[Paper Review] Virtual VNA 2.0: Ambiguity-Free Scattering Matrix Estimation by Terminating Not-Directly-Accessible Ports with Tunable and Coupled Loads

Philipp del Hougne|arXiv (Cornell University)|Sep 17, 2024

Indoor and Outdoor Localization Technologies4 citations

TL;DR

This paper introduces Virtual VNA 2.0, a method to estimate the full N×N scattering matrix of a complex system without direct access to N−NA ports by using a tunable multi-port load network (MPLN) to resolve sign and blockwise phase ambiguities. The approach enables ambiguity-free estimation using only intensity-only measurements and requires only N_S additional measurements with a simple 2-port load network, validated experimentally with a chaotic cavity at 771 MHz achieving 32 dB accuracy.

ABSTRACT

We recently introduced the "Virtual VNA" concept which estimates the $N imes N$ scattering matrix characterizing an arbitrarily complex linear reciprocal system with $N$ monomodal lumped ports by inputting and outputting waves only via $N_\mathrm{A}

Motivation & Objective

• To resolve persistent sign and blockwise phase ambiguities in scattering matrix estimation when NDA ports are terminated with individual tunable loads.
• To enable full scattering matrix recovery without explicit phase measurements, using only intensity data.
• To develop a generic, non-invasive method applicable to arbitrarily complex systems without prior knowledge of system characteristics.
• To reduce measurement complexity by scaling linearly with the number of inaccessible ports, avoiding full crossbar switch matrices.
• To enable practical characterization of large antenna arrays using only a few-port VNA.

Proposed method

• Introduce a multi-port load network (MPLN) that connects to two NDA ports simultaneously, enabling joint phase and sign calibration.
• Use a gradient-descent-based optimization to estimate the scattering matrix from intensity-only measurements, with phase alignment via median-based offset correction.
• Perform N_S additional measurements with the MPLN to resolve sign ambiguities by comparing predicted vs. measured intensity patterns.
• Apply a sequential sign alignment procedure: fix the sign of one column/row, then iteratively resolve others using cross-ports in the MPLN.
• Estimate the global phase offset θ using multiple realizations and minimize residual error across all measurable scattering coefficients.
• Leverage the fact that intensity measurements are insensitive to global phase and blockwise phase shifts, using multiple coefficients to resolve θ uniquely when NA > 3.

Experimental results

Research questions

• RQ1 Can sign ambiguities in scattering matrix estimation be resolved when NDA ports are terminated with individual tunable loads?
• RQ2 Can blockwise phase ambiguities in intensity-only measurements be lifted without explicit phase data?
• RQ3 Is it possible to achieve ambiguity-free scattering matrix estimation using only N_S additional measurements with a simple 2-port load network?
• RQ4 Can the method be applied generically to arbitrarily complex systems without prior knowledge of system geometry or low-frequency behavior?
• RQ5 Does the measurement complexity scale linearly with N_S, avoiding the exponential cost of full crossbar switch matrices?

Key findings

• The Virtual VNA 2.0 approach successfully resolves both sign and blockwise phase ambiguities using only N_S additional measurements with a 2-port load network.
• The method achieves 32 dB accuracy in full scattering matrix estimation using only intensity-only measurements in an 8-port chaotic cavity experiment.
• The phase offset θ is estimated by minimizing residual error across multiple scattering coefficients, requiring NA > 3 to ensure sufficient constraints.
• The sign alignment procedure is robust and sequential, resolving ambiguities by comparing predictions with measurements for each new NDA port connection.
• The approach is experimentally validated under intensity-only conditions, demonstrating feasibility without explicit phase measurement.
• The measurement complexity scales linearly with N_S, offering a practical alternative to costly full crossbar switch matrices in large-array characterization.