Putting Passives In Their Place

Resistors, inductors, and capacitors are disappearing from view, integrated into the circuit board itself

Illustration: R. Ulrich & L. Schaper

Open any cellphone, GPS receiver, camcorder, computer, or other consumer electronics system and what will you see? Less than you might expect, but still too much. A circuit board, maybe two, on which are mounted a few integrated circuits and dozens and dozens of tiny discrete devices—capacitors, resistors, and maybe a few inductors.

It is those so-called passive devices that, in fact, dominate the board’s real estate. Consider the Nokia 6161 cellphone, whose 40-cm2 circuit board contains just 15 ICs scattered among 232 capacitors, 149 resistors, and 24 inductors. The phone’s ICs contain millions of transistors, and yet it is those 405 passive components that mostly determine the area of the circuit board and the size of the cellphone.

Now imagine what designers could do if the passives were so small and flat that they could be inserted between layers of the circuit board itself, rather than taking up space on top of it. Phones could be thinner and sleeker than they are today, or they could contain more electronics, such as GPS receivers, or they could simply have much larger batteries and therefore longer talk time and bigger, brighter color screens. The same goes for almost every electronic device that you now stick in your pocket or carry in your briefcase, from PDAs to portable DVD players.

It gets better. These integrated passives would be a part of the circuit board itself, formed when the board was, so odds are good that their overall cost could eventually be less than what manufacturers pay today to buy and solder on discrete devices. Speaking of solder, eliminating it is another advantage of integration, because bad solder joints are one of the most common reasons electronic gear fails. Less solder also means less harm from lead waste.

The list of advantages goes on: putting the passives ”underground” leaves more room on the surface of the board for ICs, which means more design flexibility. And there are electrical benefits, too. Because current travels along a different path in integrated capacitors than in surface-mounted components, integrated capacitors can be made freer of the trace amounts of the inductance, called parasitic inductance, that plagues any capacitor and limits usefulness in high-frequency circuits. Finally, because the components are custom-made when the board is, the resistors, capacitors, and inductors can be sized to any desired value, rather than being chosen from a manufacturer’s list of available parts.

Advantages like these point to a potentially huge shift for the electronics industry. Over a trillion passive components were bonded to boards last year, according to the National Electronics Manufacturing Initiative’s (NEMI’s) road map. These devices are minuscule, and that makes putting them in place a chore. The smallest discrete passives today measure 0.50 mm by 0.25 mm; spread on a sheet of paper, they’d look like ground pepper. Such compact components are difficult to handle and attach, even for automated assembly equipment. And though the total cost of each part—including capital, assembly, and the prorated cost of the underlying board—is less than two cents on average, collectively the impact of integrated passives on system cost, reliability, and, most of all, size, could be enormous.

But for these passives to make a big dent in the US $18-billion-a-year market for discrete passive components, makers of circuit boards will have to reposition themselves as purveyors of passive electronic networks. It’s starting to happen, but slowly. Such manufacturers as Gould, Shipley, Ohmega, MacDermid, DuPont, Oak-Mitsui, 3M, and Sanmina all market products and processes for integrating resistors directly into printed-circuit boards, using at least four different technologies; and for integrating capacitors, using at least five. These sizable companies have all poured tens of millions of dollars in R&D funds into proving the concept. In the meantime, several other companies, including California Micro Devices Corp. (Milpitas, Calif.) and AVX Corp. (Myrtle Beach, S.C.), have been working on an alternative approach to integrating passives. They are selling arrays and networks of miniaturized passive devices in single IC-like packages.

In addition, a few equipment makers are starting to include integrated passives in their products. Motorola Inc. (Schaumburg, Ill.) is leading the pack with integrated resistors and capacitors in some of its newer cellphones, and several Japanese manufacturers are close to introducing products that take advantage of this approach.

In a sense, the situation with passive components today is a lot like that of active devices 40 years ago, when Intel, Fairchild, and others had just introduced ICs that combined active devices like transistors and diodes on a single substrate. But don’t expect Moore’s Law to apply to passives. These components cannot be scaled down into the submicron realm occupied by active devices. The reason, of course, is that passive components have to handle signals whose amplitude cannot be reduced arbitrarily—say, microwave signals going to a cellphone antenna or inputs for analog-to-digital conversion. Despite this fundamental limit, passive integration will make for much more miniaturization.

Integrating passives

Integrated passives are not exactly new. They have been used for decades in the ceramic substrates that underlie circuits in military, microwave, and mainframe computer systems. But those represent a specialty within the electronics market. The vast majority of circuit boards today are made using FR4, the ubiquitous green epoxy insulator reinforced with glass fiber. FR4 boards are formed by sandwiching alternating layers of insulator with etched copper circuit traces and laminating them under heat and pressure. Drilled holes, or vias, plated with copper, connect conductor segments on different layers to form circuit interconnects.

A smaller but growing portion of the circuit board market has been going to ”flex,” which are laminated stacks of unreinforced polyimide (trademarked Kapton), polyester, or layers of other polymer film, each 25 to 125 µm thick, with copper traces on one or both sides. Because the polymer layers can be thinner, enabling smaller vias, flex allows more interconnects to be crammed into a given area than is possible with FR4. But flex costs more per square centimeter than FR4.

In both FR4 and flex, the presence of organic material limits their processing temperatures to about 250 °C, far below the 800 to 1200 °C used in processing ceramic substrates. So to put passives within the layers of FR4 and flex boards, engineers had to come up with new techniques.

The components in these boards can be no thicker than a single layer of the board, maybe only a few micrometers [see illustration, "Size Matters"]. So for all intents and purposes, the devices are planar rather than three-dimensional. Manufacturers are using several different techniques, including sputtering, plating, chemical vapor deposition, screen-printing, and anodization, to deposit various film materials to produce the passives. All of those deposition methods are compatible with the 250 °C limit for FR4 and flex. Depending on the process, technicians can add material just where it is needed, or cover an entire board layer with it and then subtract material where it is not wanted.

For example, resistors can be formed by bridging two copper interconnects on the board with a resistive film. That film can be nickel phosphide plated on a board layer, carbon-loaded epoxy that is screen printed, tantalum nitride that is sputtered, or a ceramic-metal nanocomposite that is printed. There are other possibilities; those are just the most cost-effective options for making boards with a high density of resistors.

For capacitors, the main challenge is finding materials that can be deposited using techniques that are compatible with the materials and processes used on the rest of the board. For example, barium titanate, though common in conventional capacitors, is not an ideal choice because to reach its proper dielectric value—which indicates its ability to concentrate an electric field—it must be fired at a blistering 600 °C, which no polymer board could withstand. However, researchers have found a way to integrate even these high-temperature dielectrics. They can first be fired on a foil of copper, which is then processed and laminated inside the board.

Passive integration may be the only way to handle the future generations of high-power, high-frequency microprocessors

To guarantee that integrated passives will make circuit boards smaller, the material’s dielectric properties must be such that it takes only a small area of a layer of the circuit board to make up a capacitor. The average value of cellphone capacitors is typically 1 to 10 nanofarads and there can be hundreds of capacitors in each board; a manufacturer would have to pack hundreds or even thousands of nanofarads of capacitance into the board to be sure that the integration would reduce the board’s size. DuPont (Wilmington, Del.) and others are developing processes that should yield over 100 nF/cm2, a value good enough to replace many of a cellphone’s surface-mounted capacitors with integrated ones. And researchers are confident that they will soon achieve values over 1 µF/cm2, allowing integration into even smaller areas. For contrast, most current products for making integrated capacitors are limited to polymer-based low-capacitance-density materials good for only about 5 nF/cm2.

Our own company, Xanodics (Fayetteville, Ark.), is commercializing a capacitor process, called Stealth, that is based on tantalum (common in cellphone capacitors). We anodize it at room temperature to create tantalum pentoxide in a solution that is benign to the board and its copper conductors. This forms devices with capacitance per unit area higher than 200 nF/cm2. So a great many capacitors can be integrated onto the same board layer. In addition, the process makes particularly thin capacitors, 0.1 to 0.2 µm thick. This slender profile cuts down on the capacitors’ parasitic inductance, and that makes them handle high frequencies better.

In contrast to capacitors, integrated inductors are a snap to fabricate. They are nothing but spirals of interconnect metal. The challenge is not in the materials or process technology but in their design. The main problem is that any nearby metallic structures, such as interconnects or other inductors, will interfere with their magnetic fields and change their performance [see illustration, "A Very Flat Filter"].

Roadblocks on the way

There are three main barriers to bringing integrated passives into the market: too few design tools, inadequate computer models for predicting costs, and insufficient infrastructure. Better design tools are crucial because taking passives off the surface of a board and burying them generally means that more board layers are required, complicating circuit trace routing. A few design tools can take this complication into account when producing automated layouts; a couple are just now becoming available from Zuken Inc. (Yokohama, Japan) and Ohmega Technologies Inc. (Culver City, Calif.).

The lack of software to analyze costs is also a problem. Before board fabricators will get into the business of manufacturing passive components, they will want to have a pretty good idea of how it would affect their bottom lines. And cost calculations are tricky: unlike discretes, integrated passives cannot be sorted for yield and value precision; one bad component may scrap the entire board. And although most analyses suggest that passive integration can save money, the analyses are very application-specific. For instance, while there is probably a cost advantage to integrating cellphones and other small devices having a high density of components, it may not be cheaper to integrate larger boards, such as those in desktop computers, where size is not a concern. Complicating matters is the fact that the surface-mount world is not standing still; discrete components are getting smaller, cheaper, and more closely packed every year.

The problem of infrastructure is the usual chicken-and-egg story. When plotted against time, technology adoption typically takes the form of an S curve, meaning that little happens at first but eventually everyone gets on board. We are at the bottom of the curve now, but there is evidence that adoption is increasing. About a dozen products for integrating capacitors are on the market now, which is double the number a year ago. Still, board manufacturers may be squeamish until there are enough vendors in the business to guarantee a second source for their materials and processes should their first choice fail.

A killer app

Perhaps what integrated passives need most of all now is a ”killer app,” and decoupling may be it. Decoupling is used in high-frequency digital logic circuits, such as in the motherboards of laptop computers. These circuits place severe demands on power-distribution systems to supply stable, noise-free power during the clock-driven simultaneous switching of millions of transistor gates.

Decoupling capacitors help supply these large current surges, ramping as fast as 500 A/ns, to high-power microprocessor and logic ICs during the switching portions of clock cycles. This technique ensures that the logic voltage levels do not drop unacceptably as a result of the high current demands on the power supply, which may be many centimeters away and connected by unavoidably resistive and inductive conductor planes. Between cycles of current demand, the power-distribution system recharges these capacitors in preparation for the next switching cycle.

With ever-increasing clock speeds, decreasing power supply voltage, and increasing current demand, designers are finding it harder and harder to supply low-impedance, noise-free power to ICs. The main problem is that decoupling capacitors can’t deliver charge quickly, because of their intrinsic inductance.

Decoupling is an obvious first application for integrated capacitors for two reasons: they won’t take up valuable real estate near the power-hungry microprocessor, and their electrical performance is superior in this application by virtue of their extremely low parasitic inductance. Especially on digital circuit boards, surface-mounted capacitors surround the big ICs, often on both sides of the board. Since the speed of the system is often limited by memory access times, eliminating the capacitors from the surface and moving memory closer to the microprocessors would result in a smaller and faster system.

Though special discrete capacitors are being built with fairly low inductance, none of them can compare with an integrated parallel-plate capacitor using a thin dielectric located between the power and ground planes (conductor-coated layers of the board dedicated to either the ground or power supply). For example, thin-film devices that we built on flex at the University of Arkansas (Fayetteville) and Xanodics deliver several hundred nanofarads with less than 3 picohenrys of inductance and a trifling 10 milliohms of resistance. In comparison, a typical surface-mounted capacitor would have several hundred picohenrys of inductance.

Integrated decoupling will likely first appear not in the circuit board itself, but in the small piece of substrate included in the so-called ball-grid-array packaging of high-performance microprocessors. Putting the capacitance layer within the package avoids the intervening inductance of the package-to-board connection.

The intermediate step

Before true integrated passives take hold, we’ll likely see widespread use of passive arrays, in which multiple similar components (capacitors, say) are formed on the surface of a substrate and packaged into a single surface-mounted device like an IC. We’ll also see more passive networks, which combine different kinds of passives in one package. These networks include devices internally connected to form simple circuits such as filters, terminators, or voltage dividers. In either case, one mounting operation replaces many and the overall footprint of the circuit is much smaller.

These arrays and networks are a middle ground—not fully discrete but not fully integrated within a circuit board. They bring some of the advantages of full integration such as a reduced number of placement operations, fewer solder joints, and less board space. Many configurations of arrays and networks are now available in quantity from California Micro Devices, AVX, and other companies, and custom arrangements are also possible. Devices from these companies are typically fabricated on a silicon or other substrate using tried-and-true chip-making processes so the yields are high and the prices reasonable.

The technique raises some interesting possibilities. If ICs or other active devices are mounted atop a passive network, they may form so-called functional modules, such as Bluetooth or GPS subsystems. For example, a GPS module would include passives and antennas integrated on a substrate and one or more ICs bound to it, all in a single chip-scale package. The manufacturer would not have to worry about learning to design and manufacture GPS systems and could also easily upgrade or switch vendors.

On the horizon

Less than 5 percent of the trillion-plus passive devices mounted on FR4 and flex boards this year will be surface-mounted passive arrays and passive networks, and hardly any passives will be fully integrated into the circuit board.

The circuit board business, in the United States, at least, is largely a contract industry, with much of it removed from the designers of circuits and equipment makers. This gulf makes board makers a bit conservative and slow to change relative to, say, the chip industry, where all aspects of development, design, and manufacture are often in the same company. Still, integrated resistor and capacitor layers are starting to become available from reputable suppliers and a few consumer products are showing up with at least some of the passives integrated, and these should lead the way for significant market penetration in the near future.

When, if ever, will more than half the passives be integrated? It’s hard to say. The microelectronics industry is full of cautionary tales; many of us remember questions such as, ”What year will over 50 percent of ICs be made from gallium arsenide instead of silicon?” But some new manufacturing technologies do prove their economic viability and become industry standards, such as surface mounting.

Whether or not passive integration becomes an industry standard will depend on its economic viability. Certainly, it is viable for decoupling and, in fact, may be the only way to handle the future generations of high-power, high-frequency microprocessors. For discrete replacement in general, though, the best processes and materials are still being identified.

If we find suitable technologies, then passive integration will probably show a long, steady climb in use the way surface mounting supplanted through-hole mounting in the 1980s. As the infrastructure, supply chain, and industry acceptance grow simultaneously, eventually integration will gain some significant fraction of the total market and put passives in their place: hidden, ubiquitous, and cheap.

To Probe Further

IEEE/Wiley Press has just published the first book dedicated to this subject: Integrated Passive Component Technology , edited by the authors of this article. In addition, the following articles are informative:

”Moving Embedded Passives from the Lab to the Fab,” by R. Ulrich, CircuiTree, March 2003, Vol. 16, no. 3, p. 64.

”National Electronic Manufacturing Technology Roadmap” was published by NEMI (Herndon, Va.) in 2002.

”Application-Specific Economic Analysis of Integral Passives in Printed Circuit Boards,” by B. Etienne and P. A. Sandborn, Proceedings of the IMAPS Advanced Packaging Materials Processes, Properties and Interfaces Symposium, March 2001, pp. 399-404.

Advertisement
Advertisement