Soft X-ray sensitivity of a photon-counting hybrid pixel detector with a Silicon sensor matrix. A. Fornaini 1, D. Calvet 1,2, J.L. Visschers 1 1 National Institute for Nuclear Physics and High-Energy Physics (NIKHEF) Amsterdam, The Netherlands 2 Centre de Physique des Particules de Marseille (CPPM) Marseille, France Abstract We have tested a CMOS pixel readout circuit bump-bonded to a high resistivity Silicon sensor to study the noise properties of the system and to determine the lower boundary in energy for the detection of X-rays at room temperature. Our test shows that single photon counting is possible down to about 5 kev, and we obtain a mean total noise equivalent to142 electrons on the electronics input. PACS code: 87.59.-e Keywords: hybrid pixel detector, Medipix, Silicon, X-ray Further author information: Tel: +31205925177, Fax: + 31205925155, email: aforna@nikhef.nl
Page 2 1 Introduction Medipix is a low-noise, pixellated CMOS circuit, developed to be bump-bonded to a solid state pixel detector for medical X-ray imaging in the energy range above ~15 kev [1]. In this energy range, high-z sensor materials like GaAs are preferred because of their better conversion efficiency. For this material, however, several parameters like the noise generated by trapping and detrapping processes, and the charge collection efficiency, are not fully established. Recently, within the Medipix2 Collaboration at CERN [2], some Medipix samples have become available which are bump-bonded to a high-resistivity Silicon sensor. We have tested one such sample to study the noise properties of the Medipix circuit itself, and to determine the lower boundary in X-ray energy, where single-photon counting is still possible at room temperature. Our results show that single photon counting devices may be operated down to about 5 kev. 2 Experimental Set-up We use an assembly of a Medipix circuit bump-bonded to a Silicon sensor matrix of 4096 reverse biased p-implanted diodes of 170 x 170 µm 2 each. This matrix is surrounded by several guard rings. The sensor thickness is 300 µm, and we illuminate through a thin aluminized ohmic contact on the backside. The reverse bias voltage was put on this backside contact and was equal to 100 Volts, which is just above full depletion. The Medipix circuit was operated at the standard
Page 3 working point (Vdd = 3.0 V, Vdda = 3.0 V, Vgnd=1.5 V, Vbias = 1.60 V, Vcomp= 1.00 V, Vdel = 0.8 V). The Medipix option to equalise individual pixel thresholds was not used. The assembly is read-out via two commercially available I/O boards, the National Instrument s AT-AO-10 (for the analog voltage supplies and the circuit control lines), and the National Instrument s PCI-DIO-32HS (for the bi-directional digital data transfer and clock). Connection between these boards and the circuit is made via the Medipix1 re-usable Read-Out System (MUROS1) board, which translates the TTL levels of the PC to the CMOS levels of the Medipix circuit. Special care has been taken in the design of this MUROS1 board, to decouple and shield the analog connections towards the Medipix circuit. Power supply is taken from the +5V external connection of the PC. The Medisoft-3 software has been provided by the Medical Physics group in Naples [4]. As X-ray sources we have used a soft X-ray tube with Chromium anode emitting at 5.42 kev, and an X-ray tube with Copper anode emitting at 8.05 kev. The Chromium source was operated at 7.0 kv acceleration and 200 µα current. The Copper source was operated at 45 kv and 40 ma and was equipped with a Germanium monochromator having a spectral width below 1 ev. 3 Results Energy distributions as measured on a single arbitrary pixel, are shown in figure 1. They were obtained by a so-called threshold scan, by stepping the analog threshold voltage over intervals of 10 mv and differentiating the resulting count rate distribution. We observe clearly separated peaks of 5.42 and 8.05 kev and both peaks are separated from the noise region on the left in 44 out of the 64 pixels illuminated. The relation between threshold voltage and X-ray energy is assumed to be linear according to what the designers of this circuit expect for such a small region. Knowing that the charge
Page 4 collection efficiency in Silicon is very close to 100%, while the energy to create an electron-hole pair is 3.62 ev, we calibrate the horizontal scale in terms of equivalent electrons-hole pairs from the known distance of 2.53 kev between the two peaks. The total system noise (electronics noise plus detector noise plus statistical noise), per pixel then follows from the observed width of the 8.05 kev line. The results, plotted in figure 2, show a mean total noise of 142 electrons, with a sigma of 23 electrons. This is considerably better than the ~250 electrons determined for GaAs sensor matrices in [3]. Our result is obtained using 44 of the 64 pixels that were illuminated by the 8.05 kev source. The other 20 pixels were discarded because the separation between the lower peak and the region where oscillation starts, was insufficient for a good position determination. 4 Conclusion Our result indicates that the lower bound in photon energy for room-temperature operation of photon-counting devices is as low as 5 kev. Noise is lower than assumed until now, and will decrease further in the next generation of circuits, due to smaller pixel size and improved electronics circuitry. References [1] M. Campbell, E.H.M. Heijne, G. Meddler, E. Pernigotti, W. Snoeys, "Readout for a 64 x 64 Pixel Matrix with 15-bit Single Photon Counting", IEEE Trans.Nucl.Sci. 45 (3), 751, June 1998. [2] http://medipix.web.cern.ch/medipix/
Page 5 [3] B. Mikulec, M.Campbell, G Dipasquale, C.Schwartz, J. Watt. "Characterisation of a Single Photon Counting Pixel System for Imaging of Low-Contrast Objects", presented at the 11th International Workshop on Room Temperature Semiconductor X- and Gamma- Ray Detectors and Associated Electronics, October 1999, Vienna, Austria, submitted for publication in NIM-A and CERN preprint CERN-EP/99-167. [4] L. Abate, E. Bertolucci, M. Conti, G. Mettivier, M.C. Montesi, P. Russo, "Quantitative dynamic imaging of biological processes with solid state radiation detectors", presented at the IEEE Nuclear Science Symposium, Seattle, October 1999, submitted to IEEE Transactions in Nuclear Science. Figures Figure 1. X-ray energy distributions as measured in one single pixel by threshold scan. Gaussian fits are used to determine the positions of the two peaks needed for energy calibration. The upper distribution shows the response to the 5.42 kev source, the lower is measured with 8.05 kev monochromatic source. Figure 2. Total system noise per pixel, distributed over the 44 pixels in the analysis and expressed in electrons.
Fig. 1 Page 6
Fig. 2 Page 7