Scared of BGA's - justified?

[forrestc]

So, I'm hoping that some of you who have assembled BGA's on small assembly lines can either confirm or dispel my fears about BGA's.

We can assemble 0.5mm pitch TQFPs all day every day with excellent "no rework required" yields.  In addition, the very very few defects (bridges, insufficient solder, etc.) that do occur are easy to pick up with visual inspection.

We also have a few QFN's with similar pitch and have very few issues - probably better "non-rework" yields than the TQFP packages, but obviously, the defects we do have are much more difficult to visually find so they get picked up in the tester or on occasion in the field.

BGA's scare me.  First of all there are a LOT more pins on the parts I'd move to a BGA for, so more defect opportunities.  Plus, almost all of the balls are impossible to visually inspect in any meaningful way.  I've also heard horror stories about how narrow the process window is and other similar cautions.

On the other hand, the BGA's I'm looking at are 0.8mm pitch or larger, so our placement accuracy should be more than good enough.  In addition, I've heard that generally if you've got a BGA that comes out of the oven that is level and has 'settled' as expected and the balls around the edges look fine, the remaining balls should be fine.

My question then revolves around whether I'm just being unreasonably scared because I can't do what I feel is an adequate visual inspection - and that BGA's are no worse assembly-wise than other higher-thermal-mass parts  OR if they are the horrible things I envision.

Experiences?

[loki42]

We've just switched to more bga parts.  We've been improving the rest of our process but paste is the thing I'm worried about so I think SPI is useful. 

These are 0.4mm pitch bga. I've done smaller and they were annoying from a stencil cleaning point of view before I got an ultrasonic stencil cleaner.  BGAs are much easier to place than QFN for us so far,  the cameras on my machines do much better with BGAs.  programming the lighting for QFN is often fiddly. 

[forrestc]

These are 0.4mm pitch bga. I've done smaller and they were annoying from a stencil cleaning point of view before I got an ultrasonic stencil cleaner.  BGAs are much easier to place than QFN for us so far,  the cameras on my machines do much better with BGAs.  programming the lighting for QFN is often fiddly. 

I'm assuming that this will be the same for me as far as placement goes.  We find QFNs to be fairly easy and I expect to be able to use similar techniques.   

Assuming you get everything reasonably consistent in the printing and placement steps, do you see many reflow step failures?  I think that's sort of my biggest concern as things like head in pillow or other wetting failures seem to be something people gripe about,  but I have no idea if this isn't that big of issue anymore or not.

[loki42]

We haven't done enough yet to have a good idea but we don't usually have oven issues with anything else.  I'm running Indium paste and it seems to work very well.

[forrestc]

We haven't done enough yet to have a good idea but we don't usually have oven issues with anything else.  I'm running Indium paste and it seems to work very well.

That's good, as we're an indium shop.  Using 8.9HF paste here. 

I have a feeling I'm just going to have to buy some dummy/cheap BGA's and give it a shot.

[loki42]

I'm very happy with the Indium paste. I also used GC10 and GC18 which was good but had availability issues. I also tried some AIM paste that I didn't like.

These TI parts : TPS7A2033PYCJR are what we found a little challenging. 0.35 pitch, 0.18 pads. They are only a few cents each so handy for testing a process with, put down a few panels of a few 100 of them and see if they are all working okay.

[Siwastaja]

If you are fine with QFN, then BGA shouldn't be a problem. IMHO, QFN is actually a more difficult version of BGA, since the balls give a tiny bit of extra flexibility against the amount of solder. Like, if you push it down "too hard" during placement, the paste does not squeeze out and short pads like it does on a QFN package, so self-alignment during reflow has higher chances of working instead of the whole thing getting stuck at offset.

Having larger number of pins isn't the problem. If you are having any sort of yield problem, that will show even with just on a 16-pin QFN.

Yeah, you can't visually inspect. For production you want X-ray (for cheap product, random samples, for higher end stuff, every unit). In lab use / prototype, you just do functional test. Most difficult situation is if you have to manufacture small batches (tens of units) for customers quickly, and don't have an X-ray. Even if you do functional test, how do you know the units work after shipped, or in higher vibration environment? Some of the joints could be marginal and pass the initial test.

Of course if possible you want to avoid the smallest pitches. On the other hand, I hand-soldered dozens and dozens of tiny BGA 0.5mm pitch image sensors, and I'm not stellar at this kind of stuff. I didn't even have much visual clues on the correct placement, but over 95% of the boards ended up working just fine, which required basically all of the pins to function (parallel bus, many voltage supplies, etc.)