Advances in MEMS-based inertial sensors

by

Doug Sparks, CTO, Hanking Electronics

Microelectromechanical system (MEMS)-based inertial sensor commercialisation goes back around 50 years and has involved many interesting developments, new applications and acquisitions. Development of capacitive and piezoresistive silicon linear accelerometers began in the 1970s1. Airbags were the first high-volume application for these inertial sensors in the late 1980s and early 1990s. As with many MEMS devices, the aerospace industry led the development of angular rate sensors, also known as MEMS gyroscopes. Like the accelerometer, the automotive market provided the first high-volume applications for MEMS gyroscopes, starting in the mid-1990s2 with navigation assist, vehicle dynamic control (VDC) and roll-over protection (ROP) systems. Now, both of these sensors can be found in every smartphone, tablet, vehicle and wearable device.

Most micromachined gyroscopes rely on the coupling of an excited vibration mode into a secondary mode due to Coriolis acceleration. The magnitude of oscillation in the sense mode provides a measure of the input angular velocity. The history of MEMS-based resonant ring angular rate sensors stretches back more than a century. In 1890, about 50 years after the Coriolis effect was discovered, Professor G. Bryan of the University of South Wales noticed the influence of the Coriolis effect on the tone of ringing wine glasses when turned by the stem3. A little less than one hundred years later, high accuracy gyroscopes, called hemispherical resonant gyroscopes (HRGs), were being produced for aviation and spacecraft by Delco, Litton (acquired by Northrop Grumman), Safran and other companies. These led to the smaller micromachined ring gyroscope chips developed in the 1990s at the University of Michigan with Delco Electronics (spun off as Delphi Technologies)4 as well as British Aerospace and now manufactured by Silicon Sensing Systems. As shown in the centre image of figure 1, the micromachined ring formed part of the MEMS chip. It required vacuum sealing to achieve a Q (measure of resonant signal peak sharpness) of 2,000 and thus deliver improved performance. The higher the Q, the better the performance.

The single-axis MEMS gyroscopes of the 1990s had zero rate drift and noise levels of 3 to 5 degrees per second. MEMS HRGs, such as the one in the right image of figure 1, are still under development at the University of Michigan.  The single-axis glass micro HRG shown has a Q of over 5 million and a bias stability of 0.0014 degree/hour5. The new MEMS HRG work at the University of Michigan has been spun off for commercialisation as Enertia Microsystems. While these high-performance HRGs were developed for aerospace navigation, the advent of autonomous vehicles and tighter specification for advanced driver-assistance systems (ADAS) could see wider market adaptation for this ever-improving inertial technology. A single-axis ring gyroscope was coupled with a linear accelerometer on the same chip in 19986. Only a single-axis gyroscope could be made on a MEMS chip with resonant ring technology. To make multi-axis gyroscopes on MEMS chips required the use of planar, gimballed capacitive micromachines.

The development of single-axis silicon MEMS angular rate sensors, or gyroscopes, was pioneered in the 1980s by groups such as Draper Laboratory for capacitive gimbled silicon devices2 and Systron Donner Inertial (acquired by EMCORE) for piezoelectric quartz tuning fork sensors. Gyroscopic sensing in all three axes, namely yaw, roll and pitch, was available on a single MEMS chip in the mid-2000s. These were quickly combined with 3-axis linear accelerometers for 6 degrees of freedom (DOF) inertial sensors. Silicon comb fingers could electrostatically drive and sense motion in the X-Y plane while underlying conductive electrodes below the micromachine could sense motion in the Z-axis. By 2010, low-cost, consumer grade 3 and 6 DOF devices were offered by Bosch, Fairchild Semiconductor (acquired by ON Semiconductor), InvenSense (acquired by TDK), Maxim Integrated (acquired by Hanking Electronics) and STMicroelectronics. As shown in figure 2, the key to putting both vacuum packaged resonant gyroscopes and damped linear accelerometers on one MEMS chip was dual cavity wafer-level packaging (WLP).

Low Q and Q loss due to gradual pressure increases in the chamber are a failure mode for large HRGs and MEMS gyroscopes. Thin-film getters, first used 100 years ago in vacuum tubes, were integrated into the MEMS wafer process and offered one of the best ways to improve resonant gyroscope performance by raising the Q through lower cavity pressure.  Thin film getters were first incorporated into solder sealed ceramic packages by Raytheon Technologies in the mid-1990s and then into MEMS wafer level packaging by Honeywell7 and the University of Michigan8 in the late 1990s for pressure sensors and gyroscopes. The thin-film getter wafer process was commercialised as a service by NanoGetters (acquired by Materion) and SAES in the early 2000s, and so made available to the whole MEMS community.  In a 2003 paper, the Q of the same resonator was found to be 40 without getters and 21,000 with getters in the same sized cavity9. This control of cavity pressure and Qs enabled the 6 DOF IMUs in figure 2.  Resonant gyroscopes were placed in the cavity with the thin-film getter, and the linear accelerometers that required damping were placed in the higher pressure microcavity that did not have a getter patterned on the cavity wall. 

For long-term high reliability, improved methods of wafer-to-wafer bonding are required to ensure that small diameter gases such as helium do not leak through the sealing interface10. Even the 5-parts per million (ppm) level of helium in the Earth’s atmosphere has been found to penetrate some WLP vacuum seals at the high end of the operating temperature range (85 to 125˚C), and nobel gases such as helium are not absorbed by chemically reactive getters. To maintain high Q for decades of use, resonators in conventional and MEMS HRGs must take care in vacuum sealing the devices. As shown in figure 3, WLP has enabled a dramatic shrinkage in IMUs over the last couple of decades. The small surface mount large grid array (LGA) packages on the right are incorporated into smart phones and watches for personal navigation.

Further improvement in MEMS inertial sensors have employed mounting both the MEMS and CMOS chip in thin, mechanically and thermally isolated platforms inside a cavity ceramic package11. The sensing chips on the isolation platform are solder vacuum sealed, with a thin film getter patterned on the Kovar lid, in hermetic ceramic packages. These micromachined insulating platforms dampen vibration and shock and can even employ heating coils to keep the sensing chips in a constant, slightly elevated temperature range, resulting in a 10-fold improvement in performance over temperature versus conventional ceramic and plastic packages.

The printed circuit board (PCB) of the large sensor module on the left of figure 3 could accommodate barometric pressure sensors, high- or low-g accelerometers, gyroscopes and other sensors needed for automotive applications12. The small surface mount IMUs that have been developed for mobile applications in the last few years are also integrating magnetic and barometric pressure sensors. The magnetic sensors can use the Hall-effect, magneto-resistance, Lorentz force and other magnetic sensing methods. Magnetic sensors can measure the Earth’s magnetic field and are often coupled with the linear accelerometers in the IMU via sensor fusion algorithms. The linear accelerometers can detect the Earth’s gravity or tilt of the IMU. Coupling the magnetic sensors and accelerometers can provide personal orientation with respect to magnetic north, also called an e-compass. Similar to the accelerometers and gyroscopes, the three small magnetic sensors are placed orthogonally with respect to each other in the x, y and z axes as a 3 DOF magnetic sensor. By stacking the magnetic sensor chips on top of the MEMS motion sensors, 9 DOF inertial sensors have been made in small surface mount packages for smart phone applications. MEMS barometric pressure sensors, such as the one shown in figure 4, have also been incorporated in IMUs for altitude measurement and are available in 5 x 5 mm surface mount packages. They enable the user to navigate as well as be located and tracked without the need for GPS.

In the future, MEMS gyroscopes may employ an optical mechanism instead of the resonant Coriolis effect.  Optical gyroscopes were first developed in the 1970s using the Sagnac effect and coils of fibre optic cable. The Sagnac effect splits a beam of light, and the two separate beams follow the same path but in opposite directions through the coils of fibre optic cable. On their return to their starting or splitting point, the two beams are allowed to exit the ring and undergo interference. The interference fringes of the relative phases of the two exiting beams are shifted according to the angular velocity of the apparatus. These optical ring laser gyroscopes (RLGs) do not have moving parts. Improvements in RLG performance intensified in the 1990s and continue today. Higher accuracies, capable of even detecting the Earth’s rotation, have been achieved using ring diameters of 1 to 4 m. Smaller, MEMS-based RLG prototypes have been developed by several groups using oxide and nitride waveguides13. Conventional RLGs can wind three-dimensional coils of fibre optic cable with relatively large diameters, while MEMS-based RLGs are limited to two-dimensional waveguides with diameters less than 10 mm. A single-chip RLG has not yet been commercialised and one would expect the accuracy to be lower than the larger diameter RLGs that are sold commercially.

In summary, MEMS-based inertial sensors span forty years of research and commercialisation that started with single-axis linear accelerometers, followed by gyroscopes and then magnetic and barometric pressure sensors. Through advances in MEMS design, processing and packaging, more sensing features have been incorporated into a single chip or via a few MEMS chips in a small surface mount package.  Not only have multiple sensors and axes been compressed into a tiny IMU package, but the performance of the inertial sensors over temperature and in aggressive applications has also improved.  MEMS gyroscope zero drift bias has improved from degree per second errors in the 1990s to fractions of a degree per hour errors in 2020. After successful development and commercialisation, billions of inertial MEMS sensors are manufactured every year and found in drones, smartphones, tablets, toys, vehicles and wearables.

Hanking Electronics

www.hankingmems.com

References

1 Roylance, L. M.  and Angell, J. B. (1979). A batch-fabricated silicon accelerometer. IEEE Transactions on Electron Devices, volume 26, issue 12, pp.1,911–1,917.

2Weinberg, M. S., Bernstein, J., Cho, S., King, A., Kourepenis., A., Ward, P. and Sohn, J. (1994). A micromachined comb-drive tuning fork gyroscope for commercial applications. Proc. Sensors Expo, September 20–24, Cleveland, Ohio, US, pp.187–193.

3Bryan, G. H. (1890). On the beats in the vibrations of a revolving cylinder or bell. Cambridge Philosophical Society, volume VII, part III, pp.101–111.

4Sparks, D. R., Zarabadi, S. R., Johnson, J. D., Jiang, Q., Chia, M., Larsen, O., Higdon, W. and Castillo-Borelley, P. (1997). CMOS integrated surface micromachined angular rate sensor for automotive applications. Proc. International Solid State Sensors and Actuators Conference (Transducers '97), June 19, Chicago, Illinois, US, 3B1.04, pp.851–853.

5Cho, J. Y., Singh, S. S., Woo, J. K., He, G. and Najafi, K. (2020). 0.00016 deg/rt-hr angle random walk (ARW) and 0.0014 deg/hr bias instability (BI) from 5.2M-Q and 1-cm precision shell integrating (PSI) gyroscope. Inertial 2020, March 23, 2020, Hiroshima, Japan.

6Sparks, D. R., Higdon, W., Huang, X. and Johnson, J. D. (1998). Angular rate sensor and accelerometer combined on the same micromachined CMOS chip. Microsystem Technology, volume 4, pp.139–142.

7Bender, T. D. (1995). Apparatus for mounting an absolute pressure sensor. United States Patent and Trademark Office (USPTO) patent no. 5,459,351. Available at: https://bit.ly/33Ie1L4

8Chavan, A. V. and Wise, K. D. (1997). A batch-processes vacuum-sealed capacitive pressure sensor. Proc. International Solid State Sensors and Actuators Conference (Transducers '97), June 19, Chicago, Illinois, US, 4D3.01, pp.1,449–1,452.

9Sparks, D. R., Massoud-Ansari, S. and Najafi, N. (2003). Chip-level vacuum packaging of micromachines using NanoGetters. IEEE Transactions on Advanced Packaging, volume 26, number 3, pp.277–282.

10Sparks, D. (2016). Advances in high-reliability, hermetic MEMS CSP. Chip-Scale Review, volume 20, number 5, pp.36–39.

11Yang, D. M., Najafi, K., Lemmerhirt, D. F. and Mitchell, J. (2018). A micro thermal and stress isolation platform for inertial sensors. Proc. 2018 IEEE International Symposium on Inertial Sensors and Systems (INERTIAL), March 26–29, Moltrasio, Italy, pp.1–4.

12Sparks, D., Noll, T., Agrotis, D., Betzner, T. and Gschwend, K. (1999). Multi-sensor modules with data bus communication capability. Automotive Engineering, volume 107, pp.37–38.

13 Gundavarapu, S., Belt, M., Huffman, T., Tran, M. A., Komljenovic, T., Bowers, J.E. and Blumenthal, D.J. (2017). Integrated Sagnac optical gyroscope sensor using ultra-low loss high aspect ratio silicon nitride waveguide coil. Proc. SPIE volume 10323, 25th International Conference on Optical Fiber Sensors, April 24–28, Jeju, South Korea, pp. 103231A-1–A-4.