MEMS inertial sensors, such as accelerometers and gyros, are some of the most challenging types of sensors to test because their characterization requires an extensive range of physical stimuli. We recently spoke with Sascha Revel, Director of Sales at ACUTRONIC about testing of inertial MEMS devices. As a company that mainly focuses on inertial sensor testing, ACUTRONIC's perspective is unique and offers valuable insights. In this interview, Mr. Revel discusses the main types of inertial MEMS tests, reliability requirements, key challenges and associated costs.
MEMS Investor Journal: What is the main difference between MEMS device testing and IC testing?
Sascha Revel: Whereas traditional semiconductor devices use a single generic tester for mostly electrical tests, MEMS devices and especially inertial MEMS require a series of tests that include physical stimuli.
MEMS Investor Journal: What are the main types of tests for MEMS accelerometers?
Sascha Revel: MEMS accelerometers need to undergo several static and dynamic stress tests such as mechanical shock, variable frequency-vibration testing, temperature cycling and constant acceleration testing at component level. Standard qualification tests ensure long-term reliability, such as high and low temperature operating life tests also need to be passed. The most important tests for characterization are as follows: scale factor test, bias (or acceleration offset), linearity, axis alignment test, and off-axis vibration sensitivity.
MEMS Investor Journal: For each of these types of tests, what does the test do and why is it important?
Sascha Revel: The scale factor test which provides a constant defining the relationship between the raw acceleration measurements and preferred engineering units. The scale factor can be measured in many ways: (a) local gravity can be used to perform a ±1g calibration; (b) for higher “g” levels, an accelerometer can be mounted on the edge of a rotating table producing a radial acceleration proportional to the radius times the angular rate squared; (c) a linear shaker or an off axis angular vibration table can be used to measure the scale factor as a function of frequency.
Bias (or acceleration offset) which is the measured acceleration when the device is not moving and not subject to a component of gravity. Bias is characterized by initial offset after warm-up, short and long term drifting and thermal sensitivity. Bias is usually measured by nulling the device in the g-field then rotating the sensitive axis precisely 180 degrees through the g-field; the measured value at this position is twice the value of the bias.
Linearity is a measure of the scale factor error gradient over the operating range of the device. Linearity errors come in many forms which include asymmetry, saturation and rectification phenomenon. Linearity errors are measured by performing a multi-point tumble test in the gravity field or on a precision rate table using simulated radial acceleration.
Axis alignment tests measure the error between the sensitive axis and physical reference features of the package. Procedures for measuring alignment errors are generally not so simple because they involve the alignment of the rate table and knowledge of the fixture errors.
Off-axis vibration sensitivity results in measurement errors that degrade the dynamic performance of an accelerometer. These errors are often subtle but can significantly degrade the accuracy of a system that integrates acceleration over time to estimate rate and/or position motion states. The off-axis sensitivity is measured by stimulating the device with linear or rotary vibration along or about the non-sensitive axes of the accelerometer and monitoring the magnitude of the correlated signal in the sensitive axis. Since this is a dynamic phenomenon, vibration testing allows characterization as a function of frequency. Because of the architecture of MEMS devices and the utilization of suspension structures, cross coupling is more significant than developers would like to admit; often this error source is not tested and rarely compensated.
The following figure shows a sensitivity and non-linearity measurement for an accelerometer. The acceleration stimulus (over the full g-range) is generated with ACUTRONIC’s equipment (tilt axis, centrifuge or angular vibration table). The accelerometer output (in volts) is measured and plotted over the acceleration in g.
MEMS Investor Journal: What are the differences between the testing of a MEMS accelerometer and a MEMS gyro?
Sascha Revel: Unlike MEMS accelerometers, where the MEMS element is static when at rest, MEMS gyroscopes typically incorporate a vibrating mass which is continuously in movement when powered. Due to their design and tiny size, MEMS gyroscopes are much more sensitive to their environment and require more complex testing than the accelerometers.
MEMS Investor Journal: What are the main types of tests for MEMS gyros?
Sascha Revel: The most important and common tests for characterization are scale factor (offset and sensitivity), non-linearity error, bias, resolution, sensitivity to temperature (drift), and sensitivity to acceleration (drift).
Key test parameters for gyros are very similar to accelerometer test parameters, but the motion state to be measured is different: the accelerometer detects a linear displacement (which can be simulated with a rotational movement) whereas the gyro detects the angular motion. Therefore, the orientation of the sensor on the test equipment differs whether you test the accelerometer or the gyroscope, as shown on the sketches below.
Setup for an accelerometer with the test axis x by tilting Ө (left); setup for a gyroscope with the test axis x by rotating Ф (right).
MEMS Investor Journal: What are the main types of tests for MEMS based sensor combos or inertial measurement units (IMUs)?
Sascha Revel: To test a MEMS sensor combo or a MEMS based IMU, which consists of three orthogonally assembled MEMS accelerometers and three gyroscopes, it is necessary to repeat most of the tests listed above, since each stage of the inertial MEMS value chain (between single device testing and final system testing) creates more potential root causes for malfunctions which need to be verified.
For multi-axis inertial sensors,whether or not is it a single-chip combo or a multi-sensor integrated device, cross axis sensitivity needs to be characterized as well. This is typically calculated with the data measured on all axes simultaneously during the scale factor measurement.
The wide variety of applications for MEMS inertial sensors and the wide range of IMU performance requirements is a challenge for test equipment manufacturers as the qualification and characterization test requirements are not standardized.
MEMS Investor Journal: Overall, what are the biggest challenges for MEMS testing?
Sascha Revel: Always a stunning contribution is time, which is basically just the result of the involved temperatures. The gradients per se are not the problem, but soak times and stabilization times are. You don't need the most powerful oven in the world -- your pizza just needs to be exposed to a certain temperature for a certain time in order to be tasty.
MEMS Investor Journal: What other challenges do you encounter?
Sascha Revel: MEMS inertial sensors are among the first practical devices to emerge from an evolving MEMS industry. Nonetheless, MEMS inertial sensors are far from being mature as they continue to command the focus of many commercial and military development programs. MEMS sensors have been produced by many teams having expertise in the design and fabrication phases then to be challenged by their limited understanding of the testing protocols and associated test equipment. Even those organizations who have experience with testing legacy inertial devices are realizing that testing MEMS sensors requires addressing device characteristics and behaviors that may not easily be evaluated using classical motion simulation.
Self test capabilities play an important role in screening and pass-fail testing of low end commercial devices; however, physical testing of inertial sensors is always a requirement during development phases and for production testing of performance devices.
A key challenge for inertial MEMS testing is to develop hardware motion platforms and testing protocols which provide inertial MEMS developers with the tools they need to quickly make accurate assessments of devices; this applies to both product development and production testing. The high volume manufacturer needs and the startups or universities requirements cannot be satisfied with the same solutions. There is no “one size fits all” solution for inertial MEMS testing.
MEMS Investor Journal: What are the most well established methods and types of MEMS testing?
Sascha Revel: The methods and types of testing are usually defined by the final application of the system based on the MEMS sensors, and rely on different standards (JEDEC and SEMI, IEEE, MIL, etc). A successful MEMS manufacturer has only one to a few different sensor dies being used for a variety of applications. Such a sensor is then combined with an ASIC in order to fulfill its purpose. It might be sitting on a mortar for two minutes or somewhere in functional clothing for a couple of days. The required tests vary significantly and most probably cannot be tested on the same motion system. With MEMS technology opening up countless new fields of applications it won't be easy to keep the pace for test systems. And, once again, inertial MEMS are much more sensitive to their environment.
MEMS Investor Journal: How can a company calculate a testing program return on investment (ROI)? What is the best way to estimate the budget for testing?
Sascha Revel: The company needs to know to the very specific detail, what physical stimuli will be needed and what are the corresponding requirements. The final application of the sensor and its performance will define how many iterations, in what environmental conditions, and for how long the sensor needs to be tested. We ar ACUTRONIC therefore offer consulting services as part of the Inertial Test Solution laboratory in our Pittsburgh, Pennsylvania facility.
In order to estimate the return on investment, the customer should first be aware of the capital expenditure required to setup a test lab such as ours. After spending around one million dollars for the test equipment (including linear and/or angular vibration tables, single or multi-axis motion tables with or without temperature chamber and the data acquisition system) the sensor manufacturer will be able to begin their development tests, while hiring and assigning their test, mechanical and electrical engineering personnel. The team typically spends 12 to 18 months to design, develop, test, debug and validate the test facility, thus representing a few more hundred thousand to one million dollars of fix costs.
One company that ACUTRONIC interviewed confessed to have spent three years developing their MEMS test lab. Another company admitted to having spent close to $2 million to set up a facility for a single gyro product.
When using our Inertial Test Solution laboratory in the Pittsburgh facility, our customers can easily estimate the testing costs per unit by adding the rental fees, shipping and logistics costs, as well as the material and hardware costs (such as fixtures and cables) divided by the number of sensors tested simultaneously.
Furthermore, our expertise and support enables our customers to optimize their test setup and reduce the number of iterations, thus saving costs.
MEMS Investor Journal: Are there any general "rules of thumb" on what percentage MEMS testing typically constitutes of the low-end, mid-range and high-end MEMS inertial sensors final cost?
Sascha Revel: Very difficult question to answer. The MEMS Industry Group (MIG) recently conducted a survey which revealed that costs of inertial MEMS testing range from "1-2%" to "40-50%" with an average of 22% of the final device cost. It is not easy to relate these figures to the classification of the concerned sensors: low grade for commercial applications, medium grade for automotive and industrial applications or tactical grade for aerospace and defense applications.
Interestingly, the usual expectation of the low grade inertial MEMS sensor manufacturer is that cheaper sensors would require cheaper testing equipment. This is not the case at all. Especially since sensor fusion (acceleration, rate, temperature, pressure, magnetic field, etc.) is the trend, the test equipment must have the capability of multimode physical stimuli.
Like for a rental car, our daily fee decreases with the increasing number of days resulting in lower test costs for larger numbers of devices. For example, a customer of ours is currently testing up to 200 devices a day. The effective testing costs performed in our lab would represent a certain percentage of the device final selling price. Calibrating 1000 devices during a one week test campaign enables him to drop the testing costs to a significantly smaller percentage of the device cost.
*********************************************
Sascha Revel is the Director of Sales at ACUTRONIC. To contact him with further questions, please email your inquiries to [email protected].
This article is a part of MEMS Investor Journal's ongoing market research project in the area of MEMS testing and reliability. If you would like to receive our comprehensive market research report on this topic, please contact Dr. Mike Pinelis at [email protected] for more information about rates and report contents.
Copyright 2011 MEMS Investor Journal, Inc.
Comments