Pilot-less Machine Learning aided Phase Noise Estimation for 5G mmWaves Practical Deployments

The increasing demand for bandwidth has led fifth generation (5G) mobile communication systems to adopt the use of millimeter waves (mmWaves), to enable higher data rates. However, operating at such high frequencies poses technological challenges, one of the most prominent being the phenomenon known as phase noise (PN), which degrades the system, especially when combined with the orthogonal frequency-division multiplexing (OFDM) waveform. Traditionally, this effect is corrected by using exclusively dedicated pilots for PN tracking, which severely degrade the data rate. This paper presents pilotless machine learning (ML)-based methods for estimating and compensating PN with the goal of eliminating the need for pilots, leading to a substantial increase of the data rates. In addition, this paper demonstrates the feasibility of these solutions for practical mmWaves deployments through tests with real measurements in an experimental setup using an mmWave 5G-developed platform.