WEBVTT 00:00.920 --> 00:06.100 Hello everyone, my name is Meghna Patil and I am a PhD student at IBPT 00:06.100 --> 00:06.700 in KIT. 00:08.000 --> 00:10.560 And I work on beam diagnostics for CARA. 00:11.660 --> 00:15.980 Today in this talk, I will be talking about ultra-fast detectors for 00:15.980 --> 00:17.160 online beam diagnostics. 00:18.400 --> 00:22.840 One of these detectors is CALYPSO, which is a fast line array camera 00:22.840 --> 00:27.740 which can be used as a beam diagnostic tool for wide spectral range 00:27.740 --> 00:28.160 measurements. 00:28.840 --> 00:33.060 And the other is CAPTURE, which is a sampling system for ultra-short 00:33.060 --> 00:35.460 pulses generated by terahertz detectors. 00:36.500 --> 00:38.300 So, let's get started. 00:41.560 --> 00:45.000 Here is a brief outline of what I will be covering throughout the 00:45.000 --> 00:45.280 talk. 00:45.760 --> 00:50.220 First, I will introduce CARA, our research and test facility. 00:50.980 --> 00:54.520 Then, a beam dynamic phenomenon called as the micro-bunching 00:54.520 --> 00:55.140 instability. 00:55.140 --> 01:00.280 I will gather my motivation from here and introduce you all to the new 01:00.280 --> 01:02.220 versions of CALYPSO and CAPTURE. 01:03.080 --> 01:05.040 And finally, its applications. 01:08.910 --> 01:12.690 CARA is an accelerator test facility located at KIT. 01:13.590 --> 01:16.610 It has a circumference of 110 meters. 01:17.650 --> 01:22.270 It has an operating energy range from 500 MeV to 2.5 GeV. 01:22.270 --> 01:27.470 The electron bounce spacing is 2 nanoseconds and we operate in low 01:27.470 --> 01:30.730 alpha mode in addition to the other modes of operation. 01:32.990 --> 01:38.110 Highlighted on the left are a few diagnostic ports used to study micro 01:38.110 --> 01:39.070 -bunching instability. 01:41.070 --> 01:45.770 CALYPSO and CAPTURE are installed for several studies related to 01:45.770 --> 01:49.250 longitudinal, transversal and terahertz diagnostics. 01:51.350 --> 01:55.890 Now, coming back, the low alpha mode, which is achieved by reducing 01:55.890 --> 01:59.790 the momentum compaction factor, results in short punches. 02:00.770 --> 02:04.690 These short punches interact with their own radiation. 02:06.110 --> 02:09.710 The self-interaction then results in the formation of microstructures 02:09.710 --> 02:12.930 in the longitudinal phase space of the electron bunch. 02:14.230 --> 02:18.870 This phenomenon is called as the micro-bunching instability and it is 02:18.870 --> 02:20.930 extensively studied in our facility. 02:22.590 --> 02:27.110 So, I come to the motivation as to why we need ultra-fast diagnostics. 02:27.950 --> 02:32.050 To understand complex beam dynamics, which occur in very short time 02:32.050 --> 02:35.350 scales, fast real-time measurements are essential. 02:37.530 --> 02:41.250 Hence, we need detectors which can have megahertz repetition rates. 02:41.250 --> 02:45.730 High spatial resolution, wide spectral sensitivity. 02:46.290 --> 02:49.370 They should be capable of long acquisition times. 02:50.350 --> 02:53.850 They should also be capable of synchronizing with the experimental 02:53.850 --> 02:57.030 setup and also with other diagnostic tools. 02:59.410 --> 03:01.430 One such tool is CALYPSO. 03:02.370 --> 03:06.130 So, let me just briefly explain the building blocks of CALYPSO. 03:07.130 --> 03:12.290 Here, we have a microstrip sensor based on different semiconductors, 03:12.590 --> 03:15.630 which is connected to a low-noise ASIC. 03:16.570 --> 03:19.410 This ASIC operates at megaframe per seconds. 03:20.270 --> 03:25.430 When fully populated, there are 64 ADC channels which are operating at 03:25.430 --> 03:27.570 125 megasamples per second. 03:28.070 --> 03:33.210 We have several external clock inputs available to synchronize with 03:33.210 --> 03:34.390 the experimental setup. 03:35.250 --> 03:40.810 We then have an on-chip PLL, which can be programmed in order to 03:40.810 --> 03:43.810 distribute the clocks based on user requirement. 03:44.970 --> 03:49.930 This entire detector is connected to an FPGA-based readout card. 03:50.630 --> 03:56.350 This FPGA card can either be a standalone card or can be a micro-GCA 03:56.350 --> 03:57.810 -based DAQ system. 04:02.000 --> 04:05.340 Sensor is the most elementary part of CALYPSO. 04:06.140 --> 04:10.400 In this section, I will talk about the working principle of the sensor 04:10.400 --> 04:14.100 used in CALYPSO and the different kinds of sensors. 04:18.870 --> 04:23.490 We use a microstrip sensor as the sensing element in CALYPSO. 04:25.050 --> 04:29.530 This is a brief explanation as to how a silicon microstrip sensor 04:29.530 --> 04:30.010 works. 04:31.410 --> 04:36.650 Such a sensor has a bulk made up of, in this case here, n-doped 04:36.650 --> 04:37.170 silicon. 04:37.850 --> 04:40.870 The strips are made up of p-doped silicon. 04:41.990 --> 04:46.950 There are two ways to bias the sensor, either by the backplane or by 04:46.950 --> 04:49.610 punch -through from the bias pads on the top. 04:51.470 --> 04:55.850 When reverse bias is applied, a depletion region is formed in the 04:55.850 --> 04:56.190 bulk. 04:57.910 --> 05:03.710 And when a beam or a particle hits the sensor, electron hole pairs are 05:03.710 --> 05:06.930 generated or formed in this layer of silicon. 05:09.230 --> 05:13.910 These free electrons and holes are drifted by an electric field 05:13.910 --> 05:19.090 created by a pattern of anodes and cathodes interdigitized on the 05:19.090 --> 05:20.430 surface of the silicon. 05:21.090 --> 05:24.670 They are separated by a silicon dioxide insulator. 05:26.750 --> 05:31.550 My explanation here can be visualized in the TCard simulation I did a 05:31.550 --> 05:32.390 few years ago. 05:32.390 --> 05:38.430 In this current case here, the holes are collected by the strips and 05:38.430 --> 05:40.950 the electrons are collected by the backplane. 05:42.050 --> 05:48.250 The mobility of electrons is higher than that of holes, and hence they 05:48.250 --> 05:49.290 are collected faster. 05:50.950 --> 05:56.910 These charges collected are then fed into a preamplifier stage, which 05:56.910 --> 05:58.450 will be explained later. 06:02.370 --> 06:07.810 Now, let me get in detail about the different kinds of sensors used in 06:07.810 --> 06:08.290 Calepso. 06:09.110 --> 06:15.110 The most commonly used is a microstrip silicon sensor, which I just 06:15.110 --> 06:17.370 explained the working principle about. 06:19.030 --> 06:24.570 Silicon covers a wavelength sensitivity ranging from 350 nm to 1050 06:24.570 --> 06:25.210 nm. 06:26.810 --> 06:32.270 Now, silicon is basically transparent to wavelengths beyond 1050 nm. 06:33.550 --> 06:37.530 In order to extend our wavelength sensitivity, we have to use 06:37.530 --> 06:42.410 different kinds of semiconductor-based sensors, like InGaAs, lead 06:42.410 --> 06:44.430 sulfide, or lead selenide. 06:45.370 --> 06:50.190 These have sensitivities in the wide infrared range. 06:52.610 --> 06:57.590 The silicon sensor designed was taking the user requirement into 06:57.590 --> 07:02.730 account, be it the geometry or the spectral sensitivity required by 07:02.730 --> 07:02.990 them. 07:04.130 --> 07:09.010 The silicon sensor designed has an anti-reflecting coating made up of 07:09.010 --> 07:12.970 silicon nitride, which is applied in order to maximize the absorption 07:12.970 --> 07:13.570 of light. 07:14.770 --> 07:18.810 On the left, you can see the quantum efficiency measurements done for 07:18.810 --> 07:19.390 this sensor. 07:21.010 --> 07:26.410 It has been improvised for the near-ultraviolet, visible, and the near 07:26.410 --> 07:27.490 -infrared region. 07:29.730 --> 07:34.090 An initial test of the commercially available infrared sensors were 07:34.090 --> 07:38.190 done with Calepso using a 1560 nm laser. 07:39.270 --> 07:42.590 Here is a plot of the laser profile with all the three detectors. 07:42.590 --> 07:48.750 They clearly show very low noise, and they have different sensitivity 07:48.750 --> 07:54.030 due to the fact that they are sensors which have different quantum 07:54.030 --> 07:55.690 efficiency at this wavelength. 07:59.250 --> 08:02.510 Our current research is now focused on ElGuard. 08:03.250 --> 08:07.450 ElGuard is a low-gain avalanche detector which has an additional 08:07.450 --> 08:08.050 implant. 08:08.990 --> 08:13.850 When reversed biased, there is a high electric field formed at the 08:13.850 --> 08:15.950 junction of the two implants. 08:17.350 --> 08:21.010 This results in the sensor having a small internal gain. 08:22.090 --> 08:25.870 Due to this gain, the thickness of this detector can be reduced. 08:26.870 --> 08:30.850 And what we gain by reducing the thickness of this detector without 08:30.850 --> 08:36.230 compromising the signal amplitude is improving the timing response of 08:36.230 --> 08:36.970 this detector. 08:37.730 --> 08:41.750 This detector has a rise time of a very few picoseconds. 08:43.170 --> 08:47.450 Hence, this detector has a variety of timing applications. 08:50.410 --> 08:54.330 Calepso with ElGuard would yield data in hundreds of mega frames per 08:54.330 --> 08:54.650 second. 08:55.530 --> 09:00.870 This trench isolation ElGuard architecture helps in the sensor to have 09:00.870 --> 09:02.650 pitch down to 50 micrometers. 09:03.910 --> 09:09.410 This kind of ElGuard is in fabrication at FPK under the framework of 09:09.410 --> 09:12.770 RD50, under which KIT is also a part of. 09:13.850 --> 09:18.170 This sensor is being designed both for particle physics as well as for 09:18.170 --> 09:19.010 photon science. 09:23.950 --> 09:28.190 The signal from the sensor needs to be interpreted and decoded in the 09:28.190 --> 09:28.690 right way. 09:29.710 --> 09:34.610 This job is done by the application specific integrated circuit or the 09:34.610 --> 09:35.030 ASIC. 09:38.340 --> 09:43.800 Botart is a low noise ASIC designed at KIT capable of frame rates up 09:43.800 --> 09:44.880 to 12 megahertz. 09:45.640 --> 09:50.000 It has 128 input channels and 16 output channels. 09:50.880 --> 09:56.260 Initial measurements prove it to have very high linearity and very low 09:56.260 --> 09:57.240 noise performance. 09:58.160 --> 10:02.620 Due to its gain switching mechanism, we can achieve a wide dynamic 10:02.620 --> 10:03.200 range. 10:04.600 --> 10:09.540 The CSA or the charge sensitive amplifier or the preamplifier is the 10:09.540 --> 10:14.420 initial stage of this ASIC and is compatible to different kinds of 10:14.420 --> 10:15.520 semiconductor sensors. 10:18.610 --> 10:23.370 The layout has also been designed by taking radiation hardness into 10:23.370 --> 10:24.090 consideration. 10:24.610 --> 10:29.890 This is done by the implementation of P plus guard rings and enclosed 10:29.890 --> 10:32.110 layout transistors or ELTs. 10:35.470 --> 10:39.570 Here is a simple explanation as to how this ASIC works. 10:40.490 --> 10:45.110 The first stage is a charge sensitive amplifier, which converts the 10:45.110 --> 10:47.990 charges generated by the sensor to voltage. 10:49.270 --> 10:53.050 This charge sensitive amplifier, as mentioned before, is compatible 10:53.050 --> 10:55.210 with different semiconductor sensors. 10:56.210 --> 11:01.070 It is followed by a correlated double sampling stage that samples this 11:01.070 --> 11:01.570 voltage. 11:02.210 --> 11:07.410 In addition to removing any undesired offset by sampling the baseline 11:07.410 --> 11:08.110 as well. 11:09.210 --> 11:13.810 A controlled channel buffer samples the output of the correlated 11:13.810 --> 11:17.730 double sampling stage and holds it during the readout phase. 11:19.030 --> 11:23.770 The front end channels are then coupled by a multiplexer and then 11:23.770 --> 11:27.970 finally connected to a very high speed ADC or analog digital 11:27.970 --> 11:28.790 converter. 11:34.320 --> 11:39.540 Now, the interconnection between the sensor and the ASIC here is done 11:39.540 --> 11:42.580 by a wedge to wedge aluminum wire bonding process. 11:43.600 --> 11:47.860 With such a large sensor and bonds, it is very susceptible to damage 11:47.860 --> 11:50.480 and hence we encapsulate it using GlobTor. 11:51.340 --> 11:55.980 All these processes are done in-house at KIT with the support from our 11:55.980 --> 11:57.980 colleagues from the eCAPE Institute. 12:00.160 --> 12:02.440 Now is the time for data processing. 12:04.840 --> 12:09.040 Now, let us see the complete data acquisition flow, which is currently 12:09.040 --> 12:10.880 implemented for Calypso. 12:11.660 --> 12:16.060 One branch is where we use HiFlex, which is a standalone FPGA card 12:16.060 --> 12:20.180 combined with a GPU or a CPU via PCIe. 12:21.500 --> 12:26.140 We also have a new version of this card, which is compatible with 12:26.140 --> 12:27.700 micro -DCA based systems. 12:29.300 --> 12:34.780 This FPGA card is based on Zinc UltraScale Plus and it can be used 12:34.780 --> 12:39.480 with micro-DCA based systems via the Firefly optical links present on 12:39.480 --> 12:39.880 the card. 12:42.470 --> 12:46.830 Here we have our new HiFlex 2, which I mentioned in the previous 12:46.830 --> 12:47.190 slide. 12:47.750 --> 12:52.190 This will serve as our DAQ card for our future diagnostic tools, 12:52.290 --> 12:53.290 including Calypso. 12:54.190 --> 13:00.090 It has a modern Zinc UltraScale Plus system-on-chip FPGA, several 13:00.090 --> 13:04.530 optical links, DDR4 memory chip and several other communication 13:04.530 --> 13:09.950 interfaces like Ethernet, SD or USB for slow control. 13:11.250 --> 13:14.810 This card is currently being used for implementing several machine 13:14.810 --> 13:15.910 learning algorithms. 13:18.330 --> 13:23.350 Now, I will talk about our second diagnostic tool, Capture, which is 13:23.350 --> 13:27.470 capable of digitizing pulse shapes with a sampling time down to 3 13:27.470 --> 13:31.570 picoseconds and pulse repetition rates up to 1 GHz. 13:32.570 --> 13:36.950 It has been designed for bunch-by-bunch acquisition of the terahertz 13:36.950 --> 13:38.730 coherent synchrotron radiation. 13:40.230 --> 13:45.230 It is used in combination with a terahertz detector that converts 13:45.230 --> 13:49.510 photons into electrical pulse with picosecond time resolution. 13:50.850 --> 13:56.290 The higher data rate produced by this sampling system is processed in 13:56.290 --> 14:02.110 real -time by a heterogeneous FPGA GPU architecture operating up to 6 14:02.110 --> 14:05.830 .5 Gbps continuously. 14:10.420 --> 14:15.120 Now, I will explain the architecture of the Capture board. 14:15.920 --> 14:21.720 The signal of the fast terahertz detector is fed via a wideband low 14:21.720 --> 14:25.260 -noise amplifier into a wideband power splitter. 14:25.960 --> 14:31.060 The power splitter splits the detector pulse into 4 or 8 identical 14:31.060 --> 14:36.040 pulses, which is distributed into the Capture pulse sampling board. 14:36.040 --> 14:39.740 It is based on 4 sampling channels. 14:41.060 --> 14:47.060 Each channel consists of a track-and-hold unit and a 12-bit 500 MSPS 14:47.060 --> 14:48.740 analog -to-digital converter. 14:50.060 --> 14:56.340 The sampling time of each channel is individually adjustable by a 14:56.340 --> 15:00.400 picosecond delay chip with a resolution of 3 picoseconds. 15:02.100 --> 15:06.960 This enables a local sampling of the detector pulse in minimum steps 15:06.960 --> 15:08.140 of 3 picoseconds. 15:10.460 --> 15:14.620 The local sampling depends on the chosen delay between the channels 15:14.620 --> 15:18.220 and the maximum sampling rate is 330 Gsps. 15:19.880 --> 15:24.680 The newest version of Capture enables two of these cards to operate in 15:24.680 --> 15:28.760 parallel, thus providing the user with 8 sampling points. 15:30.200 --> 15:34.780 For results from Capture, please have a look at the talk from Miriam 15:34.780 --> 15:35.340 Rossi. 15:39.260 --> 15:44.800 We at KIT have also been exploring several machine learning algorithms 15:44.800 --> 15:47.720 and applications and possibilities. 15:48.820 --> 15:52.920 One such application is to control the micro-bunching instability 15:52.920 --> 15:54.800 using reinforcement learning. 15:56.040 --> 16:00.740 It is now possible to implement a closed-loop algorithm on hardware. 16:01.800 --> 16:07.840 The terahertz signal is sampled using Capture and the data is 16:07.840 --> 16:09.200 processed in the FPGA. 16:10.280 --> 16:15.840 This FPGA card also gives the inference signal to control the bunch-by 16:15.840 --> 16:20.280 -bunch feedback system of CARA, thus closing the feedback loop. 16:21.460 --> 16:25.780 For more details on this, please refer this paper from my colleagues. 16:29.090 --> 16:33.050 Now, I will give a brief overview of the applications of CALIPSO and 16:33.050 --> 16:34.650 where it is currently used. 16:35.410 --> 16:40.730 At CARA, EHER and KIT, we use an electro-optical spectral decoding as 16:40.730 --> 16:44.630 a way to measure the longitudinal profile of the electron beam. 16:45.750 --> 16:50.190 Here, you can see the current setup with the yttrium-based laser with 16:50.190 --> 16:52.490 a central wavelength of 1050 nm. 16:53.630 --> 16:59.510 This wavelength operates at a 2.7 MHz repetition rate, which is also 16:59.510 --> 17:03.470 the repetition rate of the storage ring in a single-shot mode. 17:04.750 --> 17:07.290 This is the near-field setup. 17:08.050 --> 17:12.910 In this near-field setup, the coulomb field of the electron bunch is 17:12.910 --> 17:17.430 encoded in the chirped laser pulses at the gallium phosphide crystal. 17:18.850 --> 17:24.870 By analyzing the laser pulse, we can reconstruct the longitudinal 17:24.870 --> 17:27.030 profile of the electron beam. 17:28.230 --> 17:32.070 The data seen here is the raw data taken by CALIPSO. 17:33.170 --> 17:37.670 The several structures seen here corresponds to the micro-bunching 17:37.670 --> 17:40.370 instability, which was mentioned before. 17:40.370 --> 17:45.930 The same experimental data can also be used to reconstruct the 17:45.930 --> 17:48.630 longitudinal phase space of the electron beam. 17:50.510 --> 17:54.870 You can have a look at this paper if you are interested in a detailed 17:54.870 --> 17:57.410 explanation of this reconstruction algorithm. 18:01.120 --> 18:05.980 Energy spread of an electron bunch is also a crucial parameter in 18:05.980 --> 18:07.100 micro -bunching studies. 18:07.100 --> 18:12.500 It cannot be studied directly, but it can be by measuring the 18:12.500 --> 18:18.420 horizontal bunch of the incoherent synchrotron radiation, as it is 18:18.420 --> 18:22.880 coupled to the energy spread in the dispersive sections of the 18:22.880 --> 18:23.440 accelerator. 18:25.220 --> 18:30.840 This radiation emitted at the visible light diagnostic port of CARA is 18:30.840 --> 18:35.340 in the visible spectrum, ranging from 400 nm to 700 nm. 18:36.780 --> 18:41.280 Given here is an optical setup for the study in a single turn 18:41.280 --> 18:41.640 resolution. 18:43.340 --> 18:48.400 We see the first plot, which is nothing but the raw data acquired from 18:48.400 --> 18:48.920 CALIPSO. 18:50.180 --> 18:54.880 The center and the bottom panel show the corresponding bunch position 18:54.880 --> 18:56.100 and bunch sizes. 18:58.440 --> 19:02.360 The ELCART-based CALIPSO can be very useful here in order to improve 19:02.360 --> 19:03.480 the dynamic range. 19:08.000 --> 19:13.740 One more application is laser diagnostics, in order to characterize 19:13.740 --> 19:16.700 your laser and to detect any instabilities. 19:18.300 --> 19:22.860 Due to its high frame rate, CALIPSO enables single-shot measurements 19:22.860 --> 19:27.760 that enables to correct and sort data, and to compensate for any 19:27.760 --> 19:30.020 fluctuations in your laser spectra. 19:31.560 --> 19:37.900 Here you can see the average of 100,000 turns of laser spectra, as 19:37.900 --> 19:40.080 well as the intensity over time. 19:45.960 --> 19:51.100 A similar EO setup is also found at European Exafiel for the 19:51.100 --> 19:55.320 measurement of longitudinal profile and also as a beam arrival 19:55.320 --> 19:55.820 monitor. 19:56.960 --> 20:03.560 Here they use a micro-TCA-based DAQ system, and they can measure 20:03.560 --> 20:06.840 electron bunch lengths down to 200 femtoseconds. 20:08.520 --> 20:13.600 CALIPSO has been integrated here to operate at 2.7 MHz with readout 20:13.600 --> 20:16.380 electronics based on micro-TCA architecture. 20:20.960 --> 20:23.760 Next up is fine-tuning of FELs. 20:24.560 --> 20:28.400 Here is the VLS spectrometer setup at FLASH. 20:31.120 --> 20:35.480 Here, in this animation, you can see the instabilities in the spectra 20:35.480 --> 20:36.240 at startup. 20:37.340 --> 20:42.280 And this is after using CALIPSO as a diagnostic tool to tune the 20:42.280 --> 20:45.000 machine resulting in a more uniform spectra. 20:49.480 --> 20:53.280 CALIPSO has not only applications in the accelerator physics field, 20:53.520 --> 20:57.520 but also in other fields like medical imaging and ultra-fast 20:57.520 --> 21:03.520 microscopy, like for sorting cancer cells and for other applications 21:03.520 --> 21:05.540 like optical coherent tomography. 21:07.640 --> 21:12.460 It also has applications in the field of material sciences to study 21:12.460 --> 21:16.540 the effects of carrier envelope phase-dependent currents in 2D 21:16.540 --> 21:17.220 materials. 21:18.240 --> 21:23.160 These are some of the applications which we will be pursuing in the 21:23.160 --> 21:23.460 future. 21:26.960 --> 21:32.340 Finally, I would like to conclude that beam diagnostics relies on 21:32.340 --> 21:39.000 several technological aspects, from sensors, ASICs, wire bondings to 21:39.000 --> 21:41.760 high -throughput DAQs and data processing. 21:43.060 --> 21:47.520 Quick and easy access to excellent infrastructure to test in-house 21:47.520 --> 21:51.680 developed electronics and exporting them to other accelerator 21:51.680 --> 21:54.580 facilities is an added advantage. 21:56.420 --> 22:01.220 CALIPSO and CAPTURE are such projects which enables physicists and 22:01.220 --> 22:04.960 electronic engineers to come together for a collaboration. 22:06.560 --> 22:11.000 CALIPSO and CAPTURE are now a fundamental tool to understand beam 22:11.000 --> 22:13.560 dynamics with femtosecond time resolution. 22:16.120 --> 22:18.800 With this, I would like to conclude my talk. 22:19.220 --> 22:23.220 If you have any questions, please visit me at the questionnaire room. 22:23.220 --> 22:24.800 Thank you.