Medipix Project: Characterization and Edge Analysis

Transcription

1 Medipix Project: Characterization and Edge Analysis Matthijs Damen Student ID: University of Amsterdam and Nikhef, National Institute for Subatomic Physics July 5, 2011 Abstract This thesis is a description of a Bachelor project at the Detector R&D group of Nikhef (National Institute for Subatomic Physics). The main research topic within this group is Medipix based detector systems. This report will give a short overview of the Medipix and Timepix detector systems and then focus on the presentation of the results of experiments done with these detector systems. There are roughly two parts in this project. The first part consists of the characterization of the Medipix Detector using a dedicated laser setup. This involves investigating the response of the detector while changing several laser and chip parameters. The second part is edge analysis of the silicon sensor on top of the Timepix chip using the same setup. The results are used to validate simulations of the sensors edge done by [5]. The setup appears to be well suited to be used for characterization. Especially the fact that the laser can be focused on a single pixel is very useful. Also, most of the results appear to be in accordance with the simulations. 12 EC Bachelor Thesis Faculty: Physics and Astrophysics From to Supervisor: Jan Visser Second Assessor: Auke Pieter Colijn

2

3 Contents 1 Introduction 1 2 The Medipix and Timepix Detectors Concept Generations of the Medipix chip Analog side of the pixel Modes Photon Counting Mode Time Over Threshold Mode(ToT) Time of Arrival Mode (ToA) Characterization of a Medipix Detector The setup Results of the characterization experiments Laser frequency and pile-up Pulse width of the laser THL scans and pulse width Laser intensity Intensity and pulse width Sensor Edge Analysis The setup Results of experiments Stepping to the edge of the sensor Hecht relation Systematic even-odd pixel response Pixel size as a function of bias voltage Conclusion and discussion Characterization using laser setup Sensor Edge Analysis i

4 ii

5 1 Introduction This Bachelor thesis gives an overview of two months at the Detector R&D group of Nikhef (National Institute for Subatomic Physics). The design and development of the Medipix chip is a collaboration of many institutes including Nikhef, hosted by CERN [1]. The Medipix chip and its spin off, the Timepix chip, are both hybrid pixel detectors capable of detecting photons and charged particles. A new feature is the use of one or more thresholds to discriminate certain photon or particle energies. This project consists of roughly two parts. The first part is the characterization of a Medipix detector system using a dedicated laser setup. The response of one or more pixels is monitored while changing several parameters of both the laser and the chip, such as frequency, pulse width and threshold value (THL). The second part of the project is an analysis of the edge of the sensor material. One Medipix or Timepix chip measures 1.5 by 1.5 cm. If one would like to cover a larger area, the chips need to be tiled. To do this in an useful way, the edges of the chip need to respond in more or less the same way as the rest of the sensor, or a correction for the difference in response needs to be applied. By doing several measurements on the sensors edge, these possible differences are analyzed and compared to the predicted differences. Before the experiments are explained and the results are shown, a short introduction on the Medipix and Timepix chips is given. 2 The Medipix and Timepix Detectors In this section a short overview of the Medipix detector is given. The type of detector will be discussed and the different generations of the Medipix/Timepix detectors are summarized along with their detection modes. The detailed functioning and structure of the pixels and the chip as a whole goes beyond the scope of this report. For a thorough explanation refer to [2]. 2.1 Concept A Medipix based detector is a so called hybrid pixel detector, see Figure 1. This means that the sensor material and the read-out chip are separately produced and connected with solder bumps at each pixel. The sensor can be a cube of gas or be made out of semiconductor material such as silicon or gallium-arsenide. For the latter, the material is segmented in the same pattern as the read-out. The advantage of hybrid detectors is that the sensor and electronics on the pixel can be optimized separately. Also, it makes it easy to use different kinds of sensor chips with the same read out, for instance semiconductor sensors or gas sensors. The Medipix project was started to develop a theoretically noise-free hybrid pixel-detector. The purpose of the detector would be particle tracking in high-energy experiments. The sensor is capable of detecting and counting photons as well as charged particles. The signal to noise ratio (SNR)of this chip is greatly increased because thresholds can be set in each pixel separately. With these thresholds, a energy window can be created which gives the opportunity to collect information about the energy of particles or photons or count particles within a specific energy range. 2.2 Generations of the Medipix chip The first version of the Medipix detector, called Medipix1 or Photon Counting Chip (PCC), was finished in It was a sensor with 64 x 64 pixels and a pixel pitch of 170 µm, see Figure 1

6 Figure 1: The concept of hybrid detectors with the sensor and read-out connected through bump bonds [2]. 2. Although the detector was a succes, it had some limitations. The main limitations are listed below. The spatial resolution was limited due to the pixel size. Only positive input charges (holes) could be measured. The chips could not be tiled due to the large dead area around it. These limitations lead to the development of the Medipix2. Figure 2: A picture of the electronics chip (no sensor chip on top) of a Medipix1 detector [2]. With the arrival of the Medipix2 chip, most of the limitations were removed. The pixel pitch decreased to 55 µm (256 x 256 pixels), the dead area was reduced (although tiling was still a problem, more about this later in the report) and both positive and negative charges could be measured. 2

7 After the success of Medipix1 and Medipix2 there was a need to improve the detailed charge/energy measurements and to obtain information about the arrival time of particles and photons. A modification of the Medipix2 chip was made, called Timepix. This detector was capable of measuring Time of Arrival (ToA), Time over Threshold (ToT) and/or event counting independently in each pixel. These counting modes are explained in detail in section 2.4. The newest generation of the Medipix chip is the Medipix3. The most significant improvement compared to previous versions is the capability of dealing with charge sharing (overflow of charge to neighboring pixels). It was also made possible to read and write at the same time on a single pixel, because each pixel has two counters. The experiments presented here are all done with a Medipix2 chip or the Timepix chip. 2.3 Analog side of the pixel To understand the various counting modes of the Medipix/Timepix chips, some understanding of the pixel electronics is needed. Figure 3 shows the analog part of the pixel electronics. The charge generated in the sensor material is collected at the anode, which is designated input in the figure. The charge then goes to the Charge Sensitive Amplifier (CSA) that integrates and shapes the signal. As the name suggests, the output of the CSA is charge sensitive. This means that the amplitude of the output signal is proportional to the collected charge. Because the charge in turn is proportional to the deposited energy, the amplitude of the CSA output is related to the energy. The CSA output is compared with two thresholds; Discriminators (or Disc). The lower threshold is called THL, the higher THH. If the signal falls within the (energy-)window created by these two thresholds, the discriminators will give a high output signal. Depending on the counting mode, the counter on the pixel will be incremented. Figure 3: A schematic of the analog part of the pixel electronics [2]. 2.4 Modes In this section a brief overview of the different counting modes is given. 3

8 2.4.1 Photon Counting Mode When a photon or particle hits the sensor material, it deposits energy and this results in the creation of electron-hole pairs. The amount of charges created is directly proportional to the deposited energy of the particle or photon. In counting mode, the charges created by an incoming photon or particle are collected and compared to a certain threshold (the THL and THH values). If the total amount of charge exceeds this threshold, a counter is incremented. This method thus gives information about the number of photons that hit the sensor during the shutter or acquisition time (hence photon counting mode). The threshold gives the opportunity to eliminate noise, and therefore this increases the SNR. It also makes it possible to discern lower from higher photon energies. Both the Medipix and the Timepix detector are capable of operating in this mode Time Over Threshold Mode(ToT) In the right part of Figure 4, a simulated output of a pixel in ToT mode is shown. Here the Shutter shows the time interval for which the pixel is active. The CSA amplifies the collected charge and the signal, which goes to the discriminator (Disc). If the output of the CSA is over threshold, the discriminators output is high. Then a clock starts giving pulses at a certain rate (40 MHz) to the pixel counter, which gets incremented at that rate. When the CSA output goes under threshold again, the clock stops. In this way, the number of pulses is a measurement of the time the signal was over threshold. Because the amplitude of the CSA signal is energy dependent, so is the number of counts. This means that the number of counts is a direct measurement of the deposited energy of the incoming photon or particle. Only the Timepix chip is able to count in this mode. Figure 4: Output of a pixel in time of arrival mode (left) and ToT mode (right) [2]. The top part shows the shutter or acuisition time interval. The second part shows the discriminator output. The third part shows the CSA output. Here, the straight horizontal line indicates the threshold level. The fourth part shows the clock to counter signal Time of Arrival Mode (ToA) In time of arrival mode, the counter starts to get incremented from the moment the output of the discriminator goes high. The pulses are again given by the same clock as in ToT mode. Independently of what happens during the rest of the shutter time, the counter gets incremented. The number of counts is therefore a measurement of the time the pixel was first over threshold 4

9 and thus of the time the charge arrived at the pixel. This mode is for instance used with a gas based detector system (instead of a semiconductor sensor) to capture particle tracks. Only the Timepix chip is able to count in this mode. 3 Characterization of a Medipix Detector The first part of the project was to characterize the Medipix chip using a laser setup. This means several parameters of both the chip and the laser are changed and the behavior of a single or a group of pixels is monitored. If the pixels do not behave as expected, the result is analyzed and a possible explanation is given. If possible the hypothesis is tested with another experiment. 3.1 The setup For all of these measurements a single Medipix chip is used in photon counting mode. A Fitpix USB interface is used to read out the chip. The program Pixelman [8] can preview the chip s response and is used to store the counts of the pixels in a file. In most of the experiments the laser was focused and the settings were applied such that only a single pixel was responding. (a) The laser and chip within the enclosure (b) Laser and stage/holder (c) The Medipix chip with Fitpix Figure 5: On the left an overview of the setup within the enclosure to shield it from light. In the middle the optical fiber (yellow) which is held in place by the metal cylinder and the stage which holds, in this case, the double Timepix chip. On the right a photo of the Medipix chip with the Fitpix USB interface. Figure 5 shows some pictures of the laser setup. Because the chip is used to detect photons, it needs to be shielded from unwanted light. For this reason the whole setup is inside a metal enclosure. The wires that connect the laser with the control devices go through a hole in the enclosure. The hole is filled with foam to prevent light entering. The two lasers used, one with a wavelength of 660 nm, the other of 980 nm, are inside the enclosure. The input for these laser comes from a pulse generator, which is outside. The pulses are monitored using an oscilloscope. A yellow optical fiber with a focuser at the end transports 5

10 the laser pulses to the detector. This fiber can be seen in the left and middle picture of Figure 5. The focuser is held in place by the metal cylinder. The holder is attached to a height adjustable stage which can be moved with micrometer accuracy in the z direction. The detector is placed underneath the laser holder on a precision table which can be moved with the same accuracy in the x and y direction. The movement of both the laser and the detector are controlled through a Labview program. The detector also needs a bias voltage. This is supplied using an adjustable power supply, which is also outside the enclosure. 3.2 Results of the characterization experiments Laser frequency and pile-up Figure 6: A graph of the counts of a single pixel on the y-axis versus the frequency of the laser pulses on the x-axis. The first part of the graph is fitted (red line) with a linear relation. The inset at the top right shows the values of the parameters of the fit, which has the form y = p 1 x + p 0 Figure 6 shows the counts of a single pixel as a function of the frequency of the laser pulses. Because the Medipix chip is in counting mode, the pixel will add a count when it goes over threshold. An individual photon of 660 nm is not enough to give a count, but a single pulse of the laser, consisting of multiple photons, is. This means the chip counts the laser pulses. The acquisition time of the experiment was seconds. This means that the shutter of the pixel is opened for seconds and this explains that for instance with a frequency of 200 khz, the pixel gives 200 counts. The result is fitted with a linear relation (red line) up to 800 khz. At frequencies above 800 khz the relation breaks down. This is due to the pile-up effect. Pile-up happens when one injects too much charge in one pixel or the counts come too fast after each other. This can happen when raising the pulse width, intensity or frequency of the laser too much. The amplitude of the CSA output is proportional to the injected charge. When the output of the CSA has gone over thresholds, the signal needs some time to decay, to go under threshold again, before another pulse can be measured and counted. The slope of this decay is 6

11 more or less constant (independent of the amplitude of the signal), which means that the more charge is injected, the longer the CSA needs to discharge. Now if there is too much charge injected, the amplitude of the CSA output goes far over threshold. The signal starts to decay, but before it can go under threshold, another pulse/signal is coming. This new pulse raises the output of the CSA and the pulse is not counted. Also, when the pulses are too close together in time, the same effect happens. Because the trigger for a count is the moment the signal goes over threshold, instead of counting the number of photons, or in this case pulses, the pixel will count Pulse width of the laser Figure 7: A graph of the counts of a single pixel on the y-axis versus the pulse width of the laser pulses on the x-axis. This type of measurement is also used to create the graphs of Figure 8. Besides the frequency of the laser, the pulse width can also be tuned. The pulse width is the duration (in seconds) of the pulse given to the laser. Figure 7 shows the counts of a single pixel as a function of the pulse width of the laser pulses. At low pulse widths, the charge deposited is not enough to get any counts. When there is enough charge, the counts rise fast to the maximum. The pulse at which this occurs will be called the rise pulse width from now on. At high pulse widths, the pixel goes in pile-up, and does not count anything anymore. This is called the fall pulse width. The difference between rise pulse width and fall pulse width is called pulse width durations. The lower threshold (THL) of the pixel determines the pulse width at which the pixel starts and stops counting. Rising the THL will make the graph move to the right, while lowering it will make it move to the left THL scans and pulse width At different pulse widths of the laser pulses, a scan is done of the THL (lower threshold) value. When the THL value is increased, at a certain point the pixel stops responding, because the signal does not go over threshold anymore (do not mistake this for pile-up). One of those scans 7

12 (a) Example scan with fit (b) Final result Figure 8: THL scan at different pulse widths. The graph on the left shows the counts of a pixel on the y-axis versus the THL value. The graph on the right shows the T HL 50 value on the y-axis versus the pulse width on the x-axis. is shown in Figure 8a. This curve is fitted with a s-curve (red line) for every value of the pulse width: s(x) = p p 1 exp [ p 2 (x p 3 )] + p 4 (1) The parameters of the fit are used to compute the THL value at which the pixel stops counting and the errors on this value. The mid point of the curve is taken as this THL value (which will be called T HL 50 ), while the error is given by the 5% and 95% point of the s-curve. The values of T HL 50 for different pusle widths are all plotted in a final graph, which is figure 8b. This result is fitted with a linear fit. The points seem to agree with the fit quite well, although some more points (at higher pulse widths) would have been useful to be more certain about this. As the injected charge should go linear with the pulse width, a linear relation is indeed expected Laser intensity The specifications of both lasers show how the output power behaves as a function of a DC input voltage. Because the laser is not used in continuous operation, but pulsed, the relation might be slightly different. Figure 9 shows the output of the laser, measured using a PIN diode, as a function of the input voltage. Here, the input voltage is the height of the pulses. Between 500 mv and 2000 mv the relation seems exponential. This fact is used in subsequent experiments Intensity and pulse width The figures 10, 11 and 12 all belong to the same measurement. At a certain laser intensity (given in mv), a scan is done of the pulse width (as in figure 7). This relation is fitted with the s-curve of eq. 1 and the rise pulse width, fall pulse width and width of the function is determined. The different values of the rise pulse widths obtained, are plotted against the intensity (this is related to the input voltage), as is done with the fall pulse widths and the width of the step function. This leads to the figures shown. 8

13 Figure 9: Laser intensity, on a logarithmic y-axis, as a function of the input voltage, on the x-axis. The output is measured using a PIN diode. (a) At rise (b) At fall Figure 10: The graph on the left shows the rise pulse width (y-axis) for different laser intensities (x-axis). On the right the fall pulse width. Both are for the 660 nm laser According to Figure 9 the output of the laser goes exponentially with the input voltage between 500 mv and 2000 mv. Also, the charge deposited is expected to be linear with the pulse width. This means that the graphs of the rise and fall pulse widths and the widths of the step function should decay exponentially with the input voltage. This indeed seems to be the case. Note that there are only 4 points in Figure 11a. This is because the pixel was already counting the maximum counts at the minimum pulse width, which is 5 ns. Therefore, no rise pulse width could be determined. This also means that we can only plot the width of the block for these 4 points. 9

14 (a) At rise (b) At fall Figure 11: The same two graphs as in Figure 10, but now for the 980 nm laser. (a) 660 nm laser (b) 980 nm laser Figure 12: The width (rise pulse width - fall pulse width, called pulse width duration on the y-axis) of the response curve for different laser intensities. The results for both lasers are shown. 10

15 4 Sensor Edge Analysis 4.1 The setup For the second part of this project, the same laser setup was used. Only some little modifications have been made. For the experiments done on the edge, two Timepix assemblies with edgeless sensors positioned next to each other were used instead of a single Medipix. The sensor material is provided by VTT [6]. These two chips use the Relaxd-board [7] to communicate with the computer. The chip and board are mounted on a holder, which is bolted on a stage. The construction is shown in Figure 13. A Newport x-y-stage was used with the actuators removed, because the setup was already on a movable table. An advantage of this stage is that the chip is more secure and movement is more precise. For the measurements done, the Timepix is used in Time over Threshold mode. The rest of the setup is exactly the same as before. Note that some of the experiments are done with the 980 nm laser instead of the 660 nm. (a) Newport stage with Relaxd and chip (b) Double Timepix Figure 13: Stage and chip setup used in the edge analysis experiments. 4.2 Results of experiments The difference between the edge pixels and pixels in the middle of the chip is mostly due to electric field anomalies near the edge of the sensor material. Simulations of this field have been done by [5]. These simulations show the shape of the electric field near the edge for different bias voltages, see Figure 14. The purpose of the experiments is to check whether the real situation is in accordance with the simulations Stepping to the edge of the sensor To probe the shape of the edge pixels as well as normal pixels, the laser is focused in the middle of a pixel. Then the detector is moved a (few) micron(s), while the response of the surrounding pixels is monitored. In this way a graph can be made of the behavior of the pixels; when they turn on and off. First such a graph is made for two pixels in the middle of the chip, as a reference. This is shown in Figure 15. The laser is focused on a certain pixel and then the table with the chip is moved toward the neighboring pixel (in the x direction). Halfway the experiment, pixels below the pixel of interest (so in the y direction) started to respond, possibly due to the fact that the 11

16 Y [um] Y [um] ElectrostaticPotential [V] -4.0E E E E E E E X [um] E X -1.0E+02 (a) 10V (b) 20V Y [um] Y [um] ElectrostaticPotential [V] E E E E E E E X [um] E X -1.0E+02 (c) 40V (d) 100V Figure 14: Simulations of the shape of the electric field at the edge of the sensor for different bias voltages. Note that the sensor appears upside down. The y-axis indicates the thickness of the sensor (150 µm) with the pixel electronics at 0 µm and the top of the sensor at 150 µm. The lines without arrows connect points of equal potential (field lines). The lines with arrows are perpendicular to the field lines and indicate the border between the last three pixels. chip was not perfectly aligned with the table. As this was not intended, compensation was tried by summing the response of the pixels in the same column and use this as the correct value for the pixel of interest. This procedure probably slightly increased the number of counts, which could explain the small rise in maximum pixel counts in Figure 15. The response of the pixels below the pixel of interest could be due to a slight angle in the placement of the chip. While the table is moved only in the x direction, the laser will move in the x and y direction compared to the pixels. Because the stage cannot be easily rotated, it was not possible to correct for the angle. From this experiment, the size of the laser spot on the sensor can also be determined. The pixel pitch is 55 µm, but the second pixel, indicated by black squares, responded for about 97 µm. This means the spot size is about 42 µm. The same experiment is done for the edge pixels. The laser is focused at the fifth pixel from the edge. The chip is then moved in small steps until no pixel responded anymore. Note that 12

17 Number of ToT counts [25 ns] Distance from centre pixel 104 [µm] Figure 15: A plot of the ToT counts versus the distance from the center of a pixel in the middle of the sensor. The 0 on the x-axis indicates the starting point. The difference in maximum counts is probably an artifact from charge sharing. the zero position indicates the assumed edge of the sensor. The result acquired with the 660 nm laser is shown in Figure 16. Figure 17 shows the same, but now for the 980 nm laser. As one can see the shape of the fifth, fourth and third pixel are quite normal for both lasers; the shape is the same as for the middle pixels. The shape of the second pixel is distorted and also looks different for both lasers, as is the case with the first pixel. A comparison with the simulation of the electric field indeed shows that the second pixel could be larger than normal. This effect is most present at the top of the sensor (bottom of Figure 14) and decreases when going deeper into the sensor. The second pixel seems largest in the results of the 660 nm. This is because the photons of this wavelength are absorbed in the first few microns of the sensor material. Photons with a wavelength of 980 nm are absorbed much deeper into the sensor, around 100 microns. The effect at 100 microns is much less prominent and therefore the second pixel looks more normal in the measurements done with the 980 nm laser Hecht relation The result of an experiment where the response of some pixels is monitored while the bias voltage over the sensor material changes is shown in Figure 18. The data points are fitted with a so-called Hecht relation: [ ( )] µτ Q(U) = Q 0 L 2 (U U dl 0) 1 exp (2) µτ(u U 0 ) Here, Q is the charge collected by the pixel, Q 0 is the total charge generated in the sensor material, µ is the mobility of the collected charge (electrons or holes), τ is the mean lifetime of the collected charge and U is the voltage over the sensor. L is the distance at which the charge 13

18 Mean ToT value [25 ns] Pixel coordinate (1, 160) (2, 160) (3, 160) (4, 160) (5, 160) Laser position [µm] Figure 16: The same type of plot as Figure 15, only this time the starting position is at pixel 5 from the edge of the sensor (coordinates (5,160)). The 0 position indicates the assumed edge of the sensor material. The ToT counts of all the pixels from the edge to pixel 5 are monitored until all the pixels count nothing (which means the laser is not pointed at the chip anymore). Mean ToT value [25 ns] Pixel coordinate (1, 160) (2, 160) (3, 160) (4, 160) (5, 160) Laser position [µm] Figure 17: Exactly the same measurement as is shown in Figure 16 but this one was done with the 980 nm laser. The first 3 pixels look the same, but the second pixel from the edge is clearly a bit smaller. The response of the edge pixel is better than with the 660 nm laser. is generated and d the thickness of the sensor. This relation states how efficient the pixel is collecting charge, as a function of the bias voltage (and the sensor material properties like µ and τ). The product of µ and τ gives information about the recollection probability of the generated 14

19 charge. A high value means that the probability is low, and the charge will likely be collected. A low value means the opposite. Mean ToT value Pixel coordinate (1, 160) (2, 160) (103, 160) (104, 160) Bias voltage [V] Figure 18: The ToT counts of 4 pixels versus the bias voltage. One edge pixel, a pixel in the second column and 2 pixels in the middle of the sensor have been monitored. Obviously the edge pixel does not behave normally. The second pixel is more like the middle pixels. The lines through the plots are fits using the Hecht relation of eq. 2 Figure 18 shows the Hecht relation for 2 pixels in the middle (103 and 104) and the first and second pixel from the edge. The results for the second pixel are more or less the same as for the middle pixels. This means the charge collection properties of the second pixel are about the same as for a regular pixel. The first pixel, however, collects far less charge. From the fit follows that the product of µ and τ is about 10 times lower than for the other pixels. That the collection capability of the edge pixels is so low could be due to impurities at the edge, but also due to damage (from cutting the sensor material) or the electric field anomalies near the edge. However, it is likely that an error was made in the analysis of the data. The Hecht relation is about charge collection efficiency of a pixel, under the condition that this pixel is able to collect all the charge created, were the efficiency 100%. In these measurements, the diffusion at lower bias voltages is large enough to let neighboring pixels respond. This means that the charge created is not totally collected by a single pixel, but the theory behind the Hecht relation does not take this diffusion into account. A solution could be to sum the responses of the surrounding pixels for each bias voltage. Due to time limitations of this project this has not yet been done Systematic even-odd pixel response Some of the results of the experiments done, show a hint of systematic error in the even-column pixels versus the odd-column pixels. The maximum response of the even numbered columns seemed systematically lower than of the odd numbered columns (see for instance Figure 17). To check whether this is indeed the case, the maximum response of 20 pixels in a row was measured. The result is shown in Figure 19 15

20 Mean ToT-value Pixel number Figure 19: A test of systematic even-odd pixel column response. On the y-axis the number of ToT counts is shown, on the x-axis the pixel number. The variations seem random (although quite large), so systematic (mis)behavior does not seem to be the case. In this graph the difference in maximum response does not look systematic. Although the differences in the responses are quite high, up to 800 counts, systematic even-odd pixel response difference does not seem to be the case Pixel size as a function of bias voltage Another way of measuring the pixel size, this time also as a function of the bias voltage, is to record the position at which the edge pixel starts to respond and the position at which it stops counting again. This is basically the same as Figures 16 and 17, but faster, less detailed and for different bias voltages. The results of the turn off and turn on point for an edge pixel are shown in Figure 20. The measurement is performed for bias voltages of 1, 2, 3, 4, 5, 10, 20, 40 and 60 volts and for both lasers. The 0 position indicates the edge of the sensor material. Because the laser spot has a certain extend (about 40 µm, so not a point), the pixel can still respond although the center of the laser spot is not on the sensor material anymore. The size of the edge pixels seems to increase with larger bias voltages. This is partly in accordance with the simulations of Figure 14. The border between pixel 1 and 2 indeed moves to the edge of the sensor when the bias voltage decreases. This makes pixel 2 larger and pixel 1 smaller. But according to the simulations, at 20 V the edge pixel should be totally closed by the electric field. Unless one shoots into the side of the sensor material, the edge pixel should not respond below 20 V. This is not what the results show, which means the shape of the electric field could be different than the simulations show for lower bias voltages. In section 5 a different experiment is proposed to probe the shape of the pixels in a better way. Figure 21 shows the difference between the turn on and turn off point of Figure 20. This should give the width (or size) of the pixel. The second line, which is labeled corrected, shows the same values, but now corrected for the assumed laser spot size. The laser spot size is derived 16

21 Distance from edge [µm] nm right edge left edge 980 nm right edge left edge Bias voltage [V] Figure 20: The turn off and turn on point of an edge pixel ( pixel 1 ) for different bias voltage values for both the 660 nm and the 980 nm laser. The micrometer scale on the y-axis designates the distance of the center of the laser spot from the edge of the sensor, where a negative value means not on the sensor anymore. Width [µm] nm effective width corrected 980 nm effective width corrected Bias voltage [V] Figure 21: The difference between the turn on and turn off point of Figure 20. This gives the effective width of the edge pixel. The corrected value mean that the width of the laser pulse is compensated for. from Figure 15. For the 660 nm laser, the spot size is about 42 µm, for the 980 nm laser this is about 35 µm. This corrected value should show the true size of the pixel for different bias voltages. 17

22 5 Conclusion and discussion As this project has two quite distinct parts, both will have a separate conclusion and discussion. 5.1 Characterization using laser setup The laser setup appeared to be very well suited for characterizing the Medipix chip. Especially the ability of focusing the laser spot on a single pixel makes the experiments very clear and the result easy to interpret. All the attention can go to the response of this single pixel and data reduction while analyzing the results is quite easy. An improvement of the setup could be a stage that can be used to rotate and tilt the chip. In this way the laser can also be used to shoot in from an angle, or even the side, and positioning of the chip is more accurate. Correcting for a tilt or angle in the placement of the detector is easily done. Also, a type of laser which gives a smaller laser spot in the focus would make measurements of sub pixel features possible. Both these options will make future experiments more easy. 5.2 Sensor Edge Analysis The edge analysis experiments show that the electric field at the edge is indeed distorted. The results of Figure 16 and 17 show that at bias voltages of at least 20V the effective width of the second pixel somewhere deep in the sensor is indeed smaller than at the top. This is in accordance with Figure 14c and 14d In contrast, the effective width of the first pixel at low bias voltages does not behave like the simulations shown in Figure 14b and 14a. In those Figures, the last pixel is completely closed by the electric field. The results show something different, see Figure 20. Although the pixel gets smaller for lower bias voltages, it will always show some counts. This means it cannot be closed. An experiment where one shoots into the side of the laser could give more detailed information about the border between the first and second pixel. This can even be done with only a single wavelength, which is absorbed around the assumed depth of the border. For this to be possible, the setup needs to be upgraded with the stage mentioned above. 18

23 References [1] Medipix website, available at: [2] Xavier Llopart Cudié, Design and Characterization of 64K Pixels Chips Working in Single Photon Processing Mode. Mid Sweden University Doctoral Thesis 27, Department of Information Technology and Media, Sundsvall, [3] Lukas Tlustos, Performance and Limitations of High Granularity Single Photon Processing X-ray Imaging Detectors. Doctoral Thesis, University of Technology, Vienne, Austria, [4] Jan Visser et al, Particle Detection. [PowerPoint Slides] Master course UvA, VU and UU, Amsterdam, the Netherlands, [5] Marten Bosma et al, Edgeless planar semiconductor sensors for a Medipix3-based radiography detector. Proceedings of the 13th International Workshop on Radiation Imaging Detectors, Nikhef, Amsterdam, to be published. [6] VTT, Technical Research Centre of Finland. Available at: [7] Jan Visser et al, A Gigabit per second read-out system for Medipix Quads. Nuclear Instruments and Methods in Physics Research A 633, S22S25, [8] T. Holy et al, Data acquisition and processing software package for Medipix2. Nuclear Instruments and Methods in Physics Research A 563, 254,