PCB gets warped when placed in enclosure

[cupoftea]

Hi,
A customer sent us a PCB which is 150mm by 90mm. It has some hot semi's on it.
Only 20W though.
The PCB sits in its heatsink, which is a metal extrusion.
The PCB sits on ledges which run along the long sides of the PCB.
On one side, the ledge is like a 1mm groove, into which the long edge of the PCB goes.

To get the PCB to sit on the ledges, you have to apply serious downward pressure on the PCB with your thumbs...you have to stand up
and use your body weight pressing down. When you have the PCB on the ledges...the PCB appears slightly bowed, due to the fact that its
squeezed between the two ledges.
The PCB has SMD components on it from 0402, 0603 and 0805, including 0805 ceramic capacitors.

There is one screw which holds the PCB still on the ledges, which is also the earth screw.

There is a piece of gap pad between the PCB and the heatsink, and this also is quite thick and appears to contribute to the
warping of the PCB.

Are there any circumstances under which you would just turn a blind eye to this?

 

[betwixt]

Sounds like a recipe for disaster. I assume when it is hard to push into place it is while cold so expansion probably makes matters worse. Are there also mountings, screws etc. on the PCB that further restrict expansion or anything else that would make it bow?

Brian.

 

[FvM]

You recently wrote about 3.2 mm clearance between SOT223 package and heatsink, and also about 5 (2 + 3) mm gap pad in a different post. This would well explain too much pressure on gap pad and PCB warping. You should probably try with 4 mm gap pad.

Most critical point with warped PCB is possible MLCC cracking.

 

[Mattylad]

As above, this is a recipe for disaster, joints will be under pressure, capacitors will likely already be cracked.
Either remove some of the laminate or make the enclosure bigger (the prior is easier).

I would not turn a blind eye to this, I have redesigned a complete PCB before because it was being stressed during manufacture, changing all the previous guys pad patterns, new layout rules for caps etc. It had a severely reduced return rate after this.

 

[D.A.(Tony)Stewart]

What/who goofed? The heatsink span or the PCB. There should be zero force to insert and allow fasteners to hold in place.

We once made a board for a famous jet engine controller company that looked like a drone shot of NY city for component density.
Depanelization forces caused many large MLCC's to crack because the designer put them in the wrong orientation near the edge.. We fixed it with a guillotine cutter with feedback for a new fab spin.

Put the card edge on the belt sander.. oh no