Farewell to surface mount?
The idea of placing components inside the printed circuit board (PCB), rather then on the surface, has come and gone several times in the past three or four decades.
By Keith Gurnett and Tom Adams
The idea of placing components inside the printed circuit board (PCB), rather then on the surface, has come and gone several times in the past three or four decades. Today, however, the concept is re-appearing with more promise and credibility than it has ever had before. There are even claims that the embedding of components will substantially replace surface mount. However extensive the use of embedded components becomes, the technology is likely to have implications for military and aerospace users.
There are at least two main routes to embedding components, the Chip In Polymer technology that has been developed under the project name “Hiding Dies” by a European consortium (www.hidingdies.net), and the Occam Process proposed by Verdant Electronics (www.verdantelectronics.com). Neither of them uses a standard FR4 PCB, and neither involves forming cavities in the board to drop components into. Instead, both methods start by gluing the components onto a substrate of some type, and then applying one or more build-up layers over the components. The surface of the finished board is typically bare.
For electronics users generally, and for military and aerospace users in particular, there are attractive benefits in embedding. First, performance gets a boost because the interconnections are much shorter. Second, both solder and reflow are eliminated. Problems created by the use of lead-free solders disappear because there is no solder. Third, the components embedded in the board are well protected from shock, vibration, and other hazards. It may turn out that systems made with embedded components can be stored for years with minimal degradation.
Whether embedding will make surface mount obsolete, or whether it will be used only in niche applications, no one can predict. What is presented here is the current state of development in the two chief methods.
Chip in polymer
This method has been in development for several years by a consortium of universities and companies in Europe, including Nokia (Finland), Philips (The Netherlands), AT&S (Austria), Datacon (Austria), CWM (Germany), IMEC (Belgium), and Fraunhofer IZM/Technical University of Berlin (Germany). Development work has concentrated on developing embedding processes that use existing production equipment, without demanding the creation of new equipment. Development has also focused on advanced vertical integration—meaning that the completed board should be as thin as possible.
Development work begins with wafers that are thinned to 50 microns, and sometimes even thinner. The contact pads on the die are modified to make them suitable for PCB metallization. After dicing, the unpackaged die are adhesively bonded to a board core, which may be as thin as 150 microns. This is a critical step since the adhesion must be flat and free from irregularities.
A layer of resin-coated copper (RCC) is then applied over the die. The copper is only 5 microns thick, and the resin 70 microns thick. The resin flows over and around the 50-micron thick die, with the result that the copper layer on top is perfectly flat. It’s hard to make RCC with resin thicker than 70 microns because of the risk of cracks and loose particles.
At this stage, microvias are laser-drilled through the copper and the resin down to the modified bond pads on the die. The microvias are then plated with PCB-compatible copper. A patterned resist is applied to the RCC, and etching forms the traces on the surface of the board.
Fraunhofer IZM physicist Andreas Ostmann explains that development is designed for relatively small boards, probably having not more than four embedded ICs, a number that gives a good chance of viable yield. Passive components can also be embedded, and the consortium has had experience embedding specially thinned ceramic chip capacitors.
At the same time, ordinary ceramic chip capacitors have become so miniaturized that the smallest sizes are thin enough to be embedded in boards slightly thicker than the board described here. One method, Ostmann says, is to attach the capacitors to a core substrate, and then apply a prepreg layer (more than RCC) that has cavities where the components are. The cavities do not need to fit perfectly, since during curing the prepreg will fill in gaps around components.
“In Germany today there is one small company and they embed quite thick components with a thickness of one millimeter or more,” Ostmann says. “In principle, you can embed components of any thickness, but then you would have a stack of prepregs, because each prepreg has a limited thickness, and then of course the PCB would become bulky.”
Another German company, Ostmann reports, is using a patented technology to place a flip chip onto a flex interposer. The flip chip—which is 600 microns thick because it is unthinned—is attached to a board core, with the flex interposer upward. A prepreg with a cavity is then laid on top, and vias are drilled down to the flex interposer. “This technology is much larger than what we are doing,” Ostmann observes, “but it uses conventional technologies for embedding.”
The limitation in size to a handful of ICs is related chiefly to concerns for reliability. Once a component is embedded, Ostmann notes, there no possibility for rework—it has to be right the first time. A board containing 20 ICs and a hundred or so passive components would not be feasible today because the yield from such a production line would probably be close to zero.
Early tests have given favorable results. Test boards at Fraunhofer IZM have endured 1000 hours of standard humidity testing without failures, and 6000 cycles of thermal cycling (-55 to 125 degrees Celsius), without failures.
Most recently, Fraunhofer IZM is working with Bosch and other partners to develop a 77 GHz automotive radar system using the Chip In Polymer technology. The goal is to design a low-cost, high-performance system using 100 micron pitch silicon germanium chips to handle transmission and reception in the harsh automotive environment.
The Occam Process
This method of embedding components has been proposed and is being promoted by Verdant Electronics, a San Jose, Calif., company formed in 2007 and headed by Joe Fjelstad. The company itself will eventually license some intellectual property items, but for the moment is more concerned with coordinating development with companies that have shown an interest in the Occam Process than with profits.
Verdant Electronics presents the Occam Process as a cure for practically all ills—a technology that can replace surface-mount assembly and that can make solder unnecessary. But development of the Occam Process is at a much earlier stage than the development of Chip in Polymer, and there have been few if any carefully tested prototypes.
The technology differs substantially from that developed in Europe. Fjelstad plans to use tested and burned-in components, but the components will be packaged and have leads. The active and passive components are first placed on a substrate (which may be removed later), and then overmolded with a suitable material.
Next the substrate, with components attached, is flipped over. Microvias are drilled from the surface down to the leads on the various components. Now accessible, the leads are first plated electrolessly (electroplating cannot be carried out on insulating material) with copper. Once this “seed layer” is in place, additional copper is added by electroplating. A resist layer is added and patterned, and etching creates the traces. Additional layers of build-up material and traces may be added.
Fjelstad is not focusing on specific materials. Since the development work is being done by firms in the U.S., Brazil, and elsewhere, he wants to leave those firms free to select the materials that will work for them. There are, he says, “extant materials that we have, simple encapsulants, simple epoxies, that could be laid onto it.” He adds, “There are materials and technologies that are being developed that I think would be applicable.” And while he himself envisions using packaged ICs—and particularly chip-scale packages (CSPs) the he helped develop at Tessera—he wants developing companies to be free to use whatever they choose. He predicts that the number of component package types will shrink dramatically. “At the end of the day, from a simplistic standpoint, we should be able to default technology to two types of components: the QFN and the LGA, which are the two that are the most attractive today, but the most difficult to assemble,” he says.
In some quarters, the Occam Process has drawn criticism for claiming too much. Ron Lasky, instructional professor at Dartmouth College, N.H., and senior technologist for Indium Corp. of America, states flatly: “It is absolutely never going to happen. There are just so many things fundamentally wrong with it...no prototype, no complete process description, and cost estimate.” The problem, Lasky believes, is what he terms “technical ecstasy,” in which people see so many possibilities and become so excited that they overlook details that can slow progress.
One Brazilian company is using the Occam Process to make demonstration parts for the Argentine Space Agency. Their early work, Fjelstad notes, resulted in an assembly that could be, and was, “thrown on the floor with no component pop-outs.” One of the overall goals of the Occam Process, he explains, is to progress from the more or less standard drop test to the throw test.
While work at Fraunhofer has been limited to boards with small numbers of ICs, Fjelstad sees no such restriction. If tested and burned-in components are used, and processes are carefully controlled, the Occam Process should work even with complex, high component count boards. “When you take devices that are tested and burned in,” he says, “and you plop them into an assembly—if you’ve done a good job of manufacturing, in other words you reliably made the plated through-holes—then what’s to do? You don’t worry about it, because it will work.”
To Lasky, this is an optimistic view of manufacturing. “It’s not at all a simple process,” he states, and suggests that there are plenty of processing and material glitches waiting in the winds.
Fjelstad envisions that his process will mean the end of solder as an interconnect agent in electronics manufacturing. The transition to lead-free solder, he explains, has “boxed everybody into a place where they are having trouble.” The Occam Process, named after William of Occam, a 14th-century proponent of simple-as-possible designs, could solve all the problems of lead-free solder by simply eliminating solder.
Within six months or so, Fjelstad anticipates that firms working with him will have assembled a number of Occam Process prototypes. “There will be a whole bunch of test vehicles done by then that will be in full test, daisy-chain,” he says. “The problem is that the failure mechanisms will not be the ones that we’re familiar with. Almost all failures relate to solder.
“We’re going to have to go back and determine what the new failure mechanisms will be. Likely, perhaps, some fracture of copper,” Fjelstad says. “But all the effort with microvias has shown them to be essentially bullet-proof. If you’ve done your process right, the vias don’t fail. So it might turn out that the ultimate failure could be component wearout, and that would be something that we’ve never been faced with before. That is the speculative part of it, but it’s really fascinating to me and to others as well.”
One concern is that assembly lines using the Occam Process might be significantly slower than conventional surface-mount lines. Fjelstad doesn’t believe they will be slower—the placement machines that will put the components onto the sticky substrate are accurate, and the targets for the machines will be bigger than they are for surface-mount.
An area that has drawn some criticism is the electroless deposition of the seed layer of copper in the microvias, because the process is widely regarded as slow and messy. Lasky says bluntly, “One of the filthiest processes in the world is electroless plating.” But Fjelstad believes it can be done quickly. “I used to manufacture electroless chemistry,” he explains. “You can do the electroless build-up seed layer in about 10 seconds, following catalysis, you get full coverage and you have a continuous copper film. So it doesn’t take much, although people tend to want to go much longer than that. In my own processing, I never did.” He also hints at new materials under development that would make electroless deposition unnecessary.
Fjelstad envisions the completed board as something like a brick. “After you’ve done all of your circuits and you’ve put your coatings over the top of it, you can take and encase the entire structure in metal, so it becomes hermetic,” he says. “You would essentially seal off all the plastic, so that you wind up with a metal brick.”
At that point, Fjelstad explains, it would be possible to assemble the bricks into new products. “I’m guessing that it would be possible to build virtually a supercomputer in a cubic foot,” he says.
One advantage of the Occam Process is that it lets thermal problems be solved early—a useful advantage for military systems that must endure extreme conditions. “Bernie Siegel, one of my advisors...made the observation [that] this essentially gives the opportunity to solve the thermal problem ahead of everything else, instead of letting it be the last thing. So you can actually embed the thermal solution—you can put a heat pipe in this structure, and it won’t matter.”
A drawback is that rework will be difficult if not impossible. Fjelstad believes rework might be possible. Rework was one of the concerns that Lasky had—“What happens if a component fails?” he asked. But Fjelstad points out that the same features—all components embedded, and then the board encased, if desired, in metal—make reverse engineering extremely difficult.
“I think this would be really important in the military/aerospace environment,” he notes. “Remember the Apple iPod? It came out, and two days later we had pictures of a teardown, we knew everything that was inside it.” It would be possible to tear down an Occam Process brick, he explains, “but it’ll take you a long time to figure out what’s inside it.” He thinks this would be an attractive quality for anyone developing secure military electronics.
It may turn out that electronics produced via the Occam Process have long lifetimes and high reliability. This could, Fjelstad explains, upset the existing order. “Somebody says to me, ‘Do you realize that we have a profit center that is involved in rework and repairing and field service, and it’s pretty profitable?’ With little or no rework, and with few failures because the traditional cause of failures—solder—has been eliminated, such efforts might no be needed on today’s scale.
Fjelstad believes that Occam products at the end of their useful life will be treated like a high-grade copper ore. “There’s nothing that says you couldn’t just grind it up, extract the metals back from it, and take whatever polymers are there. It’s not inconsistent with what people do today for electronics when they grind up things.”
Lasky disagrees strongly with this view. “You don’t just grind them (electronics) up into a powder and then use chemistry to separate all that stuff. The way it’s most commonly recycled is to be carefully taken apart. You re-use the electronics that you can in low-quality electronics that are using recycled components. That’s why you can go and buy a little Sudoku game in a thrift store for $2.99. It’s got used components, or components that didn’t pass some more strict tests.”
At this moment it is impossible to say in what areas and to what extent embedded will be accepted. As Fjelstad points out, new technology cannot be pushed; it needs to be pulled by people who need it. But in its broad outlines, this technology holds much promise for military and aerospace applications.