2. Yea, IDK. I'm just trying to keep the 60-100V spikes from seeing my capacitors and LDO. That's all. I figure if there is an event of 60+ V and the TVS clamps, it'll blow the 10A fuse pretty quickly as correct me if I'm wrong, but that would basically be SpikeV - ClampV / source resistance right? Which is a lot and should blow the fuse quickly, maybe with in the 600W surge/spike my TVS says it can handle. I'm almost out of board space, so a large series resistor on the input might fit, but it's getting down there in terms of space. I'm open to recommendations!
Heh...
Once again: the semiconductor protects the fuse, not the other way around.
Load dump can deliver
hundreds of joules, mainly thanks to the lengthy duration, 100s of ms. This is a time scale where heatsinking is relevant! Or at least partially so.
That "600W" diode is only rated for that, on a 1ms surge. It's rated about 1W continuous (depending on what's attached to the leads for heatsinking). Somewhere between 1ms and infinity s, the rating is less than 600W, but more than 1W. Unfortunately it's not obvious where that is...
If we assume a square pulse, 600W * 1ms = 0.6J, quite a respectable amount of energy for a semiconductor to dissipate; but again, we expect the energy capability to be higher, thanks to the long duration giving more time for heat to spread out. (The continuous rating implies infinite joules at infinite time. Somewhere between, ... you know.)
All this to say: if you have the thermal resistance curves for the device, with heatsink (if applicable), you can calculate the power rating of, well, a square pulse at least, if not an arbitrary curve necessarily. (Load dump is done with an exponential curve, of which you're clamping a roughly square-sided segment, but the power still varies a lot over that segment; so it's not obvious how to calculate the square-wave equivalent, except with the help of calculus.)
So, given that I can't offer any conclusive calculation or data supporting this claim: I can safely say a 600W diode will blow up in a couple of milliseconds, let alone a hundred or more. And not just that it will fail shorted in that time, but I wouldn't be surprised if it melts open during the transient, too!
There's really only three ways of dealing with this:
1. Brute force. Use a
big fucking absorber. Even MOVs are bad at this, actually, because low-voltage types are thin -- there just isn't much material to absorb a lot of energy -- besides having quite sloppy clamping voltage, especially in the lower voltage ratings. TVS diodes are great, but you need
big ones to do it, like $20+ worth to do it.
So, this is really only a feasible method when the load is already very large, so that other methods aren't really applicable (or the load is mission-critical such that it can't be switched off, even under such intensive conditions), and the expense of this method is therefore tolerable.
2. Shut off. LT makes transient protector controllers, or if you don't want to pay for LT, there are a few discrete approaches (e.g.,
SNVA681A). Downside: you need a big enough transistor to handle load current plus peak input voltage, making this annoying for heavy loads. Also keep in mind the tradeoff for different circuit designs: P-ch MOSFETs have 2.5 times worse performance than N, making an N-ch based circuit that much more attractive for high power loads (thus justifying the added complexity of the charge-pump or bootstrap gate drive thus required).
3. Ride through. Get an even higher voltage LDO, or use a high voltage HDO with a bypass transistor somehow to save the operating voltage drop (maybe like #2?), or anything equivalent. This is perfect for light loads, where the dissipated power during the transient is quite manageable. A DPAK transistor will easily handle a 60V-peak load dump with a load up to, say, 0.15A or so. Don't use P-ch MOSFETs: you're basically reinventing a drain-output LDO in that case, and you'll have no end of problems getting compensation right. N-ch is seemingly harder to use, but don't overlook the possibility of depletion-mode MOSFETs, which are still modestly available today, and for just these sorts of reasons! They're also great for generally limiting current flow, and have good switching performance, too (though most controllers won't be suited to their peculiar gate voltage range, of course).
3a. You can get the best of both worlds, by switching the input on and off rapidly enough that it averages out. *Cough*, in other words, just another buck converter. Well, as usual, it doesn't need to be pretty -- it doesn't need to meet EMI requirements, it only needs to be quiet enough that it doesn't trip itself up. It doesn't need to regulate load voltage or current very well, nor deliver low ripple voltage. A crude hysteretic controller will suffice. The filter inductor can be cheap and crappy, maybe a powdered iron toroid or something. Often you'll have one of these in the circuit anyway, for EMI purposes (whether that's primarily against EMI inside or outside the device!). Switching doesn't even need to be sharp, as it's not doing continuous duty.
Basically, by switching at modest (say >80%) efficiency, you -- obviously enough -- pentuple (or more) the power capacity of the switch. Instead of a DPAK, you can use a SOT-89, or even less. Or that same DPAK can protect a load of 1A+ instead of less than 0.2A.
That's what I did here:
https://www.seventransistorlabs.com/Images/SwitchingCurrentLimitUnits.jpghttps://www.seventransistorlabs.com/Images/SwitchingCurrentLimitUnits2.jpgA pair of D2PAK transistors handles a +/-30V, 20A (i.e., up to 600W) load, for 150ms. In this case, excess power is dumped into a couple TVSs when a return path is not available. The three SMDJ (3kW) series TVSs excel at this, absorbing about a dozen events before tripping the onboard thermal limit, which cools down and is ready to go again after about a minute.
Tim