Never Enough Memory
Anyone who has ever labored at a computer knows the one element you can never own enough of is memory. While the first IBM-compatible personal computer (PC) shipped in 1981 with no hard disk drive but only a 240K floppy diskette drive, today a PC equipped with a 500 megabyte hard disk drive is de rigueur. A group of engineering professors at Berkeley, in collaboration with staff researchers at International Business Machines (IBM) , are attempting to satiate our unquenched thirst for computer memory not by thinking massive, but instead, by thinking micro.
Micro-electrical Mechanical Systems (MEMS)
In the spring of 1995, the Department of Advanced Research Projects (DARPA) granted Berkeley professors Bernhard Boser, Roberto Horowitz, Roger Howe, and Al Pisano and five IBM researchers, including Long-Sheng Fan, a former Berkeley electrical engineering Ph.D., a whopping $5.437 million to design and fabricate micro-electrical mechanical systems (MEMS) to increase exponentially the storage capacity of disk drives.Howe said, "This grant was the result of the Herculean, rainmaking efforts of Al Pisano, and the critical initial seed money and ideas provided by the Berkeley Sensor and Actuator Center (BSAC)," of which Howe and Pisano are directors.
BSAC, founded in 1986, is a Berkeley cooperative research center comprised of the chemical engineering, electrical engineering, materials, and mechanical engineering departments. The National Science Foundation and an array of industry powerhouses ranging from Analog Devices to Ford Motor Company to IBM to Texas Instruments, Inc., furnish BSAC's funding.
Pisano said the historical background of the MEMS-DARPA grant may provide a prototype for future funding endeavors. In the summer of 1993, Pisano and Howe were contemplating the appropriate application for MEMS, or micro-machines. Pisano said, "we were looking for a high-tech application in a high-tech industry where we could do special things which no one else could do."
Purpose of MEMS
MEMS are designed and fabricated at BSAC for a myriad of uses. For example, Roger Howe and his graduate researcher Chris Keller developed MEMS micro-tweezers. Molded from silicon, these tweezers can pick up and move parts ranging from 20 to 100 micrometers wide (the thickness of a strand of human hair is about 100 micrometers). These micro-tweezers may be pivotal to the manufacturing process of MEMS during which infinitesimal parts must be picked up and placed onto a small silicon die.
History of MEMS
For a year, Pisano, Howe, and a group of eight graduate students met once a month to investigate how MEMS could apply to the disk drive industry. They selected this industry because of its enormous growth potential. Between 1989 and 1993, the number of disk drives, shipped worldwide grew at an astronomical whopping annual rate of 82 percent each year.
While "the idea that there might be a MEMS application in disk drives was clear in the late 1980s and early 1990s, the whole issue needed to be radically rethought and the incredible energy needed was provided by Pisano who spent hours and creative thinking revisiting the problem," said Howe.
By the summer of 1994, Pisano and Howe's ideas for MEMS disk drives had crystallized. Serendipity was the catalyst which converted these ideas into reality.
First, In the fall of 1994, DARPA announced it would entertain large-sized proposals for MEMS applications.
Second, IBM, who pioneered the disk drive industry in the 1950s, began investigating MEMS for disk drives at its Almaden, California facility.
The Berkeley team decided it needed an industry partner and connected with IBM through Fan and their long-standing collaborations with IBM in other related fields. With all the pieces in place, the Berkeley-IBM team submitted a proposal which DARPA accepted in early 1995.
The team's MEMS disk drives, composed of eight different device designs, will be revolutionary both in their design and their fabrication. Disk drives are mainly marketed in two form factors, sizes, those with 3.5 inch disk platters for desktop PCs and those with 1.8 inch disk platters for portable PCs. Other sizes, such as those with 2.5 inch and 1.3 inch disk platters exist, but they are not currently commercially successful.
Pisano said, "the market for the smaller drives hasn't blossomed probably because the cost per megabit is too high and there aren't any compelling reasons yet to justify this increased cost, however, MEMS may change all that by making these smaller disk drives more desirable because they will have greater storage density, consume less power, weigh less, and be more tolerant of vibration."
The Berkeley-IBM MEMS disk drives will increase the drive's density, regardless of the form, from the present-day one Gigabit/in2 to 10 Gigabit/in2 within the next two years. Since disk drive manufacturers traditionally upgrade their drives every six months, resulting in ever cheaper, faster, and more dense devices, the team is also aiming to achieve a 40 Gigabit/in2 density in about five to seven years in order to leapfrog over any future or current competitors.
How Will They Do It?
If you opened up your disk drive, you would see a magnetically-coated aluminum disk with a steel disk drive arm poised over it looking like the stereo receiver/phonograph album combination of yore. The arm contains a slider with a record/playback head at the tip which reads and writes data, in the form of magnetic storage bit cells, or 0s and 1s, in concentric circles on the disk. Data isn't packed densely, the width of a bit cell is between 16 and 21 times the length, because the servopositioner that controls the slider isn't very accurate.
The team proposes placing a MEMS silicon microactuator with feedback microsensors between the slider and the record/playback head to position the slider over a narrower range of the disk, at a higher bandwidth, and with greater accuracy, than is currently possible by positioning the entire disk drive arm. Horowitz and Howe have been designing similar microactuators for years.
Horowitz said the team's microactuator will be fabricated from polysilicon whereas IBM's similar microactuator will be fabricated from steel; the team will tested and fabricated both microactuators and the better design will be employed.
In addition to the microactuator, the team also proposes adding a high-voltage complementary metal-oxide semiconductor (CMOS) circuit to generate locally the necessary voltage to drive the slider microactuator. A disk drive arm manufactured from silicon will replace the current stamped steel disk drive arm. Because silicon is lighter than steel but withstands stress better the new arm should aid in making the MEMS disk drive lighter but also more tolerant of vibration, especially crucial in smaller disk drives which users may tote around in handheld or portable computers. The team foresees also adding an integrated shock and acceleration sensor array and servo system to assist the disk drive in sustaining its normal read and write disk operations in the face of environmental factors, such as being used in a car, bus, airplane, or even walking.
At the front of the disk drive arm, the team proposes adding an actuator to help avoid disk wear when users power up and power down the disk drive and adding a optimally-damped micropositioner flexure to help the actuator during interruptions of electrical power.
Finally, the team plans on replacing the current loops of copper wire connecting the slider to the disk drive arm, because these wires can accidentally hit the disk surface during vibration. Instead, they will use a gimbal, a pivoting mechanism which allows the disk arm to freely swing in multiple dimensions while maintaining its orientation. Ships' compasses were suspended in gimbals so the compass remained level when its support, the ship, was tipped or rolled. This gimbal will increase the vibrational tolerance of the disk drive. The combination of these eight advanced designs will yield a disk drive which, when used in a PC, will be faster, lighter, more dense, and more portable than the drives contained in current PCs.
Fabrication of MEMS disk drives will be revolutionary. Keller, a former IBM disk drive engineer, is a materials science Ph.D. candidate at Berkeley. He has designed a drastically new MEMS fabrication process to manufacture the disk drive arm, the manufacturing cost of which constitutes about 70% of the total manufacturing cost of the disk drive. The team also contemplates using IBM's 50 years of fabrication expertise to design at least six avant-garde manufacturing methods for the other MEMS components.
While the disk drive industry garners $20 billion in annual revenues, it's a highly competitive field and profit margins diminish daily. Disk drive parts, like the arm, therefore are stamped out of steel and mostly manually assembled using low-wage, offshore workers in countries like Singapore. This process, however, "does not permit the precision in shape, number of features, or stiffness required by the next generation disk drives," said Keller. As Howe said, "there's a limit to what humans with tweezers and microscopes can do." In addition, this process cannot currently be situated in the U.S. where labor wages make manual assembly cost prohibitive for most products.
Addressing Key Issues
Keller's system, and the team's other proposed manufacturing techniques, address these problems. In Keller's automated fabrication process, the shape of the disk drive arm is etched out of a reusable silicon wafer via a series of chemical baths and fills. Etching can be as shallow as 100 microns, the thickness of a strand of human hair. Using this sophisticated system, and other advanced MEMS manufacturing techniques which will be designed, miniature features can be manufactured, assembly can be automated and simplified, parts used can be reduced, and total assembly cost can decrease.
"This sophisticated manufacturing can be located in the U.S. helping the U.S. retain its competitive edge in fabricating high-tech devices such as Intel's Pentium chip using highly automated, productive plants" said Pisano, "this project would revolutionize the magnetic recording industry."
Because about 55 percent of employed people use a computer at the workplace and 13 percent use one at home every day, this is a revolution which will directly affect many of us. As Boser said, "it's easy to work on technology nobody understands but it's more interesting to do something people use every day."
21st Century Model of Research
MEMS disk drives will not only drastically change the personal and portable computers we now use in both our professional and personal lives, they also contain the genesis for a new, improved 21st century model of research, a collaboration between academia and industry. As Boser said, "most grants are in the $50,000 to $100,000 range, which only permits a professor or a graduate student to look at one specific function or feature but this type of proposal allows us to work on real world situations where things aren't bounded by a fence."
This multi-million dollar collaboration between Berkeley, BSAC, and IBM allows faculty members from different disciplines and engineers from industry to combine their expertise to design devices for use in our complex society where solutions are seldom single-sided.
A True Collaboration
This Berkeley-IBM MEMS efforts is a true collaboration. Team members meet about twice a month and communicate frequently by telephone, telefax, and electronic mail and the 10 team graduate students will likely spend summers at IBM pronounced the pairing "perfect." Pisano said, "we chose IBM to be our partners because they know best what is the overall cheapest and best way to manufacture technology." IBM researcher Fan said, "we chose Berkeley because we personally knew the talented faculty members (Pisano, Horowitz, Boser, Howe), the quality of the students, and BSAC's capabilities." Soon, the results of this group's experience and expertise will be spinning around in to a PC near you.