Hmm, yeah, now I read the datasheets, it's longer than I assumed - in the order of seconds. For example, a Bourns MF-R020 I was looking at is just over 2s @ 1A, and approx. 10s @ 400mA.
Yup! Meanwhile, the TVS is burning, say, 26V at 400mA+ (>10W) for those seconds.
TVSs usually fail shorted (until you reach their own, ahem, fusing current...), so the TVS can still manage to protect the circuit, sacrificing itself in the process. The fuse later clears the short, preventing a fire hazard.
It's a good system, but for almost everything, it's a "your circuit board is toast anyway" situation. Sure, you can repair it, but it's not usually worth doing.
So, the arrangement I want is: 1N4007 > fuse > TVS?
Yes. Or...
On the other hand, if you have an ESD event, the diode might be gone when the TVS starts clamping the voltage.
I've also seem huge diodes blown up when reverse polarized. It was a badly designed crowbar. Not pretty.
I would use bidirectional TVS and put the diode after the fuse and the TVS.
Either way has valid reasons!
If you're expecting a suitably dangerous amount of negative transient/surge, the series diode can get toasted. It does allow a higher reverse rating, though -- if you
really needed to handle a big reverse surge, the nice thing is, you can use a 1N4007 (which hardly drops any more Vf) to get a 1kV reverse rating, then put on basically any value MOV (say, 50V to 300V), and have that reverse surge handled better than anything.
You can employ a similar strategy in the forward direction, but you have the added difficulty of limiting the circuit's load voltage (which the LDO does to a point, but if you need considerably more than 30V, well...). If you can get a huge compliance range, though, you can do away with pricey TVSs and use a big dumb MOV there -- and gulp down mondo surges like it's nothing!
Probably not something you'd do for most any automotive application, but it would be an excellent approach for high-reliability and mission-critical (aerospace, military) gear.
Is there any point me using an LM2931 now I have a discrete reverse-polarity diode and the TVS? I initially chose it because of its built-in reverse-polarity and load dump protection, but I guess this is redundant now I have external counter-measures (because I want some 'safe' 12V as well, not just for 5V). Would I now be just as well-off using a cheap LM7805/LM78L05?
Well, an LDO will get you a lower minimum input voltage -- maybe good for low-battery or cranking conditions. If you don't need that, a regular 7805 isn't bad.
Why do radial through-hole PTC fuse packages have such a large lead stand-off height from the board?
Remember how they have to heat up to "open"?
SMT polyfuses are dubious components. Current rating depends on how they're wired up -- you want to avoid connecting to them with heavy copper pours, because you can easily double the trip current by dissipating heat. (And that's on top of the already-poor tolerance on the trip current, which is all the more reason to remember -- they're for fire protection, not protecting the attached device!)
You also want to avoid placing them near super-hot components (resistors, heatsinks?), for the same reason -- if the trip point is when the active PTC-stuff reaches ~150C, that means your trip current also goes to zero, as ambient goes to 150C! If you need an 85C maximum ambient, you need twice the (trip at 25C ambient) rating just to keep the thing running normally!
They're also rather unreliable, so don't depend on them for anything critical. It's a CYA component -- as a fuse should only ever be.
If you need limiting, put in a current-limiting or protecting circuit -- and expect to pay for it as well (reliability isn't free!).
Tim