Development of innovative fringe locking strategies for vibration-resistant white light vertical scanning interferometry (VSI)

Transcription

1 Development of innovative fringe locking strategies for vibration-resistant white light vertical scanning interferometry (VSI) Liang-Chia Chen 1), Abraham Mario Tapilouw 1), Sheng-Lih Yeh 2), Shih-Tsong Lin 3), Yi-Shiuan Lin 1) 1) Automatic Optical Inspection Laboratory Graduate Institute of Automation Technology National Taipei University of Technology 1, Sec. 3, Zhongxiao East Rd. Taipei, Taiwan 106 Tel.: + [886] ext Fax: + [886] lcchen@ntut.edu.tw Abstract White light interferometry has become an important method for measuring micro surface profiles with a long vertical range and nano-scale resolution. However, environmental vibration encountered in in-field optical inspection is usually unavoidable and it affects measurement performance significantly. Isolating vibration sources from the measurement system employing optical isolation systems sometimes is not completely effective, especially within complicated in-field fabrication environment. Therefore, in this research we aim to develop a novel method in achieving white-light fringe-locking condition during vertical scanning processes. The developed optical system consists of white light source, a Mirau objective, a PZT vertical scanner, an optical band-pass filter and two image sensing devices. The developed system generates a high and a low coherent interferograms, simultaneously captured by two charged coupled devices (s). The high coherent interferogram is employed to detect high-speed nano-scale displacement and direction of the external vibration and an active real-time closed-loop feedback is performed to compensate the vibration movement. To achieve real time fringe-locking operations, an innovative vertical scanning strategy is developed to lock fringe movement accurately without influences from all-time disturbing vibration. From some preliminary experimental results, the vibration compensation with a system response bandwidth up to 1.7 khz can be effectively achieved for fringe locking. The feasibility of the anti-vibration VSI system is verified by performing some of industrial in-field examples. Keywords: Vibration resistant, nano-scale measurement, active feedback system, real-time control system, automatic optical inspection (AOI) 1. Introduction 3-D surface reconstruction using white light interferometry (WLI) has many advantages in comparison with phase shifting interferometry (PSI). One of the advantages is the use of its large vertical scanning range to overcome the phase ambiguity problem faced by general PSI. WLI employs light source with a short coherence length to determine the zero optical path difference (OPD) for accurate surface profile reconstruction. A general problem in using optical interferometry is that it is affected by environmental disturbances mainly coming from sound or structural vibration. These disturbances are sometimes unavoidable due to in-field operation characteristics and normally significantly deteriorate the measurement result. Gao performed a comprehensive study on analyzing surface measurement errors of commercial scanning white light interferometers and identified some error sources and possible methods to resolve them [1]. Meanwhile, several approaches have been performed to overcome the vibration problems both in PSI and WLI [2-3]. In this paper, we present an innovative methodology to minimize the vibration problem in white light VSI. The underlying principle of the developed method is to maintain a constant 1-292

2 distance between an interferometric objective and the measured object within vibrationencountered in-field environment. The rest of the paper is organized as follows. Section 2 describes the optical system configuration. Following this, the principle of the proposed fringe-locking method is detailed in Section 3. According to the experimental results, Section 4 illustrates and analyzes the performances of the developed methods for 3-D surface reconstruction. Section 5 contains the conclusion of this study. 2. Optical system configuration The optical System setup is presented in Fig. 1. The light source used in the system consists of a broadband and long coherence light source. The long coherence light source is generated by using a single superluminescent white-light LED having 10-watt power. The light is ignited by a warm-type white-light LED with a two-peak light spectrum, where the first peak locates at 470 nm while another one lies on 550 nm. The system uses a Mirau interferometric objective mounted on a PZT in order to perform the vertical scanning profilometry. Then, there are two s employed in the system for capturing interferograms, in which one is a normal-speed and the other is a high-speed. The normal-speed is used to capture low-coherent interferogram images for vertical scanning while the high-speed is employed to acquire high-coherent interferogram images for fringe locking operation. A narrow optical bandpass filter with nm is employed to transform the broadband light into a high-coherent one. The high-speed is capable of detecting the high-coherent interferogram up to a bandwidth of 1.7k Hertz or higher. A software module employing a closed-loop control scheme is developed to achieve real-time fringe-locking during the vertical scanning measurement. Normal Software Module Band-pass Filter Surface profiler Collimating Lens High Speed Fringe Locking Light Source PZT Mirau Object Fig. 1. Schematic diagram of optical system configuration TM. In white-light interferometry, the unique advantage of using a long coherent light source instead of a laser one for the vibration-displacement detection is mainly in avoiding potential undesired speckles often encountered in laser interferometry. The long-coherent interferogram fringe generated by blue-led light is capable of detecting the vibration with a nano-scale resolution and a detecting range up to several-tens micrometers. Another important advantage of the proposed system is of its simple optical configuration, which is slightly different to a conventional white-light interferometer. Hardware retrofitting cost to establish the developed system is reasonably low. 3. Fringe-locking strategy The locking mechanism is performed by first detecting the phase-shifting difference in the interferogram fringe, then calculating the real travelling distance of the object from the interferogram. Figure 2 shows an illustration of interferogram fringes generated by the blue LED. The first fringe (Fig. 2(a-1)) is used as the reference at Time t. In case of occurrence of environmental disturbance, the fringe at Time t+ t may shift mainly depending on the magnitude and direction of the vibration (Fig. 2(a-2) representing for a forward movement 1-293

3 and Fig. 2(a-3) for a downward one). By detecting the actual shifting quantity of the fringe, the displacement and direction of the disturbing vibration can be determined. Reference Ia Ib Controller System Ia Ib (1) h(t0) + - dh G * dh y White Light Interferometer System I(t0+ t) Ia (2) h(t0+ T) K * di + h(t0) di - + I(t0) Ib (3) Fig. 2 Phase-shifting detection by analyzing interferometric fringes, The closed-loop process for real-time vibration detection and compensation. The interference fringe generated by the interferometric system can be generally represented by the following interference equation: I I0(1 cos ) (1) Where I0 represents the dc term of the interferometric fringe, is the modulation of the fringe, and is the phase due to the optical path difference. Therefore, by expressing the height displacement of the vibration using the phase difference, the displacement can be derived as the change of the light intensity of the interferometric fringe as follow: dh di (2) 4 I 0 sin In the above equation, the displacement of the object relative to the Mirau interferometric objective can be reasonably expressed as a linear function of the gray-level change of the interferogram fringes. Therefore, in order to track such a displacement, one fringe is captured at the beginning as a reference and then the displacement is then continuously measured using the intensity change of the successive fringes relative to the reference fringe. Using this method, it is effective to obtain the real-time vibration displacement and direction when dealing with a longitudinal vibration having a vibrating frequency less than one-fifth of the detecting frame rate of the high-speed. Figure 2 shows the scheme of the closed-loop process for real-time vibration detection and compensation. The feedback control of the developed system consists of a vibration detection mechanism relying on the principle of Equation (2) and a real-time controller. The output of the controller is automatically fed back into the WLI system by moving the PZT with the exact amount of the vibrating displacement needed for real-time compensation. Meanwhile, during the vertical scanning process of WLI, the fringe locking is employed to compensate any possible displacement errors potentially caused by the PZT. The vertical scanning step can be one half of the wavelength or a quarter of wavelength, in which the wavelength employed here, is 470 nm. Therefore, when the fringe locking scheme is applied, the vibrating detection range of the developed system is currently bounded by the coherence length of the high-coherent light source. In the developed system, the interferogram fringe is detected by the high-speed and the fringe is then analyzed by the real-time fringe analysis module to detect and track the vibration. With the vibration being accurately detected by the real-time system, a feedback control signal is automatically sent to the PZT controller to compensate the displacement in time. The success of the developed approach mainly depends on simultaneous coordination of 1-294

4 Displacement ( m) Displacement ( m) the three important modules employed in the system, such as the high speed camera, the control center with multi-thread CPUs and the bandwidth of the PZT controller. To achieve this, real-time programming scheme is applied in the development of the developed system. 4. Experiment results and analysis To test the fringe locking mechanism, a standard flat mirror was mounted on a PZT and moved in a sinusoidal modulation with amplitude of 100 nm and various frequencies. The actual displacement was measured by a calibrated external laser interferometer with a higher measurement accuracy to verify the accuracy and responsive bandwidth of the developed system. The frame rate of the was controlled to obtain an interferogram image with a sufficient contrast for ensuing success in detecting the real-time fringe movement. In the experiment, the frame rate of the was adjusted to be 100 fps as an example for detecting an external a simulated vibration with an excitation frequency of 10 Hz by using an external vibration exciter rigidly attached to the object. By using the developed system, a vibrationsuppressed displacement was continuously applied to compensate the external simulated vibration. Fig. 3 illustrates the experimental results, in which 4 shows the actual vibration detected by the laser interferometer was modulated between 50 nm when no compensation was taken. It is shown that the actual displacement was reduced to a dc level of 1 nm and an ac modulation of 3 nm when the fringe-locking scheme was employed. This clearly demonstrates the effectiveness of the fringe locking method. From the result analysis, the residual errors may mainly come from two potential sources, the measurement error of the detecting and compensating module itself and the external laser interferometer. For the WLI vertical scanning measurement, it is found that the current capability of the developed system is reasonably satisfactory for the actual demands in items of accuracy and bandwidth for infield operation. Frequency 10Hz, feedback off Frequency 10Hz, feedback On Sample (sampling time: 10ms) Sample (sampling time: 10ms) Fig. 3. Measured displacement between the Mirau objective and object when a simulated displacement of a reference mirror moved at 10 Hz and an amplitude of 100 nm: without fringe locking and with fringe locking enabled. Meanwhile, the proposed methodology was employed to measure the flatness of the reference flat mirror on a shop floor in two different vertical scanning conditions, in which the one is that the optical system was rigidly mounted on an active-type optical table while the another placed without the vibration protection of the optical table. For surface profilometry, Fig. 4 displays the surface measurement result of the standard flat mirror when the measurement was taken on the optical system and without the employment of the fringe locking method. It is verified that the maximum surface flatness was 107 nm. In contrast with that, Fig. 4 illustrates the result when the system was mounted on an inactive optical table, in which the maximum surface flatness was reduced to 79 nm. This confirms that the developed system is capable of minimizing the negative measurement errors. Without an active optical table, the measurement result by using the developed method was either comparable or better than the one obtained with the optical-table protection

5 Measurement error (P-V): 107 nm Measurement error (P-V): 79 nm Fig D surface measurement result of the developed WLI system mounted on an active optical table and mounted on an inactive optical table. The experiment results clearly confirm the effectiveness of the proposed methodology. When the fringe locking method was applied to the measurement system, vibration effect could be suppressed and reduced from ±40 to ±6 nm. This is a preliminary test result of the developed fringe locking approach. More experimental tests can be further implemented to investigate the characteristics of the developed method. A with a higher frame rate equipped with sufficient light exposure capability can be employed in the system to further enhance the performance of the system. Assuming the surface flatness of the tested reference mirror within a small surface region is perfect to be zero, the measurement deviation of the measured surface of the system mounted on the inactive optical table is 38% less than the one measured with the active optical table. The result of the measurement in an inactive optic table shows a better peakvalley error in comparison with the one on active optic table. Furthermore, in order to ensure the conclusion, comprehensive investigation in system repeatability is needed. 5. Conclusion A novel fringe locking method has been developed and implemented in WLI vertical scanning measurement for minimizing a shop-floor vibration disturbance. A WLI verticalscanning interferometer equipped with an effective fringe locking mechanism was successfully developed to maintain a constant distance between Mirau objective and the measured object within ± 3 nm deviation. The exclusive advantage employing the proposed methodology in comparison with conventional active-type anti-vibration optical system is that the hardware retrofitting cost required to establish the developed system is significantly lower. The test result indicates that the system can work without an optical table enabled to achieve a comparable measurement accuracy as the one by the traditional white-light VSI. The performance of the system currently limited by the frame rate of the camera, the coherence length of the light source and the responsive bandwidth of the developed system can be further enhanced with advancement of the hardware performance. References 1. F. Gao, R.K. Leach, J. Petzing, J.M. Coupland. Surface measurement errors using commercial scanning white light interferometers. Meas. Sci. Technol. 2008, (13pp). 2. J.Y. Liu, I. Yamaguchi, J.I. Kato, T. Nakajima. Real-time surface shape measurement by an active interferometer. Optical Review. 1997, 4 (1 Part B), pp C. Zhao, J.H. Burge. Vibration-Compensated Interferometer for Surface Metrology. Appl. Opt. 2001, 40, pp