The environment in a car is quite different from that experienced by electronic components in a box on a table in your office. There are several characteristics in particular that stress electronics components more in a car than a office environment:
- Temperature. A car still has to start on a cold winter morning in Nome or after baking in the hot July sun until mid afternoon in Yuma. Not only do these extremes pose a challenge to operating ability, but the daily thermal cycling causes mechanical stresses.
- Vibration.
- Noisy power. "12V" car power is nowhere near as nice as what a 12 V regulated power supply provides. It can drop significantly during starting, goes to 14 V when the engine is running, and can have 10s of volts of nasty spikes on it. For example, the normal 7805 regulator can be damaged by car power, even though it is rated to 35 V. There is a special regulator series that is rated for car power, which of course costs more.
- Humidity. The relative humidity is unpredictable and can be at one extreme or the other for extended periods of time.
In addition to this, the end users generally don't tolerate failures as well in cars as they do in ordinary office appliances.
I can't say how much of this is relevant to your situation since you haven't told us what you are really trying to do. However, keep in mind that there are millions of cars sold every year, so the volumes are high and there is considerable competition. If manufacturers didn't have to pay for the extra robust characteristics of the components, they wouldn't. The fact that a car company is willing to pay the extra $.50 for a relay that goes into a million cars, should tell you that it matters. That's $500k they could save if they didn't. Put another way, that's about 3-4 engineers for a year to come up with a cheaper alternative.
My guess is that all the 10nF or 100nF recommendations are for zero-crossing TRIAC applications, and that one has to be really careful to not apply that recommendation blindly to relay-based solutions.
So, I propose the following possible solutions:
1: RC snubber designed according to the calculations in How to calculate resistor and capacitor size for snubber circuits . This protects the relay from arcing and reduces EMI problems.
2: 100nF and 100 ohm RC-snubber in parallel with a MOV. The snubber doesn't protect the relay, but may reduce EMI-problems since such a snubber forms an RC low-pass filter with a 100kHz cutoff.
3: Just a MOV. May cause EMI-problems?
MOVs have a limited lifetime and must be sized appropriately to have enough longevity for the application. They may not be a suitable solution if the motor is to be turned off frequently. Placing the MOV across the relay protects the relay better, but means the MOV is always energized even when the motor is off. A short-circuit failure of the MOV will then start the motor.
The MOV may be changed to a bidirectional TVS diode. IEC 60950-1 (if it is applicable to the application) explicitly forbids TVS diodes for surge suppression. If the TVS is connected across the relay, it is basically connected the same way as a surge suppressor, and may thus not be allowed(?). IEC 60950-1 also states that MOV:s, if used, must be used together with "an interrupting means having an adequate breaking capacity", presumably a fuse.
Best Answer
You don't give much information about the transil nor the MOV, but it sounds like the stored energy in cable inductance is just too high for the MOV and transil to absorb. If there appeares to be a short when you insert your MOV, the most likely reason is that the MOV is shorted. They do fail into short when overpowered so it makes sense.
If you have selected the voltage rating on the MOV to an appropriate value, you just need to increase the energy rating or use several in parallel. The first option is better.