Tiny Springs Could Reduce Microchip Waste
A new manufacturing approach could end the junking of several chips when one fails.
Using springs and glue instead of solder to make electronic connections between computer chips could end one of the electronics industry’s most wasteful habits, say researchers at the Palo Alto Research Center and Oracle.
“The whole industry is based on nonreworkable technology like solder or tape,” explains Eugene Chow, of PARC. “If one chip in a module of several doesn’t work after you’ve soldered them down, you have to throw out the whole thing.”
Chow and colleagues are fine-tuning an alternative approach. They pattern a surface with microscale springs that compress slightly under a chip’s weight, and these form a lasting, secure electronic connection when the two surfaces are glued together. “You can turn it on, and if it works great, do a final bond with adhesive,” says Chow. “If it doesn’t work, you can just take off the die that failed and replace it.”
For now, the collaborators are developing their springy approach for the high-performance processors used in supercomputers or high-end servers. These chips are combined in closely packed groups known as multichip modules. Such modules need the processors to be packed closely together in order to speed the transfer of signals between them.
“I think it’s just a matter of course that this approach will get to the lower-end applications, too, though,” says Chow. “Eventually this could be in a high-end cell phone–everyone wants to get more chips into everything, and this can help, because the pitch [the horizontal distance between connections] can be so small.” The team has shown that their springs can be made as close together as six microns, compared to the tens of microns necessary with solder connections.
The springs are flat metallic strips that curve up from a substrate that a chip is fixed to. “Fundamentally it’s the simplest spring you can imagine,” says Chow. The spring-building process starts with the addition of a thin titanium layer to the substrate. On top of this, the spring material is deposited in such a way that builds strain into the top layer. Photolithography is used to carve out the outlines of the many springs before the titanium is etched away from underneath.
“The tension makes the springs simply pop up,” says Chow. “It’s an elegant way of making a three-dimensional structure.” The finished spring is coated with a layer of gold for added strength and a better electronic connection. The manufacturers must design the layout of the springs so that they match up to the contacts on the chips. Small sapphire balls or other peg-like structures on the surface of the substrate fit into notches in the chip to ensure the two are positioned correctly.
Last month, Chow and colleagues presented their work at the Electronics Components and Technology Conference in Las Vegas. They showed that their approach works on a test chip from Oracle that simulates the electrical and thermal behavior of a high-end processor. “It’s a test vehicle to evaluate the finished module,” Chow explains. The test chip has nearly 4,000 180-square-micron cells, each containing a thermometer, sensors to measure the power supplied to that part of the chip, and a heater so that the overall chip pumps out the same heat as a high-power processor working at full capacity.
Another reason to think beyond solder, says Chin Lee, a professor of electrical engineering and computer science at the University of California, Irvine, is the fact that it will soon limit the industry’s ability to make ever-smaller devices. “Alternatives are needed, because solder is not going to continue to shrink,” says Chin.
Manufacturers can position the electronic springs more accurately than solder, and this can boost performance, for example by letting them arrange the chips in more compact groups, says Chow. In the race to make faster chips, he says, chip makers can often overlook the ways that components are connected and packaged. “This isn’t a glamorous field,” says Chow. “Everyone focuses on transistors and components, but packaging is a real bottleneck for performance.”
Bahgat Sammakia, director of the Small Scale Systems Integration and Packaging Center at Binghamton University, agrees. “You can have the best technology in the world, but without packaging, you won’t get the best performance from them; it is what enables the creation of the finished systems we are aiming for.”
Sammakia says that although research into novel approaches to packaging chips is valuable, ultimately the market must decide whether a particular solution will work. “You can always solve a problem, but not always in a way that is commercial.”
Jennifer Ernst, PARC’s director of business development, says the project is being directly shaped by what is possible at commercial scale. “Our first priority is to get this into manufacturing,” she says. She notes that the springs are made simply, using just a few layers of metal and standard deposition and etching processes. “We are currently making these at our own fab, but expect the volume to be cost-competitive at commercial scale,” she says.
Be there when AI pioneers take center stage at EmTech Digital 2019.Register now