Rather than discuss the technical details of transient protection in MMD,
Vince Negro and Laurent Coray have communicated with each other privately
and decided to sum up this subject as it appears in the following.

For simplicity, we consider the most common case where the transient
protection is across the coil of the relay or solenoid in a direct
current (DC) system.  However, there are other configurations that can
be easier to implement (e.g., protection across the switch or relay
contacts).  Laurent is thinking of doing a web page on this entire
subject and will detail these different configurations.  The only theory
we will mention here is that when an inductive load (relay/solenoid) is
switched off, the coil will generate a voltage that tends to keep the
current after switching the same as the current before switching.

Case 1.  A diode across the coil.

This is the simplest solution.  However it also delays the release of
the solenoid or relay the most because the low voltage of the diode
doesn't contribute much energy dissipation; in a 12V DC system, the
diode only dissipates about 1/20 of the energy.  So nearly all of the
energy is dissipated by the coil resistance.  If fast release is not
needed, this is a good way to go.

Case 2. A diode plus a series resistor across the coil.

For example, consider a resistor 9 times (9x) the coil resistance.  Since
our simple theory tells us the current remains the same, the voltage
across the 9x resistor is 9 times the voltage of the coil resistance.
This combined with the constant current yields an overall energy
dissipation that is 10 times faster.  The energy dissipated by the
diode in a typical 12V DC system is about 1/200 that of the 9x resistor
and is ignored.

Case 3.  A diode plus a Zener diode across the coil.

In the example above, the voltage across the 9x resistor is 9x12 or
108V.  So for this case let's use a 100V Zener diode. Here too, the
energy dissipated by the diode is about 1/200 that of the Zener diode
and is ignored.  Although our simple theory still holds,  it is not
sufficient for this analysis.  For some typical values of dropout
current, this arrangement is from 1.5 to 2 times as fast as Case 2.

Laurent Coray & Vince Negro

 [ Good article, Laurent & Vince, and a technical web site is a
 [ fine idea.  Please keep us posted.  -- Robbie