A comparison of Fourier and POD mode decomposition methods for high-speed Hall thruster video

Abstract

Hall thrusters are susceptible to large-amplitude plasma oscillations that impact thruster performance and lifetime and are also difficult to model. High-speed cameras are a popular tool to study these dynamics due to their spatial resolution and are a popular, nonintrusive complement to in-situ probes. Highspeed video of thruster oscillations can be isolated (decomposed) into coherent structures (modes) with algorithms that help us better understand the evolution and interactions of each. This work provides an introduction, comparison, and step-by-step tutorial on established Fourier and newer Proper Orthogonal Decomposition (POD) algorithms as applied to high-speed video of the unshielded H6 6-kW laboratory model Hall thruster. From this dataset, both sets of algorithms identify and characterize m = 0 and m > 0 modes in the discharge channel and cathode regions of the thruster plume, as well as mode hopping between the m = 3 and m = 4 rotating spokes in the channel. The Fourier methods are ideal for characterizing linear modal structures and also provide intuitive dispersion relationships. By contrast, the POD method tailors a basis set using energy minimization techniques that better captures the nonlinear nature of these structures and with a simpler implementation. Together, the Fourier and POD methods provide a more complete toolkit for studying Hall thruster plasma instabilities and mode dynamics. Specifically, we recommend first applying POD first to quickly identify the nature and location of global dynamics and then using Fourier methods to isolate dispersion plots and other wave-based physics.

Nonlinear interactions of ion acoustic waves explored using fast imaging decompositions

“…Both techniques can provide insightful and complementary results. A recent preprint, reporting on the specific comparison between POD and 2D-FT applied to Hall thruster camera imaging 52 , concludes similarly. Applied to the present datasets, POD shows that the dominant physical structures are m-modes of the form e −iωt−imθ .…”