A common method of generating a tensile wave in Split-Hopkinson bar experiments is to impact a hollow striker on an anvil/collar attached to the incident bar. When the anvil/collar is attached to the incident bar using screw threads, the incident signals usually deviate from the classical trapezoidal form. In order to minimize this deviation, it is a common practice to sufficiently tighten these threads. However, the fundamental reasons for (i) the signal distortion when the threads are not tightened and (ii) how the distortion is eliminated or minimized on tightening are lacking. For a given diameter of the incident bar, the parameters of the thread that govern the incident signals are its pitch and backlash between the mating threads. The effect of these two parameters is investigated using the finite element method. The finite element results show that in the absence of a backlash, the incident signals obtained from a collar that is integral to the incident bar and a collar that is attached by screw threads are practically identical. When threaded collars are used, the pitch of the screw threads does not influence the incident wave. However, the backlash leads to a dramatic distortion of the incident signals under certain conditions. Insights from the finite element analysis are used to develop a one-dimensional analytical model that explains the distortion of the incident signals. The model is also validated against experiments. © 2022 BMJ Publishing Group. All rights reserved.