A detailed analysis of the non-deterministic scales of the flow in low-speed axial turbomachinery has revealed the presence of significant large-scales unsteadiness, with periodic features different to the blade passing frequency (BPF), which are superimposed to the turbulent structures. Introducing a frequency-based decomposition, this additional component has been segregated from turbulent phenomena so the total unsteadiness has been found to be contributed by three components: forced unsteadiness (deterministic scales), unforced unsteadiness (large-scale unsteadiness) and turbulence (small-scale randomness). Dual hot-wire anemometry has been employed intensively within the stage of a low-speed axial fan to provide a valuable experimental database, where the phaselocked averaging technique has been applied to retrieve time-resolved fluctuations and isolate the non-deterministic contribution. The present investigation shows that the presence of the unforced component is mainly related to instabilities of the rotor wakes and tip vortex structures, as well as wake-wake interactions. Moreover, typical eddies size of this component distribute their energy within the frequency range that is containing an 80% of the total unsteady kinetic energy. As a consequence, it is expected that LES schemes with accurate spatial discretizations may address this component within the resolved scales of the filter, while unsteady RANS modelling could require additional modelling of the unforced mechanisms. In addition, maps and radial distributions of every component illustrate that major flow patterns are identifiable in all of them, due to the redistribution of all-range scales throughout the energy cascade. Turbulent and forced mechanisms present important variations with the operating conditions, while the unforced component is barely affected by flow rate variations. It is shown that typical values of unforced unsteadiness reach up to a 20% of the total unsteady energy, even for nominal conditions at midspan of the rotor passage. Higher levels of all the components are found towards tip and hub boundary layers, as the total unsteadiness is reinforced by massive flow separations, tip blockages and major flow disturbances.