Precision nano-alignment system using machine vision with motion controlled by piezoelectric motor☆
Introduction
Demand for lighter, thinner, smaller and more compact electronic devices at low manufacturing cost has rendered chips that can only detect energy flow inadequate. Chips are expected to perform integrated optical-electro-mechanical functions. Hence, micro-mechanical machining aims to incorporate various functions in the minute chips, thus enhancing its value and reducing production cost. In view of the minute dimensions of chips, the line width, line distance and hole radius must be minimized in order to meet the expectation of multi-function performance. Hence, nano-technology will be a preferred choice of technology for semi-conductor manufacturing and manual operations for repeated alignment process are bound to be replaced by automated systems for greater precision and higher efficiency.
To manufacturers of electronic devices, productivity has a key role to play in the choice of technology adopted in the production. Nevertheless, quality in terms of high precision is also a prime concern. Besides semi-conductor manufacturing, the electronic industry is the other sector that incorporate machine vision in its production process, in particular the automatic alignment operation. With vision-aided automatic alignment technology, high-quality electronic devices with good precision can be produced efficiently, thus reducing the cost of production and enhancing the competitiveness of the industry.
In the literature, there have been abundant studies on automatic alignment operation with image feedback. Considering 2D rigid body transformation with size changes, Lai and Fang [1] have proposed a hybrid image alignment system, known as the Fast Localization with Advanced Search Hierarchy (FLASH) algorithm for image comparison to detect deviations in pattern localization. Although the alignment system can achieve both accuracy and efficiency, little has been said on how to apply the difference obtained to the positioning system or how the deviation can be corrected to obtain precise alignment. Lin and Lue [2] have also developed an image system for fast positioning and accuracy inspection of ball grid array (BGA) type printed circuit boards (PCB). For the wafer dicing process, Kim et al. [3] derived a two-step automatic alignment algorithm from inspection data and geometric relations on machine coordinate for detecting deviations from standard positions. According to the calculated compensation values, movements are made to minimize errors in position and orientation, thus achieving precise alignment. An analytic algorithm is employed by Nian and Tarng [4] to calculate the relationship between the rotational center (original position) of the three-axis motion control mechanism and the relative coordinates obtained by two CCD cameras. With the Δx, Δy and Δθ determined, the automatic motion control mechanism can make rotational and translational movements along the three axes (X, Y1 and Y2) to achieve fast, efficient and accurate alignment. A general model for automatic alignment in 2D space is developed by Kim et al. [5] under the assumption of precise positioning with visual inspection. The proposed formula had a simple matrix form with motions described by modified rigid body transformation and was applied to a semi-automatic dicing machine. Though experimental results show that the algorithm was valid, the derivation of the matrix form was time-consuming. To achieve alignment with greater efficiency and precision, Nian et al. [6] developed a new algorithm that makes uses of the correlation between the input signals and output feedback. Through matrix transformation, the algorithm establishes the characteristic matrix of the system function between input and output signals. Though rapid and accurate alignment can be achieved with the motion control mechanism, the precision remains at the micrometer scale. To overcome such drawback, this study attempts to couple the motion control mechanism with the piezoelectric motor. Piezoelectric motors have the advantages of high resolution, compact size, light weight, rapid response, high output, low power consumption as well as fast and accurate positioning capability. In particular, piezoelectric ceramic motors exhibit converse piezoelectric effect, making use of friction force for driving the compensation movements. This can avoid the backlash problem due to the use of traditional ball screw device. In view of their many advantageous features, piezoelectric motors have been widely applied to the positioning stage, and when coupled with robust auto-alignment algorithm, they can achieve high speed, high-precision positioning, thus enhancing the performance and consistency of repeated auto-alignment at nanometer scale [7].
In actual practice, automated manufacturing process requires repeated positioning of the objects to be aligned. However, there always exist some small but non-negligible discrepancies between the feed-in positions of the object and the target position. In this study, an automatic nano-alignment system is developed using machine vision inspection. Charge coupled device (CCD) cameras are employed to define the center of the fiducial mark in relation to the original coordinates, thus giving positional reference for motion control to achieve alignment. The auto-alignment algorithm used has its basis on system identification, which provides real-time reference commands. Information on the movement of the fiducial images captured by the CCD cameras serves as important feedback to the system. Through image processing and matrix transformation, the characteristic matrix for the system function between input and output signals can be established. With the matrix established when the object to be aligned falls within the field of view (FOV) of the CCD, positional variation between the fiducial mark and original coordinates can be calculated through matrix transformation. Then according to the difference obtained, the system can issue commands to the two-axis piezoelectric motor to make movements for compensating the above-mentioned difference, thus achieving quick and precise alignment.
Extracting high-quality features from the image captured, precise centering of the fiducial marks and accurate calculation of their centers all play significant roles in enhancing alignment accuracy. Therefore, various image-processing techniques are applied. For example, binary morphology [8] is employed to extract better and more precise features of the captured image and the best fit circle [9] is utilized to ensure accurate location of the fiducial mark center.
The rest of the paper is organized as follows. Section 2 describes the structure and driving principle of the piezoelectric motor. The control algorithm is presented in Section 3. The auto-alignment operations and algorithms involved are detailed in Section 4. Section 5 presents the application of image-processing techniques. Section 6 describes the experiment conducted to test the proposed design. A discussion on the alignments results obtained is included in Section 7. Finally, Section 8 contains the conclusion.
Section snippets
Structure and driving principle of piezoelectric motor
Piezoelectric materials exhibit two basic phenomena, namely direct and converse piezoelectric effects, permitting them to be used as sensors and motors in a control system [10]. Direct piezoelectric effect refers to the induction of electric charge or voltage in the piezoelectric material when mechanical force or pressure is applied on it. On the other hand, converse piezoelectric effect describes the generation of mechanical force and strain in the piezoelectric material when electrical charge
Control algorithm
The linear piezoelectric ceramic motor (LPCM) in this study has merits of high precision, compact size, lightweight, great torque, quick response and absence of electromagnetic interference [11]. However, they also have drawbacks of serious hysteresis behavior and highly nonlinear property, which are difficult to overcome using conventional control strategy [12]. Therefore, a rule-based control algorithm is employed to resolve the nonlinear and random problems of the LPCM in this study and
Auto-alignment algorithm
Using system identification techniques and the correlation between the input and real-time output, the proposed algorithm establishes the characteristic matrix of the system function. The information thus obtained will enable the two-axis piezoelectric motor stage discussed in Section 2 to perform alignment operation. The algorithms involved in the alignment process are as follows.
As seen in Fig. 10, P1 is the target position of CCD1. The target position is transferred from the standard
Image-processing techniques
The fiducial marks extracted from the digital image captured by the CCD provide reference coordinates for shape matching. Alignment accuracy of these reference coordinates would affect the precision of the whole alignment system. To ensure accurate matching, we employ the library of various image-processing algorithms of Vision Builder 7.0 (from National Instruments Corporation, NI) to define the exact center of the fiducial marks.
In this study, the mask serves as the object to be aligned. To
Experimental setup
To enhance the precision of alignment, the whole experimental setup was positioned on a granite stage. As shown in Fig. 14, the experimental setup consists of a machine vision system for image extraction, a CCD camera mounted on an adjustable frame, a two-axis (X and Y) linear table consisting of two HR4 piezoelectric ceramic motors, two LIE5 linear scales with 20-nm resolution, the independent X-axis table with 150-mm travel, and the Y-axis table with 100-mm travel, the Y-axis table is located
Results and discussion
This study has successfully developed a robust automatic vision-aided alignment system. Two sets of experimental results (Cases I and II) are described below for illustration. Fig. 16a shows the fiducial circle extracted from the object image captured by the CCD of Case I. The intersection of the cross indicates the target position. Table 4 shows the characteristic matrix obtained for Cases I and II. According to the image data, the difference in position between the fiducial center and the
Conclusion
In this study, a novel automatic nano-alignment system using machine vision coupled with a piezoelectric motor for motion control was developed. The alignment operation makes use of the correlation between the input and output signals. The CCD camera extracted fiducial marks from the images. These image data are employed to establish the characteristic matrix through matrix transformation. With the feed-in position of the tool, the difference in fiducial mark as well as positional variation can
References (14)
- et al.
A hybrid image alignment system for fast and precise pattern localization
Real-Time Imag
(2002) - et al.
An image system for fast positioning and accuracy inspection of ball grid array boards
Microelectron Reliab
(2001) - et al.
2-Step algorithm for automatic alignment in wafer dicing process
Microelectron Reliab
(2004) - et al.
Concepts for a class of novel piezoelectric self-locking long-stroke actuators
Precision Eng
(2002) - et al.
Precision position control using combined piezo-VCM actuators
Precision Eng
(2005) - et al.
An auto-alignment vision system with three-axis motion control mechanism
Int J Advan Manufact Technol
(2005) - Kim HT, Song CS, Yang HJ. Matrix form of automatic alignment algorithm in 2D space. In: Proceedings of the IEEE...
Cited by (45)
High-precision tracking of piezoelectric actuator using iterative learning control and direct inverse compensation of hysteresis
2019, Precision Motion Systems: Modeling, Control, and ApplicationsA new approach to hysteresis modelling for a piezoelectric actuator using Preisach model and recursive method with an application to open-loop position tracking control
2018, Sensors and Actuators, A: PhysicalCitation Excerpt :For decades, piezoelectric actuators featuring Lead Zirconate Titanate (PZT) material have been one of the most advanced actuators while it has being applied extensively in numerous fields such as robotics [1], micro/nano scale machining [2,3], vibration suppression [4,5], etc.
Designing transparent piezoelectric metasurfaces for adaptive optics
2024, Nature CommunicationsDesigning Transparent Piezoelectric Metasurfaces for Adaptive Optics
2023, Research SquareAn improved particle swarm optimization using long short-term memory model for positioning control of a coplanar XXY stage
2022, Measurement and Control (United Kingdom)
- ☆
This paper has not been published elsewhere nor has it been submitted for publication elsewhere.
- 1
Currently with Department of Mechanical Engineering, St. John’s University, Tamsui, Taipei 251, Taiwan, ROC.
- 2
Currently with Department of Automatic Engineering, Lan-Yang Institute of Technology, Toucheng, I-Lan 261, Taiwan, ROC.