Hallo Stephane,

'I will add the schematic to my website soon if you want'
Yes, I'm highly interested. It's never to late to learn from a master...

By the way, I never measured such high pulses in my car applications.
Many years ago I designed active loudspeakers. I always used a circuit made of almost so many parts as yours: Directly at the terminals a big fuse, then (in direction to circuit) big transzorb diode (1,5KE18A) accross Vbat and 0V, paralleled by 47µF/100V and 100nF/100V (MKS2), then a RLC-lowpassfilter with 0.22Ohm (5W), 180µH (self constructed on ETD34 core) and 2 x 2200µF/25V, behind that a diode, back forwarded between Vbat and 0V to avoid negative potentials from RLC filter coming at certain, very unprobable situations for not having damaged chips coming behind. Behind these elkos additional fuses to different power amp chips were inserted and additional decoupling capacitors.

It was my thought that transzorb diode in combination with battery would be sufficient to absorb energy from opening inductive loads. Rating of fuse, about 3A was such that transzorb could never becoming open-circuited, even after destruction.

Bye,
Kai

PS.: Will have a look at your website...