Megapixels @ Megahertz -- The AGIPD High-Speed Cameras for the European XFEL
Ulrich Trunk, Aschkan Allahgholi, Julian Becker, Annette Delfs, Roberto Dinapoli, Peter Göttlicher, Heinz Graafsma, Dominic Greiffenberg, Helmut Hirsemann, Stefanie Jack, Alexander Klyuev, Hans Krüger, Manuela Kuhn, Torsten Laurus, Alessandro Marras, Davide Mezza, Aldo Mozzanica, Jennifer Poehlsen, Ofir Shefer Shalev, Igor Sheviakov, Bernd Schmitt, Jörn Schwandt, Xintian Shi, Sergej Smoljanin, Jiaguo Zhang, Manfred Zimmer
MMegapixels @ Megahertz The AGIPD High-Speed Cameras for the EuropeanXFEL
Aschkan Allahgholi a , Julian Becker a , Annette Delfs a , Roberto Dinapoli b , Peter Gttlicher a , Heinz Graafsma a,e ,Dominic Grei ff enberg b , Helmut Hirsemann a , Stefanie Jack a , Alexander Klyuev a , Hans Krger d , Manuela Kuhn a ,Torsten Laurus a , Alessandro Marras a , Davide Mezza b , Aldo Mozzanica b , Jennifer Poehlsen a , Ofir Shefer Shalev a ,Igor Sheviakov a , Bernd Schmitt b , Jrn Schwandt c , Xintian Shi b , Sergej Smoljanin a , Ulrich Trunk a, ∗ , Jiaguo Zhang b ,Manfred Zimmer a a Deutsches Elektronensynchrotron - DESY, Hamburg, Germany b Paul Scherrer Institut - PSI, Villigen, Switzerland c Universitt Hamburg, Hamburg, Germany d Universitt Bonn, Bonn, Germany e Mid-Sweden University, Sundsvall, Sweden
Abstract
The European XFEL is an extremely brilliant Free Electron Laser Source with a very demanding pulse structure:trains of 2700 X-Ray pulses are repeated at 10 Hz. The pulses inside the train are spaced by 220 ns and each onecontains up to 10 photons of 12 . ≤
100 fs in length. AGIPD, the Adaptive Gain Integrating PixelDetector, is a hybrid pixel detector developed by DESY, PSI, and the Universities of Bonn and Hamburg to cope withthese properties.It is a fast, low noise integrating detector, with single photon sensitivity (for E γ (cid:39) photons at 12 . . ×
64 pixels of200 µ m × µ m. Control of the ASIC’s image acquisition and analogue readout is via a command based interface.FPGA based electronic boards, controlling ASIC operation, image digitisation and 10 GE data transmission interfaceAGIPD detectors to DAQ and control systems.An AGIPD 1 Mpixel detector has been installed at the SPB experimental station in August 2017, while a secondone is currently commissioned for the MID endstation. A larger (4 Mpixel) AGIPD detector and one to employ Hi-Zsensor material to e ffi ciently register photons up to E γ ≈
25 keV are currently under construction.
Keywords:
AGIPD, X-Ray Detector, Photon Science, European XFEL, Free Electron Laser, Hybrid Pixel Detector ∗ Corresponding author
Email address: [email protected] (Ulrich Trunk) Single particles, Clusters and Biomolecules Materials Imaging and Dynamics
1. The European XFEL
The European X-Ray Free Electron Laser (XFEL) [1, 2]in Hamburg is currently the most brilliant X-Ray source(fig. 1) in the world. It provides extremely focused, fullycoherent X-Ray pulses. Trains of 2700 of these pulses arerepeated at 10 Hz. The pulses inside the train are spaced
Preprint submitted to Nuclear Instrumentation and Methods A August 7, 2019 a r X i v : . [ phy s i c s . i n s - d e t ] A ug y 220 ns and contain up to 10 × . photonseach, while being ≤
100 fs in length (fig. 2). The highintensity per pulse will allow recording di ff raction pat-terns of single molecules or small crystals in a single shot.As a consequence 2D detectors have to cope with a largedynamic range, requiring single photon sensitivity e.g.for single molecule imaging, and registering more than10 photons / pixel in the same image for the case of liquidscattering or intense Bragg Spots . AGIPD is one of threedetectors developed to cope with the timing requirementsof the European XFEL[3]. The other two are the LargePixel Detector (LPD) [4] with 500 µ m × µ m pixel sizeand 1 Mpixel installed, and DSSC [5], a 1 Mpixel detectorwith 200 µ m × µ m hexagonal pixels, based on silicondrift diodes or Depfet sensors with Signal Compression(hence the acronym), currently under development. WhileLPD is targeted at the same energy range as AGIPD, it dif-fers from it by a lower spatial and energy resolution, butcovers a much bigger area and operates in ambient. DSSClike AGIPD operates in vacuum, but can detect lower pho-ton energies - on the expense of a nonlinear response anda lower frame rate at low energies.Future upgrades of the European XFEL planned for the2 nd half of the 2020ies include continuous wave (CW) op-eration at a pulse rate of ≈
100 kHz and a so-called longpulse mode at (cid:47)
200 kHz with 500 ms bursts repeating at1 Hz[6]. At these rates moving or exchanging solid sam-ples during a burst becomes feasible, which would reducethe radiation damage of samples considerably and thusenable european X FEL for new classes of experimentsand materials. To exploit these modes new detectors witha di ff erent readout architecture are needed, since AGIPDhardware can only operate up to ≈
16 kHz CW framerate. . Droplets and ice crystals are frequently encountered in gas and liq-uid jet sample delivery setups The firmware for the readout system would have to send commandsto the ASIC in a di ff erent order Figure 1: Brilliance of FELs and Synchrotron sources.Figure 2: Time structure of the European XFEL source. . The AGIPD Detector AGIPD [7, 8, 9, 10] is a hybrid pixel detector, developedby DESY, the Paul Scherrer Institute and the Universitiesof Bonn and Hamburg to meet the requirements for theuse at the European XFEL. It consists of 500 µ m thick Sil-icon sensors, manufactured by SINTEF, Hamamatsu andADCVACAM, to provide a high e ffi ciency (QE ≥ . ffi -ciently sheild the ASICs underneath from radiation, thesensors themselves will collect a substantial amount of ra-diation dose during their expected lifetime. Special designmeasures have been taken to mitigate radiation damageand its e ff ects [11]. Each sensor features 512 ×
128 pixels,which are 200 µ m × µ m in size and pitch as a goodcompromise. 8 × . Thehigh photon flux together with single photon sensitivityrequire the detector to operate in vacuum, in order to pre-vent the intense beam from interacting with ambient air orexit windows, which would cause a huge background. Inturn a vacuum vessel is an integral part of the detectors.The 1 Mpixel detectors at the SPB and MID experimentalstations have the 4 × . These can be arranged to form a holefor the direct beam to prevent it from hitting detector com-ponents and inflicting damage to the system. The intensebeam and the experiments envisioned also require a hugedynamic range, which can reach 10 photons / pixel / imageat 12 . The minimum distance of the individual sensor tiles are 0 . . The AGIPD 1 Mpixel detector for the HIBEF endstation features adi ff erent layout
3. The AGIPD readout ASIC
The core functionality of the AGIPD detector is imple-mented in the readout ASIC [13, 14, 15]. It is manu-factured in IBM / Global Foundries cmrf8sf (130 nm) tech-nology and contains 64 ×
64 pixels of 200 µ m × µ m.The circuit in each pixel (fig. 3) contains a charge sensi-tive preamplifier based on an inverter core with threefoldswitchable gain. This gain switching is implemented byadding capacitors of 3 pF and 10 pF to the initial preampli-fier feedback of 60 fF. A discriminator is connected to thepreamplifier output and triggers an adaptive gain selec-tion, whenever this output exceeds a selected threshold.A correlated double sampling (CDS) stage removes re-set and attenuates low-frequency noise components fromthe preamplifier output [16], such that the detection ofsingle photons is feasible. The output of the CDS, aswell as the selected gain is sampled in a capacitor basedanalogue memory for 352 images, which occupies about80 % of a pixel’s area. It is based on n-FET in n-well ca-pacitors and dual PMOS switches for radiation toleranceand low leakage, since the analogue signal must not de-teriorate during several 10 ms of storage time. For read-out each pixel features a 2 nd charge sensitive bu ff er. Lastbut not least each pixel features two sources of electricalstimuli: A constant curren source and a pulsed capacitor,which can be connected to the preamplifier input. A com-mand based interface and control circuit provides randomaccess to the memory and controls the row-wise readoutof the data via multiplexers to four di ff erential analogueports. The random access scheme allows overwriting im-age data within a bunch train, which is compliant with theEuropean XFEL’s vetoing schema to maximise e ffi ciency.The data of individual ’amplitude’ and ’gain’ frames willbe combined to single images by the readout electronics .The power consumption of an ASIC is typically around1 . . ffi cient radiation toleranceof the ASIC, individual components and building blockshave been tested for doses up to several 10 kGy, e.g. in[12]. down to ≈ ≥ σ The current firmware of the AGIPD 1 Mpixel detectors at SPB andMID reads 300 separate ’amplitude’ and ’gain’ frames, which are com-bined o ffl ine. igure 3: Circuit schematic of the AGIPD 1.1 ASIC. All AGIPD systems currently in use are based on theAGIPD 1.1 version of the ASIC, which has been thor-oughly characterised electrically and with a sensor. Sincethe medium and low gain settings are not accessible withlab X-Ray sources, an IR laser has been used for char-acterisation. This way an ENC of ≈ − was mea-sured , as fig. 4 shows. This value corresponds to a signalto noise ratio > σ for a single 12 . × e − or10 photons of 12 . .
44 % upto 5 × photons of 12 . A breif description of earlier versions and prototypes of the ASICcan be found in [12]. Equivalent Noise Charge Sacrificing dynamic range by selecting the high gain of the CDS,ENC ≈
240 e − can be reached.
4. AGIPD Calibration Procedure
Calibrating a detector system with multiple gains is a non-trivial task, since the medium and low gain settings ofthe preamplifier are not easily accessible with radiationsources. While the intensity of the direct beams at FELsand synchrotrons might be su ffi cient to reach these, thestatistics required for the necessary precision and the as-sociated time and radiation load, which is expected tovastly exceed that of experiments, render this approachimpractical. Instead a hybrid approach based on fluores-cence photons and electrical stimuli provided by the in- Figure 4: Noise performance of AGIPD [10]. An IR laser was used to ac-cess the di ff erent gains. Fluctuations of this stimulus add an amplitude-dependent contribution to the measured ENC. • Calibration of the preamplifiers high gain with fluo-rescence photons from Copper or Molybdenum tar-gets. • Scan of the preamplifiers integration time (from thenominal 150 ns to several µ s) using the in-pixelcurrent source to determine the sensitivity of themedium and low gain settings relative to the highgain.A big advantage of this approach is the independencefrom the absolute value of the current, i.e. of variationsin the ASIC’s manufacturing process or sensor leakagecurrents, whereas the indirect nature of the procedure andthe long lever arm are mitigated by the Poissonian natureof photons, which relaxes requirements for higher inten-sities.Using the DAQ system at the European XFEL, data fora full calibration can be acquired within less than 2 h, i.e.this can be conveniently performed before and after userexperiments. From this data the 3 o ff sets ( baseline ) andgains have to be calculated. In addition gain and o ff setalso depend on the size of the memory’s capacitors. Thisleads in total to (3 gains + ff sets) ×
352 storage cells × = . · constants. Calculating these Figure 5: Transfer characteristics of AGIPD, including gain switching[15]. The solid lines are linear fits to the corresponding gain setting. Figure 6: Spectrum of 7keV photons measured by AGIPD (data fromPetra III). constants requires ≤ Maxwell com-puter cluster [19], and no further reduction of constants bydisentanglement of the contributions from frontend andmemory are performed. The calculation of the two thresh-old levels required to re-digitise the the gain informationis also included in the calibration process. Remaining un-certainties in determining the gain even after calibrationled to the further improved AGIPD 1.2 readout ASIC.
5. AGIPD Detector Systems at SPB& MID
The image planes of all AGIPD cameras are composedfrom
Frontend Modules (FEM). These are constructedfrom Silicon sensor tiles, to which 8 × / Ag or Sn / Pb bumps.The sensor tiles are p-on-n type with implant sizes andmetal overhangs specially tailored to avoid dead pockets ,where charges would be trapped and contribute to radi-ation damage, as well as to facilitate bias voltages up to900 V [11]. The latter is required to overcome so called plasma e ff ects [18], when the impinging radiation e.g. ona Bragg spot creates charge carrier densities able to ef-fectively shield of the drift field. For the first user ex-periments bias voltages of 300 V or 500 V were applied.5urther parts of a frontend module are an LTCC carrierboard, to which the sensor assemblies are glued and wirebonded and the copper interposer, to which the LTCCsare bolted. For the AGIPD 1 Mpixel detectors for SPBand MID four of these frontend modules are attached toa copper cooling block to form a quadrant. This way thetemperature of the modules can be lowered to ≈ ◦ C. Thequadrants of the 1 Mpixel systems at the SPB and MIDexperimental stations are attached to a wedge-shaped ar-rangement of linear stages and mounted inside a vacuumvessel, while the image plane sticks out of it. Connectedto the experimental chamber vacuum levels down to 10 − have been reached during user operation . The movablearrangement of the quadrants permits the formation of anarbitrary hole or slot for the direct beam to pass, whilethe wedge shaped stages allow the driving motors to bemounted outside the vessel – on the expense of loosingthe orthogonality of the translation axes.The 64 analogue signals of one frontend module arebrought outside the detector’s vacuum vessel via the vac-uum board , a PCB with a flexible section to compen-sate for the motion of the quadrants, and a vacuum flangeformed by another PCB serving as the vacuum barrier.Outside the vacuum vessel PCBs with receiver ampli-fiers and ADCs provide digitisation of the data with 14 bitquantisation. This data is then collected by an FPGA and sent on via an 10 GB optical ethernet link per FEMto the data acquisition system of the European XFEL,which combines the data of individual modules to fullframes and passes them on to storage. A rendering ofthe 1 Mpixel detectors for the SPB and MID endstationsis shown in fig. 7. The power consuption of such a systemis about 1 . Low Temperature Co-fired Ceramics AGIPD can also operated at ambient pressure, but coarse vacuumlevels are – depending on sensor bias – forbidden by Paschen’s law. Printed Circuit Board Field-Programmable Gate Array Figure 7: Rendering of the AGIPD 1 Mpixel detectors for the SPB andMID endstations. To show the locations of the components, parts of thevessel and the external housing are cut out.Figure 8: The AGIPD 1 Mpixel camera at the MID experimental station.The photograph shows the Detector at the end of the 8 m long spectrom-eter arm, prior to the installation of the flight tube. . First User Experiments withAGIPD Due to its unique sampling rate, radiation hardness anddynamic range characteristics, AGIPD has been em-ployed for experiments already in its prototyping stageand was e.g. used for an experiment to determine thecoherence of the beam at PETRA III [20]. The AGIPD1 Mpixel system at the SPB station of the European XFELwas successfully demonstrated during the facility’s inau-guration and has seen continuous use since the very firstuser experiments in September 2017. In that context al-ready the very first user experiment facilitated MHz serialfemtosecond crystallography (at 1 . β -lactamase for thefirst time [21]. The figures 9 and 10 show a di ff ractionpattern from Lysozyme also recorded during that experi-ment and powder di ff raction rings from Lithium-Titanatewhich are used for the spatial calibration, including thetilt of individual sensor tiles with respect to the detectorplane.
7. New AGIPD Detector Systems forSFX and HIBEF
For the SFX instrument at the European XFEL a4 Mpixel AGIPD system is currently under construction.The image plane of this detector will consist of 56 sen-sors, arranged in a 4 ×
14 pattern, formed by two halveswith a vertical gap. These halves can be translated inthe horizontal plane. The range lateral with respect tothe beam is 30 mm, while longitudinally the image planecan be moved by 400 mm upstream – through a giant gatevalve of 800 mm diameter into the experimental chamber.Unlike the 1 Mpixel systems, this system will use doubleFEMs, consisting of two LTCCs bolted to a cooled inter-poser. To cater the side-by-side arrangement of the sen-sors, which is incompatible with the arrangement of thereadout electronics of the existing AGIPD 1 Mpixel sys-tems at SPB and MID, and to implement lessons learned Serial Femtosecond Crystallography from these systems, a new readout board was designed.
During comissioning of the AGIPD 1 Mpixel system atSPB the vast number of connectors imposed a reliabil-ity problem, which required the external housings to beopened and the boards re-seated for proper contact. Thesame held, albeit to a lesser extent, true for the power sup-ply cables, which also make moving the detector a majorendeavour. A further shortcoming was the central genera-tion of the ASIC commands by two
Master FPGA boards,which led to signal integrity issues and a limited tunabil-ity of the ADC sampling phase. In turn the basic idea ofthe new readout board was the elimination of connectorsand power supplies and a total modularity of the system.Thus the board houses all components to operate a (sin-gle) FEM as a stand-alone detector (see fig. 11). There-fore, it implements the following components: • A HV DC / DC module to generate the sensor bias. • A connector for a mezzanine board to implementbackside pulsing of the sensor as an additional meansof calibration [22]. • • Cascaded switching and linear regulators to powerthe FEM’s ASICs • A Xilinx ™ Zynq ™ SoC with DDR3 memory tocontrol the ASICs, process the ADC data and send iton via a 10 GE link. • A microcontroller to implement secondary tasks likepower sequencing, configuration and monitoring. • A multimedia serialiser and deserialiser to interfacewith the European XFEL’s clock and control (C&C)and interlock systems. • A single power connector and DC / DC converters topower the board. • A Samtec ™ FireFly ™ [23] 4-channel opticaltransceiver for communications. • Two Lemo ™ sockets for triggered stand-alone oper-ation of an FEM without C&C signals.The four channels of the FireFly transceiver are allocatedto the 10 GE data transmission, 1 GE control interface, theserialised signals of the C&C and interlock systems and System-on-chip igure 9: Di ff raction pattern from one of the lysozyme micro crystals injected with a liquid jet. It is part of a burst , i.e. an image series recorded at1 . ff raction spots and the water ring are nicely visible. The vertical stripes are masked double-sizedpixels between readout ASICs. igure 10: Powder di ff raction pattern of Lithium Titanate (Li TiO ). It shows the uncorrected amplitude data recorded by the 1 Mpixel AGIPD atSPB. Gain switching is visible as lower amplitudes (blue and white colour in between red lines) in the most intense parts of the di ff raction ringsand some of the double-sized pixels (vertical stripes) between individual readout ASICs. an ethernet channel for debugging, which is not used dur-ing normal operation. Additional FireFly transceivers on the inside and outside of a vacuum flange re-group thesesignals such, that complete 4-channel ribbons with MTP9 igure 11: Photograph of a readout board used in the AGIPD systemsfor the SFX and HIBEF endstations. connectors interface to the 10 GE DAQ system, the 1 GEcontrol system and the receiver board , implementing theserialiser / deserialiser interface to the European XFEL’sC&C and interlock systems.Since the readout boards are mounted within the vac-uum vessel, a novell cooling concept is explored: Pairs ofboards are attached to a liquid cooled heat exchanger bymeans of a ’Gap Filler’ thermal conductive plastic.For the HIBEF endstation a 1 Mpixel system, based Helmholtz International Beamline for Extreme Fields on the new readout board is under construction. The im-age plane of this detector will be a fixed stack of eightdouble modules, with a small gap between the top andbottom four, to allow the direct beam to pass. For thisdetector’s image plane only a translation along the beamaxis, i.e. to adjust the distance of the detector from thesample, is foreseen. Translations in the other 2 dimen-sions are accomplished by the detector bench , i.e. bymoving the detector’s vacuum chamber. The biggest chal-lenges for the HIBEF AGIPD detector system are thepresence of pulsed high magnetic and electrostatic fieldsand photon energies ≥
25 keV. At such energies the sil-icon sensor of AGIPD becomes transparent , i.e. ine ffi -cient and the usage of a high-Z sensor material like GaAsmandatory.
8. ecAGIPD - An electron collectingAGIPD ASIC for HIBEF
Charge carrier lifetime – especially of holes – in high-Zsemiconductor materials like GaAs and CdTe is shortcompared to elementary semiconductors like Si or Ge[24]. Recent advancements in the production of thesesensor materials [25] mitigate e ff ects like ’ afterglow ’ and’ polarisation ’ (decribed in [26]) and made compoundsemiconductor sensors the subject of investigations foran alternative to Germanium sensors in high-flux –high-energy imaging detectors at FELs [27, 28, 29]. Inaddition sensors made from high-Z materials (includingGermanium) do not show the defect mechanisms ofsilicon decribed in [30], while the high absorption ofradiation in the sensor and the low cross section of siliconat high photon energies drastically reduces the dosedeposited in the readout ASICs. However the currentAGIPD ASICs (and sensors) are hole collecting devices,and thus not suitable for high-Z materials. For this reasonan electron collecting version of the AGIPD ASIC,ecAGIPD, to equip the AGIPD camera at the HIBEFendstation with high-Z sensors is being developed.The primary di ff erence is the lowered baseline (oper-ating point at ≈
400 mV) of the charge sensitive pream-plifier. This is necessary to maintain a dynamic range of ≈ . igure 12: Spectre
Simulation results showing (from left to right) gain ( ×
29 compared to × − ≈
500 MHz) andnoise density of the preamplifier core of the electron collecting ecAGIPD ASIC (green) in comparison to the hole collecting AGIPD (red).Figure 13: Circuit schematic of the ecAGIPD, the electron collecting version of the AGIPD ASIC. The yellow frames contain the componentslocated in isolated p-wells.
Since the baseline is only minimally higher than an n-MOS’ threshold voltage, it requires the protection of theamplifier inputs and calibration stimuli sources to workfrom a negative potential, which was implemented usingthe triple well option of the GF 130 nm process. These p-wells are biased below substrate potential to allow biassources, input protection diodes, and n-MOS feedbackswitches to operate properly even in presence of largeinput charges. By these changes the performance of thepreamplifier core actually improves with respect to thehole collecting version, as the simulation results in fig.12 predict. Further changes to the circuit are minor: Besides theobvious reversal of the discriminator’s polarity requiredfor gain switching, also the levels encoding high and lowgain and the pads of the di ff erential analogue outputs wereswapped. This way ecAGIPD will deliver the same sig-nal polarities to the subsequent electronics facilitating thereuse of the existing firmware and calibration algorithms.The block schematic is shown in fig. 13. To evaluate thedesign AGIPD 0.6, a 16 ×
16 pixels prototype of ecAGIPD,has been manufactured and is currently characterised.11 . Going faster
European XFEL plans to introduce two additional bunchpatterns in the 2 nd half of the 2020ies [6]. These are fore-seen to be • CW operation at 100 kHz rate • Long pulse mode with 500 ms long bursts of ≤
200 kHz pulsesand would result in an × . . The largest benefit of the CWmode is the ability exchange (or at least move) samples in-between pulses and this way reduce radiation damage tothe sample and enable pulse-by-pulse experiments beyondjet based sample delivery.The burst frame rate of an integrating detector likeAGIPD is in principle only limited by the properties ofsensor and preamplifier. For the future CW operation ofthe European XFEL, readout bandwidth will become abottleneck. The AGIPD ASIC can cope with CW oper-ation up to a frame rate of ≈
16 kHz with original per-formance . At even higher rates (theoretically the ASICcan work at up to 96 kHz when omitting gain readout),the deterioration of the analogue readout signals due toe.g. skin e ff ect , and reflections will render operationimpractical.As a consequence, in-pixel digitisation becomes amust, while readout bandwidth remains one of the limit-ing factors. Assuming current technology (FireFly with28 Gbit / s / link), data rates of about 48 Gbit / s / cm or3 Gs / s / cm at 16 bit / sample can be reached. Fig. 14shows data rate as a function of pixel size and frame speedunder these conditions.An even more challenging problem for such detectorswill be power dissipation, since dynamic gain switching Other limits, like e.g. heat load of superconducting accelerator partsmight prevent an increase in brilliance. This mode is used during the characterisation and wafer-level test-ing of single ASICs. The readout systems of the SPB and MID AGIPDcameras require a network and firmware upgrade to implement it. At frequencies of several 10 MHz, signal currents will graduallystart to flow only on the surface of a conductor, leading to a rise ofimpedance and a non-flat frequency response. Reflections, i.e. parts of a signal traveling an electrical transmis-sion line in opposite direction, are generated at any discontinuity ofimpedance. This is predominantly visible for higher frequencies at e.g.connectors and non-matching termination resistors. Figure 14: Data rate per area as a function of pixel size and frame speed,assuming a 16 bit quantisation. requires the the preamplifier to be faster than the sen-sor’s charge collection time. Hence power consumptionis inversely proportional to the pixel area and will notrelax with the lower frame rate compared to the current4 . ≈ µ m × µ m will result in a powerdissipation of (cid:39) − and will be di ffi cult to imple-ment.
10. Summary
A 1 megapixel AGIPD detector system has been installedat the SPB instrument of the European XFEL in August2017. System fulfils all requirements, esp. in terms ofnoise, which is below 310 e − (i.e. ≤ . ≥ photons at 12 keV)and speed(4 . ×
16 pixel demonstrator has been manufac-tured and is awaiting evaluation. Since European XFELwill provide di ff erent pulse patterns in the 2 nd half of the2020ies, we are studying concepts for an ultra high fram-erate ( ≥
100 kHz) imager with in-pixel digitisation, wherepower consumption and readout bandwidth become lim-iting factors.
References [1] Europran XFEL, [2] M. Altarelli et al., [European X-ray Free Elec-tron Laser. Technical Design Report], ISBN 978-3-935702-17-1 (2006), DOI:10.3204 / DESY 06-097[3] H. Graafsma, ”Requirements for and developmentof 2 dimensional X-ray detectors for the Euro-pean X-ray Free Electron Laser in Hamburg”,
JINST
4, P12011 (2009), DOI:10.1088 / / / / P12011[4] M.C. Veale et al., ”Characterisation of the highdynamic range Large Pixel Detector (LPD) andits use at X-ray free electron laser sources”,
JINST
12, P12003 (2017), DOI:10.1088 / / / / P12003[5] M. Donato rt al., ”First functionality tests of a 64 × JINST
12, C03025 (2017),DOI:10.1088 / / / / C03025 [6] R.Brinkmann et al., ”Prospects for CW and LPoperation of the European XFEL in hard X-rayregime”, Nucl. Instr. Meth. A 786, 20 (2014),DOI:10.1016 / j.nima.2014.09.039[7] J. Becker et al., ”The Adaptive Gain In-tegrating Pixel Detector at the EuropeanXFEL”, J. Synchrotron Rad. 26, 74-82 (2019),DOI:10.1107 / S1600577518016077[8] D. Grei ff enberg et al.: ”The AGIPD detector forthe European XFEL”, JINST
7, C01103 (2012),DOI:10.1088 / / / / C01103[9] B. Henrich et al.: ”The adaptive gain integratingpixel detector AGIPD: a detector for the EuropeanXFEL, Nucl. Instr. Meth. A, 633, Supplement 1(2010), s11-S14, DOI:10.1016 / j.nima.2010.06.107[10] A. Allahgholi et al., ”The Adaptive Gain IntegratingPixel Detector AGIPD”, JINST
11, C02066 (2015),DOI:10.1088 / / / / C02066[11] J. Schwandt et al, ”Design of the AGIPD sensor forthe European XFEL”, Nucl. Instr. Meth. A 731, 252-254 (2013), DOI:10.1016 / j.nima.2013.04.031[12] U. Trunk et al., ”Architecture and designof the AGIPD detector for the EuropeanXFEL”, 2012 IEEE Nuclear Science Sympo-sium and Medical Imaging Conference Record(NSS / MIC), IEEE, ISBN: 978-1-4673-2028-3,DOI:10.1109 / NSSMIC.2012.6551175[13] X. Shi et al.: ”Challenges in chip design for theAGIPD detector”, Nucl. Instr. Meth. A 624, 387-391(2010), DOI:10.1016 / j.nima.2010.05.038[14] X. Shi et al., ”A low noise high dynamicrange analogue front-end ASIC for the AGIPDXFEL detector”, Proc. 19th IEEE InternationalConference on Electronics, Circuits, and Sys-tems , ISBN 978-1-4673-1261-5, 933 (2012).DOI:10.1109 / ICECS.2012.6463508[15] D. Mezza et al., ”Characterization of AGIPD 1.0:the full scale chip”, Nucl. Instr. Meth. A, 838, 39(2016). DOI:10.1016 / j.nima.2016.09.0071316] W. Buttler et al., ”Noise filtering for readout elec-tronics”, Nucl. Instr. Meth. A 288, 187 (1990),DOI:10.1016 / / j.nima.2012.08.008 (2012)[18] J. Becker et al.: ”Simulation and Experimen-tal Study of Plasma E ff ects in Planar SiliconSensors”, Nucl. Instr. Meth. A, 624, 716-727,DOI:10.1016 / j.nima.2010.10.010 (2010)[19] DESY Maxwell HPC Cluster, http://it.desy.de/services/computing_infrastructure/maxwell_hpc_cluster_including_gpu_computing/index_eng.html [20] I. A. Vartanyants et al.: ”Intensity interferometryof single x-ray pulses from a synchrotron storagering”, Phys. Rev. Lett. 113, 064801, / / / PhysRevLett.113.064801[21] M. O. Wiedorn et al.: ”Megahertz serial crystal-lography” Nat Communications 9(1), 4025 (2018),DOI:10.1038 / s41467-018-06156-7[22] D. Mezza et al., ”New calibration circuitry andconcept for AGIPD,” JINST
11, C11019 (2016),DOI:10.1088 / / / / C11019.[23] Samtec FireFly, [24] A. Owens, A. Peacock: ”Compound semiconductorradiation detectors”, Nucl. Instr. Meth. A, 531, 18-37 (2004), DOI: 10.1016 / j.nima.2004.05.071[25] M.C. Veale et al.: ”Cadmium zinc telluride pixeldetectors for high-intensity x-ray imaging at freeelectron lasers”, J. Physics D, 52 085106 (2019),DOI: / / / aaf556[26] M. Strassburg, Ch. Schroeter and P. Hackenschmied:”CdTe / CZT under high flux irradiation”,
JINST
6, C01055 (2011), DOI:10.1088 / / / / C01055 [27] J. Becker et al.: ”Characterization of CdTe sen-sors with Schottky contacts coupled to charge-integrating pixel array detectors for X-ray science”,
JINST
11, P12013 (2016), DOI:10.1088 / / / / P12013[28] J. Becker et al.: ”Characterization of chromiumcompensated GaAs as an X-ray sensor mate-rial for charge-integrating pixel array detectors”,
JINST
13, P01007 (2018), DOI: / / / / / P01007[29] M.C. Veale et al.: ”Investigating the suitabilityof GaAs:Cr material for high flux X-ray imag-ing”,
JINST
9, C12047 (2014), DOI:10.1088 / / / / C12047[30] J. Zhang, R. Klanner et al.: ”Study of X-ray ra-diation damage in the AGIPD sensor for theEuropean XFEL”,
JINST
9, C05022 (2014),DOI:10.1088 / / / //