5G kills! What processes and technologies are needed for the RF front end?

The market for RF devices and process technologies is heating up, especially for the two key components used in smartphones – RF switching devices and antenna tuners.

RF device manufacturers and their foundry partners continue to introduce traditional RF switch chips and tuners based on RF SOI process technology for today's 4G wireless networks. Recently, GlobalFoundries introduced the 45nm RF SOI process for future 5G networks . RF SOI is an RF version of silicon-on-insulator (SOI) technology that utilizes a high-resistivity substrate with built-in isolation.

To break the market environment, Cavendish Kinetics, a fabless IC design company, is introducing a new generation of RF products and antenna tuners based on an alternative technology, RF MEMS.

RF switches and tuners are one of the key components in the RF front-end module of mobile phones. The RF front end integrates the transmit and receive functions into the system, and the RF switch routes the signals. The tuner helps the antenna adjust to any frequency band.

Regardless of the type of device and technology, the challenges of today's RF market are daunting. Paul Dal Santo, President and CEO of Cavendish Kinetics, said: "RF was a fairly simple design a few years ago. But things have changed dramatically now. First, the RF front end must be able to handle a very wide frequency range, from 600 MHz to 3 GHz. Using advanced signal technology and 5G technology, the frequency range will reach 5GHz to 60GHz. This brings incredible challenges to front-end RF designers."

With these challenges in mind, handset OEMs must consider choosing new components. Specifically, for RF switches and antenna tuners, there are two technologies that can be attributed to RF SOI-based devices and RF MEMS.

RF SOI is prior art. RF SOI-based device capabilities are acceptable, but they are beginning to encounter some technical problems. In addition, there are price pressures in the market, and problems will arise as devices move from 200mm to 300mm fabs.

In contrast, RF MEMS has some interesting features and has made progress in some areas. In fact, Cavendish Kinetics said its RF MEMS-based antenna tuner is being accepted by Samsung and other OEMs.

Strategy Analytics analyst Chris Taylor said: "Contact RF MEMS provide very low on-resistance, which reduces insertion loss. But RF MEMS has no production record, high-capacity wireless system OEMs will not be new to technology and small suppliers. Make a huge contribution. Of course, RF MEMS as a substitute, the price must be competitive, but major OEMs want reliable products and reliable sources of supply."

RF front end

However, in a mixed-business environment for smartphones (RF switches, tuners, and other components), the RF front-end market is worthy of attention. According to Pacific Crest Securities, smartphone shipments are expected to increase by 1% in 2017, compared to 1.3% in 2016.

On the other hand, according to YoleDéveloppement, the RF front-end module/component market for mobile phones is expected to jump from $10.1 billion in 2016 to $22.7 billion in 2022. According to Strategy Analytics, the RF switching device market reached $1.7 billion in 2016.

As OEMs continue to add more RF content to their smartphones, the RF market is growing. Strategy Analytics' Taylor said: "Multi-band LTE is also extending to the lower layers, and the switch content is increasing."

In the transition to 4G or Long Term Evolution (LTE), the number of RF switching devices per handset has increased. Taylor said: "We talk about a large number of units every year, most but not all (RF switches) will enter the phone, most of which is now SOI. RF MEMS is still an emerging market, insignificant compared to RF SOI switches. ”

Despite the large number of shipments of RF switches, the market is highly competitive and price pressures are high. According to Taylor, the average selling price (ASP) for these devices is 10-20 cents.

At the same time, in a simple system, the RF front end consists of multiple components—power amplifiers, low noise amplifiers (LNAs), filters, and RF switches.

Figure 1: Simple front-end module. (Source: Globalfoundries, "Designing Next-Generation Cellular and Wi-Fi Switches Using RF SOI", May 2016)

Randy Wolf, a GlobalFoundries technician, said in a recent speech: "The main purpose of a power amplifier is to ensure that there is enough "power" to get your signal or information to the destination."

The LNA amplifies the small signal from the antenna. An RF switch routes signals from one component to another. Wolf said: "(Filter) prevents all unwanted signals from entering."

In mobile phones, the RF functions of 2G and 3G wireless networks are simple. 2G has four frequency bands, and 3G has five frequency bands. But for 4G, there are more than 40 frequency bands. 4G not only combines the 2G and 3G bands, but also carries the 4G band.

In addition, mobile operators have deployed a technology called carrier aggregation. Carrier aggregation combines multiple channels or component carriers into one big data pipeline, enabling greater bandwidth and faster data rates in wireless networks.

To handle frequency band and carrier aggregation, OEMs need complex RF front-end modules. Today's modules can integrate two or more multimode, multiband power amplifiers, as well as multiple switches and filters. Abhiroop Dutta, Qorvo's mobile strategic marketing manager, said: “It depends on the RF architecture used. The number of PAs is determined by the regional band that the mobile phone is addressing. A typical SUV global response to global multi-regional or global cellular markets through a single SKU. "Mobile phone, the frequency band coverage is wide. For the implementation of the typical RF front-end integrated module of this mobile phone, one option is to use the RF front-end with a sub-band module to solve the different requirements of high, medium and low frequency bands."

In contrast, smartphone OEMs may design regional handsets for specific markets. Dutta said: "An example is a mobile phone for the Chinese domestic market. In this case, the RF front-end needs to support the frequency band in the region."

Figure 2: 4G Front End (Source: GlobalFoundries, “Designing Next Generation Cellular and Wi-Fi Switches Using RF SOI”, May 2016)

According to Cavendish Kinetics theory, LTE handsets also have two antennas, a main set antenna and a diversity antenna. The main set antenna is used for transmitting and receiving functions, and the diversity antenna increases the downlink data rate of the mobile phone.

In actual operation, the signal reaches the main set antenna. It then moves to the antenna tuner, allowing the system to adjust to any frequency band.

The signal then enters a series of RF switches. GlobalFoundries' Wolf said: "It switches to the applicable frequency band you want to use, GSM, 3G, or 4G." The signal enters the filter from here, followed by the power amplifier, and finally to the receiver.

Given this complexity, handset OEMs face some challenges. Power consumption and size are critical. "Because of this complexity, your signal suffers more loss in the front end, which has a negative impact on the overall noise figure of your receiver," Wolf said.

Clearly, RF switches play a key role in solving this problem. In general, smartphones may contain more than a dozen RF switching devices. The basic RF switch is available in a single pole single throw (SPST) configuration. This is a simple on-off switch.

Today, OEMs use more complex switch configurations. Ron*Coff is a key indicator of RF switches. According to Peregrine Semiconductor's theory: "Ron*Coff is the ratio of losses that occur when a radio signal is in the "on" state (Ron, or on-resistance) through the switch, and the leakage ratio of the radio signal through the capacitor in the "off" state ( Coff, or turn off the capacitor)"

All in all, OEMs need RF switches with no insertion loss and good isolation. Insertion loss involves loss of signal power. If the switch is not well isolated, the system may experience interference. Qorvo's Dutta said: "Overall, the RF front-end challenge is to support the growing performance demands, which are consistent with evolving standards and increased band coverage. Also consider reducing the size of RF device packages because Mobile phones are thinning. Key indicators such as insertion loss, antenna power, and isolation continue to drive the growing drive for RF product solutions."

solution

Today, mobile phone power amplifiers mainly use gallium arsenide (GaAs) technology. A few years ago, OEMs migrated from GaAs and sapphire (SoS) to RF SOI for RF switches. GaAs and SoS are a variant of SOI and they become too expensive.

RF SOI is different from fully depleted SOI (FD-SOI) and is suitable for digital applications. Similar to FD-SOI, the RF SOI has a thin insulating layer in the substrate, which enables high breakdown voltage and low leakage current.

Peter Rabbeni, head of RF business at GlobalFoundries, said: "The mobile market continues to drive RF SOI because it offers low insertion loss, low harmonics, and high linearity over a wide frequency range, with good performance and cost effectiveness."

Today, companies such as Qorvo, Peregrine, and Skyworks offer RF SOI-based RF switches. Typically, RF switch manufacturers use foundries to manufacture these products. GlobalFoundries, STMicroelectronics, TowerJazz, and UMC are the leaders in the RF SOI foundry business.

As a result, OEMs have multiple options for component suppliers and foundry products. Typically, foundries offer RF SOI processes, from 180nm to 45nm nodes, and different silicon sizes.

Deciding which node to use depends on the specific application. Walter Ng, vice president of business management at UMC, said: "With regard to the expansion of RF SOI technology, everything is considered in terms of technical performance, cost and power, making the solution suitable for end applications."

Even with options, RF switch manufacturers face some challenges. The RF switch itself contains a field effect transistor (FET). Like most devices, FETs are affected by unwanted channel resistance and capacitance.

In the RF switch, the FETs are stacked. Typically, 10-14 FETs are stacked in today's RF switches. According to experts, as the number of stacked FETs increases, the device may experience insertion loss and resistance.

Another issue is the capacitor. In a 2014 article titled "The Latest Developments and Future Trends in SOI in RF Applications," Skyworks wrote that at least 30% of the unwanted capacitors in RF switches are due to interconnects in the device. The interconnect is a metal layer or micro-wiring scheme, including RF SOI-based switches.

Typically, in 4G handsets, the mainstream flow of RF switches is the 180nm and 130nm nodes of a 200mm fab. Most interconnect layers are based on aluminum, but not all. Aluminum interconnects have been used in the IC industry for many years, at a low price, but with large capacitance.

Therefore, copper is used for some interconnect layers in RF devices. Copper is a better conductor and has a lower resistance than aluminum. Ng said: "The traditional metal stack for 130nm RF CMOS process products includes a combination of low cost aluminum interconnect layers and high performance copper interconnect layers. This is the best solution for balancing cost and performance. RF SOI solution The solution is usually a certain number of aluminum metal layers and one or more copper metal layers."

Typically, copper is used as an ultra-thick metal option on the top layer to improve passive device performance. Ng said: "The thick top layer metal, preferably copper, can improve performance by minimizing resistive losses."

Recently, RF device manufacturers have migrated from 200mm to 300mm fabs, and their process has increased from 130nm to 45nm. Typically, 300mm foundries use only copper interconnects to process wafers.

Switch manufacturers can reduce capacitance by using only copper interconnects. However, 300mm will increase the cost of wafers, which will create some conflicts in the market. On the one hand, OEMs need lower prices in the cost-sensitive mobile phone market. On the other hand, device manufacturers and foundries want to keep profits.

Ng said: "Today, only a small number of RF SOIs are being produced at 300mm. There are many reasons, including the cost and availability of 300mm RF SOI substrates, the infrastructure to support post-silicon processing, and other factors. However, we expect in the future These challenges will be largely resolved during the year, and then most of the high-volume RF SOI applications will migrate to 300mm."

Prior to this, the industry may face 300mm supply and demand issues. Ng said: "We believe that the market will continue to be in short supply until more production migrates to 300mm. So, how fast the capacity is going online, and the demand at the time, the match between the two will be a problem."

Usually, today's RF SOI process is suitable for 4G mobile phones. GlobalFoundries hopes to jump in the 5G battlefield and recently introduced the 45nm RF SOI process for 5G applications. This process utilizes a high resistivity trap-rich SOI substrate.

5G is the follow-up to 4G. Today's LTE networks range from 700 MHz to 3.5 GHz. In contrast, 5G not only coexists with LTE, but will also operate in the millimeter band between 30 GHz and 300 GHz. 5G will make the data transmission rate reach 10Gbps or more, which is 100 times the throughput of LTE. However, it is expected that large-scale deployment of 5G will not occur until 2020.

Anyway, 5G will need a new component class. Global Foundries' Rabbeni said: "(45nm RF SOI) is mainly concentrated in the 5G millimeter wave front end, which integrates PA, LNA, switches, phase shifters, and creates an integrated millimeter wave controllable beamformer for 5G systems."

For 5G, there are other solutions. RF MEMS is a possibility. In another possible solution, TowerJazz and the University of California, San Diego recently demonstrated a 12Gbps 5G phased array chipset. The chipset uses TowerJazz's SiGe BiCMOS technology.

Who will be the winner? Time will tell us the answer. Strategy Analytics' Taylor said "It's not clear whether RF MEMS has the advantage of 5G. For SOI, monolithic integration may win at least 6-GHz bands."

What is RF MEMS?

RF SOI-based switches will continue to dominate, but new technologies, RF MEMS, also have room for existence. Cavalish Kinetics' Dal Santo said: "SOI has made incredible progress over time. The resistance drops and the linearity gets better. But the SOI switch is just a transistor on or off. It doesn't perform very well when turned on. It is not very good when it is closed."

RF MEMS has been moving forward for many years. Today, Cavendish, Menlo Micro, and WiSpry (AAC Technologies) are developing RF MEMS for mobile applications.

RF MEMS are different from sensor-based MEMS (such as gyroscopes and accelerometers). A sensor MEMS is a sensor that converts mechanical energy into electrical signals. In contrast, RF MEMS is a conductive signal.

Initially, Cavendish et al. aimed to use RF MEMS technology in the antenna tuner market, using RF SOI-based switches and other technologies.

Figure 3: Antenna tuner with switch (Source: Cavendish Kinetics)

Dal Santo said: "If the antennas are fixed, it is impossible to support them in the required frequency range. So they need to be adjusted. Now, the main method is to switch, or switch between different fixed capacitors, or switch different fixed Inductors. The problem is that the antennas are high quality factor (Q) devices. You must be careful not to load them, otherwise you will see loss of radiation performance."

In contrast, Cavendish's tuner has 32 different capacitance ranges. Dal Santo said: "They are fully programmable and have a very high quality factor (Q). So their losses are very low. You can use these to adjust the antenna to the frequency range you need to support."

Looking ahead, Cavendish plans to use RF SOI-based devices in the larger RF switch field. Dal Santo said: "If you replace it with a real switch, it must be a MEMS switch. You can see the cumulative benefit of the insertion loss of the receiver or transmitter."

But will RF MEMS devices replace RF SOI-based devices? A company called TowerJazz can provide some insights. TowerJazz offers a traditional RF SOI process and is a foundry of Cavendish's RF MEMS devices.

Marco Racanelli, senior vice president and general manager of RF and high-performance analog business at TowerJazz, said: "RF MEMS and RF SOI are competing for the same application, and there may be some small overlap between the two. In general, the two will In addition, the winner of the most demanding applications is RF MEMS, and RF SOI wins the rest of the application."

Racanelli said: "RF SOI will continue to evolve, so RF SOI is still viable for RF switching applications and some low noise amplifier markets. However, there are some applications, such as SiGe for low noise amplifiers and switches. MEMS replacement technology can provide better linearity and lower losses. RF SOI will continue to serve the expanding market, and other technologies will also evolve."

RF MEMS is gaining a foothold in the antenna tuner market. Time will tell if the technology can break the switch market. Racanelli said: "In the future, RF MEMS can increase the data rate in mobile phones by providing better linearity and lower loss than built-in RF SOI. It is understood that in RF MEMS, metal plates can be "conducted". The state directly contacts and forms a metallic, low-loss linear connection. Higher linearity allows for wider frequency bands and more complex modulation schemes, which can increase the data rate in mobile phones."