光耦合器產品品質之發展

Abstract

摘 要 全球光通信傳輸所面臨的困難是各傳輸網所需的基礎模組與元件的規格不統一。在眾多光被動元件中之光耦合器製造，因牽涉到光學、機械、電子、熱融等物理的控制，尚無法自動化，量產所必須面臨的管理與技術整合問題頗多，其中產品品質的發展是一項重要的挑戰課題。本文探討並介紹光耦合器產品品質之發展，建議藉由產品設計驗證與製程參數最佳化兩項產品品質發展工具，確保光耦合器的產品品質與生產良率改善。本文所研究的結果可提供類似光通信元件廠，欲建立具品質與成本競爭優勢的光耦合器產品的參考。 本文所建議的產品品質發展工具包含：產品研發可靠度設計驗證，製程參數最佳化。前者確保產品的可靠度能符合市場的規範要求；後者深入探討關鍵燒拉製程參數最佳化，應用（A）田口方法，（B）類神經網路與指數滿意度函數分析與（C）類神經網路、指數滿意度函數與基因法則演算法之整合分析等三項常用的方法來提昇製程良率，所建立的參數最佳化程序可取代傳統的嘗試錯誤方法，迅速滿足光通信市場上大量客製化規格需求的製程參數。總結來說，第三種整合方法最適宜用來改善燒拉製程參數最佳化。 最後，本文以新竹科學園區某光通信被動元件公司為例，說明在光通信元件產業中應用本文產品品質之發展的成效。初期以可靠度設計驗證來提昇其產品本身光特性與品質的優越性，順利完成各項嚴厲的環境試驗挑戰，不過產品燒拉製程良率僅能達成65%左右。接著，進行第二階段燒拉製程的參數最佳化，使產品的製程良率由65%改善至平均99%以上，且特優等級產品佔90%以上。本文光耦合器產品品質之發展，不僅僅獲得上述成果，對客製化產品的評估反應時間也從七至十天縮短為一至二天內完成，各種結果均顯現出令人滿意的結論。

ABSTRACT The worldwide difficulties in fiber optical transmission industry these years are coming from various developing techniques and inconsistent specifications to the basic devices and components. Among numerous passive devices, the fused fiber coupler manufacturing is most difficult to be fully automated due to its physical interactions among optics, mechanics, electrics, and heat fusion. The product quality development of fiber coupler is a critical and challenging task. This study illustrates the progress of fiber coupler quality development and recommends two tools, reliability design verification and fusion process optimization approach, to ensure the product and process quality. The success of the implementation could be provided to those manufacturers as a useful reference to establish the coupler product with competitive advantages. The two tools for ensuring fiber coupler quality are reliability design verification and fusion process optimization approach. The former provides the conformity of the product by meeting the requirements in optical industry; the latter provides the competitive cost performance by significant improvement in production yield rate. This dissertation also deeply illustrates the procedures to optimize fiber coupler fusion parameters with three different approaches, i.e. (a) Taguchi methods, (b) neural network and exponential desirability function analysis and (c) integrated neural network, exponential desirability function and genetic algorithm analysis. This study demonstrates it could easily take place of traditional trial-and-error approach, to promptly fulfill the need of frequent specification changes from market. In summary, the outcomes from the third hybrid approach are superior to the first two approaches and it is also very suitable to solve the problems for parameter optimization in fusion process. Finally, this product quality development of fiber coupler is successfully demonstrated on a commercial passive optical communication company in Hsinchu Science-based Industrial Park, Taiwan. The first illustration demonstrates product reliability by design verification. The result is encouraging that low optical performance variations has been observed under severe environmental testing as specified in optical industry. But it is unsatisfied that only 65% fusion process yielding ratio was achieved. The second illustration demonstrates process quality by parameters optimization. The recommended hybrid approach, (c), demonstrates that the exercised company not only obtain considerable yielding ratio improvement in fiber coupler fusion process from 65% to 99% but also gain supreme product performance and quality. In addition to above creditable advantages, the cycle time of specification evaluation for customized product has been reduced from 7~10 days to 1~2 days. All above evidences conclude to satisfactory results.