BGA vs. QFP pros & cons?

[Christe4nM]

Hi all,

In a project I’m working on I need to choose between using the BGA or QFP package of a certain component. There is a lot of research material on the web on either package, but those leave me with reliability questions. I would think this choice is being made by many engineers, and a comparison would be present somewhere.

What I found so far is that BGA is preferred for its smaller footprint, and better assembly yield (0.8 mm pitch BGA vs 0.5 mm pitch QFP). Also for signal integrity reasons BGA seems to be preferred since bypass caps can be placed ‘better’ and lead inductance is obviously less. QFP seem to be preferred for prototyping in some cases for easier soldering and pin access.

Yet what bothers me most is reliability. In our case the pcb will be part of an industrial measurement system. Depending on customer usage the system might be always on for >10 years. Among environmental issues are vibration and temperature changes. When long-term reliability is a concern; what package would you choose, and why?

Now I’d like to take this question beyond mentioned project, to get a more generic overview of BGA vs QFP (or TSSOP or similar packages if there are the case) for the same component. So if you have other reasons to prefer one over the other, please let me know.

Thanks,
Christean

[mikeselectricstuff]

In terms of reliability, an issue is with BGA it's much harder to know for sure that it has soldered properly, as inspection is a lot harder - x-rays are often used but you'd need to talk to someone familiar with their use about how effective they are.
It's possible you could get a joint that was only making contact by pressure, which could fail in the future.

On the other hand, QFPs are susceptable to bent pins, although at least any issues are easier to spot, and rework much easier.

[Rasz]

1 number of pcb layers required to route out signals = price

[SeanB]

BGA does have trouble of not being able to check if all joints are correct short of Xray inspection, and has been known to fail if it is not done exactly right, not right away but after a few weeks/months and in the field. QFP does at least allow inspection of all joins, and you can just do as the BGA does and put the decoupling under the board. In all cases a 4 layer or more board is recommended, with solid power and ground planes under the device.

[Psi]

The whole xray inspection issue removes BGA from consideration if the assembly or repairs are done by hand. (Unless you have a lab with BGA xray machine or you're happy to solder them blind)

If the boards get assembled professionally you're trusting that the assembler knows what they're doing and has soldered the BGAs correctly.
Bad BGA soldering can result in large failure rates months or years later so it's a good idea to send a few sample boards away to a 3rd party for dedicated BGA xray inspection every so often just to check.

I personally wouldn't use a BGA unless i had no reasonable alternative.
(So far i've never had to use one)

[cwz]

Given the choice between BGA and QFP, I'd always lean towards BGA.  First, I don't agree with comments and always needing X-ray for inspection.  Most assembly houses do not do X-ray inspection on every BGA on every PCB as it's not cost effective at all.  I can place and reflow >= 0.8mm pitch BGAs by hand without issue.  One of the nice things about BGAs is that so long as you do your pads correctly and get placement reasonably accurate, when they reflow, they will basically snap into place.  You can't of course be grossly out of alignment, but even using something like a silkscreen placement outline is usually good enough.

Second, in terms of reliability -- take a look what what's mostly on motherboards etc.  Even stuff rated for industrial use is routinely BGA.  For applications where the pin count is high, there really is not another economically viable option.  You'll occasionally see some applications that still use PGA and high temperature solder columns (rare), but they are far and few between these days. Just from a historical perspective, IBM actually created BGAs in the 60s as part of research into direct chip attach research.  See:

http://flipchips.com/tutorial/other/a-brief-history-of-flipped-chips

Ultimately of course, whatever makes the most sense for your design -- my advice is not be to be scared away from BGAs though.

[wraper]

Second, in terms of reliability -- take a look what what's mostly on motherboards etc.

And those BGA chips are one of the most often failure source of motherboards.

[NiHaoMike]

In my experience, the most common cause of BGA connection problems are due to mechanical stress, either due to board flexing or thermal cycling. Using a good heatsink mount and overdoing the cooling really helps reliability.

[ejeffrey]

Second, in terms of reliability -- take a look what what's mostly on motherboards etc.  Even stuff rated for industrial use is routinely BGA.  For applications where the pin count is high, there really is not another economically viable option.

You have hit upon the answer here.  For high pin count there is no alternative, so the downsides don't matter.  If you have the choice, QFP is going to be easier to work with and more reliable.

[marshallh]

Yes, BGA do require tight process control, but they are extremely consistent. Lead-free BGA in its infancy years ago wasn't too great but things have vastly improved since then.

If you need even more reliability you might look into underfills. It will help prevent the package from shearing off, but again in most applications where you'd have enough G forces to warrant this, there will be bigger components with even more susceptibility.

For vibration and shock reasons you may want to go to solder-mask defined BGA pads, and tighter specs. This is the alternative to the more common NSMD (non soldermask defined). The larger area of copper on the pad makes shearing less likely.

Also you probably know this, but RoHS has exemptions for certain categories of products, such as automotive, test and measurement, aerospace, military, etc. where lead is allowed.
For long term longevity lead provides a significant advantage as the presence of lead in solder largely prevents tin whiskering.

There is still research being done on the longevity of devices with leadfree joints in adverse conditions. Whiskering can happen at random, but is exacerbated by excess shock (both mechanical and temperature). Toyota and BMW have had problems with this (fly by wire pedal sensor pots having whiskers causing reliability failures, entire ECUs failing from whiskering)

Some vendors selling parts in BGA packages will provide both leaded and lead-free preballed packages. But as a whole, things are pretty much all leadfree now anymore.

And like a previous poster said - with regards to placement BGAs are far more forgiving than QFPs. I had to do some rework on a 208pinner and would've preferred reballing a BGA instead. I have placed and reflowed 0.8mm pitch bga myself at home and had no problems.

BGAs will require more pcb layers to fanout all their connections generally. It depends a lot on how many connections you need to break out, signal types (single ended, SSTL, voltage references, LVDS blah blah) but you can do up to 256balls on 4 layers, 484balls on 6 layers, 780 on 8 layers, etc. However, the overall pcb size can be greatly reduced with BGA vs the same part in QFP.

And since QFPs larger than 300 pins rarely exist (and you dont' see QFPs bigger than 200 pins really anymore) it is worth dealing with at some point.

[Psi]

I think we can all agree that a perfectly soldered BGA is pretty much as reliable as any other type of package.

The only issue is that you cant be 100% sure without xray inspection.

[NiHaoMike]

I just taken a closer look at two bad motherboards (one Athlon 64, one Pentium D) with what appear to be bad connection problems and they both were visibly warped due to the cheap (plastic) heatsink backing plate warping with age. I could actually get one of them (the Athlon 64) to boot if I flexed the board a certain way when powering it on, but they're both too old to be worth fixing so they're now in the parts bin with all the other old boards.

[Christe4nM]

Thank you all for tuning in. I sense that the consensus is to choose QFP over BGA, as long as both packages are available on a given component. While getting into a bit of further reading I basically found the same stuff marshallh mentions in his excellent post. I think Psi summarizes it quite nicely:

I think we can all agree that a perfectly soldered BGA is pretty much as reliable as any other type of package.

The only issue is that you cant be 100% sure without xray inspection.

So if you did the design properly it comes down to the EMS being able to provide high quality (control) as far as BGA reliability is concerned.