by Maarten Vraanes
Director of Consulting Services, MEMS Journal
MEMS packaging, as the entire MEMS ecosystem, is rapidly evolving. Technologies such as wafer level and 3D integration are becoming increasingly important. In this article, we outline our observations about the current and future trends of MEMS integration and packaging. Major trends focus on developing CMOS-compatible MEMS fabrication processes for monolithic integration, such as low-temperature wafer bonding. Another emerging trend is die stacking in low-cost leadless type semiconductor packaging -- these types of techniques provide smaller footprint packages with lower unit costs for high production volumes. Additionally, 3D integration enables the integration of LCR passives. Here, LCR components are embedded into the package itself to minimize external passives and accommodate smaller footprint applications, wafer bonding and vertical intra-package connectors or interposers. On the flip side, CMOS and 3D integration of MEMS devices pose challenges in modeling, testing, and reliability.
Silicon interposers and package integration
Silicon interposers are microstructures used to vertically connect two die in a stacked die or 2.5D/3D packaging configurations. Common applications include high pin count, high density connections for micro-controllers and FPGAs, but this technology is similarly adopted by MEMS in the form of though-silicon vias (TSVs).
Metal based silicon interposers are developed by DRIE etching vertical trenches of tens to hundreds of microns in silicon or glass wafers, followed by metallization to create vertical electrical conductors. The coefficients of thermal expansion (TCE) mismatches between materials (i.e. metal and silicon) can cause high levels of mechanical stress during the processing of interposer wafers. The stress can result in high wafer bow and warp (picture a potato chip) and they become very fragile and difficult to process.
Doped polysilicon can be an alternative to metal-type TSVs, especially for high-aspect ratio MEMS TSVs whose depth can go into the hundreds of microns. Other challenges for interposers include heat transfer for high-density, power intense applications, the lack of EDA tools for simulation and modeling of 3D integrated devices, assembly process reliability and access to signals during failure analysis.
Wafer level packaging
An ultimate goal in MEMS wafer level packaging is to have no package. The term, “no package” (or “silicon package”) refers to a device’s package being created during the wafer processing without the need for further assembly after wafer dicing. The main advantages are to minimize the device size and to avoid costly, traditional semiconductor assembly. Being its own package, the device is complete with metal pads for post-fabrication testing and application use (i.e. surface mounting, or SMT).
Utilizing wafer level packaging and other micro-fabrication techniques, devices can conceivably include an application specific integrated circuit (ASIC) with analog interface circuitry, combined with one or more MEMS sensors -- a complete sensor system-in-a-package (SOC). The ASIC and MEMS could be developed on the same wafer or on separate wafers, and then bonded together to form the silicon package. Wafer bonding creates a hermetic seal for protection and provides electrical interconnections between the ASIC and the MEMS. In this configuration, the ASIC wafer could include TSVs allowing electrical signals to be transferred to the bottom of the stack. This would allow metal pads to be placed on the bottom of the stack provide a medium for electrical probing/testing and surface mounting.
The wafer bonding must be a high yielding process to ensure uniform bonding quality and hermetic sealing of each individual sensor. Even the tiniest leak will change the internal environment of the MEMS, which is known to cause sensor drift over time. These leaks are hard to detect, and the tests to find them are typically destructive. Instead, sample testing and ongoing statistical process monitoring must be used to guarantee the bonding quality.
On the back-end, very small packages with their tiny pad pitches will require high-density test pins and possible even use MEMS probes for automated production testing. Finally, the device size and brittleness of silicon might prevent traditional semiconductor handling using gravity feed or pick and place handlers.
The MEMS industry has over the last few years worked towards the goal of bundling 10-degrees of freedom (DOF) into one device, boasting a full personal navigation device. The 10 DOF devices include a 3-axis accelerometer, a 3-axis gyroscope, a 3-axis magnetometer and a 1-axis barometric pressure sensor (altimeter). Although it might be technically feasible to include all these functions in a relatively small footprint package, it might not be economical for a number of reasons.
There are critical tradeoffs between the cost of integration and manufacturability. Multiple sensors means added complexity in the assembly processes where die-stacking is needed to achieve small footprint packages and film-assisted molding to create a vent hole for barometric pressure sensing. Whereas over-molding of one or two die is very common and cost-effective, multiple die stacking combined with MEMS specific processes, such as gel dispensing for mechanical shock isolation, can increase the packaging cost significantly, both in development and unit cost.
A process that is feasible in small volumes might be very challenging in mass production, where volumes can be 10s or 100s of millions of sensors and the assembly yield must be in the high 90%. Over time, MEMS product companies seem to have changed their strategies to bundle similar and complementary functions and separate others into different packages. This is meaningful from an application standpoint where a motion sensor should be located in the middle of, for example, a handheld electronic device, whereas a magnetometer is better located separately from other components to minimize the effects in the ability to measure the earth’s magnetic fields.
For production testing, motion sensors can be tested for both linear and angular acceleration in the same test jig, but pressure sensors need a completely different set of test stimuli and typically require their own test system. At the end of the day, the integration of multiple sensor functions in one package is a complicated puzzle that companies are trying to navigate as they are working to stay on top of the MEMS tidal wave in consumer electronics and other applications.
This article is a part of MEMS Journal's consulting service offerings in the area of MEMS packaging and testing. For further inquiries, please contact Dr. Mike Pinelis at email@example.com for more information.