BY PAULA DOE FROM SEMI
What the next generation of 5G cellular technology will look like depends on who you ask, but one thing is clear ─ the need for extremely high data rates (1-10Gb per second) and very low latency (1 millisecond) are essential while still preserving battery life. That likely means a big growth opportunity for high-speed, high-efficiency compound semiconductor devices, but volume markets will also bring big changes to the specialty business, and potentially more competition from silicon devices.
“People are a little blasé about smart phones currently, but there are real drivers for much higher data rates,” says Skyworks Solutions, Inc. CTO Peter Gammel, who will speak on the Compound Semiconductor program at SEMICON West 2016 on July 12.
Gammel noted that continued improvements in higher resolution displays, higher definition audio, SLR-like camera image quality, and the increased use of haptics, will all demand much higher data rates and much lower latency ─ to say nothing of emerging markets like augmented reality and autonomous cars.
The only way network operators can afford to provide this kind of service with a reasonable number of base stations, he argues, is to go to a cloud-based radio access network. Mobile service will start to look more like cloud computing, with the computing work moving to the cloud for the heavy lifting, while the device in your hand becomes a lower cost hub, which will need a robust, high speed, radio-accessed network. “I’m absolutely confident that the smart phone will become the C-phone — the cloud phone,” says Gammel.
Data rates as high as 1-10Gb/second and latencies as low as 1 millisecond will likely mean more demand for compound semiconductor devices.“The only option for the needed speed and efficiency is compound semiconductors,” he contends, noting also that stacked die packaging in SiPs will also likely be the norm, so that the best technology can be chosen for each component and the increasing analog content contained in the smallest space. “The drive for a full RF system on a chip is being reversed.”
A move to 5G is not a transition to higher data rates. It also involves higher data volumes, lower latency, a longer battery life for low-power devices and more connected devices. Source: Courtesy of Skyworks.
As eventual full 5G data rates reach 10Gb/second, the industry will need to move to millimeter wave transmission, transitioning what are now specialty, military compound semiconductor technologies into volume consumer markets. That means the specialty compound semiconductor world will move to adopt the kind of design tools and simulation used by the volume silicon business, for better model accuracy and simulation speed, and more production will move to specialty and analog foundries, he notes.
Can silicon challenge GaN and GaAs?
“The vision for this future 5G technology is pretty transformative, both for the individual and for society,” says Eric Higham, director, Advanced Semiconductor Applications, at Strategy Analytics, another SEMICON West speaker on compound semiconductors, stressing the potentially millions of IoT devices, virtual reality, autonomous cars, and remote surgery. “It’s potentially a bigger opportunity than the mobile market,” he notes. “But an order-of-magnitude jump in data rates, and in volume of devices, while still keeping power usage low, also presents significant challenges.”
Higham suggests that while the new 5G technologies are without doubt a big opportunity for semiconductor suppliers, it is not yet clear just who the winners will be. The increase in component value in each phone will be balanced by the slowing growth in total sales of smart phones as the market matures. The higher volume opportunity could also attract more investment in development from silicon suppliers for these traditionally compound semiconductor markets.
The smartphone is driving growth in the GaAs market, and could continue to do so with a move to 5G communication.
Each phone will certainly need at least three or more power amplifiers to cover the wide range of frequencies required. So far that’s a GaAs market, with the newer silicon power amps making inroads only in lower frequencies in lower end phones. But the silicon technology is improving, and if the market volume grew large enough to make investment in silicon development worthwhile, silicon could become a lower cost alternative at more frequencies.
Similarly, while compound semiconductors have the edge in efficiency and power usage needed for base stations, if 5G ended up using vastly more base stations, each using much lower power, silicon could potentially compete with GaN there as well. “The volume and opportunity will be huge, but it’s not yet clear who will capture it,” says Higham.
GaN technology is changing fast
"In the short term, GaN’s role in RF power is clear, as it’s replacing LDMOS for base stations. But it’s also starting to make inroads in other areas as well, as the technology is developing with really remarkable speed,” says Notre Dame professor Patrick Fay, who will give a tutorial on GaN technology and applications at SEMICON West.
Fay notes particularly the ongoing improvement in producing higher quality material at more affordable cost, fueled by both advancements from the LED world as well as the emergence of native GaN substrates. Researchers are working on faster growth rates bulk GaN with alternative technologies, and on approaches for removal and re-use of the growth substrate, pushing towards a goal of $0.10/Amp in high-voltage transistors.
“Native GaN substrates may still be small and expensive compared to silicon, but so are the competing SiC ones,” he notes. “And the substrate cost may recede in importance as the new growth technologies mature, and as GaN enables the needed performance at a competitive systems-level cost."
All that demand for high speed, high bandwidth data in telecommunications and data centers will ultimately also drive demand for silicon photonics technologies, but even big demand for closely integrating high speed photonics into the silicon production world may not directly translate into the high volumes of wafers needed to make sense for efficient volume foundry production.
“Demand is booming in terms of transceivers, but that still probably means silicon photonics volume of at most 10,000 300 mm wafers per year in the coming two to three years,” notes Philippe Absil, imec Department Director, PT/3D and Optical Technology, another speaker.
Absil notes that imec’s open platform, with an established PDK for generating first-time right prototypes and engineering samples, uses quite a generic 130nm CMOS-based process, requiring only the addition of more advanced lithography (193 nm “dry” lithography), germanium epitaxy, and MEMS-like deep silicon etching capability, available at many commercial foundries. The process is designed as a modular system, so various modules can be added or removed without adverse impact on the process flow. “Fabless designers would still add their own proprietary device design and circuit solutions, while the unification of the platform will provide enough volume for the foundry,” suggests Absil.
This year’s greatly expanded conference program at SEMICON West 2016, July 12-14 in San Francisco, will address compound semiconductor technology issues in sessions on Compound Semiconductors in Telecommunications and Computing, Power Semiconductors, Power Semiconductor Packaging, Photonics Semiconductor Packaging, the GaN Tutorial, and a Compound Semiconductor Pavillion.