Abstract: Noise figure, recognized as a critical technical parameter of T/R modules, governs the equivalent noise figure of radar systems. Whereas the Y-factor method predominantly employed in laboratory measurements fails to fulfill automated testing requirements for T/R modules, the cold-source method has been established as the standard solution. Automated test systems exhibiting suboptimal accuracy and consistency necessitate systematic optimization. Analysis of cold-source operational principles and noise parameters identifies source matching as a determinant factor in testing accuracy, leading to development of an optimized scheme for automated noise figure test systems. Comparative batch testing validates an approximate two-fold enhancement in consistency achieved by the optimized system.