This chapter provides an overview of the observer-based fault diagnosis framework and its historical development. It discusses major topics in the framework including observer design approaches, unknown input decoupling, and robustness issues. The chapter also outlines the key tools used in observer-based fault diagnosis research, which include linear system theory, robust control theory, and linear matrix inequalities. Major issues covered include residual generation, evaluation, and threshold computation. The chapter reviews pioneering contributions to areas like observer-based residual generation and decoupling unknown inputs, as well as more recent work on robustness and performance optimization.