CHAPTER 4 EXPERIMENTAL RESULTS
4.3 DISCUSSIONS
In Section 4.2, calculated velocity is obtained from correlation outputs accumulated over 11 frames. This is because patterns make positive contributions to shear motion when they translate from the left side of chip to the center, and make negative contributions when translating from center to the right side. The chip is 11 pixels wide from the left side to the right and thus every computation and simulation is done over 11 frames. Measurement results of Fig. 4.6 (b) to 4.8 (b) ensure this idea.
As mentioned in Section 2.2, during shifting of the previous image frame to compute ΣC(shear) and ΣC(rshear), the boundary pixels are fixed at zero for simplicity.
The boundary condition, however, does not influence ΣC(self) because of no shifting in the calculation of ΣC(self). To calibrate this boundary condition, two factors b1 and b2
are added to calculate displacement during measurement. The value of b1 is chosen to be half the number of pixels at one side of the boundary while b2 is half of b1. In this design, b1is 8 and b2 is 4.
p q
a b
a/M b/M
Fig. 4.9 A diagram used to explain the ratio of the object velocity to the image velocity
Lens is used in the setup of measurement to focus the test patterns onto the chip.
actual velocity on the chip. Fig.4.9 is a diagram used to find out the relationship between the velocity of test pattern and image on the chip. The object moves at the distant p from the lens and the image is formed on the plane which is q away from the lens. The object is length a at the time t and becomes length b at the time t+∆t. The velocity of object is obviously (b-a)/∆t. According to the theory of geometric optics [15], the lateral magnification of a thin lens is
M q
= − p (4.2)
where q is the image distance while p is the object distance as shown in Fig. 4.9.
Therefore, the image at time t is length a/M and becomes b/M after tie interval ∆t. The velocity of the image is thus [(b-a)/M]/∆t. The ratio of the object velocity to the image velocity is obtained by equation (4.2). In the case of measurement, p is about 83 cm and q is 16 cm. The magnification is approximately 5.1.
In this chapter, the measurement is focused on the first experimental chip with off-chip computations. The other one where the core circuit combined with the motion computation circuit is not tested. This part will be done in the future. A summary of the first chip is given in Table 4.1. Frame rate is limited at the low frequency because the software of computer in the measurement setup cannot speedup. This also results in the limitation of testing velocities. Finding more suitable instrument to display test patterns will help to solve this problem.
Table 4.1 The summary on the characteristics of the first fabricated chip of shear motion sensor
Technology 0.35µm CMOS 2P4M
Sampling geometry Shear motion
Pixel number 92 pixels
Pixel size 52.96 µm x 55.07 µm
Fill factor 16%
Chip size 1100 µm x 1100 µm
Clock rate 490Hz
Frame rate 3Hz
Testing velocities 0.06 mm/sec, 0.09 mm/sec
and 0.18 mm/sec
Deviation of testing velocity <±10%
Power dissipation @ 3V ~ 3 mW in dark
(except the off-chip counter used to control the decoder)
CHAPTER 5
CONCLUSIONS AND FUTURE WORKS
5.1 MAIN RESULTS OF THIS THESIS
A first shear motion sensor has been proposed, designed and tested successfully. In the proposed shear sensor, it uses the pseudo-BJT-based retinal processing circuit with adaptive current Schmitt trigger to achieve image acquisition and thus has advantages of high dynamic range, edge enhancement, and robust noise immunity. The correlation-based algorithm with modification which is inspired by a biological model has the benefits of robustness and compactness and is adopted to calculate the velocity.
With the specific pixel arrangement along with the modified correlation-based algorithm, the shear motion can be detected and other motions are dismissed. Two experimental shear motion sensors were fabricated in a 0.35 µm double-poly-quadruple-metal CMOS process. One contains only the core circuit part, and the other combines the core circuit and the motion computation circuit. The area of the first chip is 1100 x 1100 µm2 and the other is 2036 x 1765 µm2. The area of a single pixel is 52.96 x 55.07 µm2 with a fill factor of 16%. Using the first chip along with off-chip computation, shear motion is tested under 0.06 mm/sec, 0.09 mm/sec and 0.18 mm/sec and the velocity deviations are less than ± 10%. The shear motion sensor is also tested by three translating patterns and the shear motion selectivity is thus verified. The DC power dissipation is 3mW at 3V in the dark.
5.2 FUTURE WORKS
The proposed shear motion sensor is intended to detect the global shear motion as well as dismiss other global motions. The experimental results have proved the functionalities and the advantages, which includes compactness, accuracy, robustness, and real-time. To complete the shear motion sensor, the second chip is implemented and needs to be tested. The fill factor of photodiode and the area of the single pixel can be further improved to make the design more practical in the future application. The pixel array will be improved by using the upper level metal to cover the photodiodes which are not in the shear motion tracks. Therefore, the pixels can be arranged in the matrix manner and the structure will become more compact.
The relative motion between an observer and its surroundings and the 3D structure of the surroundings can be described by translation, rotation, expansion/contraction, and shear motion. In this thesis, a shear motion sensor is proposed and implemented. In the future, the detection of translation, rotation, and expansion may be integrated with this shear motion sensor to moreover analyze the 3D motion.
Besides, this proposed sensor is deal with the detection of relative motion between an observer and its surroundings. If there are multiple objects moving in different velocities and directions, the proposed sensor cannot cope with this condition. The detection of multiple moving objects can be furthermore investigated hereafter.
REFERENCES
[1] C. Y. Wu and K.H. Huang, “A CMOS focal-plane motion sensor with BJT-based retinal smoothing network and modified correlation-based algorithm,” IEEE Sensors Journal, vol. 2, no. 6, pp. 549-558, December 2002.
[2] O. Schrey, R. Hauschild, B. J. Hosticka, U. Iurgel and M. Schwarz, “A locally adaptive CMOS image sensor with 90dB dynamic range,” ISSCC Digest of Technical Papers, pp. 310-311, 1999.
[3] K. Boahen and A. Andreou, “A contrast-sensitive retina with reciprocal synapses,” Advance in Neural Information Processing 4, J. E. Moody, ed., Morgan Kaufman, 1991.
[4] C. A. Mead, Analog VLSI and Neural Systems, MA: Addison Wesley, 1989.
[5] F. Pardo, B. Dierickx, and D. Scheffer, “CMOS foveated sensor: signal scaling and small geometry effects,” IEEE Transaction on Electron Devices, vol. 44, no. 10, October 1997.
[6] J. J. Konederink, “Optical flow, ” Vision Research, no. 26, pp. 161-180, 1986.
[7] G. A. Orban and H. –H Nagel (eds.), Artificial and Biological Vision Systems, Berlin: Springer-Verlag, 1992.
[8] K.H. Huang, L. J. Lin, and C. Y. Wu, “Analysis and design of a CMOS angular velocity- and direction-selective rotation sensor with a retinal processing circuit,” IEEE Sensors Journal, vol. 4, no. 6, pp. 845-856, December 2004.
[9] C. A. Mead, “Adaptive Retina,” Analog VLSI Implementations and Neural Systems, pp. 239-246, 1989.
[10] A. G. Andreou and K. A. Boahen, “A 48,000 pixel, 590,000 transistor silicon retina in current-mode subthresold CMOS,” in Proceedings of the 37th Midwest Symposium on Circuits and Systems, vol. 1, pp. 97-102, 1994.
[11] C. Y. Wu and H. C. Jiang, “An improved BJT-based silicon retina with tunable image smoothing capability,” IEEE Transactions on VLSI Systems, vol. 7, no.
2, pp.241-248, June 1999.
[12] C. Y. Wu, H. C. Huang, L. J. Lin, and K. H. Huang, “A new Pseudo-Bipolar-Junction-Transistor (PBJT) and its application in the design of retinal smoothing network,” in Proc. IEEE International Symposium on Circuits and Systems, vol. 4, pp. 125-128, July 2002.
[13] E. R. Fossum, “Architectures for focal-plane image processing,” Optical Engineering, vol. 28, no. 8, pp. 866–871, 1989.
[14] C. Y. Wu and C. T. Chiang, “A low-photocurrent CMOS retinal focal-plane sensor with pseudo-BJT smoothing network and adaptive current Schmitt trigger for scanner applications,” IEEE Sensors Journal, vol. 4, no. 4, pp.
510-518, August 2004.
[15] Raymond A. Serway, Physics for Scientists & Engineers with Modern Physics, Saunders College Publishing, 1996.
VITA
姓 名 ： 謝 文 芩 性 別 ： 女
出生日期 ： 民國70年7月1日 出 生 地 ： 台灣省新竹市
住 址 ： 新竹市博愛街43巷16號4樓 學 歷 :
國立交通大學電子工程學系畢業 (88年9月−92年6月)
國立交通大學電子研究所碩士班 (92年9月−94年6月)
論文名稱 : 具虛擬式二極體仿視網膜處理電路之焦平面式 切變運動感測器設計
THE DESIGN OF A CMOS FOCAL-PLANE SHEAR MOTION SENSOR WITH PSEUDO BJT-BASED RETINAL PROCESSING CIRCUITS