Junko Yoshida ponders who's who and what's what in RF
LYON, France — While the mobile industry is done with its annual Mobile World lovefest — held last month in Barcelona — tech suppliers, system OEMs and mobile operators now face a host of 5G obstacles not yet overcome. In fact, they’re just getting started.
The technical issues of 5G are manifold. Among them, smart antennas and RF front-ends for 5G mmWave — typically expected to operate at frequencies such as 28 GHz, 39 GHz or 60 GHz — could seriously affect the performance of yet-to-emerge 5G mmWave mobile phones.
After returning from the Mobile World Congress, Claire Troadec, activity leader for RF electronics at Yole Développement, told us, “Although many companies such as Qualcomm, Intel, MediaTek and Samsung are using a mobile phone as a 5G mmWave demonstrator platform, we don’t believe handsets will be the first place where 5G mmWave will go.” Rather, 5G mmWave will be a stationary data modem sitting on a table or desk so that consumers can download or stream massive broadband applications, she suspected.
Given that 5G’s mmWave frequency bands are notorious for high propagation loss, directivity, and sensitivity to blockage, it’s no small feat to design a 5G handset that works all the time without losing signals. Picture consumers might well be forced to stay — literally — on their toes, turning and pacing in search of a signal.
Another challenge in deploying 5G mmWave radio in mobile handsets is battery life and death. During the PyeongChang 2018 Winter Olympics, Samsung is believed to have demonstrated its own 5G tablet. Although it worked fine, the word swirling around Mobile World Congress was a whopping caveat: the battery was toast after 30 minutes.
Asked about the rumor, Yole’s Troadec said she believes that “5G mmWave radio for cellphone has high issues related to power consumption.” She suspects that “most of the leaders are extensively looking into this area.” But she added that she has found little evidence of the remedies these technology vendors might have conceived for the obviously problematic, system-level power consumption issue associated with the 5G New Radio. Nobody was willing to discuss this further at the show, she said.
Disruptions 5G mmWave RF modules will bring to the nascent 5G market aren’t limited to changes in technologies. Deeply affected will be who’s who in the current supply chain of RF components and modules designed for 3G and 4G.
Since 5G mmWave potentially allows vendors to design front-end modules in SoC by using CMOS or SOI technology, the door will open to “advanced CMOS design and manufacturing players” currently in the cellphone architecture ecosystem, to move into the RF market, Yole explained. Candidates for this move include Samsung, Huawei and Mediatek, in addition to Intel and Qualcomm, the company added.
More bands, more RF Front Ends (RFFEs)
The mobile industry has come a long way, as technology suppliers have wrestled with complex RFFE modules capable of handling an ever-increasing number of frequency bands. As the cellular standard progressed from 3G to 4G, the bands with which the RF front end had to cope increased from four to 30, according to Troadec.
With 5G coming online, though, things are going to get even more complicated. Although 5G is, in theory, a single standard, it comes with three elements: 5G for IoT, 5G using sub-GHz, and 5G on mmWave. In terms of RF technologies, “this would mean bringing together devices that require extremely dissimilar performance,” Troadec observed.
This implies that 5G will follow “different implementation stages, and different flavors of 5G are in development in parallel.” In other words, there won’t be a single, unified 5G RFFE, but rather, “5G IoT, 5G sub-6GHz, and 5G mmWave will follow their own paths and create parallel ecosystems with their own RF SiP developments,” she said.
Asked to assess a different RFFE path for each 5G flavor, Troadec said she sees mmWave 5G technology bringing the most disruptive innovations. She expects heavy design changes and new materials to be required.
The good news is that 5G mmWave could practically end the current practice of complex system-in-package (SiP)-based front-end modules used for 2G, 3G and 4G RF. “You can design every building block — including power amplifiers, low noise amplifiers, filtering, switching and passives — based on advanced CMOS or SOI technology,” Troadec explained. This will give an opportunity to many digital chip vendors who previously had little radio expertise to develop front-end modules in SoC.
Meanwhile, as for 5G sub-6 GHz, Troadec believes that it will be built on incremental innovation. Expected are modification of current RF packaging architectures with minimal change in the bill of materials, she explained.
Since 5G IoT will use frequencies below 1 GHz, Troadec sees “little to no innovation” necessary in semiconductor packaging for RFFE. Nonetheless, the 5G IoT spec and protocols, designed to address transfer of data generated by many IoT devices, are yet to be defined and standardized.
Who’s who in today’s RF supply chain
Before diving into detailed RF solutions for 5G, let’s take a closer look at the current RF component and module suppliers.
Typically, RF front-end modules consist of such RF components as RF switches, power amplifiers (PA) /low noise amplifiers (LNA), RF filters and antenna devices (tuners and switches).
Key players in the crowded RF supply chain include: Sony, Murata (which acquired Peregrin Semiconductor late 2014), Skyworks, Qorvo, Infineon, Broadcom/Avago, Cavendish Kinetics, TDK EPCOS and more.
Each company has its own specialty RF components which often deploy varied substrates and process technologies. Their choices range from RF-SOI and BiCMOS to bulk CMOS, GaN and RF MEMS.
Because of the diverse process technologies used in different types of RF components, the path left for RF integration today is SiP, not SoC.
Today, for the frequency bands in the 2G, 3G, 4G and 5G Sub-6 GHz landscape (for all the bands below the 6GHz regime), “The only way to fulfill the stringent requirements of the radio in a smartphone is to have a SiP approach,” Troadec confirmed.
There is currently no single RF component supplier who has the best of everything. Troadec explained that in RF front-end integration, “One needs very dedicated technology for each building block: Best PA using GaAs technology, best switches using SOI technology, best filters using SAW and BAW technology, best LNA using SiGe technology.”
Asked who offers SiPs for front-end modules, Troadec said, “Broadcom, Murata, Qorvo, Skyworks, TDK/Qualcomm are the only players to provide SiP today.”
She explained that each has its own specifics such as high band Module, mid band modules, low band modules, diversity receive modules, in the form of PAMiD (highly integrated, customized modules which are performance-driven but which come with a strong cost penalty, thus only limited to players such as Apple, Samsung and Huawei to a certain extend), or FEMiD (providing a good compromise in terms of performance and cost and favored by Tiers 2 Smartphone makers such as LG, Chinese players…).
“We do see that only few companies can play in this high technology mixed environment,” she concluded.
5G Sub-GHz: Still SiP approach
As the cellular industry moves to 5G, the same principle — a SiP approach — will remain for the front-end modules for 5G Sub-GHz.
There will be some changes, however, in terms of more integrations inside SiP and packaging, according to Yole. Troadec explained that these include integration of LNA and switches on the same die based on a SOI platform, and more wafer level packaging for filters to gain in die size (today only Broadcom has such an approach, while Qorvo is working on it, for example). Also, a wafer-level approach will apply to packaging for PA (still wire-bonded today) to gain in die size.
5G mmWave: From SiP to SoC
5G mmWave front-end modules will undoubtedly alter the most intricate RF component/module supply chain. Gone are a host of complex RF components manufactured by using different process technologies. Instead, emerging on the horizon is the possibility of a mmWave front-end module in SoC, based on advanced CMOS or SOI technology.
There are many reasons why 5G mmWave makes it possible to design RF modules in SoC.
First, 5G mmWave means moving to an area of the spectrum where bandwidth is available, explained Troadec. “Thus we do not need many frequency bands to send the information. The architecture of the radio can be much simpler.”
As a result, this lowers constraints on filtering technology, she explained. “We do not need high-end filtering anymore in the module.” Troadec, however, cautioned, “We will need high-end switches (isolation, linearity) in between the different radios though (4G or 5G Sub-6GHz and 5G mmWave).”
She also pointed out that in 4G, “we use carrier aggregation with 20 MHz bandwidth per frequency band and we use many bands. Thus, we need a high-end filtering technology (with a steep curve) to distinguish each signal at each frequency band. This is only available with BAW devices (MEMS technology) today.”
Another big factor is that 5G mmWave will come with a beam-forming technology, allowing it to shape the beam to send the information to many users simultaneously. “This will lower the constraint on the power emission of the PA. This in turn means that CMOS technology can play a role.” She added, “At mmW frequency, the inductance becomes smaller, thus it becomes possible to integrate passive components with CMOS/SOI technology.”
Trodeac reiterated, however, that a limiting factor [for a 5G mmWave RF module] seems to be power consumption by the full system. “Why is that is something we need to clarify, but up to now nobody wants to tell us technically why and what needs to be done” to solve the issue.
New players coming into RF
Once the industry shifts to designing RF front-end modules for 5G mmWave in SoC by using CMOS or SOI technology, the current RF landscape will change from a seemingly cozy club of RF front-end module vendors such as Broadcom, Murata, Qorvo, Skyworks and TDK/Qualcomm.
Trodeac noted that Intel and Qualcomm are already into the modem and transceiver business for cellphones. They would very much like to master radio as well, to deliver an end-to-end solution. The goal is “A full in-house design from A to Z for the RF chain,” she said.
Had Broadcom acquired Qualcomm…
Of various product and technology segments both Broadcom and Qualcomm are in, the cellphone market is one where the two giants have very complementary businesses. Troadec observed that Broadcom is highly positioned in radio, and WiFi, Qualcomm in application processor units, modems, transceivers, WiFi/BT, NFC with NXP and their micro-controllers.
Now that Qualcomm is getting traction in the 5G mmWave space while Broadcom is focused on the Sub-6GHz regime, Troadec said their merger, had it not been blocked by the U.S. president, “would have created a very high monopoly.” She suspects, “This is why we saw Intel getting scared and trying to enter in this discussion as well.”
— Junko Yoshida, Chief International Correspondent, EE Times