r1 - 05 Nov 2007 - 09:56:34 - CarlosEscobarYou are here: TWiki >  Atlas Web  >  SiliconForwardTracker > SiAlignment > CSCGeneralStudies

CSC General Studies

Figures for fun

  • Some schematic figures presenting the features of the misalignment
    • Impact parameter shift after a global shift of the detector: jpg
    • Effect of the global shift in the reconstructed momentum: jpg
    • A clocking effect (layer rotation proportional to radius) bends the trakcs of positive and negative charged particles in a different way. The result is a bias in the momentum. jpg
    • A basic representation of he sagita effect on the reconstructed momentum. The figure presents how a sagita of a piece of the detector w.r.t. another may affect the reconstructed momentu. The picture represents two charge particles that are emmitted in oposite directions along the sagita direction. The one that flies in the direction of the sagita is reconstructed with a much higher momentum. Whilst the particle emerging in the oposite direction is reconstructed with a lower momentum. In both cases the saggita also introduces a no null impact parameter, and a change in the reconstructed phi0. This is a simple representation. The sagita has a rather more complicated effect on the momentum. For a given sagita, the overall effect depends on both Pt and phi0. jpg
    • Local X distributions after a global shift: jpg

Primary Vertex Refit

Follow the link to see our studies of the primary vertex for CSC: twiki InDetAlignVertex

Comments about Nominal vs Ideal reconstruction (as 12/Mar/07)

  • A comparison has been made for the following cases: 1) ideal geometry and error scaling off, 2) ideal geometry and error scaling on, 3) nominal geometry and error scaling off and finally 4) nominal geometry and error scaling on. The GlobalX2Fitter has been used as tracker, and tracks fitted with PIX and SCT only (no TRT). A track selection has been applied: at least 1 pixel and 4 SCT hits. Plus track is required to leave at least hit in the SCT barrel layer 3 (a way to select tracks within the SCT barrel). The numbers are extracted from 1000 CSC multimuons events in digits files.
    • track parameters: jpg
    • pixel phi residuals: jpg
    • SCT residuals: jpg
In summary: 1) when using the ideal geoemtry (so alignment known). Switching on and off the error scaling only affects to the width of the residuals. They got slighthly broad when error scalingis switched on. Almost same number of tracks is reconstructed. 2) When using the nominal geometry (so alignment unknown) the use of the error scaling has some effect in the number of reconstructed tracks, more with error scaling on, and the residuals look slightly narrower with error scaling off.

  • Check these plots presented in a Valencia-Oxford coordination meeting: pdf. In this file you may find a comparison of the track properties reconstructed using the nominal and ideal geometry (as they come no alignment yet), and using iPatRec and GlobalX2Fitter. What is presented is for all tracks delivered by the trackers. There is no further selection done by us. Besides, when the ideal geometry is used in the reconstruction then the error scaling is switched off. Whilst if the nominal geometry is used then the error scaling is switched on.

CSC: residuals (as 27/Mar/2007)

Recently there has been some mails by Florian and Alan about the CSC multimuons SCT residuals. What triggered the discussion was that the SCT residuals exhibit a side dependence, even in the case of the ideal geometry, which in principle in counter-intuitive. So not the same residuals are found for side 0 and side 1 in the same modules. One must say that we had noticed that effect time ago as it can be seen in the summary plots we produced for each of the iterations (each file contains more than 20 pages of plots).

We agree with Alan's explanation of the effect. So we must said that it is an inherent effect of the distorted CSC geometry. In fact a closer look to the SCT residuals allows to detect a dependence not only on the side but on the layer, and even more a eta-ring dependence within a layer. All that could be an effect of the built-in telescope effect (need confirmation).

  • Take a look to the z0 Vs eta plots:

It is also worth mention that some SCT residuals hic-up was observed in the SR1 cosmic simulations. As recently Grant has pointed out it turned to be some timing effect studied by Ricahrd Batley. This will provide an alternative explanation as one can not discard yet a reproduction of the timing effects on the CSC samples.

In order to verify if the observed SCT residuals "distortion" is an effect induced by the CSC geometry one has checked the sample of MultiMuons generated by Sergio some time ago (html more info here). Those MultiMuons have bassically the same generation parameters. However the ID geometry is different as it uses the ATLAS-DC3-02 geometry. So in Sergio's MultiMuons the detectors were kept at their nominal position. By repeating the same SCT residuals analysis our finding is that with the Sergio's MultiMuons SCT residuals look fine, contrary to what is observerd with the CSC multimuons (even using the ideal geometry). The same can be seen in the z0 Vs eta plots. There is a dependence for the CSC case for the nominal setup and for the ideal one, so this tells us that this is an intrinsic effect. This effect doesn't exist for the DC3 sample.

Interested reader can have a look to this pdf file pdf which shows a map of SCT residuals per plane for Sergio's MultiMuons and CSC MultiMuons, in both cases with a perfect knowledge of the geoemtry (module's location). In the first case the plots have been obtained with 25K events, while in the later 100K events. So that is why the first set of plots have a bit more of statistical fluctuation from one module to another but no general behaviour trend is seen, while in the case of the CSC MultiMuons a trend is clearly seen depending on the layer, eta ring and side.

* Residuals hangover (as 08/Feb/2007)

During the last IDSW workshop we presented the residuals (PIX and SCT) obtained when running over the CSC multimuons sample (pdf). It was observed that the residuals of the silicon (pixel and SCT) modules were different when using newTracking and iPatRec. This is now understood in terms of the biased vs unbiased residuals. In few words: while iPatRec residuals were biased, newTracking ones were unbiased.

We have processed 100K events of the CSC multimuons sample (7270) using the digits files. The exercise has been repeated 4 times with:

  • (1) iPatRec, using the ideal alignment and no error scaling
  • (2) newTracking with GlobalX2Fitter, ideal alignment and no error scaling
  • (3) iPatRec and using nominal positions for the modules and error scaling
  • (4) newTracking with GlobalX2Fitter, nominal positions and error scaling

Ideal means knowing the final positions of the modules while nominal means ignoring the real positions of the modules by placing them at their nominal location. So in a certain way ideal means perfectly aligned while nominal means misaligned. For obvious reasons no error scaling is needed for the ideal geometry, while it is needed in order to keep a decent track finding efficiency when using the nominal setup.

Tracks have been reconstructed with Pixels and SCT only (TRT excluded). It must also be said that we have used a minimum track quality cut by requiring the track to have associated at least 1 pixel hit and 4 SCT hits.

The job produces a TrkValidation.root file (with a FitterValidation/OutputTrack branch). We also produce the Richard Hawkings' ntuple (idaling). We have checked that: 1) both ntuples contain exactly the same number of tracks and 2) the residuals shape look the same. All the plots here presented are actually extracted from the idalign.root file. So in this way we had an easy acces to biased residuals for both tracking options.

In what follows we present a comparison of the results obtained for both trackers: iPatrec and newTracking. The histograms settings (range and bins) are exactly the same. They were produced running over the same sample of 100K multimuons events. They are printed exactly in the same frame. If one of them seems to have less entries is simply because it has less entries (either due to less reconstructed tracks or due to less asociated hits).

Figures 1 and 2 present a comparison of the track properties when using alternatively both tracking methods (iPatRec and newTraking-GlobalX2Fitter). Figure1 shows the result for the ideal case, while figure 2 for the nominal case. So in the ideal case (figure 1), one can appreciate an almost perfect matching between both trackers. In the nominal case (figure 2) one can appreciate some differences. The main one is that iPatRec has more entries so a higher track efficiency (iPatRec 93%, newTracking 86%).

  • jpg Figure 1: track properties when using the ideal geometry (perfectly aligned).

  • jpg Figure 2: track properties when using the nominal geometry (misaligned).

The PIX phi residuals dsitribuitons in the ideal case presented in figure 3. Figure 4 presents the SCT ones. Again, one can see an almost perfect mathcing. So in conclusion one can say that both trackers perform equally well in the case of a perfect kwonledge of the geometry.

  • jpg Figure 3: PIX phi residuals for the ideal.

  • jpg Figure 4: SCT residuals for the ideal case.

Now figures 5 and 6 present the PIX phi and SCT resiudals in the case of the nominal setup. Obviously the distributions have become broader, reflecting the ignorance of the geometry. In general iPatRec residuals and newTracking look similar, but there are certain differences specially in the pixels. This fact is due to the less number of PIX hits that the newTracking founds associated to the track. It has been checked that the narrow peak at PIX phi residuals = 0 corresponds to the tracks wiht only 1 associated PIX hit.

  • jpg Figure 5: PIX phi residuals for the nominal case.

  • jpg Figure 6: SCT residuals for the nominal case.

It would be nice if other users of the CSC sample can confirm or deny this observations and the tracking experts could help us in understanding the hit association and the resiudal distributions.

Reconstruction of simulated Events

The CSC simulated events are created with Tag ATLAS-CSC-01-02-00 (misaligned and distorted material). To reconstruct the digits we have two options:

Option 1: Nominal Geometry - Misalignment Unknown

  # Detector Description
  DetDescrVersion = "ATLAS-CSC-01-00-00"
  IOVDbSvc = Service("IOVDbSvc")

Option 2: Misaligned Geometry - Misalignment Known

  # Detector Description
  DetDescrVersion = "ATLAS-CSC-01-02-00"
  IOVDbSvc = Service("IOVDbSvc")

There is no need to specify a GlobalTag in this case. The conditions database tag will be the same as use in the simulation regardless of which geometry tag is set.

See SettingGeometryVersion for more details

-- CarlosEscobar - 05 Nov 2007

Edit | WYSIWYG | Attach | PDF | Raw View | Backlinks: Web, All Webs | History: r1 | More topic actions
Powered by TWiki
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback