適應性壓力控制系統於MOCVD製程之實現

Abstract

在生產LED元件時，金屬有機物化學氣相沉積(MOCVD, Metal Organic Chemical Vapor Deposition)是最關鍵的一道製程，由此製程所形成的磊晶均勻度與厚度會嚴重影響LED發光效率。更重要的是，MOCVD腔體壓力的穩定度又是影響磊晶薄膜品質的一項關鍵因子。故本論文期望發展出一套適應性壓力控制系統，能更精確的控制腔體壓力。 首先，經由真空壓力理論分析，推導出在固定操作點下，可代表MOCVD壓力系統動態的一階時延系統(FOPDT system)，並透過系統鑑別(system identification)實驗求出此系統的實際參數，再與理論所推導出的系統參數作比較以確認所得轉移函數與理論相符。確認受控系統的動態特性後，利用適應性史密斯預測器(Adaptive Smith Predictor)之控制架構以實現MOCVD製程的壓力控制，利用此架構消除時間延遲對系統的影響，並透過其適應性演算法，即時調變系統參數以增進系統的效能。最後，將此控制機制以PC-Based控制器實現，並與傳統的PID控制器做比較，以驗證此控制架構之實用性與其優良之控制成效。

The Metal Organic Chemical Vapor Deposition (MOCVD) is the most crucial process for LED manufacturing. The thickness and the uniformity of the epitaxy made from MOCVD will crucially affect the luminous efficiency of the LED die. Furthermore, the pressure stability in MOCVD chamber is an important factor for the quality of epitaxy. As a result, this thesis expects to develop a better pressure control system than the traditional one. First, this thesis constructs an experimental system to emulate the practical MOCVD vacuum system. Also, to figure out the system mathematical model, this thesis demonstrates how to utilize small-signal model to linearize the system dynamics as a First Order Plus Dead Time (FOPDT) model at fixed operating point. Moreover, in order to find the values of the system parameters and realize the trend of the system dynamics variation, an experiment for system identification is conducted. In this thesis, the N4SID is the method to identify the system. Second, this thesis presents the Adaptive Smith Predictor as the pressure control structure to annihilate the effect of dead-time and estimate the system parameters on-line to improve the system performance. Furthermore, a simulation is conducted to check the feasibility of the control scheme. In the end, the control scheme is implemented as a PC-based controller. Its performance is compared with the traditional PID controller to verify that it is a pragmatic and well-performing controller.