CMOS photodetectors for industrial position sensing

zyxwvutsrqp IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 43, NO. 3, JUNE 1994 489 CMOS Photodetectors for Industrial Position Sensing sion: Anssi J. Miikynen, Juha T. Kostamovaara, and Tim0 E. Rahkonen zyxwvutsrqponm zyxwvutsrq zyxwvuts zyxwvutsrq zyxwvutsrq zyxwvuts zyxwvutsrqpo where the uncertainties in the parameter measurements are defined as + Here A = (2.303/20)10k = 0.115(AT l ) , and 6f and 6s are the elementary errors in the resonant frequency measurement and the transmission coefficient measurements, respectively, with 7 1 = 0.1.2.3.. . .. The discrete character of the measurements causes a similar distribution of errors. Because there is no continuous spectrum of readings available, the error in the resonant frequency measurement h f may have a value of 6 f , / 2 or any multiple n thereof, where 6f, is the increment of the frequency change in the measuring system. The incremental error in the resonant frequency measurement was 0.025 MHz and 0.04 MHz, for the two modes, respectively, and the error in the transmission coefficient measurement 6S21 was evaluated experimentally for both modes of the cavity operation as 0.015 to 0.02 dB. The uncertainty of mass determination, as calculated from (6) and (7) for n l = 2 and 712 = 3, is approximately 2.5 mg for both modes of cavity operation. These values are consistent with those found previously 111 in which the uncertainty in moisture determination by this technique was less than 1 % moisture. They indicate that the uncertainty in the weighting of a single seed (average weight w l i 0 mg), &0.2 mg, has a negligible effect on the final accuracy of measurement. Thus, the mass of a single soybean seed may be determined with an uncertainty of 5 milligrams at the 95-percent confidence level nondestructively, without need for contact between the measuring system and the object. Similar possibilities have been demonstrated for other objects of organic as well as inorganic origin [31-[51. 111. CONCLUSIONS The moisture content and mass of single soybean seeds can be determined by simultaneous measurement of the frequency shift and change in the transmission characteristic of a resonant cavity at microwave frequencies when the seed is inserted into an empty cavity. Achieved uncertainties of less than 1% moisture content and 3% in the mass determination are sufficiently accurate for practical interest and potential development. Abstract-The properties of a CMOS-compatible pn-photodiode, phototransistor, and one-dimensional lateral-effect photodiode (LEP) for position-sensing applications are characterized. The photodiode and phototransistor seem to have properties that are comparable to typical commercial photodetectors despite the quite large variations in their spatial and spectral responses and the lower responsivity in the nearinfrared band. In addition to the above properties the LEP’s show excellent linearity, but 3-4 times larger NEP than corresponding commercial LEP’s due to low resistance of the current dividing layer. The responsivity variations have no effect on the linearity of the LEP, and the slightly lower responsivity at near-infrared has only a negligible effect on the achievable resolution (SNR). These properties, usually considered as weak points of CMOS-compatible photodetectors, are believed to have little or no effect on the properties of a position sensor, if the diameter of the light spot is small (< 100pm). CMOS-compatible photodetectors are therefore believed to be very suitable for industrial position-sensing applications. I. INTRODUCTION P OSITION-sensitive detectors (PSD’s) are widely used in optoelectronic instruments in industrial applications, which include measurements of movement, angle, straightness, object location, height, and centering. PSD’s in these instruments are applied for determining the position of a light spot on the focal plane of the receiving optics. Various types of photodetectors are used as PSD’s including bicell and quadrant detectors, lateral-effect photodiodes (LEP), and multielement arrays. In many applications, sensors with rugged construction, real-time response, high accuracy, small size, low power consumption, and simple signal processing are of the utmost importance. These include, for example, portable sensor systems for field applications. Standard CMOS technology provides one way of implementing such sensors. Various kinds of CMOScompatible optical sensors have been reported during recent years, including a single-chip video camera [l], a monolithic optical position encoder [2], and a two-dimensional visual tracking array [3], etc. [41-[71. Our goal is to develop a PSD with integrated signal processing for a portable optoelectronic instrument to be used in long-range outdoor applications. Therefore, in addition to the above-mentioned properties, the sensor should be as immune to air turbulences as possible in order to maintain the sensor inherent accuracy. This is ensured by using focused optics and a detector suitable for such focal plane processing. There are basically three possible detector types that can be used: a four-quadrant detector, a lateral-effect photodiode, and an area array detector. The need for small size, low power consumption and real-time operation excludes CCD-type detectors. Quadrant detectors have very high inherent sensitivity, but because they need to be defocused to obtain a certain measurement field, their resolution in a turbulent environment is severely impaired [8]. At the moment an LEP provides the best compactness and overall resolution due to its immunity to turbulences. An area array is believed to provide the optimal solution for our needs if small enough pixelto-pixel spacing can be implemented, because it is highly linear zyxwvutsrq REFERENCES A. W. Kraszewski, T.-S. You, and S. 0. Nelson, “Microwave resonator technique for moisture content determination in single soybean seeds,” IEEE Trans. Insfrum. Meas., vol. 38, no. 1, pp. 79-84, Feb. 1989. A. W. Kraszewski, S.0. Nelson, and T.-S. You, “Moisture content determination in single corn kernels by microwave resonator techniques,” J . Agric. Eng. Res., vol. 48, no. 1, pp. 77-87, 1991. A. W. Kraszewski and S. 0. Nelson, “Nondestructive microwave measurement of moisture content and mass of single peanut kernels,” Trans. ASAE, vol. 36, no. I , pp. 127-134, 1993. __. “Observations on resonant cavity perturbation by dielectric objects,” IEEE Trans. Microwave Theory Te;.hn.. vol. MTT-40, no. 1, pp. 151-155, Jan. 1992. -, “Resonant microwave cavities for sensing properties of agricultural products.” Trans. ASAE, vol. 35, no. 4, pp. 1315-1321, 1992. Manuscript received July 16, 1993. The authors are with the Electronics Laboratory, University of Oulu, Linnanmaa, SF-90570 Oulu, Finland. IEEE Log Number 9402235. zyxwvut 00 18-9456/94$04.00 0 I994 IEEE 490 zyxwvutsrqponmlkji zyxwvu IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 43, NO. 3, JUNE 1994 and insensitive to turbulences due to focusing and because moderate lateral resolutions can be achieved with much lower signal power than with an LEP. There are three obvious solutions to our problem: a CMOScompatible array-type detector with local computation such as proposed in papers [4] and [ 5 ] , for example; a commercially available linear (duolateral) LEP with a CMOS-ASIC for signal processing; or a CMOS-compatible tetralateral LEP the nonlinearity of which is minimized by using the anode geometry proposed in paper [9] with signal-processing electronics integrated on the same chip. As a first step towards an integrated position sensor, we have characterized the properties of some CMOS-compatible photodetectors which could be used as basic elements of the sensor chip, namely a pn-photodiode, a phototransistor, and an LEP. PD zyxwvutsrqponm zyxwvutsr zyxwvu ANODE ' 1 1 ' P+ I I~ 1 n P P1 P 11. DESCRIPTION OF THE IMPLEMENTED CMOS-COMPATIBLE PHOTODETECTORS . A. CMOS-Compatible Photodetectors LEP ANODE In a standard CMOS process there are basically two photodetector structures available: photodiode and phototransistor. The photodiode has the advantage of better linearity and faster response time whereas the phototransistor benefits from greater gain [ 3 ] .Two pn-junctions have been mainly utilized to implement photodetectors: the wellsubstrate and the well-diffused junctions. The well-substrate photodiode has the best responsivity due to a wide depletion region and because it is also able to collect the minority carriers photogenerated deeply in the substrate provided that they are generated within the diffusion length of the minority camers. The well-substrate photodiode also has the lowest capacitance, which helps to achieve a high bandwidth. The disadvantages of the well-substrate photodiode are its sensitivity to substrate noise and crosstalk from the neighboring photodiodes due to the long diffusion length of the camers [6], [7]. The CMOS-compatible phototransistors are based on vertical or lateral structure. The lateral transistor provides a higher , j , but has a complicated structure and large device-to-device variations, both of which can be detrimental to array-type detector implementation and performance. The vertical parasitic PNP bipolar transistor in the n-well CMOS process provides high compactness, moderate gain, and speed. Using a special layout, gains of over one hundred can be achieved within a large dynamic range with this structure [lo]. A lateral-effect photodiode is an analog photodetector which can provide continuous information about the position of a light spot on the detector's active surface. The LEP is a large photodiode with extended electrodes at the edges of a resistive layer. The position is derived by dividing carriers generated in the illuminated region between the electrodes in proportion to the conductance of the current paths between the illuminated region and the electrodes. Since the photon-generated current through each of the lateral contacts is a function only of the distance of the light spot centroid from these contacts, position information can be extracted by measuring the currents of the contacts. One way to construct a CMOS-compatible LEP is to use the basic well-substrate photodiode with appropriate end contacts in the well. This structure provides all the advantages that a well-substrate phototiode has, plus the highest possible resistivity for the current dividing layer (well). This means the lowest possible detector noise, because the thermal noise of the interelectrode resistance almost solely determines the noise level of the device. B. Implemented Photodetecmr. Structures CATHODE n+ \ P+ 1 CATHODE1 P+ CATHODES n+ D Fig. 1. Implemented CMOS-compatible photosensor structures: photodiode one-dimensional , lateral-effect photodiode (LEP). (PD), phototransistor (IT) polysilicon-two-metal n-well CMOS process. The operation of the detectors is based on the well-substrate junction. The substrate and the well with diffusion strapping form the anode and cathode of the photodiode, respectively. The cathode diffusion has narrow metal contact wires spaced every 75 pm. The anode contact is made to the p+ guard ring surrounding the detector. The total Si02 thickness above the Si is about 1.65 pni and the Si3N4 above that about 1 pin. The active area size of the photodiode is 0.5 x 0.5nim2. The phototransistor is a vertical, parasitic PNP-transistor having the same size and geometrical structure as the photodiode. The LEP is one-dimensional, and its active area dimensions are 1 x 0.13nini2. It uses the same structure as the photodiode except that there is diffusion only under the cathode end-contacts. The LEP was made one-dimensional, because the geometry is inherently linear (unlike two-dimensional tetralateral geometry), enabling us to better detect the amount of the technology-dependent nonlineanties caused by such factors as nonuniform spatial responsivity or nonuniformities in the current dividing layer, for example. The extemal connections of the photodiode and phototransistor were made by using standard pads including protection devices. The LEP was connected without protection devices. 111. EXPERIMENTAL RESULTS The experimental results are presented in Table I. According to the results, most of the measured properties of the photodiode and phototransistor are comparable with the commercially available detectors produced with tailored technologies. At 4.5 V reverse bias the capacitance was about 30 pF/nim2, which is sufficiently small to achieve speeds high enough for most applications. A 34 ns risetime was achieved by connecting the photodiode to a 150 MHz transimpedance amplifier. In this experiment, the speed was slowed down remarkably by the extra series resistance 1-2 kl2) of the ESD protection device, and the response time is believed to be only a few nanoseconds if the photodiode was connected directly to zyxwvuts The implemented photosensor structures are presented in Fig. 1 . The photosensors were realized using the standard 1.2 / / i i i one- zyx zyxwvutsrqp zyxwvuts 49 1 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 43, NO. 3, JUNE 1994 TABLE I EXPERIMENTAL RESULTS OF THE PHOTODIODE (PD), PHOTOTRANSISTOR (PT), AND LATERAL-EFFECT PHOTODIODE (LEP) zyxwvutsrqponml zyxwvutsrq zyxwvutsrqpon zyxwvutsr Property Response S Spectral variation of S Spatial variation of S Junction capacitance c, Dark current Id Shunt resistance Unit Comments PD PT A/W 0.47 % f20 X = 0.63 pm," ) A = 27.5 0.4 * ) 0.85pm - 0.6pm < X < 0.75 p m % f15 f8 LEP - - I pF 7 6 - r; pA 7 11 - L-, = 4.9 v MR 770 710 - UT = 0 NEP Interelectrode resistance v, ns pW/&z kQ 34 _ < 5 ns, t , > 100 ns 56 - t, - 2.5 X = 0.85pm,f = 1 kHz - 16.8 - - 900 1000 1100 zyxwvutsr 0.5 I % 600 700 800 WAVELENGTH [nm] Fig. 2. Spectral response of the photodiode. I 0.~1 R S Nonlinearity 500 400 300 = 4.5 v L't(ris) = 100 mV Rsh Risetime t,- 0.6 &0.2 90% of the active dimension i A/w 0.2 0 a transimpedance amplifier. The shunt resistance of several hundred Mn and the dark current of only 30 pA/mm2 suggest quite normal sensitivity provided that other noise sources such as substrate noise do not dominate. The most distinct differences between the tested CMOS photodetectors and detectors produced with tailored processes were found by examining the spectral and spatial responses. The average responsivity of the photodiode at 633 nm was 0.47 A/W, which is about the same as the commercial detecors have. The responsivity of the phototransistor, 27 Am, corresponds to the typical value of B achievable with the vertical transistor in this process. The spectral responsivity of the photodiode presented in Fig. 2 has a peak-to-peak ripple of about 1!~20%,in the visible band, and in the near-infrared the average responsivity is somewhat damped compared to a typical Si detector. The ripple is caused by incident light interference in passivation layers, and the low responsivity in the near-infrared is believed to be caused by the inefficient collection of the minority carriers photogenerated deep in the substrate [ 111. The spatial responsivity was measured by scanning the detector surface with a 50 pm light spot of a HeNe laser. The responsivity variations were *l5% for the diode (Fig. 3) and +8%. for the transistor. The changes are quite smooth if we compare them to the dimensions of a typical light spot (z100pm) used in position sensing, and therefore their contributions, to the nonlinearity of a position sensor are believed to be quite small, irrespective of the principle of the PSD implementation. The spatial variations are believed to be caused by passivation layer thickness variations which modulate the responsivity together with the interference effect. The properties of the LEP were evaluated by measuring the responsivity, noise, and linearity. The responsivity was measured using an infrared-emitting LED ( A = 850 nm) and was about 0.4 A m , which is slightly better than that measured for the photodiode (0.32 Am).The difference is probably due to the uncertainty caused by the ripple. The noise equivalent power (NEP) calculated on the basis of the measured noise and responsivity was about 2.5 pW/JHz ( A = 850 nm), which is 3 4 times larger than that of a typical commercial LEP. This is mainly due to the quite low 50 100 micrometre O0 450 Fig. 3. Spatial response of the photodiode. r- -2 i -3 -0.5 -0.4 zyxwvutsrqpo -0.3 -0.2 -0.1 0 0.1 0.2 0.3 POSITION OF THE LIGHT SPOT [mm] 0.4 0.5 Fig. 4. Nonlinearity of the lateral-effect photodiode. resistivity (2.4kR/D) of the well acting as the current dividing layer. The linearity of the detector was measured using a light spot the diameter of which was about 130 pm. The measured peak-topeak nonlinearity error was only &0.2'% when 90% of the active dimension was used; this clearly indicates that CMOS-compatible LEP's are suitable for high-accuracy position sensing (Fig. 4). IV. CONCLUSIONS We have characterized the properties of a CMOS-compatible pnphotodiode, phototransistor, and LEP which could be used as basic elements of a new position sensor chip and found them very suitable for our purpose. According to the experimental results the CMOScompatible photodiode and phototransistor seem to have properties that are comparable with typical silicon photodetectors fabricated with special technologies despite the quite large variations in spatial and spectral responses and the lower responsivity in the near- zyxwvut 492 zyxwvutsrqponmlkjihg zyxwvutsrqp IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 43, NO. 3, JUNE 1994 infrared band. However, in such position-sensing applications, where the light spot diameter has to be small (< 100pm), the spatial responsivity variations have little effect on the performance of the position sensor. This can be clearly seen from the excellent linearity of the one-dimensional LEP. The spectral responsivity variations affect only slightly the NEP of the detector which in many cases can be neglected. The NEP of the LEP is 3-4 times larger than that of the commercially available detectors, but this is mainly due to the low resistivity of the well. To be used in high-accuracy position-sensing applications, the CMOS-compatible LEPs should be implemented using processes with high (> 10kQ) well resistivities, or the properties of some new structures like the pinched well, for example, should be examined as the current dividing layer. In order to be suitable for two-dimensional applications the possibilities of minimizing the quite large inherent nonlinearity of the tetralateral geometry should be studied. Angular Range Extension of an Analog Integrated Circuit Synthesizing Trigonometric Functions Stephanus J. Spammer, Pieter L. Swart, and Johan Meyer Abstract-The fixed scale factors of analog integrated circuit trigonometric function converters limit their maximum input angular range. A method to increase the total input span is described. Experimental results demonstrate that the input range can be extended easily from the maximum value of f 3 n to f l G . r r . By changing the hardware resolution, it may even be increased further. Factors which determine the angular resolution and linearity of the improved function converter are discussed. I. INTRODUCTION T RIGONOMETRIC function generators are typically used in analog computation applications such as polar-to-rectangular and rectangular-to-polar conversions [ 11, [2]. Other applications of these components include the generation of quadrature signals used in phase-tracking circuits for open-loop optical-fiber gyroscopes [3], [4]. One method of generating trigonometric functions is by ROM (read-only memory) look-up tables in conjunction with D/A (digitalto-analog) and A D (analog-to-digital) converters. Another method of synthesizing trigonometric functions involves analog integrated circuits such as the AD 639 made by Analog Devices. The same trigonometric function converter can easily be used to generate sine, consine, and tangent functions and their inverses. With an ambient scale factor of 50”/V, the maximum input rating of the AD 639 is &12 V or k600” [2]. This poses a limitation on applications which require the range to exceed *600”. One way of increasing the maximum range of these components is to subtract from or to add to the input a specific value. This can be done with analog components, but they are prone to drift with temperature, or the output may become unstable at transition levels of comparators, for instance. This paper introduces an easily realizable mapping technique surmounting the deficiency of limited input range. The method involves the addition of an A/D and D/A converter which enables one to subtract quantities equal to the equivalent of multiples of 27r from the input signal of the AD 639. A theoretical model of the system which is verified experimentally is introduced. zyxwvutsrqponml zyxwvutsrqpon REFERENCES D. Renshaw, P. B. Denyer, G. Wang, and M. Lu, “Asic image sensors,” in Proc. IEEE Int. Symp. Circuits and Systems, New Orleans, LA, 1990, pp. 3038-3041. P. Aubert, H. Oguey, and R. Vuilleumier, “Monolithic optical position encoder with on-chip photodiodes,” IEEE J . Solid-State Circuits, vol. 23, pp. 465473, Apr. 1988. S. P. DeWeerth and C. A. Mead, “A two-dimensional visual tracking array,” in Proc. MIT Conf. VLSI. Cambridge, MA: M.I.T. Press, 1988, pp. 259-275. D. L. Stanley, “An object position and orientation IC with embedded imager,” IEEE J . Solid-State Circuits, vol. 26, pp. 1853-1859, Dec. 1991. W. R. Gonnason, J. W. Haslett, and F. N. Trofimenkoff, “A low cost high resolution optical position sensor,” IEEE Trans. Instrum. Meas., vol. 39, pp. 658-663, Aug. 1990. A. GNSS, L. R. Carley, and T. Kanade, “Integrated sensor and rangefinding analog signal processor,” IEEE J . Solid-Srare Circuits, vol. 26, pp. 18L191, Mar. 1991. 0. Yadid-Pecht, R. Ginosar, and Y. S. Diamand, “A random access photodiode array for intelligent image capture,” IEEE Trans. Electron Devices, vol. 38, pp. 1772-1780, Aug. 1991. A. M&ynen, J. Kostamovaara, and R. Myllyla, “Small angle measurement in a turbulent environment using position-sensitive detectors,” in Engineering Systems with Intelligence, S. G. Tzafestas, Ed. Dordrecht, The Netherlands: Kluwer, 1991, pp. 275-284. Y. Terada and K. Yamamoto, “An improvement on two-dimensional position sensitive semiconductor detector,” Japan. J . Optics, vol. 12, no. 5 , pp. 367-373, 1983. M. P. Vidal, M. Bafleur, J. Buxo, and G . Sarrabayrouse, “A bipolar photodetector compatible with standard CMOS technology,” Solid-Stare Electron., vol. 34, pp. 809-814, Aug. 1991. G. Soncini, M. Zen, M. Rudan, and G. Verzellesi, “On the electrooptical characteristics of CMOS compatible photodiodes,” in Conf. Rec. Melecon ’91, Ljubjana, Yugoslavia, May 1991. zyxwvu zyxwvutsr zyx 11. THEORY Consider a trigonometric function converter with input range [ - m 7 r , m x ] , where m = 1, 2 or 3. The output obtained can be any one of the trigonometric functions sine, cosine, tangent or their corresponding inverses. Consider an independent variable y with domain [ - hh]. . where I; is an integer which is equal to or larger than m . We want to map the variable y into a variable ?/ with a domain which is identical to the input range of the trigonometric function converter. The range of the transformed variable Q obtained by the desired mapping of y is a subset of the set [-h, k7r] given by [-m7r. m7r] where m 5 I;. The mapping Q = [y +m 7r signum ( y)] modulo (27rm) - m 7r signum ( y ) (1) Manuscript received January 5 , 1993; revised October 5 , 1993. The authors are with the Sensors Sources and Signal Processing Research Group, Faculty of Engineering, Rand Afrikaans University, Auckland Park 2006, South Africa. IEEE Log Number 9302065. zyxwvut 0018-9456/94$04.00 0 1994 IEEE