中文 English

Gaps In 5G Test

Millimeter wave technology will require a whole new approach to ensuring reliability.

popularity

Add one more industry to the long list that analysts expect 5G technology to disrupt—test. While the initial versions of this wireless technology will be little more than a faster version of 4G, concern is growing about exactly how to test the second phase of this technology, which will be based upon millimeter wave.

A number of fundamental problems need to be addressed. Among them:

  • There is no way to attach testers to antenna arrays needed to receive millimeter wave signals. The general consensus is this will have to be done over-the-air (OTA), but so far this kind of testing technology doesn’t exist.
  • It’s not clear how to test whether the receiving technology in the handset is reading the signals appropriately, due to mmWave signal attenuation and interference. Those signals can be disrupted by people, foliage or even weather, but figuring out whether it’s a hardware problem or a signal problem—and the source of that problem—isn’t trivial.
  • Likewise, the base station technology will require beam-forming technology to circumvent those objects and various types of interference. Whether that technology is working as planned, or whether it’s a problem with the handset, or both, will need to be part of the test strategy. So far, there are no clear answers yet about how to achieve this.

The test industry has made some strides. But where tests do exist for 5G technology the process today is manual, which means slow and costly. So while the United States’ FCC 28GHz spectrum auction fired the starting gun on chipmakers racing to release the first commercial millimeter-wave chips, no commercial, manufacturing-ready test-and-verification systems are yet able to do the work.

“There is a mad scramble from chipmakers, but there is a whole set of concerns about some things, especially over-the-air testing that haven’t been solved,” said David Hall, chief marketer at National Instruments. “People have been cobbling together production test systems for early silicon—some using network analyzers or higher-frequency RF analyzers. But you don’t want to try to use a fleet of those for a high-volume mobile device where you’d be verifying hundreds of millions of parts.”

Testing routines under 5G will require new techniques and automation to deal with the idiosyncratic requirements of 5G systems. Other changes will be organizational to deliver more testing officially. Others will require changes in attitude to acknowledge the increasing importance of 5G in the day-to-day lives of customers—although it is still too early to know whether 5G will fulfill the incredibly high expectations for its service.

Test providers don’t want to “be in a position where capabilities of ATE are holding the market back,” Hall said. Announcements about automatic test equipment (ATE) products show progress, but so far no fully functioning, commercially viable ATE designed for 5G exists, let alone one able to handle OTA as well as the other challenges, he says. “Customers want to know when we’re going to have a manufacturing-ready test solution that addresses a lot of issues—over-the-air testing being one of them.”

The pressure is already pushing some customers to blur the lines between development and testing. They are buying ATE to install in development labs—rather than test and verification departments—to help chip designers troubleshoot designs for prototypes more quickly, for example.

“Millimeter wave is a very different animal,” says Adrian Kwan, business development manager at Advantest. It’s been around for more than 50 years and has been used in military radar, which operates at a very high frequency range. Millimeter wave ATE was first introduced about 15 to 20 years ago in testers that go up to 50 or 60 GHz, but there were not a lot of mainstream consumer apps. And with 5G, we’re seeing up to 50 GHz—nothing more than 70 GHz. Now, as this technology moves into the consumer space, the market is trying to learn how this equipment can be adapted.”

More fundamental changes are ahead, too. “Whether you’re looking at sub-6 GHz or millimeter wave, you’re going to need a new class of devices that give you much higher power efficiency and a much more compact footprint,” says Ajit Paranjpe, CTO of Veeco. “You have to be able to power multiple antennas, and there is only so much footprint available and so much battery power available. You have to get frequency response from the silicon. It cannot deliver on the frequency or the size, so there are some restrictions there, but silicon still has an important role to play. The amplifier part will be using compound semiconductor. Whether that’s gallium arsenide or gallium nitride we don’t know. There will be a battle between the two. And then if you look at the other parts of the system, such as the modem, there probably will be direct conversion.”

What works best where remains to be seen, but devising test strategies based upon all of these changes becomes increasingly challenging, and the problem only grows with each new piece of the technology.

Beamforming, phased-array antennas
The shift to higher frequencies and the need for high-gain beamforming requires a phased-array antenna containing potentially hundreds of antenna components, each with its own RF connection. That makes the traditional practice of testing RF using a cable connecting transmitter and receiver impractical at best, and often impossible.

“The modulation scheme has changed with the demand for higher bandwidth channels,” says Advantest’s Kwan. “So we’ve gone from 20 MHz to 80 or 100 MHz per channel, and you need to combine different channels to really go to the characteristics of beam-forming.”

Beamforming also makes it impossible to test radio and antenna elements separately, as is common now, because the relevant result is not how much energy the antenna is able to put out, but how effectively it uses amplifiers and phase shifters and other components to direct the energy to a specific target and receive the result correctly, says Michael Foegelle, director of technology development at ETS-Lindgren, which sells kits and components for commercial test systems, including OTA testing of sub-6GHz 5G systems.

“5G antennas are hard to test because they don’t stop changing—they’re constantly modifying themselves for beamforming or to do MIMO (multiple input and multiple output), with advanced signal processing behind that, so it’s always changing its performance from a single angle or point of view,” according to Lawrence Williams, director of product management at ANSYS.

The only real way to test a complex antenna and its environment and take in continual change is a high-frequency structure simulator that can identify the signal-processing challenges, take the circuit design and chip configuration into account, and build a behavioral model that displays a realistic simulation of the physical processes involved, says Williams. That includes the design of the SoC, the structure of substrates and packaging, electromagnetic sources of interference and the potential for random error. It also needs to be capable of modeling and making recommendations about beamforming, power levels and data connections.

“With 5G, especially millimeter wave, you’re dealing with things that are very small, so the connector [for the RF tester] is larger than the antenna, and there are too many connections,” Foegelle says. “Worse, you’re talking about an active array with amplifiers and phase shifters and everything else, so you can’t get at the connection between the amplifier and the signal source. You have to measure from the point of view of the end user because the number [metric] you want is the sum of all the RF and antenna elements after they’ve been combined in space because some of the signal might arrive out of phase.”

Rohde & Schwarz multiport network analyzer ZNBT20 used in a 5G beamforming demo, where the device under test is an evaluation board from Analog Devices. Source: Semiconductor Engineering

Fig. 1: Rohde & Schwarz multiport network analyzer ZNBT20 used in a 5G beamforming demo, where the device under test is an evaluation board from Analog Devices. Source: Semiconductor Engineering

The number of complicating factors involved makes it difficult to come up with a single metric that rates even something as direct as how well a radio signal reaches a particular target.

Most 5G tests happen at the handset, but that is complicated because 5G signals, particularly at the higher frequencies, have trouble getting through the Styrofoam used as a base for the object being tested. Error Vector Magnitude (EVM), which measures how well a signal is modulated for broadcast and demodulated on reception, touches on only part of the whole equation. So there isn’t a good metric on which to base a test procedure designed to evaluate 5G chips, Foegelle says.

Over-the-air (OTA) testing adds another problem. This is more complicated than just testing the strength of the radio. The standard metric of wireless connection quality, EVM measures how effectively a signal is modulated for broadcast and demodulated on reception, but it doesn’t take into account things that could cause two parts of a signal to arrive out of phase, such as the handset’s CPU clock. Even simple tests like how well a handset receives a 5G signal have unexpected little traps, like the unexpected interference of the supposedly signal-neutral material used to hold up the phone being tested.

“We’re used to thinking of Styrofoam as air, so we don’t think about it for testing because it’s only holding up the phone,” Foegelle says. “The problem is that a signal at 28GHz going through 10 cm of Styrofoam shifts by half a wavelength compared to a wave that goes past the Styrofoam, so the two waves are out of phase and you have to get them to line up again.”

The phased-array and beam-forming have such a dramatic effect on the signal that most tests focus on them, but the shape or material of the device a handset or AP is made from, the plastic sheet used internally for protection and other design elements are relevant enough that testing the full system, not one component at a time, is the only real option.

Firmware
All those components are driven by firmware that makes decisions for the phased array, so the software also has to be included in verification of the array—another big departure from the norm.

Future versions of 5G will be all digital rather than generating analog signals that are converted for broadcast. That will make the whole system more efficient by eliminating the conversion, but will introduce the specific algorithms used in firmware to make decisions about beamforming as elements to be tested.

Software will play a very big part in basic functions, especially in identifying and tracking the source of connections in world of densely packed, rapidly changing cells, notes Michael Thompson, RF solutions architect in the Custom IC & PCB Group at Cadence. “You’ll need a little self diagnostic to test itself to see if there is an angle or change that will get better gain. Built-in self-testing makes sense, but you wouldn’t want it beam-searching all the time to try to get better gain,” Thompson says.

Cadence and other providers are building test and optimization functions into their existing test suites, but the expertise to code complex beamforming and gain-maximizing functions is, like many other skills required for 5G rollouts, thin on the ground. “The ratio of antenna designers to circuit designers has always been low, so it’s better to have some of that capability built in.”

In complex systems, however, the less heavy processing left to be done in software, the better,

“The very low latency requirements for 5G mean some of the things you had the luxury of doing in software will have to be done at the PHY layer,” according to Susheel Tadikonda, vice president of engineering for Synopsys’ Verification Group. “The PHY can support a lot of algorithms, so it will be much faster if you can build much of the beamforming logic into the PHY rather than leaving it in software.”

That approach does put a greater onus on designers’ ability to build complicated beamforming and RF-source tracking into PHY layers that are already extremely complicated, Tadikonda says. That alone will increase the level of complexity involved in verification, making it difficult for organizations dealing with home-grown testing tools and processes to keep up.

In 5G, however, software consists of more than just drivers to make other parts work. As designs evolve gradually from analog beamforming toward digital, the algorithms used to control the phone will become more significant parts of the RF and signal processing—which would make firmware part of the critical systems functions that have to be tested and verified, rather than just a way to tell hardware what to do when a button is pressed.

“That’s another of the changes that will completely change not only the way you design the radio but how you test it,” Foegelle says.

There are no accurate, relevant tests at the moment, and probably won’t be until more 5G devices hit the market and testing experts get the chance to evaluate not just the devices, but how well specific metrics reflect the experience of the end user of a device—and how well the tests those experts device reflect the right metrics.

“It’s a pretty big shift in assumptions and how you’d approach testing, Foegelle says. It’s difficult or impossible right now even to devise new tests or evaluate specific metrics, however, because uncertainty over how well each component will perform and how they will perform in concert make it impossible even to establish whether a metric is relevant, let alone effective.

“From the network operator’s standpoint, if I want to correct a drop in phone performance of 2 db to 2.5 db by changing my network, that equates to putting 25% more base stations on the network, so it’s significant,” Foegelle says. “Everyone is being cautious about estimates, but guesses put that number two or three times higher, so people are really cringing because the potential error in measurements means the numbers could be almost useless,” Foegelle says.

The good news, however, is that much of the uncertainty will clear up and test engineers will be able to find measurements that might not be absolutely precise but will be accurate enough.

“If you’re writing an algorithm to direct a beam to shift 45 degrees down, can you still do that with a number that is accurate at 100 feet rather than 10? I think probably yes, we can figure a way to work that out,” he notes. “Right now, the pressure to develop these test methods and define something that’s not yet defines is very much a cart-before-the-horse situation.”

Conclusion
5G is definitely a technical challenge that will require changes in test and verification organizations trying to keep up. It also is expected to become a ubiquitous, always-on data dial-tone of the kind computer companies have promised many times over the years.

If 5G is able to deliver that, end users will rely on it for more than texting and GPS. It will drive services they either cannot live without or will not tolerate losing if a signal disappears for a while—a level of service much higher than is typical of either cellular networks or WiFi networks today.

“The level of reliability needed is at a completely different level—more like autonomous vehicles than phones,” says ANSYS’ Williams. “We have to talk about how to architect a data environment, not just a network.”

—Susan Rambo and Ed Sperling contributed to this report.

Related Stories
Testing Millimeter Wave For 5G
Stumbling blocks emerge for ensuring reliability of next-gen wireless devices.
Preparing For A 5G World
Rising costs, massive complexity, and challenges in testing and simulating these devices will require some fundamental shifts in the communications market.
New 5G Hurdles
Getting 5G standards and technology ready are only part of the problem. Reducing latency and developing applications to utilize 5G have a long way to go.
Issues In Designing 5G Beamforming Antennas
Assuring quality under changing conditions with shifting standards and use models is a major challenge.



Leave a Reply


(Note: This name will be displayed publicly)