Other Parts Discussed in Thread: TPS564201, LMR51635
Tool/software:
Warning: I'm an amateur, working on a home project. My problem seems closely related to this old thread: https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/1241239/lmr51430-switch-regulator-keeps-burning-out-at-vin-net
I'm using the LMR51430 as a buck converter: power will eventually come from the 12V automotive battery on a fixed engine (stationary generator), and I step that down to 5V/3A for a Raspberry Pi Zero plus a cell modem. I made myself a circuit diagram (little snippet below) and a 2-layer PC board design (also below). Ordered three LMR51430 chips, and the first two have burned up on the first prototype assembly. As the board isn't otherwise populated yet, there is no load connected to the regulator, other than one 3mA LED. Right now, everything is being powered by an HP bench top lab supply at 14V. Both chips survived at most a fraction of a second!
Symptoms: First chip was fully shorted (about 1 ohm) between Vin and SW. The second chip has about 100 ohm between Vin and SW, about 200 Ohm between Gnd and Vin, and about 160 ohm between FB and Gnd. Haven't measured diode voltage drops between pins, but it probably doesn't matter.
I know there are several minor mistakes in the circuit diagram and PC board layout. First, Cin is way too large, 22uF (of type X5R) instead of the desired 4.7uF. Second, the high-frequency 100nF part of Cin is missing in the circuit diagram, but I added it to the prototype (by soldering an 0603 on top of the 22uF 1210). And looking at the layout, there should be more copper plane around the chip, and the two feedback resistors are too far from the FB pin. But is that enough to outright kill the chip, twice in a row? On the good side: All power traces are at least 0.5mm wide, and most of them a full 1mm. Ground plane both sides.
The obvious suspicion is transients on the input. The good news is that the input power is very clean: A lab supply, turned on by a PCB-mounted toggle switch, and then less than 2" of trace to Vin. The bad news is that the TVS diode on the Vin line (an SMBJ20CA-13-F) hasn't been populated yet. And before anyone even asks: No, I don't have scope traces of what happened. Since I only have one chip left, sacrificing the last one on that altar seems silly (plus, I'd have to borrow a storage scope). But I'll get more hardware ordered early next week (Happy New Year to everyone, by the way). Could the toggle switch be at fault?
I already have the following items on my to-do list for the next spin of the board: Change Cin to 4.7uF + 100nF, move the two feedback resistors to be right next to the FB pin, move the 100nF output cap to be right next to the two 22uF caps, and add a lot more copper and vias. Perhaps even add a 33pF capacitor to the feedback line. Any other good suggestions to make this work? Maybe move to a different chip? Since I hand-solder, SOT23 is pretty much the smallest I can do, and I want to stay away from chips that have pads underneath; but the TPS564201/8 and their 5A friends look like fine choices too, and give me some headroom on current. Are they more forgiving? Should I add something on the Vin line for transient management? I have ample room on the board, so even an active chip would fit, but I'd need a concrete suggestion.
And if this mishap ever happens in production, it will kill everything downstream of the LMR51430, so I'm thinking of adding a ~2A fuse (through-hole, the size of a 1/8W resistor) to its 12V input, and a crowbar circuit to its output; perhaps a few 6V 1A zener diodes. That would make repair easier and much cheaper.
Again, I apologize for being a clueless amateur, but finding information about good design is hard, so I thought asking the experts here might be useful.