Internship report: Characterization of GridPix detectors ...€¦ · This report is written to conclude the work placement internship, part of the program for receiving the degree

Internship report:

Characterization of GridPix detectors from test beam

measurements.

M.M. van LeestWork placement, M.Sc. Applied Physics, TU DelftNikhef Institute for subatomic physics, Amsterdam

Reactor Institute Delft, TU [email protected]

August 10, 2009

1

Abstract

In this project the resolution and efficiency of a new type of electronic Time ProjectionChamber (TPC) called GridPix is tested, involving a TimePix pixel chip in a gaseousdrift chamber. Individual electrons created in the gas by ionization due to radiation canbe detected. 3D Recording of the spatial coordinates of these electrons with the TimePixdetector allows the reconstruction of ionization tracks.

Four GridPix detectors designed for resolution and efficiency experiments (GOSSIP)were prepared and found to be working properly. For further analysis, tests were per-formed at a test beam facility at CERN using a GridPix detector with a 2 cm drift gap(DICE). Data has successfully been taken, varying several parameters. Various analyseshave been performed on the data taken with this detector. The simultaneous readout of4 detectors using a four-fold cable has proven to work.

This report is written to conclude the work placement internship, part of the programfor receiving the degree of Master of Science at the Department of Applied Physics ofDelft University of Technology. The work was carried out in the section of Detector R&Dof the Nikhef institute for subatomic physics, under the supervision of Dr.ir. Harry v. d.Graaf (Nikhef) and Prof. Dr. Pieter Dorenbos (TU Delft).

2

Contents

1 Introduction 41.1 The GridPix detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.2 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.3 Structure of report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Theory 72.1 Ionization by charged particles . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Charge transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.3 Electron multiplication and efficiency . . . . . . . . . . . . . . . . . . . . 92.4 Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.5 Choice of gas medium . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3 Experimental Set-Up 123.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.1.1 TimePix chip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.1.2 Muros2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143.1.3 Pixelman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.2 Functioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153.3 Test beam CERN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4 Test beam results 174.1 Tracking and visualization . . . . . . . . . . . . . . . . . . . . . . . . . . 184.2 Drift velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204.3 electron efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244.4 DICE in magnetic field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

5 Fourfold simultaneous readout 275.1 Single pixel chip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275.2 Quad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.3 Fourfold Cable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.4 Test beam results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

6 Conclusion 33

A MATLAB program for track visualization 36

B MATLAB program for track statistics 38

3

Chapter 1

Introduction

1.1 The GridPix detector

In subatomic physics, collisions and decay processes allow the study of elementary parti-cles. The physical processes taking place can be revealed by observing the trajectories ofthe highly energized particles involved. Energy losses per unit length and magnetic fieldsallow the identification of particles and their momentum. By identifying the products ina decay or collision process, theoretical predictions on the nature of elementary particlescan experimentally be studied. The proper observation of trajectories is crucial in theseexperiments. In this internship the functionality of a new type of digital Time ProjectionChamber is tested.

The detectors consist of a gaseous chamber, with an electron sensitive CMOS pixelchip called TimePix [1]. This detector, developed in a CERN collaboration, has 256 x256 pixels and is a modified chip based on the MediPix chip. The MediPix chips weremainly build for small scale X-ray imaging. The counter present in each pixel is used todetermine the X-ray quanta. At Nikhef, Jan Visschers suggested to use the counter toregister time steps, using an external clock source. In this TimePix chip, the counter isof importance because of the following. After ionization in the gas medium, the electronswill drift towards the detector due to an applied electric field. By using a conventionalpixel chip, only the projection on the chip will be determined and information on height islost. The counter in the TimePix detector allows the registration of this third dimensionof an event, based on the drift time of the electrons and the known drift velocity.

The MediPix chips were mainly built to detect photons for X-ray imaging. The chipscan be modified to be able to detect electrons. This is done by replacing the top photonsensitive layer on the chip by a gas medium. Since single electrons cannot be detecteddirectly, an extra high voltage grid (MicroMEGAS) has to be included to create a highelectric field for electron multiplication right above the pixels. In this way single electronscan be detected [2]. In the GridPix detectors developed at Nikhef, these grids are appliedby post-processing the bare pixel chips (InGrid [3]). During post-processing a protectiveSi3N4 layer is added as well on the chip (SiProt) to prevent discharges from damagingthe chip. A schematic representation of the working principle of the GridPix detector isshown in Figure 1.1.

The work during this internship, was to participate in the assembly and functionality

4

Figure 1.1: Scheme of the GridPix detector, showing the path of electrons towards the pixel chip, afterionization in the gas medium caused by a highly energized particle. Between the grid and the pixels, anavalanche is created.

testing of new GridPix detectors. These detectors were designed with the purpose of us-ing GridPix detectors in SLHC (and ILC) projects. For such large scale applications, itis important to keep the weight and power consumption of the detectors small, to reducecosts. The GOSSIP (Gas On Slimmed SIlicon Pixels) detectors [4] were created in orderto test if using a small drift gap (1 mm), would suffice for obtaining proper position res-olution and efficiency. Simulations revealed that this could be the case, yet experimentalproof is required as well. If experiments and simulations agree, the detectors can outper-form the conventional silicon trackers presently used in the large collision projects. Thegaseous detectors have the advantage that no radiation damage occurs in the medium,since the gas is flushed continuously. Also the weight and power consumption are muchlower compared to the silicon trackers.

Testing the GridPix detectors for resolution and efficiency, required large quantitiesof high energy charged particles with controlled features (energy, angle of incidence).Therefore tests were performed at a test beam facility at CERN. Furthermore, the ideaof reading out multiple detectors at once has been studied and a setup has been realized.

5

1.2 Organization

The work in this internship is carried out at the Nikhef institute, located in Amsterdam.Nikhef is the Dutch National institute for subatomic physics, founded on the partnershipbetween the Foundation for Fundamental Research (FOM), and four Dutch universities(UvA, VU, UU, RU). The Nikhef institute is involved in large projects in accelerator-based particle physics and astroparticle physics. This internship was carried out at thedepartment of Detector R&D, which focuses on the development and improvement ofgaseous detectors.

Experiments in high energy physics require large and expensive set-ups, which forcesinternational collaborations to be found. The main scientific research center for thesecollaborations is the European Organization for Nuclear Research (CERN), located nearGeneva. Part of this internship was to test the newly developed detectors in a test beamfacility at CERN. The MediPix collaboration is also based at CERN. The software neededfor readout of the pixel chips, was developed at Czech Technical University in Prague.The post-processing of InGrids currently takes place at the MESA+ Institute for Nan-otechnology at the University of Twente. The electronic readout device was created in acollaboration with the company PANalytical.

1.3 Structure of report

The structure of the report is as followed. First the basic physics relevant for these de-tectors is presented. Then the setup with all its components is discussed. Following theresults from the DICE detector in a test beam and the analysis performed on the dataacquired. Finally the investigations regarding the simultaneous read-out of four GridPixdetectors is presented followed by some concluding remarks.

6

Chapter 2

Theory

First the theory related to the actual ionization is discussed. After that, information onthe transport of charged particle in a gas will be given. The theory presented here israther brief, presenting only the relevant information. For more in depth information onparticle detection in drift chambers, see [5].

2.1 Ionization by charged particles

As was explained in the introduction, ionization due to high energy particles will generatethe electrons to be read out in the detector. The amount of ionizations per unit lengthis determined by the energy loss of the high energy charged particle causing the track.This energy loss is described by the Bethe and Bloch formula.

−dEdx

= −4πNAρZ

A

e4

mec2z2 1

β2

[ln

2mec2

Iβ2γ2 − 2β2 − δ (β)

2

](2.1)

Here NA is Avogadro’s number, ρ is the gas density, Z/A the effective atomic number ofthe gas, mec

2 the electron mass in rest, z the charge of the high energy particle, β = v/cits velocity, I the gas excitation energy and γ the relativistic factor. In Figure 2.1 thedependence of the energy loss per unit length on the charged particle momentum is givenfor several types of particles.

For particles with small momenta, the energy loss decreases with momentum fast.If the energy loss for a particle with limited momentum is significant, the particle willslow down and deposit even more energy per unit length. Eventually this can lead to anabrupt stop of the charged particle. Particles with high energy will not lose a considerableamount of energy in the gas and the energy deposit along the track is the same on average.There is, however, a variation of the energy loss. This variation is given by the Landaudistribution. The energy loss in Figure 2.1 only refers to an average. Since the GridPixdetectors have a small volume, these fluctuations have an influence on the variation oftotal energy deposit in the detector for similar events. So even for tracks of identical highenergy particles, the total amount of ionizations may vary.

7

Figure 2.1: Graph showing the energy loss per unit length in various materials. The correspondingenergy scales are given for different types of particles.

2.2 Charge transport

After ionization, the electrons drift towards the pixel chip for detection due to an appliedelectric field. In order to understand the working principles of a GridPix detector, thetheory behind the drift of these electrons must be applied. The electric and magneticfield applied are important parameters for the charged particle causing the track, as wellas for the drifting electrons created after ionization. The equation of motion for a particlewith mass m and charge e under the influence of an electric (E) and magnetic (B) fieldis as follows.

mdv

dt= e (E + v×B) (2.2)

An applied electric field makes the electrons accelerate. Collisions with gas moleculescause the electrons to slow down again. The average distance between collisions is given

8

by the mean free path λ or the average time by τ . Under constant conditions the electronswill have an average velocity, which is called the drift velocity.

v =e

mτE (2.3)

The mean free path/time is again a function of the type of gas (mixture), tempera-ture, pressure, the energy of the electrons and also on the applied field. Because of thiscomplexity, certainly in the case of gas mixtures, we will have to depend on documentedresults from simulations and measurements. The dependency of the drift velocity on theelectric field for the drift volume of our interest is given in Figure 2.2.

Figure 2.2: Graph showing the drift velocity as a function of the electric field. Calculated for several80:20 He:Quencher mixtures. Source: [6]

2.3 Electron multiplication and efficiency

In the GridPix detector, electric fields are applied so that the electrons generated in theionization will drift towards the pixels directly below, and the ions towards the cathode.The electrons are of interest to us and will be detected at the pixels of the chip detector,functioning as an anode plane. Between the grid and the pixel chip, the field strengthis much higher then between the grid and the cathode. The field is of such strength,that electron multiplication occurs. This simply means that the drifting electrons willbe accelerated thus, that they start secondary ionizations, thereby creating an avalancheof electrons. This allows a single electron passing through the grid to be detected on asingle pixel. The electron detection efficiency is dependent on the ratio of field strengths

9

above and beneath the grid. If the drifting electrons will go onto the grid instead of goingthrough, they will not be detected. In the situation sketched in Figure 2.3, none of thefield lines from above end at the grid, indicating that in this situation all the electronsfrom the drift volume will pass through the grid. The avalanche of electrons has to createenough electrons to exceed the threshold of the pixel, which is in the order 300 electrons.The grid potential therefore controls the electron efficiency.

Figure 2.3: A: Schematic cross-section of a GridPix detector showing the working principle and relevantparameters. B Close-up scheme of the InGrid on a chip, showing the electric field lines under normalworking conditions. Based on Figure 4.8 of [7]

The electric field between the grid and the chip must not be set too high. For gainsover several hundred thousand, the electrons in the avalanche will create an even higherfield at the front of the avalanche, boosting the multiplication even more. At a certainpoint the quencher gas can no longer absorb the photons created along the multiplicationand a plasma is formed. The plasma can cause a conductance path and thus a dischargewhen it connects the grid and the chip. This could seriously damage the chip. Nowadays,the protective SiProt layer gives protection, yet discharges remain unwanted.

2.4 Diffusion

While drifting towards the grid above the chip, the electrons will diffuse. In terms ofresolution this is an important feature and it limits the maximum height we can take forthe gas volume. In practice, the diffusion increases with the square root of the time it istraveling. Considering a constant drift time, this means that the diffusion also increaseswith the square root of the height, indicated by the mean square deviation σ. Thediffusion constant in the direction of the drift velocity (longitudinal) is slightly differentfrom the transverse direction. So the mean square deviations are

σt = Dt

√z and σl = Dl

√z (2.4)

Here D is the diffusion constant and z the average drift distance traveled by the electron(=vdriftt). For a more accurate, microscopic description of the diffusion and the effects ofdifferent quantities on the accuracy, one has to rely on simulations. In the Detector R&Dgroup such simulations were performed with the GARFIELD program, created at CERN.

10

In general one can say that the diffusion increases with the square root of the travelingdistance of the electrons. In case of the gaseous detector this implies that the accuracyis a function of the height above the chip. Ultimately this means that the height of thegaseous chamber is limited. Diffusion however, can be reduced when a magnetic field isapplied as well in the same direction as the electric field.

2.5 Choice of gas medium

Properties as drift velocity, diffusion, absorption are dependent on the type of gas thatis used in the detector. A lot of research has been done on the characteristics of gasesregarding these properties, including mixtures. For functionality it is important that thedrift velocity is constant and high and can be regulated with an applied electric field. Ingeneral a noble gas is used, to minimize losses due to absorption. Often a molecular gasis blended in, because of the following reason. In the process of ionization UV photonsare produced. Since the detector is sensitive to this radiation (sparks occur), a moleculargas is used as a quencher to absorb this UV light. Throughout this internship the com-bination of helium and isobutane is used. This mixture is known to work properly andis relatively save in operation.

11

Chapter 3

Experimental Set-Up

In this Chapter the set-up is described, in general terms and also in specific for the ex-perimental set-up in the test beam.

3.1 Overview

The general experimental set-up has the following components. A detector is connected toa Muros interface device. The Muros is connected to a PC via a built-in DAQ card. ThePC has to be equipped with, besides the DAQ card software, the readout software calledPixelman. The gas chamber(s) need to be continuously flushed with the gas mixtureof choice. The gas flow is controlled using two controllers. When the detectors areused for 3D tracking, an external trigger system has to be connected to the Muros.Two scintillators are aligned with the detector(s) which provide the trigger signal. Thescintillators are connected to a coincidence control unit which is again connected to theMuros2. The control unit sends a trigger signal only when both scintillators have a hit.After a trigger the readout takes some time, therefore a veto signal of 20 µs is created inthe control unit to prevent new trigger signals being send too early to the Muros2. Anoverview of the set-up, as it was at CERN is given in Figure 3.1. The following sectionwill go into detail on several components of the set-up.

3.1.1 TimePix chip

The TimePix chip consists of 256 x 256 square pixels with a pitch of 55 µm, thus totallymeasuring 14 x 14 mm. Each pixel element is connected to an amplifier and discrimi-nator, as well as a digital counter. The pixels can be used in three different operationmodes. In the MediPix mode, the counter is used to detect the total amount of detectedparticles during exposure. In the TimePix mode, the time is registered between shutterand detection of the first particle. In this mode the 3D constructions can be made in thecase of our gaseous TPCs. And finally there is the Time over Threshold (ToT) mode, inwhich the time that the pixel is above threshold is registered. This allows the measure-ment of the energy deposit on the pixel by the pulse height.

Each individual pixel of the TimePix device in TOT mode is connected to its ownanalog circuitry and AD converter. Thus the device contains 65 536 independent ADCs.All of them have to be individually calibrated. More details on the digital infrastructure

12

Figure 3.1: Schematic overview of the general set-up for (a) GridPix detector(s), including an externaltrigger system.

will be presented in Chapter 5.

The pixel chips are mounted and wire bonded on to a chipboard. The housing for theGOSSIP and DICE detectors can be seen in Figure 3.2. The boards allow easy mountingof the cables for readout and high voltage supply. The GOSSIP chamber still neededassembly. After the bonding of the chips, a cathode foil was placed, distanced 1 mmabove the chip, with a connection to the HV cable connector. A Kapton foil was placedover the aluminium frame shown at the top in grey of Figure 3.2A to seal of the chamber.The gas connections were made in the frame as well.

Figure 3.2: Housing for GridPix detectors. A: GOSSIP detector. B: DICE detector. Drawn by HansBand, Nikhef.

13

3.1.2 Muros2

The Muros2 board is an interface between a Medipix2 or Timepix chip and a PC. Thedevice is connected to a PC using a National Instruments SH-68-68 cable and DIO-653XDAQ board. The Muros2 can be connected to the chips, by using a VHDCI cable.The Muros2 has been developed at Nikhef as a successor to the Muros1 board, designedfor Medipix1. The Muros2 contains an adjustable clock which is used for the TimePixoperation. By checking the settings in the software, the clock frequency can be read outaccurately. The board also has a connection for an external trigger. More informationcan be found in the manual [8].

Figure 3.3: Left: Housing for Muros2, right: Muros2 board Source: [9]

3.1.3 Pixelman

Software developed at Czech Technical University of Prague, allows applying the settingsand controlling the readout of the chips, via the Muros. Special libraries were written forthe communication with Muros devices and plug-ins for special software tools. Duringstartup of the Pixelman software, it will automatically connect with the Muros and iden-tify the chip(s) connected to it via an ID number. If this procedure has succeeded, themain menu will appear (MediPix control UI), allowing the operation of the detector chip.If operation on the connected chip has occurred earlier with the same PC and program,the previous settings will be reloaded automatically. In this section, the main options inPixelman that are of importance are presented. A more extensive manual can be foundat [10].

When starting with a new chip, the following steps must be taken. In the Acquisitionmenu, settings, the polarity must be set to negative, since we are collecting electrons.Here one can also select whether an internal or external trigger is used. In the case ofusing TimePix, the external trigger is necessary, for ToT operation the internal trigger(PC controlled) will do as well.

As was told, all pixels in the chip have their own specific DA-Converter settings(DACs). Due to post-processing or problems with bonding, the pixels could be malfunc-tioning. A quick way to test the performance of the chip is to do a DACs dependencyscan. This Pixelman plug-in shows the dependency of sensed values on the selected digitalcodes of MediPix DACs. One of the most important DACs is the threshold limit (THL).

14

The threshold equalization plug-in in Pixelman, automatically matches the thresholdlimit for all pixels, so that during operation the overall threshold can be set just slightlyabove the threshold limit. Doing this, prevents information from weak signals to be lost.If pixels are not corresponding correctly, they will be masked.

For the readout, several settings are important. The acq. type can be set to ’frames’to give every acquisition separate or ’integral’ for the total sum of acquisitions. The acq.time sets the shutter time for the pixels. When the repeat option is used, the wholeprocedure readout will be repeated as long as desired. The preview option, allows thedirect observation of the pixel readout.

The frames can directly be saved in a X,Y,’value’ ASCII file. When collecting data, itis preferred to make a pre-selection of the events to store as data. This can be done byusing a filter chain. It is a list of predefined filters, which can be edited. Mainly the filterfor masking specific pixels and a minimum amount of pixel hits for writing are used. Inthe chain one also defines the folder in which the data must be stored. For large datasets, it is recommended to set the number of frames high, instead of using the repeatfunction, since the program will create a new folder for each run.

3.2 Functioning

After the assembly of the GOSSIP chambers, these detectors, including the DICE detectorwhich was made earlier, were tested for functionality. After testing the chips for basicfunctioning (DACs dependency scan, threshold equalization), the gas chambers had to beflushed for several hours with a typical gas medium, a 80:20 He-Isobutane mixture. Thelong time is necessary to make sure no oxygen is present in the chamber, which can causeabsorption of the drifting electrons. Then, steadily the cathode potential was lowered to-1000 V, while monitoring the current. Meanwhile the grid potential could be set as well,again monitoring the current. If the current becomes to large ( 1 µA), it could indicatea leak contact between the grid and the chip. Once the grid potential reaches -300 orlower it should be possible to see the first signs of Fe-55 decay products. When a Fe-55source is held near the detector, X-ray photons of 5.9 or 6.5 KeV will go through thegas causing an ionization. The freed electron will mostly take over its energy, which isenough to cause many secondary ionizations (around 160 for Helium). A Sr-90 sourcewas used as well for the testing, which produces electrons in a Fermi energy spectrumwith energies up to 2.28 MeV. With a much higher energy on average then the radiationfrom Fe-55, the energy loss per unit length is much less for these electrons. Therefore thetracks visible from the detector are less dense and occur as straight lines.

3.3 Test beam CERN

In the large colliding experiments, particles with very high energies are created. Sincethe goal of GOSSIP is to verify if the GridPix detectors can be used for inner trackingpurposes, experiments must be performed using high energy particles as well. In theNetherlands, such facilities are not present. However, CERN offers test beam periods fordetector experiments, using accelerators that have been build several decades ago. The

15

entire complex of current accelerators is given in Figure 3.4.

Figure 3.4: Overview of all the accelerators presently at CERN. The North hall, where the experimentstook place, is located in the center of the LHC ring. The beam originates from the PS synchrotron, goesto the SPS synchrotron for an extra boost and is then send to the experimental hall.

In the North Area Hall, the Detector R&D group of Nikhef had a test period as aparasitic user in section H8, downstream from another experimental group. During thefirst week, a set-up was build, baring in mind future visits. The four GOSSIP detectorswere positioned on a optical rail using standard carriers. DICE was positioned behind,placed separately on a rotatable stage with grading. The external trigger system wasused, involving the control unit and two scintillators. The scintillators were covering thesame active area as the detectors, to make sure only proper tracks were recorded.

Figure 3.5: Overview of the test beam area in section H8, North Hall Experimental Area, CERN. A:Close-up of set-up. On the left the four GOSSIP detectors are shown, on the right the DICE detector. B:The entire set-up was installed on a remote controlled xy-stage, allowing exact positioning in the beamline after the large cryogenic magnet which was used by the main users.

16

Chapter 4

Test beam results

After the installation of the set-up, the gas chambers needed to be flushed with a properHe-Iso mixture. The gas flow was regulated with two separate controllers for each gastype. The flow through the controllers was set with a control panel, with a scale between0 - 100 %. The absolute flow had to be measured using a soap bubble gas flow meter.It was thought that a 80:20 ratio He-Iso mixture would be a safe choice for doing thefirst measurements. The standard method for calibration is to measure the gas flow fora range of settings on the control panel, separately for the two gases. From the obtainedgraph, the proper settings can then be chosen. However, time in the test beam area waslimited. Therefore the gas flow was set as follows. First the gas flow of the isobutanewas measured with a 100 % flow setting. It was found to be 0.5 cm3/s. Then the he-lium flow was set iteratively, varying the controller and finally measured to be 1.7 cm3/s.Combining the gases it thus gave a mixture of 78.3 ±3% helium and 21.7 ±3 % isobutane.

The beam contained bunches of 20,000 to 30,000 pions of 180 GeV. The bunches ar-rived approximately two times per minute with a duration of about 10 seconds. Thebeam could only be turned on once the test area was clear of people. During a consid-erable time in the test period, there was no beam available. With time being scarce, itwas decided to only measure with the DICE detector, and leave the GOSSIP detectorsfor future test beams. All results presented in this chapter are thus obtained with theDICE detector.

Alignment of the detector in the beam was realized with the remote control of the xytable, which could move the entire set-up in a plane perpendicular to the beam direction.With remote desktop control, direct observations could be made with the detector, to seeif the position was right. For this the integral option of Pixelman was used. An exampleof such a beam profile, after alignment can be seen in Figure 4.1.

The alignment had to be repeated after rotating the detector. This was sometimes alsonecessary when the large cryogenic magnet upstream was turned on, which redirected thebeam slightly. The magnet was used by the other experimental group during this testperiod in the same area. After the alignment the measurements could be taken. It turnedout that the settings given in Table 4.1 for Pixelman gave the biggest amount of usefulevents.

Measurements were performed under the conditions given in Table 4.2. The cathode

17

Figure 4.1: Screen shots of the beam profile obtained with the DICE detector placed perpendicular (0o

) to the beam. A: Detector not aligned. Obtained with a long acq. time to see the scattering products.The beam was located further from the bottom right side of the image. B: An integral image in Time

over Threshold mode was taken for several minutes with the detector aligned in the beam. A short acq.time was taken to limit the contribution of scattering events.

Table 4.1: Software settings for Pixelman, used for the measurements with the DICE detector in the testbeam.

Setting ValueOperation mode TimePixAcq. type framesAcq. count 10,000Acq. time 0.001 sFilter Chain Masking dead pixels, saving for more then 10 pixel hitsTHL DACs 365

potential was always kept 600 V lower then the grid potential, except for the drift velocityexperiment in which only the cathode potential was varied. It was aimed for to have atleast 500 events per setting. This was mostly the case, however, some inconsistencies arepresent due to the limited time and irregular area access. The DICE detector also neededa third potential, for the guard plate, which was used to make the field homogeneous.The guard ring is positioned slightly above the grid, at one twentieth of the distancebetween the chip and the cathode. For creating a homogeneous field, the absolute guardpotential is thus given as follows.

|Vguard| =1

20(|Vcathode| − |Vgrid|) + |Vgrid| (4.1)

4.1 Tracking and visualization

For the reconstruction of tracks a total least squares method was used, called the prin-cipal component analysis [11]. This analysis is used to find the directions in a set of

18

Table 4.2: Overview of the settings varied for the measurements with the DICE detector.Angle Vgrid (V) Magnet0 -360/-380/-400/-420/-440 -10 -400/-420/-440 -25 -400/-420 -30 -360/-380/-400/-420/-440 -45 -360/-380/-400/-420/-440 -270 -400 0.12 T

data in which the variance is the largest. To be more specific, it is based on finding theeigenvalues and eigenvectors of the covariance matrix of the dataset, after setting thecenter of the dataset at the origin. In case of the track reconstruction, the fit equals oneof the eigenvectors. The reason for using this method, is that is simple and fast (it isa standard function in MATLAB), but most of all because the dataset is multivariate(variables in three dimensions). It minimizes the errors orthogonal to the track fit in allthree dimensions at once, instead of in just one dimension like in normal least squaresmethods. This method is therefore more adequate for the 3D data. As was told in Chap-ter 2, the spreading in electrons, caused by diffusion, is dependent on the path lengththe electrons had to drift. In order to make the track fit even more accurate, this couldbe taken into account by performing a weighted principal component analysis. However,this option is not easily implemented in MATLAB, in which the fitting programs werewritten. The study of diffusion was thus found to be beyond the scope of this internship.

In the TimePix mode, the number of clock cycles between hit and shutter at the endis registered. For tracks that go through the top and the chip itself , the z-coordinatecan be found as follows.

z = (max(tclock)− tclock) · vdrift/fclock (4.2)

Here z is the set of reconstructed height coordinates, tclock the set of clock counts for allhits, vdrift the drift velocity and fclock the frequency of the external clock. By comparingdifferent events with tracks going through the chip, it was found that the maximum clockcount was consistently around 480. This indicated that the time delay between triggerand arrival of the first electrons was consistent, meaning that the trigger was workingcorrectly. For tracks that did not go through the chip itself, it meant that this numbercould be used as the standard maximum clock count in equation 4.2.

Not all the events were useful. Sometimes secondary effects occurred, which made ithard to fit a track. Considering the scope of this internship it was decided to only selectthe events in which only a single straight track was made. This was done by setting amaximum (70) and minimum (20) amount of pixel hits and requiring that the averagefault was not larger then 0.1 mm, which would also indicate there are electrons not beingpart of the track. These values were empirically obtained and although being a roughestimation, sufficed in filtering out all the useful single event. In principle, the eventscontaining multiple tracks could be examined for further analysis as well. However, thedifficulty often is how to interpret the third coordinate. The second particle, not causingthe trigger, could have arrived at a later time. In the readout this would result in a falseinterpretation of the height of the track, since the height is based on the time betweenhit and shutter time. A nice example of an event with secondary tracks is presented in

19

Figure 4.2. Here we see one large thick track and several thinner tracks. What probablyhappened here, was that a pion interacted with the cathode top, launching a heavy ion(thick track) and several electrons. The angle of incidence was 30o from the top left, as inFigure 4.3. For the reconstruction of most of the results presented here, a drift velocityof 1.35 cm/µs was used, a value from literature [6] for an electric field of 300 V/cm.

Figure 4.2: 3D Plots of an event with secondary products. A: Side view. B: Top view of the same event.The z=0 plane corresponds to the chip level. The colors also indicate height, with blue being close to thechip, and red near the cathode top. The results are presented in the proper scale, using a drift velocityof 1.35 cm/µs for obtaining the height. Obtained using a MATLAB program, see Appendix A.

Examples of useful event plots including the fitted track and residuals are given inFigure 4.3.

4.2 Drift velocity

In this section, the idea of determining the drift velocity from the angle of incidence ofthe tracks is examined. The actual angle of incidence is known from the settings. Byrescaling the height in the data, the reconstructed angle of incidence can be matched withthe actual angle of incidence. From the scaling factor the drift velocity can be found.This procedure was used to find the drift velocity for various cathode potential settings.The results for an angle of incidence of 45o can be found in Table 4.3. The measured driftvelocities seemed to be lower then the values from literature, more so for the larger anglesof incidence. However the general trend certainly agreed with literature. The mismatchcould be due to several causes. The gas mixture was not measured very accurately dueto lack of time in the test beam area. The biggest error is probably due to the fact thatthe angles of incidence had a systematic offset. A deviation in the angle of incidencehas a larger effect on the large angles of incidence. The offset could be corrected for,by performing the same measurements with other angles and the same set of drift fields.Then the actual angles can be corrected for an offset error. Although this method couldwork, it is a bit cumbersome.

20

Figure 4.3: 3D Plots of fitted tracks for events under various angles. The z=0 plane corresponds to thechip level. The colors also indicate height, with blue being close to the chip, and red near the cathodetop. The results are presented in the proper scale, using a drift velocity of 1.35 cm/µs for obtaining theheight. Obtained using a MATLAB program, see Appendix A.

A more common way of determining the drift velocity, is by using the maximum andminimum clock count in events. This can only be done with tracks going through boththe top of the detector and the chip at the bottom. By matching this time difference withthe known drift height, the drift velocity can be found. Unfortunately, the measurementsvarying the cathode potential were only performed under an angle of 45 degrees. Theseevents did not go through both top and bottom of the detector. However for the caseof having a field of -1000 V, there are measurements with an angle of incidence of 0, 10and 30 degrees. The mean clock count difference between first and last track hits wasfound to be 131 ±1 for all these angles. This indicated that the drift velocity would be1.45 cm/µs. This leads to the conclusion that either the angles of incidence were notcorrectly set, or that the effective drift gap size is not 2 cm. Since the values from liter-ature lied in between, these drift velocities were used in the further analyses. In Figure4.4, histograms are presented for the distribution of clock counts. The window of clockcounts is the same for all events, with a sharp end at 210 clock counts and a smooth

21

Table 4.3: Measurements of the drift velocity, for various electric fields in the drift volume. Based onthe measurements performed with an angle of incidence of 45o.

Cathode potential (V) Electric drift field (V/cm) vdrift literature [6] (cm/µs) vdrift measurement-600 100 0.5 0.51 ±0.05-800 200 1.0 0.95 ±0.07-1000 300 1.35 1.26 ±0.07-1200 400 1.7 1.56 ±0.1-1400 500 1.9 1.66 ±0.13-1600 600 2.1 1.88 ±0.13

edge at around 70 clock counts. For the small angles of incidence this spread in clockcounts seems constant throughout the total window. As expected for the 45 degrees angleof incidence, the amount of hits for clock counts near the edges are smaller, since thesetracks are not going to both the top and the bottom of the detector.

Figure 4.4: Time clock histogram for datasets with different angles of incidence. The amount of hits perclock count is a summation over the entire dataset. The amount of tracks used therefore differs betweenthe histograms. Obtained using a MATLAB program, see Appendix B.

In Figure 4.5, the distribution for the angles of the reconstructed tracks are presented.The principal component analysis should be functioning well in determining the angleof incidence. A drift velocity of 1.35 cm/µs was used for the 300 V/cm drift field. Thereconstructed angles of incidence where consistently found lower, especially for the angleof 30 o. The small constant mismatch can be due to the fact that the DICE detector wasleaning forward a bit due to its weight, which decreases the actual angle of incidence.However, for the 30 o angle of incidence, it appears that the detector was not set correctly.The deviation in angles is primarily due to diffusion.

22

Figure 4.5: Results for the distribution of the angle of incidence, based on the reconstructed track fits.Obtained from events containing single tracks and a principal component analysis. A: Angle set at 0o.B: Angle set at 10o. C: Angle set at 30o, possibly mistaken for 25o. D: Angle set at 45o. Obtained usinga MATLAB program, see Appendix B.

The statistics for the angle distribution are given in Table 4.4. The mean deviation inthe reconstructed angle did not depend a lot on the angle of incidence chosen. The meandeviation is around 1.5 degrees for all measurement sets.

Table 4.4: Measurements of the drift velocity, for various electric fields in the drift volume. Based onthe measurements performed with an angle of incidence of 45o.

Angle set (x) Mean angle x (o) σx (o) Mean angle y (o) σy (o)0 1.4 1.5 -0.9 1.510 8.4 1.6 -1.7 1.630 25.0 1.2 -0.6 1.345 42.1 1.7 -1.0 1.5

In Figure 4.6 the mean position deviation between track and data points is given.It turned out that the mean deviation, orthogonal to the track is around 0.18 mm forthe small angles. For the 30 and 45 o angle of incidence, the mean deviation is larger.This could be due to the difference between longitudinal and transversal diffusion. Forthe small angles, the deviation is mainly determined by the transverse diffusion andfor 45 degrees the transverse and longitudinal diffusion play an equal role. It couldbe determined what these diffusion components actually are, but this requires furtheranalysis and comparisons with simulations.

23

Figure 4.6: Histograms of the mean deviation between track fit and data points, for different angles ofincidence.

4.3 electron efficiency

Another setting that was varied, was the grid voltage. By varying the grid voltage, themultiplication of electrons in the avalanche is affected. The rule of thumb is that every20 V of potential differences, matters a factor of two in the multiplication. Typically agas gain of 5000 is needed for proper detection. The grid potential has a direct influenceon the electron efficiency, the amount of electrons drifting towards the grid actually beingdetected in the pixels. Besides determining the gas gain, the ratio between the fields inthe drift volume and beneath the grid also determines the collection efficiency, as wasexplained in section 2.3. The electron efficiency is a very important quantity for thedetector, since loss of efficiency means loss of information and thus loss in accuracy. ForGridPix detectors, the efficiency should be in the order of 95-99 %.

In the test beam, the grid potential was varied for various angles of incidence. Forconsistency, the drift field was always kept constant, meaning that the cathode and guardvoltage were changed equally with the grid voltage. From former experience in the group,it is known that grid voltages around -400 V are of interest. An indication for the electronefficiency can be given by the total number of hits per track. Only complete single tracksare considered, using data from measurements with a small angle of incidence. Thenumber of hits for larger angles is larger because of two reasons: A The tracks withinthe detectable range are longer, B there are less multiple hits on pixels which are notdetected since the data was taken in TimePix mode. The Time over Threshold modewould have been more useful for determining the electron efficiency, since it is a bettermeasure for the energy deposit. Nonetheless the general trends can be observed in thisdata.

24

Table 4.5:0o Vgrid 0o average (Hits) 10o Vgrid 10o average (Hits) 30o Vgrid 30o average (Hits)360 28 360 - 360 41380 35 380 - 380 43400 42 400 46 400 49420 45 420 48 420 50440 46 440 50 440 52

4.4 DICE in magnetic field

Finally two magnets were added to the DICE detector, creating a homogeneous field withthe field vector directed towards the chip, perpendicular to both the detector chip and thebeam. The magnetic field was around 0.2 T. The momentum of the pions is too large tobe affected by the magnetic field, but secondary products could show curved tracks. Thedetector was placed at 90 degrees, the chip thus being parallel to the beam. In Figure4.7 some events are presented, showing secondary events.

25

Figure 4.7: 3D Plots of events with secondary products in a magnetic field. The angle of incidence is 90degrees. The z=0 plane corresponds to the chip level. The colors also indicate height, with blue beingclose to the chip, and red near the cathode top. The results are presented in the proper scale, using adrift velocity of 1.35 cm/µs for obtaining the height. Obtained using a MATLAB program, see AppendixA.

26

Chapter 5

Fourfold simultaneous readout

For applications in for instance the large LHC experiments, the usage of multiple de-tectors is of importance. The GOSSIP detectors were built for determining resolutionand electron efficiency in the reconstruction of curved radiation tracks along multipledetectors. Since there was little time during the test beam period, only experimentswere performed with the DICE detector. For the tests with the GOSSIP detectors, itwas important that a simultaneous readout could be performed. The original idea wasto put two, preferably three independent detector setups in one beam line and providethe separate Muros2 devices with the same trigger. However, the readout of the chip iscontrolled via the PC and for different setups, the timing would be difficult to synchro-nize. Instead, it would be better to read out four detectors at once with a single Muros2device and PC. This solves the timing problem and the advantage is that the Muros2and the Pixelman software are already been adjusted to be able to read out four chipsat once. This was done for a quad detector, which contains four MediPix chips tiledtogether in a single detector in order to have a larger detector area. The new thing now,is that the chips are separated in different detectors. To realize a simultaneous readout,the VHDCI cable used for the connection between detector and Muros2 was split to havefour connectors for the detectors, and one for the Muros2. In principal the same digitalarchitecture can be used, as long as the readout system can be ’fooled’ to think that aquad is connected. To make the set-up work it was required to understand the details ofthe digital architecture. In this Chapter the results of this investigation are presented.Starting with a single chip, then a quad and finally the fourfold cable.

5.1 Single pixel chip

Let’s first consider the digital infrastructure of a single pixel chip. It is not necessary tounderstand the whole picture, but the signals being send through the cables are relevant.The most important signals for operation are given in Table 5.1. Most of these signalsare of the type LVDS (Low Voltage Differential Signaling), meaning that information issend over a pair of cables in a differential way. The advantage is that common noise inthe cable does not affect the signal itself and it works best if the two signal carrying wireswithin the cable are twisted together to increase the correlation of the noise from theenvironment. This method of signaling also reduces the noise generated.

There are also signals used for setting the DACs, power, polarity and test pulses.These will not be discussed further. All the signals have been related to a pin connection

27

Table 5.1: Description of the most important signal types send to and from the chipSignal Functionf clock Fast external clock for the chipenable Token signal. Tells the chip or Muros2 when the next operation can start.data data transfer line. Data in gives the settings to the chip. Data out the actual measurements.m0 and m1 Set the operation mode chip combined: set matrix, read matrix, reset.shutter Start and stop acquisitionreset reset chip

on the VHDCI cable connector. A more detailed description on the Muros2 architecturecan be seen in [8] and for more information on the MediPix2 architecture, see [12]. Theworking principle of a single chip is shown in Figure 5.1. The block diagram shows howa readout is performed.

Figure 5.1: IO block diagram for MediPix chip. Source: [12]

28

5.2 Quad

Second step towards the fourfold simultaneous read-out, was examining the quad detec-tor. This detector basically consist of four pixel chips tiled together in order to increasethe sensitive area. In order to make such a detector operational, modifications wereneeded in the hardware and Pixelman software.

Figure 5.2: Schematic based on the original digital architecture scheme for a quad. The in and outcoming signals are given for the first and fourth chip. The middle two chips have been removed, sincetheir structures are identical. The signals selected in green go through all the four chips in series. Thesignals selected in red are delivered parallel to the chips. The blue selections concerns the power and isalso delivered parallel.

A TimePix quad was produced at Nikhef which could be used for testing. It turnedout that the detector needed an external power supply. For single chips powering via theMuros2 and the cable sufficed, but now four chips needed to be powered. The settings ofthe DACs were especially sensitive to the power. Setting the power to low, resulted in amalfunctioning in the readout.

5.3 Fourfold Cable

The idea of the fourfold cable is to split the VHDCI cable into four. Since there are64 wires involved, the splitting was organized on a chip board. The four connectors areseparated 10 cm from each other. A picture of the split cable can be seen in Figure 5.3.An external power supply connection was added using a copper rod (see middle of board),for the 2.2 V chip powering. Extra capacitors on the rod were used to create a buffer inthe power demand. As was explained, some signals go to or through the chips in parallel,some in series. The digital infrastructure is identical to that of the quad, although thesignals will have to travel a longer distance. The added time delays are in the order of

29

nanoseconds, while one clock cycle is 10 ns in standard operations.

Figure 5.3: Picture of the fourfold cable

After completion, the cable did not work with four detectors. However, using a singledetector on one of the connectors, and looping the others through, did work for all con-nectors. This implied that the cable was organized correctly. So further investigations ofthe signals being sent through were needed using an oscilloscope and differential probes(for the LVDS signals). Special connections for these probes were added on the board tostudy the signals between all the chips. By comparing with the measurements from thequad detector, the following could be concluded. The fourfold cable was indeed identifiedas a quad. Although the serial signals were adding up noise and longer times, the qualitywas good enough for normal functioning. While varying the clock frequency, it seemedthat near 100 MHz and beyond, the signals were getting messier. This can cause falsedigital interpretation of the signals. Examples of measurements with the oscilloscope aregiven in Figure 5.4.

Then the powering was studied. The chips required powering of 2.2 V. During opera-tion, four chips are requiring power. This combined results in a significant potential dropin the power, which affects the functionality of the chips. The capacitors were addedfor this reason, but did not seem to make a difference. Since the chips can cope withslightly higher voltages, the external power supply was set higher. It turned out that at2.3 V, the chips were functioning correctly. It also turned out that operating with a clockfrequency of 100 MHz was at the very edge of functioning, therefore the clock was set to80 MHz for stability. Although not all aspects were studied and understood, the fourfold

30

Figure 5.4: Measurements with the oscilloscope and differential probes. Left: Signals entering a quaddetector during threshold equalization over a span of 1 s. Procedures as setting matrix, performingmeasurement (shutter) and reading matrix were identified. Right: Signals exiting the last chip in thefourfold cable setup over a span of 100 ns.

cable was now ready to be used.

5.4 Test beam results

After the setup had proven to work reliably with the equipment used at CERN, the wholesetup was relocated to CERN again, where extra time was granted in the H8 test beamarea. This test was not part of this internship, so only a brief description will be givenon the results. In Figure 5.5 a typical result can be seen from the fourfold simultaneousreadout of three GOSSIP detectors and DICE. The latter was used as a reference. Thedifference between the two type of detectors can easily be seen and is caused by the dif-ference in drift gap.

31

Figure 5.5: Images of the first results from the GOSSIP test. On the top right the chip readout from theDICE detector is seen and the rest correspond to GOSSIP detectors. The top two readouts are rotated180 degrees w.r.t the bottom two readouts.

32

Chapter 6

Conclusion

To conclude this report, a short summary will be given on the main results and conclu-sions followed by some final words.

A theoretical background for GridPix detectors was presented and the general setupfor the readout of TimePix detectors was explained and proven to function. Data wasacquired in a test beam at CERN, using a GridPix detector with a 2 cm drift gap. Fordata analysis, programs were written in MATLAB. Using a principal component analysis,3D tracks could be reconstructed at once. Several statistical analyses have been made.They showed that the drift velocity can be determined from the data and that the angleof incidence could be reconstructed with an accuracy of at least 0.7 degrees, higher thenthe variation in the actual angle of incidence in the test beam. A detailed study of thedigital infrastructure was performed for a single chip, a quad and a fourfold cable. Thisled to a working setup for the simultaneous readout of four GridPix detectors has beenrealized, which is now being used for measurements for determining track fit resolutionof GOSSIP detectors.

Further analysis of the results from the test beam will be performed to determine theposition resolution and electron efficiency of the GridPix detectors. More data will betaken with the fourfold cable setup with different gas mixtures, and varying parametersas potentials and angles of incidence. Experiments aimed to visualize certain decay pro-cesses in GridPix detectors are also planned. A lot of the work presented in this reportwas a result of joint effort in the R&D group. A special thanks goes out to the all thepeople in this group and other Nikhef employees for their support and guidance.

33

Bibliography

[1] X. Llopart, R. Ballabriga, M. Campbell, L. Tlustos, and W. Wong. Timepix, a 65kprogrammable pixel readout chip for arrivel time, energy and/or photon countingmeasurements. Nuclear Instruments and Methods, 581:485–494, 2007.

[2] A. Fornaini, M. Campbell, M. Chefdeville, P. Colas, A.P. Colijn, H. van der Graaf,Y. Giomataris, E.H.M. Heijne, P. Kluit, X. Llopart, J. Schmitz, J. Timmermans, andJ.L. Visschers. The detection of single electrons using a micromegas gas amplificationand a medipix2 cmos pixel readout. Nuclear Instruments and Methods, 546:270–273,2005. Proceedings of the 6th International Workshop on Radiation Imaging Detectors- Radiation Imaging Detectors 2004.

[3] M. Chefdeville, P. Colas, Y. Giomataris, H. van der Graaf, E.H.M. Heijne, S. van derPutten, C. Salm, J. Schmitz, S. Smits, J. Timmermans, and J.L. Visschers. Anelectron-multiplying micromegas’ grid made in silicon wafer post-processing technol-ogy. Nuclear Instruments and Methods, 556(2):490–494, 2006.

[4] M. Campbell, E.H.M. Heijne, X. Llopart, P. Colas, A. Giganon, Y. Giomataris,M. Chefdeville, A.P. Colijn, A. Fornaini, H. van der Graaf, P. Kluit, J. Timmermans,J.L. Visschers, and J. Schmitz. Gossip: A vertex detector combining a thin gaslayer as signal generator with a cmos readout pixel array. Nuclear Instruments andMethods, 560:131–134, 2006.

[5] Walter Blum, Werner Riegles, and Luigi Rolandi. Particle Detection with DriftChambers. Springer Berlin Heidelberg, 2008.

[6] A. Sharma and F. Sauli. Low mass gas mixtures for drift chambers operation. NuclearInstruments and Methods, 350:470–477, 1994.

[7] Maximilien Chefdeville. Development of Micromegas-like gaseous detectors using apixel readout chip as a collecting anode. PhD thesis, University of Amsterdam, 2009.

[8] David S. S. Bello. MUROS2 USERS MANUAL, An inter-face board for the Medipix2 chips. NIKHEF, Amsterdam,http://www.nikhef.nl/pub/experiments/medipix/files/, July 2003.

[9] J. Visschers. Manual MUROS 2.0. NIKHEF,http://www.nikhef.nl/pub/experiments/medipix/muros.html, August 2008.

[10] Tomas Holy and Zdenek Vykydal. Pixelman manual. Czech Technical University inPrague, http://aladdin.utef.cvut.cz/ofat/Others/Pixelman, October 2008.

[11] Karl Pearson. On lines and planes of closest fit to systems of points in space. Philo-sophical Magazine, 2:559–572, 1901.

34

[12] MediPix collaboration. MediPix2.1 Users manual. CERN.

35

Appendix A

MATLAB program for trackvisualization

36

37

Appendix B

MATLAB program for trackstatistics

38

39