MEMS (micro electromechanical system) inertial sensor technology provides a major structural shift in mechanical sensing. When the performance meets their needs, system developers welcome trading complicated mechanical mounting schemes for the simple solder-reflow attachment processes used for integrating MEMS sensors.
Automotive safety systems were one of the first consumers of these products. As the MEMS manufacturing infrastructure developed to support their high-volume requirements, additional market segments, such as mobile handsets and game controllers, embraced these functions and eventually became primary drivers of product developments themselves. While each of the applications for inertial sensing is very different, they all have common requirements for this function:
- Inertial sensing (linear, angular) in small, PCB-capable packaging
- Reliable operation
- Performance levels that were appropriate for the functions they serve
- Low power dissipation
- Cost effective
- Simple Integration
Other users who value all of the above factors include industrial-system developers. One example is in motion/motor control systems, which sometime use orientation sensing to establish a reference position or to verify proper command execution. In some cases, using an accelerometer as a tilt sensor provides a valuable function that can simplify system designs.
This article highlights the process of turning this simple concept into a function that provides value for motion-control systems. The key differentiator in this case will in the “performance level that is appropriate for the purpose it serves.” The reason for this is that the accuracy required in these systems is often higher than the performance provided by the discrete sensors.
Tilt sensing overview
Tilt sensing encompasses a wide variety of approaches, which vary in both complexity and performance. In this case, the tilt-sensing system will use gravity as its only stimulus, and a MEMS accelerometer as its sensing element.
MEMS accelerometers typically employ a tiny, spring-loaded structure that is interlaced with a fixed pick-off finger structure. The spring constant of the “floating” structure determines how far it will move when subjected to a force. This distance is observed as a change in capacitance by a modulation/demodulation circuit. Since the structure is responding to a force, it does not matter if the force is due to acceleration (F = ma), or a static force such as gravity.
Figure 1 illustrates the basic, single-axis MEMS accelerometer approach to tilt sensing. “Horizon” is defined as the plane that is orthogonal to the earth's gravity. “Incline angle” refers to the tilt angle with respect to the horizon.
Figure 1: Single-axis MEMS accelerometer tilt sensor
(Click to enlarge image)
When the accelerometer is parallel with the horizon (incline angle = 0 degrees), its measurement will be zero. As the incline angle approaches 90 degrees, the accelerometer's measurement will approach 1 g , assuming perfect accuracy in the sensor.
An important consideration for a single-axis tilt sensor is that its sensitivity to incline angle decreases as the incline angle increases, eventually converging to zero when the accelerometer is in the vertical position.
Depending on the orientation system's measurement range, a two-axis approach may be useful in maintaining accuracy goals, particularly if the measurement range is greater than ±30° and the accuracy requirements are sub-1°. In this configuration, which is displayed in Figure 1, the two sensors are orthogonal to one another. As the sensitivity of one degrades, the other increases, and vice versa.
In addition, the second axis enables a full 360° of measurement capability. By itself, a single accelerometer cannot distinguish which half of the rotation cycle it is in. If a motor system rotates to an incline angle of 45°, then another 135°, the accelerometer output will be the same.
Another important note is that the plane that encompasses both accelerometers must be orthogonal to the horizon. As this plane tilts away from this position, additional error terms are introduced.
The hardware realization of this type of tilt system starts with sensor selection. There are a number of factors that will play a role in this selection process. The range of the sensor should theoretically be at least 1 g . From a practical perspective, 1.5 g would provide some margin for sensor error and a small amount of vibration—which can sometimes be filtered if it doesn't saturate the sensor. As the range of the sensor increases, the resolution will decrease, presenting a potential threat to performance goals.
Another important consideration will be in the output signal configuration, which will dictate the level of complexity in the processor interface. The three most common interfaces are analog output, pulse-width modulation (PWM), and digital. The analog output configuration requires an analog-digital interface, while the other two configurations can feed directly into a processor platform, which will perform the acceleration-to-incline-angle math described in Figure 1.
While the two direct-interface options (PWM and SPI) present less complexity, the performance of the sensor will likely have substantial influence as well. Thermal, power supply, and life stability influences on the accelerometer's sensitivity and bias are key parameters to consider in this process. In many cases these factors will make an analog-output sensor the most attractive option. If this is the case, a two-sensor tilt system may have a block diagram that looks like Figure 2 .
Figure 2: Tilt system block diagram
(Click to enlarge image)
The digital platform can be from any number of technologies, including microcontrollers, digital signal processors (DSP), field programmable gate arrays (FPGA), and programmable logic devices (PLD). The selection of these will depend greatly on the processing needs at the system level. The analog-to-digital converter (ADC) selection will depend on a number of important parameters that include resolution, power forms, reference configuration (or ratiometric), input impedance (buffering may be required), accuracy, and stability. Many microcontroller platforms provide ADCs, but they may not have adequate resolution or an interface that is suitable for ratiometric sensor outputs.
Achieving valuable accuracy
Successful development of an orientation sensing system must start with a clear performance goal. For example purposes, let's explore a system that can measure ±45° away from horizon, on two different axes, at an accuracy of ±0.5°. Using the relationship between acceleration and incline angle in Figure 1, these design goals translate into acceleration accuracy requirements of approximately 6 mg .
Assuming that the accelerometer selection is constrained to the mainstream, high-volume offerings, one of the best performing options has a sensitivity variation of ±4% and bias specification of ±25 mg . The bias variation alone is four times the allowable requirement, assuming that every other influencing factor is perfect! By itself, the sensitivity of 4% would introduce an additional error of 2.3° at an incline angle of 45°. In addition, the amplifier, multiplexer, and ADC errors will contribute as well. The bottom line is that this level of accuracy will require calibration.
One of the more practical approaches to calibrating a dual-axis accelerometer is known as the “Four Point Tumble” method. In this arrangement, the accelerometers are vertical-mounted to a motor, or other apparatus, which moves the accelerometers in 90-degree steps. Figure 3 illustrates the different positions at which the accelerometer is characterized.
Figure 3: Four point tumble positions
(Click to enlarge image)
Figure 4 illustrates the response each accelerometer would have over the entire 360° rotation range. Figure 4 shows ideal curves for both axes, along with an exaggerated measured response, for illustration purposes.
Figure 4: Accelerometer response, ideal and with error
(Click to enlarge image)
This characterization information is used to develop scale and offset factors, which are loaded into correction tables for the accelerometers. Once the accelerometer accuracy is established, incline angle accuracy follows. Ideally, the total change in accelerometer output should be 2 g , and the curves should be centered around zero. The following equations generate the appropriate scale and offset correction factors for each accelerometer.
Once the scale/sensitivity errors have been compensated, then the offset can be calculated:
In order to correct for other factors, additional characterization steps may be required. For example, power-supply variation may require doing the same calibration sequence for multiple power-supply levels.
MEMS technology offers a number of advantages to industrial system developers who are looking to integrate motion/orientation analysis into their systems. Integrating MEMS into these types of systems will require:
- Wise sensor selection
- developing the appropriate interface circuit, and
- assuring the appropriate level of accuracy through calibration.
This integration process will continue to evolve as developers continue to face short development cycles and pressure to improve value through performance increase or cost reduction. Also, the introduction of single-package options, such as Analog Devices' ADIS16201 family of products, will provide faster solution paths for achieving sub-1° accuracy levels in applications that value incline sensing.
The bottom line is that inertial MEMS sensors provide a valuable function to motion control system developers but in many cases, require extra processing to meet all of the performance criteria. The primary question that these developers face is the following: develop the required circuit and processes internally versus purchasing the capability and focusing on other issues in the design. Ultimately, there is a cost associated with achieving performance improvements. Each developer will have to weigh specific goals in order to choose between the potential for incremental component cost savings and the faster time to market and lower development costs associated with internal development.
About the authors
Mark Looney is the iSensor Application Engineer for Analog Devices, Inc. He earned a MSEE degree from the University of Nevada in 1995 and has 12 years of experience in design and applications engineering. He can be reached at firstname.lastname@example.org.
Joe Bergeron is the Director of Engineering for the MultiChip Products
Group at Analog Devices, Inc. He earned a BSEE from the University of Rhode Island and has 27 years of experience in integrated circuit and electronic component development. He can be reached at email@example.com.