Internal Calibration Process Using Chirp Pulses with Application of the Adam Learning Algorithm

TL;DR

This paper introduces a novel internal calibration method using chirp pulses and the Adam learning algorithm to effectively mitigate thermal drift in active RF components, eliminating the need for additional calibration signals.

Contribution

The paper presents a new calibration approach that leverages chirp pulses as calibration signals and employs the Adam algorithm for improved accuracy and robustness.

Findings

Achieved maximum gain difference of 0.06 dB after calibration.

Achieved maximum phase difference of 2.42 degrees after calibration.

Demonstrated superior effectiveness over conventional gradient descent methods.

Abstract

We propose a new internal calibration process using chirp pulses. Our method is utilized to mitigate thermal drift, which is unwanted changes and usually occurs in active elements such as a high power amplifier and low noise amplifier. The proposed method has advantages from two distinct aspects: calibration signal and algorithm. In respect to the calibration signal, our method does not contain an additional signal source because chirp pulses, which are normally used for remote sensing, are used as calibration signals. Moreover, our methods solve the ambiguity problem of analyzing a phase shift which occurs when sinusoidal signals are used as calibration signals. In regards to the algorithm, the Adam learning algorithm avoids learning in the wrong direction, unlike the conventional gradient descent. Using our method, mathematical forms of received signals are acquired successfully. Our method shows better effectivity compared to the conventional gradient descent algorithm. After compensation, the maximum differences of gain and phase become 0.06 dB and 2.42 degrees, respectively.