Miniaturization enables innovation—past, present, and future
Shrinking semiconductors stand out in the race to smaller products.
By Maury Wright, Editor in Chief -- EDN, September 28, 2006
Ask someone on the street what miniaturization means to them, and they will most likely indicate a mobile handset or perhaps an MP3 player. Ask an engineer, and you'll probably get a Moore's Law-centric answer. Clearly, those answers are intertwined because one enables the other. And, in fairness, compelling portable consumer devices require smaller everything—speakers, microphones, disk drives, batteries, connectors, and so forth. Complex technologies, such as disk drives, require advancements that rival the innovation in ICs.
Still, the big digital IC that lies at the heart of these products has been in the miniaturization spotlight at least since the Intel 4004 debuted, and the SOC (system on chip) promises to continue as the most important enabler of cool things for some time to come.
|
In this feature: How miniaturization beats the heat —Lewis Counts, Analog Devices The ongoing technical revolution —Gene Frantz, Texas Instruments Miniaturization history is not the full story —Walden C Rhines, Mentor Graphics |
Over the 50 years of EDN's history, few constants have remained in the tech industry, but there has been a constant march in shrinking end products, driven by the shrinking enabling technologies inside those products. We always hear that everything is smaller, faster, and cheaper, but even that saying is not necessarily true. Early transistors sold for far more than the technologies that they would usurp. Increased integration in ICs always makes for smaller products, albeit not necessarily higher performance ones. Only the smaller angle is almost universally correct.
On miniaturization, Texas Instruments Principal Fellow Gene Frantz says, "You can almost say that we are on the path to the vanishing product—where the product will be so small and insignificant in size but so significant in capability that we really don't know where we have it; we just know we have it."
Dean Kamen, founder and head of DEKA (Dean Kamen) Research, has leveraged the miniaturization trend in everything from the Segway to computerized medical instruments. "We've come to expect that electronics have gotten so small, computing power has gotten so cheap, memory has gotten so plentiful, and power consumption has gotten so reasonable," says Kamen. "You put all of that together, and, literally, we're now at a point with electronics, which we aren't with any of the other fields of engineering, where it isn't a question of what you can do; it's a question of what should you do because you know it can be done. What was unthinkable 10 years ago is virtually trivial now."
Looking back, as we at EDN have done for our Milestones That Mattered series and interactive time line, it's amazing that inventors in our industry often didn't realize the true significance of their work—at least at the time of their inventions. And third parties, including EDN editors, were often no more perceptive. For instance, in our look back at Fairchild's introduction of Micrologic Elements (see "The planar IC—revolution underestimated"), EDN originally discounted the significance of the ICs, reasoning that batteries and other components in aerospace applications would render miniature logic insignificant. Of course, that assumption ignored the multitude of consumer applications that would come to rely on digital logic.
In discussing milestones from the tech-industry history, Walden Rhines, chairman and chief executive officer of Mentor, states, "In most cases, the actual components were developed initially without knowledge of what the killer application would be." Rhines claims that such lack of foresight is a typical trait of revolutionary developments. He continues, "If you float a technology out there and make it easy enough to use, someone will discover it and match it up with the end application."
Before working at Mentor, Rhines worked at TI managing the semiconductor group. And both Rhines and Frantz relate similar stories about the birth of the DSP. In the early 1980s when the TMS32010 came to market, TI expected—but didn't find—success in the speech market. The two markets in which they would find early success—PC-graphics acceleration and hard-disk-motor control—weren't on the radar. Later, the DSP would enable the dial-up modem and the digital cellular handset, among other products.
You might wonder about recent examples of technology's success in unforeseen ways, because developers are increasingly designing products for target applications. "As an industry matures, you have less of that," Rhines admits.
Rhines is enthusiastic in looking back at his experience, both at Mentor and at TI, and offering opinions on key milestones in the miniaturization march. He points out the significance of the first germanium and then the first silicon transistors. He claims that the silicon version was especially significant because it was stable over temperature changes and therefore suitable for consumer products such as transistor radios.
Rhines also credits Intel for commercializing microprocessors and memory and especially for moving to NMOS technology, thereby enabling denser memories. And he believes the industry's move to CMOS was ultimately even more significant. Rhines relates that TI had to use low-power CMOS for its work with battery-powered devices, such as watches and calculators. But the challenge was immense. "Fundamentally, you were saying, 'I'm going to put two transistors down everywhere there is one today, and I'm somehow going to make the die smaller, faster, and lower power,'" he says. Rhines credits the Japanese conglomerates for wading through the CMOS challenge and delivering on its promise. And CMOS was clearly the best technology choice for digital ICs and even increasingly for analog ICs.
At the same time that microprocessors and memory were matriculating at Intel and elsewhere, Bob Metcalfe was already thinking about the value in connecting discrete computers. The miniaturization trend would make Ethernet possible. Metcalfe developed the LAN while working at Xerox PARC (Palo Alto Research Center) in the early to mid-1970s and went on to found 3Com and make Ethernet a commercial success.
Metcalfe recounts that the first version of Ethernet in Xerox's labs operated at 2.94 Mbps. After five years of work with Ethernet in a laboratory-type environment, his team evaluated the 2.94-Mbps network and found it to be nearly optimally loaded. The group foolishly thought that it had arrived at the exactly correct speed choice. Looking back, Metcalfe admits, "The fact was that any application that couldn't run at 2.94 Mbps didn't catch on. All the applications that could squeeze through 2.94 Mbps did." Soon, the first standardized version of Ethernet would emerge at 10-Mbps rates. "We settled on 10 Mbps for the first standard because that's how fast the Intel chips could run," he says. And Metcalfe garnered the credit for Metcalfe's Law, which suggests that the value of a network is proportional to the square of the number of users on the network.
Although silicon advancements are fairly easy to chart along Moore's Law and to apply in computers, real-world products require motors, valves, and the like. Disk drives are some of the few electromechanical products to truly ride the miniaturization curve. "Those of us in the semiconductor industry tend to try not to pay attention to the fact that the density of bits on a hard disk have increased at a slightly greater or at least at the same rate as the density of bits on an IC," says Mentor's Rhines.
Even in the real world, though, it's tough to move a miniaturization discussion away from semiconductors. Kamen of DEKA has worked on everything from tiny, implantable medical devices to his iBot, which is somewhat akin to a Segway but can also climb stairs and helps the disabled regain mobility. "The big things for us were DSPs, but high-power control stuff was also important," he says, noting that a device such as the iBot needs both powerful motors and power semiconductors and control devices that can drive kilowatt-sized loads in small packages.
The meaning of "miniaturization" also depends on the context. For a disabled person, an iBot is a miniature transportation device. For dialysis patients, the portable Homechoice dialysis machine, which DEKA designed and Baxter International sells, frees patients to travel and receive needed medical care. Kamen also notes that the lowering of the quiescent power in semiconductors is key to small medical devices, such as implantable products, which must run for extended periods.
Regarding electromechanical products, such as motors, Kamen says that performance is improving but that cost has also increased, compared with the costs in the IC market. Kamen makes the same generalization about batteries. He advocates moving mechanical functions into the electronics whenever possible.
Battery and power issues seem to resonate with everyone from users facing frustratingly short handset-battery life to design engineers managing expectations and capability. The road map of process technology for years has guaranteed power savings. These days, however, leakage current means that savings come less freely. Designers have had to get serious about power management.
"We have just figured out that power dissipation is important," says TI's Frantz, who points out that you must attack the problem both by using better batteries and by increasing the power efficiency of the circuits. "If I reduce the power dissipation of my electronics by half, I can either, with the same battery, double the battery life or, with a battery half the volume, keep the same battery life," he says. The factor of two that Frantz mentions can be significant, but he also suggests that a reduction in power requirements by an order of magnitude could make plausible a battery that is one-tenth the size of today's batteries.
You can save power by increasing the number of transistors on a chip, he points out. The savings can come not only through better power management, but also through more basic chip operation. A dual-processor chip with each processor running at 0.5 GHz from a 1V supply, he claims, offers significantly lower power dissipation than a single-processor chip running at 1 GHz from a 1.5V supply.
Even Intel has low-power-system fever. When the company launched the Intel Core 2 Duo (Picture) in August, President and Chief Executive Officer Paul Otellini claimed that the company began to focus more heavily on performance per watt when it launched the Centrino in March 2003. Intel still pushes performance. On the newest, second-generation, 65-nm strained-silicon process, Otellini says, " We've seen stunning improvement in the transistor-level performance from generation to generation." The Yonah core that Intel announced in January of this year reduces leakage current by a factor of 25 and boosts performance by 40%, he says.
The jury is still out on how much progress the industry has made on low power. Luis Pineda, senior vice president of Qualcomm Chip Technology's marketing and product-management group, is more bullish than TI's Frantz, and both work at companies with a major handset focus in which battery life matters. "We're very advanced with our power-management techniques," says Pineda. Referring to a CDMA-chip set, he says, "One of the chips is a dedicated power-management IC." The power manager can shut down portions of the chip set when necessary and controls the power amplifier, he says. Moreover, it integrates many discrete functions, such as low-voltage regulators and peripheral drivers, from earlier designs.
Pineda also claims that process technology is still helping to reduce power. He points out that Qualcomm is now shipping 65-nm ICs and working toward 45-nm ICs. "There is a lot more work than we can do," he says. He believes that the demands from the consumer—especially to play video—will escalate. Pineda claims, for instance, that the iPod Video can play at most 90 minutes and that consumers will demand more. Referring to power management, he states, "We're probably three-quarters to where the industry can be."
Turning to today and the future, where do we stand overall with enabling technologies for miniature end products, and what are the future trends? Some of the relatively new enablers are amazing. For instance, Knowles Acoustics approximately two years ago announced tiny MEMS (microelectromechanical-system)-based microphones for use in handsets and other small products. The company just followed with the MEMS-based Digital SiSonic microphone, which outputs a pulse-density-modulated bit stream. The company also offers miniature balanced-armature speakers that fit into headsets that rival the best noise-canceling headphones on the market.
Even cabling and connectors may surprise you. Molex, for instance, is shipping coaxial connectors with 0.4-mm pitch. The 40-pin connectors find use in connecting through hinges in clamshell handsets.
Grand surprises are still to come. These days, Ethernet inventor Metcalfe works as an entrepreneur and a venture capitalist at Polaris Venture Partners. He also serves as chairman of Ember Corp, which focuses on the ZigBee wireless standard. Other investments connect Metcalfe to the miniaturization angle. For example, Metcalfe invests in Unison Products, which plans to integrate speaker technology on the surface of a display. The scheme relies on tiny piezoelectric elements along the edges of a display that move a membrane stretched across the screen. Presumably, the scheme will conserve space, and the sound comes from the best possible location facing the user.
Perhaps the biggest advancement might be in energy "harvesting," or gathering energy to power a device from the immediate environment. Researchers are working on converting body heat, vibrations, stray RF energy, and light of all types into energy that could keep a relatively small battery charged. Remember Frantz's comment about a vanishing product? He finished by saying, "As much as possible, it would be good to run a product off body heat or to have no power or battery at all." But is energy harvesting feasible, and, if so, when? "If you think solar power is one of those solutions, it's here," he says, pointing out that TI has long built solar-powered calculators that use a relatively small battery that charges sporadically in lighted environments.
It's probably no surprise that Metcalfe has an energy investment and an interest in harvesting. He states, "Batteries are a big venture-capitalist kind of thing; there are a million battery deals going around." His investment, Infinite Power Solutions, focuses on tiny batteries that generate tiny amounts of power at a low duty cycle. The batteries might find use in applications such as implantable medical devices. A patient might sporadically—say, once a year—visit a doctor to receive an inductive battery charge through the skin, for example. Someday, energy harvesting might continuously charge that battery. Infinite Power is building a semiconductorlike fab to build the flat batteries using vapor deposition on a flexible substrate and is offering a 1-in.2 battery for sampling. You can solder the battery onto a pc board or build it into the board and ultimately perhaps into an IC substrate.
The discussion returns to semiconductor processes even when a battery is front and center. After all, process technology has been the constant enabler of technology. "All new fields lead off with innovations in material science," says Mentor's Rhines, pointing out that mastering silicon-purification techniques was the obstacle that TI, Fairchild, Intel, and others cleared in launching the IC industry, and material scientists drove the innovation.
Meanwhile, other engineering disciplines await such a breakthrough. "We're starting to enter an age now in which electronics might have been a couple of decades ago, when material science, surface finishes, nanostructures, building compounds, materials, and surfaces are getting so extraordinary that we are going to see some big changes in everything from electrical properties to thermal properties to strength and structural properties of materials that are going to be pretty exciting," says Kamen.
He still counsels that you should leverage the maturity of electronics in every way possible. "There are opportunities to move some of the issues out of the world of the mechanical into the world of the electronic and take advantage of the high-performance electronics," he says, citing the transition from the phonograph to the CD to the MP3 player as an example. The CD eliminated contact with the media, and a flash-based MP3 player eliminates the mechanics.
Metcalfe is also a staunch believer in the IC as the key enabler. "Moore's Law is the fundamental; all of the rest comes along," he says. His biggest interest these days is in embedded microcontrollers and especially connecting those microcontrollers, again exploiting Metcalfe's Law. He defines a layer of embedded networking below Wi-Fi, Bluetooth, or cellular in which machines communicate with machines.
|
Ember, with ZigBee, is a player in embedded networking, and Metcalfe believes that wireless is the key. "You can't run cable very well among lots of embedded micros; you had better do it wireless; hence, the development of CMOS radios," he says, claiming that vendors annually ship approximately 10 billion microcontrollers. "That number is going to go up thanks to the arrival of embedded networking," he says.
Not surprisingly, with his semiconductor and EDA background, Rhines believes the future of innovation is in chip design and the number of design engineers that have access to chip-design technology. He points out that, when ASICs debuted, custom-IC folks laughed off the upstart technology because ASICs would offer lower performance and be bigger than custom ICs. During the custom-IC era, however, only a few thousand IC designers existed, he points out. ASIC technology allowed tens of thousands of engineers to design ICs. Rhines claims that FPGAs or another disruptive technology will result in the emergence of 5 million chip designers by 2010 to 2020. "In itself, that should generate more innovation," he says.
