Simple buck design - occasional failure

[phil from seattle]

I have a board that incorporates a TI buck regulator (TPS562202DRLR). The circuit is direct from their Webbench power designer with a slight tweak to the feedback resistors to avoid weird values.  It powers 5V circuitry (3 74AHCT541s and an AMS1117-3.3 which powers an RP2350). The 5V draw is generally less than 200 mA. Overall the design works fine. In my testing I used a programmable load to pull up to 2A and the buck regulator performed as expected (got hot) with no failures.  At 1A draw, the buck regulator reaches about 40C (ambient 18C) so it appears to be well behaved with intended use. With about 80 boards in user's hands, I have had reports of 3 failures where the TI chip has completely blown (see photo). After the chip blows, the boards still work fine with an alternate source of 5V (USB). I'm not a power expert so am hoping for some help.

I spec'd the input for a max of 12V and all 3 users claim they used a decent quality 12V PSU.  Also, they claim the failure happened when they powered the board up. It is possible they did something bad like reversed polarity but they all claim it worked previously and they just powered it up again. I spent way too much time trying to repro this. did 100s of power cycles to see if I could get it to fail. It kept working, no failure. Tried poor quality wall warts. Overvoltage (24V input) causes a failure but not a smoking, melted chip. I'm wondering if the failure mode is some sort of surge. The fact that the chip blew up makes me wonder. Should I be using a TVS? Would that even help? Is there something I should do to make this circuit more resilient?

I will admit, I suspect it is user error - stray piece of metal shorting the 5V rail to ground.  But I want to be sure I haven't missed something. FWIW, I know I should incorporate polarity reversal protection but the users all say they connected it correctly.

[tooki]

I'd be willing to bet they hot-plugged it into the power supply. With only ceramic input caps, the inrush current is huge, and when they finish charging, the inductive kick from the supply wires can cause the voltage to surge to damaging levels. (I killed a few ebay DC-DC converter boards this way.)

https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf describes this in detail.

If it is this, the fix is easy: add an electrolytic cap in parallel, at the board's input. (TVS diodes or series resistors are other solutions; the application note goes into various fixes.)

[phil from seattle]

Thanks.  That makes sense.  I wonder if a resistor in-line (say 5 to 10 Ohm) would be enough to reduce the inrush. Not sure what wattage but it probably doesn't have too big as the inrush would be relatively short lived.

[edit]  I'll add hot plugging to my test scenarios.

[mtwieg]

Hot-plugging the input is definitely one option. I would also consider assembly defects, a cold or cracked solder joint can work for months and spontaneously decide to fail.

[BadeBhaiya]

Can 10uF + 100nF MLCCs really cause such a massive inrush current to cause an inductive spike which would go damage the chip? That's not much capacitance, would the parasitic inductance even let the current ramp up that much before the caps are charged enough to make the current very low or zero?

[PGPG]

My assumption is that IC supply voltage specified in 'Absolute Maximum Ratings' should be minimum 2x the nominal supply voltage.
When I specify 12V supply I expect user can use 12V accu so 13.6V or even 14V. So I have to use transil for working voltage 15V and break voltage 18V. Such transil during surge pulse limits input voltage to about 24V so IC should be for 2x nominal voltage.

The other question is soft start.
What happens if someone connects supply by hand and he makes short brak after device has already soft-started. In my opinion it can happen that the second start is without soft. What happens to inductor current in such situation? Will it reach a saturation current. If so than current limiting can be not fast enough to protect IC from damage.

[Someone]

Can 10uF + 100nF MLCCs really cause such a massive inrush current to cause an inductive spike which would go damage the chip? That's not much capacitance, would the parasitic inductance even let the current ramp up that much before the caps are charged enough to make the current very low or zero?

In reality? Yes, a significant problem:
https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

[chasnc]

If they are powering the buck with a supply that has really bad startup overshoot, that could over-voltage things. many years ago I had a server PSU that was killing my own hard drives like popcorn. I checked the 12V rail during startup, and it was overshooting to almost 20V! Not sure how they missed that. It is a known characteristic to PI and PID control loops that during startup you tend to get overshoot. I have seen some benchtop PSUs do horrible things too.

I would ask them to use teh exact setup and equipment that they did before and get a scope capture of the input voltage with the feeding PSU powering up. Or even better - go there yourself.

[thm_w]

Can 10uF + 100nF MLCCs really cause such a massive inrush current to cause an inductive spike which would go damage the chip? That's not much capacitance, would the parasitic inductance even let the current ramp up that much before the caps are charged enough to make the current very low or zero?

What you see is not the damage from the overvoltage spike, that only initiates a conduction path, after that the main 12V supply dumps through the IC at probably 10 or 20W or more, which causes the burning.

https://www.ti.com/lit/wp/scaa124/scaa124.pdf?ts=1762287330550

[HwAoRrDk]

Also bear in mind old transformer-based unregulated wall-wart PSUs can put out significantly higher than their nominal output voltage when lightly loaded. For example, I've seen a nominally "12V" PSU put out 18V with a very light load.

[phil from seattle]

If they are powering the buck with a supply that has really bad startup overshoot, that could over-voltage things. many years ago I had a server PSU that was killing my own hard drives like popcorn. I checked the 12V rail during startup, and it was overshooting to almost 20V! Not sure how they missed that. It is a known characteristic to PI and PID control loops that during startup you tend to get overshoot. I have seen some benchtop PSUs do horrible things too.

I would ask them to use teh exact setup and equipment that they did before and get a scope capture of the input voltage with the feeding PSU powering up. Or even better - go there yourself.

Unfortunately, these are electronic tech types.  They barely have a DMM.  Wish I could go there - they are spread out. US, Germany, UK.

[phil from seattle]

Also bear in mind old transformer-based unregulated wall-wart PSUs can put out significantly higher than their nominal output voltage when lightly loaded. For example, I've seen a nominally "12V" PSU put out 18V with a very light load.

That was my first thought - crappy wall warts.  But when I probed for the PSU, none of them claimed to be using wall warts.  Given that the Max Vin of the chip is 17V, that is a very real issue.  I've documented to use quality PSUs but who pays attention?

I am considering moving to a buck regulator with a higher Vin max - 28-30V.  But that may be a ways off.

[Psi]

Anything with monkeys playing with it needs to be protected from overvoltage and reverse polarity, especially if it has a DC jack on it.

Most people will just plug random DC jacks into devices until they find one that works. So plan for your device to get 5VDC then 18VAC then  24VDC, then 9VAC, then 12VDC as people try random adaptors. 

Also should have something to protect the chip from spikes above its max voltage, like a TVS diode.
This can be tricky if you have very little voltage between normal input and max input.

[phil from seattle]

My assumption is that IC supply voltage specified in 'Absolute Maximum Ratings' should be minimum 2x the nominal supply voltage.
When I specify 12V supply I expect user can use 12V accu so 13.6V or even 14V. So I have to use transil for working voltage 15V and break voltage 18V. Such transil during surge pulse limits input voltage to about 24V so IC should be for 2x nominal voltage.

The other question is soft start.
What happens if someone connects supply by hand and he makes short brak after device has already soft-started. In my opinion it can happen that the second start is without soft. What happens to inductor current in such situation? Will it reach a saturation current. If so than current limiting can be not fast enough to protect IC from damage.

Yeah, that is the hot plug issue called out earlier.  A flakey connection to the supplying PSU could do exactly what you said.

[phil from seattle]

Anything with monkeys playing with it needs to be protected from overvoltage and reverse polarity, especially if it has a DC jack on it.

Most people will just plug random DC jacks into devices until they find one that works. So plan for your device to get 5VDC then 18VAC then  24VDC, then 9VAC, then 12VDC as people try random adaptors.

Also should have something to protect the chip from spikes above its max voltage, like a TVS diode.
This can be tricky if you have very little voltage between normal input and max input.

The board has screw terminals so it requires a bit more thought to plug it in. 

You point out the exact issue I have with a TVS - the gap between nominal and over voltage is 5V (12-17).  This is why I am seriously looking at something with 28-30V Max Vin.

[Psi]

For 12V automotive stuff guidelines say you need to handle 24V for 1 hour (from memory).
And TVS diodes with a working voltage of 24V usually clamp around 36V.
So you normally use a chip with absolute max around 35-40V.
If you're not doing automotive you can probably reduce that quite a bit.
But if you need to handle DC at that voltage, instead of spikes, then you need a polyfuse or something to protect the TVS diode.

[Bud]

The typical application circuit in the datasheet has a 10K resistor on the EN pin. Maybe you should add it to the next board revision. Who knows if it has to do with what may be going on in that path during spikes on the input.

[tooki]

Can 10uF + 100nF MLCCs really cause such a massive inrush current to cause an inductive spike which would go damage the chip?

Absolutely! The application note I (and later, Someone) posted explains it. And it's been experienced by many people on the forum, myself included.

[phil from seattle]

The typical application circuit in the datasheet has a 10K resistor on the EN pin. Maybe you should add it to the next board revision. Who knows if it has to do with what may be going on in that path during spikes on the input.

The webench design just has it tied to Vin. What would be the failure mode from that? The pin that shows the most scorching is the SW pin (connected to the inductor).   

[Psi]

When you get these high voltage transients caused by ceramic caps, where's the best location to check for them at power up?  Just scope across the first ceramic input cap?

[Someone]

When you get these high voltage transients caused by ceramic caps where's the best location to check for them at power up? 
Just scope across the first ceramic input cap?

Differential scope probe repeated across parts of interest (you shouldn't be worried about it blowing up? right? your design is good!).

[Psi]

why differential scope probe?

[Someone]

why differential scope probe?

Because the end user doesn't operate the device with that added grounding path.

[JohanH]

Most people will just plug random DC jacks into devices until they find one that works.   

Here's anecdotal evidence of that statement. One of my wifi access points power supplies failed once, so I grabbed a random 12V from my shelf full of adapters. Soon there was a smell of burned electronics... It was an AC adapter, not DC! Fortunately, the wifi AP had a protection diode on the input, and it was that one that was overheating and smelling. Once I changed it for a DC adapter, the wifi AP worked fine. Suffice to say, my AC adapters are labelled nowadays.

[Psi]

Yep, a friend had a 12V adaptor get plugged into a 5V USB hub once.
We're not sure how long it was plugged in like that but it was probably days. I suspect a TVS somewhere was getting very hot and the 12V psu was going into hickup mode.
Eventually the TVS died and let 12V into the PC. it took out the PC motherboard and also the replacement motherboard before the issue was discovered.