The data sheet for the FPF3042 indicates that V_IN_SEL is bidirectional. When EN is low, then Vin_sel becomes an output, which can obviously be ignored. The relevant truth table appears on page 5 of the data sheet. On both v1.2 and v1.4, EN and DF_IN are tied to ground, so the chip will prefer V_IN if both power lines are above V_UVLO. Otherwise it will take whichever one is above V_UVLO. The choice will be indicated on V_IN_SEL.
It looks like FPF3042 is bidirectional between its Vout and Vin / Vbus pins(!), depending on which input is selected. (3 paragraph of datasheet description section.) So I could imagine a chain of events like:
- Attach USB & External power, in either order. Probably everything is fine.
- FPF3042 allows external power onto VCC_INT, all the capacitors on VCC_INT up to the external supply voltage. I count up ~70uF on VCC_INT.
- External power is removed. FPF3042 detects this, disconnects Vout from Vin, and then connects Vbus to Vin (bidirectionally!). A bunch of capacitance at 12V (or more!) is then connected to VCC5P0_USB. Yeah, C272 & C375 (10uF + 100nF) can try to absorb some of it, but it's already charged to 5V. The rest will try to flow up the USB cable (host is probably protected), and through FR38, the ferrite bead reported as damaged.
I'm not a real EE, and don't even play one on TV, and what I've described may not even be possible/damaging. But whenever my Lime shows up, I will do the following:
- Remove USB before external power. Perhaps arrange for USB to not even provide power if I'm going to use external, that should keep the FPF3042 from ever switching over to it.
- Keep external power low. 6V should be plenty for the board; anything more is just that much more heat being generated on the expensive SDR board.
- Definitely be careful of the power supply attached to external power. If it has excursions above the overvoltage limit (14.2V), the FPF3042 will disconnect it (at the worst possible time!), and you'll probably never know.
As for this fixation with how the driver doesn't have an option to "safety eject device", as far as I know that's entirely there for USB storage media, where portions of the filesystem may be cached in memory. I've never seen it used or referenced for anything else. And I don't see any control lines from the FX3 into the power subsystem, so there's nothing the hardware could do the power even if you told it you were going to disconnect it.
Absent the LMS7002 needing some shutdown process (I don't know, I haven't read the data sheet), I don't see what need the hardware could have for being "safely ejected", and suspect that whole line of thought to be a distraction.