Multiport VNA Self-Calibration of a Partially Leaky Error Model with a Reciprocal Standard

The present paper provides a new self-calibration procedure, applying the well-known unknown thru techniques to a multiport network analysis with a partially leaky error model characterized by 31 unknowns. The error networks of the pairwise coupled ports can be calibrated separately with an arbitrary direct calibration or self-calibration method, that is suitable for the 15-term error model. Afterwards, a widely unknown but reciprocal network can be connected between these pairs to link the error networks for a subsequent system error correction. The method is presented for a four port network analyzer and can be extended for 2N-port VNAs (N = 2) with pairwise coupled reflectometers. In this context, a double reflectometer configuration at each port is supposed. After a detailed presentation of a vectorized partially leaky error model of a four port setup, a cascade matrix representation of the error networks and the device under test (DUT) is obtained. Within this expression, the impact of a reciprocal standard vanishes when applying the determinant. Experimental measurements with the suggested method have been performed with pairwise artificial crosstalk, which show an almost perfect match relating to results of a common TLR self-calibration in a crosstalk-free setup.