Glad it's better than 105C/2000h

IMHO is says what I have been preaching for years: Do not use BGA packages unless you really, really, really have to.

Not 100% sure since I don't use any but on BGA's it's probably the balls themselves. Micro-fracture and so on. Having a large and totally rigid package soldered to a flexible material (circuit board) is something that I always thought was a bad idea. Except for all the reballing service folks that sprouted up, of course :-)

Our experience with BGAs has been fantastic. Production yields are 99+ per cent and field failures are nil. Much better than fine-pitch gullwing parts.

It's the only way to get 1000 pins out of a package, too.

On equipment that goes into 19" racks and remains largely stationary BGAs can serve well. But most of my designs have to fly on aircraft for umpteen years, ride in cars, or are banged around in clinics, on oil rigs and in heavy industry. Usually failure creates problems of varying magnitude. That's where BGAs can become a real nightmare. Underfill helps a bit but adds cost and possibly hampers repair efforts.

Le Thu, 20 Jun 2013 14:05:50 -0700, Joerg a écrit:

My package is an LQFP which gives 9 years at 110°C. And the graph has nothing to do with vibration but it seems with just temperature exposition.

Lets see what they have to say...

Yeah, that one is a bit surprising. They say that this is explicitly for their TMS320. Maybe some thermal cycling inside that is particularly extreme with this package?

Let us know what they say.

Have you seen a lot of field failures from bad BGA connections? Can you correlate that to vibration or temperature cycling or whatever?

ROHS an issue?

On Thursday, June 20, 2013 4:42:00 PM UTC-4, Fred Bartoli wrote:

The two figures are not related ( except by part numbers they describe of c ourse). FIT rate of the part does not include the package failure mode. For package failure mode TI does have this to say in a separate paper:

---------------------------------------------------------------------------

--- Reliability modeling is another important tool used to predict package performance in an application. Thermal, electrical, and thermo-mechanical modeling, verified by experimental results, provide insight into system behavior. This modeling process also shortens package development time, predicts system lifetimes, and provides an important analytical tool. In applications such as BGAs, where interconnections are made through solder balls, the useful life of the package usually depends on the useful life of the solder itself. Because this area has been studied extensively, accur ate models exist, both for predicting solder behavior and for interpreting accelerated life testing. TI methodology includes extensive model refinement and constant experimental verification. For a given package, a detailed 2D finite elemen t model (FEM) is constructed that performs 2D plain strain elastoplastic analysis to predict areas of high stress. These models also account for the thermal variation of material properties, such as modulus of elasticity, coefficient of thermal expansion (CTE), and Poisson?s ratio as a function of temperature. These allow the FEM to calc ulate the thermo-mechanical plastic strains in the solder joints for a given ther mal loading. Package and board-level stress analysis is performed using finite element modeling, which provides full 3D nonlinear capabilities for package stress, component warpage, and solder joint reliability studies.

---------------------------------------------------------------------------

---

But figure 2 is still worthless unless you have some idea of your operating environment to which you apply the failure rate "pi" multipliers of classi cal reliability analysis. NASA and some other organizations spell out most of these environmental factors but I don't find anything with actual number s. It looks like no one is using tables and charts anymore, and you now hav e to use much more thorough prediction software, such as this:

It's your headache, and you can have it.

Field failures, yes. Correlation is difficult because it isn't logged what they have been exposed to. Vibration has a very different effect on solder joints at higher temps versus lower. Unfortunately this topic has not been very thoroughly researched yet. But there are some studies such as these:

This was not really a surprise to me since BGA also fare much worse under temperature cycling, although not as bad as QFN:

Yes, but only from hear-say (I rarely design for RoHS), and in an interesting way. It can improve BGA reliability for low level vibration but at high shock load it makes it worse:

Of course, this does not include failures due to tin whiskers and other RoHS pathologies. Unfortunately much of this goes undetected with BGA because optical inspection is next to impossible.

Long story short, I've got nothing against BGA or QFN as long as they are small. Opamps, mixers, FETs, RF parts. But I will not use any large BGA or QFN on any hi-rel design. Or pretty much on any design if I can help it. Even with small packages, if they simultaneously offer an MSOP and a DFN like they do on the LT3757, I always go for the MSOP if I have available real estate.