# Group delay of non differentiable unwrapped phase

As we know group delay as a function of frequency, computed as differentiation of unwrapped phase, but what happen at those frequencies where unwrapped phase is non differentiable?how to calculate delay at that particular frequency?

If there are phase discontinuities after all $$2\pi$$-jumps have been removed, then these discontinuities are usually caused by zeros of the frequency response. The phase jumps by $$\pi$$ at these frequencies, and the group delay doesn't exist, or, if one prefers, is non-finite. Note that the group delay is meaningless anyway at frequencies where the frequency response is zero.