As mentioned above, tracking software usually makes use not only of space points, but of the uncertainty in the position, in the fit. Therefore, it is important that the uncertainty be understood and estimated hit-by-hit. Also mentioned was the fact that the magnitude of the uncertainty can depend on quantities not known at the hitfinder level (i.e. crossing angles), so that uncertainties will have to be ``updated'' at tracking time.
Any estimate of the position resolution of the cluster/hitfinding software depends strongly on the degree to which the simulations are realistic. In our case, it is known the simulator does not incorporate so-called ``finite electron'' effects, due to the fact that the number of primary ionization electrons liberated by the track as it passes over the padrow is not infinite. Studies by Howard Wieman have shown that finite electron effects will worsen position resolution by about 20% [4]. This will affect the resolution along the padrow (in the ``pad direction'') as well as along the time direction. Also, ExB distortions and other field distortions are not taken into account in the simulations. Hence, the resolutions obtained from this study are overly optimistic. Furthermore, we note that the accuracy of the current standard versions of STAR Geant and TPC slow simulator have been questioned, in particular about the handling of delta electrons and energy fluctuations. In order to focus on the cluster/hitfinder, we use these standard versions in the resolution study, with the understanding that updates to the information given here will come as needed in the future.
The residual along the padrow and in the time (z) direction was calculated for each hit that could be associated with a Monte Carlo padrow crossing. These were then transformed to residuals in cm in the x, y, and z directions. These redidual distributions for Au-Au events are shown in Figure 21. Position resolution for hits coming from a one-hit cluster are typically 15% better than those for overlapping hits.
Figure 21: The resolution in the global x-, y-, and z-directions
is shown for a simulated Au-Au event. Significant non-Gaussian tails
on the distributions lead to a large RMS value
of the residual distribution.