Pages Menu
Categories Menu

Posted in Top Stories

Getting the Most Out of ATE Test Seconds in the Chiplet Age


This article is a condensed version of an article that appeared in
EE Times on March 27, 2024. Adapted with permission. Read the original article here.

By Ira Leventhal, VP, U.S. Applied Research & Technology, Advantest America

In the 2023 movie BlackBerry, the head of AT&T Cingular explains to the BlackBerry CEO about Cingular’s strategy to sell data plans with the newly introduced Apple iPhone: “Do you know what the problem with selling minutes is? There’s only one minute in a minute.”

As someone working for a company selling “test seconds” for semiconductor devices, this got me thinking that there’s only one test second in a second. And while our customers may accept some additional test seconds to get a device to market quickly, those seconds need to settle back down to typical levels once devices ramp to high volume—or else those additional test seconds are coming right out of our customer’s bottom line. 

In the pre-chiplet/pre-heterogenous integration world, tester resources needed to be faster and more accurate than the device-under-test. While this requirement has certainly provided significant challenges over the years, we now face the added challenge of having to be smarter than the complex, multi-chip system-under-test.

And we must meet this added challenge as 2.5D and 3D packaging are reducing direct access to device pins, and general-purpose processors are giving way to artificial intelligence (AI) processors that serve multiple specialized applications. If you don’t take advantage of AI-based approaches, the companies that do are going to take away your business.

Making it happen

How do we squeeze more than one second of value out of one test second? The data collected during those test seconds can be combined with data from across the semiconductor value chain, enabling feed-forward and feed-backward applications for optimizing design, manufacturing, and test processes.

Multiple successes have already been achieved by connecting data from two or more manufacturing or test steps and taking advantage of improvements in machine learning and edge compute technology to gain more insight from this data. Continuing to build on these successes can achieve a critical mass that will fuel further development of the enabling technologies.

I believe we are just cracking the surface in terms of the additional value that can be extracted from the data collected during those test seconds. The creation and widespread adoption of innovative approaches for extracting this value will be a key requirement for success in the chiplet age.

Read More

Posted in Top Stories

Concurrent Measurement on Wave Scale RF8 Enables Wi-Fi 7 EVM Test Improvements

By Joerg-Walter Mohr, Product Definition Expert, Advantest Corp.

The November 2022 issue of GO SEMI & Beyond included an article detailing how Advantest has evolved our Wave Scale family of test cards for the V93000 test platform by readying Wave Scale RF8 to address the test challenges associated with the forthcoming Wi-Fi 7 standard. This article provides further details on how to augment the Wave Scale RF8 solution by putting an additional splitter on the load board. With this splitter, two RF measurement systems can be used concurrently/simultaneously to improve the EVM measurement of one DUT RF signal.

The Wi-Fi 7 standard covers the now less congested frequency range between 6 GHz and 7.125 GHz, which was opened as an extension to Wi-Fi 6 and called Wi-Fi 6E. Wi-Fi 7 utilizes this wide 6 GHz Band with channel bandwidth up to 320 MHz and going up to 4096 Quadrature Amplitude Modulation (QAM) schemes.

The Wave Scale RF8 solution addresses the test challenges associated with Wi-Fi 7 by providing the frequency range and the bandwidth needed in an industry-proven instrument (Figure 1). 

Figure 1.  Wave Scale RF8 enables measurements that fully address the frequency and bandwidth expansions associated with the Wi-Fi 7 standard.

The Error Vector Magnitude (EVM) of the modulated/demodulated symbols is an important figure of merit to describe the signal fidelity. EVM is measured either in % or in dB, where dB is used for smaller values for better readability. Wi-Fi 7 standard allows device transmitters to reach an EVM of -38 dB at most. Thus, devices will have better specs and production test equipment must be able to check those, e.g. -41 dB must be checked, which means test measurement unit should provide at least an EVM test capability of -44 dB if the device has a 3dB margin. Of course, for characterization purposes it is desired to have a test measurement unit with an EVM better than -50 dB.

Figure 2.  Wave Scale RF8 concurrently measures with two RF subsystems one DUT Tx signal by adding a splitter on the load board.

Wave Scale RF8 EVM measurement performance can be improved by putting an additional splitter on the load board to enable the use of two RF subsystems simultaneously, as shown in Figure 2. By evaluating for each demodulated symbol the EVM, which is commonly detected by both RF subsystems, and neglecting the individual contributions to EVM of each RF subsystem, -50 dB EVM measurement capability is enabled for Wi-Fi 7 160 MHz 4096 QAM as shown in Figure 3.

Figure 3.  EVM concurrentMeasAB test -50 dB for highest carrier frequency of Wi-Fi 7 160 MHz. Result with and without splitter for 160 MHz plus common error.

The reduced dynamic range due to the loss via the splitter can be compensated for by averaging in cases where the focus of the test is the detection of impairments caused by non-linearities. With averaging even for Wi-Fi 7 320MHz -50 dB EVM test capability can be achieved as shown in Figure 4.

Figure 4.  EVM concurrentMeasAB + averaging test -50 dB for different frequency of Wi-Fi 7 320 MHz.

Summary

Using the high parallelism of the Wave Scale RF8 together with advanced test method addresses higher test requirements of the new standards like Wi-Fi 7 in the sub 8GHz frequency range. This might even be true for some upcoming FR3 applications in 5G/6G bands.

Read More

Posted in Top Stories

The Future of Data Analytics and Semiconductor Testing

This article is adapted with permission from a recent Advantest blog post.

By Michael Chang, Vice President & GM ACS, Advantest

The world is changing more rapidly than ever. With the explosion of Artificial Intelligence (AI), Machine Learning (ML) and data analytics, semiconductor manufacturers now have the opportunity to extract valuable insights from the massive amounts of data being generated throughout the silicon lifecycle. By leveraging AI algorithms and ML, semiconductor manufacturers can now optimize silicon design, assembly, and testing processes. It is through the analysis of these vast amounts of data, AI can quickly identify patterns, predict failures, and optimize quality. 

So, what does this all mean? It means that we now have the capability to greatly improve yield rates, reduce production costs, and accelerate time-to-market. Ultimately, the goal is to create an end-to-end utilization of analytics throughout manufacturing and test operations so that data analytics and ML will enhance the speed and accuracy of the testing process, reduce the risk of defects, and help the entire industry move ever closer to its goal of zero defects.  

This is where Advantest is revolutionizing the test industry. We just announced Advantest’s ACS Real-Time Data Infrastructure (ACS RTDI™), a solution that offers advanced analytics, including machine learning capabilities and future-proof, real-time, automated production control. The Advantest ACS ecosystem integrates all data sources across the entire IC manufacturing supply chain, a revolutionary first in the industry. In fact, ACS has been collaborating with multiple major data analytics companies as part of an industry-wide collaboration to accelerate data analytics and AI/ML decision-making within a single, integrated platform. These partnerships will help customers take advantage of new levels of data integrity and security across different test nodes and benefit from proven infrastructure solutions that will enable them to achieve new levels of operational efficiency. 

This is how Advantest is unlocking the intrinsic value of AI in semiconductor testing. 

The ACS RTDI platform integrates data sources across the entire IC manufacturing supply chain while employing low-latency edge computing and analytics in a secure True Zero Trust™ environment. This innovative infrastructure minimizes the need for human intervention, streamlining overall data utilization across multiple insertions and supporting customers’ databases. Because security remains a top concern among customers, the ACS RTDI platform has been architected to be reliable and safe, ensuring hassle-free OS revisions, while protecting data from unauthorized access or loss. This is accomplished by leveraging True Zero Trust™. Overall, the new ACS ecosystem will enable customers to boost quality, yield, and operational efficiencies, and to accelerate product development and new product introductions for years to come.

To fully support ACS’s revolutionary strategy, we also offer the ACS Solution Store which enables customers to choose from a comprehensive collection of software solutions designed for the digital age, addressing major challenges facing the semiconductor industry and that can be tailored to individual customer needs. Customers can select from the ever-expanding catalog of solutions in an easy-to-navigate browsable online catalog ― from a growing list/ team of partners joining the Advantest open solution ecosystem revolution.

Figure 1: Semiconductor Integrated Workflow and Benefits

Learn more about how Advantest is improving the technological world by setting a new standard in the semiconductor industry on our website: https://www.advantest.com/acs/overview/.

Read More

Posted in Top Stories

Solving High-Energy Testing Challenges

This article is a condensed version of an article that appeared in the November 2023 issue of Electronic Specifier. Adapted with permission. Read the original article here, p. 12.

By Fabio Marino, Managing Director, CREA, an Advantest company

Although the global semiconductor market is currently experiencing a slowdown, the automotive sector remains solid fueled by the demand for EVs. What used to be a niche market is now rapidly expanding to the mainstream, and companies that supply power IC technology must increase production volume to meet growing demand. 

Last year, CREA leased a new building to expand production capacity and keep up with ongoing business growth. This will allow CREA to produce test equipment for a wide variety of power semiconductors, including insulated-gate bipolar transistors (IGBT) and silicon carbide (SiC) and gallium nitride (GaN) semiconductors. SiCs’ advantages over traditional IGBTs include higher thermal conductivity, better ability to tolerate high voltages, higher switching speeds and lighter weight. Wide-bandgap technology such as SiC is the key to developing more efficient advanced battery systems that will enable new electric vehicles (EVs) to go farther and faster. 

Addressing high-power test requirements

Parasitic inductance and capacitance, which play an important role in the measurement, can create conditions that may damage the tester. Thus, testing high-powered SiC devices requires highly refined, specialized test equipment. CREA’s low-stray-inductance and probe card interface (PCI) technology enables engineers to minimize parasitic values. This allows the performance of specialized tests needed to ensure reliability and quality, facilitating the development of efficient batteries for new EVs. 

To meet customer demands for lower cost, CREA is expanding its bare-die test capabilities. Bare-die test utilizing the PCI and thermal control technology holds the key to expanding dynamic test to the wafer level. Package test is simpler, but if a single switch malfunctions, the entire package must be discarded. Bare-die test is more cost-efficient and creates less waste—the only challenge is that a probe card is needed to perform the test. Probe cards are fragile, and the high amount of energy generated during dynamic test can break the probe card and damage the tester itself.

CREA’s PCI technology monitors each probe needle for abnormal current distributions, shutting off the tester when such an abnormality is detected to prevent damage. CREA also developed a chamber for bare-die test that moderates temperature by controlling airflow to prevent sparking that can occur while working with high voltages, ultimately reducing the threat of harm to the ATE.

SiC technology provides many benefits over traditional IGBT technology, as noted earlier. While many major semiconductor companies are investing in R&D to support SiC technology, SiC is very different from silicon wafer technology. It requires completely different equipment, and the automated tools that factories currently have are designed for silicon wafers and will not work with SiC. Because SiC is a maturing technology, production yields are low. This creates a significant opportunity for test companies to deliver SiC-optimized test equipment. CREA continues to refine its power IC testing technology to increase yield and help customers maintain sustainable business models that can keep up with rising demand.

Conclusion

Fueled by major investments from the global semiconductor community, the power IC industry is evolving quickly. This creates significant opportunities for companies like CREA that are building solutions to address high-power specs and overcome industry challenges.

CREA’s patented LSI™ and PCI™ technology provides specialized testing solutions for power ICs found in hybrid and EV automotive engines. These solutions will accelerate the shift from 400V to 800V batteries, accommodating the testing specs needed to develop cutting-edge EV technology. Today, CREA engineers are developing techniques to run high-energy tests in parallel – increasing yield and helping to accommodate rising global demand for SiC and other advanced power semiconductors. 

 

CREA’s Power Device Testers

Read More

Posted in Top Stories

Improving Debug Time and Test Coverage with Parallel Validation Strategies

This article is excerpted from an article that appeared in the July/August 2023 issue of Chip Scale Review. Adapted with permission. Read the original article here, p. 28.

By Adir Zonta, Product Marketing Manager, V93000 Engineering Solutions, Advantest

Test data volumes are exploding as the number of transistors per chip increases along with the number of test vectors needed to test each transistor. A recent article [1] described how traditional methods of device validation and characterization, ATE structural and functional test, and system-level test no longer suffice due to increased device complexity, and introduced innovations in pre-silicon verification, first silicon bring-up, and post-silicon validation (PSV) that are necessary to meet today’s challenges.

This article looks further at these innovations and the systems necessary to implement them, including how to equip an engineering lab with automated parallel test stations to speed up test engineering tasks such as pattern validation. It also describes how a new standard helps bridge the gap between electronic design automation (EDA) and ATE and how Cadence and Advantest have collaborated on an initiative to put the standard into practice.

Test-pattern validation

One of the challenges that the explosion in test data imposes on test engineering is the ever-lengthening time required for test-pattern validation, which is impacting time to market. Test-pattern validation determines whether the patterns are generated correctly, that the expected responses are accurate, and that they have enough margin to account for parameter variations (for example, in voltage and frequency) in production.

Generating test patterns 

The test patterns include structural scan patterns generated by automatic test-pattern generators or functional test patterns generated manually from a test specification or automatically using random or constraint-based test-generation methods or other techniques linked with EDA tools. Test patterns from the EDA tools are generally in a standard format such as STIL (Standard Test Interface Language) or WGL (Waveform Generation Language).

Structural test patterns target specific fault models, such as “stuck at” faults or timing faults, whereas functional test patterns confirm the performance of the device under test (DUT) in its end use. Functional test vectors are particularly important in automotive and other industries where performance and safety are critical. The key aspects of generating test patterns are summarized below.

Cyclized test vectors. The patterns in STIL or WGL from EDA tools are converted to cyclized test vectors for the target ATE system by adding timing and control information to synchronize the patterns with a specific ATE system’s clock and control signals. This can require extensive development time.

Error causes. Inevitably, the cyclized test vectors will experience errors resulting from design defects percolated through the cyclization process, the cyclization process itself, or corner cases that the original design did not take into account. Regardless, the PSV process must identify and correct any errors.

Correcting test-pattern errors. When errors are detected during the pattern validation process, they must be corrected through manual or a combination of manual and automated methods.

Automated parallel test stations speed up the process

Speeding up test pattern generation requires a test lab with the equipment necessary to run parallel pattern validation, minimizing the time spent on pattern debugging while assuring sufficient test coverage. A solution such as the Advantest V93000 EXA Scale EX Test Station, an engineering platform for complex device bring-up that supports structural and functional test, provides this parallel test capability without requiring a lot of floor space because it is designed to fit under the company’s single-site M4171 automated handler. Complete with integrated active thermal control (ATC) over a range of -45 to +125°C, the handler brings automated device loading, unloading, and binning into the laboratory environment. As shown in Figure 1, six test cells can fit within a 5m by 5.5m laboratory space.

Figure 1: Six EX test stations with M4127 handlers in a 5m by 5.5m laboratory space can speed up test-pattern validation and other engineering tasks.

There are two primary challenges involved in creating functional test on ATE:

  1. The need to convert the functional test content into a production test vector pattern, which requires tooling and extensive development time.
  2. Typically, there is no native software debugging environment on a typical tester, making it very difficult for the test case developer to debug any issues in support of the test engineer. Excessively long, unpredictable debug cycles are inevitable.

Pre-silicon methodologies and an ATE instrument can work together to seamlessly and interactively validate the functional test content to help to meet these challenges.

PSS links EDA and ATE

Reuse of pre-silicon verification test content can ease the transition from the pre-silicon verification stage to first silicon—including bring-up, bare-metal test execution, and ATE stage. To that end, the Accellera Systems Initiative, an organization focused on the creation and adoption of EDA and intellectual property (IP) standards, has developed the Portable Test and Stimulus Standard (PSS), which specifies a single representation of stimulus and test scenarios that span simulation, emulation, and post-silicon [2].

PSS enables the once-siloed EDA and ATE disciplines to work together. However, while structural test dominates the ATE side, rising quality expectations are driving a need for more functional test to ensure the chip will perform properly in its end-use mode. However, as previously mentioned, converting functional test content into production test vectors requires extensive development time, and a typical ATE system lacks a native software debugging environment that could speed up the process [3].

Joint EDA-ATE PSS implementation

A joint cooperative initiative between Cadence and Advantest involved a combination of the PSS and HSIO approaches. The companies have developed a solution that involves PSS-based test content creation, an interface to ATE software, the loading of parameterized test content, test execution on ATE hardware, and debug and analysis (Figure 2). The Cadence Perspec System Verifier automates the process of extending the PSS models used in pre-silicon validation to the ATE environment, reducing the complex use-case scenario development time. A container file labeled FDAT in Figure 2 provides an efficient interface between Perspec and the Advantest SmarTest 8 software for its V93000 ATE systems. 

Advantest’s Link Scale ATE instrument interacts natively with the DUT using low pin-count HSIO, such as USB and PCI Express interfaces running in full-protocol mode, without pattern cyclization. Collected test traces can be viewed in a SmarTest viewer or imported into Cadence’s Verisium Debug AI-powered debug tool for correlation with the original PSS tests. In addition, Link Scale can host embedded software debuggers such as the Lauterbach TRACE32.

Figure 2: PSS enables interfacing of EDA and ATE to optimize test validation.

Device validation best practices

Going forward, one key will be smoothing the transition from the lab environment with engineering test stations to the production floor. The single-load-board strategy, in which a multisite load board for high-volume production can be used in the lab with only a single site enabled, makes it unnecessary to develop one board for engineering activities and another for high-volume manufacturing (HVM).

The engineering environment should be as close as possible to the HVM environment. The EX Test Station achieves this goal because it uses our Xtreme Link technology, designed to provide high-speed optical data connections, embedded computing power, and card-to-card communications for high-volume production ATE. The station is also suitable for testing initial engineering batches efficiently. In addition, it helps to ensure seamless flow between the engineering and HVM environments.

Conclusion

The semiconductor industry has a long and successful history of testing increasingly complex devices, continually enhancing structural, functional, and system-level test to minimize test escapes. Advances continue as the industry contends with an exploding amount of test data necessary for silicon bring-up, PSV, and other test engineering tasks. A key innovation is a laboratory equipped with engineering workstations that can operate in parallel to speed up tasks such as pattern validation. In addition, EDA and ATE companies are cooperating to leverage standards such as PSS to bridge the pre- and post-silicon verification stages, and they are leveraging HSIO to allow ATE to apply test patterns without cyclization. Finally, engineering workstations are incorporating the load-board, compute, and communications technologies of production ATE systems, thereby speeding the transition from the lab to HVM.

References

  1. D. Armstrong, “Device validation: the ultimate test frontier,” Chip Scale Review, Nov-Dec. 2022, p. 26.
  2. Accellera Board Approves Portable Test and Stimulus Standard 2.0,” Accellera Systems Initiative, April 14, 2021.
  3. M. Rubin, A. Zonta, Pre and Post-Silicon Verification Have Never Been Closer! Leveraging Portable Stimulus for Automatic Test Equipment (ATE),” Cadence Design Systems Inc., May 4, 2023.

 

Read More

Posted in Top Stories

Deploying Cutting-Edge Adaptive Test Analytics Apps: Innovation Based on a Closed-Loop Real-Time Edge Analytics and Control Process Flow into the Test Cell

By Ken Butler, Senior Director of Business Development, Advantest, & Guy Cortez, Senior Staff Product Manager, Synopsys

Semiconductor test challenges abound in this era of AI. As such, semiconductor test engineering is increasingly moving towards fully adaptive test where each device receives the “right” test content to assess its correctness. Advantest and Synopsys have partnered to provide new cutting-edge real-time adaptive test applications at the test cell based on complete closed-loop analytics and control process flow. Our solution leverages a high-performance, highly secure real-time data infrastructure combined with advanced analytics derived from a comprehensive silicon lifecycle management (SLM) platform. For example, when test measurement data is combined with on-die sensor readings using a very fast and secure computing platform, the solution provides an in-situ adaptive test with milliseconds latencies.

Figure 1. Semiconductor test challenges for the AI era

First, let’s review the Advantest ACS portion of the solution. The Advantest ACS Real-Time Data Infrastructure (RTDI) is a platform that provides low latency and highly secure data access and system control for test operations (Figure 2). It consists of the following components:

  • ACS Container Hub™, a web registry for the management and distribution of open container initiative-compliant AI/ML and statistical workloads.
  • ACS Unified Server™, a multi-purpose, reliable and scalable platform that serves as a gateway and local mirror for containerized applications.
  • ACS Edge™, a high-performance and highly secure computing solution for the execution of complex analytical workloads for real-time applications in production test.
  • ACS Nexus™, the communications backbone for ACS RTDI which allows for streaming access to test and test cell data as well as real-time control of tester operations.


Figure 2. ACS Real-Time Data Infrastructure

Two example use cases to which adaptive test flow can be applied are:

  1. Reduce test time – Save cost and improve throughput by eliminating some tests that appear useless (no parts failing).
  2. Reduce DPPM – Improve quality control by adding additional tests for some “risky” parts.

Our first use case focuses on this second method of improving quality. Figure 3 below is an image of a stacked wafer map that highlights specific zonal regions on the wafer that have an excessive number of failures, shown in purple. A typical stacked wafer map consists of 25 wafers or one lot’s worth. The remaining good die in this region are suspect due to the amount of failures on this part of the wafer. The larger and darker the purple identifier per x, y coordinate is, the more prominent failures there happen to be at that x, y coordinate across all wafers analyzed.

An application is available to identify which packaged die from select x, y coordinates are considered risky based on a specific failure threshold set by the product engineer. Additional test(s) will later be applied to those parts labeled as risky during final test (a form of ZPAT at final test), thereby improving the overall quality of the chip but with minimal impact on total test time.

Figure 3. Wafer stack map

A second application is adaptive limit setting, i.e., the adjustment of test limits during test program execution. The method shown here utilizing sensor data provides higher accuracy in limit management compared to existing methods such as dynamic part average testing (DPAT), because sensors embedded in the chip provide additional key information that enables the monitoring of the chip’s operational metrics such as power and performance. This example highlights the use of sensor data that characterizes the process and environment information to enable more accurate limits on speed/power consumption during testing, thus resulting in lower DPPM and higher quality.

Figure 4 below shows a comparison of the two adaptive test limit approaches. First, the DPAT method shown is a standard univariate approach based on the die population results for a given test. Next, the sensor-aware method incorporates a bivariate correlation between the data measured from sensors and the results of a specific VDD consumption test. The second method can identify at-risk die that would be missed by conventional DPAT analysis.

Figure 4. Traditional univariate DPAT vs. sensor aware bivariate method

Conclusion: In this article, we describe a real-time, highly secure data infrastructure plus a pair of complex, high-value analytical applications that consume both test response and on-die sensor data to produce inferences for true adaptive test decision-making with low milliseconds latencies. The analytics and associated applications are available as part of an open solutions ecosystem, which allows users to either develop their own solutions or procure and deploy them from Synopsys or other providers. The result is the democratization of machine learning driven applications, making them available to everyone in the semiconductor test community.

Related Links:

The Advantest ACS Solution Store

Advantest ACS

Advantest Talks Semi Podcast

Read More