EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH

Transcription

1 EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH CERN-TOTEM-NOTE September 2015 Timing performance of diamond detectors with Charge Sensitive Amplifier readout M. Berretti, E. Bossini, N. Minafra Abstract CERN-TOTEM-NOTE /10/2015 Research on particle detectors based on synthetic diamonds has always been limited by the cost, quality and availability of the sensitive material. Moreover, the read-out electronics requires particular care due to the small number of electron/hole pairs generated by the passage of a minimum ionizing particle. However, high radiation hardness, low leakage currents and high mobility of the electron/hole pairs make them an attractive solution for the time-of-flight measurements and the beam monitoring of new high energy physics experiments where the severe radiation environment is a limitation for most of the technologies commonly used in particle detection. In this work we report the results on the timing performance of a mm 2 sccvd (single crystal Chemical Vapour Deposition) sensor read-out using a charge sensitive amplifier. Both sensors and amplifiers have been purchased from CIVIDEC Instrumentation. The measurements have been performed on minimum ionizing pions in two beam tests at the PSI and CERN-PS facilities with two different detector capacitances. In particular, a time resolution of the order of 200ps has been obtained. CERN and INFN Sezione di Pisa INFN Sezione di Pisa and Museo Storico della Fisica e Centro Studi e Ricerche E. Fermi CERN, University of Bari and INFN Sezione di Bari

2

3 Timing performances of diamond detectors with Charge Sensitive Amplifier readout 1 Introduction Particle detectors based on artificial diamond as sensitive material are characterised by high radiation hardness, low leakage currents and high mobility of the electron/hole pairs [1]. Such properties make them an attractive solution for time-of-flight measurements and beam monitoring of new high energy physics experiments where the severe radiation environment [2 4] is a limitation for most of the technologies commonly used in particle detection. Thanks to the continued developments in the single crystal Chemical Vapour Deposition (sccvd) diamond production technology, detectors with high charge collection efficiency are nowadays available, with affordable prices if the surface to be covered by the detector is small. However, due to the high effective energy which is needed to create an electron/hole pair ( 13eV) the charge released in sccvd sensor by a Minimum Ionized Particle (MIP) is small (on average e for a thickness of 500 µm) and a low noise amplifier is therefore needed. In this work we report the results on the timing performance of a 4.5x4.5 mm 2 sccvd sensor read out by a Charge Sensitive Amplifier (CSA). Both sensors and amplifiers have been purchased from CIVIDEC Instrumentation 1. The measurements have been performed on MIP pions in two beam tests at the PSI and CERN-PS facilities with two different detector capacitances. The time resolution measured in this work is, to our knowledge, the best result so far obtained on MIP particles with commercial diamonds and amplifiers. Better performance can be obtained by developing a single PCB hosting both the diamond sensor and a transconductance preamplifier. The latter has to be placed as close as possible to the sensor in order to minimize the parasitic capacitance of the connection. This approach, proposed in [5] allowed to obtain a time resolution on MIP of about 110 ps, which is the best result in the literature. The document is organized as follows: section reports the experimental setup used to perform the timing measurement. Section reports details about the diamond sensor and the CSA. In section the measurements of the time resolution, obtained with different analysis techniques, are finally reported. Experimental setup The data used in this article have been taken in two test-beams. The first set of measurements has been collected at the πm1 beam facility of the Paul Scherrer Institute (PSI, Villigen, Switzerland). The πm1 beam is a quasi-continuous high-intensity secondary beam generated on a fixed target 22 meters upstream of our detector. Although the beam is mainly composed of pions, a non negligible contamination of electrons and muons is present. The particle momentum can be selected in the range MeV/c with a resolution better than 0.1%. The detector consists of an RF-shielded box (30x30x60 cm 3 ) housing both the diamond detectors and the trigger system. The trigger system is composed of two fingers of plastic scintillator (3x3x30 mm 3 ) directly coupled to two single pixel SiPM (HAMAMATSU S C), as shown in fig1. The two detectors were placed one upstream the diamond detectors and the other downstream in cross configuration, selecting particles in a spot of 3x3 mm 2. Amplification to the SiPM signal was provided through custom amplifier boards and the amplified signal sent to a NIM crate. The trigger efficiency, defined as number of triggered events with signal from at least one of the diamond detectors was around 40%, depending on the alignment of the detector with respect to the beam. The intrinsic efficiency of the diamond can be instead considered to be around 100%. To check this we tried to repeat the alignment until we reached an efficiency of almost 100% during one data taking run (since this fine alignment was time consuming we can not do it for every setup we tested). Moreover we took data triggering on electrons: the trigger efficiency in the electron samples was the same as with the π, meaning that there was no inefficiency due to the lower energy release of the π in the detectors. 1

4 2 M. Berretti, E. Bossini, N. Minafra Fig. 1: Picture of the scintillator fingers (a) and SiPM (b) used for the trigger and one assembled detector (c). The diamond configuration for all test-beams will be presented in the next section, while a schematic representation of the entire setup can be found in fig.2 Fig. 2: On the left the box with the two diamond detector in the middle. On the right a scheme of the complete setup; the top scintillator was used only at PS and will be described below. It was possible to identify the different particles thanks to the possibility of changing the momentum of the beam: electrons are already relativistic at the minimum energy, hence their TOF is not depending on the beam momentum. On the other hand, the TOF of the pions is the most affected. Therefore, to perform particle identification we made use of the RF signal of the beam and of the trigger timing. RF provides the time of the impact of the primary beam on the production target, while the trigger signal gave the arrival time of the generated particle at the detector. The trigger system was able to detect the particle arrival with a time resolution of 1 ns, enough to trigger only on desired particles. The distribution of the arrival time with respect to the RF signal for different beam energies is given in fig.3. The particle flow was reduced by closing the beam collimator to obtain a trigger rate of 1 khz in order to prevent more than one particle passing through the detectors at the same time. Data at 250 MeV/c triggering on different particles were collected using an oscilloscope Agilent DSO9254A, placed in the experimental area as close as possible to the Diamond amplifier output. Almost the same configuration was used during the data taking at the T9 beam line [6] at CERN, served by the Proton Synchrotron (PS) accelerator. The T9 beam line is able to provide charged particles with momenta up to 12 GeV/c, mainly π ± and protons, randomly distributed in slots of 400 ms. For our purpose we selected negative particles with momenta of 10 GeV/c. Selecting particles with negative sign is possible to get rid of the proton component, leaving an almost pure π beam. To check contamination with electrons and muons a gaseous cherenkov counter placed near the detector was used. The cherenkov gas pressure was tuned to detect 10 GeV/c particles lighter than π and put in coincidence with the trigger system. A contamination well below 1 was detected.

5 Timing performances of diamond detectors with Charge Sensitive Amplifier readout 3 Fig. 3: Distribution of particle time of flight (TOF) with respect to the beam RF signal. Particle peaks generated from the same collision are indicated; since TOF difference between different particles type is bigger than the collision period (20 ns) multiple collisions are overlapped. As expected the e time of flight does not vary with beam energy. The selection of negative particles had as side effect a very low luminosity beam and thus forced us to enhance the trigger efficiency by placing another detector in front of the box. This detector was a 10x10 cm 2 tile of plastic scintillator with a hole of 1 mm aligned with the beam coupled to a PMT. The signal from it was used to generate a veto signal on the trigger. Thus a maximum efficiency of 60% was reached. Data collection was performed with two different oscilloscopes, the Agilent of the previous test-beam and a Tektronix DPO73304, still placed next to the detector. Diamond detectors with charge sensitive amplifiers Several previous works [7] [8] have shown the relevance of Cividec Instrumentation on the market of Diamond Detectors. The selected diamonds are a pair of identical sccvd diamonds 4.5x4.5x0.5 mm 3 provided by Cividec Instrumentation with a single pixel metallization of 4x4 mm 2. Each diamond is protected by an RF shield. The layout has been designed to work with Cividec amplifiers that provide the bias voltage and read the signal through an SMA connector. Among the others, the Cividec C6 Fast Charge Amplifiers have been proven to be the best solution for timing measurements. The tolerances on the components used by the company result in small differences in the behaviour of the amplifier that can lead to different results of the time resolution. For this reason, several amplifiers of the same type have been tested. The main parameters to characterize an amplifier are the Signal to Noise Ratio (SNR) and the Rise Time (RT); these parameters can be measured using a synthesized input and studying the output. The resistance of the diamond is of the order of PΩ and can be modelled as a capacitance and a current generator. However, for the purpose of this work, the amplifiers were tested using the diamond as signal source to be sure that the amplifiers are working in the right conditions. Usually, the rise time t 10% 90% is measured as the time difference between the instants when the signal is equal to 10% and 90% of its maximum. This value is related to the bandwidth of the signal, assuming a single-pole amplifier: BW = 0.35 t 10% 90% (1)

6 4 M. Berretti, E. Bossini, N. Minafra Fig. 4: General schema of an amplifier with the equivalent circuit of a diamond detector. However, it is possible to measure the rise time t 20% 80% between 20% and 80% of signal s maximum that is more robust for low SNR. In the same way, it is possible to relate it with the bandwidth of the signal: BW = 0.22 t 20% 80% (2) On the other hand, to compute the SNR, the usual procedure is to find the Most Probable Value (MPV) of the maximum of the signal and divide it by the RMS noise: SNR = MAX{v out} RMS{v noise } (3) In figure 6 we can see the distribution of signal maximum, used to compute SNR. The capacitance of the detector has a non-negligible influence on the behaviour of the output signal; in fact, the commercial PCB was modified together with Cividec with the goal of the lowest possible capacitance for a non-integrated system. The improvement in the rise time using the modified PCB is of the order of 20% for one of the amplifiers, as shown in fig. 5. Fig. 5: Rise time t 20% 80% distribution of one of the tested charge sensitive amplifiers; the modified PCB allows a rise time 20% faster, using the same CSA. The modified diamond board has a similar SNR compared to the commercial version, of about 17 for the MPV of a MIP.

7 Timing performances of diamond detectors with Charge Sensitive Amplifier readout 5 Fig. 6: Distribution of output signal maximum of the two CSA. The black line indicates the thresholds that will be used in the analysis described in sec.. From the measured values of t 20% 80%, combining (1), (2), we can obtain t 10% 90%. Using the SNR we can extimate a timing resolution on the order of 230 ps [5]: σ t t 10% 90% SNR Measurement of the time resolution 1.6t 20% 80% SNR 230ps (4) The time resolution σ T has been measured with several off-line algorithms. For each recorded event the time of a particular point in each of the two waveforms is determined (t Det1,t Det2 ). The time resolution is obtained as σ T / 2 where σ T is the standard deviation of the T 12 distribution, with T 12 t Det1 t Det2. The time resolutions obtained with different techniques are shown in Tab 1. The results are reported for minimum ionizing π, by using the two charge amplifiers CIV A, CIV B described in section and the diamonds with the smallest connecting capacitance ( 5 pf). The results are obtained by selecting waveforms having V peak > 10 mv for the first diamond and V peak > 8 mv for the second. Offline method T 12 fitted-value, ( T 12 RMS), [resolution] 1 Simple Threshold 1450 (1490) [1025] ps 2 Position of the Maximum 719 (754) [508] ps 3 Normalized Threshold (70%) 467 (491) [330] ps 4 Normalized Threshold (50%) 353 (359) [250] ps 5 Normalized Threshold (30%) 336 (341) [238] ps 6 Fitted Normalized Threshold (35%) 308 (315) [217] ps 7 Offline CFD 306 (298) [210] ps 8 Extrapolation of normalized Threshold 277 (281) [196] ps Table 1: Time resolution summary In method 1 a threshold of 12 mv is set and t Det1,2 is taken at the crossing point of the threshold. Method 2 selects the time t Det1,2 as the one where the maximum of the waveform is reached. The Normalized

8 6 M. Berretti, E. Bossini, N. Minafra Threshold (X%) method (3 to 5) takes the time t Det1,2 from the point in the rising edge of the waveform where the amplitude reaches X% of the maximum. In the method 6, in addition a polynomial fit (of 6 th order) of the waveform rising edge is performed and the Normalized Threshold value is measured on the fitted function. The Offline Constant Fraction Discriminator (CFD, method 7) emulates the standard technique widely used in the electronic circuits to reduce the time walk effects. More details on this method are reported in [9]. In short, each waveform is inverted, attenuated, delayed and finally added to the original one. The time for which the resulting function has a null amplitude is taken as an estimate of t Det1,2. The value of the attenuation factor and of the time shift used in this method is 0.45 and 7 ns respectively. Method 8 is obtained by performing a linear extrapolation to X=0 of the time values obtained by using a fitted normalized threshold method at 25% and 60%. In all our methods a constant fit has been performed on a 20 ns time interval prior to the diamond signal, aiming to remove the effect of the baseline fluctuation on the diamond signal. We didn t find any important variation of the results (< 5%) by introducing a filter based on the FFT transform or by smoothing the signal with a moving average or a Butterworth filter. The smallest time resolution, 196 ps, is obtained with method 8. The corresponding T 12 distribution is shown in fig. 7 with a Gaussian fit superimposed. It has been also found that the reported time resolutions are better by about 15% with respect to the ones obtained in the first beam test where the detectors had a larger capacitance (C 15 pf, see section ). The analysis has been repeated by reducing the minimum value of V peak of the analyzed waveforms to 4.5 mv. The fit results are stable at 1% level while outliers appear in the T 12 distribution so that the RMS gets worse by about 10%. In general, by selecting the signal where the highest amount of charge is released, the time resolution improves. As an example, by not including in the analysis the 20% of the diamond signals having the smallest V peak the time resolution is about 5-10% better, depending on the method. Fig. 7: Distributions of the time differences T 12 = t Det1 t Det2, obtained with method 8 (see tab. 1). A Gaussian fit is superimposed. Summary In this work we measured the time resolution of a timing detector based on commercial sccvd diamonds and charge sensitive amplifiers. A S/N of about 17 has been measured with a typical rise-time of about 2.5 ns. The time resolution

9 Timing performances of diamond detectors with Charge Sensitive Amplifier readout 7 was measured for MIP particles with several off-line methods, the best of which gives a result of about 200 ps. After the first tests, the capacitance of the detector was found to be critical and was improved with a custom PCB made in collaboration with Cividec. On this way, we plan to improve our results by assembling a setup similar to the one presented in [5], with a unique PCB hosting both the diamond sensors and a discrete-component front end electronics. Acknowledgments We thank Cividec, in particular its CEO Erich Griesmayer, for providing us the state of the art components used in this work and for their customizations to our needs. We are also grateful to Tektronix for giving us one of their top-line oscilloscopes. This work has been possible thanks to the TOTEM Collaboration. References [1] S. Koizumi, C. Nebel, M. Nesladek. Physics and Applications of CVD Diamond. Wiley, ISBN: , [2] M. Albrow et al. (CMS and TOTEM collaboration). CERN-LHCC ; TOTEM-TDR-003 ; CMS-TDR-13, [3] A. Goriek et al. Nucl. Inst. Meth. A, 572(1):67 69, [4] J. Pietraszko et al. Nucl. Inst. Meth. A, 618: , [5] M. Ciobanu et al. IEEE Trans. Nucl. Sci. 58(4), , [6] L. Durieu, A. Mueller, and M. Martini. Optics Studies for the T9 Beam Line in the CERN PS East Area Secondary Beam Facility. Conf.Proc., C : , [7] Nunzio Randazzo, Sebastiano Aiello, Gabriele Chiodini, Giuseppe AP Cirrone, Giacomo Cuttone, M De Napoli, V Giordano, Simon Kwan, Emanuele Leonora, Fabio Longhitano, et al. Comparative timing performances of s-cvd diamond detectors with different particle beams and readout electronics. In Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), 2012 IEEE, pages IEEE, [8] M Osipenko, S Minutoli, P Musico, M Ripani, B Caiffi, A Balbi, G Ottonello, S Argirò, S Beolè, N Amapane, et al. Comparison of fast amplifiers for diamond detectors. arxiv preprint arxiv: , [9] D. Lucsanyi. CERN-STUDENTS-Note , 2014.