TECHNET Archives

February 2008

TechNet@IPC.ORG

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
"Igoshev, Vladimir" <[log in to unmask]>
Reply To:
TechNet E-Mail Forum <[log in to unmask]>, Igoshev, Vladimir
Date:
Mon, 25 Feb 2008 10:20:55 -0500
Content-Type:
text/plain
Parts/Attachments:
text/plain (442 lines)
Nice cross-sections. It does look like a typical case of Brittle Fracture. I'm wondering, though, what a "yellowish" layer on top of the E-Ni layer is (image 1 failed 1000x)? Another thing is that it seems that intermetallics is much thicker in the failed ball. If that is the same package and the BGAs are close by, then I'd expect both of them to have a similar (thickness-wise) layer.

Vladimir

-----Original Message-----
From: TechNet [mailto:[log in to unmask]] On Behalf Of Steve Gregory
Sent: Monday, February 25, 2008 9:37 AM
To: [log in to unmask]
Subject: Re: [TN] Metallurgists, need assistance, closing this one

Good Morning Everyone!

I've got Phil's photos up, pretty nice pictures. They're at:

http://stevezeva.homestead.com/files/_1_failed_200x.jpg

http://stevezeva.homestead.com/files/_1_failed_1000x.jpg

http://stevezeva.homestead.com/files/_1_not_failed_200x.jpg

http://stevezeva.homestead.com/files/_1_not_failed_1000x.jpg

http://stevezeva.homestead.com/files/_2_not_failed_200x.jpg

http://stevezeva.homestead.com/files/_2_not_failed_1000x.jpg

Steve



-----Original Message-----
From: TechNet [mailto:[log in to unmask]] On Behalf Of Phillip Bavaro
Sent: Sunday, February 24, 2008 7:17 PM
To: [log in to unmask]
Subject: Re: [TN] Metallurgists, need assistance, closing this one

Hi Inge,

I think I know what the problem was with these BGAs, although we are still running down all the loose ends before finalizing the report.

This seems to have been the result of a soldermask defined stress line combined possibly with a marginal adhesion possibly due to not enough heat being applied during reflow.

The solder mask aperture for this mask defined BGA had been erroneously reduced down to .3 mm and for a .45mm ball, I think that was just too much for it to survive.

The piece of data that helps me support this theory is that once we fixed the aperture of the mask on the next spin of the board, there are no failures at all, even after ESS testing.  Used the exact same reflow settings and tried not to change any other parameters.  Shear test values indicate good adhesion.

I appreciate everything that all of you contributed, as it certainly opens ones eye's to all the possibilities.

And Inge, I certainly can feel what must have been a lot of frustration for you to delve so far into something that doesn't always have a clear cut root cause.

I will keep you all posted and look forward to hearing what you may see in the six photos that I have sent to Steve.  

1) Entire ball crack, 200x, mask defined stress line crack?
2) Ball crack, 1000x, mask defined stress?
3) Good Ball, 200x, reduced mask defined, same as 1,2 but from different CCA
4) Good joint, 1000x, reduced mask defined, same as 1,2 , but different CCA
5) Neighboring BGA good ball, 200x, on same CCA as 1,2
6) Neighboring good joint, 1000x, on same CCA as 1,2

Sorry it took so long to get the pictures up!

Thanks in advance,

Phil



Hernefjord Ingemar <[log in to unmask]> wrote: Phil, If you want to make a real deep dive, I recommend that your start with Dave's brilliant lesson in the art of understanding the soldering mechanisms:
 
The Impact of Reflowing A Pbfree Solder Alloy Using A Tin-Lead Solder Alloy Reflow Profile On Solder Joint Integrity
 
I recognize your cross section and your dilemma. I also agree with some guys that your cross section is 'obscure' (strange use of the word here). However, whatever you do, you'll get headache.  George Wenger got a disc from me, a 69 page report, in which you can follow my battle to understand why some BGAs just jumped off the boards, due to poor or no true solder joints at all. Maybe George can share the content with you. But you will get headache!
 
As for your question about IMC, I think that the creation of that layer or those layers 'depends' on a lot of factors. The IMC can occur :
 
as homogenous and even 0.5 um to 1.0 um thick lines, clearly and non disputably seen in a mildly etched cross section.
 
as inhomogenous and irregular 0um to 5 um thick lines, appearing as rounded structures,  reminding of cauliflower
 
as burtst or needles stretching several micrometers from the solder interface, or fingers, as you say.
 
I think that the appearance of the IMCs are much dependent on the microstructure and chemistry of the nickel, and also of the solder constituents. Is the nickel plating lamellar grown or columnar grown? Are brightners involved? What size of nickel grains?  What size of the nickel conglomerate? What contaminations in  the nickel plating? What organic adds? What oxygen level during soldering? What activation grade of the flux? What solder particle size and distribution? Then you have the soldering process parameters.  And a lot more..
 
WHAT YOU SEE IS  DEPENDENT ON THE DEGREE OF PREPARATION AND SCIENTIFIC INSTRUMENTS YOU USE.  That's why the discussion becomes endless sometimes, we see the subject with different eyes. Don't give up, Phil, one day you'll see your  problem clearly, but doing it "my way" as sang Sinatra.
 
Back to practice: start with the article I mentioned in the ingress.
 
Inge
 
PS. A LOM and a SEM photo on the balls and the pads AFTER prying, would be interesting too. Use both SE and BSE mode.
 
 

________________________________

From: Phillip Bavaro [mailto:[log in to unmask]]
Sent: den 15 februari 2008 19:22
To: TechNet E-Mail Forum; Hernefjord Ingemar
Subject: Re: [TN] Metallurgists, need assistance, closing this one


Ok, I know this one is old, but I just happened to notice something I did not understand.  

In Inge's cross section, VD08073_improved_cross_section, my assumption is that there is copper and nickel plating on the board, and then there is an intermetallic layer sitting on top of the nickel, followed by the solder ball.

What is that intermetallic if the PWB was ENIG and the ball/paste is SAC305?

And is it supposed to be a complete layer covering the nickel or merely to  be thin "fingers" which grab the solder.

My gut feeling is that the longer the connection is held at soldering temperature, the more complete that IMC will become until it forms a uniform bond/thickness between the nickel and solder.  

If the IMC grain structure gets too long, then it becomes brittle so there is a definite window of opportunity with respect to the reflow times.

Am I close?

Thanks in advance,

Phil



Hernefjord Ingemar  wrote: 

 Thanks George,
 
 Such words from you are technical vitamins. Indeed, I had to finish before the winning post. I'm nicknamed the "terrier" here, usually don't let go anything once I clench the teeth, but my client thinks he can take over here. I have already asked for MTBF or similar that could give a better background, but he does not like that. Your proposal to swap is brilliant, and is already in the report, posted to another country. 
 "Wie get's mit ENIG?" 
 "Wenig." ha-ha
 Inge 
 
 -----Original Message-----
 From: TechNet [mailto:[log in to unmask]] On Behalf Of Wenger, George M.
 Sent: den 11 maj 2007 00:04
 To: [log in to unmask]
 Subject: Re: [TN] Metallurgists, need assistance, closing this one
 
 You've done an extremely thorough job of analyzing the Ingemar,
 
 You've done an extremely thorough job of analyzing the issue associated with randomly lifted FPGA pads and you've listed many logical conclusions. However, your email seems to indicate that you are getting tired of looking through the kaleidoscope and going around in circles. That doesn't sound like the "old Ingemar" I interacted with several years back. However, I do understand and I think Werner would agree that all customers are not the same and some have a tendency to make their technical talents feel like they are going around in circles. The one thing I did not get out of your first email posting was a sense of how random the randomly lifted FPGA pad is. In order to do an informed risk assessment one needs the data to input into the risk assessment. You've certainly provided the technical knowledge of possibilities but what is missing is the frequency or percentage of lifted pads. I'd like to make one other suggestion that you might want to consider in the report to
 your customer. Since your customer is using ENIG PCBs the only real method I know of to determine if the lifted pads are related to ENIG is to run a comparison to another surface finish. I would recommend that your customer have a percentage of his boards fabricated with immersion silver and track the occurrences of lifted pads. If the frequency of occurrences stays the say with ENIG boards and you don't see lifted pads with immersion silver boards your customer with then have "real" data to allow them to make a decision of how they want to proceed; work with their ENIG board vendor to improve the PCB surface finish quality or switch to an alternate surface finish. 
 
 Regards,
 George
 George M. Wenger
 Andrew Corporation Wireless Network Solutions Senior Principal FMA / Reliability Engineer 40 Technology Drive, Warren, NJ 07059
 (908) 546-4531 [Office] (732) 309-8964 [Cell]
 
 -----Original Message-----
 From: TechNet [mailto:[log in to unmask]] On Behalf Of Steve Gregory
 Sent: Thursday, May 10, 2007 4:50 PM
 To: [log in to unmask]
 Subject: Re: [TN] Metallurgists, need assistance, closing this one
 
 Hi Inge!
 
 Well folks, another Swede adventure comes to an end, but he has left in his wake yet one more image in a kaleidoscope of SEM's and cross-sections that has all of our eyeballs spinning: 
 
 http://stevezeva.homestead.com/files/VDO8073_IMPROVED_CROSSECTION_2.jpg
 
 I think if you look through the many images that Inge has sent for me to post on my web page over the years, there a few gems that could be blown-up and framed as art...
 
 Steve Gregory
 
 
 -----Original Message-----
 From: TechNet [mailto:[log in to unmask]] On Behalf Of Hfjord
 Sent: Thursday, May 10, 2007 2:42 PM
 To: [log in to unmask]
 Subject: [TN] Metallurgists, need assistance, closing this one
 
 Well, well, anything funny comes to an end. I have to write a report and I will do my best to convince my client. If anything goes wrong, I will say : this is what the TechNet gurus decided....he-he
 
 Seriously, I think I have focused on wrong objects. Gone in circles. Not unusual in our profession. I have considered and weighted what you gnus had to say, and made this logical conclusion:
 
 General joint strength: Good
 Leads plating: OK
 Solder pad plating: OK
 Solder paste: OK
 Solder phases near lead: OK
 Solder phases in middle of joint: OK
 Solder phases near solder pad: OK
 IMC at lead interface: OK
 Solder voids: caused by insufficient peak dwell Solder voids vs. failure: not primary cause Solder void content: debris caused by micro shockwaves in water, polishing grains acts as abrasive, soft Lead partly removed mechanically.
 Solder voiding countermeasure: diamond milk, low rotation speed.
 Fracture at lead interface: NO
 Fracture in the middle of joint: NO
 Fracture at solder pad surface: NO
 Fracture at lead/IMC interface: likely
 Fracture frequency: random
 Fracture pattern from field : random
 Black pad: not likely
 Skip plating: not likely
 Dewetting: not likely
 IMC at solder pad interface: much varying in thickness, partly missing!
 Consequence of poor IMC: local weakness in joint strength Cause for poor IMC on solder pad: Insufficient soldering Primary failure cause: varying flow of coating under FPGA, thermomechanical impacts overloads regions with missing IMC.
 Countermeasure: No coating allowed under FPGA, improved solder profile.
 
 The confidence of this conclusion is better than 50/50, best I could do.
 Look at last photo, which I send to Steve. Follow the IMC from right to left. Seen even better in SEM, but this is THE END.
 
 Thanks all
 
 Inge
 
 
 
 
 -----Ursprungligt meddelande-----
 Från: TechNet [mailto:[log in to unmask]] För Hfjord
 Skickat: den 9 maj 2007 20:04
 Till: [log in to unmask]
 Ämne: [TN] SV: [TN] SV: [TN] SV: [TN] Metallurgists, need assistance
 
 Dave,
 
 Goofy, agree. No, I have not confirmed the ENIG finish, because my client did not send naked boards to me. I will ask for a fresh board so that we can check the condition. That's a good thought, to investigate the parts before assembly, the board, the solder paste and the FPGAs. My client denies there could be anything wrong here, but it can, of course.
 I'll call them tomorrow.
 
 As you could read, one guy pointed out that the voids can be created at the polishing process, because of Lead dissolved by the water. Sounds goofy too. Suppose that could happen, why don't I get that when polishing the good references then. I've seen SnPb cross sections for years, but never seen Lead being superseded by voids!
 
 OK, can try oil polishing instead...
 
 Inge
 
 -----Ursprungligt meddelande-----
 Från: TechNet [mailto:[log in to unmask]] För David D. Hillman
 Skickat: den 9 maj 2007 14:12
 Till: [log in to unmask]
 Ämne: Re: [TN] SV: [TN] SV: [TN] Metallurgists, need assistance
 
 Hi folks! I agree with Vlad - I don't see any "B" word characteristics in the nickel plating. Looking over the SEM photos, something just looks goofy. As Vlad pointed out the voids don't seem to be from a flux outgassing root cause - they contain rough, jagged features. Inge - have
 
 you run a SEM EDX analysis to get a chemical ID on the structures inside
 
 the voids? My guess is that the cause of the poor solder joint integrity
 
 is linked to the formation of your strange void structures. I was thinking maybe gold embrittlement but if you are using ENIG then you don't have enough gold present. Have you confirmed that the ENIG finish was correct?
 
 Dave
 
 
 
 Vladimir Igoshev Sent by: TechNet 
 05/08/2007 08:01 PM
 Please respond to
 TechNet E-Mail Forum ; Please respond to Vladimir Igoshev 
 
 
 To
 [log in to unmask]
 cc
 
 Subject
 Re: [TN] SV: [TN] SV: [TN] Metalurgists, need assistance
 
 
 
 
 
 
 Hi Inge,
 
 The images are nice and clearly prove we won't have to use the "B" word any more (at least not in that case). 
 
 But I got even more confused :-)
 
 My uneducated guess would be that the first two images were taken in the backscattered electron topo mode. That would explain why there are plenty of blubs/hillocks (or whatever one'd prefer to call them) on the surface. 
 That is how surface CAVITIES (not protrusions) can look like in that particular mode.
 It's sort of an optical illusion. 
 
 Your other pictures are in the secondary electron mode and here the cavities do look like cavities.
 
 If I'm right, then I'd have to understand what those black spots are, as
 
 they don't look like voids (voids woud look like protrusions and I'd expect voids to be more rounded).
 
 The interface on the board side looks normal on all images (and from my understanding that is the interface you are having the problem with). 
 
 I don't see any Ni on the lead side and I don't like how the intermetallics looks like there (there is kind of a "demarkation" line inside the layer, which might bring up a couple of speculations on is origin).
 
 The intermrtallics is most probably (Ni-Cu)/Sn, not Ni/Sn as you have a huge source of Cu (cut off lead). It even looks like Ni-Cu/Sn, not Ni/Sn.
 
 And now we got down to the most "misterious" part :-) - those cavities in the surface. I might be wrong, but I think I either heard or read somewhere that Pb is etched away during polishing (reaction with water).
 
 That would explain why you see the cavities. To prove (or disprove) my point, I'd take pictures from the same areas in backscattedred COMPOSITIONAL (not topo) mode.
 
 And after all, I'd love to see how a failed joint look like in cross-section (with the lifted lead) :-)
 
 BTW, did you see those Sn spheres? :-)
 
 Regards,
 
 Vladimir
 
 
 
 -----Original Message-----
 From: Hfjord
 To: Vladimir Igoshev; 'TechNet E-Mail Forum'
 Sent: Tue May 08 14:44:42 2007
 Subject: SV: [TN] SV: [TN] Metalurgists, need assistance
 
 Vlad & TN gnus,
 
 Hey, curious? I'm still concerned, but at a higher level, as said Newton. To begin with, I Xsectioned samples from failed part AND known good references. Both finished with 0.1 micron diamond milk. Etched for 10 second, 5 seconds, 1 second and no etching at all. Same result: voids in the failed parts, absolutely flat with references. I'll send photos to Mr Wallman, alias Steve.
 
 Pic new xsections_1 : typical solder joint with void concentrations under the lead's foot. Not etched.
 
 Pic new xsections_2 : details from under the lead's foot. Nickel plating seen clearly, and a very thin nickel IMC layer. All bright areas are predominated by Tin, very little Lead. Black spots are voids that contains a little of everything debris from polishing. Not etched.
 
 Pic SEM_3 shows following: from bottom copper pad / solder joint with voids / lead with nickel finish. Not etched.
 
 Pic SEM_1 shows upper part of solder joint, with lead copper uppermost.
 Not etched.
 
 Pic SEM_ 2 shows typical solder joint at higher magnification with predominating Tin and also little Lead and Copper. Not etched.
 
 As I said earlier, the references show no voids at all like the above.
 
 Preliminary, I have excluded Black pads and Skip plating. Furthermore, I don't think there are brittle copper IMCs either, even if there are relatively high Cu peaks everywhere. Think Cu peaks come from preparation. Nor is it likely with Ni IMCs, they belong to the interface parts. So, what is this? Someone mentioned a insufficient soldering process, and that is finally what I begin to think too. Against this speaks the very experienced and big company that makes these boards (one of the big elephants, not Ericsson, not American). On the other hand, even the best can fail. 
 
 The failure appears randomly around the world, much annoying for the customers, and the technicians scratch their heads. 
 
 Finally, I remind you all about the fact, that the fractures start close to the solder pads and spread to the middle of the solder joints. And only on the large FPGAs. No other components are hit.
 
 Inge
 
 
 
 -----Ursprungligt meddelande-----
 Från: Vladimir Igoshev [mailto:[log in to unmask]]
 Skickat: den 8 maj 2007 15:47
 Till: TechNet E-Mail Forum; Hfjord
 Ämne: RE: [TN] SV: [TN] Metalurgists, need assistance
 
 So inge,
 
 What the cross-section looked like? :-)
 
 Vladimir
 
 -----Original Message-----
 From: TechNet [mailto:[log in to unmask]] On Behalf Of Hfjord
 Sent: Monday, May 07, 2007 4:57 PM
 To: [log in to unmask]
 Subject: [TN] SV: [TN] Metalurgists, need assistance
 
 Diet Coke! Man! Does not remove black pads on your teeth..
 New samples are ordered, will try both oxides on paper and diamond on nylon.
 If the customer pays, I may try even FIB or TOFSIMS. Both are superior, and even if the time tax is high, I must say that polishing, and repolishing and endlessly repeated polishing can be quite expensive as well.
 Inge
 
 -----Ursprungligt meddelande-----
 Från: TechNet [mailto:[log in to unmask]] För David D. Hillman
 Skickat: den 7 maj 2007 14:23
 Till: [log in to unmask]
 Ämne: Re: [TN] Metalurgists, need assistance
 
 Hi George! Good comments! It does seem that the industry, at times, spends more time in describing a phenomena rather than resolving the root cause
 
 so that we can avoid a reoccurrence. As for Inge's Monday dilemma, I would wait and see what the cross-section results reveal. If the observations are that the nickel plating grain boundaries look bad, then I would work
 
 with my pwb fabricator on what went wrong in the plating bath. It the observations are that the nickel plating grain boundaries look good, I would chase either a poor solderability or a poor soldering process root
 
 cause. As you noted, the plating chemistry suppliers have worked extremely hard on making the ENIG plating process more robust and "real" black pad
 
 problems are less prevalent today. I have always viewed the solution to black pad as one of two choices: don't use ENIG or work very closely with the pwb fabricator. Here is an article reference which I think really does the best job to date, of describing the root cause of black pad (one of the contributors is our good fried Dick Coyle!).
 
 Zeng, Steirman, Abbott, Murtuza, "The Root Cause of Black Pad Failure of
 
 Solder Joints with Electroless Ni/Immersion Gold Plating", JOM, June 2006, pp. 75-79
 
 The reality for us may be that no amount of plating bath control will completely eliminate the potential of having a black pad situation and that a risk assessment on a product basis is the final solution. Ouch, way too much thinking on a Monday before I have had a Diet Coke!
 
 Dave
 
 
 
 "Wenger, George M." 
 05/04/2007 07:56 PM
 
 To
 "TechNet E-Mail Forum" , cc
 
 Subject
 RE: [TN] Metalurgists, need assistance
 
 
 
 
 
 
 David,
 
 I hate the "B" word because it describes an optical appearance rather than a failure mechanism. I've heard the "B" word as well as terms like "weak Nickel" and "Black Line Nickel" used many times to describe ENIG failures. The reason I don't like these terms is because they are adjectives to describe what one thinks they see and don't describe the failure mechanism that causes the brittle interface failures. We use ENIG on an old telecommunications legacy product because it wouldn't be profitable to do any engineering evaluations to qualify another surface finish. However, since the root cause of the ENIG solder joint failure mechanism has never been understood or resolved to the point where one can turn it on or off, we avoid the use of ENIG surface finish on all new products. Even though the suppliers of ENIG chemistry have done a great job of monitoring their chemistry and educating board shops on the proper controls to reduce the occurrences of brittle solder joint
 failures they haven't been able to eliminate these type of failure.
 Ingemar is one of the few people I know of on TN that not only has a great understanding of solder joints but appreciates the process control details one needs to follow in PCB fabrication as well as PCBA to avoid un-reliable solder joints. His current problem (i.e., not being able to understand the failure mechanism or predict when it is going to happen) is the major reason we avoid the use of ENIG surface finish. 
 
 Let's just say Ingemar does a cross section and confirms that there is "mud flat" nickel grain boundary attack, what does he do on Monday?
 Your advice about having a beer because it is Friday and not worrying about this until Monday only will help this weekend but come Monday what does he do. I assume that he's already using a "good" board shop and a "Good" ENIG chemistry. I know what we did and it worked for us but as an engineer I'd really like to what the ENIG failure mechanism is and how to avoid it.
 
 Regards,
 George
 George M. Wenger
 Andrew Corporation Wireless Network Solutions Senior Principal FMA / Reliability Engineer 40 Technology Drive, Warren, NJ 07059
 (908) 546-4531 [Office] (732) 309-8964 [Cell]
 
 -----Original Message-----
 From: TechNet [mailto:[log in to unmask]] On Behalf Of David D. Hillman

=== message truncated ===

---------------------------------------------------
Technet Mail List provided as a service by IPC using LISTSERV 15.0
To unsubscribe, send a message to [log in to unmask] with following text in
the BODY (NOT the subject field): SIGNOFF Technet
To temporarily halt or (re-start) delivery of Technet send e-mail to [log in to unmask]: SET Technet NOMAIL or (MAIL)
To receive ONE mailing per day of all the posts: send e-mail to [log in to unmask]: SET Technet Digest
Search the archives of previous posts at: http://listserv.ipc.org/archives
Please visit IPC web site http://www.ipc.org/contentpage.asp?Pageid=4.3.16 for additional information, or contact Keach Sasamori at [log in to unmask] or 847-615-7100 ext.2815
-----------------------------------------------------

---------------------------------------------------
Technet Mail List provided as a service by IPC using LISTSERV 15.0
To unsubscribe, send a message to [log in to unmask] with following text in
the BODY (NOT the subject field): SIGNOFF Technet
To temporarily halt or (re-start) delivery of Technet send e-mail to [log in to unmask]: SET Technet NOMAIL or (MAIL)
To receive ONE mailing per day of all the posts: send e-mail to [log in to unmask]: SET Technet Digest
Search the archives of previous posts at: http://listserv.ipc.org/archives
Please visit IPC web site http://www.ipc.org/contentpage.asp?Pageid=4.3.16 for additional information, or contact Keach Sasamori at [log in to unmask] or 847-615-7100 ext.2815
-----------------------------------------------------

---------------------------------------------------
Technet Mail List provided as a service by IPC using LISTSERV 15.0
To unsubscribe, send a message to [log in to unmask] with following text in
the BODY (NOT the subject field): SIGNOFF Technet
To temporarily halt or (re-start) delivery of Technet send e-mail to [log in to unmask]: SET Technet NOMAIL or (MAIL)
To receive ONE mailing per day of all the posts: send e-mail to [log in to unmask]: SET Technet Digest
Search the archives of previous posts at: http://listserv.ipc.org/archives
Please visit IPC web site http://www.ipc.org/contentpage.asp?Pageid=4.3.16 for additional information, or contact Keach Sasamori at [log in to unmask] or 847-615-7100 ext.2815
-----------------------------------------------------

ATOM RSS1 RSS2