[GriffinRU]

Originally Posted by Pdelcast

You'll find there are much better (faster and more robust) solutions than TVS.

Notice you never see those high voltages (115V) in a MMM or MM, because the body diodes clamp WAY before avalanche occurs. But there are several improvements that can be made beyond what the MMM has. First -- the body diodes do clamp to rail +- 1.5V or so, but they are slow. The snubbers are faster, but can't handle enough power -- bigger snubbers make a huge difference.

The whole idea is giving that energy a place to go that is LOW resistance path, not a high resistance path like a TVS. That energy of the inductor doesn't have to be wasted in a resistive part -- it can be recaptured, re-routed, restored.

That is where software comes to play...
Hundreds of ps are not that slow (it is possible to get faster devices but then substrate size (to maintain current capacity) will increase capacitance, thus slow you, limit you) , and there is no need to clamp all the way to zero, just to the safe Fet's level, and fet's are not that fast. But overtime both of them degradate, TVS first (easy to replace) fet second. Temp also change their response. Diodes are also rated to max Fet’s voltage, they are by-product of Fet. So you will exceed diode ratings if you do not clamp it earlier with something rated at higher voltage. You got the idea.
There is no much energy there to recuperate, but plenty to damage. Just switching fronts I am talking, nothing else.