"The penetration of 5G is happening," Verizon CEO Hans Erik Vestberg said in the company's most recent earnings report.
Some beg to differ.
"Up until now it has been more hype than reality," says Mike Noonen, the chief executive of wireless chip startup MixComm, regarding 5G.
"What has been rolled out so far is not ready for prime time," argues Noonen. "A scattering of headlines: it's too hot and you can't find it; it's too costly."
There is evidence indeed that 5G has so far not fulfilled its promise.
As early reviewers found in the past year, not all 5G is equal. Tests show 5G has not yet realized, at least in a reliable fashion, the promise of being "ten times faster than home WiFi."
A study, for example, by the bandwidth measurement firm Ookla in July found that T-Mobile was the fastest at 5G in the U.S., and had the best availability of a 5G signal. The best speeds, however, on T-Mobile, topped out at under 100 megabits per second, with limited availability.
Even believers on Wall Street don't think 5G is really happening yet.
"We're still in the twilight of the LTE era," wrote analyst Craig Moffett of the eponymous MoffettNathanson research firm last month.
"Perhaps the best we can hope is to white-knuckle our way through until the new iPhone, and the real 5G era, arrive."
The new iPhone 13, unveiled last week, broadens 5G support in some markets, but it doesn't extend 5G to new geographies as hoped, such as Japan. The iPhone 13 seems not to be the solution.
To MixComm's Noonen, there is a broad and deep technological issue that must be addressed.
"It's a problem to be solved and an opportunity to be exploited," said Noonen in an interview via Zoom from his home in Truckee, California.
MixComm, the four-year-old startup Noonen runs has received roughly $15 million in venture capital, including a B round this year, to develop a family of chips that can operate in devices ranging from smartphones to base station equipment.
The premise is that the chips will fix what ails the technology known as millimeter wave.
Millimeter wave is one portion of the electromagnetic spectrum. The "extremely high frequencies" of millimeter wave can transmit many more bits per second than other spectrum. All the carriers are deploying millimeter wave as part of their 5G efforts.
But deployment has been spotty. Another study in July, by research firm Opensignal, also in the U.S., found that across Verizon, T-Mobile and AT&T, the average time that users were able to connect to millimeter wave signals in various locations was always less than 1% of their total connection time.
"Nobody has been able to make millimeter wave behave," said Noonen. Without it, there will never be the promised speed boost of 5G, he insists. "Somebody has to put the 20% more "G" in 5G," says Noonen.
Millimeter wave refers to a swathe of electromagnetic radiation where the "wavelength," the distance from peak to peak, or from trough to trough, of a wave measures a millimeter or multiples of a millimeter.
GSMA, the cellular industry's standards setting body, specifies millimeter wave frequencies from 24 gigahertz to 29.5 gigahertz, as well as spectrum in the 37 gigahertz to 43.5 gigahertz range.
It's also expected that still higher frequencies will be used, in particular unlicensed spectrum at 60 gigahertz. The potential range of millimeter wave is as high as 300 gigahertz.
Most cellular communications today use what are typically considered microwave frequencies with wavelengths that are much longer, as long as a meter, and frequencies that are much lower, mostly 3 gigahertz and below, the so-called S-band.
Verizon and others, while building out millimeter wave, have mostly emphasized something in between, what is called C-band spectrum, with wavelengths that are several centimeters long and frequencies of 4 gigahertz to 8 gigahertz.
The carriers have collectively spent $80 billion to lease rights to C-band from the FCC, and the spectrum has been widely characterized by some as the savior for 5G.
Noonen, however, is in the camp that believes the true promise lies up the frequency scale, with millimeter.
"It's the lowest-cost way to deliver a gigabyte of data," Noonen told ZDNet regarding millimeter waves. Efficiency really is the key, because "5G is more about the service provider than the consumer," says Noonen. It's all about streamlining costs.
The appeal of millimeter wave over S-band and C-band is obvious, and so is the challenge. Any time you Increase the frequency and decrease the wavelength of spectrum, it boosts bandwidth because the more frequent the periodic cycles of a wave, the more symbols can be encoded per second.
Millimeter wave can theoretically offer raw throughout speeds of several billions of bits per second versus mere tens or hundreds of millions of bits at C-band and below.
While banking on C-band, Verizon and all other carriers know that millimeter wave is crucial. Millimeter wave is the enabling technology for what Verizon is calling its "Ultra Wideband" service, which it hopes will drive subscribers to take its "Premium" tier of service.
As of the latest quarterly report, 27% of Verizon's customers are on the Premium plan, Verizon's CFO, Matthew Ellis, said at an investment banking conference last month. The company aims to boost that amount "significantly" over the next two years, he said.
However, millimeter wave cannot reach very far, perhaps a kilometer in most settings, because it is absorbed by molecules in the atmosphere. Nor cannot it penetrate easily through walls and other obstructions. Those shortcomings mean that millimeter wave is dependent on so-called small cells, a radius around a cellular base station that is much, much smaller than the typical cell site.
That means tons of capital spent to put many more radios in many more locations throughout the land, with particular challenges in rural areas where population density doesn't always justify such a capital investment in terms of subscriber returns.
"What's been rolled out so far hasn't been ready for prime time," says Noonen of current millimeter wave deployments.
"There are three challenges: poor range, insufficient power, and then, this stuff is just too expensive," Noonen told ZDNet.
Enter MixComm. The company has perfected, says Noonen, production of what are called front-end radio frequency integrated circuits, using a special class of semiconductor material known as "silicon on insulator," or SOI.
A front-end RF IC is the chip that takes a signal from the antenna and processes it using mixers, filters and digital-to-analog converters, and then sends it on to a separate intermediate frequency (IF) integrated circuit, where the millimeter wave frequencies are converted to much lower frequencies that can then be decoded by the digital baseband processor.
The same process happens in reverse when a device is transmitting, sending the signal to the front-end RF IC to be boosted by the amplifier and then placed onto the antenna and transmitted.
The key architectural issue of a millimeter wave front-end IC is that the antenna needs to be very close to the IC, Noonen explained.
"Up until now, that's been very difficult to do because the ICs ran too hot, they would cook the antenna," said Noonen. But if one lowers the power of the IC, to reduce heat, the whole system would be too weak to boost the millimeter wave signal to reach the tower.
The leader in wireless chips, Qualcomm, has of course for years been able to do millimeter wave in standard silicon chips, known as CMOS. Noonen argues Qualcomm can't handle the power requirements needed for chips in wireless infrastructure such as base stations.
"Qualcomm has done a terrific job for handsets, that is working great," says Noonen. "But it's one hand clapping," he says, leaving infrastructure adrift. "They would love people to use their products for infrastructure, but if all you have is a hammer…"
What is needed for infrastructure is a high-frequency material that is different from Qualcomm's CMOS. The RF industry has long worked with alternative materials that have greater power efficiency, known as mixed-signal chips, such as gallium arsenide and silicon germanium. Unlike CMOS, however, they can't integrate the digital functions that need to be on-chip to process signals.
The solution for MixComm is silicon on insulator, a broad class of materials that place a thin film of ordinary silicon on top of a substrate that is not semiconducting, but rather acts to block current, an insulator.
MixComm's roots are a program at Columbia University's High-Speed and Millimeter-wave Integrated Circuit Lab, which received $94 million in funding from DARPA over a decade.
One scientist from that project, Harish Krishnaswamy, a tenured professor of electrical engineering, is a co-founder of the company. Krishnaswamy was able to make RF SOI work in a practical fashion. "Harish over two years worked out all the fundamentals to get RF SOI to work," says Noonen.
The company's first implementation, a front-end RF IC called "Summit," has been shipping since the fourth quarter of last year. This summer, MixComm announced it had achieved the crucial integration of antenna and IC, by combining Summit with MixComm's own antenna modules.
Those modules can put sixteen antennae in the same physical package as the Summit IC. The approach is known as "antenna in package," or "AiP." Most manufacturers assemble circuit boards with multiple heat sinks to combine ICs and antennae; MixComm's package crams four ICs and the sixteen antennae into an impossibly small area, a square of fifteen millimeters on a side.
The cost of the integrated package can be 30% to 70% cheaper than what is currently on the market in the form of multiple chips and boards, claims Noonen.
The result, said Noonen, is a sweet spot, an integrated device that has sufficient output power for the amplifier but also "a level of integration and efficiency where you really hit the key money specs," meaning, cost-effective for equipment makers.
There is an additional dimension, however. The RF IC includes fast SRAM memory circuits that hold information about "beam shifts," the way that the individual antenna elements are directed to achieve the most focused signal.
Beam shifts are a kind of magical way to steer a wireless signal, known as beam forming. Krishnaswamy, along with fellow co-founder Frank Lane, who was for nine years vice president of engineering for cellular chip giant Qualcomm, have patents on technology for beam-forming.
Beam forming takes multiple antennae and directs their energy in a coordinated fashion in a narrowly focused area. By being focused, a wireless signal is able to arrive at the receiver with greater power than if it were spread out in space. More power at the receiver means less error in transmission, and therefore higher data rates.
To focus those antennae into a tight beam, the on-board SRAM is larger in capacity than normal, says Noonen. Its integration into SOI material is another MixComm breakthrough.
The package still requires a baseband processor. Qualcomm dominates the baseband chip market and there's no way they'll partner with MixComm, because Qualcomm aims to get its next billion dollars by selling its own RF ICs.
Fortunately, there are a number of competitors to Qualcomm that MixComm is looking to partner with, including Marvell, Mediatek, and Samsung. There is a multi-vendor initiative of wireless chip vendors such as Skyworks and Qorvo, along with Intel and others, called OpenRF, which aims to get around Qualcomm's dominance.
"Our job is to be the loyal opposition," says Noonen.
The point of all this is for the chip to use power in a very selective fashion. The Summit chip, connected with the antenna array, has "twice the efficiency of anything out there," says Noonen, and four to ten dBm (decibel milliwatts) more output power.
"Because you have better output power, you can reduce the number of antenna elements needed" in order to "hear" a smartphone, he explains. For example, Samsung's network infrastructure business will deploy 1,000 antennae for an area, "but we can do it in a fourth of that."
"If you can boost the power by 2 dBm (decibel milliwatts), you can reduce the capital expenditures per square kilometer by a million dollars," Noonen explains. That's because "we make small cells bigger" by letting the radios transmit farther with focused beams.
"Think of how many base stations are going to get deployed; if you can save a million dollars a kilometer, that adds up really quick."
That could speak to one of the main "pain points" that have hampered Verizon and others. They have taken "baby steps" in deploying millimeter wave because it has been expensive to build the many small cells needed for the short range of the frequencies.
"Up until now, we really haven't had the horsepower to cover an area cost-effectively," he says. "Anyone who can reduce the cost of deployment, that's a big, big help," he says. "The capex avoidance is tremendous."
The AiP can span a product range from handsets to base stations. As many as four AiP antenna arrays can be combined by customers, to put as many as 256 antennaa in a product.
On top of the raw capex savings, millimeter wave, done right, will lead to vast efficiencies, says Noonen. The higher bandwidth of the frequencies means "We can use more efficient modulation schemes so you can cram more bits into the same space," he says.
There is another, hidden benefit: power savings.
"It ends up being a power savings, because you can transmit more in a short amount of time," explains Noonen. That means that despite higher-power chips, "millimeter wave actually is going to save battery life rather than consume it just because of the duty cycle of getting the job done."
The stage is now set for millimeter wave. The pandemic delayed the roll out of 5G by a year, said Noonen, as carriers were "frozen in time."
That was a blessing in disguise, for it allowed MixComm to engage with its customers. "They realized they needed something better, and they used this downtime to get ready for a pent-up demand for better connectivity and a new way of living," he said, referring to the newly nomadic workforce.
The coming 18 months will be "when we start to ramp, and you will see significant deployments on a global basis."
An enormous base of Apple and Android phones with millimeter wave will create momentum from the client side, says Noonen.
"There are going to be a half a billion smartphones that are millimeter-wave capable in people's hands," he observes.
Apple's iPhone 12, released last year, already supports C-band and also millimeter wave, specifically millimeter wave bands formally designated by the GMSA for 5G, called "n260," and "n261," for 39 gigahertz and 28 gigahertz, respectively. With the iPhone 13 unveiling last Tuesday, Apple announced the iPhone 13 will add a new millimeter wave band, "n258."
Driven by the pandemic scattering of staff from offices, and the new hybrid future of work, it is possible millimeter wave will have an even greater role to play in 5G.
Noonen sees a market not just in telecom, but in private installations as well. "Our basic thesis is this is going to move into other domains such as factory floor and defense and automotive," says Noonen.
Things such as robotic arms and autonomous vehicles will be big consumers of millimeter wave bandwidth.
"This is not going to be much bigger than a WiFi access point, so enterprises, companies such as corning, can get better connectivity on private networks than they would on WiFi 6," he contends.
The pandemic has pushed information workers to expect bandwidth in many more locales, he notes.
"The way people are using their devices has changed forever," says Noonen. "We'll never work every again in one silo," he notes. As people are able to get a gigabit of bandwidth — Ethernet speeds — on their phone, the true premise of 5G, "nobody is going to give that back."
For Noonen, who is a semiconductor industry veteran, the kinds of fundamental innovation going on in SOI represent a rebirth of the semiconductor industry after decades of relative lack of innovation.
Before coming to MixComm two years ago, Noonen for years ran sales for giant semiconductor firms including NXP Semiconductor, Rambus, enterprise communications technology firm 8x8, and chip foundry Global Foundries.
Noonen has seen many eras of the chip industry. He can remember programming on one of the early AI computers, a workstation from 1980s startup Symbolics, in the scientific language LISP, back in his NCR days.
Noonen was also a co-founder in 2014 of Silicon Catalyst, a consortium of investors that emerged to encourage venture investments into semi startups at a time when the field was deeply out of favor with investors.
Given that background in the chip field, it is important to Noonen that the achievement of breakthrough 5G technology is being nurtured by an American startup company.
"It flies in the face of conventional wisdom that China and Huawei has all the 5G smarts," said Noonen. "Maybe for sub-6 [gigahertz frequency], but millimeter wave? Absolutely not."
Those geopolitical elements are less important than how the whole spectrum map may be shaken up in coming years. Existing spectrum is filling up to deal with the spiraling amounts of bandwidth demand.
"Even Sub-6 and C-band will be oversubscribed in every metropolitan area the next two years," says Noonen. "Moving up-spectrum is inevitable."
Edge computing will start to exploit more frequencies of spectrum as cost-effective access points spread around the land. LTE radios could be decommissioned, or "re-farmed," something that Noonen rates a distinct possibility.
"There's going to be that opportunity," says Noonen.
Down the road is 6G, transmitting in the terahertz range. Although 6G is "only 16% more G," quips Noonen, it will require substantial innovation.
"There is lots of invention that is still going to happen," says Noonen. Millimeter wave is a kind of stepping stone to that. "To tackle terahertz, you have to master gigahertz."
"We are at early days, the very beginning, this is the first inning of what's going to be more than a nine-inning game," he says, "and there's lots of opportunity ahead of us."