Automotive input protection, a recap

[liteyear]

I'm doing some blank sheet designs for a couple of doodads that will be powered from a 12V vehicle supply. I figured this was probably a paint by numbers exercise by now, but no, it seems not.

Here's what I found:

• ISO 16750-2 defines the quintessential test voltage profile.
• 
• There's a bit of doubt the load dump section of this is relevant any more: few vehicles use an unregulated alternator anymore, and there is a lot of distributed surge suppression in a car's electrical network already. But certainly it can't be ruled out.
• Good discussion on that here.
   
• Some OEMs now follow LV 124 instead.
• In that standard, load dump peaks at 27V!
• Some good discussion here.
• Some silicon vendors claim that protection is not necessary any more, because their SMPS controller ICs have such a wide input range. Of course, that assumes these ICs are satisfactory for your design.
• Or deal with just the negative voltage part without a TVS, using the LM74701.
• Texas Instruments's SNVA717 from 2014 still seems to be going strong, and recommends a series diode and a crowbar from some dozen discrete components.
• 
• Analog Devices has a guide from 2020 that recommends an all-in-one IC... plus a 4-switch, two dozen discrete solution.
• 
• A couple of vendors have TVS that people report as satisfactory for automotive, though the datasheets themselves usually fall short of making specific claims, making it hard to do parametric searches.
• DIOTEC MR2525
• "5kW" parts like the YAGEO 5.0SMDJ and ST LDP01-xxAY (whose datasheet *does* make a specific ISO16750 claim).
• The 7kW and 11kW PAM7KSMDJ24CAP and PAM11KLD8S24CAP from ProTek.

But what I really find interesting is that as soon as you add a TVS, you inevitably have an unprotected region.

For example, suppose you have a 5A load and use the 20Vwm TVS, 5.0SMDJ20. By 22 to 25V (at 25°C) it will be conducting 1mA. By 32V it will be conducting 155A. Between those voltages it's a bit hard to be sure. But the TVS is only good for 6.5W continuous, so once it starts conducting a quarter of an amp, it will cook itself if sustained.

Worse, with such a low Vwm, your doodad becomes the surge arresting sink for the whole vehicle! You could easily get hundreds of amps flowing through the PCB and connectors and supply wiring, and need to design accordingly. Some suggest going for as high a Vwm as you can withstand, so you're only a surge dump of last resort. But that makes the power dissipated by the TVS in moderate over-voltage even higher.

In theory, of course, this situation will not be sustained. In practice, it's a pity the protective measure cannot protect itself. A 5A fuse will not blow. Even a 100mA fuse might not blow quick enough if there's only 250mA flowing.

The thing is, without a crowbar it seems a TVS in this configuration will always have this vulnerability. Maybe you could thermally couple the fuse and TVS, but it would be difficult to fully qualify. If you choose a PTC resettable fuse instead, and can get away with a very low rating, it can actually regulate instead of "blow" like a badly behaved resistor-Zener combination!

It took me a while to convince myself this fairly simple conclusion is indeed the case. Am I missing something? Is a "standard" TVS front end inevitably susceptible to a sustained over-voltage scenario? Is there an alternative standard front end that's a bit closer to set-and-forget, without requiring twenty discretes?

[Psi]

I use a 24V 128.5A 5000W SMC package TVS that clamps at 38.9V (Nice voltage as it clamps before 40V regulators fail).
There's no crowbar in the design. Power input is just the TVS then it splits in two. One feed to 2x 10uf caps in series and an automotive grade 5V reg. The other to an automotive high-side load driver.

The TVS size might be a bit overkill, but I've not had a single return where the TVS is popped.

This is the TVS i use.
https://www.digikey.com/en/products/detail/comchip-technology/ATV50C240JB-HF/5013675

The automotive product drives a 10A load so has power/ground wiring that can handle 10A continuous. (16 or 18AWG from memory.)

[liteyear]

I use a 24V 128.5A 5000W SMC package TVS that clamps at 38.9V

Nice. That's almost exactly where I ended up. Experiences like yours lead me to believe that the "sustained overvoltage" scenario is simply not worth worrying about. Not because it's impossible, but because it's so rare and hard to handle that blowing a TVS is an acceptable outcome. I'm sure you could come up with a double jump start scenario that will cook that TVS. But in practice, not worth worrying about?