Machine Learning-Based Calibration Algorithm for Harmonic Rejection N-Path Mixers
Résumé
This paper introduces a calibration approach for an RF front-end harmonic-rejection N-path mixer. Designed for low-power multi-standard receivers with a broad bandwidth, this mixer prioritizes simplicity. However, inherent variations in the fabrication process can lead to gain and phase mismatches within the system that dramatically reduce the performance of the mixer. To mitigate this, this work proposes a machine learning-based calibration method to safeguard against these nonidealities and maintain optimal system performance. To verify the feasibility of the approach, a 0.17-1.2 GHz RF front-end with calibration capabilities has been fabricated in a 28 nm FDSOI technology. After calibration, it achieves a harmonic rejection higher than 45 dB for the 3 rd , 5 th , and 7 th LO harmonics with a conversion gain of 13 dB and a total power consumption of 22 mW.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |