TPSM84209: Burned out Vin pin

Hi Chad,

Thanks for the details you've provided.

Is it possible the 24V supply is exceeding the abs max 30V rating during turn on? What is supplying the 24V input?

EN is floating, right?

Can you please share your schematics and layout? If you prefer not to post them on the forum, you can send to me through email:  kristoffer.flores at ti.com.

On the working boards you have, what current is being drawn from the 24V input supply? Is the current consistent between boards, or some boards are drawing notably more current?
As a side note, from your PCB image, I see that you have exposed copper, i.e. no solder mask, for ground down the middle under the device. The datasheet (p.30) shows the example layout with top metal underneath the middle of the device, but exposed copper just under the GND pad itself. In your layout, I am wondering if perhaps VIN is getting shorted to GND underneath the IC, though that would not explain how a board that is initially working would suddenly burn out.

Kris

Hi Kris, thanks for your response.

https://www.digikey.com/product-detail/en/puls-lp/CP10.241/1736-1054-ND/6591556

This is the 24v supply the engineer is using, and the customer will be using in their final setup with this board.  We haven't verified yet if that actual supply is going over 30V, but the datasheet states a maximum overshoot of 200mv.  I will ask the engineer to put a scope on it to verify when he gets a chance.

EN is floating.

I can't share the entire schematic or layout as it's confidential, but here is the 24v supply and the corner of the board from Orcad.  The 24v comes in through the connector on the right.  The 6v goes out through the copper pour to the other regulators off screen above this screenshot.  The only associated traces that are not on the top layer are coming from R88 and R89 to the feedback pin.  There are thermal vias in the Vout and Vin pins, as well as the ground strip in the center.  The Vin and Vout vias don't connect to any other layer as 24v and 6v only exist on the top layer.  Only the ground vias connect to ground planes inside the board.  There are clearances in the ground planes where the vias pass through.  

Our bench supply current display only goes to one decimal place, but the entire board draws 0.1A at 24v.  The current draw seems consistant between boards, at least to the resolution of the supply display.  As I stated before, at 6v it draws 0.5A.  0.1A at 24v is 2.4W, 0.5A at 6v is 3.0W.  Without more a more accurate current measurement (decimals on the display), that seems within reason of power in equals power out.

The footprint I used I downloaded from a site like SnapEDA.  I didn't catch the exposed ground pad you pointed out until after we sent the gerbers off to manufacture a rev B of this board, this is rev A.  There didn't seem to be solder bridges when we removed the burnt regulator, but there could have been.  But like you said, that wouldn't explain why the board worked, then suddenly burnt the supply IC.

I do have a question about the 2 exposed metal circles on the bottom of the chip.  The datasheet shows it as a keepout area around these.  I have the keepout in the footprint, but if you look there is a small finger that I missed in the area of one of those circles.  If that shorted, could it cause this kind of catastrophic failure?  I cleaned up those areas in rev B thinking it could be a problem.  Too bad I missed covering the ground pad.

Later today I plan to put a high watt/low value resistor in series with the 24v and look at the voltage drop with a scope to see if there is a short high current draw as the board powers up.  If we were drawing too much current during the soft-start period (5ms by the datasheet I believe), would the overcurrent protection still work in that time?

Thanks for any help you can provide.

Chad

Hi Kris,

We can't risk another board being damaged, so we're going to have the engineer just leave the one he has working powered up.  We ordered another PULS 24v supply for testing.  It won't tell us if the other PULS is defective, but it will tell us if the supply isn't suitable for this purpose.  I'll get back when I have more information.

Thanks,

Chad

Chad, any findings or updates? Please let me know if you're still digging into this. -Kris

Hi Kris,

We have done quite a bit of digging in the past few days.  We explored all of the bench power supplies and the PULS supply that we are using and haven't found anything out of the ordinary.  We have looked into possible grounding issues and haven't found anything.

Our current theory is this, I'd like your input on it.  The ground pin down the middle of the footprint was not covered by soldermask, my mistake.  We think it's possible that slight inaccuracies in placing the ICs on the board, and then the solder reflowed, that the exposed pins on the bottom of the IC settled very close to the ground pad.  You mentioned that those pins were connected internally to the switch node.  We think that if they were close enough to the ground pin, without touching, that when the IC powers up a small arc could occur, possibly from an inductive spike in the internal inductor or possiblly through some impurities trapped under the IC.  If one of those pins were shorted to ground, it could cause the destruction we have seen here.

We were able to catch the board fabricator in time to change the footprint so that the soldermask is correct with the datasheet on our rev B.  So we are hoping that will solve the problem.

Any thoughts?  I was going to reply with my findings later today, you were just a step ahead of me. :)

Thanks

Chad