Improving Control with MEMS Inertial Sensors

Recent advances in sensor technology have revolutionized the design of robotics and other industrial systems. In addition to robotics, applications where inertial sensors have the potential to improve system performance or functionality include: platform stabilization, industrial machinery motion control, security/surveillance equipment, and industrial vehicle navigation. The motion information provided by such sensors is very useful, not only to improve performance, but also to increase reliability, safety, and reduce costs.

However, to achieve these benefits, some obstacles must be overcome, especially the harsh physical environments in which many industrial applications are located, where the effects of temperature, vibration, space constraints, and other factors must be considered. For engineers, in order to obtain consistent data from sensors, convert it into useful information, and then respond within the system's timing and power budget, engineers must have knowledge and experience in multiple technical fields and follow good design practices.

Understanding the Problem
Information from inertial sensors can be processed and integrated to provide many different types of motion, position, and orientation outputs (see Figure 1). Each type of motion involves a host of application-specific complexities that must be understood. Industrial control applications are a good example where some form of pointing or steering equipment is useful. Tilt or angle sensing is often the core task in such applications, and in the simplest examples, a mechanical bubble sensor will suffice. However, before defining the sensor requirements, the full motion dynamics, environment, life cycle, and reliability expectations of the final system need to be analyzed.

Figure 1. Today’s inertial sensors are capable of detecting a variety of motion types.

If the system's motion is relatively static, a simple angle sensor may be sufficient, but the actual technology decision depends on response time, shock and vibration, size, and performance drift over the entire service life. In addition, many systems involve multiple types of motion (such as rotation and acceleration) and often operate on multiple axes, which requires considering combining multiple types of sensors.

Once the right sensor type and technology are known, the challenge shifts to understanding and ultimately compensating for the sensor's response to the environment (temperature, vibration, shock, mounting position, time, and other variables). Environmental compensation involves additional circuitry, testing, calibration, and dynamic adjustments, and each type of sensor, and even each sensor, is unique, which in turn carries the additional risk of under- or over-compensation unless the engineer has a good understanding of the sensor characteristics. This last point drives many design engineers to adopt fully integrated sensor solutions to eliminate barriers to adoption and implementation.

Linear Rate or Angular Rate
There are many types of inertial sensors. MEMS (micro-electromechanical systems) sensors are one of the most well-established sensor types and have benefited numerous applications. 15 years ago, MEMS linear acceleration sensors (accelerometers) revolutionized automotive airbag systems. Since then, a variety of unique features and applications have been enabled, from laptop hard drive protection to more intuitive user motion capture in game controllers.

Based on the principle of a resonator gyroscope, MEMS structures can also provide angular rate sensing. Two polysilicon sensing structures each contain a "disturbance frame" that is electrostatically driven into resonance to produce the necessary motion to generate the Coriolis force during rotation. At the two outer extremes of each frame (orthogonal to the disturbance motion) are movable fingers that are placed between fixed fingers to form a capacitive pick-off structure to detect the Coriolis motion. As the MEMS gyroscope rotates, the position changes of the movable fingers are detected by changes in capacitance, and the resulting signal is fed into a series of gain and demodulation stages to produce an electrical rate signal output. In some cases, this signal is also converted and fed into a proprietary digital calibration circuit. The
level of integration and calibration around the sensor core is determined by the final performance requirements, but in many cases, motion calibration may be required to achieve the highest performance levels and stability.

Conditioning and Processing
In the industrial market, applications such as vibration analysis, platform calibration, and general motion control require highly integrated and reliable solutions, and in many cases the sensing element is embedded directly into existing equipment. In addition, sufficient control, calibration, and programming functions must be provided to make the device truly self-sufficient. Some application examples include:

● Machine Automation: By improving the accuracy of position detection and more rigorously correlating this information with remotely controlled or programmed movements, autonomous or remotely controlled precision instruments and robotic arms can be made more accurate and efficient.

● Condition monitoring of industrial machinery: More practical value can be obtained by embedding sensors deeper into the machinery and leveraging sensor performance and embedded processing to detect signs of state changes earlier and more accurately.

● Mobile communications and surveillance: Whether it is a land, air or sea vehicle, inertial sensors help it achieve stability (antennas and cameras) and directional navigation (dead reckoning using GPS and other sensors).

The industrial sensing market is extremely diverse and must support a variety of performance, integration and interface requirements by integrating embedded adjustable features such as digital filtering, sampling rate control, status monitoring, power management options and dedicated auxiliary I/O functions. In other more complex cases, multiple sensors and multiple types of sensors are required. Even seemingly simple inertial motion, such as motion limited to one or two axes, may require both accelerometer and gyroscope detection to compensate for gravity, vibration and other unconventional behaviors and effects.

Sensors may also have cross-sensitivities that often need to be compensated for, or at least understood. Further complicating matters is the fact that there are many different standards for inertial sensor performance metrics. When specifying angular rate sensor requirements, most industrial system designers are primarily concerned with gyro stability (bias estimation over time), which is not typically specified for consumer-grade gyros. Even a good gyro bias stability of 0.003°/s may be meaningless if the sensor has poor linear acceleration performance. For example, assuming a linear acceleration characteristic of 0.1°/s/G, this would add 0.1° of error to the 0.003°/s bias stability in the simple case of a ±90° (1 G) rotation. Accelerometers are often used in conjunction with gyroscopes to detect the effects of gravity and provide the necessary information to drive the compensation process.

In order to optimize sensor performance and minimize development time, a deep understanding of sensor sensitivity and application environment is required. Calibration plans can be tailored to the factors that have the greatest impact, reducing test time and compensation algorithm overhead. Application-specific solutions that combine the appropriate sensor with the necessary signal processing, if cost-effective and with readily available standard system interfaces, will eliminate implementation and production barriers that many industrial customers have faced in the past.

Acceleration, vibration analysis
In some use cases, relatively simple sensor outputs may be sufficient, but in other applications (e.g., condition monitoring via vibration analysis), considerable additional processing is required to achieve the desired output.

An example of a highly integrated device built around an inertial sensor is the ADIS16227 (see Figure 2), a fully autonomous frequency-domain vibration monitor. Such a device may not provide a relatively simple g/mV output, but rather an application-specific analysis. In this case, its embedded frequency-domain processing, 512-point real-valued FFT, and on-chip memory can identify and classify individual vibration sources, monitor their changes over time, and react based on programmable thresholds.

Figure 2 ADIS16227 has detection and analysis capabilities to simplify design, with functional block diagram on the left

Being able to detect and understand motion can have application value in almost every imaginable field. In most cases, people want to control the motion occurring in a system and use this information to improve performance (response time, accuracy, operating speed, etc.), enhance safety or reliability (system shutdown in dangerous conditions), or obtain other value-added features. However, in some cases, the absence of motion is critical, so sensors can be used to detect unwanted motion.

These features or performance upgrades are often implemented on existing systems, and the small size and low power consumption of MEMS inertial sensors are undoubtedly very attractive, considering that the power consumption and size of the final system are determined or must be minimized. In some cases, the designers of these systems are not experts in motion dynamics, so the presence or absence of fully integrated and calibrated sensors may be the most critical factor in deciding whether to upgrade the system.