A Precision-Enhanced 1× De-Embedding Method Incorporating a Hybrid Transformer–CNN Network

The conventional 1x automatic fixture removal (AFR) de-embedding method gains wide application due to its simplicity and efficiency, which requires only a single de-embedding standard. However, impedance discontinuities on the fixture can significantly compromise its accuracy. To address this limitation, this work proposes a precision-enhanced de-embedding method that integrates a hybrid Transformer-CNN network into the conventional 1x AFR framework, particularly for fixtures with impedance discontinuities at their terminals. By learning and compensating for the non-idealities introduced by fixture impedance discontinuities, the proposed method substantially enhances the accuracy of the traditional 1x AFR method without the need for additional de-embedding standards. This hybrid neural network-assisted de-embedding method offers a practical and scalable solution for high-precision characterization in scenarios where fixture discontinuities are non-negligible.