Central Fiber Tracker Resolution (in Wet and Dry Air)

Please refer to the CFT homepage for a general description of the Central Fiber Tracker project. The CFT Trigger homepage also has a lot of information.

down to:

back to:

Introduction

General System Test

Because the production of the Front End boards for the CFT has fallen too far behind schedule, it will not be possible to test the performance of the system in the real detector before we have beam. Clearly, it is important to test somewhere, at least once, what the performance of the CFT is likely to be in order to avoid (if possible) any huge problems. The prototype cylinder 3b was tested

The Wet and Dry Air Issue

The D0 silicon vertex detector dissipates about 3600W of power which is removed with -10oC water. Since this temperature is low enough to cause dew point condensation or even ice formation, it was decide to run the tracker in dry air (100% nitrogen). Ice formation and dew point condensation are serious concerns. Unfortunately, there is a conflict: the CFT needs to run in humid air with about 30% relative humidity because it was built in humid air. The builders of the CFT are convinced that drying it out will cause the fiber positions to shift enough to make the map which was made of the exact fibers positions useless (hence the only way to do alignment will be using tracks). Perhaps even more alarming, there is a CFT group prediction that the internal stress on the 1st CFT cylinder dried out will be at 80% of the value needed to physically (and therefore irreversibly) break it.

To test how drying out the CFT might change it's performance, the Lab 3, cylinder 3, proto-type CFT was dried out, and data was taken as it dried to check for deformation. Preliminary analysis results follow; they should be repeated once the new CMM data is available.

Results: (skip down to the resolution section to see plots)

  1. There are 10 micron global shifts in the mean position of the fibers as it dried out which are consistent with some kind of rotation of the cylinder (local? global?). Great care was taken not to move the cylinder and to keep the temperature constant during the tests in order to not introduce systematic shifts, so these shifts might be related to dryness.
  2. When one looks at the profile distributions (which are sensitive to the tails of the distribution) along the length of the ribbons, there are local shifts of up to 100 microns; conversely, when one fits Gaussians to the peaks, the local shifts are much smaller.
  3. The resolution improved slightly with dryness.

My preliminary conclusion is that something was happening as the cylinder dried out, but I can't yet say exactly what.

back to the top

Setup

Clyinder 3b was a prototype cylinder. The setup consisted of flat ribbons on the top and bottom of the cylinder. The flat ribbons were used for tracking are were not dried out or moved during the test.

back to the top

Cluster Definition

Here is a blown-up view of a section of a ribbon. Notice how there are two, half-fiber offset layers with space in between them. The spacing between the fibers was obtained using a mold and is maintained with glue. I've drawn a line which delineates the basic cell or "local fiber coordinates". Clearly, a muon trasvering the ribbon vertically will either hit one or two fibers, often called a singlet or doublet, respectively, by the CFT group.

For my analysis, there is no need to make a distincition between signlets and doublet, one simple looks for local maxima and checks if the local maxima passes a threshold cut or not (Incidentially, my alogrithm is slightly more efficiency than the Level 1 trigger algorithm which requires a strict isolation cut on the two neighbors making it more sensitive to noise). Nevertheless, as a cross-check, one can look at the local fiber coordinates and clearly see when the "singlets" and "doublets" are.

PLOT OF:Local Fiber Coordinates

The singlet peak, in the previous plot, centered at zero is what one would expect from simple geometrical considerations.

For reference, here is a plot of a typical PE Distribution (for this particular plot, there was a 2 PE pre-selection cut made to suppress the noise).

back to the top

Resolution versus Dryness

The following table summarizes the global offsets and resolutions versus the number of days that the cylinder was dried. The fit error on all the number is the table is about 1 micron.

timeTop Ribbon OffsetTop Ribbon Resolution Bottom Ribbon OffsetBottom Ribbon Resolution
humid, day 0 01110122
dry, day 10 -111076121
dry, day 25 -51055120

Now, after seeing the global shifts, one can look at the shifts along the length of a ribbon:

PLOT OF:Mean of residual versus the ribbon length (i.e. a Profile histogram).

PLOT OF:Gaussian fit to the mean of the residuals versus the ribbon length (i.e. a FitSlice histogram).

In the previous plots, one sees how the ribbons are not perfectly straight. There is structure. What is important, however, is the relative shifts. The profile method shows large local shifts whereas the fitslice method does not. At present, I do not understand the difference; it must somehow be related to the tails of the distribution. Hopefully, once the new CMM data is ready, the results can be clarified.

back to the top

Light Yield

In the axial layers, the odd fibers are the ones at smaller radius. The means given in the following tables are in units of photoelectrons. There appears to be know even and odd structure, i.e. the upper and lower layers are showing approximately the same light yield.
time / fibers L1 (flat)L2 (flat)L3 (axial)L4 (axial)L4 (axial)L5 (flat)
day 0 / even 7.42 7.53 7.26 6.98 7.82 6.56
day 0 / odd 7.20 7.52 7.40 6.96 7.69 6.42
day 10 / even 7.24 7.52 7.63 7.47 7.61 6.40
day 10 / odd 7.01 7.48 7.73 7.28 7.41 6.27
day 25 / even 7.22 7.49 7.59 7.22 7.53 6.53
day 25 / odd 7.06 7.39 7.70 7.15 7.40 6.39

Here are a few plots of the # of photo-electrons versus fiber # for all five of the layers:
L1 Flat ribbon
L2 Flat ribbon
L3 Cylinder 3b Axial
L4 Cylinder 3b Axial
L5 Flat ribbon

back to the top

Efficiency .vs. Light Yield

PLOT OF:Efficiency versus Light Yield in units of photoelectrons.

What one extracts from the previous plot is that the noise level must be low enough so that we can set a 1 PE threshold. Due to cassette production constraints, different sets of cassettes had to be used during the three runs. That may be the reason that the efficiencies vary for the three runs. Each VLPC has its own characteristic gain and spread of gain, and unless all of the bais voltages are correctly set for high efficiency (which can mean higher noise), the efficiency can suffer.

back to the top

Misc. Details

There were 3 datasets taken:
Days Delta Days Condition Triggers Selected Preselected (strong) Preselected (weak)
Aug 22-24 0 humid 1383312 43266 75319 107923
Sept. 2-3 10 dry 1577729 51650 71212 99501
Sept. 16-18 25 dry 1388932 46118 61742 78900

Each ribbon is connected to 4 SVX chips. One does see occasional common mode noise on a chip by chip basis.

back to the top

Send comments and suggestion to brsmith@fnal.gov

Last modified: Dec 1, 2000