Machine-learning based tuning algorithm for duplexer systems

Abstract

This disclosure provides techniques for impedance matching. A radio frequency (RF) device includes a power detector to determine a transmitter leakage and a post-processing unit to determine a receiver leakage, and determines if isolation is acceptable based on the leakages. The RF device may include a device for measuring antenna impedance. Otherwise, the RF device may select multiple tuner settings (e.g., capacitor values) for test signals to be transmitted and received at a target frequency, determine multiple sets of leakage values, determine multiple reflection coefficients based on the multiple sets of leakage values, and determine an estimated antenna impedance at the target frequency based on the reflection coefficients. The RF device then determines impedance tuner settings based on the measured or estimated antenna impedance. Alternatively, the RF device determines impedance tuner settings using an inverse machine-learning model based on a determined matching impedance.