster of hit strips rather than a single hit strip improves the point resolution. Just taking the center of a cluster as an estimator of the track crossing coordinate is better than the single-strip center:
For comparison (5th plot m29), sigma is 18.6 for normal-incident, single-strip hits. Also note that this distribution is flat, and making the assumption of a gaussian error distribution will negatively affect track fitting algorithms that assume measurement errors are gaussian. Since all this depends on angle of incidence, I added for this section 5 runs, with z-vertices at -10, -5, 0, 5 and 10cm. Further resolution improvements will result from making ADC threshold cuts, and ADC-weighted cluster fits. | |||||||||||||
|
Thresholds
We have to ignore cells where the signal is below some threshold. A particle traversing 300 um of silicon typically generates 24000 electrons. Setting the threshold at 2000 electrons, about 10% of this is an equivalent minimum path length of 25 microns. Strips that are traversed by less than the threshold path are not considered 'on'. | |||||||||||||||||||||||||||||||||||||||||||
Applying the threshold cut is expected to somewhat improve the
resolution, and this is what we see:
For example, for the most prevalent cluster size of 3, sigma goes from 9.876 to 8.633, a 15% improvement. The weighted mean is now 8.92 um.
Note that the threshold cut (25) is at 1/4 of the most probable path length (~100). | |||||||||||||||||||||||||||||||||||||||||||
Increasing the threshold cut is expected to make the cluster size smaller by discarding more of the first and last hits in a cluster. As expected, the number of clusters of size 1 goes up, the mean cluster size goes down, the ratio of the 2 most frequent cluster sizes (2,3) goes down, but the resolution (for simplicity, the sigma of the most popular cluster size) doesn't suffer much. |
|
||||||||||||||||||||||||||||||||||||||||||
Increasing the strip size while keeping the threshold at approximately 1/4 of the most frequent path length is expected to decrease cluster size and thus data volume. The resolution (defined as above) suffers accordingly. |
|