Quote Originally Posted by dpgoldenberg View Post
Here is another function that does include a shoulder:

f(x) = a1*log(1+exp(a2*x+a3))/(a4+log(1+exp(a2*x+a3)))
David, your curve implies that the shoulder is as abrupt as the toe, which is generally not the case. It also requires that you catch the complete shoulder region with your measurement data points or you risk numerical instability in your curve fit.

For these reasons I strongly advise against this model and recommend you take another look at my square model.