BGA question

Unless these things are being shot out of guns, I'd suspect the reflow quality.

Orient the things to the direction of shock? Other than that, I think you're going to have to find a lighter package.

Interesting. We have exactly the same problem with big QFN (64 pads). Some of the pads are loose and some of the chips came off. We are going to die-bond them for productions. Are your chips available in wafer/die?

Potting, or a small heatsink with mechanical screw mount.

But if your BGA's are seriously falling off the board, then something is probably amiss with your soldering and/or resin quality. Or you've simply chosen the wrong package for such an environment.

Dave.

Forget about the BGA, I've heard of one application that BGA's will not survive. TQFP or Die mount is the way to go.

Cheers

I was going to say exactly that but didn't because whenever I do they throw tomatoes at me in a German NG, calling me a luddite and that sort of thing ;-)

TQFP would not be any better. Actually, QFN (including center pad) and BGA have more soldering area. If they can't hold, TQFP would not either.

Yes.

They should be held as tightly as they need to be to survive any shock test just on the solder balls alone. That should get you well into 40G range.

Check the freshness and brand choice (quality) of your solder, and examine your solder process as well. You may be using an old lead solder process profile with new ROHS lead free solders, which require a completely different profile (well, not completely). The other way is to switch your cooling mode to conduction methods, where you compress a profiled plated and a specific thickness of thermal compound together and against the parts. The whole PCB gets encased in a thermal plate clamp. NOTHING shakes off. Cool the plate externally. You also, of course, first conformally coat everything in a high quality coating.

Put it in a can that clamps all the goods together.

Are you using glue under them? I recall reading epoxy can be injected underneath.

Must be a nightmare for rework. ;-)

Tim

-- Deep Friar: a very philosophical monk. Website:

Making my suggestion even more viable.

Referred to as underfill, mandatory for medium shock survival, along with good PCB support to minimize flexing. For really high shock you simply do not support components by their leads, balls or otherwise, and proper potting would be required for BGA survival.

The other usual method is to isolate the sensitive PCB from the shock and vibration with compliant supports, sometimes referred to as shock mounts or vibration isolators.

Generally there is no rework possible for potted or underfilled, make a new one. By the time a PCB fails from shock or vibration other joints are fatigued and closer to failure, so rework is not even desirable and the item is sold on eBay "as is, untested" :-).

Bare die bonding is a lot more cost effective. I was told that they can even rework and move the die to new board. The problem is getting prototype quantities of them (less than 5k). They also need to be handled with care.

If you potted the module with a potting compound of the same density as the parts, they'd 'float' with full support on all surfaces, regardless of the acceleration of the assembly.

That would depend on the 'firmness' of the potting media.

I think it is called 'modulus of elasticity'.

Anyway, it essentially means that rubbery potting compound will not 'assist' much under high G stresses. It needs to be firm to provide structural support assistance.

OR a profile following conductive cooling 'can' or cap and thermal compound sheet compressed against it to dampen or stop vibrational shock issues, and allow for thermal conduction of the heat it may generate.

In many cases, this can be cheaper than potting (especially from a serviceability POV) and also allows for precise thermal management.

Almost a prerequisite for high altitude high power devices since there is not enough air 'up there' to carry the heat off of a finned heat sink. Conductive cooling methodologies are just about the only way to go in such cases.

It is a lot quieter too, generally. :-)

Until the board warps.

If the shock waves have a wavelength comparable to the components, the elastic properties matter; I was speaking of the effect of lower frequency (millisecond, not microsecond) disturbances. Presumably, elastic mounts can take out vibrations in the ultrasound region.

Unless you treat the board as a 'part', and instead of connecting it with fixed bolts, float it, too. A stiff (torsion box) frame with compliant board supports won't deflect much, nor put bending forces on the board. Even if it were a Kapton flexible circuit board: floated in a soft near-liquid, it'd be as safe from bending as a yolk inside an unbroken egg.

The usual reason for a board to bow under acceleration is that the mass of the center is unsupported except by the strain of the board; the addition of floatation medium changes that.

I deal with high shock and high vibration environments for PCBs, also. We could not use BGAs, at all. There is a new(?) package (at least the connections) that apparently works for high vibration/shock

-- or so says our ME staff. That package is a Column Grid Array -- not to be confused with a Pin Grid Array.

Look into replacing the balls with columns. Also, look into using metal bars to stiffen up the PCB and reduce the shock loading to the place on the board where the CGA is placed.

Tom P.