5G Brings New Testing Challenges

As 5G nears commercial reality, makers of chips and systems that will support 5G will need to take on the standard burden of characterizing and testing their systems to ensure both performance and regulatory adherence.

Millimeter-wave (mmWave) and beamforming capabilities present the biggest testing challenges.

“5G is expected to have the extended coverage plus the bandwidth to harness the world of IoT that’s coming into our industries,” said Adrian Kwan, business development manager at Advantest. “By 2025, there’ll be about 40 billion devices connected to this ecosystem.”

Jeorge Hurtarte, wireless product marketing strategist at Teradyne, noted that smart phone data will continue to increase over the next five years. “It is expected that by 2025 about 45% of data traffic will be on 5G networks,” he said.

Those phones contribute even more when it comes to data. “Smartphones generate over 95% of the mobile data traffic today, and continue to increase,” Hurtarte added. And new applications mean tighter specs.

And depending upon the application, that will require consistent high bandwidth and low latency. “We’re going to need ultra-low latency for mission-critical uses like autonomous driving,” said Daniel Bock, customer RF applications specialist for the RF Product Group at FormFactor. “You need to have 1-ms latency or it could cause serious accidents.”

Whether at wafer probe, final test, or system test, test equipment and test methodologies will need to evolve in order for chips and systems to be cost-effective.

5G ups the ante
5G is expected to be 10 to 100 times faster than 4G, and potentially as much as 1,000 times faster with 1% of 4G’s latency. That kind of speed will place higher demands on the testing regime. RF testing for cellular chips and systems is not new, but the focus until recently has been on, at most, 6 GHz. The low-GHz range is crowded, but the jump to mmWave is large, moving straight from 6 GHz to 28 GHz. The highest anticipated frequencies are in the 40- or even 60-GHz range.

It may be awhile before mmWave hits high volumes. “It’s going to take some time for millimeter wave infrastructure to develop,” said Bock. “It’s going to be more expensive, and the ranges are so short. Almost every telephone pole would have to be a 5G base station to be able to do it. There are towers and light posts that have [the base station] integrated inside of them. For the higher frequencies to be functional, that’s what you’re going to have to do.”

But even within the lower-frequency ranges, the greater usage of beamforming will prove challenging to test. Many chips will have phased antenna arrays built-in (referred to as antenna-in-package, or AiP, as contrasted with antenna-on-package, or AoP). They will work with the multiple cell towers and with multiple antennas on a single tower. By focusing the transmitted energy in the direction of a specific receiver – either on a mobile device or on a cell tower — less energy is required because most of it is no longer wasted in radiation away from the intended receiver. In addition, more mobile devices can be serviced, because they won’t all be bathed in discernible messages for others.

Those antennas may have to support up to 36 RF channels, Bock said. That requires high accuracy and precision, but it also creates challenges for running tests quickly in order to keep test throughput at an economic level.

The new RF features will require significant new silicon content in high-end 5G handsets. In fact, most of the new content in handsets will be dedicated to supporting advanced RF capabilities. According to Kwan, the amount of RF front-end content for a midrange 5G phone will be approximately equal to that of a high-end LTE phone. In a high-end sub-6-GHz 5G phone, that content will increase by 1.5X. And phones supporting mmWave will increase that content by 2.5X. Kwan also noted that as much as 90% of the testing at the module level will be for RF.

Fig. 1: Challenges for testing 5G handsets. Source: Advantest

Starting with the wafer
Bock pointed to the challenges of wafer test, saying that probe card design makes a big difference on the frequencies that can be achieved and referred to four different ways FormFactor uses to construct a probe card:

• A membrane-to-CBI approach (where CBI is “core-to-board interface,” the “core” being the mechanical block supporting the probes as shown in orange in the figure below) has probe tips that are integrated into the membrane. It supports up to 80 GHz. But the probe tips can’t be repaired, making it a less mechanically robust solution.
• Vertical MEMS-based probe tips are more mechanically stable. These cards can support up to about 10 GHz because of the multi-layered organic (MLO) construction. They provide a large probe field of about 75mm².
• A vertical-MEMS approach can be combined with a membrane. It also works, at present, up to 10 GHz, but the pins can be repaired and it supports a probe field of about 12 x 75 mm.
• FormFactor further refined the latter approach to get it to operate at 45 GHz.

Fig. 2: Four different probe-card styles and their respective bandwidths. Source: FormFactor

Since today’s development efforts don’t yet exceed the 45-GHz range, this provides sufficient coverage for much of the early 5G work. The large probing area makes multi-site testing possible, critical for maintaining high throughput.

Signals need to travel to and from the chip, and they do so via coaxial cable. A critical consideration here is inductance. It’s not necessarily the total inductance that matters, but rather what is called “residual inductance.” Some of the inductance is compensated by the capacitance that the coax shielding provides to keep a consistent, controlled impedance. So that part of the inductance isn’t an issue.

What matters is the uncompensated – hence residual – portion. That is contributed by the parts of the signal path, like the probe needles, which are not shielded. Those distances typically are measured in microns. FormFactor considers this residual inductance, or the length of uncompensated signal path, to be a good figure of merit for a probe card.

The required tests include antenna array testing, spectral measurements, RF calibration, constellations, and bitstreams. But in order to figure out what tests are most important, Bock identified areas of focus for several types of engineers.

• RF engineers are focused on insertion and return loss.
• Digital engineers are looking at eye-diagram amplitude.
• Wireless communications engineer will look at constellation diagrams and the error vector magnitude (EVM), which gives data-recovery guidance similar to that of an eye diagram. The more points in the constellation, the tighter the EVM must be to ensure adequate separation between the points.

Fig. 3: Critical areas for 5G testing. Source: FormFactor and 5G Summit

The goal for wafer test, as with any other silicon, is to optimize downstream yields while not over-testing. “We want to make sure that at the silicon level we do as much as possible functional, DC, and digital, to maximize the yields,” said Hurtarte.

But careful wafer inspection during manufacturing can pull up that yield, starting at the wafer level and proceeding all the way through the packaging level — in particular for those devices that will use advanced packaging. “5G continues to quickly propel advanced packaging designs and processes forward. With the increased investment, innovation, complexity, and various configurations, having a robust, 100% inspection and metrology process in place is important. Effectively inspecting and measuring copper pillars, wafer bumps, solder balls and bumps in advanced-packaging applications is key to preventing reliability failures,” said Tim Skunes, vice president of research and development for CyberOptics.

Subodh Kulkarni, CyberOptics’s president and CEO, reinforced that point. “The semiconductor industry is experiencing incredible transformation with the stacking of chips in advanced packaging applications. This presents unique opportunities and challenges, such as trying to find the right inspection solutions where manual processes aren’t ideal. It is critical to control their processes, ensure high yields, quality, and long-term reliability. 100% inspection and metrology is vital vs. sampling and manual methods,” he said.

Testing the RF
An emerging requirement for beamforming is over-the-air (OTA) testing, an area of focus for Hurtarte’s Semicon West presentation. “If you have a phased antenna array with 4-by-4 or 8-by-8 antenna elements and there’s no longer a way to contact them, then they have to radiate. We have to figure out ways to test this [through] the air.”

Kwan agreed. “We will see a lot of packaging of semiconductor devices that will encompass the antenna that’s being built in onto the device,” he said. “Over-the-air testing is becoming a trend in the next two to three years.”

Kwan noted that there are many testing opportunities along the manufacturing flow for OTA. “We have IC testing to be concerned about, module testing, sub-assembly, and final test,” he said. “Different tests will require different over-the-air strategies.”

OTA testing may not make sense for wafer sort. “Everyone says they want to do over-the-air testing,” said Bock. “And people have been saying we should do that at wafer test. The problem that you run into is that over-the-air works well when you have the full antenna, because that’s where it’s optimized. On the wafer, it isn’t really optimized. In addition, all of the test hardware for handling a wafer isn’t designed for that. The spacing isn’t big enough to fit an antenna.”

In addition, the test equipment for wafer test needs to be redesigned to handle OTA at wafer sort. “There’s just no way to easily do that without a total rework of the technology,” Bock said. “And most companies, with the amount of hard iron they have, that’s just a massive investment that they’re not going to want to do.”

Simpler measurements also might make wafer testing easier. “A lot of the semiconductor manufacturers don’t want to pay to be able to measure both phase and magnitude, which you need [in order] to do the beamforming test,” said Bock. “They just want to measure magnitude. They’re saying that they can get enough yield control by just asking, ‘Can we get a signal out with some amount of power?’ Even if the phase is off a little bit [due to] process variation, they can do some amount of in-situ calibration of the phase control circuitry [after packaging].”

Of particular concern here is how far away from the device the testing will be done. Hurtarte illustrated the three “fields” possible:

• Near-field reactive, which is close in enough to where physical obstructions can affect the signal;
• Near-field radiative, where an obstacle can still distort the signal; and
• Far-field, where the signal simply radiates out with a square-law power decay.

Testing of radiated signals also raises the challenge of ensuring that other stray signals don’t muddy up the test results. “How do we do the testing in a functional environment [so that] we do not have any interference from the outside, or any signals that are being propagated externally in the environment?” asked Kwan.

This suggests the use of testing chambers, but the sizes of chambers used for testing will depend on which of the three field regimes (near reactive, near radiative, and far) is to be measured. The transitions between the zones are mathematically determined, although the transitions between them are not abrupt.

Hurtarte also addressed the different production stages at which testing is required, or at least should be considered:

• Silicon testing is mandatory, in particular to ensure that basic DC and digital functional checkout will be sufficient to ensure good downstream yields.
• Modules must also be tested for “OTA continuity.”
• The final product must be OTA tested.
• It’s also likely that, at the module level, assembly and functional OTA testing will be important.

Hutarte considers sub-assembly OTA testing to be optional.

One new requirement in the test equipment is an increase in the number of RF ports to support the level of beamforming testing that will be needed. This will impact the cost of the equipment, as well as the cost of the units if other test cost savings can’t be found.

Handlers are another item that will need attention. “There are different variations of form factors today in the SiP (system-in-package) world, and we need to have handling capabilities to take care of these devices,” said Kwan.

Kwan listed a large number of items that may need testing (See Fig. 4). Costs will dictate which of these go into full high-volume production.

Fig. 4: Numerous items that need to be tested at the intermediate frequency and baseband level, the transceiver level, the front-end module, and for beamforming. Source: Advantest/Semicon West

Conclusion
All in all, test engineers will have their hands full for the next few years if high-frequency 5G equipment is going to become available at costs that the industry will be willing to pay.

“Millimeter wave instruments are very expensive,” said Kwan “They’re not produced in massive volumes.”

He said that significant effort will be required to reduce the costs of the testers. “[You can significantly move down the cost path] when you drive up multi-site testing,” he added.

Related Material
The Growing Challenges Of 5G Reliability
Rapid changes in next-gen wireless technology and standards are only adding to the complexity.
New Challenges In Testing 5G Devices
Complex hurdles in characterizing performance and optimizing signals.
What’s After 5G
The path to 6G will require some radical changes to both infrastructure and use models.
An Inside Look At Testing’s Leading Edge
FormFactor’s CEO peels back the covers on AI, 5G and HBM test issues.