INITIAL PERFORMANCE STUDIES OF THE FORWARD GEM TRACKER A THESIS SUBMITTED TO THE GRADUATE SCHOOL IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

Transcription

1 INITIAL PERFORMANCE STUDIES OF THE FORWARD GEM TRACKER A THESIS SUBMITTED TO THE GRADUATE SCHOOL IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE MASTER OF SCIENCE BY MALORIE R. STOWE DR. DAVID GROSNICK BALL STATE UNIVERSITY MUNCIE, INDIANA DECEMBER 2012

2 Table of Contents List of Figures iii List of Tables vii Abstract viii Chapter I Introduction Chapter II Detector Chapter III Data Analysis III.A Cosmic Ray Data: High Voltage Off III.B Cosmic Ray Data: High Voltage On III.C STAR Data Chapter IV Conclusions Appendix A Acknowledgements References ii

3 List of Figures Figure 1 Feynman diagram showing the interaction between quarks and antiquarks creating W bosons Figure 2 The facilities of Brookhaven National Laboratory.. 7 Figure 3 The STAR Detector Figure 4 Gem foil hole pattern Figure 5 Gem foil holes and electric field lines Figure 6 Identification format of the disks and quadrants of the FGT. 11 Figure 7 Example of quadrant strips Figure 8 The first disk of the FGT labeled with quadrant names, FEE assemblies, and APVs Figure 9 R and φ strip locations read by a single APV Figure 10 Cosmic-Ray Test Setup Figure 11 Pedestal histogram of channel number 048, time bin 02, and APV 02 with Gaussian fit Figure 12 Total ADC distribution of the Gaussian fit-means for all combined channels, time bins, and APVs Figure 13 Distribution of Gaussian fit-mean per time bin.. 19 Figure 14 Average of Gaussian fit-mean distribution per time bin. 19 Figure 15 Difference in average pedestal fit-mean values across the time bins for different channels in APV Figure 16 Fit-means from time bin 00 subtracted from time bin 04 as a function of channel number iii

4 Figure 17 Pedestal fit-mean values for time bin 05 subtracted from time bin 04 fit-means as a function of channel number.. 22 Figure 18 Pedestal fit-mean distribution per APV Figure 19 The mean of APV distribution as a function of APV.. 24 Figure 20 Total pedestal fit-sigma distribution Figure 21 Fit-sigma distribution per time bin Figure 22 Mean of pedestal fit-sigma distribution for each time bin. 26 Figure 23 Fit-sigma distribution as a function of APV Figure 24 Mean from fit-sigma distributions per APV as a function of APV number Figure 25 Examples of poor Gaussian fits to the pedestal.. 28 Figure 26 Figure 27 Pedestal width ADC distribution after common mode noise subtraction Raw ADC spectrum of 128 channels from APV01 and time bin Figure 28 APV channel number as a function of ADC channel.. 31 Figure 29 Histogram of ADC values for time bin 04 and APV Figure 30 APV channel number as a function of ADC channel.. 32 Figure 31 Histogram of ADC data with the six-sigma cut applied. 33 Figure 32 Ratio of the signal to total entries as a function of the location on the ADC pedestal cut Figure 33 Mean from pedestal APV fit-mean distribution.. 35 Figure 34 Fit-mean distribution for time bin Figure 35 Fit-sigma distribution for time bin iv

5 Figure 36 ADC channel as a function of time bin Figure 37 Figure 38 Difference in pedestal means as a function of channel number in APV Difference in pedestal means as a function of APV channel number for APV Figure 39 Pedestal fit-mean as a function of electronic ID.. 39 Figure 40 Pedestal fit-sigma as a function of electronic ID.. 40 Figure 41 Map of R and φ strips for APV Figure 42 Figure 43 Figure 44 Pedestal fit-means and fit-sigmas as a function of strip length in the phi plane for APV Pedestal fit-means and fit-sigmas as a function of strip length in the R plane for APV ADC distribution showing a typical pedestal for channel 095, time bin 01, and APV Figure 45 ADC distributions of bad pedestals considered good. 44 Figure 46 ADC distribution showing the pedestal for channel 057, time bin 02, and APV Figure 47 Frequency plot of the ratio of RMS to the mean of the histogram Figure 48 Pedestal fit-means as a function of time Figure 49 Pedestal fit-sigmas as a function of time Figure 50 Figure 51 Pedestal fit-mean as a function of voltage added to nominal high voltage Mean (of the histogram) of the fit-mean distribution per time bin and per APV as a function of voltage added to nominal high voltage v

6 Figure 52 Figure 53 Figure 54 Figure 55 Figure 56 Pedestal fit-sigma as a function of voltage added to nominal high voltage Mean of fit-sigma distribution per time bin and per APV as a function of voltage added to nominal high voltage Average of the mean of the histogram from the fit-sigma distribution per APV as a function of voltage added to nominal high voltage Frequency plot of 70:30 gas mixture pedestal fit-means divided by 90:10 gas mixture pedestal fit-means.. 52 Frequency plot of 70:30 gas mixture pedestal fit-sigma divided by 90:10 gas mixture pedestal fit-sigma.. 53 vi

7 List of Tables Table I Average and standard deviation for the pedestal fit-mean values of the time bins subtracted from pedestal fit-means of time bin 04 for all channels of APV Table II Failure States for pedestals Table III Failure rates for the pedestals Table IV Nominal high voltage values per quadrant Table A.I Cosmic-ray test data Table A.II STAR data Table A.III Pedestal study in time Table A.IV Pedestal study with change in high voltage Table A.V Pedestal study with change in gas mixture vii

8 Chapter I Introduction The proton is a subatomic particle, comprised of gluons, sea quarks, and three valence quarks. The proton is known to have a fundamental property called spin, with the spin of a proton measured to be ½ћ. It is natural to expect that the spin of the three valence quarks would make up the spin of the proton, but measurements made in deepinelastic scattering (DIS) experiments have shown that these quarks only supply about 30% of the total spin of the proton [1]. This discovery led to what is known as the proton spin crisis, and the need to find out what constituents are responsible for the remaining amount of spin. The rest of the proton spin is thought to be produced by several other constituents, including the sea quarks and antiquarks, gluons, and the orbital angular momentum of all these potential contributors. To investigate the contribution of these other constituents to the spin of the proton, experiments are conducted at the Relativistic Heavy Ion Collider (RHIC) at Brookhaven National Laboratory. RHIC has the ability to produce, accelerate, and collide polarized-proton beams at energies of 500 GeV in the center of mass. It has been the goal of the RHIC spin physics program to measure the contribution of gluons and various sea quarks to the spin of the proton. This goal may be accomplished by 1

9 using large particle detectors, such as the Solenoidal Tracker at RHIC (STAR) to study collisions of polarized protons. The resulting particles created in the collisions provide a probe into the proton to investigate each of the possible spin contributors [2]. For example, measuring the asymmetry, which is the difference between the numbers of particles in two different spin states divided by the sum of the number of particles in the two different spin states, of direct photons produced in polarized proton collisions will provide information to calculate the gluon contribution to the spin. The Forward GEM Tracker (FGT) is a detector that will be used to help determine the spin contribution of the sea quarks by measuring the asymmetry of the production of W bosons from collisions at 500 GeV center of mass. Specifically, the FGT will improve previous tracking capabilities of leptons from W boson decays in the STAR detector. The FGT operates by using gas electron multiplication (GEM) technology. Charged particles (e.g. leptons) ionize the material they traverse and the liberated electrons from this process are multiplied during collisions with gas-mixture atoms within a given electric field. All of these electrons produce a pulse on two planes of strips, providing two-dimensional information on the hit position of the charged particles. Six sets of these planes (disks) are located at distances from the intersection region. Combining hits in each disk allows the track of a lepton to be reconstructed. The FGT is a newly-constructed detector in STAR and therefore its initial performance needs to be studied in great detail. The FGT will be instrumental in observing the decay leptons from W bosons, created by the interaction of the quark of one proton and the sea antiquark of another 2

10 in a proton-proton collision with a center of mass (CM) energy of 500 GeV. This high CM energy is needed to create W bosons, which have a mass around 80 GeV. Figure 1 shows that the W boson formed during the collision depends on the quark/antiquark interaction of the colliding protons. If a valence down quark of one proton interacts in a collision with an up antiquark of another proton, it can produce a W -. A W + is created when a valence up quark of a proton interacts with a down antiquark of another proton. These W bosons survive for only a fraction of a second until they decay into a lepton and a neutrino (about 30% of the time). A W - decays to an electron and an antineutrino, where a W + decays to a positron and a neutrino [3]. Neutrinos are very rarely detected, so a signature in the detector for a W boson is a single, isolated lepton with an energy of one-half of the W boson mass in its CM frame. The other one-half of undetected missing mass of energy is carried off by the neutrino at 180 o from the lepton in the CM of the W boson. The lepton energy is measured by calorimeters. The sign of the lepton can be recognized by the bend of its trajectory in the magnetic field of the STAR Fig. 1. Feynman diagrams showing the interaction between quarks and antiquarks creating W bosons. 3

11 detector. By knowing the sign of the lepton s charge, the W boson produced in the collision can be identified and then, ultimately which quark/antiquark interaction occurred (see Fig. 1) [4]. The asymmetry in the production of the W bosons from polarized-proton collisions can be measured, and from that, the spin contribution from the sea quarks can be calculated. The FGT is one part of a tracking upgrade for the STAR detector. Before the installation of the FGT, tracking (finding the trajectory of charged particles) was done primarily by the Time Projection Chamber (TPC) [5], which covers an angular range of, where η (in units of pseudorapidity) is the angle measured from the vertical axis, perpendicular to the beam line at the collision point, towards the incoming beam direction. As the angle η increases, the tracking capabilities of the TPC decrease due to a decrease in tracking volume (discussed in Chapter II). The installation of the FGT covers the increasing η range that is missed by the TPC [6]. This thesis reports on the analysis of data taken from the initial operation of the FGT. Chapter II provides a description of the production of high-energy polarized protons, the STAR detector, and specifically the FGT detector. Chapter III describes the analysis of the FGT performance, which includes investigating correlations of quantities, such as electronic pedestal position and width, measured with both high voltage off and on, using cosmic ray test data and data after the FGT was installed into STAR. Chapter IV discusses the conclusions from this analysis and the possibilities of future analysis. 4

12 Chapter II Detector The Relativistic Heavy Ion Collider (RHIC) at Brookhaven National Laboratory (BNL) was the first facility to collide heavy ions. Normally for heavy-ion experiments at RHIC, gold ions are accelerated, but polarized proton beams can also be accelerated and used in spin physics experiments [2]. For a proton beam to be polarized means that the proton spins are preferentially aligned in a particular direction. The polarizations of each beam are measured at several locations in the acceleration process by polarimeters, and the total polarization ranges on a scale from 0-100%. Typical polarization measurements are around 50%, meaning there is a 50% excess of protons with spin in a desired direction for the proton beam. The magnitude of the beam polarization dilutes the measured spin effects from polarized proton collisions, so it is important to have as large of a polarization value as possible. An asymmetry ε, as measured by a detector system, is related to the analyzing power A, which indicates the effect of spin on particle production through interactions, by the equation,, where P is the beam polarization. The asymmetry is defined as the difference between the number of particles measured in one state and those in an opposite or different state divided by the sum of those states, as described by 5

13 . (1) In Eq. (1), N + is the number of particles measured with beam proton spins aligned parallel, and N - is the number of particles measured with spins aligned antiparallel. Asymmetry measurements lead to an understanding of the spin effects in collisions, such as the spin contribution of constituents to the overall spin of the proton [7]. A diagram of the RHIC facility [8] is given in Fig. 2. Polarized protons are generated at more than protons per bunch at the optically-pumped polarized ion source [9] and are injected into a linear accelerator, where the protons are accelerated to 200 MeV, before being sent to the booster synchrotron. In the booster, the protons are accelerated to 1.5 GeV and then continue to the Alternating Gradient Synchrotron (AGS). In the AGS, polarized protons are accelerated to about 25 GeV. The protons are finally injected into RHIC and are split into two beams, called the blue and yellow beams. Protons in the blue beam travel clockwise around RHIC, while the protons in the yellow beam go counterclockwise. The protons reach relativistic speeds and each beam has an energy up to 255 GeV [10]. Helical magnets called Siberian snakes [11] are located in RHIC to keep both beams polarized. Each time the beam of protons pass through a snake magnet, the spins are rotated by 180 o to prevent accumulated depolarization as the proton beams travel around the RHIC ring. For half of the RHIC ring, the protons are polarized spin up and the other half, with spin down. Spin rotator magnets are used to longitudinally polarize the protons just before the intersection or collision points of the two beams [9]. 6

14 Fig. 2. The facilities of BNL showing the linear accelerator, the Booster Synchrotron, the Alternating Gradient Synchrotron (AGS), and RHIC [8]. The RHIC ring has five separate polarimeters to measure beam polarization, while the AGS, the Booster Synchrotron, and the linear accelerator each have a polarimeter for measuring the beam polarization during each step of the acceleration process. The RHIC ring is 2.4 miles in circumference and has up to six intersection points for the two beams, as shown in Fig. 2. The 6 o clock intersection point is the location of the STAR detector (Solenoidal Tracker at RHIC) [12]. There are several important components to the STAR detector: the Time Projection Chamber (TPC) [5], which is used for detecting particle tracks after the collision; and the Barrel Electromagnetic Calorimeter (BEMC) [13] and Endcap Electromagnetic Calorimeter (EEMC) [14], each of which measures the energy deposition of particles produced in proton collisions [10]. A 7

15 new addition to the STAR detector is the Forward GEM Tracker (FGT) [4], which also provides the tracking of charged particles from the intersection region in the forward direction. Figure 3 shows a cut-away view of the STAR detector with the collision point located in the center of the figure. One of the main goals of the experiment is to determine the quark/antiquark contribution to the spin of the proton, which comes from a measurement of the asymmetry in the production of W bosons. Both the FGT and TPC are needed to provide tracking within the magnetic field to identify the momentum and the sign of the lepton from the W boson decay. This selects the boson and hence which quark and Fig. 3. The STAR Detector. Components include the Forward GEM Tracker (FGT), the Time Projection Chamber (TPC), the Barrel Electromagnetic Calorimeter (labeled EMCal), and the Endcap Electromagnetic Calorimeter (labeled Endcap EMCal) [15]. 8

16 antiquark were involved in the interaction. Before the FGT was installed into STAR, tracking was done predominantly by the TPC, in which the tracking of particles after the collision was inadequate at large η (forward angles). The FGT provides the needed tracking at large η for charged particles before they reach the EEMC, where the energy deposited by the charged particle is measured. The FGT is comprised of six disks. Each disk is divided into four quadrants, and each quadrant consists of two high voltage planes with three GEM (gas electron multiplier) foils in between. A typical operating voltage of the high voltage planes is around 3800 V. The GEM foils are made with chemically-formed holes, each having a pitch of 140 μm. The holes have a double conical shape with a 50 μm inner diameter and a 70 μm outer diameter [4], as shown in Fig. 4. The GEM foil technology was originally developed at CERN [16], but the GEM foils used in the STAR FGT were manufactured by Tech-Etch Inc. [17]. The foils are made from a 5 μm copper coating on a 50 μm kapton insulator. The gas flowing through the holes is a mixture of argon and carbon dioxide (Ar-CO 2 ), and was designed to contain 70% argon and 30% carbon dioxide [18]. The gas in the FGT is monitored by a bubbler to indicate flow. Fig. 4. Gem foil hole pattern [4]. 9

17 The FGT detects charged particles using gas electron multiplier (GEM) technology [16]. When a charged particle traverses the FGT detector, it ionizes the gas, freeing electrons. These electrons drift in an electric field of the quadrant until reaching holes that are located in the GEM foils. A strong electric field is produced in each GEM foil hole, as seen by the density of electric field lines in Fig. 5. The electrons are accelerated in the electric field of the GEM foil holes and collide with more gas-mixture atoms, creating an electron multiplication. The electrons proceed through three layers of GEM foils, thus creating a large number of electrons. All of the electrons drift towards two planes of strips providing two-dimensional readout for data acquisition. As the electrons approach the readout plane, a charge is induced and creates the signal. The large number of electrons can induce a charge on many strips, and these strips can then be clustered to form a hit in the plane [19]. A cluster groups nearby strips with energy deposited on them. This cluster of energy in a quadrant indicates a hit in the FGT disk. By connecting hits from the other quadrants and disks, a particle track can be determined. Fig. 5. Gem foil holes and electric field lines (shown in red) [4]. 10

18 The FGT is designed to have six, triple GEM foil disks and each disk be divided into four quadrants. The GEM disks are numbered 1 through 6, with one being closest to the collision point and 6 being the farthest from the collision point. The quadrants are lettered A through D in a clockwise manner beginning with A at the one o clock position. Figure 6 indicates the naming convention and the placement of the GEM disks and quadrants around the beam line. Each quadrant contains ten APV25-S1 readout chips [21]. The signal pulse from the electron multiplication is sent to a preamplifier, inverter, and shaper within the APV chip to create an analog signal. The two-dimensional readout planes of the FGT have been divided into radial (R) and azimuthal (φ) strips, with each strip corresponding to one channel. Each APV chip serves 128 channels and each channel has seven time bins. One time bin holds readout data for each strip in a period of time of approximately 25 ns. The seven time bins can be used to map out pulse shapes and adjust timing. This design in the readout planes results in a total of 720 R strips and 560 φ strips for each quadrant [4]. Figure 7 gives an example of the strip arrangement for one quadrant. Fig. 6. Identification format of the disks and quadrants of the FGT [20]. 11

19 Fig. 7. Example of quadrant strips [22]. A front-end electronics (FEE) assembly, that contains the APV chips, is located at the border of adjacent quadrants in a disk. Each APV chip reads data from a defined region of R and φ strips. A differential analog pulse is driven from each APV chip to the APV readout module (ARM). In the ARM, the pulses are digitized by an analog digital converter (ADC). The data are collected from the ARMs by the APV readout controller (ARC) and then sent to the STAR data acquisition system. The ARC also sends commands to the ARM cards for setting up the APV chips. Figure 8 shows the first disk of the FGT with these pieces of hardware labeled. The time required to read out one time bin of each channel of one APV is approximately 7 μs. 12

20 Fig. 8. The first disk of the FGT labeled with quadrant names, FEE assemblies (asmb), ARMs, ARCs (shown as RDO), and APVs (numbers) [22]. Each quadrant uses ten APV chips and each chip reads data from a certain region of R and φ strips within the quadrant. Figure 9 gives two examples of the R and φ strip locations read by a single APV in a single quadrant. (a) (b) Fig. 9. R and φ strip locations read by a single APV. Shown are the strips read by (a) APV 0 and (b) APV 1 [22]. 13

21 The FGT was installed into STAR in 2011 with the first disk instrumented with all four quadrants and the remaining five disks with only two quadrants each. These two quadrant disks are aligned so that reconstruction of tracks can still be made. Data from this configuration were analyzed and described in the next chapter. 14

22 Chapter III Data and Analysis As the Forward GEM Tracker was being built, quadrants were individually tested using cosmic rays. The cosmic-ray test setup had one or more quadrants of the FGT placed in between two large-area scintillator detectors. Figure 10 shows a photograph of the FGT cosmic-ray test setup. The trigger consisted of a signal coincidence between the two scintillators, which caused the FGT quadrants to be read out. Scintillator Quadrant Quadrant Quadrant Scintillator Fig. 10. Cosmic-Ray Test Setup. The black arrows indicate the location of the quadrants and the red arrows indicate the location of the scintillator detectors [23]. 15

23 Data were analyzed to verify that the detector works, to investigate how it actually works, and to be sure the detector still operates in the same manner once it is installed into STAR. The analysis also involved looking for any trends or correlations that may exist between the channels, time bins, or APVs. These trends or correlations could arise within individual electronic channels or by features inherent in an APV or time bin. As an example, the order in which data are read within an APV chip may affect the electronic pedestal width. The first data sets were collected with the high voltage to the GEM foils turned off. This means that no signals from actual cosmic rays are read by the FGT. III.A Cosmic Ray Data. High Voltage Off The first priority in the analysis was to ensure that a given quadrant was working properly. One way to check this operation is to investigate the readout of the electronic pedestal. A pedestal is an ADC reading of a channel in the electronics, when no energy is deposited in the detector. Any signal from charged particles passing through the detector should deposit energy above the pedestal value. When the high voltage on the GEM foils is turned off, there is no signal produced by charged particles, so the only data comes from the electronics. Figure 11 shows an ADC distribution for APV channel 048, time bin 02, and APV 02, and the peak shown is the pedestal. Ideally, this data would be a very narrow Gaussian peak, meaning that the electronics read a single value or ADC channel, but instead the pedestal is observed as a wider-width Gaussian distribution. 16

24 This distribution is the result of noise from the electronics and small shifts in the pedestal location in ADC channel. In Fig. 11, a Gaussian fit was made of the peak in the distribution and the mean and width were calculated. Understanding the pedestals is a fundamental step in testing a detector. If the pedestal is not understood, then problems interpreting a signal in the detector could arise. These studies were very helpful to the collaborators building and implementing the FGT. This analysis gave an initial look at the output of the electronics to see if behavior is as expected or if any adjustments or replacements needed to be made before the FGT detector was installed into the larger STAR detector. The data used for this analysis includes 30 APVs or three quadrants that were placed in a configuration between the scintillator detectors (see Fig. 10). The analysis involved looking at the ADC distributions for each channel (128 total for each APV), time bin (7 total for each channel), and APV (30 total). The ADC spectra, which consist of Fig. 11. Pedestal histogram of channel number 048, time bin 02, and APV 02 with Gaussian fit. The fit-mean is at ADC channel number 745 and the fit-sigma (width) is ADC channels. 17

25 pedestal peaks only with high voltage off, are fit with a Gaussian curve with a fit range set to 400 ADC channels from the mean of the histogram. From this fit, three fit parameters were found: the peak value, the mean of the fit, and the width of the fit. The fit-mean and fit-sigma values for each channel, time bin, and APV, were used for the subsequent analysis. Figure 12 shows the fit-mean distribution that includes the fit-means from every channel, time bin, and APV of three quadrants, totaling 26,880 fits. Figure 12 illustrates the wide range of pedestal mean ADC channel values. The total distribution of the fitmeans can be separated into the seven time bins (numbered 0-6) to look for any patterns or correlations. The fit-mean distribution per time bin is shown in Fig. 13 for two time bins, 00 and 03, for all channels and APVs. Only two of the seven time bins are shown in Fig. 13, but all seven were similar in shape, meaning the fit-mean distributions did not change per time bin and all have a Fig. 12. Total ADC distribution of the Gaussian fit-means for all combined channels, time bins, and APVs from three quadrants. 18

26 Mean (ADC channel) (a) (b) Fig. 13. Distribution of Gaussian fit-mean per time bin. ADC distributions are shown for (a) time bin 00 and (b) time bin 03. similar shape to the total fit-mean distribution in Fig. 12. Figure 13 includes data from all channels (128 for each APV) and all APVs (10 for each quadrant) from three quadrants, which results in a large number of values. Any possible trends across the time bins may be hidden or diluted within this data set. Figure 14 shows the mean of the ADC spectra for each time bin, including the two shown in Fig. 13. Figure 14 indicates that there is no correlation between the pedestal fit-means and the time bins. As previously mentioned, Fig. 13 and Fig. 14 contain all channels and APVs, so the time bins needed to be investigated individually by channel and APV Time Bin Fig. 14. Average of Gaussian fit-mean distribution per time bin. 19

27 Normailized fit-mean (ADC channel) The next analysis focused on the ADC spectra from one APV and examined each time bin of every channel. The analysis involved the pedestal fit-means from selected channels in APV02, and the fit-means for each time bin were compared to time bin 04. APV02 appears to be a normal APV, based on other analysis. Time bin 04 was selected because it is expected to be the middle of the seven time bins in the APV channel. Figure 15 shows the pedestal fit-means of each time bin (subtracted from time bin 04) for five channels of APV02 and illustrates an oscillatory behavior between even- and odd-numbered time bins for certain channels. Even though this pattern exists, the difference in pedestal fit-means between the time bins for APV02 is roughly 30 ADC channels maximum. Figure 15 only shows five channels of one APV, but it includes the different levels of variation in fit-mean values across the time bins seen across all the channels. Figure 15 indicates that time bin 04 and time bin 00 have similar ADC values across the channels. This pattern is also seen in Fig. 16, which shows the fit-means of Time Bin Channel 17 Channel 34 Channel 62 Channel 95 Fig. 15. Difference in average pedestal fit-mean values across the time bins for different channels in APV02. The ADC values are compared to time bin 04. Lines are included to guide the eye. 20

28 Normalized fit-mean (ADC channel) time bin 00 subtracted from time bin 04 for all channels. In Fig. 15 and Fig. 16, the error bars are smaller than the size of the points. This flat line seen in Fig. 16 is the same in all even numbered time bins in comparison to time bin 04, meaning that time bins 00, 02, 04, and 06 all have similar fit-mean values across the channels for APV 02. Odd-numbered time bins subtracted from time bin 04 show a correlation in Fig. 17. Figure 17 displays a downward-sloping trend in the normalized fit-mean values across the channel numbers for time bin 05. At channel numbers greater than about 70 in Fig. 17, the ADC values of time bin 05 are most likely greater than the ADC values in time bin 04. This same trend is seen in all the odd-numbered time bins, but the crossover point is not at the same channel number. In Fig. 17, the error bars are smaller than the size of the points. Table I lists the average and standard deviation values of the difference in the pedestal fit-mean values subtracted from the pedestal fit-mean values of time bin 04 for all the channels of APV 02 for all time bins APV Channel Number Mean = St. Dev. = 1.47 Fig. 16. Fit-means from time bin 00 subtracted from time bin 04 as a function of channel number for APV

29 Normalized fit-mean (ADC channel) APV Channel Number Fig. 17. Pedestal fit-mean values for time bin 05 subtracted from time bin 04 fitmeans as a function of channel number for APV 02. In the previous analysis presented above, the total Gaussian fit-mean distribution in Fig. 12 was separated by time bins. The next analysis separated the same distribution (Fig. 12) by APVs. The data have a total of 30 APVs (numbered 0-29) and each APV has 128 channels. Figure 18 shows two examples of the fit-mean distributions that include all channels (128 per APV) and all time bins (7 per channel). APV11 and APV28 were selected to show that the APVs individually cover a range of about 200 ADC channels, but this range varies in ADC channel location. For example, APV11 has an ADC range between about ADC channels, and the ADC range of APV28 is between about ADC channels. Table I. Average and standard deviation for the pedestal fit-mean values of the time bins subtracted from pedestal fit-means of time bin 04 for all channels of APV02. Time Bin Subtracted from Time Bin Average (ADC channel) Standard Deviation (ADC channel)

30 (a) (b) Fig. 18. Pedestal fit-mean distribution of (a) APV11 and (b) APV28. These distributions include all 128 channels and seven time bins. Taking the mean values from the 30 APV fit-mean histograms, such as those shown in Fig. 18, and plotting them as a function of APV number will illustrate any variation in the means across the APVs. Figure 19 displays the mean value of the ADC distribution for an APV as a function of APV number for three quadrants used in the cosmic ray tests. The data appear similar in groups of five APVs. This pattern correlates to the experimental setup of the APVs; each quadrant has ten APVs, in two groups of five, located at the quadrant boundaries (see Fig. 8). The first five APVs are connected as stated in Chapter II, and the next five are connected using a crossover cable that connects the two sides of the FEE assemblies to the readout. This pattern then repeats itself in all three quadrants. The error bars are on the order of a few ADC channels in Fig. 19, so the variation is statistically significant. It is not expected for the APVs to have the same mean (pedestal) values because each APV uses different capacitors, but the pattern repeats based on the similar physical connections in hardware. 23

31 Mean Values (ADC channel) APV Fig. 19. The mean of APV distribution as a function of APV. The ADC spectra of the pedestals were fit with a Gaussian curve and, besides the mean of the distribution, the widths (sigma) of the distribution were found within the fit range. Plotting all the pedestal fit-sigma values per channel, time bin, and APV gave the total fit-sigma distribution, shown in Fig. 20. Most of the pedestal widths have a typical fit-sigma value in the ADC channel range, but there are some with much larger values. A large fit-sigma demonstrates a wide pedestal or a poor Gaussian fit to the pedestal. A wide pedestal could be the result of large amounts of electronic noise or just a shift in the pedestal mean value. Separating the ADC pedestal width values in Fig. 20 into time bins or into APVs, just as was done with the pedestal fit-means, provides additional information about the performance of the FGT electronics. Figure 21 gives two examples of the distributions of the pedestal fit-sigma values separated by time bin. The same issue of the dilution of any possible effect holds for the pedestal fit-sigma distribution per time bin, as with 24

32 Fig. 20. Total pedestal fit-sigma distribution. The distribution in ADC channels contains data from each channel, time bin, and APV. the fit-means per time bin. Figure 22 displays the mean of the histogram from the pedestal fit-sigma distribution histograms per time bin, including those in Fig. 21, as a function of time bin. Figure 22 shows little variance in the average pedestal fit-sigma across the seven time bins. (a) (b) Fig. 21. Fit-sigma distribution per time bin. Distributions are shown for (a) time bin 04 and (b) time bin 05 for all channels and APVs. 25

33 Mean of the pedestal widths (ADC channel) The total pedestal fit-sigma values in the distribution shown in Fig. 20 can also be separated by APV, and these results are displayed in Fig. 23. Here it is clear that APV 28 produces pedestal fit-sigma values over a wide range, where APV 10 gives more expected sigma values. Both Fig. 23 (a) and (b) do however have multiple peaks. This implies that each APV has channels or time bins that produce pedestals at a range of different widths. A narrow pedestal width could mean that a channel is not working properly, and a wide pedestal could be the result of large amounts of noise from the electronics or a shift from instability. Figure 24 shows the distribution of the mean of the pedestal fit-sigma distribution histograms per APV, including those shown in Fig. 23. In Fig. 24, it is again possible to see the groupings of ten APV chips by quadrant for the three quadrants studied. It also shows that the APV28 mean value does produce an abnormal distribution of pedestal widths, as shown in Fig. 23 (b), in comparison to the other APV chips Time Bin Fig. 22. Mean of pedestal fit-sigma distribution for each time bin. Each point includes data from 30 APVs and 128 channels per APV. 26

34 Mean from Sigma Distribution (ADC channel) (a) (b) Fig. 23. Fit-sigma distribution as a function of APV. Distributions are shown for (a) APV 10 and (b) APV 28. Each distribution includes all channels (128) and time bins (7). Some fluctuations in the data, such as a shift in the pedestal location, could result in poor fits to the pedestals. Some abnormal ADC distributions result in the pedestal being double-peaked or asymmetric, as displayed in Fig. 25. These shapes could imply a possible shift in the pedestal during the run APV number Fig. 24. Means of the fit-sigma distributions from each APV as a function of APV number. 27

35 (a) (b) Fig. 25. Examples of poor Gaussian fits to the pedestal peak shown (a) for channel number 14, time bin 06, APV 28 and (b) for channel number 3, time bin 05, APV 02. Another study involved looking into common mode noise. In the FGT, the operating conditions could change and cause fluctuations event by event, creating a type of shift in the pedestal data common to the whole detector. To investigate this behavior, an average ADC value from all the channels of a single APV is calculated and that average value is then subtracted from a given channel s raw ADC value for each event and time bin. This subtraction should result in a reduced width in the pedestals, i.e. a smaller sigma value, because it would remove any systematic variations common to all the channels. The ADC distribution histograms were re-created after this subtraction for each channel, time bin, and APV. The Gaussian fit is applied over the range of 400 ADC channels from the mean of each pedestal histogram. Figure 26 shows the new total pedestal fit-sigma distribution, after the subtraction for common mode noise was made. A comparison of the distribution from Fig. 20, which shows the total raw fit-sigma distribution, to Fig. 26, which shows the fit-sigma distribution after 28

36 Fig. 26. Pedestal width ADC distribution after common mode noise subtraction. The distribution shows the pedestal fit-sigma values of all channels, time bins, and APVs. the common mode noise subtraction, gives evidence of common mode noise in the FGT. The pedestal values decreased in width and the main peak is taller after the subtraction. This change means that more pedestals are now in the width range of ADC channels, where before, as in Fig. 20, there was a bump around a width of 60 ADC channels. III.B Cosmic Ray Data. High Voltage On By supplying high voltage to the GEM foils, ADC values larger than the pedestal can be observed in the two-dimensional histograms of APV channel as a function of ADC value. These larger ADC values are due to the multiplication of electrons stemming from cosmic rays that pass through the quadrants. The same 30 APVs from the three 29

37 quadrants used in the cosmic ray test apparatus and from the pedestal analysis were used for these tests. Figure 27 shows the raw ADC distribution of all the channels in APV01 and time bin 04. The dense region around ADC channel number 700 in Fig. 27 is produced by the pedestals and all the ADC values larger than that band represent hits from cosmic rays. One feature of Fig. 27 is the slant or systematic shift in the pedestal location across the channels. This characteristic presents a problem when trying to separate the pedestal from signal. Normalizing the pedestal means to a certain value from a single channel will straighten this distribution and make it easier to separate the pedestal from signal. The lowest pedestal fit-mean value of each channel in a given APV was used as the normalization factor. Figure 28 represents the same data as shown in Fig. 27, only now the data are normalized to the lowest pedestal fit-mean value of a channel for each APV. Fig. 27. Raw ADC spectrum of 128 channels in APV01 and time bin

38 (a) (b) Fig. 28. APV channel number as a function of ADC channel for time bin 04 and APV 01. The values are normalized to the lowest fit-mean of the channels with (a) high voltage on and (b) high voltage off. The separation of the signal from the pedestal ADC values can be investigated by analyzing the projection of Fig. 28 (a), which is given in Fig. 29 (a). Since the pedestal dominates this histogram, Fig. 29 (b) shows the same projection with an enlarged scale so that the signals above the large pedestal peak can be observed. An estimate was made on a pedestal cut at ADC channel number 800 for time bin 04 and APV01. This cut limit (location in ADC channels to remove the pedestal) needs to be several sigma away from the pedestal mean to avoid any variation that is still a part of the pedestal. A cut on a constant value of the ADC across all the time bins or APVs cannot be applied due to variation in the pedestal means (see Fig. 27). The previous study has shown that the pedestals are not in the same location for each time bin or APV, so the cut may be made at a few sigma away from the pedestal mean. A data cut on ADC channel number 800, for the time bin and APV shown in Fig. 29, is 31

39 (a) (b) Fig. 29. Histogram of ADC values for time bin 04 and APV 01. (a) This is the projection of ADC channels of Fig. 28. (b) The same projection is shown with an enlarged scale. approximately six-sigma away from the mean of the histogram. This six-sigma cut may then be applied across all APVs and time bins. Figure 30 displays the same data as Fig. 28, but now the six-sigma cut has been applied to the data, which removes the pedestal from the histogram. The projection of Fig. 30 on the x axis is shown in Fig. 31, in combination with all other time bins of one APV. Fig. 30. APV channel number as a function of ADC channel, for time bin 04 and APV 01. Only ADC values larger than six-sigma are shown. 32

40 Fig. 31. Histogram of ADC data with the six-sigma cut applied. The distribution includes all time bins of APV 01. The change in the number of data points representing the signal can be studied by varying the sigma cut from the pedestal location. Figure 32 displays the ratio of signal values to the entire ADC data sample as a function of the ADC pedestal cut. The conclusion of this study is that as the sigma cut is changed, the number of signal values does not change by more than 0.1%. The six-sigma cut was also applied to the data with the high voltage off, to see how many of the entries above six-sigma still remain due to the pedestal. For example, the APV shown in Fig. 27 has zero entries six-sigma above the mean with the high voltage off, while other APVs, such as APV15, have entries greater than six-sigma still coming from the pedestal with high voltage off. In addition, the location of the pedestals did not shift by more than a few ADC channels between high voltage on and off for the cosmic ray test data. 33

41 Ratio of signal to total entries (%) Sigma Cut 0 APV 00 1 APV 01 2 APV 02 3 APV 03 4 APV 04 Fig. 32. Ratio of the signal to total entries as a function of the location on the ADC pedestal cut. The data points are summed over all time bins and are shown for the first five APVs (00-04). III.C STAR Data. Once the FGT was installed into the STAR detector, some portions of the analysis from cosmic ray data were repeated to see if the detector behaved in the same manner. The new analysis of the data from the FGT located within the STAR detector is referred to as STAR data. Pedestal data taken with the high voltage off for the cosmic ray test setup and for the STAR data were compared, and the results were examined to see if any shift or change occurred. It is important to note that in this comparison analysis between cosmic ray data and STAR data, it is actually the electronic ID numbers that remain the same, not necessarily the same quadrants or hardware. Because of this, more changes could have been introduced than just the location of the FGT, such as 34

42 Mean from APV fit-mean distribution (ADC channels) changes in the electronics or APV chips. Therefore, the following analysis can only give a rough comparison between cosmic ray test data and STAR data. Starting with the ADC distributions for both cosmic ray test data and STAR data with the high voltage off, Gaussian fits were applied to the pedestal peaks of each channel, time bin, and APV. These data sets included 30 APVs, each with 128 channels and each channel with seven time bins. Taking each pedestal fit-mean from the ADC distributions of each channel, time bin, and APV, the total fit-mean distribution of the pedestals can be plotted. The total fit-mean distribution separated by APV will show if the distribution has changed. Figure 33 shows the means from the APV pedestal fitmean distributions as a function of APV. The cosmic ray data in the figure are the same as Fig. 19. The pedestal means are now smaller and more tightly-bunched with the STAR data compared to the cosmic ray test data, and the groupings of five APVs are no longer prominent. This difference in the data could be the result of adjustments or replacements in the APV chips. The error bars are on the order of a few ADC channels APV STAR COSMIC Fig. 33. Mean from pedestal APV fit-mean distribution. Each point includes all channels (128) and time bins (7). The cosmic ray data are the same as that in Fig

43 Figure 34 shows the fit-mean distributions from STAR data and cosmic ray data for time bin 04. These histograms include every channel (128 per APV) and APV (30 total) for one time bin. The pedestal means now fall within a smaller range of ADC channels for the time bin shown. This difference could be because of better timing of the pulse in the time bins or a different APV chip. The widths of the pedestal ADC distributions indicate that the fit-sigma values remain about the same for the time bins. Figure 35 shows the fit-sigma distributions for time bin 04 for STAR data and cosmic ray test data, and indicates that the pedestals did not change width between the two data sets. Figure 15 previously revealed an oscillatory pattern between the even- and odd-numbered time bins in the cosmic ray data. Now using STAR data, Fig. 36 shows the same channels, and the fit-means were normalized to time bin 04, just as in Fig. 15. The patterns across the time bins in Fig. 36 are quite different from those in Fig. 15, and (a) (b) Fig. 34. Fit-mean distribution for time bin 04. (a) The fit-mean distribution for STAR data and (b) the same distribution for cosmic ray data are shown. 36

44 Difference in fit-means (ADC channel) (a) (b) Fig. 35. Fit-sigma distribution per time bin (04 shown). (a) The fit-sigma distribution for STAR data and (b) the distribution for cosmic ray data are shown. do not present a clear pattern as they did previously. This difference again could be the result of different APV chips used with the STAR data than those from cosmic ray test data. The error bars are smaller than the size of the points in Fig. 36. The pedestal means were also investigated as a function of channel. The pedestal means in time bin 00 were subtracted from pedestal means in time bin 04 (as Time Bin 17 Channel Channel Channel Channel Channel 121 Fig. 36. ADC channel as a function of time bin. The fit-means were normalized to time bin 04 and the ADC values are shown for the same five channels of APV 02 as in Fig

45 Difference in pedestal fitmean (ADC channel) Difference in pedestal fit-mean (ADC channel) shown in Fig. 16 for cosmic ray data). In Fig. 37, the shape of the curve is still similar for STAR data. Looking at the pedestal means in time bin 02 subtracted from the pedestal means in time bin 04 in Fig. 37(a), a different shape is observed at low channel numbers. The error bars are smaller than the size of the points in both plots of Fig. 37. Comparing the correlation between the difference in pedestal means between time bin 04 and 05, a positive slope for the STAR data and a negative slope for cosmic ray data can be seen in Fig. 38 and Fig. 17. This slant in the data was present in both cosmic ray test data and STAR data and requires more investigation. The sloping trend in the channels may be a result of the electronic ID number from the data acquisition. The electronic ID refers to the order in which the APV channels are read by the data acquisition electronics. Using the ADC fit-means from the pedestals and plotting them as a function of electronic ID, a similar sloping trend is indicated, as seen in Fig. 39. Figure 39 contains data from time bin 04 of APV00 with (a) APV Channel Number (b) APV Channel Number Fig. 37. Difference in pedestal means as a function of channel number in APV 02. (a) Pedestal means in time bin 02 subtracted from pedestal means in time bin 04 for STAR data and (b) the same formula using cosmic ray data are shown. 38

46 Fit-mean (ADC channels) Difference in fit-mean (ADC channel) APV Channel Number Fig. 38. Difference in pedestal means as a function of APV channel number for time bin 05 subtracted from time bin 04 for APV 02. the high voltage off for the FGT in the STAR detector. The pedestal fit-sigma values for time bin 04 and APV00 do not show a strong correlation with electronic ID, as seen in Fig. 40. Fig. 39 and Fig. 40 both have error bars smaller that the size of the points. Even though data from only APV00 are shown here, further analysis indicated that several other APVs had similar plots to those shown above Electronic ID Number Fig. 39. Pedestal fit-mean as a function of electronic ID for time bin 04 of APV00. 39