(All-in-one DC-DC Buck converter) MPM3630 burns at startup. Layout error?

[alirabarber]

Hi,

I am struggling with a working layout for the MPM3630 from MPS.
The input voltage is 16 V (4S LiPo Battery) and the desired output voltage is 6 V.
There is no load connected.

When I connect VCC it works fine at first.
However, when I try disconnecting and reconnecting the Battery multiple times, then eventually the MPM3630 burns at the VIN pin.
I am relatively sure that the internal MOSFET power stage burns.
It actually lights up and burns really beautifully.   

I did the same experiment with the corresponding Eval-Board: It works fine.
Also with additional input capacitance as I have it in my design.

So, I am pretty sure, that I have made a significant layout error.
But I cannot seem to find it.
And I am quite lost about what is actually happening here and causing this fail.

Does anyone have a hint as to what might be the error, or what is happening here?
I attached screenshots of my design and a picture of the burned circuit.

Thanks!
Alex

[Psi]

Putting 16.8V (4.2*4) into an 18V switchmode chip is playing it way to close.
Switchmode converters can generate a lot of ringing/noise.  And you only have 1.2V before it's at 18V.

See if you get the issue with a 3S lipo

[alirabarber]

That's a good hint, thanks!

The actual ringing should not be influenced by the board layout, correct?
I thought, that the entire commutation loop of the power stage was inside the package.

I can imagine, that the actual U_DS_max voltages of the MOSFETs vary from part to part.
In that case, the eval board would have a "lucky" part.

What could have an influence, is the different cable length to the battery between the two setups.
The eval board setup has a significantly longer cable length.
It is a much bigger board and I did not solder the cables to it.


I will repeat the experiment with a 3S battery.

[Phoenix]

However, when I try disconnecting and reconnecting the Battery multiple times...

This might be causing lots of arcing and voltage spiking.

[alirabarber]

I just tried it again, first with 2S and then with 3S.
It works fine.

The ringing as explanation makes sense for me.
It seems, that the eval-board layout somehow works better to suppress ringing than my layout.
I suppose the big VCC and GND areas can have something to do with it.

When I connect and disconnect VCC multiple times, there are for sure a lot of arcs and current spikes.
I am not completely sure about voltage spikes - because of the input capacitors.

To solve the issue I will limit the max voltage of VCC to a 3S LiPo.

If anyone has ideas on how to reduce the ringing, I am open for suggestions.
Other than that I consider the problem as solved and am really relieved. 

Thanks for your time and effort!

[RiZsho]

Hi!

Voltage ringing on input destroying DC DC converters is nothing new and limited to this chip only. I know LT had an app not, but can't remember the title. The jist of it is the following: Long leads on input can have some inductance to them, so when you connect them to the input of the DC DC they tend to form a LC tank with the input caps (low esr helps ringing a lot in this case) and can cause massive spikes upon connecting. The app note tested three configurations: only ceramic, ceramic+electrolytic and finally ceramic+RC snubber in paralell (1uF||4.7Ohm). Basically keeping the ringing damped by increasing series resistance seen on empty caps. Almost all solutions helped a lot in some form or other ie. lowering the spike significantly.

Hope this helps

[mikerj]

This is a classic failure mode caused by the ceramic decoupling capacitors on the input to your circuit.  Have a look at this application note:

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

[RiZsho]

That's the one!

[alirabarber]

Oh, very cool!
That definitely helps!

Now that I think about it, it absolutely makes sense.
I will take a minute and read it in detail.

Thanks alot!