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

Printed Circuit Board Assembly & PCB Design Forum

SMT electronics assembly manufacturing forum.


BGA problem: open after reflow

Thomas Ballhausen

#2575

BGA problem: open after reflow | 1 November, 2000

Using convection type reflow we observed following problem: One ball of a 313-PBGA appears to be dewetting. This happens randomly, i.e. no specific ball location, failure rate is rather low (3-4%), but since there are 12 PBGAs on one board the failure rate at board level is as high as 40%. So it's a pain.

While no evidence can be found by x-ray, the system fails at electrical test. Investigations revealed an open; detailed investigation exposed the ball has lost contact to its pad (substrate), while the joint to PCB-pad is OK. The substrate pad apears dry and reballing failed.

According to substrate vendor pad is constructed by copeer with Ni/Co layer then protected by thin Au layer (imerson).

The pad does not appear to be dark; is it still possible to be related to "dark pad" phenomenon?

Has anybody comments or ideas of what may cause a solder ball to dewet from its pad?

reply »

Dason C

#2576

Re: BGA problem: open after reflow | 2 November, 2000

Please check with Fab house, what is the thickness of the immersion gold if the gold thicknesss is less than say 2.5 micron, you will found dewetting or open. It is recommend the gold thickness is around 5 micron.

Good Luck

Dason

reply »

Thomas Ballhausen

#2577

Re: BGA problem: open after reflow | 3 November, 2000

Thank you for your hint. According to the substrate supplier the thickness is less than 1 micron. After Ni/Co layer there is water rinse, then "gold strike", another rinse then "gold plating". But still the total Au layer shall be less than 1 micron. Do you have any background, why the gold thickness should be around 5 micron? DO you have any further experience on this? Is there a chance for oxidation of Ni/Co layer, if the protective gold layer is not thick enough?

best regards, Thomas.

reply »

Dean Stadem

#2578

Re: BGA problem: open after reflow | 6 November, 2000

We have seen this problem several times. It always goes back to pad solderability, very seldom does it have anything to do with ball size. If there is no evidence of wetting on that pad, but all of the other pads did solder, you need to ask yourself why. Typically, every time you remove the BGA, do the site prep on the bga pads by removing the solder and leaving a flat coat of solder plating, and then replace the BGA the rework is successful. This is because the oxidation/contamination of the pads was removed during the rework. If these pads were in a solderable condition in the first place, you would not see these individual opens coming out of the reflow oven. Check the plating process, no matter what type it is. That is where the problem usually is.

reply »

#2579

Re: BGA problem: open after reflow | 6 November, 2000

Thomas: Wheeee, a 40% failure rate. Now, that make your boss chase you around the room won�t it?

As you indicated, x-ray won�t do dip about helping to identifying problems [or ball cracks either]. Consider using an ERSA scope [er whatever they call it] to inspect for your problem.

Dason is giving it to you straight. If your total gold thickness is 1 thou, you are getting too little gold. And this shortage will almost certainly give you a solderability problem, sooner or later.

A typical ENIG thickness specification spec is: 3-8 uinches of gold over 150-250 uinches of nickel.

The two counter-balancing issues with gold thickness are: 1 Gold too thin allows the nickel to corrode, making it very difficult to solder. 2 At below 0.5% of gold in solder no difficulties in soldering are encountered, but at higher concentrations, solder becomes sluggish when cooling. Concentrations of greater than 4% gold in alloyed solder, under plastic deformation, exhibit brittle fracture.

Finally [while not strictly on topic, but minutia close to being on topic], immersion gold self limits around 12 uinch.

reply »

Thomas Ballhausen

#2580

Re: BGA problem: open after reflow | 7 November, 2000

Thank you for your kind contribution and efforts in helping me. I just found out we have been talking different units; my microns were micro meter ...? So, let me get things right now: the substrate supplier confirmed the following data: Cu foil + Cu plating is 22um, equals 88ui Ni-Co plating is 5um, equals 20ui (equals 20 thou, isn't it?) Au plating is performed in two steps - first "strike" (?what's that?) then "immersion" with a total thickness of 0.3 ~ 0.7 um, equals 12 ~ 28 ui. Based on your experience - 'gold is critical if thicker than 20ui' - then 12 ~ 28 ui appears to be a quite wide spec; can we request the supplier to optimize his process and to limit it to 20ui max? Would this be reasonable? Comments or further hints welcome. cheers, tb.

reply »

#2581

Re: BGA problem: open after reflow | 7 November, 2000

Cheers to you

Your gold plating is specified for the gold wire bonding between the die of your BGA and the pads on the BGA substrate. For solderable surfaces this is heavy on gold for use on pads. Gold SB thinner on pads to limit the amount of gold available to form tin gold intermetallics.

Consider requesting a gold pad spec described in an earlier posting on this thread for the solder side of the substrate.

Types of gold coatings available are: * Electrolytic is usually thicker (50+ uinches). It is often used for mechanical contacts (edge connectors), although I have seen electrolytic gold used for solderable surfaces. * Immersion (or electroless) gold is the more common finish for solderable surface mount boards. The copper pads are first plated with about 50+ uinches of Ni. The resulting gold coating is about 3 to 12 uinches thick.

I used to think I knew what "flash gold" meant, but now I think answer is either "yes" or "no" depending on whom you are asking. Specific terminology for electroless plating are sometimes not the same from person to person or region to region. Clarify the term with the person before starting a discussion to prevent any misunderstanding. For example immersion gold is also called: * Self Limiting Gold * Metal Replacement Gold * Electroless Gold * Flash Gold

So, when your supplier says "first �strike� then �immersion�", I don�t have a clue, because I think of "strike" as flash, which I think of as immersion. So, ...

Returning to your original point, it�s possible that the gold in your soldered connections is high enough to embrittle the solder connection, causing it to fracture when you smack the board with your test probes. Using metallographic analysis to determine the alloy composition of your solder connection might be something to consider in assessing this theory. So if you know a good failure analysis lab, it's time!!!

reply »

Arul

#4993

Re: BGA problem: open after reflow | 17 January, 2001

We have few failures similar to the discussion in this thread, on a 50 Mil pitch, 30 Mil mill Ball PBGA. Would like to understand the mechanism of the BGA reflow process bit more. Appreciate your response to the following.

1.We sheared few solder balls from a raw BGA. We sheared a BGA laminate from the ball attached to the PCBA leaving the ball on the PCBA. Now we compare the surface of the material left on the BGA laminate pad.

We find that the raw BGA's pad sheared surface has the SnPb solder at various Z heights whereas the assembled BGAs pad leave a smooth layer of SnPb along the pad.

Like to understand the cause for the differnece. 1.During reflow, when the ball melts, does the ball separate from the BGA? If yes Where does intermetallic layer stay? In the molten ball or the pad? 2. Can the flux from the solder paste flow to the BGA's Pad during reflow? 3. Does this flux play a role in "remaking" solder joint made betweent the BGA's substrate and solder ball? 4. Does this SMT reflow degrade the shear strength of the joint between the BGA ball & BGA laminate after SMT reflow? 5. What is the metallugical characteristics of the NiSn Intermetallics? What is the melting point/ eutectic point of a NiSn alloy?

During the SAM analysis we don't find the Gold in both cases. We Assume that the gold had migrated to the solder.

Thanking in advance for your valuable response.

Regards, Arul

reply »

#5002

Re: BGA problem: open after reflow | 17 January, 2001

>We find that the raw BGA's pad sheared surface has the SnPb solder at various Z heights whereas the assembled BGAs pad leave a smooth layer of SnPb along the pad. Like to understand the cause for the difference.

Who could guess? Solder joint strength in pull/shear varies with lead geometry, solder volume, lead metal/metallization, and the way the test is done. It is very difficult to shear solder constantly. Shear tests are [in my opinion] senseless. The shear stress you measure depends more on the shear rate and on the point where the force is applied than on the grain structure. When shearing a component, you not only apply shear stress, but also roll the component over. This means that both shear and normal stress are present, again making the results very dependent on where you apply the load.

What�s the composition of the materials on the two solderable surfaces? Are they different? You mentioned gold. Are these both ENIG coated pads? Is your solder near-eutectic?

Without having a clue of what you�re doing [or how you�re doing it], I�d guess the rough surface was soldered well and the smooth surface was not. Do you get the same results popping the ball off with a screw driver [or torquing it off with a pair of pliers]? For me, this is the true value of shear testing � "bump pull" tests and the like only confirm quality, but do not assure reliability.

Can you post real nice pictures on-line on driveway, idrive, or someplace like that?

>1. During reflow, when the ball melts, does the ball separate from the BGA?

Which reflow? What kind of balls? What pads? What pad solderability preservative? What solder? Hel-low-oh!!!

When soldering the balls of a BGA to the assembly, I expect the ball to remain wetted to the [nickel plate over???] the copper pad on the BGA interposer. With a PBGA, like yours, I expect the ball will melt, collapse, and wet to the solderable surface on the assembly PCB. If you find the balls separating from the BGA, something is wrong, either with the BGA fabrication process or your material selection / process control.

> If yes, where does intermetallic layer stay? In the molten ball or the pad?

Intermetallic, which intermetallic? The gold should be dissolved in the ball and will form gold / tin intermetallic[s] where ever it chooses. Nickel / tin intermetallics should form along the boundary between the pad and the solder. Similarly if copper pads are involved, copper / tin intermetallics should form along the boundary between the pad and the solder.

>2. Can the flux from the solder paste flow to the BGA pads during reflow?

You bet, I�d be surprised if the solder paste flux did not flow to the BGA pads during reflow.

>3. Does this flux play a role in "remaking" solder joint made between the BGA's substrate and solder ball?

Yes, the solder connection at the BGA interposer and the ball will reflow, assuming the paste and solder ball alloys, and the solder on the BGA have similar liquidous temperatures, reach that temperature, and are solderable. While not quite the same, we have talked previously on this Forum about vias next to the BGA pads getting hot enough during wave soldering to reflow balls of a BGA located on the primary side. Cool [hot], eh?

>4. Does this SMT reflow degrade the shear strength of the joint between the BGA ball & BGA laminate after SMT reflow?

Listen, the pad will give it up before a properly designed and formed solder connection.

>5. What is the metallurgical characteristics of the NiSn Intermetallics?

The appearance and composition if the intermetallic layer depends on the temperature of formation. The brittle Ni3Sn4 forms on the nickel side. It�s tough to say what forms on the tin side. Some people say NiSn3, others say Ni3Sn2. I�m not taking any bets.

>What is the melting point / eutectic point of a NiSn alloy?

http://cyberbuzz.gatech.edu/asm_tms/phase_diagrams/pd/ni_sn.jpg

>During the SAM analysis we don't find the gold in both cases. We assume that the gold had migrated to the solder.

SAM??? You use SAM for metallurgical analysis? We use it to find package leaks and cracked balls. Please tell us more about this analysis technique. Who makes the machine are you using?

Anywho. Yes, gold dissolves into solder very quickly. You should be able to find the more common Au3Sn intermetallic with routine metallographic analysis.

reply »

Arulvanan P

#5013

Re: BGA problem: open after reflow | 18 January, 2001

We find that the raw BGA's pad sheared surface has the SnPb solder at various Z heights whereas the assembled BGAs pad leave a smooth layer of SnPb along the pad. Like to understand the cause for the difference.

Who could guess? Solder joint strength in pull/shear varies with lead geometry, solder volume, lead metal/metallization, and the way the test is done. It is very difficult to shear solder constantly. Shear tests are [in my opinion] senseless. The shear stress you measure depends more on the shear rate and on the point where the force is applied than on the grain structure. When shearing a component, you not only apply shear stress, but also roll the component over. This means that both shear and normal stress are present, again making the results very dependent on where you apply the load.

What�s the composition of the materials on the two solderable surfaces? Are they different? You mentioned gold. Are these both ENIG coated pads? Is your solder near-eutectic? >>>>>Dave, The BGA substrate surface finish is NiAu, while the PCBA pad surface is HASL. The solder ball composition is SnPb 63/37 and the solder paste used on the PCBA is also the same but with 2% Silver.>>> Dave, sorry I don't have the picture with me now to post. After sodering the BGA on to the PCBA, we peeled off the BGA's subsrate which left the ball well in tact on the PCBA, and we analysed the surface left on the peeled BGA's subsrate pad. That was smoother than a raw BGA's removed ball by pulling it away using a plyer.1. During reflow, when the ball melts, does the ball separate from the BGA?

Which reflow? What kind of balls? What pads? What pad solderability preservative? What solder? Hel-low-oh!!!

>>>Dave,Forced air Convection reflow after the BGA's Placement. Ball 63/37, No preservative HASL 63/37 pad, Solder is same with 2% Silver.>>>>Dave,I have kept the picture at driveway of the analysed surface left on the BGA's pad. The crystals that you see in the picture are analysed to be of copper rich . From this can you tell whether it is due to the BGA fab process or SMTA process ?? If yes, where does intermetallic layer stay? In the molten ball or the pad?

Intermetallic, which intermetallic? The gold should be dissolved in the ball and will form gold / tin intermetallic[s] where ever it chooses. Nickel / tin intermetallics should form along the boundary between the pad and the solder. Similarly if copper pads are involved, copper / tin intermetallics should form along the boundary between the pad and the solder.

>>> >2. Can the flux from the solder paste flow to the BGA pads during reflow?

You bet, I�d be surprised if the solder paste flux did not flow to the BGA pads during reflow.

>3. Does this flux play a role in "remaking" solder joint made between the BGA's substrate and solder ball?

Yes, the solder connection at the BGA interposer and the ball will reflow, assuming the paste and solder ball alloys, and the solder on the BGA have similar liquidous temperatures, reach that temperature, and are solderable. While not quite the same, we have talked previously on this Forum about vias next to the BGA pads getting hot enough during wave soldering to reflow balls of a BGA located on the primary side. Cool [hot], eh? >>>>

>4. Does this SMT reflow degrade the shear strength of the joint between the BGA ball & BGA laminate after SMT reflow?

Listen, the pad will give it up before a properly designed and formed solder connection.

>>>>

>5. What is the metallurgical characteristics of the NiSn Intermetallics?

The appearance and composition if the intermetallic layer depends on the temperature of formation. The brittle Ni3Sn4 forms on the nickel side. It�s tough to say what forms on the tin side. Some people say NiSn3, others say Ni3Sn2. I�m not taking any bets.

>What is the melting point / eutectic point of a NiSn alloy?

http://cyberbuzz.gatech.edu/asm_tms/phase_diagrams/pd/ni_sn.jpg

>>>

>During the SAM analysis we don't find the gold in both cases. We assume that the gold had migrated to the solder.

SAM??? You use SAM for metallurgical analysis? We use it to find package leaks. Please tell us more about this analysis technique. Who makes the machine are you using?

>>>

Anywho. Yes, gold dissolves into solder very quickly. You should be able to find the more common Au3Sn intermetallic with routine metallographic analysis.

reply »

#5021

Re: BGA problem: open after reflow | 19 January, 2001

Intermetalic Layer (IL). A compound formed at the interface of two different metals, whose atoms have an extremely high natural attraction for each other, so high, that they do not bond to other elements by any other means. Also, intermetalic compounds have fixed stoichiometric ratios. Whereas, alloys can have varying ratios of metals.

Bingo!!! You shorely do have some great looking intermetallics. [http://www.driveway.com/share?sid=9e6a021.123344&name=BGAFA&view=0 ] I did not expect see crystals like that!!!! Something is very, very wrong. Bothersome things are: 1 There is no connection between the ball and the pad 2 I see no Ni/Au over coat on the copper pad 3 Wetting angle of the solder connection is poor 4 All those needle shaped crystals. An analysis of the needles will probably show Au3Cu.

Here�s what I�d guess happened � Someone at your BGA fabricator was in a rush or dearly wanted to save money in a sneaky fashion. They skipped the nickel barrier that SB between the copper pad and the gold overcoat. The copper and gold did their thing by forming an intermetallic. They soldered balls on the BGA, which helped the intermetallic grow. And then you ran it through another heat cycle to help them grow further.

To validate this hypothesis, do a similar sectioning and SEM analysis, both on * Ball of a raw BGA as received. * Ball of a raw BGA that you ran through the reflow cycle that the assembled board goes through, but not attached to the board.

reply »

#5025

Re: BGA problem: open after reflow | 21 January, 2001

I just can not get over the size of these crystals. What is the chronology of and the span of the thermal cycles this BGA has seen.[Just guess at the fabrication cycles, if it's difficult to determine]

reply »

Arulvanan P

#5084

Re: BGA problem: open after reflow | 30 January, 2001

Dave, To do further analysis we don't have the BGAs of the batch from which we had this issue. We don't face this issue now in the current production lots. The problem got solved while trying to find the root cause.

Will keep updated if it occurs again or if there is any other findings.

Thank you very much for your efforts. Regards, Arul

reply »

Arulvanan P

#5085

Re: BGA problem: open after reflow | 30 January, 2001

This BGA could have gone through the reflow cycle two times as we understand from the BGA manufacturer, during the BGA ball attachment process. (If any ball is found missing, they replace the missed ball manually after applying flux manually and reflow the BGA again).

The SMT PCB assembly is single sided, hence it should have through the reflow only once in the SMT process.

The posted pictures are are from a PCBA which was found to fail during functional test from the normal production, which had not gone through any other reliability tests.

reply »

reflow oven profiler

Reflow Oven