KangWon Lee

Quick Wafer Alignment Using Feedforward Neural Networks

This paper proposes a wafer alignment method using a feedforward neural network (FNN). A wafer is placed on a vacuum table and its misalignment is inspected by a frame grabber. The alignment of the wafer is repeatedly corrected by adjusting its kinematic positions until the misalignment becomes zero. The training set is composed of the misalignment and the xy¿ compensation. Fifteen sets of misalignment data were measured by the conventional alignment method. The FNN has four normalized inputs as mark locations in the field of vision and three normalized outputs for the kinematic compensation. The macro and micro steps share the FNN by means of a multiplexer (MUX) and a scaler. The training rule was back-propagation (BP). The proposed method was applied to a wafer manufacturing machine and its alignment performance was compared with that of conventional methods. The alignment time was reduced by the FNN model which was especially efficient in the case of macroscopic alignment.

An electro-magneto-pneumatic spring for vibration control in semiconductor manufacturing

An active vibration isolator was developed for semiconductor manufacturing machines using electro-magnetic(EM) levitation force. The spring is composed of a pneumatic elastic chamber and a magnetic actuator. The actuator has a coil and a permanent magnetic plate and is installed inside of the chamber. The mechanical and electronic parts are designed to operate under a weight of 2.0 tons. An air mount is constructed for the experiment with a stone surface plate, 4 active air springs, 4 gap sensors, a DSP controller, a linear power amp and vibration measurement systems. The impulse response of the air mount was monitored before and after the active control. The time duration was reduced by 77% in the impulse response. The maximum peak in frequency domain was reduced by 62%. We also found that the active system can avoid the resonance caused by the natural frequency of the passive system.

A 3 DOF Model for an Electromagnetic Air Mount

A 4 × 4 matrix model with three degrees of freedom is proposed as a means for controlling microvibrations and applied to an electromagnetic isolator. The model was derived from an assumption based on small- and low-frequency vibrations. The coordinates of the 3 DOF was composed of the 4 variables, representing a vertical position, pitch, roll, and a proof term. The coordinates were calculated from the 4 position sensors in the isolator and formulated into a 4 × 4 matrix, which possesses inversive full rank. The electro-magnetic isolator was built for a simulated machine in semiconductor manufacturing and consisted of a heavy surface plate, sensors, amps, a controller, and air springs with electromagnets. The electromagnets are installed in a pneumatic chamber of the individual air spring. The performance of the 3 DOF model was experimented and compared with that of a 1 DOF model in an impact test. The settling time in the result was reduced to 25%.

An Electro-magnetic Air Spring for Vibration Control in Semiconductor Manufacturing

One of the typical problems in the precise vibration is resonance characteristics at low frequency disturbance due to a heavy mass. An electro-magnetic(EM) air spring is a kind of vibration control unit and active isolator. The EM air spring in this study aims at removing the low frequency resonance for semiconductor manufacturing. The mechanical and electronic parts in the active isolator are designed to operate under a weight of 2.5 tons. The EM spring is floated using air pressure in a pneumatic elastic chamber and actuated by EM levitation force. The actuator consists of a EM coil and a permanent magnetic plate which are installed inside of the chamber. An air mount was constructed for the experiment with a stone surface plate, 4 active air springs, 4 gap sensors, a DSP controller, and a multi-channel power amp. A PD control method and operating logic was applied to the DSP. Simulation using 1/4 model was carried out and compared with the experiments. The time duration and maximum peak at resonance frequency can be reduced sharply by the proposed system. The results show that the active system can avoid the resonance caused by the natural frequency of the passive system.

Capacitive Absolute Pressure Sensor with Vacuum Cavity Formed by Bonding Silicon to Soiwafer for Upper Air Observations

We present a capacitive absolute pressure sensor with a large deflected diaphragm that was fabricated with a sealed vacuum cavity formed by removing handling silicon wafer and oxide layers from a SOI wafer after eutectic bonding of a silicon wafer to the SOI wafer. The deflected displacements of the diaphragm formed by the vacuum cavity in the fabricated sensor were similar to simulation results. This result was estimated because of the dense interface produced between cavity-formed Si and the top Si layer of the SOI wafer by the Si-Au eutectic bonding process. Initial capacitance values were about 2.18pF and 3.65pF under normal atmosphere, where the thicknesses of the diaphragm used to fabricate the vacuum cavity were 20 μm and 30 μm, respectively. Also, it was confirmed that the differences of capacitance value from 1000hPa to 5hPa were about 2.57pF and 5.35pF, respectively.

A method of vertical and horizontal force estimation by using air-filled material and camera for soft physical human-robot interaction: fundamental experiments

A method of vertical and horizontal force estimation for soft physical human–robot interaction by using an air-filled material and a camera was proposed in this study. First, we considered hypotheses for vertical and horizontal force applied to the air-filled material. Second, the prototype of air-filled material and vision-based force-sensing device (AVFSD) was constructed. Based on the result of preliminary experiments, the relationships between force, pneumatic pressure, and contact surface displacement were formulated. The vertical and horizontal force could be estimated using the AVFSD with percentage errors of 3.6 and 5.8%, respectively, and the hypotheses were verified through the experiments. Graphical Abstract

A 3 DOF model for an electro magnetic air mount

A model with three degrees of freedom is proposed as a means for controlling for controlling micro vibrations in semiconductor manufacturing. The model was derived from an assumption based on small and low frequency vibrations. The coordinates of the 3 DOF was composed of a vertical position, pitch, roll and a proof term. The coordinates of the 4 variables can be calculated from the 4 positions in an isolator. This coordinate can be written as a 4 × 4 matrix, which possesses inversive full rank. The proposed algorithm was applied to an electro-magnetic isolator. The active isolator consisted of a heavy surface plate, sensors, electro-magnetic actuators, amps, air springs and a DSP controller. The electro-magnetic actuator was installed in a pneumatic chamber of the individual air spring. The active isolator was experimentally used as a base structure under the semiconductor manufacturing machines. The performance of the 3 DOF model was compared with that of a 1 DOF model in an impact test. The settling time was reduced to 25% in the experiment.

A self-learning method for automatic alignment in wafer processing

We propose a self-learning method for automatic wafer alignment in the semiconductor manufacturing process. A feed forward neural network is trained by and used for wafer alignment. The network determines the movement of kinematic parts from the misalignment inspected by machine vision. However, it is time-consuming and inconvenient to obtain training data in this way. So, we built an automatic learning rule to gather the data and train the network. The network may determine wrong outputs and cause other misalignments at first, but the error can decrease as the training proceeds. The training sets consisted of a variation of misalignment data and the movement of an alignment stage. Five recent sets are used for training and others are dismissed or forgotten. This re-trained network tried aligning, measured misalignment, and made new training sets. This sequence makes it possible to acquire alignment skill and automate the process. After learning, automatic alignment accomplished sub-pixel accuracy for several cases of misalignment. The result showed that the proposed method could be applied to the semiconductor manufacturing process. Its performance improved about 6% compared with conventional algorithms.

Design and fabriction of micro-viscometer using the propagation of acoustic waves in micro-channel

A micro-viscometer for measuring viscosity changes in small amount of liquid in real time is designed and fabricated. The merits of the device are to use minimal liquid and to increase the sensitivity for measuring the viscosity. The micro-viscometer is composed of two chambers connected by several micro-channels. Each chamber has a unimorphic piezoelectric diaphragm for driving and sensing sound waves. The micro-channel plays a role as a waveguide and an attenuator of sound waves. Attenuation of sound wave is a function of the viscosity of the liquid. So, the output voltage of sensor is also changed as the viscosity of liquid is changed. After fabrication through micromachining, the device was tested on several viscous liquids, and then it was demonstrated that the device can play a role of a viscometer with high sensitivity. In this paper, we mainly discuss on the comparison between experimental results and newly simulated results.

Software Architecture and Subclassing Technique for Semiconductor Manufacturing Machines

In this paper, we proposed a software architecture for operating an automatic semiconductor manufacturing machine. Nowadays, machines used for semiconductor processing require a high level of automation in the areas of motion control, machine vision, data acquisition and networking. These functions are assured by industrial equipment that is generally installed in a computer. This equipment occupies a large part of the system resources and generates a large computational burden, so the software should be designed for efficiency. The proposed architecture consists of four layers and virtual equipments(VEs). The VEs are made by subclassing the physical equipments(PEs), and the layers are coded into a thread which updates the status of the VEs. Subroutines in the program p to the pointers to the VEs, and direct access to the physical equipment is prohibited. The number of accesses(NOAs) to the PEs in a typical industrial application was simulated for an unlimited access structure and the proposed structure. The result showed that the proposed structure was more efficient than the typical one irrespective of the subroutines. This architecture was also applied to design machine operating software for performing automatic wafer dicing.