05191280

1084 IEEE SENSORS JOURNAL, VOL. 9, NO. 9, SEPTEMBER 2009 Single-Photon Avalanche Diode CMOS Sensor for Time-Resolved Fluorescence Measurements David Stoppa, Member, IEEE, Daniel Mosconi, Lucio Pancheri, and Lorenzo Gonzo, Member, IEEE Abstract—A single-photon avalanche diode–based pixel array for the analysis of fluorescence phenomena is presented. Each 150 m2 pixel integrates a single photon detector 180 combined with an active quenching circuit and a 17-bit digital events counter. On-chip master logic provides the digital control phases required by the pixel array with a full programmability of the main timing synchronisms. The pixel exhibits an average dark count rate of 3 kcps and a dynamic range of over 120-dB in time uncorrelated operation. A complete characterization of the single photon avalanche diode characteristics is reported. Time-resolved fluorescence measurements have been demonstrated by detecting the fluorescence decay of quantum-dot samples without the aid of any optical filters for excitation laser light cutoff. Index Terms—CMOS, fluorescence lifetime, single-photon avalanche diode (SPAD), time-gating. I. INTRODUCTION N recent years, there has been a growing interest in fast and low-cost biological testing. However, existing analysisequipment are expensive, bulky and time-consuming, making their use exploitable for research applications only. Among the many methods used for biological testing, optical detection is the most commonly used. In particular, fluorescence lifetime imaging is an investigation tool of paramount importance in imaging of molecular processes in physics and life sciences research [1], allowing the mapping of many cell parameters (such as pH, ion concentrations, etc.) and the detection of pathologies or DNA sequencing [2]. A typical fluorescence lifetime experiment uses a pulsed or modulated laser pulse to excite the fluorophores and the emitted light is revealed by means of intensified CCD cameras or photomultipler tubes [3], in order to achieve the required time-resolution and light sensitivity specifications. The performance of these laboratory instruments are excellent in terms of time accuracy and spatial resolution but they are expensive and bulky, while there is an increasing demand for portable and inexpensive biosensors for environmental and biomedical diagnostics. In the last few years, there have been many efforts in realizing integrated circuits aimed at biosensing applications [4], [5], [11]–[15]. The use of a CMOS integrated optical I Manuscript received October 02, 2008; revised December 17, 2008; accepted February 03, 2009. Current version published August 05, 2009. The associate editor coordinating the review of this paper and approving it for publication was Dr. Krikor Ozanyan. The authors are with Fondazione Bruno Kessler—Centro per la Ricerca Scientifica e Tecnologica (FBK-irst), I-38100, Povo (TN), Italy (e-mail: stoppa@fbk.eu; mosconi@fbk.eu; pancheri@fbk.eu; lgonzo@fbk.eu). Digital Object Identifier 10.1109/JSEN.2009.2025581 sensor based on simple photodiodes has been proposed for fluorescence decay detection [5], but the time resolution was limited by the RC time constant of the photodiode. High accuracy time resolution can be achieved exploiting single-photon avalanche diodes (SPADs) [6], but most of the efforts in the implementation of such systems reported in the literature regard the integration of the SPAD with quenching circuits, and of a simple voltage comparator for the digital conversion of the received Geiger pulse [7]–[9]. Only a few examples of SPAD arrays with integrated readout channels have been reported so far. In [10], the authors report a linear SPAD array with an in-pixel Time-to-Amplitude Converter, used to detect the photons arrival time. In [12], a 128 128 pixel array and a column Time-to-Digital Converter have been demonstrated. A 16 4 SPAD pixel array with two pixel-level gated counters have been recently proposed for fluorescence lifetime measurement [13], and also integrated with micropixellated light-emitting diodes in a very compact analysis system [14]. We have recently presented a chip capable of performing time-gated fluorescence detection [15], which integrates, at the pixel level, an actively quenched SPAD and a 17-bit digital counter. The chip, which includes a 7 2-pixel array, was fabricated in a high voltage CMOS technology. 0.35In this paper, a complete characterization of the chip proposed in [15] is reported, using an FPGA-based control board and a USB interface for data acquisition. Using this system, lifetime measurements of quantum dot fluorophores without the use of optical filter for laser light cutoff have been successfully performed. The time-gated fluorescence lifetime measurement technique is briefly reviewed in Section II, while an overview of the pixel architecture is carried out in Section III. The chip architecture is presented in Section IV. Finally, a complete characterization of the system is reported in Section V. II. MEASURING TECHNIQUE The adopted measuring technique is sketched in Fig. 1. A sub-ns laser pulse illuminates the biological sample containing the fluorophores, and the low signal associated with the fluorescence emission is detected by the sensor. The adopted measuring technique is based on a time-gated detection method, where the light signal is detected by using two or more observation windows (OWs). Each window has an externally programmable time width and can be delayed with respect to the trigger laser pulse by a user-defined time value. The time offset between the laser pulse and the beginning of the observation 1530-437X/$26.00 © 2009 IEEE STOPPA et al.: SINGLE-PHOTON AVALANCHE DIODE CMOS SENSOR FOR TIME-RESOLVED FLUORESCENCE MEASUREMENTS 1085 Fig. 2. Cross section of the implemented SPAD. Fig. 1. Adopted time-gating measuring technique. window offers the possibility of suppressing unwanted background signals like scattering and auto-fluorescence. This improves the signal-to-background ratio when the light intensity is measured. The measurement starts setting the first OW synchronized with the laser trigger. If an avalanche event is generated within this window it will be detected by the in-pixel event counter. The measurement is then repeated for a programmable number so that a significant statistical population can be obof times tained and, finally, the pixel-array is read out. After that, other measurement cycles are performed by using time delayed observation windows. At the end of the full measurement, it is possible to sketch a histogram reporting the number of detected events within each OWs. With a proper choice of the OW time width, it is possible to estimate not only the light intensity (given by the integral of the histogram), but also the time constant of the fluorescence process [16]. If only one window is used, the sensor provides an average intensity measurement which is often satisfactory for many applications (e.g., DNA micro-arrays). III. PIXEL ARCHITECTURE The proposed sensor was designed and fabricated in a high CMOS technology. The structure of the fabvoltage, 0.35ricated SPADs is similar to that presented in other works in 0.8high voltage CMOS technologies [7]–[10]. A schematic cross section of the device is shown in Fig. 2. The active region of the device is made by a over deep junction. A implantation is formed using the guard-ring surrounding the layer which is available inside a deep in this kind of fabrication processes. The geometry of the device is square to optimize the area occupation, but the corners are smoothed so as to avoid electric field peaks at the junction corners. The active area is defined by means of an optical window opened in the top-metal light shield only in correspondence with the region where avalanche multiplication occurs. Fig. 3. Schematic diagram of the proposed pixel. The schematic diagram of the proposed pixel is sketched in Fig. 3. It basically consists of a SPAD, a quenching diode-connected transistor Mp2 to limit the avalanche current, a voltage comparator (INV1) for the avalanche event detection, and a digital counter able to count and store the number of avalanche events. To avoid the use of high-voltage transistors, expensive in terms of area occupation, the cathode of the SPAD is connected , and the biasing through the quenching transistor to of the SPAD is assured above the breakdown voltage biased at a negative by means of the external line . The main voltage whose typical value is drawback of this approach is that the maximum excess voltage available is below . The feedback loop, consisting of INV1 and Mn4, realizes an active quenching mechanism, able to force the input node to ground as soon as an avalanche event is detected. The two control signals Precharge and SPADoff are used to implement the active reset and quenching functions, respectively. When a photon is absorbed by the SPAD, the avalanche is triggered and the node OUT is pulled up. The feedback loop forces this state until the arrival of the next SPADoff pulse, at the end of each repetition period (quenching-function). This prevents the input node from unwanted recharging. At every rising edge of the signal CKstop, the event-counter is incremented only if OUT is high. By so doing, an avalanche event is counted only within the observation window defined by the rising edges of Precharge and CKstop. To reduce the area occupation, the in-pixel 17-bit digital events counter consists of a pseudorandom number generator, based on a LFSR (Linear Feedback Shift Register). With a suitable choice of the bits used for the XOR mask [17], the different states, so designed LFSR allows generating 1086 IEEE SENSORS JOURNAL, VOL. 9, NO. 9, SEPTEMBER 2009 Fig. 4. Die microphotograph and pixel layout. m SPADs with V = Fig. 6. Dark count rate distribution measured on the 20  . 4V Fig. 5. Block diagram of the chip testing board. that a maximum of 131 k events can be detected in the proposed pixel. The 17-bit shift registers are based on a dynamic latch [18] where the initial logic state is set by means of an internal reset switch. A more detailed description of pixel architecture and working principle are reported in [15]. IV. CHIP ARCHITECTURE The chip core is based on two different arrays of SPAD sensors, each one composed by seven pixels. SPADs of the first row have an active area of about , while in the second row devices have been used. An internal logic generator provides all the waveforms necessary to the pixel operation, while a configuration register has been implemented for full programmability of the on-chip digital part. Digital control signals can optionally be generated externally and provided to the chip. The microphotograph of the fabricated sensor and the pixel layout are shown in Fig. 4. V. EXPERIMENTAL CHARACTERIZATION The proposed sensor has been mounted on an field programmable gate array (FPGA)-based board, which is schematically shown in Fig. 5. The FPGA was used to generate timing and configuration signals as well as to manage the communication with a PC via a USB channel. The use of an FPGA allows the operation of the SPAD in both asynchronous and synchronous mode using the same hardware, implementing only firmware modifications. In asynchronous operation, each SPAD is recharged after a fixed dead time from the occurrence of an avalanche event. In synchronous operation, the SPAD is activated only during the observation window and is held in an off state for the rest of the cycle. The characterization of the SPAD performance has been conducted in asynchronous operation, while fluorescence measurements have been performed in synchronous operation mode. The breakdown voltage of the SPADs, which has been measured on several devices, varies from 27.5 to 28.4 V, and no dependence on the device area has been found. SPADs is in the order The dark count rate of cps at room temperature with an applied excess bias of and 1 dead time. This is indeed a very large value, which renders this detector not suitable for asynchronous operation, even considering cooling of the device. However, these large SPADs can be used synchronously, activating the devices for the small amount of time necessary to observe the desired event and shutting them off immediately after the event has finished. In a 10-ns OW, for example, the probability of observing a dark event will be about 1% and could be further decreased with device cooling. On the contrary, smaller SPADs can be used also in asynchronous operation, thanks to the lower dark count rate. HereSPADs is reported. after, the characterization of Dark count rate has been measured on several samples, to take into account its variability across different SPADs (Fig. 6) the complete DCR characterization of 17 devices tested on different . chips with an excess bias voltage Almost 60% of the devices have a dark count rate lower than 4 kHz, which is a fairly good value for a SPAD of this size fabricated in a CMOS technology. Fig. 7 reports the temperature dependence of dark count rate. At high temperatures, the slope of the curve is typical of SRH generation, while at low temperatures the dominant generation mechanism is tunneling [20]. The corner temperature, at which the two contributions have the same weight, is around 10 . Afterpulsing probability density was measured using a digital oscilloscope to histogram the time delay distribution between two subsequent pulses, and is shown in Fig. 8. As can be seen, the detrapping time constants are quite long, in the order of 1 . STOPPA et al.: SINGLE-PHOTON AVALANCHE DIODE CMOS SENSOR FOR TIME-RESOLVED FLUORESCENCE MEASUREMENTS m SPADs with Fig. 7. Dependence of dark count rate on temperature for a 20  . V = 4V 1087 Fig. 9. Photon detection probability measured at different excess bias voltages. Fig. 8. Afterpulsing probability density. Fig. 10. Measured SPAD response to an optical power sweep under a wide spectrum illumination. The SPAD is biased at with 1- dead time and 25-ms time window. V =4V Total afterpulsing probability can be reduced by increasing the dead time, as shown in [8]. In our device, this figure is 4.5% at setting a 500-ns dead time, but it can be reduced down to 0.5% by increasing the dead time to 2 . The spectral response of the SPAD has been measured using an electro-optical bench composed by a white-light source (QTH lamp, 250 W), a monochromator filter and a calibrated reference detector. With this setup, spectral response can be measured with a maximum relative error of 5% within the wavelength range 350–1000 nm. The measured spectral dependence of photon detection probability at different applied excess bias voltages is shown in Fig. 9. The maximum is reached at about 450 nm, because of the shallow depth of the /ntub active junction. The dynamic range of the SPAD in asynchronous operation has been measured using a wide spectrum lamp. An illumination power density spanning more than seven orders of magnitude was obtained on a dedicated electro-optical bench using a set of neutral filters and varying the distance of the sensor from the light source. The measured counts in a 25-ms time window as a function of optical power density incident on the SPAD is shown s in Fig. 10, together with the measured noise, obtained as the standard deviation over a large number of acquisitions. This curve can be corrected at low count rates by subtracting the dark counts, In this way, at small illumination power the dynamic range is limited by the Poisson noise related to dark counts. At large count rates, when the sum of all the dead times becomes a relevant fraction of the measuring time window , the curve can be corrected by using the relation (1) is the corrected counts, and where is the measured counts, is the dead time. The noise relative to the corrected counts can be obtained by applying error propagation to the measured noise. Both the corrected signal and noise are shown in Fig. 10. It is worth noting that, with this linearity correction, the sensor has a dynamic range of six orders of magnitude with a linearity error lower than 5%. 1088 IEEE SENSORS JOURNAL, VOL. 9, NO. 9, SEPTEMBER 2009 Fig. 11. Measured time resolution using a 470-nm pulsed laser with 80-ps pulse width. m Fig. 13. Fluorescence decay measurement performed with a 20- SPAD coupled to a TCSPC instrument and with our time gated readout channel. Fig. 12. Fluorescence measurement setup. The time resolution of the SPAD has been measured using a pulsed semiconductor laser at 470 nm as a light source. The laser pulse width was 80 ps FWHM and the histogram was accumulated using a TCSPC instrument [21]. The measurement is shown in Fig. 11. at A fluorescence decay measurement was performed using the setup shown in Fig. 12. A glass capillary (internal diameter of ) containing a fluorophore solution (CdSe/ZnS quantum 550 dots in toluene [22]) was mounted in close proximity to the chip surface and a pulsed laser light (470-nm, 80-ps FWHM, 1-MHz repetition frequency, 13-pJ pulse energy) was used to excite the fluorescence. To demonstrate the sensor capability of discriminating the fluorescence light from the laser excitation, and to characterize the sensor performance independently from the collection optics, a very simple setup has been used as shown in Fig. 12. As can be seen, neither focusing optics nor optical filters are present. To validate this setup, the photon arrival time histogram has been measured with a commercial TCSPC instrument [21] (see Fig. 13). Although the sharp laser line, due to scattered laser photons, is visible in the measurement because of the absence of optical filters, fluorescence decay could still be measured accurately. Fig. 14. Fluorescence decay measurement at different fluorophore concentra, performed with a 20- SPAD and time gated readout channel. tions C m A second measurement was then performed with the proposed system using the FPGA for data acquisition and transmission to a PC. In this measurement, a 10-ns observation window has been swept at 60-ps time steps across the time range of interest. The 60-ps time delays could be generated with very good accuracy by the FPGA. In this way, the convolution of the light signal with the 10-ns observation window has been measured and the result is shown in Fig. 13. Because of that, while the TCSPC instrument produces a narrow line in correspondence of the laser pulse, the proposed system generates a 10-ns knee in the left-part of the graph. The average extracted lifetimes are in good agreement, being 19.54 ns 50 ps and 19.97 ns 40 ps for the TCSPC and time gated systems, respectively. Fluorescence decay measurements performed on fluorophore solutions with different concentrations are shown in Fig. 14. As STOPPA et al.: SINGLE-PHOTON AVALANCHE DIODE CMOS SENSOR FOR TIME-RESOLVED FLUORESCENCE MEASUREMENTS expected, fluorescence intensity changes with fluorophore concentration, while the decay curve remains the same. Quantum dot fluorophores exhibit a multi-exponential decay [23], but the main signal contribute is within the first tens of nanoseconds, where the decay curve approximates a mono-exponential decay. It is worth noting that both intensity and lifetime information can be extracted from the measurement. The lifetime measurement setup presented so far offers the possibility of array integration, at a cost that is much lower than currently used TCSPC instruments. Moreover, time gated operation allows laser light cutoff without the use of optical filters, with a relevant simplification of measurement setup. The proposed readout channel can be further improved to include multiple gating needed to maximize the overall collection efficiency, thus minimizing acquisition times and reducing fluorophore photobleaching. VI. CONCLUSION A pixel architecture implementing a SPAD detector and a dedicated read out circuitry for fluorescence lifetime measurements has been presented. Each pixel is capable of single photon detection and measures the number of events generated within a user-defined observation window. A preliminary test-chip, consisting of a 14-pixel array, has been fabricated CMOS technology. The characterin a high-voltage 0.35ization of the proposed sensor with an FPGA-based control board has been shown. The SPADs have shown a six orders of magnitude dynamic range under asynchronous operation. Time-gated lifetime measurements of quantum dot fluorophores without the use of optical filter for laser light cutoff have been demonstrated. ACKNOWLEDGMENT The authors would like to thank Austria Micro Systems for its support in designing layout structures for the SPADs implementation and Dr. G. Pedretti for his valuable assistance in the sensor testing. REFERENCES [1] J. R. Lakowicz, Principles of Fluorescence Spectroscopy, 2nd ed. New York: Springer-Verlag, 1998. [2] R. Cubeddu, D. Comelli, C. D’Andrea, P. Taroni, and G. Valentini, “Time-resolved fluorescence imaging in biology and medicine,” J. Phys. D: Appl. Phys., no. 35, pp. R61–R76, 2002. [3] M. Straub and S. W. Hell, “Fluorescence lifetime three-dimensional microscopy with picosecond precision using a multifocal multiphoton microscope,” Appl. Phys. Lett., vol. 73, no. 13, pp. 1769–1771, 1998. [4] H. Eltoukhy, K. Salama, A. El Gamal, M. Ronaghi, and R. Davis, “A 0.18 um CMOS 1E-6 lux bioluminescence detection system-on-chip,” in Proc. ISSCC, Feb. 2004, pp. 222–223. [5] G. Patounakis, K. L. Shepard, and R. Levicky, “Active CMOS array sensor for time-resolved fluorescence detection,” IEEE J. Solid-State Circuits, vol. 41, no. 11, pp. 2521–2530, Nov. 2006. [6] K. K. Ng, Complete Guide to Semiconductor Devices. New York: Wiley, 2002. [7] A. Rochas, M. Gani, B. Furrer, P. A. Besse, R. S. Popovic, G. Ribordy, and N. Gisin, “Single photon detector fabricated in a complementary metal-oxide-semiconductor high-voltage technology,” Rev. Scientific Instruments, vol. 74, no. 7, pp. 3263–3270, 2003. [8] F. Zappa, S. Tisa, A. Gulinatti, A. Gallivanoni, and S. Cova, “Monolithic CMOS detector module for photon counting and picosecond timing,” in IEEE Proc. ESSDERC, 2004, pp. 341–344. 1089 [9] C. L. Niclass, A. Rochas, P. A. Besse, and E. Charbon, “A CMOS single photon avalanche diode array for 3D imaging,” in Proc. ISSCC Digest Tech. Papers, Feb. 2004, pp. 120–121. [10] D. Stoppa, L. Pancheri, M. Scandiuzzo, L. Gonzo, G.-F. Dalla Betta, and A. Simoni, “A CMOS 3-D imager based on single photon avalanche diode,” IEEE Trans. Circuits Syst. I, vol. 54, no. 1, pp. 4–12, Jan. 2007. [11] D. E. Schwartz, E. Charbon, and K. L. Shepard, “A single-photon avalanche diode imager for fluorescence lifetime applications,” in Proc. Digest Symp. VLSI Circuits, 2007, pp. 144–145. [12] C. Niclass, C. Favi, T. Kluter, M. Gersbach, and E. Charbon, “A 128 128 single-photon imager with on-chip column-level 10b time-todigital converter array capable of 97 ps resolution,” in Proc. ISSCC, 2008, pp. 44–45. [13] B. R. Rae, C. Griffin, K. R. Muir, J. M. Girkin, E. Gu, D. R. Renshaw, E. Charbon, M. D. Dawson, and R. K. Henderson, “A microsystem for time-resolved fluorescence analysis using CMOS single-photon avalanche diodes and micro-LEDs,” in Proc. ISSCC, 2008, pp. 166–167. [14] B. R. Rae, C. Griffin, J. McKendry, J. M. Girkin, H. X. Zhang, E. Gu, D. R. Renshaw, E. Charbon, M. D. Dawson, and R. K. Henderson, “CMOS driven micro-pixel LEDs integrated with single photon avalanche diodes for time resolved fluorescence measurements,” J. Phys. D: Appl. Phys., no. 41, pp. 1–6, 2008. [15] D. Mosconi, D. Stoppa, L. Pancheri, L. Gonzo, and A. Simoni, “CMOS single-photon avalanche diode array for time-resolved fluorescence detection,” in Proc. ESSCIRC, 2006, pp. 564–567. [16] H. C. Gerritsen, M. A. H. Asselbergs, A. V. Agronskaia, and W. G. J. H. M. Van Sark, “Fluorescence lifetime imaging in scanning microscopes: Acquisition speed, photon economy and lifetime resolution,” J. Microscopy, vol. 206, pp. 218–224, 2002. [17] P. Kitsos, N. Sklavos, N. Zervas, and O. Koufopavlou, “A reconfigurable linear feedback shift register (LFSR) for the bluetooth system,” in Proc. IEEE Conf. Electron. Circuits Syst., Sep. 2001, vol. 2, pp. 991–994. [18] J. M. Rabaey, Digital Integrated Circuits. Englewood Cliffs, NJ: Prentice-Hall, 1996, pp. 351–353. [19] D. Stoppa et al., “A single-photon avalanche-diode 3D imager,” in Proc. IEEE ESSCIRC, 2005, pp. 487–490. [20] A GulinattiI. Rech, P. Maccagnani, M. Ghioni, and S. Cova, “Large-area avalanche diodes for picosecond time-correlated photon counting,” in Proc. IEEE ESSDERC, 2005, pp. 355–358. [21] PicoQuant PicoHarp 300, [Online]. Available: www.picoquant.com [22] [Online]. Available: www.evidenttech.com/products/evidots.html [23] U. Resch-Genger, M. Grabolle, S. Cavaliere-Jaricot, R. Nitschke, and T. Nann, “Quantum dots versus organic dyes as fluorescent labels,” Nature Methods, vol. 5, pp. 763–775, 2008. 2 David Stoppa (SM’97–M’02) received the Laurea degree in electronics engineering from Politecnico of Milan, Milan, Italy, in 1998, and the Ph.D. degree in microelectronics from the University of Trento, Trento, Italy, in 2002. In 2002, he joined the Integrated Optical Sensors Group, Fondazione Bruno Kessler (FBK-IRST, formerly ITC-IRST), as a Research Scientist, working on the development of CMOS image sensors for advanced applications. Since 2000, he has been teaching at the Telecommunications Engineering faculty of the University of Trento courses on analogue electronics and microelectronics. His research interests are in the design of high-performance CMOS imagers and biosensors. He has authored or coauthored more than 40 papers in international journals and presentations at international conferences, and holds two patents in the field of image sensors. Dr. Stoppa was a recipient of the Best Paper Award at the 2006 European Solid-State Circuits Conference. Daniel Mosconi received the Laurea degree in electronics engineering from the University of Padua, Padua, Italy, in 2002. In 2003, he was with the Altran Group, where he worked as Consultant in Selex Communications, as a DSP Firmware Engineer. In 2005, he joined the Integrated Optical Sensors Group, Fondazione Bruno Kessler (FBK-IRST), as a Researcher working on integrated optical sensors and biosensors in CMOS technology. Since July 2007, he has been with Zobele Holding Spa as an Electronic Engineer. He is currently the Supervisor of the Electronic Center of the Zobele Group. 1090 Lucio Pancheri received the Laurea degree in materials engineering from the University of Trento, Italy, in 2002, and the Ph.D. degree in information and communication technologies from the University of Trento in 2006. During his Ph.D., he has worked on the design and characterization of novel silicon photodetectors for 3-D imaging applications. In 2006, he joined the Integrated Optical Sensors Group, Fondazione Bruno Kessler (FBK-IRST), as a Research Scientist. He is currently a Postdoctoral Fellow at FBK-IRST, where is involved in the development of optical sensors for advanced imaging and biomedical applications. Lorenzo Gonzo (M’97) received the degree in physics from the University of Trento, Trento,Italy, in 1983 and the Ph.D. degree in physics from the Technical University of Munich, Munich, Germany, in 1992. IEEE SENSORS JOURNAL, VOL. 9, NO. 9, SEPTEMBER 2009 He is a Senior Scientist at FBK-IRST and Head of the Research Unit Smart Optical Sensors and Interfaces—SOI. In 1998, he joined the Department of Surface Sciences, FBK-IRST, where he has been working on tunneling microscopy and surfaces investigation techniques applied to microelectronics. Since 1993, he has been working with the Integrated Optical Sensor Group, FBK-IRST, where he was involved in the development of integrated optical sensors in CMOS technology. His main research activity has been focused on sensors for imaging both in 2–D and 3–D and with on chip signal processing. Since 2006, as Head of the Research Unit SOI, he is coordinating a research activity on high sensitivity optical sensors, single-photon avalanche diodes, time gated and multispectral imaging. He has been working also in the field of metrology with particular attention on 3–D measurements for realistic multiscale 3–D modeling of objects, buildings and land. He has authored or coauthored more than 70 papers in international journals and presentations at international conferences, and holds two patents in the field of image sensors.