The fast depolarization phase at the onset of the cardiac action potential (AP) is known to be auto-regenerative, i.e. cell membrane in this phase acts as a depolarizing current source for the surrounding cells. Auto-regenerativity during the late plateau phase of cardiac AP has been described in the past only in terms of all-or-nothing-repolarization, which is achieved during anodal current injections that make the membrane a repolarizing source. By means of a patch clamp protocol which combines current and voltage clamp and which I have recently simulated in a number of mathematical models of different cardiac cell types, I have demonstrated that late phase of cardiac AP repolarization can be intrinsically auto-regenerative without need of any current injection. Comparison between different APs is made straightforward by using 3D time-voltage-current surfaces obtained through this simulated protocol. The time window when membrane becomes a repolarizing source, which I have named Auto Regenerative Repolarization Phase (ARRP), is not always present in all cardiac cell types, and is shown to be significant for what concerns propagation and entrainment of cardiac repolarization. I present here simulations on several cardiac human cellular AP models which include SA-node, atria, Purkinje fibers, and ventricles (epicardial and M - type). I show that only ventricular cells are endowed with ARRP, which furthermore is differently distributed across the ventricular wall. I investigate the mechanism underlying spatial ventricular gradient of ARRP, and extensively discuss physio-pathological consequences.