Edge localised modes (ELMs) are quasi-periodic relaxation events, during which plasma is ejected to the walls, which is a major concern for future fusion reactors. Each ELM crash is preceded by transient density and magnetic field fluctuations inside the last closed flux surface, known as precursors. This paper presents a spectral wavelet analysis of ELM-precursors in mega ampere spherical tokamak. Type-I (totalling 77) ELMs and type-III (26) ELMs have been analysed. The analysed data originate from beam emission spectroscopy, Outboard Mirnov Vertical coils, and Outboard Mirnov Array High-frequency Acquisition coils. The precursor mode is detected at kHz. The wavelet cross-correlation analysis could indicate two types of precursors: a linear type with low wavenumbers cm−1, length scales and mode numbers ; a non-linear type with distinct ranges cm−1, and n > 10. Poloidal velocities tend to be of the order of km·s−1. The wavelet bi-coherence analysis suggests that precursors could arise from two types of non-linear couplings: kHz and kHz; kHz and kHz. The wavelet statistical analyses demonstrated that different tokamaks share generally similar precursor parameters.