Foundries Enable Integrated And 3D RF Chips
UMC has developed a process to create 3D RF chips, while Tower has enabled integrated Wi-Fi devices.
By Mark LaPedus
A pair of foundry vendors have put the processes in place to develop a new class of 3D and highly-integrated RF chips, paving the way towards next-generation systems with more functions and bandwidth.
United Microelectronics Corp. (UMC) has developed a process to create 3D RF chips, while Tower Semiconductor has enabled integrated Wi-Fi wireless devices.
Both foundry vendors have separately found ways to manufacture these newfangled RF chips using a technology called RFSOI. RFSOI, or radio frequency silicon-on-insulator, is a specialized process used to manufacture select RF chips. In the market for decades, RF chips in general are the critical devices that handle the wireless communication functions in smartphones and other products. RFSOI is different than FD-SOI, which is used to process other chip types.
Basically, foundry vendors manufacture chips for other companies in large facilities called fabs. Nonetheless, Tower, along with chip supplier Broadcom, have enabled highly-integrated RF devices for Wi-Fi 7, the latest Wi-Fi wireless technology that is faster than the previous versions.
Then, in a separate development, UMC has developed a 3D IC solution for RFSOI, which addresses several issues in smartphones and other products. In recent times, smartphone vendors have been scrambling to develop new mobile products that support a growing number of frequency bands and wireless protocols. In response, vendors are incorporating more and different RF chips in their systems. Generally, RF chips are placed inside a module, known as an RF front-end module, in the smartphone. Front-end modules incorporate the components that transmit and receive data in phones and other systems.
In UMC’s 3D IC process, select RF chips are stacked and bonded together in a 3D-like fashion inside the front-end module, which reduces space in the system without impacting the RF performance. Basically, a 3D IC, or 3DIC, means different things to different people, but it generally involves stacking different dies on top of each other, creating a three-dimensional chip. Nonetheless, 3DICs are fast becoming an important technology in the market.
Source: UMC
These and other breakthroughs are needed, and for good reason. Many consumers want faster mobile devices and other wireless products with more bandwidth. In addition, autonomous vehicles, vehicle-to-vehicle communications, factories and other segments require faster wireless networks at lower latencies.
These and other segments require new and faster processors at lower power. In addition, there is also a need for better memory with more bandwidth.
Moreover, the industry also requires new RF devices and architectures. This in turn brings in a new set of challenges for the RF design and device community. They not only need to develop new RF devices with more functionality, but they must also find ways to contain the size and cost of their products. New die-stacking and chip integration technologies are steps in the right direction to deal with the complexities of today’s RF architectures, particularly front-end modules.
4G/5G RF front-end module structure. Smartphones have limited space. A front-end module, which resides inside a smartphone, works as an RF receive path for incoming signals. Each module consists of a multitude of tiny RF chips/devices. The module reduces the size of the components on a board. Diagram source: Balteanu F (2024) RF circuit techniques for transition to 5G advanced. International Journal of Microwave and Wireless Technologies, 1–19.
RF chip ecosystem
By some accounts, the RF semiconductor market is projected to grow from $17.4 billion in 2020 to $26.2 billion by 2025. Radio frequency (RF) technology itself, which has been around for ages, is an important innovation that is often taken for granted. RF is the mechanism that enables smartphones and other systems to communicate with each other over wireless networks.
“RF is a term that describes a range of electromagnetic waves with frequencies from 20kHz to 300GHz,” according to Keysight. “These frequencies form part of the electromagnetic spectrum that is utilized for various wireless communication technologies. The RF spectrum is divided into several bands, each with its own specific applications.”
According to Rohde & Schwarz, RF communications are used in three areas: transfer information over distances (TV and radio broadcasts, satellite communications, smartphones, and Wi-Fi); detecting objects (radar, body scanners); and industrial applications (medical products, microwave ovens).
These and other systems require RF chips. In the RF world, there are a multitude of different and complicated RF device types to address a wide range of applications in the market.
In the semiconductor ecosystem, there are a number of companies that develop and sell RF chips under their own brand names. Broadcom, Qualcomm, Qorvo, Skyworks and many others fit in this category. Some companies manufacture their own devices within their own fabs.
Others have their RF devices and other chips manufactured by foundry vendors, such as GlobalFoundries, Samsung, SMIC, Tower, TSMC, UMC and others. Each foundry vendor offers various process technologies, which are basically the recipes to manufacture a given chip line in a fab. For example, many foundries offer various logic processes, which are used to make AI chips and processors.
Many foundries also offer specialty processes, such as analog, mixed-signal and RF. The most advanced logic devices are manufactured using leading-edge processes. In contrast, analog, mixed-signal and RF devices are made using more mature processes.
2G to 5G
Nonetheless, the wireless communications market has undergone a vast number of changes over the years. In 1973, Motorola demonstrated the first mobile phone, a bulky system that weighed 4.4 pounds.
In 2007, Apple introduced its first iPhone, which helped fuel the growth for cell phones. The first iPhone supported 2G, the second-generation wireless network standard. The system supported the older GSM/GPRS/EDGE 2G wireless standards as well as Wi-Fi. 2G enabled transfer speeds of 40 kilobits-per-second (Kbps).
Fast forward. In 2019, Samsung introduced the first 5G phone. Apple and others soon followed. 5G, the 5th generation mobile network standard, is faster than the previous version, dubbed 4G. 5G delivers up to 20 gigabits-per-second (Gbps) peak data rates and over 100 megabits-per-second (Mbps) average data rates. 5G is designed to support more traffic capacity than previous technologies.
5G smartphones are complex systems that incorporate a multitude of components, such as an application processor, camera modules, memory, modem, RF devices and others. For example, Apple’s latest iPhone 16 incorporates the new A18 application processor. Built using TSMC’s 3nm process, Apple’s A18 integrates a 6-core processor with a 16-core Neural Engine for use in running large generative AI models. For years, Apple has designed its own applications processors in-house. Then, Taiwan foundry giant TSMC manufactures Apple’s processors within its own fabs in Taiwan.
Nonetheless, there are different ways to explain 5G, but many split the frequency ranges for 5G networks into two sets:
*Frequency range 1 (FR1). The sub-6GHz range, covering 410MHz to 6GHz. FR1 carries the bulk of today’s voice and data traffic.
*Frequency range 2 (FR2). The range that covers 24.25GHz to 71GHz. FR2 involves the millimeter-wave (mmWave) spectrum.
In a recent paper, Florinel Balteanu, a technical director at Skyworks, indicated that there are other standards in the works, including:
*Frequency range 3 (FR3). Includes bands from 10GHZ to 20GHz, which are in discussions for future 5G deployment.
*Frequency range 4 (FR4). Covering 52.6GHz to 71GHz, FR4 is expected to be used in autonomous vehicles and automotive radar.
*Frequency range 5 (FR5). Covering 95GHz to 325GHz, FR5 will be part of the next-generation standard--6G.
New and future systems based on 5G and possibly 6G are expected to be faster with more bandwidth. But smartphone vendors and other system houses face a multitude of challenges to develop these new and future products.
For example, Apple’s next-generation application processors are expected to be built using TSMC’s upcoming 2nm process. This process supposedly enables faster chips. But the chips incorporate new and complicated transistor structures, which are difficult and expensive to manufacture.
The RF community also faces challenges. In just one example, mmWave technology, which has been out in the market for some time, can deliver speeds 16x faster than standard 5G at the sub-6GHz range.
But mmWave technology has a relatively short range. Plus, the signals can be blocked by cars, buildings, trees or any object. “Just several smartphones carry mmWave modules due to size, power consumption, higher RF propagation loss, and extra cost,” Balteanu said in the paper.
That’s not the only challenge. To deal with the current and future frequency bands, a system will require more RF content (RF chips/devices), driving up the complexity and cost of a system. Years ago, cell phones were simple devices with little RF content. Generally, in today’s smartphones, the processor and memory devices represent 55% of the total bill of material (BoM) costs within the system, according to Yole Group. The analog RF devices represent 30% of the BoM cost, according to Yole.
Making RF chips and modules
As stated, 5G smartphones are complex systems that incorporate a multitude of components. Generally, a smartphone incorporates several RF device types, including filters, low-noise amplifiers (LNAs), power amplifiers, RF switch chips, tuners and others.
These devices are typically assembled inside an RF front-end module, which is then placed on a board in the phone. Each RF device type performs a different function. And each type is manufactured using different processes and materials.
For example, the power amplifier, the workhorse RF device in systems, amplifies a signal in a phone or other product. Generally, to make a power amp in a fab, a semiconductor manufacture first obtains a substrate based on a III-IV compound material called gallium arsenide (GaAs). Then, the device maker takes the GaAs substrate and manufactures an RF device on the substrate in the fab.
In this case, it’s a GaAs-based power amp. Power amps can also be manufactured using other processes and materials, such as gallium nitride (GaN), RFSOI and silicon germanium. It depends on the application.
Antenna tuners, RF switch chips and others are manufactured using an RFSOI process. LNAs are built using RFSOI or other processes. In systems, an LNA amplifies faint signals, while switch chips are devices that control the path of a signal.
In the RFSOI process, a device maker or foundry obtains a specialized RFSOI wafer from a company called Soitec. Then, you take the RFSOI wafer and manufacture an RF device on top of the substrate. RFSOI wafers are used to develop RF devices with better linearity as well as lower losses and crosstalk. "RFSOI substrates developed more than 10 years ago by Soitec is present in 100% of 5G phones and many other wireless devices, enabling outstanding RF front-end performance for 5G and Wi-Fi connectivity," said Jean-Marc Le Meil, executive vice president for the Mobile Communications Division at Soitec.
Smartphones and other products also incorporate components called filters, which help prevent different RF signals from interfering with each other. Filters are manufactured using a different process.
5G complexity
As stated, the various RF devices are manufactured and then assembled into RF front-end modules. Years ago, the modules incorporated a few RF chips. But as cellular technology evolved and became more complex, the modules required more and different RF devices.
In the 3G era starting in the early 2000s, for example, the modules in phones were relatively simple and incorporated a few filters, power amps and switches. The modules become more complex in 4G and 5G. For example, 5G phones not only need to support 5G, but also 4G and the legacy standards. “With the 5G band proliferation, there are several (RF front-end) modules in a mobile device covering more than 50 bands,” Skyworks’ Balteanu said in the paper.
“5G requires more RF bandwidth and therefore an increase in the number of components such as RF switches, acoustic filters, and power amplifiers integrated in few RF front-end modules. Also, there is an increase in the number of RF radio transmitters and receivers operating at the same time,” Balteanu said. “These additions and the requirement for one/two stock keeping units (SKUs), add a lot of pressure to keep a balance between increased functionality and additional cost and size associated.”
In 5G, a smartphone may incorporate several--possibly five to seven--small RF front-end modules, all in the same system. Each module incorporates at least one power amp, 10 to 14 filters, and four to seven RF switches.
5G smartphone RF front-end module architecture. A module, which resides in the smartphone, consists of the components needed to interface between antennas and the digital section (i.e. processor, modem) in the phone. The components work together to ensure signal integrity. Diagram source: Balteanu F (2024) RF circuit techniques for transition to 5G advanced. International Journal of Microwave and Wireless Technologies, 1–19
In total, a 5G phone could incorporate more than 50 filters and 30 switches. That’s just for the front-end modules alone. That doesn’t include the antennas, which reside outside the module and near the edge of the system. Today’s mobile devices may have six to nine separate antennas, which support the 5G frequency bands, as well as Bluetooth, GPS and Wi-Fi.
Fortunately, there are several new solutions to address the complexity and growing number of RF devices in systems. In the paper, Skyworks’ Balteanu proposes a new RF architecture. It involves an envelope-controlled power amplifier principle, along with a novel simplified calibration architecture designed for 5G/5G+ operating under 6GHz, as well as for FR2-range mmWave power amps.
UMC, meanwhile, has devised a 3D IC solution for RFSOI. Basically, the technology involves taking RFSOI-based RF devices and stacking them. Stacking RF dies isn’t new. Over the years, companies have been stacking RF devices in one form or another.
UMC has apparently put a new twist on the technology. It has implemented an advanced wafer-to-wafer bonding process. Generally, in this process, the first step is to manufacture RFSOI-based RF chips on one wafer. Then, the process is repeated on a second wafer. The two wafers are aligned and then bonded using a wafer-to-wafer bonding process. Then, the dies on a wafer are diced, resulting in separate stacked chips. That’s a simple way to describe a complex process.
Utilizing wafer-to-wafer bonding technology, UMC’s 3D IC solution for RFSOI resolves the common issue of RF interference between stacked dies. Source: UMC
By stacking dies, UMC is able to reduce the overall die size and space in the front-end module. The process is available on UMC’s 55nm RFSOI platform. “Typically, there are separate tuner modules, LNA+switch modules and PA modules,” said Raj Verma, associate vice president of technology development at UMC, in an e-mail exchange. “And in each case, we can reduce the size by more than 45% with minimal RF interference with our 3D IC solution.”
The antenna is a different story. “The antenna may reside on the same PCB board, but it won't be between the two stacked wafers as UMC does not make antennas,” Verma explained. “The antenna may also reside on top of stacked dies and that depends upon the dielectric thickness between antenna and the stacked dies. It also depends on how the end customer wants the whole module assembled and capabilities of the assembly house.”
Meanwhile, partnering with Broadcom, Tower has enabled integrated Wi-Fi 7 front-end module devices on a single RFSOI die. This process reduces chip area despite the complexity of having to support new features and frequency bands.
“A fully integrated Wi-Fi front-end module on a chip consists of a power amplifier for the transmit function, a low-noise amplifier for the receive function and a high-frequency switch device to switch between those two amplifiers and the antenna,” said Ed Preisler, vice president and general manager of Tower’s RF Business Unit, in an e-mail exchange. “What sets RFSOI apart is the quality of the RF switch - no other technology can achieve the same performance for the switch as RFSOI. The challenge with RFSOI compared to other processes has historically been the performance of the power amplifier, but with Tower’s new technology, these limitations have been solved.”
Like 5G, Wi-Fi 7 is important. Wi-Fi 7 doubles the bandwidth compared to the previous versions--Wi-Fi 6 and 6E. The new version also enables up to 2.4 times more throughput and reduced latency. Wi-Fi 7 also extends the 6GHz transmit range in both indoor and outdoor environments by using optimal spectrum allocation.
What’s next? On the product front, Huawei recently unveiled the world’s first tri-fold phone. The phone is an expensive but intriguing product. Autonomous cars are being deployed.
On the technology front, 6G is in the works. Perhaps a next-generation Wi-Fi standard is also being debated. Clearly, the industry needs a way to improve mmWave. These and other technologies should keep the RF community busy for the foreseeable future.