The Growing Challenges Of 5G Reliability

The test field is getting more complicated as chips become larger, more heterogeneous, and subject to almost constant changes. Nowhere is this more evident than in 5G, where standards are still evolving and use cases are still being defined.

Without passing test, no technology advances. But those definitions are subject to change, and they can change again over time.

The communications industry’s primary goal with 5G is to get more traffic on limited spectrum and keep the traffic flowing at high speed with zero latency. Secondary goals are lowering power use and costs. Reliability, convenience, and security are certainly also important to the success of 5G.

To reach the primary goal, the industry is using high-capacity frequencies that were previously off limits and new complex electronic systems with new sophisticated chips. The bag of technical tricks to choose from includes massive MIMO (multiple-input, multiple-output) antenna arrays that pack more traffic on a signal and beamforming to precisely steer signals at targets. It also includes spectrum sharing systems that tell mobile phone users when to vacate a piece of shared spectrum, and an infrastructure buildout — also known as network densification — of macro cells, metro cells, base stations, small cells and distributed antenna systems to move signals around.

Not only will the large number of devices using the network cause congestion, but a huge amount of data moving securely through front haul and backhaul portions of the network will also tax networks.

5G has many frontiers for test because so much of 5G is a frontier with standards still settling into place. Chip and system testing must include a range frequencies in addition to the 5G, which will use spectrum in LTE frequency range (600 MHz to 6 GHz) and in millimeter wave bands (24 to 86 GHz). Test equipment has to handle higher bandwidth in more complex signals.

Often the focus of concern is testing of any devices using millimeter wave, which will increase over the air (OTA) testing done in a test chamber. Spectrum sharing testing is just beginning to exist, and the standards for 5G New Radio are still being solidified. The good news is tests can be done with existing tools, but the way tools are used may change.

“The current frequencies clearly aren’t going away, so we have to continue to test those in the way that we have done in the past,” said Neill Mullinger, product manager in the vertical market solutions group at Mentor, a Siemens Business. “But then you augment that with new solutions to deal with the expanded frequencies of 5G. It’s an area where, again, the extant family of testers allows you to do a lot of that and bring it virtual, which means we can set up a test environment for the lab or the field and run the exact same environment. It’s an area that’s going to evolve over time.”

He noted this was one of the drivers behind Mentor’s acquisition of Sarokal Test Systems last year. Sarokal makes 5G front-haul test equipment.

“They do testers for the field and in the lab, and we can now bring that forward and virtualize what they have so it can be used pre-silicon,” Mullinger said. “As you are respinning things, you are able to do a lot of the analysis that you do in the lab much sooner in the process and try out all different kinds of scenarios.”

This is particularly difficult for antenna arrays, where there are no exposed leads, and it becomes particularly difficult in the millimeter wave space.

“It starts from the mathematical from the tools that are doing the antenna design,” said Peter Zhang, R&D manager in Synopsys’ solutions group. “But people do have to prototype and figure out a way to be able to test it in a more realistic environment so they can see the calculated gain they are accomplishing with the antenna. This is actually the same thing applied to the digital design part of the ASIC, because we can all have very nice simulated results with the formulas. We can have a very nice model describing what is the radio environment around us, and there are so many models available. But when you go to 5G and you run data in the millimeter wave, which is a lot more complex than we have before, do the models really represent what we’re going to deal with when our system is outside?”

Different versions of 5G
5G millimeter wave testing is a different beast. “If engineers are used to working at lower frequencies on these earlier cellular applications, and they transition to working on 5G at higher frequencies, all of sudden all the rules are more stringent. All the rules of thumb go out the door, and you have to do a more thorough design,” said Mike Leffel, an application engineer at Rohde & Schwarz.

5G New Radio has two types of waveforms — cyclic-prefix OFDM (CP-OFDM) for downlink and uplink, and discrete Fourier transform spread OFDM (DFT-S-OFDM) for uplink only. This waveform resembles LTE’s single-carrier frequency division multiple access (SC-FDMA). The waveforms are wider, which makes it challenging to create and manage these waveforms during testing. Therefore, wideband test equipment is needed.

Many of the beamforming ICs and devices only can be tested over the air because they don’t have a physical connector for test equipment. (Using a physical connection would not produce a realistic test, anyway.) The OTA tests need to be done in chamber, and getting that to scale up for production-level speed and accuracy is a challenge. It doesn’t help that testing standards are still being formed. 3GPP, the international standards organization for 5G NR, is still working on testing standards.

“It’s still pretty early, and everybody’s trying to figure out what’s the best production-grade solution for testing these devices, which offer no traditional RF connectors, over the air at different stages,” said Alejandro Buritica, senior solutions marketing manager at National Instruments. “Either at the chip level, package level, module level, system level, we are trying to figure out some valid alternatives to do it consistently, accurately, and to be able to repeat it for millions of units at a reasonable cost.”

Parametric testing
For 5G chips, however, using parametric testing — where all aspects of the spec are tested—may help with the complex chips that include RF and analog components. “Part of the purpose of the parametric testing is to identify defects before you get to packaged part. One of the challenges with some of the advanced chips, specifically with 5G, is that there’s so much complexity, with the not only the RF measurements, but just in the chip itself,” said David Hall, head of semiconductor marketing at NI. “You have to do a lot of the low-level parametric measurements, but then you also have to do a lot of packaged part testing, as well, because it’s typically insufficient to identify defects.”

Parametric testing isn’t just traditional parametric IC testing, where the probes touch the wafer. “The peer of parametric test is functional or system-level test, where parametric means you’re looking at specific measurement parameters, such as power or gain or modulation quality,” said Hall. “For final production tests of RF and 5G chips, that’s a really important aspect of ensuring the chips’ quality.” Some parametric testing is done in both the wafer tests and the final package tests. “Everything in the spec document gets measured as part of a parametric measurement,” said Hall.

Fig. 1: Component characterization and validation require more testing. Source: National Instruments

The parametric tests can be done starting in the design stage and continue throughout the process.

“We work very closely with the designers along the design process of the ICs to take these measurements and enable their labs to make the initial measurements on their first silicon samples,” said NI’s Buritica. “Then, as they move to slightly higher volume and they’re building reference boards and reference designs, we continue to provide, using the same platform, automated tests and instrumentation for automating the tests to measure these devices.”

Throughout the process the design team uses the same tests to verify their IC performs well in different contexts, such as in a system or a surrounding set of circuitry for specific functions, such as a power amplifier or for beamformer. With advanced nodes, advanced packaging and systems on packages, relying on built-in self test becomes more important.

“You have to rely a lot on whatever is built-in self-test to make sure there is correct, inner working between different dies,” said Buritica. “They might have been tested independently. And they might have parametric data on those. But then, as they work together as part of a package device such as a SiP, finding those defects gets harder unless you have some sort of system-level test and a set of specific analytics that look for specific types of defects. At that point, creating some sort of regression test makes sure that the small change they introduced is now covered in that kind of approach. But it’s becoming increasingly difficult to do these tests because you’re now testing a system of various systems within the package device.”

Beamforming testing
The beamforming antenna equipment is tricky to test because of its complexity. The beamformer steers signals to a target. “For a beamforming device — or let’s call it an antenna-in-package or antenna and module type of device — you’re going to have a highly integrated processing unit that’s able to tell the analog components, the phase shifters, the gain states to go to a certain state to steer the beam in a particular direction,” said Buritica. “And so you’re going to need coverage to make sure that digital part is working for these beamformers. But then you don’t know how much your signal is being changed as it goes through, so you have a design you think is going to be one way, but then it needs to be highly verified.”

That’s just the beginning. “Once you surround that device with a screen, plastic casing, ground traces going around it and so on, it starts to detune the antenna. It starts to change how it behaves, and so what you thought was going to be a 45° steering now becomes a 52° steering, and that’s no longer good enough. So now you have to check the performance of these devices at the system level. They need to be very, very well characterized,” said Buritica. “It’s impossible to do that if you don’t have proper analog signals coming through. Your test and measurement equipment must provide the right stimulus, and you have to be able to trust that the response you’re measuring in terms of the phase of a very high frequency signal like a millimeter wave is within a couple of degrees.”

There are other potential pitfalls, as well. Any metal close to the antennas can change the behavior of the waveform. That includes a screen, a battery, a screw, or a PCB. All of that needs to be taken into account in the design and tested at the system level.

Wafer test analytics
Because of the amount of testing that must be done, and the amount of data that must be analyzed, 5G test will benefit from machine learning.

“Let’s assume that we analyze the images and now we have a whole bunch of parametric data that we have extracted,” said Sam Janaidi, vice president of automotive solutions at OptimalPlus. “We start doing everything possible with our customers’ current data, but then of course we quickly improve that by identifying additional high-value data that is not available. We have to make sure that it becomes available. Or sometimes we synthesize the data, maybe by triangulating off of a couple of other indirect data points that could produce data of interest.”

That data, in turn, can be looped back into the production flow to improve the consistency of the antennas.

“To produce consistent antennas that are high-performance, we have to do what we call AC test,” Janaidi said. “We have to make sure they manage the high speed and high frequencies very well. This is where it gets quite interesting, because now we can do impedance checks, which actually run certain AC characteristic tests on the circuit and bring all that data in. The net benefit is to enhance the ability of flex circuit manufacturers to produce homogeneous antennas through collecting all the data in a historical fashion, creating a statistical template for their performance, and then being able to tighten that over time so most of the antennas operate and function with as little variation as possible.”

Others agree. “There are a lot of data that are not really used right now from the fabs,” said Oreste Donzella, executive vice president at KLA. “But having all these available data running in a machine learning model can drive the inspection practice to be much more intelligent.”

Donzella noted that when 5G makes its way into cars, it will be scrutinized even more. Analytics from different test phases will be essential.

Conclusion
Implementing a validation or production test strategy for new wireless standards is difficult enough, but with 5G millimeter wave it’s even more complicated. Establishing a methodology and applying analytics to data will help significantly.

“If you think of 4G as mobile broadband, you should start thinking of 5G as mobile cloud,” said Durga Malladi, senior vice president at Qualcomm, in a recent presentation. “It’s a way of actually accessing the cloud.”

But nothing works in the market until it’s proven in test, and test will make it possible or prove the model flawed.

Related Stories

Solving 5G’s Thorniest Issues

Gaps In 5G Test

5G Drives New Test Approaches