Inside The 5G Smartphone

Amid a slowdown in the cell phone business, the market is heating up for perhaps the next big thing in wireless—5th generation mobile networks or 5G.

In fact, major carriers, chipmakers and telecom equipment vendors are all rushing to get a piece of the action in 5G, which is the follow-on to the current wireless standard known as 4G or long-term evolution (LTE). Intel, Samsung and Qualcomm are among the players in the 5G sweepstakes.

Several organizations are still hammering out the standards for 5G, although preliminary trials and demonstrations have already started. Hoping to satisfy the insatiable thirst for more bandwidth, 5G will enable data transmission rates of more than 10Gbps, or 100 times the throughput of LTE. Compared to 4G, 5G is expected to provide 1,000 times more capacity and one-tenth the latency.

But questions are beginning to surface about 5G. The actual 5G rollout is projected to commence in 2020, but the deployment could take much longer. Interoperability and other technical issues could arise. And it will take billions of dollars of investments to deploy 5G networks worldwide.

In fact, 5G could get pushed out by new and faster versions of LTE, dubbed LTE-Advanced (LTE-A). “Although the year 2020 seems to be the favorite timing for 5G to commence, I see it as a date of first experimentation, not rollout,” said Will Strauss, president of Forward Concepts, a research firm. “Between now and then, LTE-A with carrier aggregation will fulfill the growing market need for higher data speeds. By 2020, Cat 10 LTE-A will likely be the global norm for premium networks. We probably don’t need 5G in 2020. Nokia is talking about 5G in 2030, which is probably more realistic.”

But if 5G does happen on time or shortly thereafter, the industry will require new chip and system breakthroughs. Operating at frequencies of 6-GHz and above, a futuristic 5G smartphone alone will need faster chips at lower power.

“The expectations for 5G are fairly aggressive,” said Aaron Thean, vice president of process technologies and director of the logic devices R&D program at IMEC. “It will require a 10X lower energy consumption at a 100X peak data rate, and a significantly reduced radio-link latency.”

So what will futuristic 5G smartphones look like? And what types of chips will they require? It’s difficult to predict the future, but Semiconductor Engineering asked experts to list some of the future requirements for a hypothetical 5G smartphone, including the application processor, memory and RF front-end.

Why 5G?
So what exactly is 5G and why do we need it? Today, many carriers have deployed Category 4 (CAT 4) LTE-A mobile networks, which enable a data downlink speed of up to 150Mbps and an uplink speed of 50Mbps. In the United States, Sprint has deployed CAT 6 LTE-A, which enables a data downlink speed of 300Mbps and an uplink speed of 50Mbps. Then, in the future, carriers are expected to deploy CAT 10 LTE-A, which supports data downlink speeds of up to 450Mbps and uplink speeds of 100Mbps.

At some point, though, 4G may hit the wall, prompting the need for a faster wireless technology like 5G. “Video is becoming such a dominant component in Internet traffic,” said Simon Segars, chief executive at ARM. “And besides that, you will scale up to 4K or 8K displays. That is something that will need a lot of bandwidth.”

Others agree. “The expectations for 5G communications will continue to grow based on the rising demands for smart devices, cloud services, smart home technology, and Internet of Things,” said Chang Yeong Kim, head of the DMC R&D Center at Samsung, at a recent event.

Still, there will be a number of challenges to deploy 5G networks. “We consider 5G to be a transformation of how networks are constructed and how radio resources are used,” Kim said. “To support 100 times greater throughputs at a fraction of the latency, we need to consider more than just a single network component. We need to look at how everything works together. At the same time, the evolution toward 5G must be an incremental process, introducing new technologies in the short- and mid-term that can be tried, tested and proven on commercial networks.”

In fact, 5G will require several new and complex pieces. For example, Japan’s NTT DoCoMo hopes to launch the world’s first 5G wireless network in time for the 2020 Olympics in Japan.

For its 5G network, NTT will deploy a centralized radio-access-network (C-RAN) architecture. In this technology, a central data center is connected to multiple sites. RF signals are sent over lines between remote radio equipment and the central center.

NTT’s C-RAN technology handles the processing for carrier aggregation within a centralized baseband unit, thereby reducing the load on the core network. In addition, 5G will require a massive MIMO, which is a method for multiplying the capacity of a radio link using multiple antennas.

That’s just the tip of the iceberg. “The 5G network topology will change quite a bit in the coming years,” ARM’s Segars said. “It’s about low latency and many data connections. The processing moves into the network, as opposed to being just in the cloud. You will see more bay stations, small cells and macro cells.”

Inside a 5G phone
To power a 5G network, the industry will require a new class of chips. A hypothetical 5G smartphone, for example, will need to process a staggering amount of data. So, a 5G phone will likely require souped-up chips for the digital portion, which includes the applications processor and a baseband.

“There are already application processors announced with 10 processor cores. We’ll keep adding more to the point of diminishing returns,” Forward Concepts’ Strauss said. “10nm application processors are coming by 2020, but there will be analog front ends that could never get to such small geometries anytime soon.”

The modem, according to Strauss, is the biggest issue. “The 5G modems to ultimately feed such monster application processors will be the big challenge,” he said.

For 5G phones, OEMs will likely incorporate application processors and other digital chips based on finFET transistors. “I believe (finFETs) will be relied upon for the data processing part of 5G,” Imec’s Thean said. “We also see that finFETs, in their current form, may not be the best for all aspects of RF applications due to their parasitics. This is currently being addressed by innovating the spacer and source/drain structures targeting 7nm. This will be in line with the 5G deployment timeframe.”

There are other issues. “At the system level, I believe that a heterogeneous mix of technologies will still be the case to optimize between performance, functionality and cost,” Thean added.

Memory
Meanwhile, for mobile systems, OEMs currently are using LPDDR3- and LPDDR4-based mobile DRAMs, which are specialized, low-power versions of PC DRAM.

Some hope to extend planar-based mobile DRAMs and are developing an evolutionary scheme called LPDDR5, which should be ready by 2017 or 2018.

“We envision most smartphones to be using LPDDR4 technology primarily by the year 2020,” said Ken Steck, senior product marketing manager at Micron. “LPDDR5 and other advanced memory technologies will most likely be used by the high-end smartphones, but may not be as ubiquitous. Memory densities by this time are expected to be approximately 2X of today’s phones.”

OEMs could also move towards a 3D DRAM technology called Wide I/O-2. Another possibility is the next-generation memory types, such as  and ReRAM.

RF front-end
Radio-frequency (RF) technology is another critical part of the wireless infrastructure. Today’s 4G networks operate from 700 MHz to 3.5 GHz. In contrast, 5G schemes will need to operate at the unlicensed or mmWave bands, which provide 10 times more bandwidth than 4G networks. 5G may operate anywhere from 6 to 60 GHz, or higher.

In today’s 4G cell phones, the RF front-end must support more than 40 frequency bands, 3 carrier aggregation bands and an 8 x 8 MIMO. In comparison, the RF front-end in a 5G smartphone needs to support 5 carrier aggregation bands, 50 frequency bands and a massive 64 x 8 MIMO, according to RF chipmaker Skyworks Solutions.

“We are talking about lots of bands,” said Aniruddha Joshi, senior technology director of wafer foundry engineering at Skyworks. “So we need better linearity so that you don’t distort the signal. We want devices with a lower Ron-Coff. We want better efficiently. And we don’t want to waste any power.”

The shift toward 5G will require changes in the RF front-end of a cell phone. The classic RF front-end includes three main parts: the power amplifier, antenna switch and filter. Today’s RF front-ends are moving towards multi-mode, multi-band power amps. Typically, the power amp is based on gallium arsenide (GaAs) heterojunction bipolar transistor (HBT) technology. The power amp amplifies RF signals in the phone.

“GaAs has not run out of steam yet,” Joshi said. “But since we want to operate at higher frequencies (with 5G), we may have to switch to something different like indium phosphide.”

Another key component, the RF switch, routes signals between the antenna and the handset. Typically, the RF switch is based on RF-SOI. On that front, Soitec is ramping up a next-generation RF-SOI technology, dubbed Enhanced Signal Integrity (eSI) SOI. Based on a high-resistivity substrate, eSI also includes a trap-rich layer to reduce loss and improve linearity, according to Bernard Aspar, senior vice president and general manager of Soitec’s Communication and Power Business Unit.

“RF-SOI has become mature. More and more foundries are offering it. Prices are coming down,” Skyworks’ Joshi said. “But at some point, we need a big jump in improvement, especially when we go to higher frequencies.”

RF-SOI may do the job for 5G. If RF-SOI runs out of gas, the industry may opt for MEMS technology in RF switching applications.

“5G will likely shift the industry towards higher frequencies. That, in turn, will push all of the figures of merit (FoM) higher and to the right. This march towards improved FoMs will not only require continuous innovation within the SOI technology, but also the invention of disruptive switch technologies,” said Marco Racanelli, senior vice president and general manager of the RF/High Performance Analog and Power Business Group at TowerJazz. “MEMS technology has been shown to provide the lowest insertion loss and highest linearity among various semiconductor technologies available to us today.”

But RF-MEMS face some challenges, particularly cost. “Until RF-MEMS technology reaches that stage of maturity, RF-SOI will likely remain the workhorse of wireless networks for the foreseeable future,” Racanelli said.

Still, the RF switch or other components may not be the biggest challenge for 5G. “In our opinion, the biggest challenge for 5G networks remains interoperability,” Racanelli said. “The users must be able to migrate seamlessly and efficiently across multiple high-speed wireless standards without compromising on the quality of service.”

All told, it will take a monumental effort to make 5G happen. “5G networks will require a far closer collaboration across the entire semiconductor ecosystem than what is present today,” he added.