//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>
There is no doubt that 5G’s higher bandwidth, lower latency and higher availability make it well suited for a range of applications. However, the higher-frequency bands, particularly millimeter-wave (mmWave), also bring challenges around delivering these improvements. As a result, power IC manufacturers are looking at more efficient technologies, such as wide-bandgap (WBG) semiconductors like gallium nitride (GaN) and silicon carbide (SiC), to enhance performance and lower costs in these new networks.
What does the mmWave band bring to the table? First, it expands the spectrum for wireless data communications, which helps increase the amount of data that can be transmitted and lowers latency, and secondly, it is ultra-fast. Yet industry players said that to deliver high, 20-Gbits/s data rates of 5G, the mmWave spectrum will need to be rolled out.
However, mmWave has several drawbacks: an ultra-short range and the signals easily being blocked by things like buildings, walls, trees and even adverse weather conditions like rain.
The real benefits of 5G mmWave are high download speeds and low latency, but the downside is poor signal propagation because of the higher frequency, making it easily blocked by windows and walls, and general distance makes it more difficult propagation-wise, said James Edmondson, senior technology analyst at IDTechEx.
While silicon technologies like traditional LDMOS semiconductors still provide high performance at lower frequencies, WBG semiconductors like GaN-based amplifiers that can handle up to the 100-GHz range will be needed for mmWave frequencies. GaN semiconductors already offer benefits in the sub-6-GHz range, leveraging its higher density and efficiency and lower parasitic capacitance. But there are still some developmental challenges that GaN presents before it can be fully adopted into mmWave networks.
5G comes in two flavors: sub-6 GHz that supports mid (about 3.5 GHz to 7 GHz) and low (<1 GHz) frequencies, now being deployed in mobile networks, and the ultra-high–frequency band (mmWave), typically noted as between 24 GHz and 100 GHz, which is still at the very early stage of development.
Most of the 5G deployments today are based on sub-6 GHz, according to IDTechEx, due to its high throughput and cost. Despite its ultra-fast speed, mmWave is still too expensive to deploy and it has disadvantages around line of sight. About 1 million mmWave antennas are currently deployed, according to the market research firm, with a forecast of 50 million by 2033.
What will speed up 5G mmWave deployment?
Although many industry players say it is simply too early for 5G mmWave deployments, mostly because of cost, GaN power devices will play a role in developing more cost-effective networks and achieving the promise of 5G.
One of the roadblocks to fast-tracking 5G mmWave is a lack of market demand. “When 5G was first talked about, there were a lot of these killer potential applications like remote surgeries and the metaverse, but not all of that has taken off yet, so there is a bit of a market pull lacking at the moment,” said Edmondson.
GaN RF power IC suppliers like NXP Semiconductors also cited the lack of demand as a challenge. NXP offers a wide selection of RF power devices, including GaN-on-silicon, silicon LDMOS, silicon germanium (SiGe) and gallium arsenide (GaAs) technologies.
“What we saw is a lot of hype and interest because of these enormous swaths of bandwidth that became available, and there was a lot of initial excitement with everyone rushing to get products out, including NXP,” said Geoff Tucker, director of RF systems engineering in the Radio Power business unit at NXP Semiconductors. “It kind of fizzled a little bit when it came to actual shipments. I think it still has a role in networks, but as of yet, it doesn’t seem like it’s truly found its niche, whether it be for a mobile technology or fixed wireless access.”
However, how this technology evolves, given the large pieces of spectrum, which is the gold of the industry, and where that might be going is of interest to NXP, he added.
Despite the lack of market demand today, power IC manufacturers are working to solve the biggest technology challenges, including higher integration, to deliver high-efficiency power amplifiers, which will be needed to reduce power consumption and shrink form factors in radios and antennas. GaN-based devices are at the top of their list. Although most of them are working with GaN-on-SiC, there is some work under development for GaN-on-silicon for mmWave.
Yole Intelligence reports a lot of activity around GaN-on-silicon. “The company OMMIC is proposing GaN-on-Si beamformers for 5G mmWave, and this solution is expected to be seen penetrating the market in the coming years,” said Cyril Buey, technology and market analyst for RF Devices and Technologies at Yole Intelligence, part of Yole Group. “We also see the startup Finwave developing FinFET GaN-on-Si for mmWave technology, as well as STMicroelectronics and GlobalFoundries working on GaN-on-Si devices, so a lot of activities on GaN-on-Si.”
However, Yole is less optimistic about GaN-on-SiC devices for 5G mmWave. “Qorvo has a portfolio of GaN-on-SiC power amplifiers in mmWave frequencies, but for now, at Yole Intelligence, we don’t see 5G mmWave products using GaN-on-SiC devices,” he added.
Buey thinks there is room for GaN-on-Si for 5G mmWave applications but not for GaN-on SiC, mainly due to the high level of integration needed for mmWave devices. “In the end, GaN-on-SiC might be suitable for backhaul applications, also at mmWave frequencies, where the system architecture is simpler.”
Why GaN for mmWave?
It is well known in the industry that SiC and GaN offer several advantages over silicon power devices, including lower switching and lower conduction losses. SiC also offers reduced thermal management, while GaN delivers higher switching frequencies.
The primary advantage of GaN is its higher power density, which allows for a smaller-form factor and thus a reduction in overall system size, at the same performance. This can benefit mmWave base stations by allowing a signal to be transmitted with more power, translating into a wider coverage area.
LDMOS is good up to 4 GHz, but above that, it starts to become quite inefficient to operate, said Edmondson. “The big challenge for GaN, especially initially, was more toward the cost of the material, so silicon technology is very mature and very cheap. But because of the high power density, you can get away with using a lot less material, so there is somewhat of a tradeoff. Typically, GaN has been more expensive, but also on top of that, there is a general lack of industry experience with the material.”
“There are some physics involved here, but what GaN does is concentrate more of the power into a smaller area to achieve the same goal from the amplifier, and therefore, it is a nice technology to help us densify the overall designs,” said Tucker. “You still have to put all the other analog functionality in there—switches, gain blocks, attenuators and whatever else we decide to throw in a given amplifier—but it does help us to miniaturize things if it can be done successfully.”
The biggest challenges of mmWave around line of sight, range and signal propagation (high losses) will need to be solved. Some of the answers will require changes in the RF circuitry and power amplifiers, although there are technologies like massive MIMO, miniature antenna arrays and smart active repeaters being used to resolve some of these issues.
“When you move to mmWave, the antennas shrink, which increases your power density per device, so the actual number of antenna elements within a device increases a lot,” said Edmondson. “You can be looking at thousands of antenna elements in a package, and what that actually means is that the power demand on each individual amplifier goes down. I think that’s been a big reason why we haven’t seen GaN adopted significantly in mmWave 5G yet. If you can get away with using the existing silicon technology, then that’s probably going to be the easier way to do it.
“There will be more adoption of GaN in the future in mmWave, but it does face some extra challenges like component integration and the fact that there’s actually less power demand on each individual amplifier,” he added.
Ultimately, the choice of technology—silicon- or WBG-based—comes down to the application. “For an antenna design, it’s the specifications of the amplifier that is going to sell it to me,” said Edmondson. But at the same time, if the GaN device requires extra work around mounting it on the board or dealing with thermal management, those are obviously design tradeoffs, he added.
The No. 1 design challenge is the architecture of the radios themselves and what feature sets get put into them because it is still at the very early stage, said Tucker.
First-generation radios are purely linear in nature, and the amplifier final stages don’t take advantage of high-efficiency techniques that are found elsewhere in the industry, Tucker said. “That’s slowly changing, where you start to see higher-efficiency architectures and digital predistortion coming online for these higher-frequency radios, but still, it’s very simplistic. It’s not nearly as mature as well-established communication systems.”
There is a trend toward higher power and fewer transmit paths, which is important when it comes to GaN and what it can actually do with high frequencies, he added.
Tucker also noted cost as a design challenge. “We see a simplistic architecture being employed for these mmWave radios that are purely analog beamforming, which is probably not the preferred way to do it. It’s certainly not what we use at sub-6 GHz.”
The reason is simple economics, Tucker explained. These high-order types of transmitters have a lot of analog functionality with a lot of integration and pairing with digital front ends that are used in modern radios, he added.
GaN-on-SiC, which NXP uses today, will be very competitive and work extremely well at up to 30–40 GHz, but beyond 40 GHz, you will start to lose some of that “superb” efficiency, said Christopher Dragon, director of device engineering in the Radio Power business unit at NXP Semiconductors.
Despite some challenges in the higher-frequency ranges, Dragon believes GaN-on-SiC will still have a play based on research in the industry that is looking to extend the frequency range. One area of research and discussion is N-polar GaN-on-SiC. “I think the research will lead us in a direction that will work,” he said.
Silicon LDMOS, which would rule the industry for base stations for years, started to lose its efficiency at about 2–3 GHz, and that is where there is a transition to GaN, Dragon said. “I see that happening [loss in efficiency] with GaN-on-SiC in the 30- to 40-GHz range.”
As the frequency goes up, with all of the parasitics in the device, you start to lose that efficiency, and that is what kills off the usability of the technology, Dragon said. “That is where you’re into problems trying to design power amplifiers. You really need those things to run efficiently and you need them to be linear. The power-added efficiency is going to be what gets you.”
Designers also have to consider thermal management.
“GaN is great: It’s got a lot of power density per millimeter of periphery in the devices and it’s really cranking out the watts, but you’ve got to manage all that heat, which is why the SiC is so important,” Dragon said. “It’s very good normally, but as you approach these mmWaves, there are actually big debates around moving to GaN-on-Si for a couple of reasons. One is that you don’t necessarily need the SiC for the thermals anymore and silicon is going to be much cheaper than GaN-on-SiC.”
“Thermals still are something that need to be strongly considered in these radios because you have an area-per-watt problem that needs to be solved and the thermal design of the entire radio is still very important and a very strong consideration for designers,” added Tucker.
Monolithic microwave integrated circuits (MMICs) will also become more important at those higher frequencies, Dragon said. Integrating all of those different components, making them more repeatable and reliable, will become really important at those higher frequencies, he added.
“At higher and higher frequencies, integration becomes important—we can’t use traditional chip and wire interconnects,” agreed Tucker. “If we were to go take GaN-on-SiC and do a full-blown MMIC type of design, that would be horribly expensive. It would probably be wonderful for the power amplifier but pretty poor for all the other features and functionality that wind up on these chips. So the trick to GaN at a higher frequency is how we can put it together with another technology that supports the other analog functionality that we need in a cost-effective way.”
Whether it will be a chiplet type of approach using GaN-on-SiC die along with a SiGe die or GaN-on-Si integrated on a bigger chip, that all remains to be seen, but these are the sorts of things that are being discussed at conferences and in universities that are doing research to tackle exactly that problem, he added.
“It is more difficult to do, but we will see more of it as we go up higher in frequency, without a doubt,” Dragon said.
“I think GaN has a good place in those mmWave ranges,” he added. “In terms of the R&D aspect, whether it moves over to N-polar or GaN-on-Si because of the integration piece, those are interesting questions, but I think GaN will absolutely be there. But traditional silicon LDMOS types will not be there at all.”