Cavitation is among the most severe problems that arise inside fluid power components. It leads to performance reduction, material erosion, and noise. Since new standards and regulations impose reductions on both energy losses and noise generation, cavitation detection, and its reduction are becoming increasingly important for every hydraulic component, even the ones that are notoriously prone to cavitation, i.e. proportional valves. The present work proposes an in-depth study of cavitation inside a 2-way, 2-position spool valve through the analysis of noise generated by cavitation occurrence inside it. The valve has been placed inside an acoustic enclosure to isolate the background noise, and high-frequency acoustic signals have been recorded under different valve operating conditions. The derived Sound Pressure Level (SPL) signal has then been post-processed by means of modal analysis, more specifically, by applying a Proper Orthogonal Decomposition (POD) on the signal in the frequency domain, to recognise characteristic patterns that can arise from the cavitation regime. Through the analysis of acoustic measurements, this study aims to find a reliable method for cavitation detection and to distinguish different regimes of cavitation conditions (incipient, full, and severe cavitation). The work is part of an ongoing project that aims to develop techniques that allow to detect cavitation and to automatically control valve components in order to avoid this unwanted phenomenon.