Muon Offline Meeting -- Minutes ================================= 20 - March - 2002 ---------------------------------------------------------------------- David Dagenhart -- Muon Fiducial Function ====================== David discussed the work that he has done to create the "new FIDMU", a function that will tell you whether a given track is expected to be in the fiducial volume of a given muon subsystem. There is a new class, MuonFiducialTool, in the Muon package. This class has a method isFiducial(CdfTrack track, MuonDigiCode::system, distance), where "distance" is an optional argument. At the moment, CMX is the only valid system. If "distance" is included in the call, the tool will fill the variable with the track displacement in the local X-Z plane to the nearest point on the fiducial boundary. The value will be negative if the track is fiducial, and positive if not. Thus, a user can find out just how close or far a track is from the fiducial-volume boundary. The tool has been tested for an extremely wide range of track parameters. At the moment, the code uses the MExtrapolator to take the track to the chambers, but it wouldn't be hard to switch to the GEANT extrapolator if desired. It is designed to allow for a connection to a dead-channel database, so that a dead chamber is not considered part of the fiducial volume. The code uses transformations from the standard geometry to determine whether a trajectory hits each drift chamber, and then counts the number of chambers that are hit to determine if that trajectory is fiducial. It was suggested that this code could be turned into something that would tell you what piece of geometry a given point of space belongs to, but that is probably a lot more work. David looked at the results of applying this tool to some single-muon simulation. In the great majority of cases the tool gets it right -- a track that is called fiducial has a stub present, and a track that isn't doesn't. In most of the other cases, it looks like there was a reasonable amount of multiple scattering (these were 9 GeV/c muons) which would knock the muon off its predicted trajectory. There was some debate, then, on whether it is proper to return a simple boolean value from this function -- as you go lower in momentum, scattering becomes more important, and the tool is more likely to screw up. The boolean is really only useful for high-momentum muons. In any case, more testing should be done. David will be extending the reach of this tool to the CMU, CMP and BMU, and he was encouraged to make this his top priority. He also plans to add additional functions to return separate X and Z distance, and to return additional information about the internal calculation, such as the position and direction of the extrapolated trajectory, and the MuonDigiCode of the nearest chambers. ---------------------------------------------------------------------- Slava Krutelyov -- The G3XEx Is Fixed? =================== What was supposed to be a simple signoff on the G3X extrapolator turned out to be yet another abattoir for this next advance in muon reconstruction, and Slava properly followed Hinchliffe's Rule in his title. Slava exposed several different problems related to the use of GEANT. One is in the simulation. Our step sizes are often too large, especially in the wall and plug calorimeters. The step sizes are set more reasonably in G3X, which leads to an inconsistency in results. This can easily be seen at low momentum (about 2 GeV/c) in the CMU, but even at very high momentum in the CMX, where of course the muons are going through the calorimeters in question. But this is only part of the story -- fixing the simulation improves matching at higher momentum, but there is still trouble at low momentum. The problem is deep inside GEANT; there are some cross-section tables that get initialized quite differently depending on whether you tell GEANT to have energy-loss fluctuations on or off. This was a huge surprise to our GEANT experts (i.e. Andreas), and needs further exploration. The bottom line is that GEANT must be initialized quite differently for simulation and reconstruction, and this raises big issues for running both processes in the same executable. Somewhere in this discussion we talked about the magnetic field map in GEANT, which certainly plays a role here. We're using the Run 1 map, and we don't actually know how correct it is for our current configuration. Later, a smaller group of experts agreed on plans for G3X in the short and long terms. In the short term, we will try to get the initialization done correctly, and in such a way that you can flip from simulation to reconstruction intialization via a talk-to. If and when release 4.5.0 is used to reconstruct on the farms, we will get the talk-to switches thrown. In the longer term, we'll try to understand what is really going on with GEANT. ---------------------------------------------------------------------- Ken Bloom -- G3X in High-pT Muons Redux ========================== Despite the above bad news, we pressed on with talks that showed that G3X was working just fine. Ken repeated all of his studies with high-pT muons, looking at both single-muon simulation and some pre-shutdown data. In all cases, things looked good for G3X. In the simulation, we see dx distributions that are even more centered on zero than for the simple extrapolator, and with the same width. The dx distribution stays centered on zero down to pT of 5 GeV/c, as low as Ken looked. The data also looked fine, with CMU dx means and widths entirely consistent with the old extrapolator. As far as Ken was concerned (at least before Slava's talk), there were no problems with G3X, except for getting the defaults in SimInitManager. ---------------------------------------------------------------------- Mario Spezzegia -- G3X in the J/psi Sample ======================= Mario looked at matching quantities in the J/psi sample, using the G3X extrapolator. He too saw very similar results to what was obtained with the simple extrapolator. There did seem to be some loss in total event yields, but after a sideband subtraction, the J/psi yield remained the same. So here too, all seemed well. Jaison Lee showed a similar set of plots, and came to the same conclusion -- it looks like G3X is working fine in the data, despite the issues that Slava discovered. ---------------------------------------------------------------------- Henry Frisch -- Proposal for Limited Volunteers in the High-pT Samples ====================================================== Henry came running in from the top-dilepton meeting (now if only we could get them to move to a different time!) to show his revised proposal for getting volunteers into our data streams. The hope is that the volunteer rate will be zero; if not, something is wrong and needs to be investigated. For the short term, at least, Level 3 is going to be running all reconstruction on every event, in response to the e-g-g-met trigger debacle. Thus, there is no longer a need to try to shield uninteresting events from full reconstruction, and Henry's proposal becomes much simpler. One simply takes the global OR of all Level 2 accepts, and then runs the usual high-pT muon filter at L3. If all is well, there will be no events accepted on this path that are not accepted on the usual path. No objections were voiced to Henry's proposal. ---------------------------------------------------------------------- K.Bloom