5 \page README_rec Reconstruction
7 The reconstruction is a multistage process, driven by the AliMUONTracker and AliMUONReconstructor classes
8 via the AliReconstruction class, which is divided into three parts:
9 - the digitization of the electronic response,
10 - the clustering of the digits to locate the crossing point of the muon with the chamber,
11 - the tracking to reconstruct the trajectory of the muon in the spectrometer from which we can extract the kinematics.
13 All the adjustable options and parameters used to tune the different part of the reconstruction are handled by the class AliMUONRecoParam.
16 \section rec_s1 Digitization
18 - We read the RAW data, convert them (convert them back for simulated data) to digit (object inheriting from AliMUONVDigit
19 stored into containers inheriting from AliMUONVDigitStore). This conversion is performed by the class AliMUONDigitMaker.
20 - We calibrate the digits, via AliMUONDigitCalibrator, by subtracting pedestals and multiplying by gains. All the calibration parameters
21 (pedestals, gains, capacitances and HV) are read from the OCDB and stored into AliMUONCalibrationData objects.
22 - We create the status of the digit (e.g. pedestal higher than maximum or HV switched off), using AliMUONPadStatusMaker.
23 - We create the status map for each digit, i.e the global status (good/bad) of that digit and of its neighbords, using AliMUONPadStatusMapMaker.
24 - Calibrated digits might be saved (back) to TreeD in MUON.Digits.root file.
27 \section rec_s2 Clustering
29 - We convert the digits having a positive charge into pads (AliMUONPad objects), which also contain information about the digit geometrical
31 - We loop over pads in the bending and non-bending planes of the DE to form groups of contiguous pads. We then merge the overlapping groups
32 of pads from both cathodes to build the pre-clusters that are the objects to be clusterized.
33 - We unfold each pre-cluster in order to extract the number and the position of individual clusters merged in it (complex pre-clusters are
34 made of a superimposition of signals from muon, from physical background (e.g. hadrons) and from electronic noise).
35 - We finally determine the MC label: take the one of the simulated track that contribute the most to the total charge of the 2 (bending and the
36 non bending) pads located below the cluster position. This is possible only if we perform the reconstruction from simulated digits (which contain
37 the list of MC track contributions). We set it to -1 when reconstructing from raw data or in case of failure.
39 Several versions of pre-clustering are available, all inheriting from AliMUONVClusterFinder, with different ways to loop over pads to form
41 - AliMUONPreClusterFinder
42 - AliMUONPreClusterFinderV2
43 - AliMUONPreClusterFinderV3
45 Several version of clustering are available, all inheriting from AliMUONVClusterFinder, with different degrees of complexity:
46 - AliMUONClusterFinderCOG simply compute the Center Of Gravity of the charge distribution in the pre-cluster.
47 - AliMUONClusterFinderSimpleFit simply fit the charge distribution with a single 2D Mathieson function.
48 - AliMUONClusterFinderMLEM uses the Maximum Likelihood Expectation Minimization algorithm.
49 This is a recursive procedure which determines the number and the approximate position of clusters into the pre-cluster that are needed
50 to reproduce the whole charge distribution. It assumes that the charge distribution of each single cluster follow a 2D Mathieson function.
51 If the estimated number of clusters is too high (>3), the pre-cluster is split into several groups of 1-2 or 3 clusters selected having
52 the minimum total coupling to all the other clusters into the pre-cluster. Each group of clusters is then fitted with a sum of 2D Mathieson
53 functions to extract their exact position.
54 - AliMUONClusterFinderPeakCOG is a simplified version of the MLEM clusterizer, without splitting and computing the Center Of Gravity of the
55 local charge distribution to extract the position of every clusters found in the pre-cluster.
56 - AliMUONClusterFinderPeakFit is another simplified version of the MLEM clusterizer again without splitting. The pre-cluster is fitted with
57 a sum of 2D Mathieson if it contains less than 3 clusters or we switch to the above COG method.
59 The cluster recontruction is driven by the class AliMUONSimpleClusterServer, inheriting from AliMUONVClusterServer.
60 It can be performed either before or during the tracking. In the first case, all the chambers are fully clusterized and the clusters (objects
61 inheriting from AliMUONVCluster stored into containers inheriting from AliMUONVClusterStore) are saved to TreeR in Muon.RecPoints.root file.
62 We use the class AliMUONLegacyClusterServer (also inheriting from AliMUONVClusterServer) read back the TreeR and provide clusters to the tracking.
63 In the second case, we clusterize the chambers only in the region where we are looking for new clusters to be attached to the track candidates.
64 This makes the clustering faster but the clusters cannot be saved to the TreeR.
67 \section rec_s3 Tracking
69 The MUON code provides two different algorithms to reconstruct the muon trajectory. In both cases the general tracking procedure is the same,
70 the only difference being the way the track parameters are computed from the cluster positions. The "Original" algorithm perform a fit of the
71 track parameters using the MINUIT package of Root, while the "Kalman" algorithm compute them using analytical formulae. The classes driving
72 the tracking are AliMUONTrackReconstructor and AliMUONTrackReconstructorK for the "Original" and the "Kalman" algorithms respectively,
73 both inheriting from AliMUONVTrackReconstructor. The reconstructed muon tracks are objects of the class AliMUONTrack.
75 The general tracking procedure is as follow:
76 - Build primary track candidates using clusters on station 4 and 5: Make all combination of clusters between the two chambers of station 5(4).
77 For each combination compute the local position and impact parameter of the tracklet at vertex and estimate its bending momentum given the averaged
78 magnetic field inside the dipole and assuming that the track is coming from the vertex. Also compute the corresponding error and covariances of
79 these parameters. Then select pairs for which the estimated bending momentum and the non-bending impact parameter at vertex are within given limits
80 taking into account the errors. Extrapolate the primary track candidates to station 4(5), look for at least one compatible cluster to validate them
81 and recompute the track parameters and covariances.
82 - Remove the identical track candidates (i.e. the ones sharing exactly the same clusters), and the ones whose bending momentum and non-bending
83 impact parameter at vertex are out of given limits taking into account the errors.
84 - Propagate the track to stations 3, 2 then 1. At each station, ask the "ClusterServer" to provide clusters in the region of interest defined in
85 the reconstruction parameters. Select the one(s) compatible with the track and recompute the track parameters and covariances. Remove the track if
86 no good cluster is found or if its re-computed bending momentum and non-bending impact parameter at vertex are out of given limits taking into
88 - Remove the connected tracks (i.e. the ones sharing one cluster or more in stations 3, 4 or 5) keeping the one with the largest number of cluster
89 or the one with the lowest chi2 in case of equality. Then recompute the track parameters and covariances at each attached cluster (using the
90 so-called Smoother algorithm in the case of the "Kalman" tracking).
91 - Find the MC label from the label of each attached cluster (if available): more than 50% of clusters must share the same label, including 1 before
92 and 1 after the dipole. Set it to -1 when reconstructing real data or in case of failure.
93 - The reconstructed tracks are finally matched with the trigger tracks (reconstructed from the local response of the trigger) to identify the
94 muon(s) that made the trigger.
96 The new clusters to be attached to the track are selected according to their local chi2 (i.e. their transverse position relatively to the track,
97 normalized by the convolution of the cluster resolution with the resolution of the track extrapolated to the cluster location).
98 If several compatible clusters are found on the same chamber, the track candidate is duplicated to consider all the possibilities.
100 The last part of the tracking is the extrapolation of the reconstructed tracks to the vertex of the collision. The vertex position is measured
101 by the SPD (the Silicon Pixel layers of the ITS Detector). In order to be able to perform any kind of muon analysis, we need to compute the track
102 parameters assuming the muon has been produced in the initial collision as well as the track parameters in the vertex plane. The first set of
103 parameters is obtained by correcting for energy loss and multiple Coulomb scattering in the front absorber (we force the track to come from the
104 exact vertex position (x,y,z) by using the Branson correction), while the second one is obtained by correcting for energy loss only.
106 The final results of the reconstruction - from which we will perform the physical analyses, compute detector efficiencies and perform calibration
107 checks - are stored in objects of the class AliESDMuonTrack and saved in AliESD.root file. Three kinds of track can be saved: a tracker track
108 matched with a trigger track, a tracker track alone and a trigger track alone (unused data members are set to default in the last two cases).
109 The complete list of MUON data saved into ESD is given in section @ref rec_s5.
112 \section rec_s4 How to tune the muon reconstruction
114 Several options and adjustable parameters allow to tune the entire reconstruction. They are stored in the OCDB in the directory MUON/Calib/RecoParam.
115 However, it is possible to customize the parameters by adding the following lines in the reconstruction macro (runReconstruction.C):
117 AliMUONRecoParam *muonRecoParam = AliMUONRecoParam::Get...Param();
118 muonRecoParam->Use...();
119 muonRecoParam->Set...();
121 MuonRec->SetRecoParam("MUON",muonRecoParam);
124 Three sets of default parameters are available:
125 - <code>GetLowFluxParam()</code>: parameters for p-p collisions
126 - <code>GetHighFluxParam()</code>: parameters for Pb-Pb collisions
127 - <code>GetCosmicParam()</code>: parameters for cosmic runs
129 Every option/parameter can be set one by one. Here is the complete list of available setters:
130 - <code>SetCalibrationMode("mode")</code>: set the calibration mode: NOGAIN (only do pedestal subtraction),
131 GAIN (do pedestal subtraction and apply gain correction, but with a single capacitance value for all channels),
132 GAINCONSTANTCAPA (as GAIN, but with a channel-dependent capacitance value).
133 - <code>SetClusteringMode("mode")</code>: set the clustering (pre-clustering) mode: NOCLUSTERING, PRECLUSTER, PRECLUSTERV2, PRECLUSTERV3, COG,
134 SIMPLEFIT, SIMPLEFITV3, MLEM:DRAW, MLEM, MLEMV2, MLEMV3.
135 - <code>SetTrackingMode("mode")</code>: Set the tracking mode: ORIGINAL, KALMAN.
136 - <code>CombineClusterTrackReco(flag)</code>: switch on/off the combined cluster/track reconstruction
137 - <code>SaveFullClusterInESD(flag, % of event)</code>: save all cluster info (including pads) in ESD, for the given percentage of events
139 - <code>SelectOnTrackSlope(flag)</code>: switch to select tracks on their slope instead of impact parameter at vertex and/or bending momentum.
140 - <code>SetMinBendingMomentum(value)</code>: set the minimum acceptable value (GeV/c) of track momentum in bending plane
141 - <code>SetMaxBendingMomentum(value)</code>: set the maximum acceptable value (GeV/c) of track momentum in bending plane
142 - <code>SetMaxNonBendingSlope(value)</code>: set the maximum value of the track slope in non bending plane (used when selecting on track slope).
143 - <code>SetMaxBendingSlope(value)</code>: set the maximum value of the track slope in non bending plane (used when selecting on track slope).
144 - <code>SetNonBendingVertexDispersion(value)</code>: set the vertex dispersion (cm) in non bending plane (used for the original tracking and to
145 select track on their non-bending impact parameter at vertex).
146 - <code>SetBendingVertexDispersion(value)</code>: set the vertex dispersion (cm) in bending plane (used for the original tracking, to compute the
147 error on the estimated bending momentum at the very begining and to select track on their bending impact parameter at vertex (used when B=0)).
148 - <code>SetMaxNonBendingDistanceToTrack(value)</code>: set the maximum distance to the track to search for compatible cluster(s) in non bending
149 direction. This value is convoluted with both the track and the cluster resolutions to define the region of interest.
150 - <code>SetMaxBendingDistanceToTrack(value)</code>: set the maximum distance to the track to search for compatible cluster(s) in bending direction
151 This value is convoluted with both the track and the cluster resolutions to define the region of interest.
152 - <code>SetSigmaCutForTracking(value)</code>: set the cut in sigma to apply on cluster (local chi2) and track (global chi2) during tracking
153 - <code>ImproveTracks(flag, sigma cut)</code>: recompute the local chi2 of each cluster with the final track parameters and removed the ones that
154 do not pass a new quality cut. The track is removed if we do not end with at least one good cluster per requested station and two clusters in
155 station 4 and 5 together whatever they are requested or not.
156 - <code>ImproveTracks(flag)</code>: same as above using the default quality cut
157 - <code>SetSigmaCutForTrigger(value)</code>: set the cut in sigma to apply on track during trigger hit pattern search
158 - <code>SetStripCutForTrigger(value)</code>: set the cut in strips to apply on trigger track during trigger chamber efficiency
159 - <code>SetMaxStripAreaForTrigger(value)</code>: set the maximum search area in strips to apply on trigger track during trigger chamber efficiency
160 - <code>SetMaxNormChi2MatchTrigger(value)</code>: set the maximum normalized chi2 for tracker/trigger track matching
161 - <code>TrackAllTracks(flag)</code>: consider all the clusters passing the sigma cut (duplicate the track) or only the best one
162 - <code>RecoverTracks(flag)</code>: during the tracking procedure, if no cluster is found in station 1 or 2, we try it again after having removed
163 (if possible with respect to the condition to keep at least 1 cluster per requested station) the worst cluster attached in the previous station
164 (assuming it was a cluster from background).
165 - <code>MakeTrackCandidatesFast(flag)</code>: make the primary track candidates formed by cluster on stations 4 and 5 assuming there is no
166 magnetic field in that region to speed up the reconstruction.
167 - <code>MakeMoreTrackCandidates(Bool_t flag)</code>: make the primary track candidate using 1 cluster on station 4 and 1 cluster on station 5
168 instead of starting from 2 clusters in the same station.
169 - <code>ComplementTracks(Bool_t flag)</code>: look for potentially missing cluster to be attached to the track (a track may contain up to 2
170 clusters per chamber do to the superimposition of DE, while the tracking procedure is done in such a way that only 1 can be attached).
171 - <code>RemoveConnectedTracksInSt12(Bool_t flag)</code>: extend the definition of connected tracks to be removed at the end of the tracking
172 procedure to the ones sharing one cluster on more in any station, including stations 1 and 2.
173 - <code>UseSmoother(Bool_t flag)</code>: use or not the smoother to recompute the track parameters at each attached cluster
174 (used for Kalman tracking only)
175 - <code>UseChamber(Int_t iCh, Bool_t flag)</code>: set the chambers to be used (disable the clustering if the chamber is not used).
176 - <code>RequestStation(Int_t iSt, Bool_t flag)</code>: impose/release the condition "at least 1 cluster per station" for that station.
177 - <code>BypassSt45(Bool_t st4, Bool_t st5)</code>: make the primary track candidate from the trigger track instead of using stations 4 and/or 5.
178 - <code>SetHVSt12Limits(float low, float high)</code>: Set Low and High threshold for St12 HV
179 - <code>SetHVSt345Limits(float low, float high)</code>: Set Low and High threshold for St345 HV
180 - <code>SetPedMeanLimits(float low, float high)</code>: Set Low and High threshold for pedestal mean
181 - <code>SetPedSigmaLimits(float low, float high)</code>: Set Low and High threshold for pedestal sigma
182 - <code>SetGainA1Limits(float low, float high)</code>: Set Low and High threshold for gain a0 term
183 - <code>SetGainA2Limits(float low, float high)</code>: Set Low and High threshold for gain a1 term
184 - <code>SetGainThresLimits(float low, float high)</code>: Set Low and High threshold for gain threshold term
185 - <code>SetPadGoodnessMask(UInt_t mask)</code>: Set the goodness mask (see AliMUONPadStatusMapMaker)
186 - <code>ChargeSigmaCut(Double_t value)</code>: Number of sigma cut we must apply when cutting on adc-ped
187 - <code>SetDefaultNonBendingReso(Int_t iCh, Double_t val)</code>: Set the default non bending resolution of chamber iCh
188 - <code>SetDefaultBendingReso(Int_t iCh, Double_t val)</code>: Set the default bending resolution of chamber iCh
189 - <code>SetMaxTriggerTracks(Int_t val)</code>: Set the maximum number of trigger tracks above which the tracking is cancelled
190 - <code>SetMaxTrackCandidates(Int_t val)</code>: Set the maximum number of track candidates above which the tracking abort
192 We can use the method Print("FULL") to printout all the parameters and options set in the class AliMUONRecoParam.
194 RecoParams can be put into OCDB using the MakeMUONSingleRecoParam.C or MakeMUONRecoParamArray.C macros.
196 \section rec_s5 ESD content
198 Three kinds of track can be saved in ESD: a tracker track matched with a trigger track, a tracker track alone and a trigger track alone (unused
199 data members are set to default values in the last two cases). These tracks are stored in objects of the class AliESDMuonTrack. Two methods can be
200 used to know the content of an ESD track:
201 - <code>ContainTrackerData()</code>: Return kTRUE if the track contain tracker data
202 - <code>ContainTriggerData()</code>: Return kTRUE if the track contain trigger data
204 The AliESDMuonTrack objects contain:
205 - Tracker track parameters (x, theta_x, y, theta_y, 1/p_yz) at vertex (x=x_vtx; y=y_vtx)
206 - Tracker track parameters in the vertex plane
207 - Tracker track parameters at first cluster
208 - Tracker track parameter covariances at first cluster
209 - Tracker track global informations (track ID, chi2, number of clusters, cluster map, MC label if any)
210 - TClonesArray of associated clusters stored in AliESDMuonCluster objects
211 - Trigger track informations (local trigger decision, strip pattern, hit pattern, ...)
212 - Chi2 of tracker/trigger track matching
214 The AliESDMuonCluster objects contain:
215 - Cluster ID providing information about the location of the cluster (chamber ID and DE ID)
216 - Cluster position (x,y,z)
217 - Cluster resolution (sigma_x,sigma_y)
221 - TClonesArray of associated pads stored in AliESDMuonPad objects for a given fraction of events
223 The AliESDMuonPad objects contain:
224 - Digit ID providing information about the location of the digit (DE ID, Manu ID, Manu channel and cathode)
225 - Raw charge (ADC value)
227 - One saturation bit and one calibration bit to say whether it is saturated/calibrated or not
230 \section rec_s6 Conversion between MUON/ESD objects
232 Every conversion between MUON objects (AliMUONVDigit/AliMUONVCluster/AliMUONTrack) and ESD objects
233 (AliESDMuonPad/AliESDMuonCluster/AliESDMuonTrack) is done by the class AliMUONESDInterface. There are 2 ways of using this class:
235 1) Using the static methods to convert the objects one by one (and possibly put them into the provided store):
236 - Get track parameters at vertex, at DCA, ...:
239 AliESDMuonTrack* esdTrack = esd->GetMuonTrack(iTrack);
240 AliMUONTrackParam param;
241 AliMUONESDInterface::GetParamAtVertex(*esdTrack, param);
244 - Convert an AliMUONVDigit to an AliESDMuonPad:
247 AliMUONVDigit *digit = ...;
248 AliESDMuonPad esdPad;
249 AliMUONESDInterface::MUONToESD(*digit, esdPad);
252 - Convert an AliMUONLocalTrigger to a ghost AliESDMuonTrack (containing only trigger informations):
255 AliMUONLocalTrigger* locTrg = ...;
256 AliMUONTriggerTrack* triggerTrack = ...;
257 AliESDMuonTrack esdTrack;
258 AliMUONESDInterface::MUONToESD(*locTrg, esdTrack, trackId, triggerTrack);
261 - Convert an AliESDMuonTrack to an AliMUONTrack:
264 AliESDMuonTrack* esdTrack = esd->GetMuonTrack(iTrack);
266 AliMUONESDInterface::ESDToMUON(*esdTrack, track);
269 - Add an AliESDMuonTrack (converted into AliMUONTrack object) into an AliMUONVTrackStore:
272 AliESDMuonTrack* esdTrack = esd->GetMuonTrack(iTrack);
273 AliMUONVTrackStore *trackStore = AliMUONESDInteface::NewTrackStore();
274 AliMUONTrack* trackInStore = AliMUONESDInterface::Add(*esdTrack, *trackStore);
277 2) Loading an entire ESDEvent and using the finders and/or the iterators to access the corresponding MUON objects:
278 - First load the ESD event:
280 AliMUONESDInterface esdInterface;
281 esdInterface.LoadEvent(*esd);
284 - Get the track store:
286 AliMUONVTrackStore* trackStore = esdInterface.GetTracks();
289 - Access the number of digits in a particular cluster:
291 Int_t nDigits = esdInterface.GetNDigitsInCluster(clusterId);
294 - Find a particular digit using its ID:
296 AliMUONVDigit *digit = esdInterface.FindDigit(digitId);
299 - Find a particular cluster in a given track using their IDs:
301 AliMUONVCluster* cluster = esdInterface.FindCluster(trackId, clusterId);
304 - Iterate over all clusters of a particular track using an iterator:
306 TIterator* nextCluster = esdInterface.CreateClusterIterator(trackId);
307 while ((cluster = static_cast<AliMUONVCluster*>(nextCluster()))) {...}
310 Note: You can change (via static method) the type of the store this class is using:
312 AliMUONESDInterface::UseTrackStore("name");
313 AliMUONESDInterface::UseClusterStore("name");
314 AliMUONESDInterface::UseDigitStore("name");
315 AliMUONESDInterface::UseTriggerStore("name");
319 \section rec_s7 ESD cluster/track refitting
321 We can re-clusterize and re-track the clusters/tracks stored into the ESD by using the class AliMUONRefitter. This class gets the MUON objects
322 to be refitted from an instance of AliMUONESDInterface (see section @ref rec_s6), then uses the reconstruction framework to refit the
323 objects. The reconstruction parameters are still set via the class AliMUONRecoParam (see section @ref rec_s5). The initial data are not changed.
324 Results are stored into new MUON objects. The aim of the refitting is to be able to study effects of changing the reconstruction parameter, the
325 calibration parameters or the alignment without re-running the entire reconstruction.
327 To use this class we first have to connect it to the ESD interface containing MUON objects:
329 AliMUONRefitter refitter;
330 refitter.Connect(&esdInterface);
334 - Re-clusterize the ESD clusters using the attached ESD pads (several new clusters can be reconstructed per ESD cluster):
336 AliMUONVClusterStore* clusterStore = refitter.ReClusterize(iTrack, iCluster);
337 AliMUONVClusterStore* clusterStore = refitter.ReClusterize(clusterId);
340 - Re-fit the ESD tracks using the attached ESD clusters:
342 AliMUONTrack* track = refitter.RetrackFromClusters(iTrack);
343 AliMUONVTrackStore* trackStore = refitter.ReconstructFromClusters();
346 - Reconstruct the ESD tracks from ESD pads (i.e. re-clusterize the attached clusters). Consider all the combination of clusters and return only
349 AliMUONTrack* track = refitter.RetrackFromDigits(iTrack);
350 AliMUONVTrackStore* trackStore = refitter.ReconstructFromDigits();
353 The macro MUONRefit.C is an example of using this class. The results are stored in a new AliESDs.root file.
356 This chapter is defined in the READMErec.txt file.