Calculation of the Performance of Communication Systems from Measured Oscillator Phase Noise

TL;DR

This paper establishes a mathematical link between oscillator phase noise measurements and communication system performance, providing bounds on error metrics and analyzing the impact of different noise types and bandwidths.

Contribution

It introduces a statistical model of phase noise, derives bounds on system performance, and analyzes how noise characteristics affect high-speed communication systems.

Findings

White noise far from carrier has less impact at high symbol rates.

1/f^3 noise is more predictable and less detrimental than 1/f^2 noise.

Performance bounds depend on the noise spectrum and communication bandwidth.

Abstract

Oscillator phase noise (PN) is one of the major problems that affect the performance of communication systems. In this paper, a direct connection between oscillator measurements, in terms of measured single-side band PN spectrum, and the optimal communication system performance, in terms of the resulting error vector magnitude (EVM) due to PN, is mathematically derived and analyzed. First, a statistical model of the PN, considering the effect of white and colored noise sources, is derived. Then, we utilize this model to derive the modified Bayesian Cramer-Rao bound on PN estimation, and use it to find an EVM bound for the system performance. Based on our analysis, it is found that the influence from different noise regions strongly depends on the communication bandwidth, i.e., the symbol rate. For high symbol rate communication systems, cumulative PN that appears near carrier is of relatively low importance compared to the white PN far from carrier. Our results also show that 1/f^3 noise is more predictable compared to 1/f^2 noise and in a fair comparison it affects the performance less.