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

Printed Circuit Board Assembly & PCB Design Forum

SMT electronics assembly manufacturing forum.


BGA attach eval.

rdr

#23984

BGA attach eval. | 1 April, 2003

Hi all, I've got a question that hopefully someone can provide insight. We have a gold immersion assembly that we believe is experiencing fractures or non wetting on a certain BGA. To be sure that it is a fracture and not a wetting issue I would like to perform the following analysis. I would like to "grind down" the package where the opens are so I can remove the part without heat and check to see if the open is at the BGA or PCB interface and if we have solder covering the pads or if it is still gold or black or whatever. My question is, when I get to the ball and pad is it possible that the gold could have dissolved into the joint without wetting to the pad?

Thanks for your input

reply »

#23987

BGA attach eval. | 1 April, 2003

You're entirely correct. The gold is most likely gone away. [Don't forget you can check if the surface is still solderable with a hand soldering iron.]

We're not sure it it will give you the results you seek, but consider using a dye penetrant failure analysis. Here's what we do [It's based on a paper by P-E Tegehall & B Dunn, ivf.se/elektronik. You can probably find it on the web]: 1 Application of Dye 1.1 In order to facilitate the application of the dye: * Create a dam around the component using a modeling wax or clay. * Apply tape over via holes beneath the component on the opposite side of the board to prevent the dye from escaping through them. * If that is not possible, place the sample in a glass beaker. 1.2 Fill the dam with dye [We just use ITW Dymon, DYKEM Steel Red, 800.443.9536. There are 'real' dyes out there that are very good.] or alternatively pour the dye into the glass beaker to completely submerge the sample in the dye. 1.3 Place the board or the glass beaker in a vacuum chamber. 1.4 Make two evacuations down to 100 mbar to remove any air entrapped in the cracks. Since solvents in the dye evaporate during this process, evacuate as fast as possible to prevent the viscosity of the dye getting too high. 1.5 Pour the surplus dye from the sample. 1.6 Dry the sample at 100�C for 15 minutes. 2 Remove the part 2.1 Close and secure vice grips on the part. 2.2 Gently twist the part on a plane parallel to the board. 2.3 Move the vice grips to the other two sides and twist as above. 2.4 Repeat steps 2.1 through 2.3 as required. 3 Assess the metal surfaces

reply »

rdr

#23989

BGA attach eval. | 1 April, 2003

Dave, thanks for the info. This sounds like it would verify the presence of an open. We have verified that these balls are open and I am trying to figure out if we are fracturing or not wetting. causing these opens. I would use the die method on new builds however I cannot afford to scrap to many of these parts since they are relatively expensive. this problem appears to be at 1-2 percent of total production runs. So I am still trying to figure out a economical and fast method to differentiate between fractured joints and non-wetting. So if I understand you right, It could be possible that gold dissolved but the ball still didn't wet to the underlying nickel. which would foil my attempt at validating a fracture by the non-presence of gold on pad. I wa shoping that if the pad had solder present on it it would mean fracture, If it was gold we obviously didn't wet. Before any of these questions arise I should note that the reflow profile has been verified at the bare board level to 225 deg. These opens are also ocuuring at the corner locations of the part. Review of gerberber data and such reveals no unusual copper layers or anything that would cause a massive difference in relow temps at this location or any surrounding area.

reply »

#23995

BGA attach eval. | 1 April, 2003

R1: So if I understand you right, it could be possible that gold dissolved, but the ball still didn't wet to the underlying nickel. D1: Correct. The gold dissolves into the solder first. The solder wets to the nickel second. If you nickel is corroded, the gold is gone and you�ll have a pretty flat, not irregular fracture surface on the pad.

R2: It would foil my attempt at validating a fracture by the non-presence of gold on pad. D2: Correct. But if you think about it a little differently, you might still pull this off. Say, for discussion purposes, you torque a ball from a BGA, generally the pad surface is very rough, compared to a pad with just solder. It the pad doesn�t take solder or dewetts, it�s still going to be pretty flat and smooth.

R3: I was hoping that if the pad had solder present on it would mean fracture, if it was gold we obviously didn't wet. D3: See the gold just prevents the nickel from oxidizing and that makes it pretty easy to solder to the nickel. If you saw gold, which is not impossible, it would mean something is wrong with the gold.

R4: These opens are also occurring at the corner locations of the part. D4: Now you�re making sense. This must be a BIG BGA. One of my favorite topics. Look here http://www.smtnet.com//forums/index.cfm?fuseaction=view_thread&CFApp=1&Thread_ID=5713&#Message22497

reply »

rdr

#23997

BGA attach eval. | 1 April, 2003

Well, here is what I have found so far. I have removed the top of the BGA and found that the ball had detached from the package. Three balls were removed with the two adjacent balls pulling the pads off the board and stayed attached to the BGA while the open ball remained attached to the PCB. inspection of the three locations showed that the pad with the open was pretty flat with 3 small solder bumps remaining. I am now curious about possible dewetting of this ball from the package. I suppose you could have a fracture that would look like this but I am not sure.

What exactly can cause dewetting?

By the way this is not a large BGA it is a 27mm 1mm pitch full array Alter APEX with a solid metal top attached to the FR4 substrate.

Dave, I read the link (smart media connector?) I should point out that it is not the outermost corner ball but very close like pins 20-23 out of 25. We have not been able to verify if 24 and 25 are open since they tie to ground with a lot of others within this package.

Reflowing these components fixes the problem. I am looking at our process to find stressing operations at this time and I hope that it is a simple matter of fracturing joints.

reply »

iman

#23999

BGA attach eval. | 1 April, 2003

from my humble experience :

1) check if there is nickle contaminate in the gold bath? shd have report data from the PCB fab house.

2) check if your production did 48hr bake at 125degC, IF the BGA did not come in its "original" MSD cozy packaging.

reply »

#24001

BGA attach eval. | 2 April, 2003

RDR

The most common cause of dewetting is from a fabricator using any form of abrasive (e.g. pumice, abrasive brushes etc.) on copper. With this process, abrasive particles get implanted into the copper (often actually swaging copper over some of the particles). Extending this, it would be strange to see dewetting on a single pad.

That you have solder on the pad indicates little as far as understanding you dewetting issue. [Is the pad gold? ;-)] That you are reflowing these connections indicates that it is not dewetting.

Yes, the smart media connector thread.

It's interesting that your failure occured at the package. You say there was three small solder bumps remaining on the pad. Where is the solder that you'd expect to find on a solder fillet like this located?

reply »

scottf

#24008

BGA attach eval. | 2 April, 2003

I agree with Iman.

Verify the integrity of the component for moisture.

Most definately place an amount of ownership on your PCB supplier. Request tank analysis results and copies of the lot travelers. Look specifically for any in process rework procedures such as an aluminum scrub.

Do you have under BGA scope capabilities?? If so you should be able to determine wetting. An experienced Xray technician can identify an open at the interconnect as opposed to a floating void.

If your profile is of a longer cooler nature then it is possible that the gold hasnt time to completely intermix during your reflow, hence leaving a higher concentration of gold at the interconnect area (Gold Embrittlement).

Fractures at the package/die interconnect area are most likely caused by undo stress within your post SMT operations. Are these assemblies processed through ICT? Post assembly hardware installed in or around the noted areas?

Component location i.e. center of the board can cause your outer ball phenom, as well as moisture. Look to a Moyet analysis to verify.

Hope this helps.

reply »

rdr

#24010

BGA attach eval. | 3 April, 2003

Thanks Scott,Iman, The problem has been isolated to detachment from package so I believe that there are no fabrication issues related to the Gold on the PCB. What is a Moyet analisys? where can I find more info?

Dave, I am still waiting on Altera to find out if the BGA pads are gold, I would doubt it however. I am also trying to find out if they use abrasives as you mentioned. To answer the question "where was the solder" I beleive that it was in/on the ball. with a forceable removal, could one expect to see a flat pad? I think we have seen this during rework process when we remove the part "a little early" We see the ball left on the PCB with a flat top and the BGA pad appears to be virgin. On these we do know that these balls were not defective or open so I guess I might be seeing the same thing but it leads me to maybe think about profile again. We do not have great cooling in our oven (2-3 deg. C /sec.) and I wonder if that could do something wierd.

reply »

#24012

BGA attach eval. | 3 April, 2003

R1: Contacting Altera to find out if the BGA pads are gold. D1: Bet money the pads were ENIG.

R2: Contacting Altera to find out if they use abrasives. D2: If abrasives were the problem, you�d see it in many pads, especially those close to the problem pad.

R3: Could one expect to see a flat pad with a forcible removal? D3: We�d be surprised not see lifted pads. Lifted pads are flat, right? Lifted pads indicate a good solder connections. Mmmmm, flat non-lifted pad doesn't sound good.

R4: We see the ball left on the PCB with a flat top and the BGA pad appears to be virgin. D4: Not good. The flat top indicates that the ball melted at one time or another. This pad and ball have seen at least two solder cycles with solder and flux. Something is wrong. Will this pad take solder with an iron? [If only Mikie was still here for a riff on virgins.]

R5: It leads me to maybe think about profile again. D5: Dunno. Why wouldn�t that affect beau coupe balls?

reply »

rdr

#24015

BGA attach eval. | 3 April, 2003

Well, first off now I'm just starting to believe that we are fracturing again. But here is some more input

D1 ENIG? I would not have thought that they would be but as usual you are probably right.

D2 Again I would have to agree about it should affect more pads if it was an abrasive problem.

D3 this makes me feel better about the process since the bad BGA during removal had all pads lifted at either the board or The BGA except the bad ball which stayed with board.

D4 We don't use flux when removing BGAs (could that be what I'm seeing?) I am going to start removing in the future with flux and see if the same phenom occurs. The "virgin" pads readily take solder however in this condition

D5 DUNNO either, thinking maybe CTE with the thick metal top maybe fracturing joints during cooling? again why only 1 ball (not the same) on the entire package and only on the outer row A20-25? still DUNNO.

Overall I really appreciate your insight and help into this matter. None of it is making sense to me at this time.

reply »

#24017

BGA attach eval. | 3 April, 2003

R4: We see the ball left on the PCB with a flat top and the BGA pad appears to be virgin. D4: Not good. The flat top indicates that the ball melted at one time or another. This pad and ball have seen at least two solder cycles with solder and flux. Something is wrong. Will this pad take solder with an iron? [If only Mikie was still here for a riff on virgins.] R4: We don't use flux when removing BGAs (could that be what I'm seeing?) I am going to start removing in the future with flux and see if the same phenom occurs. The "virgin" pads readily take solder however in this condition D4: The "virgin" pads take solder? That�s interesting, assuming you're using the same flux when reflowing and this touch with the iron. Well then, that gets me to profile also. Consider thermalcoupling these locations during reflow and comparing that recipe with well flowed locations. There�s something screwy about your board design that�s sucking the heat from the solder connection.

R5: It leads me to maybe think about profile again. D5: Dunno. Why wouldn�t that affect beau coupe balls? R5: DUNNO either, thinking maybe CTE with the thick metal top maybe fracturing joints during cooling? again why only 1 ball (not the same) on the entire package and only on the outer row A20-25? still DUNNO. D5: Correct.

R6: Well, first off now I'm just starting to believe that we are fracturing again. But here is some more input D6: This pad is not fractured. If it�s virgin flat, it�s not fractured. Typically, a fracture drives through the center [half of the solder remains on the pad, half is on the ball] of the solder connection to the bottom center of solder [underneath the center bottom of the ball]. It may breakout parallel to the pad or up through the center of the solder on the pad.

R7: None of it is making sense to me at this time. D7: Ask Altera if they will do a failure analysis of your BGA with the flat pad.

reply »

PCB Soldering Tools

Fluid Dispense Pump Integration