Minutes of the Muon Software Meeting, Nov 10, 2004 ================================================== 1. Dale & Ingyin: t0 measurements for the modified CMX preamps ---------------------------------------------------------------- This is work in progress. We need to have the final numbers before we start to get beam again. The radekas on the preamps for some of the CMX stacks have been modified with the addition of capacitors, to reduce the effects of noise. The boards are in the Miniskirt, Keystone and two stacks in the SW arch. The addition of the capacitor effects the t0. Dale has measured the t0 using a scope. He defines the t0 as where the signal reaches its half height. He measures the t0 before and after the the change. The average t0 difference (dt0) is 35ns. The measured dt0's in the miniskirt are seen to be the same as for the arch. However an the dt0 is seen to be larger on the West side than the East side. Nobody can think of any reason why this should be. Dale and Ingyin are going to investigate further. We agreed that if this situation is not resolved before we start taking data that we will add +35ns to the t0 of the stacks were the Radekas have been attached. The shutdown ends at the end of this week, so any measurements will have to be have been made by then. 2. Victoria: Explaining the dxCMU distribution ----------------------------------------------- [Please see my updated transparencies with a better drawing of the CMU geometry at: http://lotus.phys.nwu.edu/~vmartin/cdf/talks/10Nov04/mytalk.ps] Victoria presented the current understanding of the CMU hit and stub reconstruction, and how this can be used to describe some of the features of the dxCMU distribution at high-pt. For each CMU signal, all we know is from which pair of ganged chambers the signal comes from, and the drift distance. Two things can go wrong in reconstructing the stubs: 1. We can reconstruct the hit in the ganged chamber instead of the chamber where the muon really was. We decide on the chamber by looking at the signal read out of the ganged chamber pair. We reconstruct the hit in the chamber with the largest width. 2. One we have decided where the hits are, we fit potential stubs and work our the chisq of the fit. Two potential stubs are fitted. We keep only the one with the best chisq. The chambers are slightly staggered, so the "real" stubs should have a better chisq that the fake stubs. We can describe the behavior of these two effects as follows: 1. The stub will be reconstructed in the neighboring stack. So dxCMU will peak at around +/- 6cm. 2. The wrong stub will be kept. This will place the reconstructed stub wrt the real stub within +/-6cm. So this would give a dxCMU distribution peaking at 0cm, and falling on either side and stopping at +/- 6cm. And, of course, both of these effects could happen at the same time. So using these two hypothesis Victoria fitted the dxCMU distribution for W->mu nu candidate events to the sum of four functions: 1. A central Gaussian for the correctly reconstructed CMU stub. 2. A logarithmically falling function, falling on both sides, from a maximum at the mean of Gaussian 1. 3. A Gaussian centered at about -6cm 4. A Gaussian centered at about +6cm. This not the ideal fit, but quite a good descriptions of what's going on. It fits with X^2/ndf = 180/89. Comments: If hit is reconstructed within 2.5cm of the end of the chambers, then we put this hits in both chambers. This 2.5cm came from run 1. We haven't tuned this. In run 2 neighboring chambers are ganged together, however in run 1 there was one intervening chamber between the ganged pair. So there is no good reason that 2.5cm is still suitable for run 2. This arguments does not take account of the CMU drift model, which is a radial drift model. Slava has seen that a lot of muons with large dxCMU tend to have a drift distance right on the edge of the radial part of the drift model. To reconstruct a stub we require at least 3 hits. If we reconstruct 0 or 1 hits in the neighbouring chamber we will reconstruct the stub in the correct chamber. If we reconstruct 3 or 4 hits in the neighbouring chamber we will reconstruct the stub in neighbouring chamber. However we may reconstruct 2 hits in the correct chamber and 2 hits in the neighbouring chamber. In this case we will not even reconstruct a CMU stub! This is a real inefficiency in the CMU reconstruction algorithm. What is the effect at low-pt vs high-pt. At high-pt we could recover some of the inefficiency by substantially loosening the the dxCMU cut. Now we understand this a lot better, we all agreed that this would be acceptable. At cut of |dxCMU| < 7.5 cm is probably the best motivated value. At low-pt the effect of the CMU misreconstruction is seen for pt>~8 GeV. Being able to cut tighter at low-pt would be reject a significant amount of background. We all agreed that we should revisit the CMU stub reconstruction. Some increase in both reconstruction and ID efficiency could be made by tuning the 2.5cm cut, and keeping both stubs in the chambers in some circumstances. If this is not enough then we may have to reconstruct the philosophy of the CMU stub reconstructing being independent of the other detectors. Using information from the CMP and tracking might be necessary. We would really like some manpower to work on this issue! Next meeting is scheduled for the day before Thanksgiving. We are canceling this meeting.