A model based soft sensor that estimates the location of the in-cylinder pressure peak from the ion current is described. The soft sensor uses a neural network algorithm and has been implemented in a SAAB 9000 low-pressure turbo production car. It estimates the pressure peak location, in real time, during normal highway driving with an error of 2-3 crank angle degrees. The soft sensor has been tested during normal Scandinavian weather conditions, with a relative air humidity of about 50%, as well as when water is sprayed into the intake manifold, resulting in approximately 100% relative humidity. The neural network based soft sensor is significantly better than that of another method, based on nonlinear Gaussian curve fits, for the same task.