Estimating momentum resolution with fast simulator

Understanding, parametrizing, and correcting for momentum resolution is important for HBT in E895, as shown by this discussion.
Here, we use RQMD events through the fast simulator to get the p^- momentum resolution. Obviously, saying it is "1%" or something is way too simplistic. It will vary with the track rigidity, and with the number of padrows the track passes over in the detector.
Also, it is important to get the resolution for each component! There is no reason the momentum resolution along the beam direction, in the drift direction, and in the bend direction must be the same percentage of the total momentum.

Here, we use only tracks with small DCA ( < 1.0 cm because the fastsim gives better DCA than the data), and look at the reconstructed vs. input momentum for negative pions. We do this for windows on |p| and on # padrows crossed. For each cut, I have checked, and the difference distribution for each component looks (roughly) Gaussian, so I characterize the resolution with a RMS. Another point, in these simulations the target is OUT. I wanted to see what the effects other than MCS and straggling in target was. I take care of them seperately.

Here are the results. Black stars are when the vertex refit is on, and red is when it is off. (Thanks to Chris for bringing this back to life!)
Plotted is the RMS of (R'_i-R_i)/R_i, where R' is reconstructed, and R is input, and i=x,y,z. This means, for example that in the plot for R_x, the resolution for long tracks goes from ~ 1.3 percent of the total momentum to ~0.5% of the total momentum starting at about 0.5 GeV.

Update

So, when we correct the correlation functions using the parameterized resolution, how do the fit parameters compare with our previous estimates using a 1% and 3.5% tracking resolution?

Here are the plots. The point at "0" is raw (uncorrected), and the point at "-1" is using the parameterized resolution.