By Tom Adams, Consultant; Sonoscan, Inc.
The idea of embedding components, both active and passive, inside a printed circuit board is nearly as old as printed circuit boards themselves. One great advantage is that embedded components are very well protected; another is that embedding makes it possible to dispense with solder, reflow, and most other trappings of surface mount assembly, although some form of interconnection is still needed. Another great advantage, especially significant for high-performance systems, is the ability of embedded designs to greatly shorten connection distances between the capacitor and the active elements, which thereby reduces the effect of resistance in the supply line. Most recently, the concept of embedding took on new life when a project known as HIDING DIES (High Density Integration of Dies into Electronics Substrates) was launched by a European consortium in 2004. The consortium, composed of the Fraunhofer Institute IZM and other academic and commercial entities, planned to develop techniques for embedding active components. One requirement was that any techniques developed must use existing assembly equipment, and not require specialized new equipment. The consortium envisioned their techniques, when fully developed, would be useful for embedding active components in small numbers, for example, a board or module having at most perhaps five or six active components. One compelling reason to stick with small assemblies is that rework of embedded components is essentially impossible. The difficulty of rework, along with a normal level of process anomalies, mean that it would be very difficult to achieve an economically feasible level of reliability with an embedded board having, say, 40 components.
Capacitors and resistors already had a long history of being embedded, first with the resistor materials marketed by Ohmega, and more recently with capacitor planes marketed by Oak-Mitsui, although these materials have fairly small values unless a relatively large area is available. So, at least in theory, the capabilities existed to embed both active and passive components. But, at the same time the European consortium was developing their methods, multilayer ceramic chip capacitors were being produced in smaller and smaller sizes in order to meet the demands of small circuit boards, such as those in cell phones, where miniaturization in all three dimensions is needed. As cell phones, active components, and capacitors all became thinner, the capacitors eventually reached the point of thinness that would permit them to be embedded.
The embedding process developed at the Fraunhofer Institute IZM aimed at achieving very thin board thickness.
Instead of cutting cavities in an existing board, the process began with the core of a board, which could be as thin as 50 to 150 μ. Active components, thinned to 50 μ, were then bonded directly to this core, without wirebonds or other electrical connections. A resin-coated copper sheet was then applied to the populated core.
The copper was only about 5 μ thick, but the resin was about 80 μ thick, and was fluid enough to flow over and around the 50 μ components. To make the electrical connections to the now-embedded chip, a laser drilled vias down to the chip’s bond pads, and these vias were plated with copper. Selective etching of the 5 μ copper at the surface completed the circuitry.
Fraunhofer physicist Andreas Ostmann observes, in theory, a ceramic chip capacitor of almost any size and
thickness could be embedded if the overlying lamination (which might be more than one layer) is thick enough.
0201s, he notes, are about 300 μ thick, while 01005s are about 150 μ thick.Both are thicker than the 50 μ limit
dictated by the 80 μ resin layer in the Fraunhofer design, but one of the purposes of that design was to pave the
way for other designs, including those that would use thicker components. For components thicker than 50 μ,
Ostmann suggests a different approach, shown in Figure 1. “The process is to take a core substrate, and
then attach the capacitors by solder-
-ing or by an SMT adhesive. Then embed the capacitors in a prepreg layer [laminated onto the top of the assembly] that has cavities in the positions where the dies are. “The cavities do not necessarily have to surround each individual capacitor; they can surround a group of capacitors. During lamination, the prepreg will fill these slightly larger gaps.”
One company in Germany is already using this process to successfully embed components up to 1 mm or more in thickness— a method that opens the door to some of the larger capacitor sizes. Would it be possible to use embedding to create much thicker assemblies? Ostmann thinks that a practical limit would be reached. “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 very bulky.” Other problems would also show up in a very thick assembly. For example, the top diameter of blind vias laser-drilled to reach the embedded components would become larger in area, since their aspect ratio typically is up to 1:1. But, at the moment, it appears that multilayer ceramic chip capacitors up to 1 mm in thickness can be embedded.
Currently, there does not appear to be any manufacturers in the U.S. or in Europe who are embedding ceramic chip capacitors as a part of normal production. However, there is considerable interest in the idea, and it would be somewhat surprising if reports of embedded ceramic chip capacitors were not circulating within six months to a year. Does the embedding of discrete capacitors pose a threat to existing makers of passive planes such as Oak-Mitsui and Ohmega? Bruce Mahler, Ohmega Vice President, doesn’t see a big threat, and thinks there could be an advantage. He comments, “It just points more toward the demand for getting actives and passives off the surface of the board—one more reason to go internal.”
There will be some applications, Mahler points out, where it will make sense to use discrete capacitors and resistors, and there will be applications where planes are more suitable. However, he thinks the extremely small size of some discrete passives could be a problem. “The smaller these little critters get, the more issues there are of reliability, the more issues there are of handling. Whether it’s on the surface or whether it’s embedded becomes a major issue.” Mahler believes that more widespread use of embedded discretes probably adds credibility to the whole idea of embedding components. “If anything, it makes people more comfortable with the idea of embedding,” he notes, “whether it’s a discrete or as a plane. And I think it’s more complementary than it is competitive.” But Mahler doesn’t believe laminating and other processes that make it possible to embed both passive and active components will replace surface mount technology any time soon. “We’re an embedded product, so I believe that embedding is going to be a growing market,” he explains. He expects that embedding might become important in specific applications, such as cell phones.
Will embedding eventually replace a certain amount of surface mount assembly? Mahler thinks it could. “But you’re still talking about a few percent of the overall electronic business. It could be 10, 20, or 30% in the next decade or two, but I don’t see it taking over the entire world because people are going to do what’s economically most viable. Why go ahead and go through the hassle of embedding a component inside of a board if you don’t need to?”
Readers Who Read This Article Also Viewed: (1) Integrated Passive Devices: World Markets: 2003-2008 ISBN # 1-893211-61-4 (2003) (2) LTCC Components & Modules: World Markets, Technologies & Opportunities: 2002-2007 Technical-Economic Analysis ISBN # 1-893211-11-8 (2002).