Evolution of MEMS Devices and Effect on Packaging and Test

By Pradeep Chakraborty & Aanchal Ghatak

MEMS has been driving innovation at the silicon level, and, in turn at the packaging and testing of devices. Here, Pradeep Chakraborty and Aanchal Ghatak interview key players who discuss the evolution of MEMS devices, and how they are changing packaging and test strategies.

According to Stephen Whalley, strategic advisor, SEMI-MEMS & Sensors Industry Group (SEMI-MSIG), MEMS devices are the driving force in system innovation across multiple markets, such as smart transportation and vehicles, smart cities and buildings, smart agriculture, smart medical, and the Internet of Things, etc.

“This explosive growth in MEMS is also driving innovation at the silicon level, and consequently, in the packaging and testing of the devices,” he explained. “The vast majority of MEMS devices require a cavity of some sort, which can be categorized as vacuum, hermetic at various atmospheric conditions with inert gas, or open cavities that sense the surrounding environment (examples are acoustic devices and chemical sensing devices).”

Allyson Hartzell, managing engineer, Veryst Engineering, elaborated: “Also, coupled in many packages are other die. Examples are ASICS and other MEMS devices. Innovative methods for creating the cavities by chip-to-chip solder and metallic bonding keep a ‘lid’ on the thermal budget of sensitive transistors on ASICs and other CMOS devices that are housed in the MEMS packaging.

“When it comes to standard packaging aspects, wire bonds and solder bumps are still popular options. Capping with sealing materials, such as glass and metals/solders are also used. TSVs are now commonplace.

“Wafer level packaging (WLP), over molding into a JEDEC configuration, SOIC, and high-reliability metal and ceramic packaging are varied, but also used in semiconductor packaging. Therefore, testing strategies have a starting point through the semiconductor industry.

“As packaging becomes smaller and smaller, the cavities are smaller, and so, their internal environment is harder to control. Thus, testing requires more steps than the electro-mechanical testing of yesteryear’s single chip inertial packaged devices. MEMS process development requires cavity environmental gas testing, as nano-liter cavities are not easily tested by earlier methods such as the residual gas analysis (RGA). RGA could go by the wayside in lieu of humidity and pressure sensors patterned into the nano cavities themselves.

“Traditional getters are no longer easily fitting into tiny cavities. MEMS process development includes getters patterned right into the cavities itself. Gettering in small cavities that are not vacuum cavities could be required to reduce the moisture that causes MEMS device drift.

“Yield and reliability testing strategies are different as the cavity size reduces, as more chips (including MEMS) are integrated into a package, and handling the tiny packages during testing to keep proper orientation, without mechanical shock exposure, has to be designed into the test protocol. Packaging with interposer type methods and TSVs can use fan-out test strategies, already developed in semiconductor testing, for testing in a format that mirrors WLP and JEDEC type package dimensions and solder balls.”

“Yet, for some parts, such as environmental sensors, using methods such as tape-on-reel must be vetted for adhesive outgassing that can tie up surface states and reduce sticking coefficients during the operation of the devices. Testing of environmental devices requires some controlled testing environments with analytical chemistry characterization methods incorporated into the test platform. The devices drive the test strategies, and cost is reduced if the standard packaging outlines and the electrical contacts are designed from the beginning.”

John Stabenow, director of Marketing at Tanner EDA, part of Mentor, a Siemens business, said: “Multi-MEMS devices, as part of system-in-package solutions, are becoming more main stream. SiP devices allow edge device designers a better plug-and-play opportunity for complex subsystems that include not just the sensing (MEMS) technology, but, as well as MCUs, RF and power management components in a single unit. The modular design of a SiP solution allows for faster time-to-market, and more predictable design outcomes.”

Mary Ann Maher, CEO and founder, SoftMEMS, noted: “Key issues for packaging and test are the need to support multiple sensors in the same package. IC-typical applications require multi-sensor fusion. MEMS on flexible substrates is driving new packages and test strategies. MEMS on flex must be tested under bending and flexing.”

Adrian Arcedera, senior VP, MEMS, Sensors & WB BGA Products, Amkor, elaborated: “The growth of MEMS devices — from automotive and industrial markets into the consumer market — drove the need for shorter package development time, and the steep ramp up to mass production, while continuing to require the management of packaging stresses to the MEMS structure. These market changes ushered the need for standard packaging platforms.”

Gerard John, senior director, Advanced Engineering, Amkor, added: “Test strategies change and evolve at the same rate as the packaging strategies. From a test perspective, the accuracy and performance of low-cost MEMS devices are coming close to those of expensive counterparts. This calls for test stimulus that can provide 10x higher accuracies than the device under test. To reduce the cost of test (CoT), test strategies include, increasing the sites tested in parallel, reducing the number of test insertions, and designing multi-functional stimulus.”

Opportunity for OSAT companies
Certainly, the OSATs (outsourced semiconductor assembly and test) providers, must have a role to play here. OSAT companies have huge opportunities to provide standard packaging outlines and testing. What is inside the packaging is the clever part, and getting these various chips and reliable cavities into the package to fit the form factor is challenging.

The OSAT company that can create these paths will be very busy indeed. Interposers, creative methods for new TSV-like technologies, new materials for assembly, and human factors incorporated into robotic handling of the small packages will result in new business. Yield and reliability are the key factors in OSATs. Reliability is an area where they can grow in order to provide a reliable solution for the marketplace.

OSAT companies have shown they can make hermetic TSVs (this is still a challenge for some companies) and can build custom-test equipment with high yield using some industry standard base equipment.

Flexibility in test fixture and equipment development is a market brimming with possibilities. An OSAT that hires creative and experienced scientists and technicians will be ahead of the game. Experience in MEMS testing, semiconductor testing, and optical testing will be key for future test development.

John Stabenow, Tanner EDA, noted, “With an expectation of a trillion connected sensors online in the next decade, the OSAT suppliers who can deploy MEMS sensor calibration, as part of testing at large volumes, will have competitive advantage in the market.”

Adrian Arcedera, Amkor, said that as the MEMS and sensor market continues to grow, an OSAT company like Amkor, with a long experience in MEMS, can provide packaging technology for assembly and test, allowing IDMs and fabless companies to focus and invest on front-end/MEMS silicon design.

Move toward common testing protocols
It would be interesting to evaluate how the maturing MEMS sector is moving toward common testing protocols.

Stephen Whalley, SEMI-MSIG, said that while certain elements of the MEMS supply chain mature, there are other ever-changing areas in the ecosystem where the industry collaboration is needed to ensure smooth and scalable growth to meet the demand of new markets.

While standards are not always the answer, they are certainly needed where multiple suppliers are providing common solutions for multiple devices, markets and customers. SEMI has a history in standards stretching over four decades with nearly 1,000 standards and safety guidelines published to date. SEMI and its members are active in the following packaging and test communities designed to unearth, discuss, and address these critical supply chain challenges:

• Advanced Packaging & Test Committees
• SEMI integrated Packaging, Assembly & Test (SiPAT)
• Collaborative Alliance for Semiconductor Test (CAST)
• The Heterogeneous Integration Roadmap

These working groups will be integral to driving cross industry collaboration and solutions for MEMS packaging and testing methodologies.

Mary Ann Maher, SoftMEMS, felt that common test protocols are being followed based on device type. For example, inertial sensors, like acceleration sensors, are often tested in a similar way. The so-called “shake and bake”. Also, the application may dictate test protocols that are common to any MEMS addressing that application. RF MEMS components, for example, must pass the qualification tests for the application.

Gerard John, Amkor, added that in general, testing protocols for a particular type of MEMS devices are standardized, irrespective of the type of transducer used. For example, a MEMS microphone, irrespective of the type of transducer – capacitive, inductive or piezo-resistive, would require a sound source that is calibrated and generate a signal in the 20Hz to 20KHz range.

The testing protocol would be to measure sensitivity, SNR, THD and frequency response as a minimum. Of course, every MEMS stimulus has its own design peculiarities that would need a custom test, that is outside the standard protocol.

More efficient testing and process development
Another area of interest is: how would all these developments make the MEMS process development and testing more efficient in future.

Gerard John, Amkor, felt that the best way to make testing more efficient is to reduce the amount of testing, using the “reduced test set” (RTS) approach. In this approach, most of the parameters are guaranteed by design and only a few tests are run during high volume manufacture.

Allyson Hartzell, Veryst, said: “MEMS process development and testing are commonly done in a company’s development facility, a university fab, with consultants and/or with some OSATs that provide such services. The use of common materials, common cavity sizes, and common interconnect technologies can speed up the time to market by eliminating some cycles of learning.

“Using known and reliable methods can allow for integration with new unproven sensor technologies – the latter will be the area for extensive development and reliability testing. Yield can be designed into the MEMs process and packaging using finite elemental simulation techniques and numerical applications. Reliability is also now being designed early into devices using simulation techniques and understanding the physics of failure.

“Using standardized testing with some special tweaks for each new MEMS application is easier than starting anew each time. Standard building blocks for testing should be easily identified so that test systems can be kitted easily, saving the development time for specific applications. Kits for inertial, chemical, acoustic (examples) testing should be available in various resolutions and variables.”

Mary Ann Maher, SoftMEMS, added that the use of standard unit processes and design for test will create more efficiencies. There will still be plenty of custom MEMS processes, but, we are seeing some standard processes too, with process design kits, as well as custom process sequences built from standard unit processes.

In the opinion of John Stabenow, Tanner EDA, one key enabler of mass producible MEMS devices will be the inline inspection for defects using ML and AI techniques. To achieve large volumes, there will need to be better wafer-level inspection, as MEMS devices pass through various manufacturing steps.

He added: “Currently, this is a very high-touch process, requiring a master craftsman approach to visual inspection. The innovation coming will be inline inspection systems that can model specific, known defects and then, detect them accurately and quickly, as part of a volume production process.”

Enabling future wafer test systems
Finally, are MEMS technologies enabling the future wafer test systems? Mary Ann Maher, SoftMEMS, felt that MEMS technologies have forced the evolution of wafer test systems, particularly, in the area of environmental test and the use of mechanical stimuli/measurements.

John Stabenow, Tanner EDA, noted that MEMS-based FT-IR spectroscopy is one innovation that may find its way into standard wafer test systems and flows. As part of a monolithic mechanical-optical-electrical device, it may be bring advantages, like better core component alignment over traditional solutions using discrete components.

Amkor is one of the pioneers of MEMS wafer testing. Gerard John said, “We continue to work in this area with our vendors and customer in improving the measurement quality and the number of sites being tested in parallel.”

Allyson Hartzell, Veryst, added: “MEMS wafer test has developed to include options like vacuum test, testing with sound, and acceleration chucks for silicon wafers. Yet, how are we to perform various environmental inputs on say, wafer-level packaging on flexible substrates? Substrates that bend, twist, and curl are coming into popularity. The drive to place known good die on substrates that are no longer only PCBs, is in the forefront today. Placement and accurate analysis of bonding, to flex circuitry, thin bendable glass substrates, and even paper substrates, are being developed and will be coming soon.

“The drive to reduce size and power is also key. This will introduce challenges such as how to test a MEMS device with a good signal-to-noise ratio.”