BGA/CSP Soldering: More Consistency Needed

As electronic assemblies pushed to higher circuit density, array packages arrived achieving higher circuit density compared to perimeter leaded packages such as the QFP. One concern at that time (>10 years ago) was that the BGA solder joints were not visually inspectable. This concern however was overcome by the promise of much higher yields for array packages. At that time, fine-pitch QFPs did suffer from solder assembly yields from solder bridging and other process problems. Consortiums were formed, papers were written and degrees were given. The conclusion was that with ordinary process optimization, the yields would be so high that the array packages would produce more consistent, high yield solder joints.

As of this writing in 2010, based on a reading of industry publications (ref 1, 2, 3),. BGA solder issues are quite common and the lack of visual inspection is a great hindrance to diagnoses. In fact, there is now an instrument (ERSAscope) which was developed specifically to address BGA solder problems. Meanwhile the fine-pitch QFPs show very high yields thanks to improvements in solder paste and processes. Although one might think this is just a yield issue, the difficulty in test and inspection raises reliability concerns.

When RoHS was introduced, pad cratering surprised the industry as the higher temperature PCB laminates became more brittle and therefore prone to cratering. Recently, Head-and-Pillow seems to be a particular problem judging from references 1, 2 and 3 below. This is partly due to increase warpage of components at higher reflow temperatures although it could happen at lower tin/lead processing temperatures as well. Warpage of BGAs as not been focused on as much in the industry including by component suppliers. There are also other BGA solder issues that are beginning to appear. As the BGA/CSP packages become finer in pitch (0.8 mil and below) and the pad areas get smaller, any contamination or marginal solderability on the substrate becomes a problem. Other packages such as SOICs and QFPs use a larger pad area and are not as sensitive to these conditions. The prediction from this blog is that more BGA soldering studies will be performed, more papers will be written, more tutorials will be given.

References

1. Head-and-Pillow SMT Failure Modes, Intel, SMTA International

2. Telecommunications Case Studies Address Head-in-Pillow Defects and Mitigation through Assembly Process Modifications and Control, Acatel-Lucent, AEPX 2010

3. Awakening from Head-and-Pillow: A Novel Pre-Production Test Method for BGA non-wet Issues, Senju Metal, SMTA International

A Galactic Failure

In my blog entry of June 4th, I described increasing concern of wear-out of multilayer ceramic capacitors. This increasing concern is driven by demands for increasing perfomance (CV) and minizaturzation of of chip sizes. Another concern in electronic systems is the increasing susceptitibility of soft errors from radiation sources. The two major sources of radiation upset are cosmic radiation and semiconductor packaging materials. I find it interesting to think about the sea of (mostly) protons which with were created by exploding stars millions of years ago and then traveling millions of miles, finding their way to the solar system, to earth, to North America, to a specific city, to a specific street, to a particular electronic system, to a single IC and then cause a high reliability system to fail. These particles were well on their way before any electronic systems were even thought about. Although the discovery of radiation effects on electronics was known in the 1990s, it has become an increasing concern is the last several years as IC geometies go below 90nm and Vdd voltages are at 1.5v and below. The secondary neutrons are more likely to cause a soft upset when the critical charge, Qc, gets smaller. Also, the natural radiation found in semiconductor packaging materials which previously did not result in a major problem, now must be calculated in the total system reliability. Most of the electronic systems in use now are consumer electronics where a device upset every 6 months is of no concern. However high reliability systems such as in medical electronics must carefully consider soft error rates. All high reliability system manufacturers should have some expertise in this area. IC suppliers must be asked to provide data on soft errors and then the data must be interpreted by a knowldegable engineer. The design of the IC can have an important effect on SER and the suppliers should have either tested or modeled small geometry devices and provided data to the system designer. This data should include effects from both cosmic radiation and semiconductor packaging materials such as molding compounds, lead solder and BPSG passivation.

Reference: SER - History, Trends and Challenges, Cypress Semiconductor, 2004

Note: Some of the early incidents of radiations effects were, in retrospect, somewhat humorous. For example a semiconductor fab facility was shut down for months due to radioactive phosphoric acid derived from radioactive bat droppings from Jordan Mountain which is near an old uranium mine. An apparantly insignificant decision cause a major problem (the "butterfly effect")

The Reliability Cycle?

DfR Solutions has published a white paper entitled The Reliability Cycle: Understanding the Booms and Busts of Reliability in Electronics. In the 1950s and 1960s, large OEMs built-up impressive reliability programs in response to numerous reliability problems. The thesis of the paper is that large OEMs are reducing reliability staffs to reduce overhead costs. The paper further statest that companies are now vulnerable to major reliability problems as component and contract manufactuer suppliers are entrusted with product reliabilty. After some major problems, the reliabilty cycle pendulum will swing back to the other side.

Certainly no one has a crystal ball in this area. While I do see increased vulnerability in depending on suppliers, the reliabilty cycle (?) will not swing back to the 1960s. The days of reliability predictions based on a generic formula are gone (limited usefullness). The days of standard reliability qualification testing for either parts or systems is also gone (can't get the product to market). Instead what is needed is knowledge - or as Edward Deming would say profound knowledge. The best way to proceed in these fast paced times is hire, develop, nurture reliability masters with a deep knowledge of the physics of components and failure mechanisms. This type of knowledge is more important than knowledge in traditional reliability statistics. From this profound knowledge come keen observations and indentification of tests pin-pointed to answer specific questions. Some of these tests are simply ad hoc tests for rapid asessments. Assessments come in a timely manner. Useless tests can be ignored. Where to find these individuals? Since this specialty area is not taught in schools, first look for the old timers - those who have seen many failures. Secondly, start a mentoring program for the younger engineers who want this to be their life's work. Avoid those who are the fast track to management and will quickly leave the field. Also, encourage networking in professional organizations so that we can all learn from other's experience. The old does sometimes cycle back but it usually returns in a different form.

Wear-out of MLCC?

Passive Component Magazine, March 2010, reported on the 2010 CARTS held in New Orleans on March 16. Clive Hendricks, Intel Corporation, is quoted regarding high value multi-layer ceramic capacitors from 2.2uF to 100uF. The dielectric thicknesses have decreased to sub-micron thickness.

Clive Hendricks - "Up until recently, failure due to dielectric wear-out was not a concern for the capacitors used to support CPUs, in fact, our reliability models showed that the capacitors could be used for thousands of years before the insulation resistance would begin to degrade ... in the last five years we have noticed a disturbing trend - as the capacitance density has increased, the usable life has reduced to hundreds, then tens and now even less than five years".

As the market is demanding more energy density in capacitors of all types, the result is decreasing margins for reliability. Historically, the multi-layer ceramic capacitor (MLCC) use-rating is 10% of the breakdown voltage to achieve a reliability margin. This means that the breakdown voltage must exceed 80V/um (ref. 1). As the dielectric spacings are reduced there will be pressure to reduce that reliability margin. At higher temperatures, the use-rating must be even more conservative. There is also a decreasing trend in the number of dielectric grains per layer (ref 1). Fewer grains per layer can result in increased probability of a defect which can bridge the internal electrodes. Reducing the grain size also reduces the K of the dielectric.

It is sometimes assume by the circuit designer that all the parts found in the supplier's catalog have equal probability of failures due to defects. It would be prudent however not to design-in the parts with the highest CV values as these parts are pushing the technology envelope.

With regard to inherent wear-out mechanisms, the same unsettling trends are also appearing in high density VLSI design as well PCB technology (the subject of the future blog entry). Only 10 years ago, electronic assemblies did not have any inherent wear-out mechanisms except in a few special circumstances such as power electrolytic capacitors and batteries or extreme environments. Now it appears that OEMs must pay more attention to wear-out mechanisms. The component suppliers will also be challenged to supply predictive reliability models along with the component. As these reliability concerns become more known, could we be seeing a renewed emphasis on reliability engineering within electronic design companies.

Reference 1: Thin Film MLCC, M. Randall, et al.,2007 CARTS Symposium Proceedings, March 2007

Electrical Component Reliability Blog

The intention of this blog is to comment on industry trends and challenges in the area of electrical assembly reliability with particular focus on electronic component reliability. I believe there are a number of challenges that are quite interesting. As anyone who has written an article can attest, there is nothing quite like writing about something to sharpen one's thoughts. If you want to learn then either write or teach. For myself, this blog is somewhat of an experiment and it will be interesting to see where it will lead.