Power Supply very hot and not functioning

I'm still a noob but trying to learn and improve as much as I can. I'm hoping someone here can set me on the right track to fix a schematic that went wrong somewhere.

I made a power supply sub-system (image below) for my PCB about a month ago. My project is powered by a 3.6V 13Ah LiSOCl2 battery. Battery-VIN is connected to D1. The purpose of this power supply system is to ensure that I would always have a steady 3.6V output even during high current draws (up to 800mA) and cold-weather conditions (down to -20C). Previous tests without this power supply system showed that under these conditions the battery-voltage could drop as low as 2.8V which would create frequent brownouts on my microcontroller (Particle Electron).


I based my schematic + parts on this datasheet (page 1).

Parts:
D1: SBR05U20LP-7
L1: MLP2012H2R2MT0S1
U1: TPS610995YFFR
C1/C2/C3: 0402ZD106MAT2A

Note that since VIN is 2.8-3.6V, and VOUT should be 3.6V, I went with the 610995 model with FB connected to GND. W1+W2 are two wires with a JST-PH2 connector that can be inserted into my microcontroller to power it.


My DRC clearance is set to 6mils, and object clearance is set to 10mils (as per manufacturer's recommendation).

My problem:
Yesterday, I finally received this PCB (10 copies). I tested it and while it worked for 3 copies, for the remaining 7 copies the parts became extremely hot and VOUT was 0. Clearly, something went terribly wrong. I looked at my parts again, and here's my thoughts:

(1) The maximum rating for D1 is 500mA, which is too little for my project. Nonetheless, just by powering up my PCB these components should draw nowhere near that amount.
(2) The saturation current for L1 is 1A, which ideally should be higher. But again, simply powering up my PCB should draw nowhere near that amount.
(3) C1+C2 = 20uF, but the maximum here can be 100uF. To better mitigate current drops, stronger capacitors should be used.

Basically, after seeing that my PCBs didn't work, I looked into the parts and came up with several things that can be improved, but nothing so far that could explain why my system entirely didn't work.

I'm really hoping someone here can give me a push in the right direction as I'm not sure what else to look at.

If three worked and the rest didnt, it suggests there may be some build errors such as a bridged track, capacitors or chips in the wrong way around.

which bit got hot- that might be a clue

Check for microscopic copper shorts (left over from manufacturing) on the PCB.
Your battery symbol top right hand side of the schematic is screwed up.

Show us a good image of one of the bad boards.

hammy:
If three worked and the rest didnt, it suggests there may be some build errors such as a bridged track, capacitors or chips in the wrong way around.

which bit got hot- that might be a clue

I tried to check thoroughly to figure out which part got hot - all of them seemed to get hot, but D1/L1/U1 definitely were worse off than the Capacitors. Since U1 is an extremely small part (the pads are only 6mils apart) do you think it's possible that the manufacturer (PCBA Store) screwed up here? I find it unlikely since they assured me 6mils should be fine (and it's a very common clearance that's used) but anything's possible I guess.

larryd:
Check for microscopic copper shorts (left over from manufacturing) on the PCB.
Your battery symbol top right hand side of the schematic is screwed up.

Show us a good image of one of the bad boards.

I'll take a better picture. I made the battery symbol myself but messed up the spacing of the pin names; the schematic is correct but the names are on the wrong sides - will fix it.

Shoot, the image in the OP is the best my phone camera can do. I can't get a better picture for you guys right now.

hammy:
If three worked and the rest didnt, it suggests there may be some build errors such as a bridged track, capacitors or chips in the wrong way around.

which bit got hot- that might be a clue

Tested again.

C1/C2/C3 - slight temperature increase
D1 - slightly warm
L1 - hot. still bearable to touch.
U1 - extremely hot. can't touch for more than a second

Check for PCB shorts!

Check for proper capacitor polarity.

larryd:
Check for PCB shorts!

Check for proper capacitor polarity.

Thanks. As far as I understand, 0402ZD106MAT2A (C1/C2/C3) does not have polarity although I mostly base this off of the absence of such information + lack of markings on the physical part; I can't find any conclusive information about it. It is also my general assumption ceramic capacitors don't usually have polarity.

I'll search for possible shorts. U1 is my prime suspect since it has the most pads/pads are closest together.

C1-3 are labeled as 10 µf.

99.9% chance that these will have a polarity!

Appears they may not have polarity markings, try removing C1 and C2 on a bad board.

MLCC capacitors:
These capacitors can cause anti-resonance effect because of their low ESR.
“The low ESR is a feature of the MLCC, but it is so much lower compared to an aluminum electrolytic capacitor that on the contrary, the output voltage of the DC-DC converter becomes unstable and causes oscillations to occur.”

You may have to try a non MLCC type of capacitor.

larryd:
C1-3 are labeled as 10 µf.

99.9% chance that these will have a polarity!

Appears they may not have polarity markings, try removing C1 and C2 on a bad board.

MLCC capacitors:
These capacitors can cause anti-resonance effect because of their low ESR.
“The low ESR is a feature of the MLCC, but it is so much lower compared to an aluminum electrolytic capacitor that on the contrary, the output voltage of the DC-DC converter becomes unstable and causes oscillations to occur.”

You may have to try a non MLCC type of capacitor.

Removing them didn’t fix the boards unfortunately. How could a polarized part not have markings though? I had these boards assembled by a fab house, but if there’s no way for them to identify the polarity then who’s able to commercially use these parts? Just to clarify; it’s possible for SMD Ceramic capacitors to be polarized? I thought they always came as non-polarized.

Anyway, I’ll make sure to use non-MLCC capacitors to avoid any potential issues.

Couldn’t see what capacitors you were using in the image.

If they are MLCC, MLCC capacitors have no polarity markings.

However, you may be experiencing oscillations as metioned, this can account for things getting hot.

You can be having problems with copper shorts as mentioned also.

Here's the capacitors (same used for all)

Thanks for bringing the oscillation to my attention.

Since these boards were assembled by the fab house, I expected the margin of error to be very low (especially when using their recommended clearances). Still, I decided to swap out U1 for the exact same chip but with a WSON package (instead of BGA). I'm hoping that this will significantly reduce the possibility of shorts.

Other than that + the MLCC capacitors, I'm not sure where else something could've gone wrong.

If you can, look at the output with a scope.