Tactile surfaces are commonly found on a big number of devices nowadays. Generally, this kind of interfaces lacks haptic feedback, which lowers the efficiency and perceived quality of human interaction. Time reversal of elastic waves has been studied and implemented for vibrotactile stimuli generation achieving stimuli within the tactile range. Generally, manual impulse generation methods are used to create the signals that are reversed in time, but these methods are not very efficient nor repeatable. This paper describes the development of a linear impactor and an automated dataset acquisition system to enable the future development of novel methodologies for haptic feedback generation on surfaces using deep learning algorithms.