All these would be infinitely close in the nonstandard characterization. I just picked logistic because it was easy and step is discontinuous so it shows off the approach’s power. If I started with delta instead I would have done Gaussian and integrated that and ended up with erf.