This paper presents an indirect time-of-flight (TOF) measurement technique with an impulse photocurrent response of a lock-in pixel. By using a short-pulse laser, the generated photocurrent can be presumed to be an impulse response. This facilitates the utilization of the full high-speed performance of the photodetector and gives high range resolution. As a proof-of-concept, a test chip with a lock-in pixel based on draining-only modulation was implemented using 0.11 μm CMOS image-sensor technology. The test chip achieved a range resolution of 0.29 mm in a 50-mm measurable range, which corresponds to a time resolution of 1.9 ps and the successful acquisition of a 3-mm example step.
© 2014 Optical Society of America
Indirect time-of-flight (TOF) range measurement techniques using two-dimensional CMOS lock-in pixel imagers [1–4] are useful for implementing a small range measurement module suitable for three-dimensional (3D) scanner applications. However, the range resolution of conventional TOF range imagers is limited to several millimeters in a measurable range of a few meters. For a contact-less 3D scanner, sub-millimeter or higher range resolution is necessary.
The reported indirect TOF range imagers can be classified according to the light source: continuous wave [1,2] or pulsed illumination, as [3,4] shown in Fig. 1. The range resolution of those methods is inversely proportional to the square root of the number of signal electrons. In addition, it is inversely proportional to the modulation frequency, fmod, for the former method, and proportional to the width, TP, of a light source for the latter. For high resolution, the value of 1/fmod or TP should be chosen to be as small as possible. Nevertheless, in conventional techniques, these parameters are chosen to be much larger than the photocurrent response, τ, of a lock-in pixel, in order to maintain the range accuracy.
This paper introduces a new indirect TOF measurement technique that employs an impulse photocurrent response . By using a light source with a very short pulse, i.e., TP << τ, the photocurrent response can be regarded as an impulse response. As a result, the range resolution is determined only by the photocurrent response, which facilitates the utilization of the full high-speed capability of the lock-in pixel. The high-speed photocurrent response can be achieved through draining-only modulation (DOM) lock-in pixels [5,6]. A prototype range imager that employs the indirect TOF measurement technique, in combination with a column-wise skew calibration scheme, has been implemented and described in , but that paper does not provide the details of TOF calculation and lock-in pixel design. In this paper, we discuss the fundamental principle of the proposed TOF measurement technique and the design of high-speed lock-in pixels with 3D device simulations. Experimental results with a test-fabricated chip demonstrate the capability for intensity-independent TOF measurement and high range resolution.
The remainder of this paper is organized as follows. In Section II, the principle of the proposed TOF measurement technique is described. Section III presents the implementation of a test chip and a DOM detector based on lock-in pixels. Section IV gives the experimental results obtained with the test chip. Section V provides a concluding summary.
2. Proposed TOF measurement principle
Figure 2 shows the timing diagram of the proposed TOF measurement principle. Instead of a squared or sinusoidal light source, a short-pulse laser with about a hundred picoseconds is used in the proposed method.
The design of the lock-in pixel including DOM detectors [5, 6] is shown in Fig. 3. In the DOM detector, a draining gate (TD) is formed along the channel, which controls the channel potential, in order to perform lock-in detection. There are no transfer gates in the signal path, which facilitates high-speed modulation and loss-less accumulation.
For the TOF measurement, a light pulse is emitted to a target object, and the reflected light generates a photocurrent, Iph, in a photodiode. (The response time associated with the photocurrent is assumed to be much larger than the light-pulse width.) The time of flight, ttof, is then measured by the response time, τ0, associated with the photocurrent. The response of the photocurrent is assumed to be linear with respect to time, and is expressed asFig. 2.
For Toffset − τ0 < ttof ≤ Toffset, accumulated charges, N1, N2 and N3, which are generated by the time windows TW(1), TW(2), and TW(3), respectively, are given by
In the preceding analysis, a background light was taken into account for generality. In our current target, however, the laser power is supposed to be much higher than that of the background light. For this reason, the background light is assumed to be negligible in the following discussions. If the shot noise is dominant and the background light is zero, the range resolution, σL, is given by
Although the photocurrent response is assumed to be linear in time, the actual response is more complicated and can be modeled only by a high-order equation. Such an equation can be used for the range measurement, but this is beyond the scope of this paper.
Figure 4 shows the operation of a pixel. Since the DOM pixel produces a signal for a single time window, three measurements and readouts for N1, N2, and N3, with three different time windows, TW(1), TW(2), and TW(3), respectively, were carried out in the experiments.
With the proposed technique, the measurable range is limited to several centimeters. Nevertheless, this limited range is sufficient for many applications in contactless 3D scanners. Such scanners are often used in combination with a robot arm such as [8, 9], which is commonly used in coordinate measuring machines (CMMs), and in such machines, a robot arm has its own position sensors that would compensate for the limited range of the proposed TOF range imagers.
For a proof-of-concept, a test chip with lock-in pixels based on DOM detectors was fabricated using a 0.11-μm CMOS image sensor (CIS) technology. As shown in Fig. 3, the implemented pixel consists of DOM detectors, a reset transistor (RST), a row select transistor (SEL), and an in-pixel buffer. The DOM detector consists of a pinned photodiode (PD), a TD gate, and a floating diffusion (FD). The channel potential is modulated by the lateral electric field due to the TD gate bias, which enables lock-in detection.
When the TD gate is closed, through the application of a low gate bias, all electrons that are generated at the PD are transferred and stored in the FD. This occurs within the time window. When the TD gate is opened, the generated electrons are drained out, and the photocurrent in this phase does not contribute to the signal stored in the FD. So this operation effectively realizes a virtual switch between the PD and FD, as shown in Fig. 3. The DOM detector does not have any transfer gate in the signal path, which facilitates high-speed charge modulation and loss-less repetitive accumulation . In the read phase, an SEL is first activated and the signal level is then read out. After the FD is reset, the reset level is read out. To extract the photogenerated signal, the difference between the two levels is taken at the peripheral circuits.
To shorten the charge-transfer path while maintaining sensitivity, the DOM detector is designed to have a very small unit size (2.8 × 2.8 μm2), and 48 DOM detectors are implemented in a pixel, all of which are connected with each other. The pixel size is 22.4 × 22.4 μm2. The rest of the pixel area is filled with MOS capacitors and in-pixel circuits. The fill factor of a DOM unit is 12%, which is 9% at the pixel level. The fill factor does not take into account a light focus with microlenses.
To exploit the high-speed capability of the DOM detector, the layout should be optimized to create a large lateral electric field. Figure 5(a) shows the layout of DOM detectors with triangular and rectangular PDs. Figure 5(b) shows the corresponding simulated potential and electric field distributions in the channels (PD-FD) when the TD gates are closed. With the rectangular PD, the electric field becomes very low at the X position of 1.0 to 2.0μm. In the triangular PD, since the width of the PD gradually increases, a high-drift field is created in the entire region of the PD [10, 11]. Figure 5(c) shows simulation results of the transient photocurrent response to a light pulse of 100 ps width for the rectangular and triangular layouts of the PD. In the simulations, the wavelength of incident light was set to 440 nm. As shown in Fig. 5(c), the photocurrent response in the triangular PD is much higher than that in the rectangular PD.
Figure 6 shows the simulated photocurrent responses for incident lights of 440 nm and 870 nm wavelengths. Each of the photocurrents shown has been normalized by the maximum value of the corresponding photocurrent. Recent TOF range cameras commonly utilize a near-infrared (NIR) light source because of its invisibility to the human eye. NIR light is, however, absorbed in the deep substrate region and so gives a slow photocurrent response. Since a fast photocurrent response yields a high range resolution, in the proposed TOF measurement technique, a wavelength as short as 440 nm is desired.
4. Experimental results
Figure 7 shows the results of the linearity and resolution of range measurements with the implemented TOF detector chip. In the measurement setup, the sensor board provides a laser-emission trigger signal to a 443 nm laser via a digital-delay generator (DDG), which can accurately control the delay time of the laser trigger. Since a change in delay time is equivalent to that in the target distance, the horizontal axis is expressed in terms of equivalent target distance. The repetitive frequency and the number of accumulations are set to 7.5 MHz and 4×104, respectively. The estimated distance is calculated using Eq. (5), without any corrections. The non-linearity is below ±2.5%FS for a 50-mm range. The range resolution is less than 0.29 mm, which corresponds to a time resolution of 1.9 ps.
Figure 8 shows normalized pixel outputs, N1, N2, and N3, in terms of the trigger delay, together with the differential of N2, which is equivalent to the photocurrent response. The modulation contrast is 85%. Although the estimated photocurrent response is different from that of the linear model of Fig. 2, the linear region in the measured photocurrent response can be used for the TOF calculation.
Figure 9 shows the measured signal-charge ratio, r, of Eq. (5) as a function of N1. In this measurement, a variable neutral density (ND) filter was included in the measurement setup of Fig. 7. N1 was swept by changing the attenuation rate of the ND filter while keeping the point of trigger delay, with the measurement repeated for every five points of trigger delay, i.e., Delay 1st to 5th. As expressed by Eq. (5), the change of distance is estimated only from r because τ0 and Toffset are independent of the object distance. As shown in Fig. 9, the value of r is independent of N1 in the available range. Therefore, the proposed TOF calculation does not depend on the light echo intensity. For the region outside the available range, the signal is limited to the readout noise or saturated, thus causing a change in the r value.
Figure 10 shows measured results of a one-axis range profile of a mirror target having 3-mm step height. The profile was measured using a mechanical stage to move the target, and the figure shows the measured one-axis profiles, using a single measurement, and the average of 10 measurements. This result shows that the proposed TOF measurement technique is effective for 3D range mapping with sub-millimeter resolution.
When a white noise such as shot noise is dominant, the range precision can be improved by taking the square root of the averaging number. In our measurement system, however, a large 1/f noise is observed, which degrades the precision improvement by averaging.
In this paper, a new TOF measurement technique for sub-millimeter or higher range resolution by means of very short light pulse has been presented. In the technique, the TOF calculation is determined only by a photocurrent response, and not by the light-pulse width. The use of DOM pixels provides high-speed charge modulation and loss-less accumulation. The test chip demonstrates the high range resolution of 0.29 mm at 50-mm measurable range; a one-axis depth profile with a 3-mm step was successfully acquired.
The authors thank M. Fukuda for his support in the measurements and T. Akahori and Y. Kaneko, Brookman Technology Inc., for their help in the chip design. This study was partly supported by the Grant-in-Aid for Scientific Research (S), No. 25220905 of the Ministry of Education, Science, Sports and Culture (MEXT), supported by the Center of Innovation Program from Japan Science and Technology Agency, JST, and supported by VLSI Design and Education Center (VDEC), the University of Tokyo in collaboration with Cadence Design Systems, Inc., and Mentor Graphics, Inc.
References and links
1. R. Lange and P. Seitz, “Solid-state time-of-flight range camera,” IEEE J. Quantum Electron. 37(3), 390–397 (2001). [CrossRef]
2. D. Stoppa, N. Massari, L. Pancheri, M. Malfatti, M. Perenzoni, and L. Gonzo, “A range image sensor based on 10-μm lock-in pixels in 0.18-μm CMOS imaging technology,IEEE J. Solid-State Circuits 46(1), 248–258 (2011). [CrossRef]
3. S. Kawahito, I. A. Halin, T. Ushinaga, T. Sawada, M. Homma, and Y. Maeda, “A CMOS time-of-flight range image sensor with gates-on-field-oxide structure,” IEEE Sens. J. 7(12), 1578–1586 (2007). [CrossRef]
4. S.-J. Kim, D. K. Kim, B. Kang, and K. Lee, “A CMOS image sensor based on unified pixel architecture with time-division multiplexing scheme for color and depth image acquisition,” IEEE J. Solid-State Circuits 47(11), 2834–2845 (2012). [CrossRef]
5. K. Yasutomi, T. Usui, S. Han, M. Kodama, T. Takasawa, K. Kagawa, and S. Kawahito, “A time-of-flight image sensor with sub-mm resolution using draining only modulation pixels,” in Proc. 2013 Int. Image Sensor Workshop (2013), pp. 357–360.
6. Z. Li, S. Kawahito, K. Yasutomi, K. Kagawa, J. Ukon, M. Hashimoto, and H. Niioka, “A time-resolved CMOS image sensor with draining-only modulation pixels for fluorescence lifetime imaging,” IEEE Trans. Electron. Devices 59(10), 2715–2722 (2012). [CrossRef]
7. K. Yasutomi, T. Usui, S. Han, T. Takasawa, K. Kagawa, and S. Kawahito, “A 0.3mm-resolution time-of-flight CMOS range imager with column-gating clock-skew calibration,” in Dig. Tech. Papers, IEEE Int. Solid-State Circuits Conf. (ISSCC) (2014), pp. 132–133.
8. VECTORON, Kosaka Laboratory Ltd., http://www.kosakalab.co.jp/english/product/precision/3_dimensions/
9. FaroArm, FARO Technologies Inc., http://www.faro.com/
10. C. Tubert, L. Simony, F. Roy, A. Tournier, L. Pinzelli, and P. Magnan, “High speed dual port pinned-photodiode for time-of-flight imaging,” in Proc. 2009 Int. Image Sensor Workshop (2009), pp. 357–360.
11. H. Takeshita, T. Sawada, T. Iida, K. Yasutomi, and S. Kawahito, “High-speed charge transfer pinned-photodiode for a CMOS time-of-flight range image sensor,” Proc. SPIE 7536, 75360R (2010). [CrossRef]