Seems that cause is spikes due to long and thin tracks where pulsed current flows. This is common mistake in first SMPS. Spikes caused overvoltage and/or IC latch.

@Vladimir thanks. How wide would you advise tracks should be? Note that the drivers blew with the CTRL pin floating, so the drivers should have been 100% duty cycle (no pulsing).

Track inductance also matters. Main idea is to keep tracks with pulsed current as short as possible, contours with current as narrow as possible. Try to search dc dc layout suggestions, there are many articles as from IC manufacturers as engineers.

They recommend 33μH to 100μH. Why 330μH? You should view waveforms on scope and post results.

@Vladimir thank you! I'll read up on that. I'm a mechanical engineer but definitely in the range of "don't know what I don't know" for electrical stuff - always looking to learn.

@StainlessSteelRat good question. For several channels, input voltage is much higher than LED voltage (48v vs 18v), and I was worried about efficiency. The datasheet says "Higher inductance is recommended at higher supply voltages", which I read to mean that 33-100uH is generally sufficient but that greater inductance could be used to boost efficiency. Was this a bad interpretation? Afraid I just dabble in electronics so I don't have a scope. Also, the circuit is cooked and I don't have spare drivers so there's not much probing I can do at the moment (power supply instantly trips now).

I'd interpret it as closer to 100μH than 33μH. I always check out eval boards for what the design engineers recommend. AL8862EV1 USER GUIDE Note track widths. They use 68μH and 1000mA. And a scope is essential for this stuff. Always populate 3 protos for when the magic smoke is released. And never go full load on initial power.

@StainlessSteelRat thanks. I guess I'll switch to 100μH in my next iteration. Yeah, looks like their tracks are about twice as wide as mine (and much shorter). I'll definitely beef mine up. Is there a short answer why trace width is important (outside of thermal concerns)? It looks like my traces have inductance on the nH range, which is nothing compared to the μH inductor I'm using. Maybe I'll see if I can use a scope in my school's EE lab. What signal would you look at? And yeah, I should have started at 0 and slowly ramped the power, but I was being lazy.

Forgive me this stupid idea, but anode (3-7) and cathode (8-12) connections have reverse order. If you would accidentially interchange the LED connections, current sensing would fail and the AL8862 would allow 3A to flow with around 4W internal heat.

I just noticed their prototype board has two caps in parallel between VIN and GND - why would that be? The driver's datasheet only has a single cap.

There's no reason you should target or use a 40°C temperature rise for tracks you have the available space to make wider. The reason to make tracks wider is to lower impedance.

@Jens are you talking about the connections on J1? That's a good thought - let me verify I haven't mixed up the wires there.

@JYelton I made my traces an order of magnitude larger than was required for 40C temp rise. IPC-2152 says the traces will experience <3C temp rise. I thought that was already overkill, but I now see I was mistaken.

You want a big one for sustained demand and a small one for instantaneous. Also the further the source is from board. If you are close to source, then all you need is a small one since power supply will have it's own.

Check out calculator on website Use it to size your inductor, diode.

@StainlessSteelRat thanks - wonder why their datasheet doesn't mention needing dual caps. Do you think that's crucial, or could I get by without it (assuming a decent power supply)? I didn't think to check their website for more docs beside the datasheet! That calculator is very helpful and suggests a 427uH inductor. Either they forgot to code a check to make sure it's in the 33-100 range in the datasheet, or larger values are okay. However, I do see that the theoretical efficiency is actually better with a smaller, lower-resistance inductor (at the expense of slightly greater ripple).

Put it on pcb. You don't have to populate it. If you find you need it, you have it. Governing factor is distance between chip and power supply (and quality of power supply). Better quality manufacturers have eval boards and models.

Thanks. There's ~2m of wire between the board and PS (and it's just a budget Mean Well LRS), so I expect a second cap would be beneficial. Any suggested reading to figure out sizing guidelines for the pair of caps?

That via under the IC also seems highly suspect to me. If I'm reading it right, that via connects the SET pin to the sense resistor and LED+. If the solder from the ground pad has infiltrated into that via (e.g. if it wasn't fully masked off) it would be shorting SET to GND.