Terecircuits’ unique manufacturing process could create a new class of custom, stackable SiPs and reduce the bottlenecks in semiconductor manufacturing.
In between the microscopic world of the semiconductor and the macroscopic domain of the printed circuit board, there lies the realm of the System-in-a-Package (SiP), which combines bare semiconductor devices with tiny discrete components to produce a single functional unit. SiPs are compact, use minimal material, and offer high reliability, making them an attractive alternative to their PCB counterparts. One problem with manufacturing SiPs is that it’s still largely a pick-and-place application, like making a PCB but on a smaller scale, and we’re currently reaching the physical limits of robotic pick-and-place capabilities. Another issue is the delicate nature of its components, which can be damaged by robots. Finally, pick-and-place is a linear operation, with one component at a time being placed on the package.
California-based Terecircuits may have the solution. The company has developed a chemical process that works with small, fragile devices and places thousands of SiP components in a single pass, resulting in fewer errors and shorter manufacturing times. Engineering.com spoke with Terecircuits CEO Wayne Rickard, who provided details on the company’s process and explained how it streamlines the supply chain and enables future technology like ultra-high-resolution displays made of microLEDs.
Chemical vs mechanical placement
Unlike the traditional pick-and-place assembly-line process, in which components are moved from a carrier tape onto a PCB or substrate using a thin needle to “poke” them into place, Terecircuits’ method employs a thin (1µm or less) photopolymer layer on a transparent plate to hold and transfer the components. When they’re ready to be placed on the substrate, a UV laser disintegrates the polymer into a gas that gently and precisely “pushes” the component into place, offering more accurate parts placement and reducing the possibility of damaging fragile components. The company claims that its process works with components as small as five microns with sub-micron placement accuracy. A mask that matches any size or shape component allows the release of multiple parts in a single cycle.
“We can do it without damaging the small and thin components and we can do it in parallel,” Rickard says. “So instead of transferring one component, like a typical pick-and-place tool would, we can transfer maybe 10,000 components in a single operation, so you get higher throughput and the ability to handle things on a much smaller scale.”
The ability to insert thousands of devices in one pass could cut manufacturing time down by one to four orders of magnitude, and Terecircuits says the process has been independently tested by display industry OEMs, semiconductor manufacturers and production tooling companies.
Ultra-high-res displays
OLEDs represent state-of-the-art display technology, but their ability to provide enough brightness in daylight settings is a limitation. They also draw a lot of power, which decreases battery life in portable devices like phones, smartwatches and VR headsets. The microLED, roughly the size of a red blood cell, offers a promising replacement technology, delivering ten times the resolution, higher brightness and contrast, less temperature-induced degradation and lower power consumption. In addition to personal electronics, microLEDs will be found in smart vehicle displays, where they can be molded to fit the curved contours of a dashboard, as well as ultra high-resolution TVs.
According to Rickard, the challenge is that these displays can’t be assembled with pick-and-place technology. There are several approaches that are being attempted, and he thinks the industry is converging on laser-based tools that can do a parallel transfer, because they can place thousands of components at a time in order to achieve the necessary throughput for high-volume manufacturing.
One proposed method of assembling a microLED display is PDMS (polydimethylsiloxane) stamping technology. A 64-inch TV requires 6,000,000 microLEDs. Using the PDMS process, it would take nearly three hours to assemble a single display. Rickard says that Terecircuits’ process can do it in ten minutes. He emphasized that these numbers are hypothetical, as commercial scale tools have not yet been built for either process. He expects the Terecircuits “lab-to-fab” transition to be ready by 2025.
Just a little SiP
SiPs combine chiplets, components and ICs into a single package. Terecircuits envisions a day when engineers can combine a particular embedded controller, tiny sensors and analog circuitry into one very compact package. Unpackaged silicon and non-silicon components would arrive at the manufacturing facility, ready to be custom assembled for each unique application. To shrink the footprint, the trend is to build vertically by stacking parts on top of each other, perhaps ten layers high. This means the components need to be thinner, making them even more fragile.
“Some of the components might actually be sensors that aren’t even built with silicon, so the industry’s looking at doing things like 3D stacking,” Rickard explains. “Instead of laying out the components next to each other and interconnecting them, they literally stack them one on top of each other and connect them by drilling through the stack and running a conductor vertically, so you get shorter traces.”
Rickard says that this approach results in less capacitance and better coupling of signals between layers, though there are added challenges like heat dissipation. Altogether, it creates unique opportunities for designers. “This ability to go in the third dimension is equivalent to building elevators instead of roads,” he says.
Rickard believes that the future of electronics is in 3D stacking, with some products using ultra-thin devices stacked ten layers high. Advanced CPUs, high-capacity memory chips, MEMS, sensors and other peripherals will be packaged to fit low-profile devices like 5G phones and smart watches.
“The industry is looking at different ways to build semiconductors,” Rickard says. “It really plays right into our core competency, which is to disaggregate the functions so that not everything needs to be on a 3nm process node for the latest and greatest performance.”
Rickard notes that even the new gallium nitride (GaN) and silicon carbide (SiC) power electronic components will become thinner, making Terecircuits’ process useful to manufacturers of power converters and affect the EV and renewable energy industries. “You have to thin the components down, which doesn’t necessarily change their electrical characteristics, but it does make them more fragile, and that’s where our process comes into play,” he says.
Streamlining the supply chain
Terecircuits has applied for funding through the U.S. CHIPS Act, which is intended not necessarily to onshore semiconductor manufacturing, but to address the vulnerabilities in the supply chain, partly through strategic partnering.
For example, European countries excel in production tools, Japan is big in process chemicals, and packaging and assembly are primarily done in China and southeast Asia. Terecircuits believes that it can maximize the advantages that an Asian country might have in terms of labor contribution by taking a labor intensive serial assembly process and turning it into a parallel process.
Custom ICs for everyone
Terecircuits isn’t a manufacturer. The company’s business model is to sell its IP to a new class of foundries that can assemble chiplets and other components from a variety of manufacturers in order to build custom semiconductors for a wide range of products. In a way, they intend to create a new class of ASIC—one that’s affordable to niche markets as well as high-volume applications. In the past, building an ASIC was cost-prohibitive to anyone but the largest manufacturers. Soon, building an ASIC will be akin to assembling a PCB from off-the-shelf components.
“It’s an engineering toolkit that creates all kinds of possibilities,” Rickard says. “Because now you’ll be able to combine, for example, a GPU from Nvidia, memory from Samsung, and a cool new sensor from this little startup company, and build something unique.”
He concluded, “It’s a great time to be in the semiconductor industry.”
But you knew that already, didn’t you?
