Problems with TI DSP modules

1- DF1: WOW!!! How do people get these cool toys? CB: I am with the University at Buffalo Electronic Packaging Lab. http://www.packaging.buffalo.edu We have pretty much everything that is sold on the market as well as Laser inspection technologies we develop in house. DF2: Nice picture. What do inspect with lasers? Is that part of that Moire Interferometry jazz?

2- DF1: What makes you say a 20-micron intermetallic is necessary in a good package? CB: Intermetallic is a metallic diffusion region. You need thermal energy to facilitate diffusion. When we compare good packages with bad packages we see that good packages (long fatigue life) have a thicker IMC region. We see a lot of packages from different manufacturers. DF2: Where has this been published? It is not conventional thinking. Conventionally, IMC are brittle and fatigue cracks appear along the lead interface to IMC interface when temperature cycles that are fast enough to bring enough shear stress in the intermetallics to break them. For instance, a CALCE study showed there is drop off in strength after the intermetallic thickness exceeds approximately 8-10 microns a standard Sn60 solder joint.

Taking a step backward, I�ve been generalizing about copper-tin IMC, which is wrong in this case. Hopefully there is no copper available to form an IMC on the top of the ball, where the nickel SB blocking all copper reactions. And there is no way a nickel-tin IMC could get anywhere near 2-micron thick, much less 20-micron. So, you must be talking a gold-tin IMC. * Gold-tin IMC can easily get to be 20-micron, but at that thickness it would be so brittle, it would fall from the board in a stiff breeze. * There�s no way that TI would intentionally give you enough gold to grow an IMC that thick.

So, I�m double confused!!! First point of confusion is this �thick IMC is better� thing. Second point of confusion is about the composition of the 20-micon IMC that you are seeing at the ball interposer interface.

3- DF1: Why isn�t a thinner intermetallic was better in the received component, since we�re just going to make it thicker as we reflow the board a couple of times? CB: No actually you don't reflow the previous IMC when you reflow the package on the board side. IMC from the previous process always and should have a higher melting point than the next reflow temp. If you did reflow the previous IMC you would not be able to control the positioning on the pad and you would have misregistered solder ball. DF2: Sorry, I didn't mean to suggest that IMC would melt at reflow temperatures. [I can still read my phase diagrams: Both Cu-Sn IMC melt above 415�C. Two of the three Au-Sn IMC melt above 309�C. The third Au-Sn IMC melts above 190�C, but it doesn't occur all that much.] The point had to do with both the exponential time dependence and Arrhenius temperature dependence of IMC growth, which at its nub says: the more you heat an IMC above absolute zero, the thicker it gets.

5- DF1: What does the delaminated pad look like? CB: I can send you a SEM picture if you want. DF2: I shudda known. I�ll take the SEM pix. What composition of materials do you get?

7- DF1: What�s your thinking on �black pad�? CB: What do you mean by black pad? Different sectors in packaging industry use different terms for same things. I am sorry you have to explain it a bit. DF2: Black pad just about counter-balances all the good things anyone can say about ENIG. The immersion gold works by corroding the nickel. If it is too aggressive, it takes away the nickel and leaves the phosphorous behind. This makes it look like the phosphorous level is too high [GT 9% w] in the nickel bath. Black pad can not be identified visually, you need to do surface analysis. What you will see is obvious dewetting. Since the nickel appears in varying shades of gray, it is commonly called "black pad".

Cogent black pad articles: Nick Biunno's article: http://www.nukcg.org/downloadfiles/Hadco%20on%20Immersion%20Gold%20failures.pdf George Milad's article: http://www.circuitree.com/ct/cda/articleinformation/features/bnp__features__item/0,2133,11666,00.html

However, let's not forget that dewetting can be caused by things other than those things traditionally associated with "black pad" (hyper corrosion of the nickel).

reply »