TPC-only sDCA vs. phi studies
G. Van Buren
July 9, 2007
*** INTRODUCTION:
I have examined the signed DCA (sDCA) as a function of azimuth (phi) for the east and west halves of the TPC separately for 2007 data. I divided phi into 72 bins and looked at low luminosity data from FF (run 8120057) and RFF (159048). Here are sDCA [cm] vs phi [radians] for east (left, red) and west TPC (right, black), where for each phi bin I have fit to a gaussian (FitSlicesY, using the range sDCA = [-0.5,0.5]). The upper plots are all 72 bins, the lower plots are "rebinned" with 3 bins -> 1 bin which helps you see the trends (note that rebinning by 3 bins causes the vertical scale to be incorrect by a factor of 3). I have some basic quality cuts on tracks/events: pt = [0.3,2.0], |eta|<1, |sDCA|<4.0, innerTPC hits>=5, outerTPC hits>=10, |z-vertex|<50.
sDCA vs. phi (FF)
A few things to note:
1) There are edge effects at the sector boundaries
2) There are some odd sectors (e.g. sector 18 (east at phi~4.5-5.0), probably due to RDO outage in that sector)
3) Some sectors show more of a systematic trend across their phi coverage than others (e.g. sectors 22-24 (east at phi~0.5-2.0))
4) There are some overall systematic trends (seen more clearly in the rebinned plots)
5) One can re-compose to say that east and west distributions come from common (west+east) and differential (west-east) source.
*** DIFFERENCE:
These plots were made after the twist distortion correction, so the difference between east and west due to that distortion should no longer be present. Below I show west-east for before (left) and after (right) twist correction (the curve represents my guess at what is "twist"):
sDCA vs. phi (FF, west-east)
We can see that the twist correction now in place reduces most of the west-east difference, but there seems to be some remaining. My initial interpretation of this is that there is some misalignment of either sectors 19, 20 or 4,5 (at phi=~5-6) and the calibration is compensating for it by leaving some remnant twist.
I don't think sector alignment (with respect to each other / the primary vertex) has been done since about the 2000/2001 time frame.
Here's the RFF data wth my best guess at the twist corrections (using 2005 Twist parameters with 2006 geometry, which is what came out of the FF calibration):
sDCA vs. phi (RFF, west-east)
Perhaps it takes the FF data to "guide the eye", but the RFF looks roughly similar, indicating that a misalignment in the phi=~5-6 range may be real.
*** SUM:
Here is the (west+east)/2 for FF (left) and RFF (right):
One small note: the RFF data is on average above zero, and I think this may be due to some small SpaceCharge ionization in the RFF low luminosity runs.
Clearly, there is a systematic azimuthal trend which is common between west and east TPC (it can be seen in the east and west separated plots above as well, so it's truly not due to just one side of the TPC), and is similar for FF and RFF!!! I have been brainstorming over what could be common between west and east and have some thoughts:
1) It cannot be an ExB distortion (central membrane or endcap tilts, for example) as that would flip sign with B field.
2) It is consistent with a primary vertex being reconstructed in the wrong place.
In both FF and RFF, the reco vertex is at (x,y) ~= (0.49,-0.17). If the true vertex were at ~ (0.053,-0.18) that would explain the trends.
3) It cannot be systematic sector misalignments as that would cause the vertex reconstruction to match the track projections (we would be blind to it here).
4) So it seems to me it must be a primary vertex reconstruction bias towards (0,0) and we're off by ~0.04 cm.
However, Minuit starts with a vertex seed of (0,0) and I tried setting it to (0.6,-0.2) and it only moved by 2e-6 cm. Moving the seed to (1.6,-0.6) moved the reconstructed vertex by 2e-5cm (0.2 microns). So it doesn't seem to me as though the Minuit primary vertex finder is significantly biased due to its seed.
Another possibility I thought of to explain a primary vertex bias is the use of the "dcaGeometry" of tracks based on their helix parameters at their DCA to (0,0). So I tried altering this from (0,0) to a few other values and was able to move the primary vertex by about 0.006 cm (60) microns. Interesting that it's a larger effect than the Minuit seed, but still not big enough to account for my problem.
So I'm left still wondering what this is...
*** CONCLUSIONS:
1) There appears to be a true sector misalignment at phi=~5-6. We need to make sure we know which side it's on, fix it, and re-run the twist calibration.
2) We need to figure out what's causing the common to east&west/FF&RFF azimuthal trend as this could cause some effects on alignment of silicon detectors.