Printed Circuit Board Assembly & PCB Design SMT Electronics Assembly Manufacturing Forum

Printed Circuit Board Assembly & PCB Design Forum

SMT electronics assembly manufacturing forum.


BGA Rework

We've encountered a little problem with BGA's. There was a p... - Apr 09, 2003 by

Chrissie

#24061

BGA Rework | 9 April, 2003

We've encountered a little problem with BGA's. There was a process problem in the build of the boards which has ment the BGA are failing open circuit due to the lack of solder paste. Any-one else had this problem? Any suggestions on solutions? We've tried replacing one BGA, but when the unit has been rebuilt and tested another BGA open circuit has been found on a different BGA. We also have the added problem of the boards being difficult in the sense we have to reflow the rework through the over rather than locally. This obviously reworks all other parts on the board. So we can only do it once. What does anyone think to my proposal to removal all the BGA's on this board (22) and replace them all?

Help!

reply »

#24063

BGA Rework | 9 April, 2003

LITTLE problem!!!

Consider removing and replacing the all BGA in one pass. Doing them one at a time [if that's the future you see for yourself] could get boring pretty quickly.

reply »


RDR

#24074

BGA Rework | 9 April, 2003

If you really think the vast majority of these are going to fail I think that doing them all in one shot is the best thing. That way, as you stated you only have to reflow the remaining parts one more time instead of 22!! i hope you only have one board like this. As dave stated I think it is going to get real boring just removing and cleaning the 22 all at once. As a side note, be careful when you are removing them you will probably want some cooling intermissions now and then.

Russ

Russ

reply »

#24075

BGA Rework | 9 April, 2003

This just makes my head hurt.

Why do these 22 BGA with hundreds of solder connections each have "open circuit due to the lack of solder paste"? Something just doesn't sound Kosher. Tell us about: * Stencil design * Solder paste * Board design in the vacinity of the BGA pads * Board fabrication * Other stuff we need to know

reply »

#24082

BGA Rework | 10 April, 2003

Hi, that post sounds very bad.But I thing you shoul make an order for that problem.First the problem as I can see is not single.If I am right you should do exams step by step: 1.Design of the board-examine very carefuly(ask the manifacturer for help) 2.Printing-most common problem(includes type of the paste,stencil design,pads reduction,paste heihgt,complanarity.....) 3.Try with accuracy and oven profile. 4.If it's not one of the top you are in a big trouble.Could try with the technology parameters(specials criterions on chip mounting(given by manifavturer),incompatibality chips and board or board with solder paste or solder paste with the balls of BGA's. Wish you good luck!

reply »

#24088

BGA Rework | 10 April, 2003

Thanks for your replies guys, there's some good info there.

Dave, answers to your questions:

The stencil was initially a big problem as with the Egineering Model's (EM's) and Engineering Qualifing model (EQM's) we kept getting short circuits due to too much paste. To fix this we reduced the pad size (these are all teardrop shape pads). We then found that the reliabily of the print went right down because the stencil was right on the limits of what could be done. We introducted a critical inspection of the solder paste print at at least times ten mag. This was unfortunatly ignored in build of approximately 30 boards. All these board have been brought into dispute and we've reworked several.

The boards can not be point to point tested on the outputs/inputs of the BGA's so we run a boundary scan on the boards which should pick up any faults. We have since found that this DOES NOT pick up ALL faults. Big problem.

These are all space boards and have the ESA spec on them, they can't be reworked too many times, and on most of the boards a bunch of FPGA's and line drivers have to be taken off then put back on after rework.

Another problem is that the boards are double sided and therefore each of the BGA's have already offically been though one rework. The boards (polyamide) are a bit notorious for the pads leaping off anyway without rework.

ANOTHER problem is that the board are super-duper space hardware and therefore dissapate heat so quick you can hardly measure it, this makes sure you cannot put the BGA's back on locally as the heat profile cannot be achieved and is not space qualified. We have just about managed to qualify taking the BGA's off locally and thankfully this seems to work, and we even have evidance in some cases which clearly shows there was no solder paste present.

The bottom line of the problem is the operator of the SMT machine who was trusted to follow the work innstructions didn't, and we have no recovery plan! Every board we have reworked has had more than one failure than went undetected from the boundary scan.

I even went as far as putting the board inthe X-Tec x-raying machine to see if you could see the change in shape of the solder fillet, which you can, but only if you know what you are looking for, on which ball, on which BGA, which when we have 240 connections on each BGA and 22 BGA's on a board - it doesn't work out well to spent all that time trying to get the right angle in the x-ray machine!

The only conclusion i could come to was the removal of all BGA's to ensure that i have enough confidence in the board to tell the customer it will survive launch and then work for 15 years!

I do believe there are machines that measure the solder paste once it is printed, and some even measure the hight of the solder paste, had a quick look but haven't checked with Siemens, who made our SMT machine.

Any thoughts on the above welcome!! Chrissie

reply »

#24092

BGA Rework | 10 April, 2003

Chrissie,

You have my deepest sympathies! But I'm very curious about your business takling a space contract. The quantity of boards that you mentioned hints at a GPS satellite systems, am I correct?

In any space quality manufacturing, every step of fabrication must be triple-checked. That printer operator you mentioned should be fired and his family exiled to Antartica (just kidding!)

The way I see it, these boards are so critical that reworking anything on it just degrades their specs, but they are so expensive that throwing them out is not an option.

So how about removing everything from the board and start again with the board clean? That is if all pads survives the cleaning and desoldering process.

One way to proceed building these kind of board is to process all those critical parts first (all FPGA)using printer and pick and place. Once all Fpga are inspected, checked and approved, then all other parts are "hand assembled" on the boards by "QUALIFIED" hand soldering experts. This is the way SPAR Aerospace do things.

I am sorry this doesn't help you much because the damage is done already.

reply »

#24093

BGA Rework | 10 April, 2003

Thanks for all the posts, i'm off on holiday for the next two weeks, but i know this problem will be around after i get back so hopefully i will be able to update you as to whether the board passes test or not!

reply »

#24105

BGA Rework | 10 April, 2003

Two weeks!!! Wow, what a place to work. Any openings???

reply »

jimcorr

#24137

BGA Rework | 14 April, 2003

They are letting you go on holiday with these problems outstanding??!! But seriously after reading about your problems I think that any further work is basically throwing good ��$$ after bad and reworking BGA almost always ensure that they will fail later certainly within your 15 years. If I've understood the situation you have 30 rouge PCB assemblies so get someone senior enough to bite the bullet and scrap the boards after stripping off the really expensive parts. Then do a major review on your process control. Jim

reply »

Jon Fox

#24138

BGA Rework | 14 April, 2003

This may be a little too late, but ever heard of a missing "silver". Had one one time (missing ground plane)and wow look at those TANT caps explode...might need some safety glasses! If they are all open look towards the design a little closer just to be sure that you can rule it out.

reply »

Jon Fox

#24139

BGA Rework | 14 April, 2003

That's nothing! A former aerospace customer of mine ordered their pick and place equipment and had it installed in August. Their first run of 5, yes 5, boards was not due to testing until the following May. They just wanted to make sure they were using up their government budget for each year. Must be nice to work in the aerospace industry. They ended up using it for research projects anyway in the mean time.

reply »

#24153

BGA Rework | 15 April, 2003

If I did the math correctly, we're talking about over 600 BGAs to rework. I have nothing of relevance to add other than Chrissie should be given a medal for not jumping off a bridge or tall building. Let's save that fate for the operators who didn't properly inspect the boards, eh?

J

reply »

mk

#24156

BGA Rework | 15 April, 2003

I cant help with the rework but I can help with putting them on right the next time.

Look into ssd. Contact me off line if it looks like it could help.

Matt Kehoe mkehoe@sipad.net

reply »

#24313

BGA Rework | 28 April, 2003

Hey, i'm back from the holiday.

Progress on this board is quite good actually, althought the guy who i got to take off and replace all the bga's won't even look me in the eye!! He said he had a really tough job re-doing all the paste. 22 BGA's in about 4 hours. He said he almost gave up after 4 and i think i was lucky not to be in the building!

Anyway, all the BGA's have been replaced, and then x-rayed (i haven't seen the x-rays yet, but i have the greatest confidance in my guy replacing them) and the board is being rebuilt. It should get to test today or tomorrow and it takes about 40 hours to test it. will it pass?!

Chrissie

reply »

#24314

BGA Rework | 28 April, 2003

The BGA rework process has been qualified for space flight. Which basically means someone sensible over in ESA has been replacing BGA's and life testing them. I agree the best way is to build clean boards without any mistakes, but i can honestly say that i have never seen one of these!! It's like any rework. It gets qualified by ESA, we use it, the customer buys it.

Chrissie

Oh, scrapping the 30 boards would probably cost approx �25 million.

reply »

#24315

BGA Rework | 28 April, 2003

I think i am just going to try and rework this one board. As you can well imagine we are going to have some interesting meetings trying to get the customer to accept it. The other boards actually have replacements (of the ones that have failed, and please don't even mention Thermal Vac to me as i'm sure some the results from there will be intersting!). This one didn't have a replacement and had a lead time of 7 weeks just for the PEC. Not to mention the 14 man days it then takes to build it up to test!

Chrissie

reply »

#24316

BGA Rework | 28 April, 2003

Chrissie,

I'm glad you (and your people) made it through the painful process of reworking all those BGA's. Hope all your boards sucessfully pass test. Now on to the larger issue.

In my experience, the solder paste screen printing process is (along with Reflow) the most important part of the whole surface mount process. After all, if the boards (prints) don't look good coming out of the screen printer, they certainly aren't going to get any better on down the line !!

I'm actually shocked to hear that you aren't doing anything but a visual inspection - post print, given the nature of the products you're building.

As to your comment "I do believe there are machines that measure the solder paste once it is printed, and some even measure the height of the solder paste, had a quick look but haven't checked with Siemens, who made our SMT machine." There are many such machines available.

With this market you could pick up used equipment very cheap. I have had significant experience with the Cyberoptics (www.cyberoptics.com) Laser Section Microscopes (LSM). These are stand alone units (LSM 3) that are very operator dependent.

What you might want to look into is the Cyberoptics SE200 (Cybersentry), which is an inline system which you could program to measure paste height and volume on your most critical areas around the board. Such as, oh I don't know, BGA locations maybe ? This could take the operators out of the equation by providing you with a "Go, No Go" capability. The Cybersentry's are plentiful on the used market now.

There are of course much more expensive options avialable to you (Inline AOI) depending on your needs and budget.

Hope this helped in some way. Good luck.

DaveM

reply »

#24378

BGA Rework | 1 May, 2003

IT PASSED! no problems, and the customer has accepted it. Never underestimate the stretching of the rules when the scheldule is super tight. Thanks for all the replies on here, much appreciated.

Right, onto the next problem.....

Chrissie

reply »

Kevin Simpson

#24446

BGA Rework | 8 May, 2003

Nice sugestions from all.. but one simple solution that you should practice from here on out.... CFM... If you would have built up your process, you would have found the flawes there early allowing you to correct the process flow. I would use CFM on a job worth a dollar much less 25 million. Good luck in the future! Whatever those boards goes in hopefully will stay clear of my house.

Kevin Simpson Technical Services Manager Qualcon www.qualcon1.com

reply »

#24448

BGA Rework | 8 May, 2003

Don't worry Kevin, the boards will be in space, so about as far from you house as possible! The process is mapped out and proved, and was recorded as a six sigmas process. The operator of the process, on the otherhand, wasn't. Chrissie

reply »

#24449

BGA Rework | 8 May, 2003

Hi Chrissie,

This might be old news to you but I just wanted to make sure that you are aware of all the implications of BGA rework relative to the control of moisture-sensitive components. For example, you need to keep track of the total exposure time once the components are removed from their sealed dry bag, all the way to final reflow, including the additional exposure time if the boards go through a double-side reflow process. And storing unused component in dry cabinets does not necessarily stop the clock of exposure time.

When you get to the rework itself, if you want to safely remove the BGAs for re-balling or failure analysis you need to bake the complete boards, for up to 10 days at 90C prior to removal. And of course you want to make sure that the replacement BGAs are still within their floor life. (Ref: J-STD-033A, section 6 Rework)

As you are probably aware, the issue of internal delaminations and cracks caused by moisture expansion during reflow is a problem that can cause a significant reliability exposure because these types of latent defects cannot be inspected or detected by any standard equipment, including X-ray.

Do not hesitate to call or e-mail if you have any questions on this subject.

Best Regards,

Francois Monette Cogiscan Inc. Tel : 450-534-2644

reply »

Phil J

#24457

BGA Rework | 9 May, 2003

Could you explain "dry cabinet storage may not stop the clock!!!"

reply »

#24458

BGA Rework | 9 May, 2003

Phil, here is why "dry cabinet storage may not stop the clock" :

The moisture diffusion inside a component is actually quite complex. When previously exposed components are returned to dry storage, the moisture gradient that is already absorbed in the plastic body continues to make its way inwards, towards the die interface. It may eventually exceed the critical limit after a certain time in dry storage. In order to account for this phenomenon, the IPC/JEDEC standard J-STD-033A provides the following guidelines to account for dry storage :

"4.1...Placing SMD packages, which have been exposed to factory ambient conditions for greater than one hour, in a dry cabinet or dry pack does NOT necessarily stop/pause the floor life clock. However if the conditions of 4.1.2 are met te floor life clock can be stopped or reset."

In summary, section 4.1.2 provides a series of rules to account for the time spent in dry storage. The rules are based on the specific MS level of the components, as well as the duration of prior exposure and dry storage : For MSL 2-4 : Rule 1 : For MSL 2, 2a, 3 : at less than 10% RH the floor life clock will stop/pause, but the cumulative floor life must not exceed the limit. This does not apply to Level 4 and higher. and Rule 2 : If exposure to ambient is less than 12 hours, followed by dry storage at less than 10% for 5X the prior exposure time, the floor life clock can be reset. (this is equivalent to a bake cycle at room temperature)

For MSL 5-5a : If exposure to ambient is less than 8 hours, followed by dry storage at less than 5% for 10X the prior exposure time, the floor life clock can be reset.

The above rules are applicable for previously exposed components. For previously dry components, once again it depends on the MS level and dry storage conditions. The floor life clock will remain at zero when the humidity is less than 5%RH. At less than 10%RH, some components with MSL 4 and higher will have a limited shelf life of 2 to 18 days in dry storage. This varies based on the body thickness. (Re : table 7-1)

I know that this sounds very complicated, but it is actually a simplification of a much more complicated phenomenon. There is a wide variation across a range of different components and the only other alternative is to take a conservative approach based on a worse case scenario. In this case one would assume that the floor life clock never stops in dry storage and this would result in a very large number of unnecessary bake cycles with the associated degradation in solderability.

Regards, Francois

reply »

PCB Soldering Tools