The curious case of the suicidal buck regulator

[4flyer]

So, I have been working on troubleshooting a buck regulator on a design of mine for weeks now. I am using a Texas Instruments LMQ61460 regulator on a custom board. I used the TI design tools to determine passive components and read the datasheet throughly. I designed my PCB and got to testing. The board worked!

The bad news is that the board did not work for long. The board would work no issues even up to 5A of load but at some point in testing, the board will fail. I even had two running for 150 hours straight (1 with constant 2.5A load, another with a toggling load). If the boards were in a constant state, they could run. If I put them on the power supply though, they would at some point fail when toggling power to the board via the power supply. When the baord fails, a short between the output and input of the regulator is formed.

Output of the SW node looks fine. I probed everywhere. I am worried that this is a transient issue but with only a 60MHz scope, there is only so much that I can see. I added some TVS diodes onto the board via a bodge and that did not work.
I contacted TI and asked them to look at the schematic. They said that they could not see any issues with it. Go figure.

I am increasingly thinking that this is a layout issue. I will attach screenshots of the layout. I am particularly worried about the ground paths around the regulator.

Any help or guidance will be appreciated!

[T3sl4co1l]

How many layers is this?

There are some pretty enormous slots/holes in your polygons, and what looks like entire arrays of vias to nowhere.  I don't know if that's fatal, but yes, it can be improved.

What are C15, 16?

Tim

[langwadt]

checked that the powersupply is well behaved when you turn it on and off?

[daqq]

Is the inductor 1.8uH or 18uH? The schematic resolution makes it unclear.

edit:

If I put them on the power supply though, they would at some point fail when toggling power to the board via the power supply.

Might help if you add a series diode between the power source and the converter. Some DC DC converter will die if you short their input (or discharge their input caps) fast enough.

See: https://www.analog.com/media/en/technical-documentation/data-sheets/8620fa.pdf , page 17

[4flyer]

All, thanks for the replies!

How many layers is this?

This is a 4 layer board.

checked that the powersupply is well behaved when you turn it on and off?

This is a great thought. I did in fact did an in-depth search into this. I was in fact using a PS that was putting out some transients. Lo and behold, changing power supplies will still result in the regulator failing. I even used a PS that had a pretty impressive soft start to it. Next step will be to test the board with a lipo to see if it is truly a PS issue.

Is the inductor 1.8uH or 18uH? The schematic resolution makes it unclear.

18uH. I'll check out the link

[daqq]

18uH. I'll check out the link

That's a pretty high inductance - looking over the LMQ61460 datasheet, the highest value I'm seeing even mentioned is 4.7uH. Also, you are supposed to connect the RT pin to a resistor, are you not?

[HwAoRrDk]

Also, you are supposed to connect the RT pin to a resistor, are you not?

I think OP made a typo in their initial post. Their schematic says they are using the LM62460, not the LM61460.

The datasheet for that IC says "connect to VCC for 400 kHz" so I think, OP is okay there.

[georges80]

OP, so what is your input voltage and what are the cable lengths from power supply to your buck?

You appear to have no transient input protection.

Your schematic shows 28V input. Reg is rated at 36V max?

Have you measured for transients right at the input/ground of your buck chip as input power supply is cycled?

cheers,
george.

[T3sl4co1l]

What are C15, C16?

Tim

[daqq]

Also, you are supposed to connect the RT pin to a resistor, are you not?

I think OP made a typo in their initial post. Their schematic says they are using the LM62460, not the LM61460.

The datasheet for that IC says "connect to VCC for 400 kHz" so I think, OP is okay there.

Could be, but still, the inductor is larger than mentioned anywhere. It will hold more energy at the same current. Not saying it won't work, but it could be an issue?

[m98]

I'd also be interested to know what's up with some TI buck converters. Recently used a LM60440 on a custom power supply module, they would just randomly either work flawlessly over hundreds of power cycles or blow up/short input to output after one or a few power cycles. Nobody found any issues with the schematics, PCB layout, components, or behavior during turn-on or turn-off. Failure rate of a run of 60 was like 10 % of the modules either blowing their fuse on first power-up, our just not regulating the output. The rest are still working totally fine after months of operation.
Never had anything like this, and I would've suspected we where just duped by the distributor if they weren't directly ordered from TI.

[KE5FX]

I'd also be interested to know what's up with some TI buck converters. Recently used a LM60440 on a custom power supply module, they would just randomly either work flawlessly over hundreds of power cycles or blow up/short input to output after one or a few power cycles. Nobody found any issues with the schematics, PCB layout, components, or behavior during turn-on or turn-off. Failure rate of a run of 60 was like 10 % of the modules either blowing their fuse on first power-up, our just not regulating the output. The rest are still working totally fine after months of operation.
Never had anything like this, and I would've suspected we where just duped by the distributor if they weren't directly ordered from TI.

Similar experience here with an Analog Devices part of known good provenance, an ADP222 LDO.  The 1.8V section came up OK at startup time, but over the next minute its output voltage would fall to about 900 mV.  Weird.  No likey.  Factory-new parts aren't supposed to act like that. 

In the OP's case, I'd hang a scope from pin 2 to ground and see if you're getting spikes due to lead inductance going back to the supply.  If so, an additional input filter capacitor with some series R can take care of that.

[thm_w]

What are C15, C16?

Tim

Electrolytics. But 15uF seems small considering how large the footprints are. Could try higher value as you are likely implying.

Also some weirdness with the layout, even if it might not be the cause. C16 is kind of off nowhere with a thin trace connected to the other cap. But the net is SYS_PWR_PROT, so why is there a plane of the same net right beside it that the trace does not connect to? Odd.