In this paper, we report on these developments, with an emphasis on the challenges involved building readout systems and the potential of the Medipix family ASICs in this field of research.
Radiation Measurements 140 (2021) 106459 Contents lists available at ScienceDirect Radiation Measurements journal homepage: http://www.elsevier.com/locate/radmeas X-ray imaging at synchrotron research facilities Cyril Ponchut a, *, Nicola Tartoni b, David Pennicard c a Detector Unit, European Synchrotron Radiation Facility (ESRF), 71 Avenue des Martyrs, F-38000, Grenoble, France Diamond Light Source Ltd., Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, UK c Photon Science Detector Group, Deutsches Elektronen Synchrotron (DESY), Hamburg, 22607, Germany b A R T I C L E I N F O A B S T R A C T Keywords: X-ray detectors Hybrid detectors Solid-state detectors Synchrotrons At synchrotron facilities, many X-ray imaging and diffraction experiments require pixel detectors with minimal noise, high speed and reasonably small pixel size, all of which can be achieved with the Medipix ASIC family So, ESRF, Diamond Light Source and DESY have developed detector systems based on Medipix In this paper, we report on these developments, with an emphasis on the challenges involved building readout systems and the potential of the Medipix family ASICs in this field of research Introduction In the late 1990s most large-area 2D X-ray detectors at ESRF and at other 3rd generation synchrotron facilities were based on chargecoupled (CCD) sensors optically coupled to X-ray converter screens These detectors were, and still are, successfully used for example for Xray diffraction, imaging (Bravin et al., 2004) or spectroscopy with energy-dispersed beams (Labiche et al., 2007) However some short comings of large-area CCD systems, like the presence of electronic noise or the limited speed of the charge transfer readout, make them nonideal for applications where a large dynamic range, a high detectivity at low signal level, and/or high frame rates are required Lag effects as well as light scattering in the X-ray converter screen also contribute to reduce the achievable temporal and spatial resolutions By providing noise-free, direct detection as well as readout times below ms, the Medipix2 (Llopart et al., 2002) and the Timepix (Llopart et al., 2007) photon-counting readout chips were then appearing as a strong alternative to CCDs in these cases Then the great leap forward achieved by Medipix3 (Ballabriga et al., 2013) with higher frame rates, higher count rates, as well as improved spectroscopic capabilities owing to the charge summing mode, broadened even more the field of possible applications at synchrotrons Due to their smaller pixel size than other photon-counting pixel de vices available at that time, Medipix chips also provided the highest spatial resolution Moreover, thanks to their compatibility with high-Z sensor materials working in electron collection mode like CdTe, CdZnTe, or GaAs, Medipix chips are also enabling synchrotron experi ments at medium and high energies which is of crucial importance in particular for ESRF which operates several high-energy beamlines By opening the path to an ever extending range of new applications Medipix chips were therefore undoubtedly promising to a bright future on synchrotron beamlines This motivated the development and commissioning of Medipix-based pixel detectors systems at ESRF, Dia mond Light Source, and DESY, as depicted in the next sections Medipix at ESRF 2.1 MAXIPIX detector system In the early 2000s, soon after joining the Medipix2 collaboration, ESRF started the development of the MAXIPIX photon-counting detector system (Ponchut et al., 2011) based on Medipix-2 and later Timepix readout chips This project was aimed at bringing the advantages of Medipix2 and Timepix chips to ESRF beamline users: the noise-free photon-counting detection, the fast readout time, the small pixel size, and the wide energy range thanks to the chips compatibility with high-Z sensor materials like CdTe, CdZnTe, or GaAs A modular hardware ar chitecture was retained, in order to facilitate adaptation to specific beamline requests A version of MAXIPIX with five Timepix chips (256 × 1280 pixels) is shown on Fig This explained the rapid deployment of MAXIPIX detector systems, nowadays totalling about 21 commissioned units operated in user mode on ESRF beamlines, plus a few units operated at DESY, Diamond light source and Soleil synchrotrons., The next sections are a non-exhaustive review of the MAXIPIX detector –and therefore of Medipix2 and Timepix chips– applications at ESRF * Corresponding author E-mail address: ponchut@esrf.fr (C Ponchut) https://doi.org/10.1016/j.radmeas.2020.106459 Received February 2019; Received in revised form September 2020; Accepted September 2020 Available online 28 October 2020 1350-4487/© 2021 The Authors Published by Elsevier Ltd This is an open (http://creativecommons.org/licenses/by-nc-nd/4.0/) access article under the CC BY-NC-ND license C Ponchut et al Radiation Measurements 140 (2021) 106459 Fig MAXIPIX detector, × version (70 × 14 mm2 input field, 327 kpixels) Left: detector head Right: Detector module implementing five Medipix2 or Timepix chips in a row assembled on a monolithic silicon sensor (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) Fig Raman spectrometer at ID20 Left: spectrometer gantry Right inset: single-chip MAXIPIX detector module 2.2 Inelastic scattering The first Medipix2-based X-ray detector commissioned at ESRF was the focal plane detector of the diced crystal analyzer on ID16 inelastic scattering beamline (Huotari 2006) This detector consisted of a Medi pix2 chip with 500 μm thick silicon sensor coupled to the parallel port of a PRIAM (Ponchut et al., 2007) readout interface, and an ESRF Linux acquisition workstation Photon-counting detection was desirable for efficient detection of the low available X-ray signal at the detector input By exploiting the energy-dispersive property of the analyzer and the small pixel size of Medipix2, the effective energy resolution of the in strument could be improved from 190 meV to 27 meV (Huotari et al., 2006), with a detector contribution of only meV thereby leaving room for further improvement Later, the Medipix2 module was replaced by a Timepix module Successful use of Timepix in this application led the ESRF to retain MAXIPIX/Timepix as the detector platform for the new inelastic scat tering beamline ID20, consisting of two end stations as detailed below: The ID20 Raman scattering end station consists of a gantry embed ding six multimirror analyzers, the sample stage and an X-ray focusing optics as shown in Fig Each analyzer is fitted with a custom detector module implementing one single Timepix chip with a 500 μm thick silicon sensor (Huotari et al., 2017) As the instrument operates in near-backscattering geometry, a compact detector module with a small distance between the pixel matrix and the physical edge of the module Fig RIXS spectrometer at ID20 had to be designed All detector modules are connected to MAXIPIX controllers with m long VHDCI cables in order to enable independent motion of each analyzer in a wide angular range The detectors are operated simultaneously from the acquisition workstation The 10 ms C Ponchut et al Radiation Measurements 140 (2021) 106459 detector readout dead time imposed by the cable connection is negli gible compared to the typical 1–10 s exposure times in this application The ID20 Resonant Inelastic Scattering (RIXS) end station imple ments a five-mirror analyzer fitted with a 5x1 MAXIPIX detector mod ule, as shown in Fig (Moretti-Sala et al., 2018) Focusing of each mirror on a different chip enables individual centering and quality checking of each focal spot This instrument can achieve an energy resolution of 25 meV, about 50% of which is the pixel size contribution 2.6 Imaging A MAXIPIX system implementing Timepix chips and a CdTe sensor with 512 × 512 pixels was used to quantify potential benefits of X-ray phase contrast imaging (XPCI) for X-ray mammography in terms of patient dose (Diemoz et al., 2016) Use of phase contrast instead of absorption contrast allows higher X-ray energies to be used, thereby reducing the attenuation cross-section of the observed tissue and thus the radiological dose The beam energy was 60 keV monochromatic, instead of the 15–25 keV typical energy with conventional mammog raphy X-ray tubes The mm thick CdTe sensor provided nearly 100% detection efficiency at this energy and the Timepix1 chip spatial reso lution of 55 μm was ideally suited for breast imaging The high signal-to-noise ratio even at low X-ray flux provided by Timepix1 was also a key feature for dose reduction As a result, compared to a con ventional mammography unit the XPCI method with the CdTe MAX IPIX/Timepix detector achieved a dose reduction from 3.5 mGy to 0.12 mGy on the same breast sample, for images of equivalent diagnostic value 2.3 X-ray photon correlation spectroscopy (XPCS) XPCS is a technique exploiting the coherence of the synchrotron beam for dynamical studies at nanometre scale of complex fluid systems like colloidal suspensions, surfactant solutions and thin-film structures (Gruebel et al., 2004) (Leheny 2012) Initially, XPCS experiments per formed at ID10 ESRF beamline were using direct detection CCDs (Gruebel et al., 2004), providing optimum spatial resolution as well as high sensitivity However their readout times in the range of seconds as well as their gradual and irreversible damage under X-ray radiation were severe limitations Replacing direct detection CCD by a 512 × 512 pixels MAXIPIX system revolutionized the XPCS technique at ID10 by improving data statistics and dynamic range (Caronna et al., 2008) and by giving access to time scales in the ms range (Orsi et al., 2012), (Fluerasu et al., 2010) The smaller size of Timepix pixels as compared to any other available photon-counting device also better matches the typical transverse coherence lengths of 10 μm × 100 μm (Gruebel 2004) of the ESRF SR beams used for XPCS experiments 2.7 High-Z sensor material studies Medipix2 and Timepix readout chips in combination with finely focused monochromatic synchrotron X-rays, were also used at ESRF and other synchrotrons as tools to explore the properties of semiconductor pixel sensor materials: detailed analysis of CdTe sensors instabilities and spatial defects (Ruat et al 2012, 2014), measurement of charge trans port properties and analysis of the fine structural details of GaAs:Cr sensors (Ponchut et al., 2017) 2.4 Coherent X-ray diffraction imaging (CXDI), ptychography Medipix3 based detector developments at Diamond Light Source These techniques make it possible to perform 2D or 3D imaging of small isolated objects (CXDI) or of extended objects scanned with a focused beam (ptychography) at the nanometre scale Images of the objects in the direct space are reconstructed from diffraction patterns acquired in the far-field region (Favre-Nicolin et al., 2017), using inverse Fourier transform and phase retrieval algorithms A CXDI experiment carried out at ID10 using a 2x2 MAXIPIX system provided 3D images of a sample with a voxel size of 24 nm (Chushkin et al., 2014) Photon-counting detection combined with frame accu mulation provided the necessary dynamic range to accurately record the intensities of the diffraction patterns which can extend over more than orders of magnitude in the same frame Moreover the small pixel size allowed the far-field detection geometry to be kept within reasonable size One possible concern with this technique is the high incident in tensity close to the beam axis which may exceed the counting rate range of Timepix chip This can be overcome by inserting a calibrated atten uator in the central beam path between the sample and the detector 3.1 Summary and motivation Diamond Light Source (DLS) has developed a suite of detector sys tems to exploit the Medipix3 ASICs in photon science experiments These systems are Excalibur, a large area position sensitive detector that tiles 48 ASICs, Merlin, a light and very compact position sensitive de tector that tiles up to ASICs, and Lancelot, a beam position and in tensity monitor that uses one ASIC Excalibur is described in section 3.2 and Merlin and Lancelot are described in section 3.3 The development of a large area position sensitive detector is the reason why DLS joined the Medipix3 collaboration By the time DLS joined the collaboration in late 2007 it was clear that the photon counting hybrid pixel detector technology presented many advantages with respect to the position sensitive detector technologies in use at synchrotron facilities until then (fundamentally scintillating plates coupled to CCD imaging sensors by tapered fibre optics bundles) Particularly important were the capability of providing quantum limited image noise and to be able to run at much faster frame rate However, the only commercially available detector systems had a pixel size too large for a number of applications One of these applications at DLS was the Coherent Diffraction Imaging beam line I13 that was under con struction The ideal pixel size for I13 use case was 50 μm which was far smaller than the 172 μm size of the aforementioned commercially available system Thus, a clear science case emerged to develop a de tector system based on the Medipix3 ASIC which enables to use sensors whose pixel size is 55 μm and therefore close enough to the ideal requirement of the beam line I13 Among the drawbacks of hybrid de tector technology are the relatively limited size of the semiconductor sensors and the restrictions due to 3-side buttable read-out ASICs The use case of I13 required at least in one direction 2000 pixels with no gaps and at least 1500 pixels in the other direction The development of a detector system with such a high channel density was a major challenge DLS met such a challenge by developing Excalibur in collaboration with 2.5 Surface and interface diffraction The characteristics of Timepix chip proved to be particularly well suited for the study of interfaces, and for this reason most experiments carried out on ID01 and ID03 beamlines use a MAXIPIX detector The fast readout time of Timepix chip combined with the noiseless photoncounting detection allowed novel high data throughput scanning tech niques to be employed, which would otherwise be impossible due to prohibitive acquisition times and/or poor data quality In ID01 for instance a 2x2 MAXIPIX detector operated at 300 Hz frame rate was used to acquire 2D diffraction data at every point of a sample scan in X, Y and ω (incidence angle) positions (Chahine et al., 2014) This experi ment was totalling more than × 105 data frames acquired in about h, therefore well within a beamtime shift (8 h) In that case the time bottleneck was the data processing, not the detector performance C Ponchut et al Radiation Measurements 140 (2021) 106459 ASICs at Advacam in Finland The bump bonding technology uses solder (Pb/Sn) bumps The hybrid assemblies are then glued on the top of a molybdenum base plate and wire-bonded to a printed circuit board (carrier board) that is glued on the opposite side (rear side) of the base plate The carrier board routes the control, data, and power lines to the downstream electronic equipment (data acquisition card, trigger and power board) via two 300 way connectors Each connector reroutes the data control and power lines of a row of ASICs The base plate provides mechanical stability, accurate positioning, and removes the thermal power generated by the ASICs The three base plates are mounted on the heat sink and positioner that guarantees their accurate relative positioning Chilled water flows inside the heat sink and maintains the structure to a constant temperature Fig shows Excalibur partially assembled with the three modules that are mounted and clearly visible Further details on the Excalibur front-end design and fabrication can be found in the literature (Marchal et al., 2013) Fig The three modules of Excalibur mounted on the heat-sink The flexirigid boards on the back of the sensors are clearly visible and are to be con nected to the FEM cards the UK Science and Technology Facility Council (STFC) A future application of the Medipix3 ASICs at DLS will be an arc detector with CdTe sensors for the X-ray Pair Distribution Function beam line I15-1 The development of the arc detector is ongoing and it is due to be mounted on the beam line in 2021 3.2.3 Controls and data acquisition channel The data lines of each row of ASICs are linked through a 300 way connector and a flexi rigid adapter to an FPGA data acquisition card, known as Front End Module (FEM), developed by STFC as data acqui sition electronics for the Large Pixel Detector project of the European XFEL (Coughlan 2011) Each FEM card has Spartan3 FPGAs and a Virtex5 FPGA Two of the Spartan3 FPGAs are the interface of the board to the Medipix3 ASICs and are used to distribute and multiplex the signals The Virtex5 FPGA is the main engine of the card and it is used for sequencing, data decoding and data buffering The Virtex5 FPGA has dual PowerPC440 embedded processors used to manage the DDR2 memory controller The third Spartan3 FPGA is used as configuration controller and boot device The ASICs are driven with a clock frequency of 200 MHz and since the eight LVDS data lines of each Medipix3 ASIC of the row are read out in parallel, a set of registers of the entire detector (over million 12 bit registers) is downloaded in 492 μs The data are then sent from each FEM card to a Linux PC cluster via a 10 Gbit/s fiber optic link A Gbit/s copper Ethernet link of each card is used as control channel and to stream the diagnostic data In the first version of Excalibur the data acquisition and control ar chitecture of the detector system consisted of parallel read-out chan nels each of those controlling and receiving data from a row of ASICs A data acquisition channel was made out of a FEM card, a 10 Gbit/s fiber optic link and a Linux PC A cluster of Linux PCs was necessary to drive the detector The data were sent from the PC cluster to the remote storage and recorded in HDF5 format Each node opened the same file on the file system and wrote its portion of data in the correct location of the file so as the full image could be reconstructed (Thompson et al., 2012) After a couple of upgrades of the infrastructure (new storage file system and faster data link connecting the beam line to the storage) the parallel writer data acquisition system managed to be able to deliver a sustained frame rate of up to 275 frames per second The frame rate was only limited by some bottlenecks in the FEM cards and not by the data acquisition system However, in order to make the data acquisition system still more flexible it was finally upgraded to a different hardware and software architecture In the new data acquisition architecture the 10 Gbit/s data links from the FEM cards are now sent to a deep buffer switch instead of being connected directly to the Linux PCs The data acquisition cards that send the data through the various links send the UDP packets related to a specific image frame to the same address Consequently, the deep buffer switch routes the packets related to the same frame, but coming from different links, to a single link downstream the switch that is connected to a Linux PC Every frame is addressed by the data acquisition cards to a different Linux PC in a round robin fashion In this way each Linux PC operates on the full frame The Linux PC grabs the frame and writes it to 3.2 The Excalibur detector system 3.2.1 General detector architecture In order to meet the requirements of the I13 use case, Excalibur tiles 48 Medipix3 ASICs These ASICs are flip-chip bump-bonded to three monolithic silicon sensors The pixel pattern of each sensor is a rectan gular matrix of 2048 × 512 pixels with 16 ASICs bump-bonded in an × matrix The overall sensitive area of Excalibur is therefore 2048 × 1536 pixels In one direction there are more than 2000 pixels with no gap, as required by the use case, and in the other direction more than 1500 pixels with two gaps between the three sensors These gaps are indispensable to guarantee the clearance for the control, data, and power lines which have to be connected to the ASICs through their pe riphery Although Medipix3 supports the through silicon via (TSV) technology, at the time of the approval of the Excalibur project the TSV technology was not still mature enough to be used without adding considerable risk, time, and cost The ASICs are read out by data acquisition cards Each card reads a row of ASICs of a sensor and sends the data to a Linux PC server that in turn sends the data to the data storage The data coming from each data acquisition card therefore are a portion of the overall image (a strip of ASICs in a sensor) and the full image has to be reconstructed before it is sent to the data storage This turned out to be a major challenge A precision mechanical structure provides the accurate relative po sitions of the three sensors and acts as a heat sink to remove the heat generated by the 48 ASICs Synchronization and power electronics, di agnostics of temperature and humidity, and a chiller to provide the flow of cooling water complete the detector system 3.2.2 Sensors, hybridization, interconnections, mechanical structure The silicon sensors are fabricated at the Olen plant of Mirion Tech nologies in Belgium on inch wafers The thickness of the wafers is 500 μm The pixels are square with 55 μmpitch in the two orthogonal di rections so as to match the pitch of Medipix3 The pixels at the boundary of two ASICs are larger to provide enough clearance to accommodate two contiguous Medipix3 devices and at the same time to avoid dead areas in the sensor The edge pixels are rectangular with dimensions 55 μm × 137.5 μm while the corner pixels are square 137.5 μm × 137.5 μm The overall dimensions of the sensors including guard structures are 30.3 mm × 115.7 mm Only two sensors of such dimensions could be accommodated in a inch wafer The sensors are diced and flip-chip bump-bonded with the Medipix3 C Ponchut et al Radiation Measurements 140 (2021) 106459 and the temperature of the hybrids are monitored through appropriate diagnostic sensors located on the carrier printed circuit board The data of the diagnostic sensors are passed to the control software through the FEM cards that receive them through I2C bus channels If either the temperature or humidity exceeds preset values the detector is auto matically shut down A comprehensive trigger system enables the users to synchronize the detector with external events or to synchronize external equipment with the frames being acquired Fig shows Excalibur fully assembled with the FEM cards and the rest of the electronics 3.2.5 Systems deployed at DLS The first version of the Excalibur used the first version of Medipix3 (Tartoni et al., 2012) Because of some bugs in the first version of Medipix3 there were some restrictions to the functionality of the de tector and namely the simultaneous read-out and counting could not work The detector in I13 was then upgraded to the bug free version of the ASIC, Medipix3RXv2, and at the same time a second detector head was built and rolled out at the Hard X-Ray Nanoprobe for Complex Systems beam line I14 After that the data acquisition system ODIN-DAQ was developed and rolled out to the two detectors in I13 and I14 ODIN-DAQ enabled to exploit at best the capabilities of the detector head hardware A third Excalibur detector is under construction and will be rolled out at the Microfocus Spectroscopy beam line I18 complete with the ODIN-DAQ Finally a single module version of Excalibur was developed (16 chips, 2048 × 512 pixels) and has been rolled out in the Surface and Interface High Resolution Diffraction beam line I07 Fig shows an image of the shadow cast by a poppy taken by Excalibur during the commissioning at I13 Fig The Excalibur detector fully assembled The three modules on the left are visible through an aperture of the enclosure flushed with nitrogen They are read out by the FEM cards, in the middle of the frame, that in turn stream the data to the Linux PC cluster the data storage independently on the other servers by using the HDF5 Virtual Data Set libraries The HDF5 VDS present the data to the final user as if they were a single data set This architecture is called ODINDAQ (Yendell et al., 2018) It is possible to operate Excalibur in burst mode where a number of image frames is acquired at a frame rate up to 2000 frames per second, stored in the local RAM on the FEM cards, and then streamed to disk after the acquisition 3.3 Merlin and Lancelot detector systems 3.2.4 Ancillary equipment The detector is assembled in a holder that provides mechanical sta bility and hosts the sensors, the FEM cards, power supply electronics, high voltage bias supply, and the ancillary trigger electronics The sensors are located in an enclosure flushed with dry air or nitrogen in order to prevent excessive moisture The humidity level of the enclosure 3.3.1 Merlin Merlin is a readout system for up to four Medipix3 ASICs based on the National Instruments (NI) PXIe chassis and embedded controller and an FPGA card (Horswell 2011; Plackett 2013) The FPGA card is connected to the detector head (up to ASICs mounted on a chip board and an appropriate enclosure) through an adapter card that is plugged into the Fig Shadow of Ni fluorescence X-rays (7.5 keV) cast by a poppy The shadow of the tape that keeps the poppy in place is also visible in the upper part of the image Definition of the details, contrast and dynamic range of the image, and noise level are well in line with the performance expected by photon counting detectors The image is flat field corrected In the bottom left corner of the image there is a non-working super-column because the related wire was broken C Ponchut et al Radiation Measurements 140 (2021) 106459 Fig Merlin ladder (4 × 1) detector head A small amount of electronics is located on the detector head This makes the head very light and compact Fig Photoelectric X-ray absorption efficiencies for typical thicknesses of different sensor materials (Berger 2010) FPGA card and through a cable up to 10 m long The adapter card was developed at DLS as well as the Merlin software and firmware written in the LabVIEW graphical programming language The ASICs on the de tector head are powered up through the connecting cable and the volt ages are supplied by the PXIe chassis through an extra module developed at DLS and plugged into the PXIe chassis The sensor is also biased through the connecting cable and the high voltage bias supply is pro vided by the aforementioned extra module that is controlled by the Merlin software through the PXIe bus Merlin started as a pilot project to gain the knowledge of the Medi pix3 ASIC necessary to implement the Excalibur detector However it was soon recognized that Merlin could be used as a detector system in synchrotron experiments where the space and weight are at premium Since the bulky read-out electronics (PXIe chassis) can be located rela tively remotely with respect to the detector head it is easy to place the latter close to the samples and mount it on equipment with weight constraints such as the diffractometer arm The Merlin detector heads come as single ASIC, quads (2 × ASICs), and ladders (4 × ASICs shown in Fig 7) Merlin detectors have been rolled out at DLS in some cases augmenting the capability of certain beam lines It is the case of I16 where diffraction imaging experiments can now run A TCP/IP interface was added to enable the integration into the standard control and data acquisition system of the beam lines of DLS The TCP/IP interface enables also the system to be controlled over network con nections from simple clients written in Matlab, LabVIEW, Python, or other languages The Merlin system has been commercialized by Quantum Detectors for use at synchrotron experiments and more recently for electron mi croscopy experiments To date 29 systems have been sold worldwide to organizations other than DLS Fig LAMBDA detector module A 42 mm × 28 mm Cr-compensated GaAs sensor bonded to Medipix3 chips is mounted on the left side of the detec tor head Medipix and high-Z sensor materials at DESY 3.3.2 Lancelot The Medipix3 ASIC found use at DLS also as a beam diagnostic In collaboration with the University of Manchester, Lancelot was devel oped, which is a pinhole camera imaging the radiation scattered by a foil The sensor is a silicon sensor bump-bonded to a single Medipix3 ASIC and the entire Lancelot camera is contained in a single enclosure mounted on the beam line (Kachatkou et al., 2014; Rico-Alvarez et al., 2014; Garcia-Nathan et al., 2017) Lancelot is controlled with the same TCP/IP protocol that is used by Merlin It operates the ASIC in 6-bit mode and can run at a maximum speed of 245 frames per second Lancelot has been deployed on two beam lines and it is used effectively as beam position and beam profile monitor (Chagani et al., 2017) The system can measure displacements of the beam of μm 4.1 Hard X-ray experiments at synchrotrons The high speed, dynamic range and sensitivity of photon-counting detectors such as Medipix have made them the technology of choice in many X-ray experiments at synchrotrons, particularly X-ray scattering Many of these experiments study small samples which not strongly absorb X-rays, such as protein crystals So, these experiments can be performed at X-ray energies where standard silicon sensors provide high detection efficiency; for example, 12 keV photons provide a sufficiently short wavelength to resolve atoms (0.1 nm), and have an absorption length in silicon of 215 μm However, some experiments require higher X-ray energies; for example, experiments studying the internal structure C Ponchut et al Radiation Measurements 140 (2021) 106459 of more absorbent samples (e.g buried interfaces in fuel cells), and experiments taking place inside sample environments (e.g highpressure experiments or in-situ industrial processes) For hard X-ray experiments, it is advantageous to use “high-Z” (high atomic number) semiconductors such as GaAs, CdTe or Ge, which have much higher efficiency as shown in Fig In particular, investigating changes in samples on short timescales requires this combination of high quantum efficiency, low noise and high readout speed offered by photon counting detectors The PETRA-III synchrotron at DESY has a high storage ring energy of GeV, and as a result most beamlines can reach X-ray energies of 30 keV, with specialised hard X-ray beamlines going up to 250 keV So, development of Medipix detectors at DESY has placed an emphasis on building systems using high-Z sensors The LAMBDA readout system (Pennicard 2013) built by DESY is designed to work effectively with a range of high-Z materials, and LAMBDA systems have been built using CdTe and GaAs as sensors In particular, large multi-module systems with megapixel area have been built using GaAs Fig 10 Flatfield X-ray response of a 1000 μm-thick CdTe sensor with 768 × 512 pixels of 55 μm size, and ohmic contacts harmed by high temperatures, so bump-bonding of CdTe is typically done with softer, low-melting-point solders such as In alloys The re sistivity of CdTe is sufficiently high that CdTe sensors with ohmic con tacts (Pt metal) can be used with Medipix3 readout around room temperature without excessive leakage current However, it is also possible to use other metals to form a Schottky diode Fig 10 below shows the flat-field X-ray response of a CdTe sensor with 768 × 512 pixels of 55 μm pixel size (42 mm × 28 mm area), bonded to Medipix3 readout chips and read out by LAMBDA The sensor thickness was 1000 μm, and ohmic contacts were used The X-ray tube had a Mo anode and 35 kV voltage The response is not perfectly uni form, with a visible network of lines present; these are likely due to dislocations in the material distorting the electric field However, the yield of working pixels is high The main limitation of CdTe is that its response to X-rays can change with time and X-ray illumination Firstly, temporary changes can occur due to polarization; trapping of charge carriers can lead to a build up of space charge in the material, changing the electric field and altering the X-ray response In particular, high illumination can cause a collapse of the electric field, leading to low response For example (Ruat 2014), reports distortions in the response of a CdTe detector after several hours of illumination with a relatively modest illumination of 7.105 pho tons/mm2/s Permanent damage effects after much higher levels of irradiation are also reported These instabilities are also influenced by the choice of ohmic or Schottky contacts (Astromskas 2016) The impact of these instabilities depends strongly on the experiment being per formed For example, X-ray diffraction from a single crystal will produce intense diffraction spots (Bragg peaks) whose intensities must be measured to determine the crystal structure So, polarization can pre vent this information from being accurately measured However, be tween the Bragg peaks there are much weaker diffraction signals containing information about crystal shape and defects, and the high quantum and photon counting capability of this sensor can improve measurement of these signals 4.2 LAMBDA readout system LAMBDA (Large Area Medipix-Based Detector Array) is a modular readout system for the Medipix3 chip The LAMBDA module, shown in Fig 9, is designed to have a large area (6 by Medipix3 chips), highspeed readout at 2000 frames per second, and to be compatible with different high-Z sensors Multiple LAMBDA modules can then be tiled together to build larger systems The detector head of a LAMBDA module consists of one or more sensors bonded to Medipix3 chips, glued to a ceramic circuit board, which in turn is glued to a copper mechanical block The detector head was designed to mount up to 12 Medipix3 chips in a 6-by-2 arrangement, giving an array of 1536 by 512 pixels of 55 μm size and an area of 85 mm by 28 mm High-Z materials are typically only available in wafers of inch size, which means that the largest Medipix3-compatible sensor design available is a 3-by-2-chip layout of 42 mm × 28 mm So, this detector head design makes it possible to mount either two high-Z sensors or one large Si sensor The Medipix3 chips are then wire bonded to the ceramic circuit board A ceramic with many thermal vias was used, rather than an FR4 board, in order to ensure good thermal conduction from the Medipix3 chips to the cooling block, and to mini mise thermal stresses in the system by matching the thermal expansion coefficient of most semiconductors The detector head is then connected to a “signal distribution” board, which provides powering to the detector head and routes the signals from the Medipix3 chips to a readout board This readout board has a Virtex5 FPGA which controls and reads out the detector head The board incorporates two 10 Gigabit Ethernet optical links, which makes it possible to read out the Medipix3 chips at their full speed of 2000 frames per second (12 bits per image depth) and continuously send the data to a control PC over the links The module is designed so that the readout electronics fit behind the detector head, allowing multiple modules to be tiled together to cover a larger area In multi-module systems, all the detector heads are attached to a single large water-cooled copper block, and an additional board is used to fan out control signals from a “master” module to all the others so that the modules take images simultaneously 4.3.2 Gallium arsenide Gallium arsenide is widely-used for applications such as optoelec tronics, and is readily available in large wafers, which makes it a potentially appealing option for X-ray detection However, most GaAs does not provide the required combination of long carrier lifetime, high resistivity and large active layer thickness In particular, high-resistivity GaAs typically has many electron traps, leading to a low carrier lifetime In recent years, most GaAs X-ray detectors used with Medipix have been made from Chromium-compensated GaAs (Budnisky 2014) This material is produced by taking n-doped GaAs material with a reasonable electron lifetime and doping it with Chromium to compensate the excess electrons The resulting material has sufficiently high resistivity that it 4.3 High-Z sensor materials 4.3.1 Cadmium Telluride Cadmium Telluride is a well-established high-Z sensor material, with a higher quantum efficiency at extreme X-ray energies than Ge or GaAs It is commercially available in inch wafers, with carrier lifetimes in the order of microseconds (Takahashi 2001), which means that signal loss from charge trapping has little effect in Medipix CdTe detectors It is physically more fragile than Ge and GaAs, and its performance may be C Ponchut et al Radiation Measurements 140 (2021) 106459 Fig 11 Flatfield X-ray response of a LAMBDA module constructed from two 500 μm-thick GaAs sensors, each with 764 × 512 pixels of 55 μm size and ohmic contacts megapixel GaAs detectors (Pennicard 2018) As shown in Fig 12, these are constructed from modules, each with GaAs sensors tiles of 42 mm × 28 mm, giving a total pixel count of 1528 × 1536 pixels Each module incorporates two 10 Gigabit Ethernet readout links as described above, allowing images to be continuously taken at up to 2000 frames per second The data is received by a pair of server PCs During exper iments at the maximum frame rate, it is not possible to process the im ages and save them to disk in real time, so the data is saved temporarily in RAM instead; this means this frame rate can be sustained for approximately 50s, until the RAM is full This is sufficient for most high-speed experiments These systems have been used primarily at PETRA-III beamline P02.2, studying materials under extreme pressure in diamond anvil cells In these experiments, a sample is compressed between a pair of diamonds, which can exert pressures comparable to the earth’s core (hundreds of gigapascals) The structure of the sample can then be probed by X-ray diffraction, typically using an X-ray energy of 42 keV at P02.2 The high speed and single photon counting capability of the LAMBDA system make it possible to record useable diffraction patterns at 2000 fps, compared to the 10 fps rate normally achieved by typical flat-panel detectors As a result, experiments studying rapid changes in pressure can observe changes on much shorter timescales than before (Marquardt 2018) Fig 12 A 2.3-megapixel GaAs LAMBDA detector, constructed from six GaAs tiles each of 42 mm × 28 mm can be used as a photoconductive sensor at room temperature by depositing Ohmic contacts Currently, the Cr-compensation process can be applied to wafers of up to inches in size Fig 11 below shows the flat-field X-ray response of a Lambda module built from two Cr-compensated GaAs sensors mounted in a single de tector head Like the CdTe sensor above, each individual sensor has 764 × 512 pixels of 55 μm pixel size (42 mm × 28 mm area) and is bonded to Medipix3 readout chips The sensor thickness is 500 μm, and there is a 10-pixel-wide gap between the two sensors The GaAs has greater pixelto-pixel variation in response than the CdTe, with a distinctive cellular structure Microbeam scans on another sensor show that this variation is not due to signal loss, but rather to variations in effective pixel size caused by nonuniformities in the electric field (Ponchut 2017) These field nonuniformities are due to variations in dopant and defect con centrations which develop during growth While the raw images from GaAs show greater nonuniformity than CdTe, the behaviour of the material is more stable with time and irra diation (Hamann 2015) and the image uniformity can be greatly improved with flat-field correction As a result, the material can be effectively used for X-ray scattering experiments, including some dy namic experiments where the instability of CdTe would cause problems However, to perform flatfield correction it is necessary to illuminate the detector with uniform (or at least smoothly varying) X-rays of the cor rect energy, with high counting statistics This can often be inconvenient when performing experiments at a beamline Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper References Astromskas, V., Gimenez, E., Lohstroh, A., Tartoni, N., 2016 IEEE Trans Nucl Sci 63, 252 Ballabriga, R., Alozy, J., Blaj, G., et al., 2013 J Inst 8, C02016 Berger, M., Hubbell, J., Seltzer, S., et al., 2010 NIST XCOM Photon Cross Sections Database available http://www.nist.gov/physlab/data/xcom/index.cfm Bravin, A., Fiedler, S., Coan, P., et al., 2004 IEEE Nuclear Science Symposium, Satellite Workshop on Synchrotron Radiation Detectors, Rome, Italy Budnitsky, D., Tyazhev, A., Novikov, V., et al., 2014 J Inst 9, C07011 Caronna, C., Chushkin, Y., Madsen, A., et al., 2008 Phys Rev Lett 100, 055702 Chagani, H., et al., 2017 J Inst 12, C12044 Chahine, G.A., Richard, M.-I., Homs-Regojo, R.A., et al., 2014 J Appl Crystallogr 47, 762 Chushkin, Y., Zontone, F., Lima, E., et al., 2014 J Synchrotron Radiat 21, 594–599 Coughlan, J., Cook, S., Day, C., Halsall, R., Taghavi, S., 2011 J Inst 6, C12057 Diemoz, P.C., Bravin, A., Sztr´ okay-Gaul, A., et al., 2016 Phys Med Biol 61, 8750–8761 Favre-Nicolin, V., Chushkin, Y., Cloetens, P., et al., 2017 Synchrotron Radiat News 30 (5), 13–18 Fluerasu, A., Kwasniewski, P., Caronna, C., et al., 2010 New J Phys 12, 035023 Garcia Nathan, T.B., et al., 2017 J Inst 12, C10011, 2017 Gruebel, G., Zontone, F., 2004 J Alloy Compd 362, 3–11 Hamann, E., et al., 2015 J Inst 10, C01047 Horswell, I., et al., 2011 J Inst 6, C01028 Huotari, S., Albergamo, F., Vank´ o, G., et al., 2006 Rev Sci Instr 77, 0503102 Huotari, S., Sahle, Ch J., Henriquet, Ch, et al., 2017 J Synchrotron Radiat 24, 521–530 Kachatkou, A., Marchal, J., van Silfhout, R., 2014 J Synchrotron Radiat 21, 333–339 4.4 Large area high-Z systems The largest high-Z LAMBDA systems built to date by DESY are 2.3 C Ponchut et al Radiation Measurements 140 (2021) 106459 Labiche, J.C., Mathon, O., Pascarelli, S., et al., 2007 Rev Sci Instr 78, 091301-091308 Leheny, R.L., 2012 Curr Opin Colloid Interface Sci 17, 3–12 Llopart, X., Ballabriga, R., Campbell, M., et al., 2007 Nucl Instrum Methods A 581, 485–494 Llopart, X., Campbell, M., Dinapoli, R., et al., 2002 IEEE Trans Nucl Sci 49 (5), 2279–2283 Marchal, J., et al., 2013 J Phys Conf 425, 1–5 Marquardt, H., Buchen, J., Mendez, A.S.J., et al., 2018 Geophys Res Lett 45, 6862 Moretti-Sala, M., Martel, K., Henriquet, C., et al., 2018 J Synchrotron Radiat 25, 1–12 Orsi, D., Cristofolini, L., Baldi, G., et al., 2012 Phys Rev Lett 108, 105701 Pennicard, D., Lange, S., Smoljanin, S., et al., 2013 J Phys Conf Ser 425, 062010 Pennicard, D., Smoljanin, S., Pithan, F., et al., 2018 J Inst 13, C01026 Plackett, R., et al., 2013 J Inst 8, C01038 Ponchut, C., Cl´ ement, J., Rigal, J.M., et al., 2007 Nucl Instrum Methods A 576, 109–112 Ponchut, C., Cotte, M., Lozinskaya, A., et al., 2017 J Inst 12, C12023 Ponchut, C., Rigal, J.M., Cl´ement, J., et al., 2011 J Inst 6, C01069 Rico Alvarez, O., et al., 2014 J Inst 9, C12063, 2014 Ruat, M., Ponchut, C., 2012 IEEE Trans Nucl Sci 59 (5), 2392–2401 Ruat, M., Ponchut, C., 2014 J Inst 9, C04030 Takahashi, T., Watanabe, S., 2001 IEEE Trans Nucl Sci 48, 950 Tartoni, N., et al., 2012 IEEE NSS/MIC Conference Record, pp 530–533 Thompson, J.A., et al., 2012 Proc ICALEPCS’11 894–897 Yendell, G., et al., 2018 Proc ICALEPCS’17 966–969 ... in combination with finely focused monochromatic synchrotron X-rays, were also used at ESRF and other synchrotrons as tools to explore the properties of semiconductor pixel sensor materials:... mounted on the heat sink and positioner that guarantees their accurate relative positioning Chilled water flows inside the heat sink and maintains the structure to a constant temperature Fig shows... Timepix1 chip spatial reso lution of 55 μm was ideally suited for breast imaging The high signal-to-noise ratio even at low X-ray flux provided by Timepix1 was also a key feature for dose reduction