1st buck converter design help

[trilerian]

I created my first buck converter circuit and I keep frying the buck IC.  I created the design in TI webench, copied the schematic over to KiCad and then designed my pcb.  The goal behind this was to be able to have JLCPCB assemble the boards for me.  Normally I assemble my own boards, but due to amount these will be sold for, ie not much, I want as little as my own time in them as possible. 

So what is this?  This is a power distribution block for 12V DC and 3x USB A connector for 5V usb devices.  The idea is to plug the input into a 12V DC power supply, then you have 4 pair of 4mm connectors to power your chargers and other items for RC racing. A lot of racers complain that the cheap USB connectors they get don't supply enough current so I wanted to design my USB 5V with a buck converter that could supply high current. 

The webench design parameter was 8A and 5V, and it conveniently spits out a circuit for me.  I then went through BOMs of the designs and chose one that JLC had the components in stock for, with the exception of the inductors.  This was really the only change I made from the webench design.  I chose a 2.2µH inductor instead of the 3.3µH in the design, however I did look at the datasheet of the TPS51396ARJER and it suggests a 2.2µH inductor for use. 

Long story short, the completed boards that JLCPCB sent me short the VIN as soon as power is applied.  No load being applied through the USB connectors, it shorts on power up. The IC burns up with a little flame, and melts the solder mask above the traces (yes the IC is placed correctly).  When I remove the component the short is gone.  So there is not a short in the traces.  This happened with all 5 prototype boards I got from JLC.  Kind of expensive failure since this was done in 2oz copper and ENIG finish.

I am attaching the webench design, my schematic, my pcb layout and the BOM.

I hope this is some minor mistake I missed that can be easily fixed, thanks for any help.

-Andrew

[tooki]

Was the power supply already turned on when you connected it? If so, this might be classic “inductive kick” into ceramic input capacitors: https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

Do you have any new ICs you can install? If so, then try adding a much bigger electrolytic cap (like 1000uF) in parallel to the ceramic input caps, and turn off the power supply before connecting it. If this powers up successfully then you have a decent idea what’s wrong. Then you can consider whether to fix it by adding series resistance, or a big electrolytic, or a TVS diode.

[JustMeHere]

Those 2 pins that say NC but you have connect to ground could be the issue.  Never mind, I see the datasheet says it's okay to ground them.

Also make sure the inductor you choose has the right (something).  I'm not sure what (something) is but the first buck converter I made wasn't stable and Vout was way too high. I swapped it out everything started to work.  Maybe it's the resonant frequency. 

[trilerian]

Was the power supply already turned on when you connected it? If so, this might be classic “inductive kick” into ceramic input capacitors: https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

Do you have any new ICs you can install? If so, then try adding a much bigger electrolytic cap (like 1000uF) in parallel to the ceramic input caps, and turn off the power supply before connecting it. If this powers up successfully then you have a decent idea what’s wrong. Then you can consider whether to fix it by adding series resistance, or a big electrolytic, or a TVS diode.

Well that is interesting.  And yes, the power supply was already turned on.  I have connected those $1 Chinese buck converters up to the power supply while it is already turned on without issue.  And they use ceramic input caps as well, figured I should be safe. 

I can order some more ICs and test though. I would like to verify that I didn't make any other mistake in the pcb design first. 

[trilerian]

Those 2 pins that say NC but you have connect to ground could be the issue.  Never mind, I see the datasheet says it's okay to ground them.

Also make sure the inductor you choose has the right (something).  I'm not sure what (something) is but the first buck converter I made wasn't stable and Vout was way too high. I swapped it out everything started to work.  Maybe it's the resonant frequency.

Here is the datasheet for the inductor I used:  https://www.lcsc.com/datasheet/lcsc_datasheet_2108142230_KOHERelec-MDA1040-2R2M_C2847549.pdf.  The datasheet for the TPS51396A https://www.ti.com/lit/ds/symlink/tps51396a.pdf?ts=1725728959354&ref_url=https%253A%252F%252Fwww.mouser.com%252F says for 600kHz and Eco mode to use a 330kΩ and a 5.1kΩ voltage divider network for the mode pin.  Which is what I am using, and later in the datasheet it says for 5V and 600kHz to use a 2.2µH inductor.  Like I said, the webench design called for a 3.3µH inductor, but I made sure the resistance of the inductor I chose was lower than the webench design and the saturation current was at least twice what the buck IC could output.  I would have just chose the same inductor as the webench design, but JLC didn't carry the inductor component from any of the webench designs. 

[Electrodynamic]

So what is this?  This is a power distribution block for 12V DC and 3x USB A connector for 5V usb devices.  The idea is to plug the input into a 12V DC power supply, then you have 4 pair of 4mm connectors to power your chargers and other items for RC racing. A lot of racers complain that the cheap USB connectors they get don't supply enough current so I wanted to design my USB 5V with a buck converter that could supply high current. 

For buck, boost and cuk converters I like to use cheap micro controllers like the Arduino nano.

We can set the operating frequency with a timer on an output pin and monitor the output voltage and current using using input pins and interrupts.

Like this, https://maker.pro/arduino/tutorial/how-to-build-an-arduino-based-buckboost-converter

The problem is what I call "dumb electronics" which don't have the ability to self monitor themselves. For example, I like to throw in a couple super cheap thermistors to monitor the inductor and switching mosfet temperature. If it ever exceeds my overheat set point it turns off. I also like to stage high power applications so as one inductor and mosfet becomes overloaded I can switch in another identical inductor/mosfet section to take up the slack. I do this because small ferrite toroids are dirt cheap and larger ones much more expensive. We can also add a $5 OLED display to monitor stuff which also helps.

I would also overbuild the input and output buffer section which I found lacking on most electronics. They cheap out using undersized caps and any transients tend to wipe them out.

I'm also into electric RC cars, airplanes and helicopters for about 30 years.

[trilerian]

I ordered some more of the buck ICs.  I'll do more testing, I have some electrolytics I can put on the VIN as well to test.  Maybe the ICs were a bad batch, who knows. 

[trilerian]

I hooked my scope up to the VIN of the circuit and measure the voltage spike when flipping the switch on the circuit to turn it on while connected to a DC supply, probe tested at the positive side of C3.  The Max Voltage was 28V.  The IC I am using has a max VIN of 24V, so maybe after repeated use the 28V transient my destroy the chip, but I doubt that is the case for the pyrotechnics I got when power it up with the IC connected.  It really acts like a dead short on power up. 

Here is the scope output.

[tooki]

Was the power supply already turned on when you connected it? If so, this might be classic “inductive kick” into ceramic input capacitors: https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

Do you have any new ICs you can install? If so, then try adding a much bigger electrolytic cap (like 1000uF) in parallel to the ceramic input caps, and turn off the power supply before connecting it. If this powers up successfully then you have a decent idea what’s wrong. Then you can consider whether to fix it by adding series resistance, or a big electrolytic, or a TVS diode.

Well that is interesting.  And yes, the power supply was already turned on.  I have connected those $1 Chinese buck converters up to the power supply while it is already turned on without issue.  And they use ceramic input caps as well, figured I should be safe. 

I can order some more ICs and test though. I would like to verify that I didn't make any other mistake in the pcb design first.

Well, cable length is a factor, as is the total input capacitance, as is the supply voltage, and you don’t know if the Chinese modules use a chip that tolerates bigger spikes. (FYI, I have blown up Chinese DC-DC converter modules this way. I now add an electrolytic in parallel with their inputs if there’s any chance they’ll be hot-plugged.)

[tooki]

I hooked my scope up to the VIN of the circuit and measure the voltage spike when flipping the switch on the circuit to turn it on while connected to a DC supply, probe tested at the positive side of C3.  The Max Voltage was 28V.  The IC I am using has a max VIN of 24V, so maybe after repeated use the 28V transient my destroy the chip, but I doubt that is the case for the pyrotechnics I got when power it up with the IC connected.  It really acts like a dead short on power up. 

Well, once the chip is fried, you don’t know what all it shorts together internally. It very well could result in the failure mode described.

[trilerian]

I hooked my scope up to the VIN of the circuit and measure the voltage spike when flipping the switch on the circuit to turn it on while connected to a DC supply, probe tested at the positive side of C3.  The Max Voltage was 28V.  The IC I am using has a max VIN of 24V, so maybe after repeated use the 28V transient my destroy the chip, but I doubt that is the case for the pyrotechnics I got when power it up with the IC connected.  It really acts like a dead short on power up. 

Well, once the chip is fried, you don’t know what all it shorts together internally. It very well could result in the failure mode described.

I added a 220µF in parallel, this was the biggest electrolytic I had laying around with a 25V rating.  Attached is the scope shot of that hot powering up.  This I think should definitely be fine.  This is being made to be used with 12V DC supply, There is a little variance there like mine that is 14V, but even with mine this doesn't spike with that cap.  So I will assume this will be good enough for testing.  I can redesign the board with bigger caps for a safety margin. We'll see what happens when the parts come in I guess.