Low yield / large failure rate in prod.10 layer BGA boards. What can I do?

[userx2]

Hello

We make a lot of boards using 10 layer boards and ATMEL 9G45 processors DDR2 etc.
Heaps of BGA stuff.

Sometimes we get a great yield of say 4 failures in 500 but then at other times (now), we get 36 failures in 500. That is a lot of money.

The failures range from RAM failures to boards not booting etc.
They are mostly not repairable as we often can't find the cause.
We have to write these boards off.

I have to date not been able to find the actual cause of these problems but I suspect a problem with the actual pcb manufacture.

Now the question is, what do you other folks do to avoid these issues?

Is there some sort of bare board test I could specify to make sure the boards are at least 100%?

I do specify bare boards testing but it is unclear what the manufacturer actually does regarding that.

The board assemblies are done by subcontractors.

Any help or hints will be greatly appreciated.

Best regards
X

 

[Warpspeed]

Any decent bare board PCB supplier should be able to probe the board using the PCB net list to ensure no shorts, open circuits, and integrity of vias, etc...
Naturally this service adds to the cost, and you have to judge the economics.

It may not be the bare boards at fault, but the assembly process, or even the chips themselves may be failing from over temperature during soldering operations.

I suppose you could try to use different board suppliers, different chip suppliers, and a completely different assembly line to try an narrow down the problem.

 

[userx2]

Any decent bare board PCB supplier should be able to probe the board using the PCB net list to ensure no shorts, open circuits, and integrity of vias, etc...
Naturally this service adds to the cost, and you have to judge the economics.

It may not be the bare boards at fault, but the assembly process, or even the chips themselves may be failing from over temperature during soldering operations.

I suppose you could try to use different board suppliers, different chip suppliers, and a completely different assembly line to try an narrow down the problem.

We have used various board suppliers and it then seems like hooray, it is better but then, the next time, we may have issues again. the chips are mostly single source so there is not much to be done there.

I am not sure to what extend board manufacturers do bare board testing.
I am just wondering that this may require specific written instructions perhaps. I am not sure but I stumbled across IPC-9252.

I have also just found some articles on moisture potentially causing issues like we have.

Dunno.

Regards
X

 

[Warpspeed]

I am not sure to what extend board manufacturers do bare board testing.

You could always ask, but generally its a quick visual check at each stage of manufacture. The guys (and girls) on the production line are usually pretty damned good, but they are not infallible. With a ten layer board, the chances of a few duds sneaking through on bad days increases. Only way to be completely sure is to pay the extra for a fully automated (no bad days!) probed test of the finished boards and see if your yield increases.

Its a real pain, but its the only way to be absolutely certain.

Another thing to enquire into are the design rules for the PCB. Hair thin tracks and ultra fine clearances are going to be more problematic than bold tracks and larger clearances. Your CAD design people should be able to tell you. Drawing up a PCB that is very difficult to manufacture is not going to do you any favours.

As far as chips go, buy direct from the chip manufacturer, and make sure you have an honest purchasing officer. Counterfeit "Dragon/Panda brand" may be cheap, and come with a really nice Christmas present for your purchasing officer, but it spells doom to your business. He may feel he is doing his very best by saving your company money, but it may not be a bargain. Enquire into this...

As far as assembly and soldering goes, your methods could range from truly awful, to superb. No way of really knowing until you eliminate other possibilities.
I guess it all finally distilles down to professional know how, work ethic, and integrity throughout the entire chain.

One final thought.
ESD damage during manufacture due to poor methods of component handling is not unknown. Do staff have adequate training and the equipment to do the job ?

 

[mtwieg]

I've never done 10 layer designs but IMO if you're dealing with lots of BGA parts then my suspicion would fall on the assembly process, not the bare board fabrication. Bad solder joints, bowing, misaligned components, cracked capacitors, etc.

I would perhaps invest in getting an Xray imager (or borrowing someone else's) to carefully examine a collection of bad boards.

 

[marce]

May be bad solder joint or it could be via barrel cracking, the latter can be more prevalent in boards that have not been laminated at the correct temperatures and pressures, this may be why you are getting differences in the batches of boards.