Making 5G More Reliable

The rollout of 5G is a complex and monumental effort involving multiple separate systems that need to function flawlessly together in real-time, making it difficult to determine where problems might arise, or how and when to test for them.

Investments in 5G have been underway for the better part of a decade, and the technology is considered the next huge growth opportunity for mobile devices — from smart phones to cars and trucks — as well as smart infrastructure and cities, smart manufacturing, and even industries such as farming, where autonomous farm equipment can be controlled more precisely.

The challenges involved in this rollout are significant, though. And despite some initial successes and technology improvements, those challenges still aren’t completely solved. Among them:

• There is no single implementation of 5G. In addition to the two most obvious flavors of this technology — sub-6 GHz, which is essentially a faster version of 4G LTE, and millimeter wave (mmWave), which uses higher frequency bands — there are localized implementation challenges based upon regional differences in available wireless spectra, terrain, and available capital to build and monitor the infrastructure.
• Depending upon the frequency, signals will attenuate at different distances and rates. The shorter the distance, the more repeaters that will be required. All of this equipment, from base stations to repeaters, will be subject to weather. In extremely hot environments, the lifespan of the chip will be shortened due to accelerated chip aging. And in most applications, temperature variation will require continual testing and adjustments to account for signal drift.
• The higher the frequency in the mmWave spectra, the more susceptible signals are to interruption. That means the same base station system-in-package or SoC developed one country may not work as well in another. And because those spectra allotments are changing (the U.S. FCC plans to free up spectra from as low as 26 GHz to as high as 90 GHz, for example), chips developed today may not work as well in the future.

Despite these issues, communications companies and equipment makers are convinced mmWave technology is viable. Their investments in this technology are a reflection of that commitment, and the potential benefits are significant. At higher frequencies, much more data can be streamed in less time than with previous generations of chips, and it all can be done using less power.

Challenges ahead
It’s well understood that mmWave signals cannot bend around corners like 4G or sub-6 GHz 5G. With 6G, the next rev of wireless technology, these kinds of issues will be even more pronounced. But even before the technology reaches that stage, some basics need to be addressed, such as testing how the various components will work — base stations, signal extenders (repeaters), and handsets or other end devices.

Testing chips for quality and reliability during manufacturing is well understood. What’s different with 5G is the chips also have to be tested in the context of the other pieces and use cases. This has been a significant challenge, because none of the three key elements is fully developed.

“Every supplier of a chip has a different way they want to test their wireless links to make sure they’re working properly and verify their chips meet their own quality standards,” said David Vondran, wireless product manager at Teradyne. “You also have to fact in that every supplier has their own strategies about the position of that chip inside the ecosystem. Those are influenced by economics and quality. You could test the functionality of a chip for a premium device to make sure the sensitivity of the receivers is working properly. You might want to make sure the transmitters are operating in the right spectrum without contamination to any out-of-band performance. And you might want to collaborate on these things.”

Fig. 1: 5G-intermediate frequency signal distribution in smart phone. Source: Teradyne

Collaboration across the ecosystem is essential. “Today, there are millimeter wave phones around, but you only can use them in certain areas because the base stations have not yet been deployed,” said Adrian Kwan, senior business development manager at Advantest. “You’ll see 5G become real over the next year or two.”

In most regions, the rollout of 5G handsets and base stations continues to be spotty. But over the next few years, that is likely to accelerate. SK Telecom, for example, completed a trial earlier this year for a standalone 5G network, which allowed signals to revert to 4G when 5G speeds were unavailable. That combination will be necessary when signals are interrupted or there is insufficient 5G infrastructure. But a big challenge is how to test the interactions of all of this equipment in real time, and over equipment lifetimes, to prevent any service interruptions.

“With test, there are two different approaches,” said Kwan. “First, there are devices doing conductive testing for beamforming. There also are customers doing over-the-air (OTA) beamforming testing.”

Which way the market ultimately heads is unknown at this point. Advantest believes this ultimately will be a conductive ATE solution. Teradyne believes there will be a strong play for OTA. It’s not clear if both will be used for different pieces of the system, or whether one will win out. In either case, though, the goal is to improve reliability and reduce test time, which ultimately will bring the cost down and improve throughput in the fab.

Still, ensuring reliability requires the whole system to work, not just the components. In handsets, the phased antenna array is being built into a system-in-package, and quality of service may vary depending upon how the end device is used and how much noise other applications create. This requires both flexible prioritization of data, and intelligent partitioning of power.

A second challenge involves continuity of service. Disruptions may be an annoyance with a smart phone, but they can cause more serious problems when it comes to automotive applications or smart infrastructure.

“You need ultra low latency and multi-link topologies,” said Kwan. “Autonomous vehicles are very reliant on this kind of setup for continuous communication that will not drop. To make that work, customers will need to deploy repeaters, which may be on lamp posts or traffic lights.”

That adds yet another challenge, because some of these chips are being developed at advanced nodes. Under extreme conditions, circuits will age in well-understood ways. They can suffer from electromigration, time-dependent dielectric breakdown (TDDB), or inconsistent memory retention. In addition, quickly software can become bloated with a series of security updates, which can affect the performance in all of these devices.

“In the past, semiconductors went into very controlled environments — laptops, PCs, and data centers,” said Danielle Baptiste, vice president and general manager for software at Onto Innovation. That’s no longer the case, and there will need to be a constant loopback of monitoring data to improve reliability. “The chips are going to be reporting back so we can start to understand if there’s some sort of unpredictable result once it’s out in the field, and what that really means. Then we can start to feed that data back into the manufacturing process. What’s happening in the field becomes really compelling and interesting.”

That field also is getting wider and more varied, too. “Everyone knows that with the fifth generation of the cell phone we have the advantage of higher speeds,” said Andy Heinig, group leader for advanced systems integration and department head for efficient electronics at Fraunhofer IIS’ Engineering of Adaptive Systems Division. “We can reach speeds of up to 20 gigabits per second, and this is possible because of the new frequency bands. Compared to older generation technologies, we have massive input and output and real-time capabilities, and this works well for automotive and industrial applications. You can save energy if you have the same speed of communication, and in automotive you also can ‘look behind hills’ and ‘see around corners’ using car-to-car communication.”

There are less-obvious benefits, too. So instead of large antennas, at higher frequencies the size of those antennas can be reduced into 4 x 4 arrays in advanced packaging. And instead of putting everything on a board, some of the passive components can be moved into those packages, which provides improved performance.

Quality assurance
All of this needs to be viewed in the manufacturing and packaging flow, as well. Test is only one of a number of processes involved in 5G, and most of them become more complicated in this context. DFT is challenging enough with complex devices, but it’s more challenge when those devices need to work in sync.

Likewise, simulation, emulation, and prototyping in the context of other components requires much bigger data sets, as well as massive amounts of compute power. It also requires a continuous loop between the different players in an ecosystem because of the speed at which this technology is moving.

“The fact that the spec changes very often for 5G, and now 6G is the same problem we saw years ago with switches and routers,” said Jean-Marie Brunet, vice president of product management and product engineering at Siemens Digital Industries Software. “This requires you to have a model that you can modify and apply to the spec and see how it reacts virtually. These designs are also very big. We’re seeing a lot of designs with protocols switches, where there a lot of possible combinations. They need to be verified with multiple configurations, specs, formats, and interfaces. There are a gazillion combinations, The only way to do all of this is with hardware-assisted verification.”

The challenge here is comparing results may vary by time of day and ambient environmental conditions. “If someone is crossing a road at midnight, it’s going to be a very different simulation than if someone is doing that during the day with heavy traffic,” Brunet said. “It’s also going to vary, depending on the weather. You need to simulate those changes in communications. Otherwise, you have to test for everything. There are a lot of companies already using 5G, and they are emulating for their particular use.”

Fig. 2: Industrial 5G router. Source: Siemens

With advanced packaging, which is required for tight form factors, potential issues like die shift and structural weaknesses may not show up until the package is sealed. While this may seem a world away from how a smart phone works, getting this wrong can have reverberations around the globe. The problem is how to find them, and there is something of a slow-moving race underway among technologies that in the past were largely confined to research settings or low-volume manufacturing. This includes various ways to connect different chips in a package, as well as how to do metrology and inspection that is deep enough to identify latent defects and still fast enough to make it cost-effective.

“The reliability question continues to come up, and the typical answer is more metrology,” said Hector Lara, director and business manager for microelectronics AFM at Bruker Nano Surfaces. “With AFM, we’re seeing a ‘dusting off’ of conversations. We haven’t played a lot in packaging because the throughput requirements are so stringent. You would need a 300X improvement to match any other metrology tool, for example, and the reasonable throughput for this technology is in the 5X range. But companies are finding they can get rid of two or three other metrology steps with AFM that does x,y,z in packaging, and as they ramp up new fabs they want to go with new approaches.”

Conclusion
It takes time to work the kinks out of any new technology. Early cell phone coverage was spotty at best, and in highway traffic even low-frequency signals were dropped regularly because base stations couldn’t handle the volume. The challenge with 5G is not that calls will be dropped, but that the highest-speed connections may be interrupted, forcing devices to revert to lower-frequency communication.

How to plan for those kinds of interruptions and reduced performance is where all of this reliability testing, simulation, and ongoing monitoring will play a significant role for the foreseeable future.

“It’s a multi-disciplinary problem, and that’s the challenging part of all of this,” said Teradyne’s Vondran. “In this millimeter-wave frontier, finding all of that expertise in one person is very difficult. We’ve developed multi-disciplinary teams of experts — system integrators, so to speak — who have dedicated their careers to do this kind of stuff. And while it’s an extension of doing the conducted test of the past, they’re now applying it to innovative problems in the future, like beamforming antenna structures. It’s more than a team. It now requires an entire tribe.”