Electric arcs pose an ignition hazard in the presence of flammable gas mixtures. Electrical equipment to be used in such hazardous environments shall therefore satisfy strict safety requirements, through the use of internationally standardized explosion protection methods as e.g. "Intrinsic Safety", documented in the IEC 60079-11 Standard. This "Intrinsic Safety" is verified by using a stochastic empirical procedure, using a "Spark Test Apparatus" connected to an electrical energy source. The apparatus generates electrical discharges between a separating tungsten anode and a cadmium cathode enclosed in a test cell filled with flammable gas atmosphere. The connected electrical circuit is considered intrinsically safe if no ignition results within a defined number of contacts. However this procedure suffers from variability and poor reproducibility. The goal of this work is to investigate the relationships between such arc discharges and the ignition of the gas. The measurement is technically challenging, as the physical processes occur in different time scales (ns, mu s, ms) and the physical dimensions are small (gm). Additionally the ignition process can also damage the experimental equipment. An important prerequisite for such discharges is a sufficient degree of electrode surface wear. Microscope images show such electrode surfaces as well as the presence of metal whiskers. These inhomogeneous rough surfaces are compared by the means of Abbott curves. A spectral analysis of the radiation from the electrical discharge shows, that the main substance is cadmium vapour. The electrical characteristics of these arcs are characterised by voltage, current and length curves. For constant currents from 74 to 270 mA up to a voltage of 40 V, the transient arc lengths, voltages and currents were measured for arc discharges generated in a 21% Hydrogen-Air gas mixture. These initial results appear to correspond qualitatively to published curves for static arcs, however the accuracy of the measurements has to be improved. Knowledge of these relationships between electrical, mechanical and ignition processes will ultimately make it possible to recognize discharges which most likely cause an ignition by their current and voltage waveforms. This will allow the development of a more reliable alternative, where discharges are electronically simulated. (C) 2017 Elsevier Ltd. All rights reserved.