Electrooptical measurement system for the DC characterization of visible detectors for CMOS-compatible vision chips

1 Electro-optical measurement system for the DC characterization of visible detectors for CMOS compatible vision chips Elisenda Roca1, Fabián Frutos2, Servando Espejo1,2, Rafael Domínguez-Castro1,2 and Angel Rodríguez-Vázquez1,2 (1) Instituto de Microelectrónica de Sevilla - Centro Nacional de Microelectrónica Avda. Reina Mercedes s/n, (Edif. CICA) E-41012, Sevilla, Spain (2) Area de Electrónica, Departamento de Electrónica y Electromagnetismo, Universidad de Sevilla, Avda. Reina Mercedes s/n, E-41012 Sevilla (Spain). IEEE Transactions on Instrumentation and Measurements. Submitted for publication, 1998. © 1999 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE. This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author’s copyright. In most cases, these works may not be reported without the explicit permission of the copyright holder. 2 Electro-optical measurement system for the DC characterization of visible detectors for CMOS compatible vision chips Elisenda Roca1, Fabián Frutos2, Servando Espejo1,2, Rafael Domínguez-Castro1,2 and Angel Rodríguez-Vázquez1,2 (1) Instituto de Microelectrónica de Sevilla, Centro Nacional de Microelectrónica, Avda. Reina Mercedes s/n, Campus Universidad de Sevilla, E-41012 Sevilla (Spain). (2) Area de Electrónica, Departamento de Electrónica y Electromagnetismo, Universidad de Sevilla, Avda. Reina Mercedes s/n, E-41012 Sevilla (Spain). Submitted to: IEEE Transactions on Instrumentation and Measurements. 3 Abstract An electro-optical measurement system for the DC characterization of visible detectors for CMOS compatible vision chips is presented which can help designers to characterize these detectors. The measurement system has been designed to be versatile, fast and easily expandable and used. Two different set-up’s for the measurement of the spectral response and the optical dynamic range of the detectors are described in detailed. Measurements of the spectral response are done with a fully computer controlled set-up, avoiding tedious and inaccurate measurements. A description of the different detectors available in a CMOS process is also given, together with the parameters affecting their response and a set of test structures which can be useful for the characterization of the detectors. 1. Introduction In the last years there has been a growing interest in the development of CMOS compatible imagers which can substitute CCD cameras in imaging applications [1], overcoming the limitations of this technology when small sizes, low power consumption and random access are required [2]. CMOS imagers also allow on-chip processing of the image data which is very complicated and area expensive in a CCD camera [2]. This last property has been of great importance for the development of vision chips which not only detect the light intensity and transduce it into an electrical variable (voltage or current), but also process in real-time the signals from an array of photodetectors by including parallel analog signal processing on-chip. A good revision of vision chips including signal processing circuitry can be found in [3]. 4 One major problem encountered when designing a vision chip in standard CMOS technologies is that foundries do not deliver characterization data for the available photosensors. Thus, designers are forced to use simplistic behavioral models based on the idealized CMOS-compatible photosensor descriptions found in literature. This may be enough for chips intended to process binary (black-and-white) images [4]. However, the characteristics of real devices largely differs from their idealized descriptions, being, on the other hand, strongly dependent on the fabrication process. Consequently, using these idealized models yields very inaccurate results whenever the analog content of the incoming light (regarding both intensity and wavelength) is significant for signal processing; i.e. chips whose behavior is anticipated on the basis of such simplified models will, most likely, not accomplish the specifications 1. In this scenario designers of vision chips are confronted to the necessity to fully characterize the CMOS-compatible photosensors by themselves. Unfortunately, this is not a simple task because it requires instrumentation apparatus and expertise which are not common to chip designers. This paper is intended to help designers to fill this gap. Based on a set of CMOS-photosensor test structures, we present a measurement system (including equipment and procedures) for their DC characterization. This system has been conceived to achieve two major targets: first, it should be used by people with a minimum optical background (i.e., by chip designers); second, it should be versatile (meaning that it should allow for different types of measurements set-up’s), easily expandable, and fast. This is achieved by making the different optical and measurement parts to interact in such a way that the whole operation can 1. This risk cannot be undertaken because vision chips may be very large, up to 1cm2.The fabrication cost of a 1cm2 chip as part of a MCP in a standard 0.7µm technology is about 27.500 $. 5 be computer controlled, eliminating tedious and inaccurate measurements. In section 3, the set of test structures useful for the characterization of visible detectors will be presented. Before a background on junction detectors available in a CMOS process will be given in section 2. This basic knowledge of the junction structures used for detection and the main parameters affecting their response will help to better understand the design of a dedicated chip for the characterization of the photodetectors and the measurement system used. A detailed description of the elements of this measurement system and the different setup’s mounted for their electro-optical characterization will be given in section 4. In this section the results obtained in the characterization of one of this detectors, which is a good candidate for the realization of vision chip, will be presented. 2. Theoretical background on CMOS junction photodetectors 2.1 Junction photodetectors In a standard CMOS process, several junction structures can be used to convert light into an electrical signal. These structures are depicted in Figure 1 for an n-well process (the complementary structures are obtained in a p-well process). The first three structures are p/n diodes, whereas the last one is a vertical bipolar junction transistor (BJT). Diodes are formed by well/substrate (n-/p-) junctions, source-drain diffusion/substrate (n+/p-) junctions and source-drain diffusion/well (p+/n-) junctions. Vertical p/n/p BJT’s are obtained by the structure formed by a p+ source-drain diffusion, an nwell and the psubstrate. When light is incident on a diode, photons absorbed in the depletion region of the p/n junction create electron-hole pairs which will be separated by the high internal electric field 6 formed across it. This charge can be detected as an increase in the reverse current of the device or as a change in the voltage across it. Carriers generated in the silicon bulk at a distance within a minoritary carrier diffusion length from the depletion region can also contribute to the detected signal, increasing the sensitive volume of the detector. The detected signal in a diode can be represented by a photogenerated current [5]: (1) where A is the detector area, P the incident light power per unit area, λ the wavelength of the incident light in microns and ηthe quantum efficiency of the process. For a phototransistor which is polarized in the active forward region with the base terminal (the well) left floating, the incident light generates electron-hole pairs in the reversed biased base-collector junction that induce a current which is equivalent to a base current. This base current is amplified by the gain of the transistor, resulting in an emitter current which is (1+βF) times the detected current. The emitter current is expected to be equivalent to the detected current in a well/substrate diode of the same well area, but amplified by (1+βF). The photogenerated current will be given by, (2) where A is the area of the transistor base. Comparing both types of detectors, photodiodes have the advantage of simpler structure and faster response, whereas phototransistors benefit from greater gain, although they also Iph A P λ η⋅ ⋅ ⋅ 1.24 ----------------------------= Iph A P λ η⋅ ⋅ ⋅ 1.24 ---------------------------- 1βF +( )= 7 require proportionally more power to operate. The optical dynamic range of a photodiode or a phototransistor is normally defined as the range of incident light power over which the detected signal behaves as a linear function, according to equation (1) or (2), respectively. Usually its lower limit is determined by the dark current of the detector, which is the current flowing through the device without illumination, whereas the upper limit may be device or external circuit limited. For the detectors available in a standard CMOS process the optical dynamic range is about 7 decades of incident light intensity for photodiodes and 5 decades for phototransistors [6]. The difference in photoresponse between each type of sensor is determined by the fact that the absorption of light in silicon is strongly dependent on the wavelength. The absorption length, L(λ), is shorter for shorter photon wavelengths [7]. This means that photodetectors formed with junctions which have different junction depths will have different spectral response. According to this, the three different types of diodes introduced in Figure 1 will present different spectral responses. From literature it is known that the n-/p- and n+/pdiodes behave in a similar manner [8], increasing their response with photon wavelength. The p+/ndiodes are more sensitive at shorter wavelengths as collection of carriers is limited by the volume bounded by the surface and the n-/pjunction, in contrast with the other diodes where the collection of carriers is limited by the diffusion length of minoritary carriers in the silicon bulk. Phototransistors response is very similar in shape to the spectral response of the n-/pdiodes but increased by the amplifying factor (1+βF) [8]. 2.2 Technology and area scaling effects The response of these detectors is also affected by other parameters, such as technology 8 scaling and detector size and shape. The effect of technology scaling is mainly caused by the wavelength dependence of the absorption length. When technology dimensions shrink, the spectral response of the devices shifts to shorter wavelengths, reducing their spectral detection range. This is due to the fact that for scaled processes the vertical dimensions get shorter, and photogenerated carriers contributing to the detected signal are those closer to the surface. Longer wavelength photons can contribute only if the junction is deep enough. This means that the detector response degrades for long wavelengths when the technology is scaled down. It is also known that in submicron technologies the doping level of the junctions is substantially higher than the corresponding value in a technology larger than 1µm, therefore affecting the junction characteristics and hence the response of the detectors. Technology scaling also implies a voltage scaling which imposes important limitations in the dynamic range due to the decrease of the full signal charge capacity [2]. Technology scaling is certainly followed by a pixel scaling. The trends showed during the last years indicate that pixel size and fill-factor (ratio of sensor area to pixel size) scale linearly with the minimum feature size [2]. This reduction in sensor area implies a reduction in the photogenerated current, as can be seen from equation (1). Nevertheless, this linear relationship is no longer valid when the area of the detector is very small. At this point the volume of the lateral depletion region is comparable to the volume of depletion region determined by the detector area, therefore increasing the expected detected current. For this reason, it is important to characterize all possible detectors for different sizes independently of the circuitry to obtain the photogenerated current levels for each case. These effects have to be considered when selecting a detector for an imaging application. 9 All possible detectors have to be characterized for each technology in order to select the optimum device to match the requirements of the signal processing circuitry. Therefore the need for an electro-optical set-up which allows the characterization of these detectors. 3. Photodetectors test structures The characterization of the above described detectors needs in most cases of the realization of a dedicated chip with test structures which can be used to determine the actual response of the detectors processed in a given technology. The response of these detectors can be affected by different parameters which can be either process or lay-out dependent. Process dependent parameters (i.e. junction depths, doping concentration ...) cannot be modified by the designer, opposite to lay-out dependent parameters (detectors size and shape). Therefore, when designing a test chip to characterize visible detectors both types of parameters have to be considered. On the one hand, all possible devices or structures available in the technology to detect visible light have to be included in the chip to obtain their actual behavior in the visible spectrum. On the other hand, these detectors should be tested for different sizes and shapes to account for the lay-out dependent parameters. Ideally, each detector should be tested for as many as possible sizes and, for each size, different perimeters (shapes) should be included. Furthermore, the chip should include as many as possible instances of each detector configuration to obtain statistical information on their response. However, the cost of fabrication of such a chip would be very high since detectors have to be connected to bonding pads to have direct access to the detected signal. A more realistic approach would consist on including a limited number of discrete detectors with different sizes and shapes, covering all the possible device structures, and several arrays of 16 the detectors could be easily determined when using a LED (Light Emitting Diode) or a low power visible laser which can be modulated at different frequencies. 4.2 Electro-optical measurements Two different types of measurements are done to characterize the detectors: measurements of the spectral response and measurements of the optical dynamic range. Both types of measurements will be described now. 4.2.1 Spectral response measurements The spectral response of the detectors is obtained by mounting a very similar set-up to that depicted in Figure 4, although in this case the neutral density filters are not used. A 250W halogen-quartz lamp coupled to the monochromator is used as a monochromatic light source. Light going out of the monochromator is collimated before being splitted by the beamsplitter. The transmitted light impinges on the D.U.T. a current which is measured by the electrometer. The reflected light arrives to the calibrated detector, giving a measurement of the optical power incident on the D.U.T. (a small correction is done when calculating the quantum efficiency to take into account the ratio of transmitted to reflected light by the beamsplitter which is 36%/ 33%). Diodes are reverse polarized by the HP3245A power supply, and the current flowing through them is measured by the 6512 Keithley electrometer. First the dark current is determined to eliminate this component of the measured current with illumination and obtain the actual photogenerated current. This substraction of the dark current is done by software. The wavelength of the incident light is controlled by the HP-VEE program through the 17 IEEE controller of the stepper motor, therefore eliminating the manual change of the output wavelength. For each wavelength the program reads the optical power and the photocurrent from the optical powermeter and the electrometer, respectively. The program directly operates these figures to depict in real-time the quantum efficiency of the detector on a graph, taking into account both the dark current and the ratio of transmitted to reflected light by the beamsplitter. The use of the beamsplitter helps to reduce the inaccuracy in this type of measurements since the optical power can be measured almost simultaneously to the photogenerated current in the detector. If the beamsplitter is not used, the spectral output power of the lamp has to be measured before or after the spectral response of the detector, introducing an error in the measurements since the output power of the lamp could have slightly changed. The spectral response of the arrays of detectors can also be measured by using the HP- 8175A digital signal generator. Then, once the wavelength is selected, the photocurrent is measured for all the different detectors of one or more arrays. The HP-VEE program is slightly modified for this measurement by introducing the selection of the different detectors with the digital word generator. As an example of this type of measurement, the spectral response of a 100µmx100µm p+/ndiode can be seen in Figure 5. The ripple observed is due to multiple light interferences in the dielectric layers deposited on top of the detector, that depend on the dielectric layer materials and thicknesses. As expected, the quantum efficiency shows a maximum at short wavelengths (around 480nm) and then decreases with photon wavelength. 4.2.2 Optical dynamic range Measurements of the optical dynamic range are done using the set-up shown in Figure 6. 18 A 250W halogen-quartz lamp is used together with an interference band-pass filter to select the light wavelength. One or more neutral density filters are also used to reduce the light power arriving to the detector. Once the power supply of the lamp is set, the optical power is measured with the calibrated detector in the exact position where the D.U.T. will be placed. Then the photogenerated current is measured with the electrometer without the neutral density filter to obtain the first data point. Afterwards this current is measured placing different calibrated neutral density filters in between the light beam and the D.U.T. to reduce the light power arriving to the detector. In this case the filters are placed manually since in some cases a combination of two or more filters is used. The optical dynamic range for a p+/ndiode is shown in Figure 6 for different wavelength. As it can be observed, the optical dynamic range measured is 4 decades of incident light. Higher light power could not be reached due to the limitations of our lamp; however, it is expected that the dynamic range will improve in about 3 decades of light intensity, corresponding to the values known from literature. 5. Conclusions An electro-optical measurement system for the DC characterization of visible detectors is presented which can be used by vision chip designers to characterize these detectors. This measurement system is versatile, fast and easily expandable and used. Two different type of measurement are done: spectral response and optical dynamic range measurements. For the first type of measurements, the set-up is computer controlled, avoiding tedious and inaccurate measurements. A set of detector test structures for the characterization of a technology process is also given which can help designers in the selection of the detector better suited to their application. Results are presented for both types of measurements for diffusion/well diodes 19 processed in a double poly, double metal 0.8 µm technology. Acknowledgments This work is supported by NICOP Grant N68171-98-C-9004 and CICYT Grant No. TIC96-1392-C02-02. References [1] E. R. Fossum, “CMOS image sensors: Electronic camera on a chip,” Int. Electron Device Meetings, p. 17, 1995. [2] H.-S. Wong, “Technology and device scaling considerations for CMOS imagers,” IEEE Transactions on Electron Devices, Vol. 43, No. 12, pp. 2131-2142, 1996. [3] A. Moini, Vision Chips or Seeing Silicon, Technical Report, Centre for High Performance Integrated Technologies and Systems, The University of Adelaide, March 1997. [4] R. Domínguez-Castro, S. Espejo, A. Rodríguez-Vázquez, R. Carmona, Péter Földesy, Ákos Zarándy, Péter Szolgay, Tamás Szirányi and Tamás Roska: "A 0.8µm CMOS 2-D Programmable Mixed-Signal Focal-Plane Array Processor with On-Chip Binary Imaging and Instructions Storage". IEEE J. Solid-State Circuits, Vol. 32, No. 7, pp. 1013-1026, July 1997. [5] S.M. Sze, Physics of Semiconductor Devices, 2nd edition, John Wiley & Sons, 1981. [6] J. Mann, "Implementing early visual processing in analog VLSI: light adaptation", Proc. of SPIE, Visual Information Processing: from neurons to chips, Vol. 1473, pp. 128-132, 1991. 20 [7] W.C Dash and R. Newman, “Intrinsic optical absorption in single-crystal germanium and silicon at 77K and 300K,” Physical Review, Vol. 99, pp. 1151-1155, 1955. [8] T. Delbrück, Investigations of Visual Transduction and Motion Processing, Ph. D. Thesis, Computational and Neural Systems Program, California Institute of Technology, 1993. [9] C.H. Aw and B.A. Wooley, “A 128x128-pixel standard-CMOS image sensor with electronic shutter”, IEEE Journal of Solid-State Circuits, Vol. 31, No. 12, pp. 1922-1930, December 1996. 21 Figure captions Figure 1. Junction devices available in a standard n-well CMOS process for visible light detection. The complementary structures are obtained in a p-well process. The active junction is shaded for each device. (a) Well/substrate (n-/p-) diode. (b) Diffusion/substrate (n+/p-) diode. (c) Diffusion/well (p+/n-) diode. (d) Vertical bipolar transistor (p+/n-/p-). Figure 2. Lay-out of a dedicated test chip for the characterization of junction photodetectors. The metal2 layer, which covers most of the chip area, is not shown to allow a clear image of the lay-out Figure 3. Lay-out of the junction devices available in a standard CMOS process for visible light detection. (a) Well/substrate diode. (b) Diffusion/substrate diode. (c) Diffusion/well diode. (d) Vertical bipolar transistor. Figure 4. Schematic representation of the different elements of the electro-optical measurement system. The set-up indicated in the figure is used for spectral response measurements, although for these measurements the neutral density filter is not included. Figure 5. Measurement of the spectral response of a p+/ndiode. The ripple observed is due to 22 multiple light interferences in the dielectric layers deposited on top of the detector. Figure 6. Schematic representation of the set-up used for the measurements of the optical dynamic range. Figure 7. Optical dynamic range for a diffusion/well diode at different light wavelengths. 23 Figure 1 p-type substrate n-well P+ out p-type substrate P+ p-type substrate n-well P+ Vddout Vdd out p-type substrate n-well P+ GND out GND out GND P+ GND out out N+ out N+ P+ N+ (a) (b) (d) (c) 24 Figure 2 row decoder column decoder array decoder discrete detectors discrete detectors discrete detectors arrays output VSS VDD discrete detectors test pixel test pixel discrete detector 25 Figure 3 (a) (b) (c) (d) n-well n+ contact ring substrate contact ring output (metal 1) n+ diffusion substrate contact ring output (metal 1) n-well substrate contact ring diffusion output (metal 1) p+ diffusion n+ contact well output (metal 1) p+ diffusion substrate contact ring emitter output (metal 1) n-well