The potential drop technique is a powerful method to monitor the crack length in fatigue research. The DC potential drop method relies the fact that a constant DC current is passed through the test specimen in such a way that a change in crack length alters the potential difference of suitably placed potential probe points. The technique is simple, robust, highly stable and suitable for automation. Disadvantages of the technique are: a calibration curve is necessary (simple experimental or complex numerical calibrations), if crack faces short an underestimation of the crack length is obtained, thermal electromotive forces can cause problems and high DC currents can cause heating of the specimen. The heating effect of DC currents can be lowered by using a pulsed DC current. This pulsed current is applied to the specimen just before and during the determination of the crack length. Crack underestimation can be eliminated by measuring at the maximum opening of the crack. Instead of the normally used potential drop difference between two pick-up points, two potential differences are used. This eliminates the effect of variations in the current density. Influences of electromotive forces can be eliminated by measuring the potential difference with current on and substracting potential differences with current off. Measured potential differences are divided and used as a function of the crack length. Requirements for a suitable calibration are: 1) accuracy or how well does the calibration curve approximate the relationship between potential change and crack length, 2) reproducibility or the effect of small probe variation, 3) sensitivity or the experimental measurement of potentials, and 4) measurability or the electric noise levels. In this paper optimisation of the four potential probe and current input positions is made by means of finite element calculations. The numerical obtained calibration curve is compared with an experimentally obtained calibration curve. Transactions Engineering Sciences vol 13, © 1996 WIT Press, www.witpress.com, ISSN 1743-3533