On the choice of cascade de-embedding methods for on-wafer S-parameter measurement

Performance of thru-only cascade de-embedding methods and their variants that use a Π- or a T-equivalent to represent and bisect a symmetric THRU is assessed. The results from the Π- and T-based methods are reasonable at low frequencies. However they are shown to deviate noticeably from the correct results as the frequency gets high or, equivalently, when the length of the THRU approaches an effective wavelength λ. A better alternative at high frequencies is TSD (thru-short-delay), which, when THRU is symmetric, requires only THRU and LINE. TSD gives correct results except in the periodically appearing `dead zones', provided that the characteristic impedance, Zχ, of the transmission line (TL) in the LINE is known. A Π-based method could be used to extract Zχ at low frequencies, from which Zχ can be extrapolated to higher frequencies.