High-Volume Production Test of AiP Modules for 5G Applications

This article is a condensed version of an article published in the May-June 2020 issue of Chip Scale Review, p. 20. Adapted with permission. Read the original article at http://fbs.advantageinc.com/chipscale/may-jun_2020/22/. 

By Jose Moreira, Senior Staff Engineer, SOC R&D, Advantest

The arrival of 5G promises enhanced mobile broadband (eMBB), massive machine-type communication (mMTC), and ultrareliable low-latency communication (URLLC). But as 5G rolls out, the test community faces challenges and opportunities. That’s particularly true regarding the antenna arrays that will connect handsets to base stations.

5G New Radio (5G NR) defines two ranges, frequency range 1 (FR1) and frequency range 2 (FR2). FR1 includes the sub-6-GHz frequencies in use for previous generations of cellular technologies, but FR2 opens up mmWave frequencies above 24 GHz for 5G deployment. 5G NR leverages the FR2 frequencies to achieve larger modulation bandwidths (for example, 800 MHz). However, the high transmission losses at these frequencies require the use of antenna arrays for multiple-input and multiple-output (MIMO) functionality and to focus the transmission beam (beam forming) in both the base station and the consumer’s handset. These arrays come in the form of antenna-in-package (AiP) modules.

For the handset, these AiP modules will usually have an array of dual-polarized patch antennas for top firing and, in some instances, an array of dipole antennas for side firing (Figure 1). 

Figure 1. This example of a generic antenna array module comprises 12 dual-polarized patch antenna elements and seven dipole antenna elements.

To minimize RF losses to the antenna radiators, the AiP modules include RF integrated circuits that provide modulated mmWave signals to the AiP antenna array with the needed gain and phase to each radiating element. The modules then usually only require power, digital control signals, and modulated intermediate-frequency (IF) signals.

AiP modules for 5G handsets must be small to fit into the modern cellphone form factor, and multiples of them need to be used in a single cellphone because the user’s hand position has a significant impact on the transmitted beam loss. Also, the AiP modules in a cellphone might not be all equal, but in fact have different antenna configurations depending on the handset design.

Regardless of configuration, these AiP modules must be tested. The 3GPP standard defines three methods for the over-the-air (OTA) standard compliance testing of AiP modules: direct far field, indirect far field, and near-field to far-field transformation. Each of these methods have advantages and disadvantages, but they all require relatively large test chambers and a complex manipulator to rotate the AiP device under test (DUT) or the measurement antenna.

These methods are neither practical nor necessary for high-volume production testing, where the objective is to check the functionality of the antenna under test (AUT), not its compliance. Low cost of test is critical because most of the end applications are consumer oriented. Also, to keep costs down it is important to reuse as much as possible the test-cell infrastructure already deployed for testing RF integrated circuits.

This paper presents three possible options for the production OTA testing of AiP modules with ATE: far-field testing, radiating near-field testing, and reactive near-field testing. The book Theory and Practice of Modern Antenna Range Measurements¹ provides details on the transition from the near field to the far field. For the purposes of this discussion, it suffices to note that there is no hard boundary between near and far field, but a continuous transition where the radiated waves become locally more planar as they propagate away from the radiating antenna. From an antenna-measurement perspective, the far-field region is the best because the radiated waves are locally planar, and the measurement antenna is too far away to have an impact on the AUT. 

But the far-field distances also imply large dimensions for the measurement setup, and radiating and reactive near-field testing approaches offer more compact alternatives.

Figure 2. These examples of OTA ATE far-field measurement setup show a motorized linear stage (left) and a static setup (right).

OTA far-field testing

Figure 2 shows two examples of a simple far-field measurement setup on an ATE system. This approach is excellent for an initial start with OTA testing on ATE because one can start in the safety of the far-field measurement range while doing correlation and debugging of the AiP DUT using the ATE system. Calibration on a far-field setup is also trivial using standard antenna measurement calibration procedures1. The problem arises when considering high-volume production by integrating a far-field OTA methodology on a standard ATE test cell.

The mechanical dimensions required for a far-field OTA test solution prevent the usage of standard ATE test-cell commercial handlers, thereby requiring custom robotic handlers and creating additional costs. Cost reduction through multisite implementation on ATE is also nontrivial with a far-field OTA ATE implementation.

OTA radiated near-field testing

One approach to integrating an OTA measurement setup into a standard ATE test cell is to move the measurement antenna into the radiating near-field region. Figure 3 shows low-cost radiating near-field test sockets for a patch-type antenna array AiP. In this example, the measurement antenna is 11 mm from the DUT AiP antenna array. A radiating near-field antenna test has the advantages of easy integration within a standard ATE test cell along with easy multisite implementation.

Figure 3. These examples of low-cost radiating near-field OTA sockets support manual ATE-based OTA testing.

Because in a production test environment the objective is to identify failed AiP modules and not to characterize them, one could assume that there would be some easy correlation between good AiP modules tested in a far-field setup with failing AiP modules tested on a near-field setup, assuming a comprehensive list of performed tests. This is a valid thinking, but one needs to be aware of two important drawbacks on a radiating near-field measurement setup. The first is that the measurement antenna is now so close to the AiP DUT antenna array that it will have an impact on the DUT AiP antenna elements (antenna detuning) and can even result in a standing-wave effect.

The second drawback is shown in Figure 4. 

Figure 4. In this example, the distance from the measurement antenna to the different antenna array elements on an AIP DUT differs.

Because only one measurement antenna is used, depending on the DUT AiP antenna array geometry, the distance of each DUT antenna array element to the measurement antenna will be different. This can have a significant impact on a worst-case scenario^2,3. Finally, calibration in the radiating near-field is nontrivial. If a golden-device calibration is used, results are critically dependent on the golden device’s performance, and absolute measurements are not possible.

OTA reactive near-field testing

An alternative is to measure the DUT AiP antenna array in the reactive near field. In this case, a classical measurement antenna cannot be used because in the reactive near-field range it would have a dramatic effect on the DUT AiP antenna elements. To measure on the reactive near field, the antenna or probing element needs to be very small. Figure 5 shows one reactive near-field probing concept for OTA ATE that has been patented by Advantest using two very thin parallel needles to probe the electric or magnetic field on the DUT AiP reactive near field. The main advantages are that each element of the DUT AiP array is individually measured (power and phase) and that the probe size is very small to minimize the disturbance of each radiating element. This concept is explained in more detail in other papers^4,5. 

Figure 5. This near-field probing concept for OTA ATE uses two very thin parallel needles to probe the electric or magnetic field on the DUT AiP reactive near field.

Figure 6 shows an example of a prototype reactive near-field socket3. Here, measurement of a dual-polarized 2×2 AiP array results in eight individual signals. To keep ATE resources to a minimum, a solid-state relay switches each of the antenna/polarization signals in series to the ATE measurement instrument. A parallel measurement approach is also possible but requires eight ATE measurement instruments. The optimal setup will depend on a detailed cost-of-test analysis.

Figure 6. In this prototype reactive near-field socket, a dual-polarized 2×2 AiP array is measured resulting in eight individual signals.

Summary

For OTA testing with ATE of AiP modules, there is no right or wrong answer. Depending on the testing requirements and testing stage (for example, initial ramp-up or mature high-volume manufacturing), the OTA test strategy might be different. Figure 7 shows a high-level comparison of the different OTA test strategies presented in this paper.

In a future paper we will use a custom-designed 28-GHz 2×2 path antenna array in a 0.4-mm-pitch BGA package to compare the different approaches in terms of OTA measurement results with the Advantest V93000 Wavescale Millimeter CardCage ATE system.

Figure 7. This chart shows the advantages and disadvantages of three OTA test strategies in an ATE environment.

ACKNOWLEDGEMENTS

We would like to thank Natsuki Shiota, Aritomo Kikuchi, Hiromitsu Takasu, Hiroyuki Mineo, Sui-Xia Yang, and Frank Goh from Advantest for their support and collaboration on the OTA project development. We would like also to thank Prof. Jan Hesselbarth from the University of Stuttgart.

REFERENCES

1. Clive Parini, et al., Theory and Practice of Modern Antenna Range Measurements, IET, 2014.
2. Jose Moreira, Jan Hesselbarth, and Krzysztof Dabrowiecki, “Challenges of Over The Air (OTA) Testing with ATE,” TestConX China, Shanghai, October 29, 2019.
3. Natsuki Shiota, Aritomo Kikuchi, Hiroyuki Mineo, Jose Moreira, and Hiromitsu Takasu, “Socket Design and Handler Integration Challenges in Over the Air Testing for 5G Applications,” TestConX 2020, May 2020.
4. Jan Hesselbarth, Georg Sterzl, and Jose Moreira, “Probing Millimeter-Wave Antennas and Arrays in their Reactive Near Field,” 49th European Microwave Conference, 2019.
5. Utpal Dey, Jan Hesselbarth, Jose Moreira, and Krzysztof Dabrowiecki, “Over-the-Air Test of Dipole and Patch Antenna Arrays at 28 GHz by Probing them in the Reactive Near-Field,” To be presented at the 95th ARFTG Microwave Measurement Conference, August 6, 2020.

Wave Scale RF8 – Enabling the Next WAVE in RF Communications Test

By Dieter Ohnesorge, Product Manager, RF Solutions, Advantest Corp.

Over the years, Advantest has remained at the forefront of test innovation through close collaboration with customers and partners, and by keeping our finger on the pulse of industry and market trends. We launched the Wave Scale family of test cards for our V93000 system-on-chip (SoC) test platform just over four years ago, and in that short time, we have greatly expanded the line with new products designed to meet burgeoning test demands – e.g., Wave Scale RF, Wave Scale MX, and Wave Scale Millimeter.

The test economics of state-of-the-art smartphones, tablets and routers demand highly parallel RF test. Now, we are addressing this next wave in RF communications test, enabled by Wi-Fi 6E, operating in the 6GHz band and coming up to 7.125GHz. This forthcoming update to the Wi-Fi standard will extend the features and capabilities, including higher performance, lower latency, and faster data rates—for this higher band. Our new Wave Scale RF8 card enables parallel test capabilities for Wi-Fi 6E, as well as for 5G-NR transceivers, LTE-Advanced Pro and internet of things (IoT) devices. 

The extension to Wi-Fi 6E will make available 1200MHz of additional bandwidth in the unlicensed frequency spectrum (see Figure 1). Compared to these 1200MHz, the 2.4GHz band has just three non-overlapping channels with a total bandwidth of 60MHz, and is already crowded with multiple users competing for bandwidth. Even with 25 channels and an additional 500MHz of bandwidth, the 5GHz band gets filled up quickly – a problem that has become even more apparent with many people in close proximity tapping into Wi-Fi to work in from home or attend school remotely. 

Figure 1. The chart illustrates the importance and benefits of Wi-Fi 6E. Wi-Fi at 2.4GHz uses 60MHz total bandwidth with up three non-overlapping channels, while the 5GHz band adds 500MHz and up to 25 channels. Wi-Fi 6E now adds a massive 1200MHz to the existing Wi-Fi bandwidth, adding up to 59 additional channels.

With the 6GHz band, however, comes added channels and a substantial extension of 1200MHz in usable bandwidth. Moreover, all three bands can be used simultaneously – e.g., users can read and send email in the 2.5GHz band, place Wi-Fi calls in the 5GHz band, and download streaming content in the 6GHz band. Good news for users, this nevertheless creates new challenges with respect to testing these communication devices.

Wave Scale RF 8 is capable of both highly parallel multisite and in-site parallel testing, providing a new dimension of test coverage and economics. Testing both the send and receive channels takes a fraction of the time that would be required using a traditional test flow, and it can perform high multisite testing using native ATE resources, all within the V93000 test head. Advantest is the first in the industry to enable such high multisite parallelism for these applications, providing an unmatched test time benefit (see Figure 2).

Figure 2. The benefits of massively parallel test are illustrated here. Stacking tests test in a parallel test flow rather than one-by-one serial test drastically reduces test times; parallel mission-mode tests are reduced by up to 50%.

The card’s RF-optimized architecture comprises four complete RF subsystems to achieve high-throughput testing. Within each subsystem is an independent modulated source, a waveform generator/digitizer, scattering parameters, and a test processor that can make multiple RF measurements in the shortest possible time. Each card includes 32 RF ports, true parallel stim measurement ports, and – as mentioned earlier – operates at up to 8GHz with a modulation bandwidth of 200MHz. Its wide-frequency capability is a vital aspect of Wave Scale RF8 – the Wi-Fi6E standard can actually go up to 7.125GHz, so is well covered by the 8GHz capability of the Wave Scale RF solution.

Continually staying ahead of the industry curve is an important aspect of Advantest’s brand promise to our customers. We focus on having the solution in place that customers will need in order to adapt to new test requirements. With Wave Scale RF8, we have made sure that we can accommodate the massively parallel testing that advanced communications devices demand. With multiple independent subsystems in a single card, Wave Scale RF8 delivers the cost-efficient production solution for next-generation Wi-Fi 6E and cellular devices.

System-Level Test Methodologies Take Center Stage

By Fabio Pizza, Business Development Manager, Advantest Europe

Note: System-level test (SLT) continues to expand in importance throughout the industry. In past newsletters, we have published articles looking at the company’s efforts in this space, primarily for the storage market, as it offered the most immediate opportunity for implementing SLT. Now, rising industry demand, driven by mission-critical applications, has put SLT at the forefront for Advantest company-wide.

Because electronic systems for all applications in end-user markets must provide the highest possible reliability to match customers’ quality expectations, semiconductor components undergo multiple tests and stress steps to screen out defects that could arise during their lifecycle. Due to new semiconductor devices’ increasing design complexity and extreme process technology, increased test coverage is needed to meet stricter quality requirements.

To solve this problem, system-level test that mimics a device’s real-world operating conditions is increasingly being adopted as one of the final steps in the production testing process for complex systems-on-chip (SoCs). In the past, system manufacturers typically implemented SLT on a sample basis, plugging devices into systems to check that the devices would function in an application. Semiconductor companies have now adopted SLT methodology throughout the test process to increase test coverage and product quality for mission-critical applications (Figure 1).

Figure 1. Advanced technology is driving changes in test requirements, creating the need for integrated SLT approaches throughout the test flow.

Advantest provides customers with an end-to-end test solution, from ATE to SLT, in line with the company’s Grand Design, created to ensure that Advantest remains at the forefront of our industry. The central vision of this corporate-wide plan is for Advantest to strengthen its contributions to customer value in the semiconductor business by enriching, expanding and integrating our test and measurement solutions throughout the entire value chain, as shown in Figure 2.

Figure 2. System-level test is crucial to the mission of Advantest’s Grand Design – “Adding Customer Value in an Evolving Semiconductor Value Chain.”

Recent market and financial analyst commentary supports Advantest’s view that SLT is the way of the future and that our expertise in this area provides new growth opportunities. Following our briefing on SLT in June, VLSI Research CEO Dan Hutcheson wrote in the July Chip Insider newsletter that the session prompted him to think that SLT “may well be the next major revolution in test equipment…The essential argument is that test is becoming a more important enabler going forward versus its decades-long position as a cost center to be pushed down. What has changed is the increasing complexity of SoCs and SiPs, the introduction of advanced packaging, chiplets and high-bandwidth memory.”

A July report issued by Mitsubishi UFJ Morgan Stanley Securities noted, “Recently, we have seen an increase in demand for the testing of semiconductor devices at the system level, in addition to the wafer and package levels, as temperature and voltage fluctuations place them under severe stress when they are used in applications such as data center servers. There is similar testing demand from the makers of storage and mobile devices and automotive systems, and we believe this will provide a fresh source of growth for Advantest.”

The mega-markets shown in Figure 3 represent mission-critical applications for SLT. Advantest has established itself as a leader in SLT solutions for the computing, memory and storage, and mobile markets, with systems in production performing massively parallel SLT for these applications, and we continue to sustain and grow our leadership in these areas. The automotive space is a new domain where we are now focused on expanding our SLT business.

Figure 3. Memory & storage, computing, mobile and automotive markets are the four mega-markets driving system-level test.

We are already working with leading customers in Europe, the U.S. and Japan who are seeking automotive SLT solutions, primarily for advanced driver-assistance systems (ADAS) and infotainment. One customer developing automotive microcontrollers is experiencing some returns from the field that were not detected in the standard traditional final test steps. They must expand test coverage to close these gaps. Unlike with mobile phones, one failure per million devices can be disastrous or even deadly in the automotive space, so chipmakers must be able to ensure the quality of their devices when installed. Quality over time is particularly important, as the final product lifetime can be 10 years or more.

Advantest’s SLT capabilities 

Advantest SLT test cells are based on modular building blocks, as shown in Figure 4. The first step involves collaborating with the customer to develop a customized application board to ensure accurate reproduction of the system environment’s conditions while optimizing for high volume production. Next comes automation, the degree of which differs, depending on target production test time and required parallelism. High-volume devices require a much greater amount of parallel testing to meet cost-of-test objectives.  

Figure 4. Advantest’s SLT approach involves modular test-cell building blocks.

The third piece is the thermal environment, which depends on device power and test stress requirements. As the figure indicates, Advantest offers a range of thermal-control technologies: pure passive ambient, tri-temperature active thermal control (ATC) with air cooling, and tri-temp ATC with liquid cooling using rapid temperature switching methods (RTS). Devices are tested independently at controlled temperature. As newer-generation devices tend to consume high power, each needs its own thermal controller and sensors to ensure stable test temperature and prevent device failure. Examples include HPC devices, which can consume over 300W each. ADAS applications require a great deal of power to process data generated by vehicle cameras. When tested, these automotive processors must be heated up without exceeding the maximum junction temperature of 125-130 degrees.

Our SLT solutions also share a common software framework called ActivATE™, which enables test programs to be reused easily. ActivATE™ comprises an integrated development environment (IDE), a test sequencer, and a device manager, and allows test engineers to rapidly create and deploy test programs using standard programming languages.

These building blocks have been assembled by combining our existing proven SLT offerings with some strategic acquisitions. In late 2018, the semiconductor test division of Astronics became part of Advantest, adding massive parallel test solutions to our arsenal. Parallel testing is essential for minimizing the cost of test for SLT, as is mitigating handling limitations of pick-and-place technology. Astronics developed systems with slots that can test hundreds of devices in parallel with virtually 100-percent multi-site test efficiency.

This is a must-have for high-volume manufacturing of mobile and high-performance computing (HPC) products. While automotive volumes are not as high, the electronics in cars are increasing, so here, the requirement is covering multiple variations of devices – i.e. a main design with some customization. This requires the ability to test more small lots with diversified packages and variations of a main device family, and we can now handle these different packages and fully parallel-test them in one system. 

Exemplifying our building-block approach, we developed in less than one year the 5047, our dedicated SLT test cell consisting of our standard M4841 logic handler docked to a 547 SLT system to perform SLT for lower-volume automotive devices with limited parallelism requirements (x8 or x16). These devices run at low power with short test times (tens of seconds to a few minutes), so the standard pick-and-place handler can cover them satisfactorily. Its tri-temperature thermal environment (-55 to +155°C) supports both hot and room temps; cold temps require some further design accommodation for condensation abatement. 

This past January, we also acquired Essai, Inc., adding its test sockets and thermal-control units to our portfolio. The same macro trends pushing processors to higher speed, higher power and higher complexities demand that our SLT platform be tightly integrated with the socket design.   We are currently integrating Essai’s offerings into our end-to-end solutions and will soon be able to offer SLT test cells with socket-accuracy and performance assurance.

Figure 5. Advantest is uniquely qualified to provide all aspects required for high-volume SLT.

As SLT demand becomes more widespread, it is an exciting time to be part of the test industry.  As Figure 5 depicts, Advantest is uniquely situated to provide our valued customers SLT cells with the right communication protocols, power, automation, active thermal control and worldwide service and support.  We look forward to continuing to share our progress in further building this already-vital part of Advantest’s business.