testing approaches for characterization and selection of ...meptec.org/resources/9 -...

Testing Approaches for Characterization and Selection of MEMS Inertial Sensorsby Dino Smajlovic and Roman Tkachevby Dino Smajlovic and Roman Tkachev

© 2016, ACUTRONIC © 2016, ACUTRONIC 1

Table of Contents

Summary & Introduction 3Summary & Introduction 3

Sensor Parameter Definitions 5

Selection of Sensor Parameters for Testing 9

Test Methodology 14Test Methodology 14

Test Procedures 21

Test Results 29

Conclusion 33

© 2016, ACUTRONIC 2

Conclusion 33

Content Summary

The presentation is dealing with the question of selecting proper MEMS inertial The presentation is dealing with the question of selecting proper MEMS inertial sensors for your device/application: how to choose sensors that best fit your needs? Sensor Parameter definitionSensor Parameter definition

This section will covers a variety of parameters that typically need to be characterized.

Test methodology overview gy This section will cover how to determine parameter characterization.

Sample test process A sample test report is reviewed demonstrating the parameters and p p g p

methodology discussed previously.


Introduction

Due to a large number of manufacturers of the inertial MEMS sensors it becomes obviousDue to a large number of manufacturers of the inertial MEMS sensors it becomes obvious selection process is more complicated than ever before.While some of the characteristics of the sensors are easily compared (price, size, number of axes, etc.), performance parameters are typically only partially defined by the manufacturers and can vary greatly from one to another and over different environmentsmanufacturers, and can vary greatly from one to another and over different environments.

Because the expected use of the products with the embedded inertial sensors (smartphones, wearables, smart helmets, VR sets, etc.) can vary greatly the best approach to select MEMS inertial sensors is by conducting characterization tests for a defined set ofto select MEMS inertial sensors is by conducting characterization tests for a defined set of parameters.

In this presentation we are describing the process of testing inertial MEMS sensors (accelerometers and gyroscopes) selection of the parameters and choosing test(accelerometers and gyroscopes), selection of the parameters and choosing test procedures to collect relevant data. Once the data is collected and the product application is considered selection of the appropriate sensors can be made.


Sensor Parameter Definitions

Inertial sensors measure object’s orientation and position in space Inertial sensors measure object’s orientation and position in space Gyroscopes (gyros), accelerometers and magnetometers are considered

inertial sensors Typically inertial sensor parameters can be divided into two groups: dynamic Typically inertial sensor parameters can be divided into two groups: dynamic

and static. Static measurements include:

Noise and zero input offset informationNoise and zero input offset information Dynamic tests include

Scale factor error and linearity, cross-axis sensitivity, misalignment, full scale range and bandwidth testingg g

Most of these parameters can be tested over temperature to identify any temperature sensitivity.


Sensor Parameter Definitions

In order to properly define each parameter we recommend using IEEE 2700 In order to properly define each parameter we recommend using IEEE 2700 Standard for Sensor Performance Parameter Definitions by the IEEE Standards Association

In 2012, Intel Corp and Qualcomm Technologies, Inc. published a document to establish an industry standard minimum set of MEMS performance parameters: Standardized Sensor Performance Parameter Definitions (Rev 1)

Later, that working standard become an officially sanctioned standard by IEEE: 2700-2014 - IEEE Standard for Sensor Performance Parameter Definitions

Currently work for is underway for the next edition of the standard

For more information on inertial parameters IEEE Standards dedicated to


inertial sensors only are released and maintained by Gyro and Accelerometer panel

Some of the Sensor Parameter Definitions from IEEE 2700

Noise – The smallest measurable change in rotation rate expressed as the root mean square (RMS) and calculated as the standard deviation of a minimum of 10,000 sample points under vibration isolation and zero rotational input

Allan Variance – Allan variance is a time domain analysis technique used to determine the character of the underlying random processes that give rise to the data noise It helps identify the source of a given noise term present in thethe data noise. It helps identify the source of a given noise term present in the data

Zero Rate Bias – Zero rotation rate output deviation from expected zeroZero Rate Bias Zero rotation rate output deviation from expected zero rotation rate output value for each sensing axis


Some of the Sensor Parameter Definitions from IEEE 2700

Sensitivity – The change in rotational rate input corresponding to 1 least significant bit (unit in which digital values are counted) change in output.

Non-Linearity Error – Maximum deviation of measured output from the best fit straight line

Cross-Axis Sensitivity – Ratio of the measured rotation rate for an axis to the input rotational rate along each axis orthogonal to the measured axis.

F ll S l R P k t k t f th h Full Scale Range – Peak to peak measurement range of the sensor per each orthogonal axis.


Selection of Sensor Parameters for Testing

So which tests are appropriate for your sensor?– It all depends on the application!

In AHRS systems at least zero rate bias (over temperature) and sensitivity/non-linearity should be tested → they are the biggest error contributors to the final orientation angles

Noise and Allan variance measurements should be performed to identify “goodness” of the sensors, impact of noise on error budget, and long term tendency of the sensortendency of the sensor

For high speed rotation/high dynamic applications, bandwidth and full scale range are important → they define sensors ability to track the motion and

© 2016, ACUTRONIC

range are important → they define sensors ability to track the motion and identify the point at which the saturation of the output occurs

9


It is very important to keep in mind the end use of the device as that information can drastically decrease the complexity and duration of the tests

– In general, all MEMS sensors, including inertial, are temperature sensitive and performance will be affected by variations in temperature

– If the device is known to work only over a defined temperature range then testing should be performed over that range

E i t ith hi h ib ti t t l ti l i t f– Environments with high vibration content also negatively impact performance; if a device containing inertial sensors is expected to encounter high levels of vibration, then sensors should be tested over vibration environments



When doing thermal testing most companies perform testing at a limited number of discrete temperature points

– Most often 3 temperature points are used (maximum, minimum and room temperature)

– However, if the device is expected to be used over a broader temperature range it may be appropriate to perform thermal testing over the entire range while slewing, at a predetermined temperature rate, from one set point to anotheranother.

– This step could be used during characterization of the sensors



Benefits of doing more thorough thermal testing– It allows for screening of the unusual behavior of the sensors that would

otherwise be undetected by testing at a few discrete temperature points

– Slewing at a predefined ramp rate, for example 1 or 2 °C/min, could mimic the real environment that the device sensor will experience

– If you choose to collect data during the ramp up or down you will get a better model of sensor performance between the set temperatures.


As a recognized leader in

Test Selection Test Parameters Test Description Test Conditions Units

gthe inertial sensor testing industry, ACUTRONIC has developed their own standard suite of tests and conditions to comply with the

Noise and AllanVariance

Static Output for Sensor Measured over Long Time Duration

At room temperature, no physical stimulus, all axis

deg/s RMS, g RMSdeg/s/√Hz

Zero Input Bias Static Sensor Over -40 to +85 °C, deg/s, gconditions to comply with the standardized parameters specified in IEEE 2700 standard

Output with no physical stimulus applied

all axes

Sensitivity Change in Sensor Output per Change in Physical Input

Over -40 to +85 °C, all axes

deg/s/LSB, g/LSB

In order to provide a direct product test performance comparison, ACUTRONIC applies their own, standard set of test conditions to

in Physical Input

Non-Linearity Error Maximum deviation in sensor outputfrom a best fit line


% of Full Scale Range

set of test conditions to every test product.

The list of tests, conditions, and parameters is provided

Cross-Axis Sensitivity

% change in off-axis output to primary sensitive axis


% of Full Scale Range

Full Scale Range Peak to Peak Over -40 to +85 °C deg/s g

© 2016, ACUTRONIC

p phere:

13

Full Scale Range Peak to Peak measurement range of sensor output

Over 40 to +85 C, all axis equipment permitting

deg/s, g

Test Methodology

Several hardware platforms that can be used to test both the dynamic and static properties of MEMs sensors. It is possible to use 1-, 2- or 3-axis rate table test systems. Since the market trend is towards full 6 (9) degrees of freedom IMU, a 2-axis rate table provides enough flexibility to perform almost all test groups on all axes in a reasonableprovides enough flexibility to perform almost all test groups on all axes in a reasonable time frame. This section discusses test principles applied to such IMUs.

2-AXIS RATE TABLEACUTRONIC AC216


ACUTRONIC AC216

Test Methodology Static Tests25 ºC

-40 ºC85 ºCTemperature

Cycle

Zero Input Bias– For gyros, the zero input offset tests can be done with

sensitive axis in any orientation and no rotational motion of the rate table The data can be collected over a 3 minute period at

0 ºC45 ºCrate table. The data can be collected over a 3 minute period at each chosen temperature. At least 5 temperature points should be used, both extremes, ambient and two more temperature in between the 3 others.For accelerometers the sensitive axis has to be placed– For accelerometers, the sensitive axis has to be placed parallel to Earth’s gravity thus providing a zero G input. As such, the motion platform would have to be repositioned at least once to align the 3rd IMU axis parallel to the Earth’s gravity The same temperature profile can be used as forgravity. The same temperature profile can be used as for gyros.

– Both accelerometers and gyros can be tested at the same time during the static tests with the exception being one axis of

l h ld d ddi i l li i


an accelerometer that would need additional sampling time after proper alignment.

Test Methodology Static Tests25 ºC


Cycle

Noise Characteristics– These tests should be run at higher sensor sampling rates, as

close to sensor ODR as possible.O er a long period of time (3 to 6 ho rs) to allo eno gh data

0 ºC45 ºC– Over a long period of time (3 to 6 hours) to allow enough data

for statistical analysis.– No reorientation of IMU is needed and data sampling for

accelerometers and gyroscopes can be done at the same time.

– Noise is calculated in terms of the smallest change of rotation measurements in a rotation-resistant environment with no input rotation.

– The Allan Variance computation calculates the random variation in the sensor output due to white noise, random walk, bias instability, etc.

– All the above tests can be done over temperature as well thus


– All the above tests can be done over temperature as well thus making this a very lengthy portion of the testing.

Test Methodology Dynamic Tests25 ºC


Cycle

The dynamic tests typically consist of non-linearity, sensitivity, full scale range and bandwidth of a sensitive axis. In an IMU, it is also common to find factors such as cross axis sensitivity between sensitive axes which shows mechanical misalignment.

0 ºC45 ºC All of the above mentioned tests should also be performed over temperature to expose any temperature sensitivity. Conveniently, the results for each test parameter can be

obtained from a common motion profile performed on a 2 axis t d iti t bl l ll i l d i thrate and positing table as long as all axis are sampled in the

IMU.– For gyroscopes, the motion profile involves performing 5 to 10

rotation at a given rate going from negative maximum rate to positive maximum rate with a preset increment of some value ofpositive maximum rate with a preset increment of some value of deg/s.

– For accelerometers, the motion profile is even simpler and consists of a series of position indexes in the field of Earth’s gravity for each sensitive axis. As little as 4 index positions can be

d iti i t d t l h t i th t t


used per sensitive axis to adequately characterize the test parameters for that axis.

Test Methodology Dynamic Tests

One parameter that takes an exception to a single motion profile is full scale range.– For accelerometers, that have a full scale range

above 1G. Centrifugal force has to be used in order to obtain data at higher Gs. Centrifugal force is defined as the product of the radius and the square of the angular rate Figure to the rightthe square of the angular rate. Figure to the right shows an IMU at a radius to center of rotation. In this case Y axis will experience the G forces.

– For gyroscopes, remounting of the x and y axis maybe necessary to be in plane with table axis if the tilting axis is not able to achieve the same rates as the table axis. In the figure to the right z axis is in plane with the table axis

© 2016, ACUTRONIC

axis is in plane with the table axis.

18


Another parameter that usually takes an exception to using a rate table as a test platform all together is bandwidth

Since most accelerometers and gyros have higher bandwidth (in 100s of Hz) specification, the best tool for this job would be a linear and an angular vibration platform for accelerometers and gyroscopes respectivelyaccelerometers and gyroscopes respectively


ACUTRONIC 105-AVT


Bandwidth– The test profile for each

type of inertial sensor consists of a logarithmicconsists of a logarithmic or linear frequency sweep from 5 to 2000 Hz on each motion platform while recording the IMU output at maximum sampling rate possible (as close to sensor ODR as possible).


Test Procedure

Following the methodology described in previous slides ACUTRONIC has Following the methodology described in previous slides ACUTRONIC has tested three 3-axis MEMs gyroscopes from Bosch Sensortec, InvenSenseand STMicroelectronics

The devices used were BMG160, MPU-3050 and L3G3200D

Bandwidth tests were not performed on these devicesBandwidth tests were not performed on these devices


Test Procedure# DevicesModel Manufacturer # Devices Available* Gyroscope Evaluation/Shuttle Board

BMG160 Bosch Sensortechttp://www.bosch-sensortec.com 3

MPU-3050

InvenSense Inc.www.invensense.com 4

L3G3200D STMicroelectronics

www.st.com 3


Test Procedure – UUT Orientation

ACUTRONIC 2-Axis AC277 Rate Table

MountedBosch BMG160

Axis 1 - CW - Positive (+) Z Negative (-)

MountedOrientation

Onboard ChipOrientation

Axis 1 - CCW - Negative (-) Z Positive (+)

Axis 2 - CCW - Positive (+)* Y Positive (+)

Axis 2 - CW - Negative (-)* Y Negative (-)

Axis 3 - CCW - Positive (+)** X Negative (-)

Axis 3 CW Negative ( )** X Positive (+)Axis 3 - CW - Negative (-)** X Positive (+)

*These results assume ACUTROL Axis 1 is at 0 º.

**The AC277 only has two axes. In order to test the 3rd gyroscope axis, we rotated the ACUTROL Axis 1b 90º d d A i 2


by 90º and used Axis 2.



MountedInvensense MPU3050


MountedOrientation



Axis 2 - CCW - Positive (+)* Y Positive (+)

Axis 2 - CW - Negative (-)* Y Negative (-)

Axis 3 - CCW - Positive (+)** X Negative (-)

A i 3 CW N ti ( )** X P iti ( )Axis 3 - CW - Negative (-)** X Positive (+)







MountedSTMicroelectronics

L3G3200D


MountedOrientation



Axis 2 - CCW - Positive (+)* Y Negative (-)

Axis 2 - CW - Negative (-)* Y Positive (+)

Axis 3 - CCW - Positive (+)** X Positive (+)

A i 3 CW N ti ( )** X N ti ( )Axis 3 - CW - Negative (-)** X Negative (-)





The testing is divided into three sessions:

Test Procedure 25 ºC

The testing is divided into three sessions: – First Session:

1. Data is recorded during all stages of the test at 50 Hz via I2C.a) All devices are mounted on a rate table inside a temperature

-20 ºC

-40 ºC85 ºC

65 ºC

TemperatureCycle

chamber. Starting at 25 degrees C, the temperature is controlled at +/-1 °C/min. Data is recorded during the transition periods. 0 ºC45 ºC

b) The temperature cycles negative and then positive stopping at the followingthen positive stopping at the following key temperatures in Celsius: 25, -20, -40, 0, 45, 85, 65, 25. At each stable temperature, a soak of 1.5 hours with no rotational input is recorded.

b) After each soak, before progressing to the next temperature step, the following input rates are applied to the devices for 5 revolutions per rate:


-2000 to 2000 deg/s (at +/-100 deg/s intervals) for Z axis. -500 to 500 deg/s (at +/- 100 deg/s intervals) for X and Y axes.

Test Procedure

Second session: Second session: 2. Data is recorded during each segment of the test at 50 Hz via I2C.

a) The devices are physically rotated to test each device axis at the full scale range of +/-2000 deg/s. Data is recorded for 2.5 minutes at each extreme.


Test Procedure

Third session: Third session: 3. Data is recorded at device native frequencies of 1 kHz via I2C.

a) Devices are placed in a temperature and rotational resistant environment. Data is recorded for 90 minutes with no input rotation.


Test Procedure – Results Noise Tests

RMS Noise calculated for each type of sensor


Test Procedure – Results Noise Tests Allan Variance calculated for each type of sensoryp

More data on:Noise and Allan VarianceNoise and Allan VarianceZero Input Bias SensitivityNon-Linearity ErrorCross-Axis SensitivityPerformance over temperatureavailable from ACUTRONIC


Test Procedure – Results Dynamic Tests Full Scale Range for each type of sensorg yp


Test Procedure – Results Dynamic Tests Cross-Axis Sensitivity for each type of sensory yp

BMG160 MPU3050 L3G3200DParameter Units Mean Mean MeanNoise gyroX deg/s RMS 0.243 0.094 0.202Noise gyroY deg/s RMS 0.269 0.091 0.247Noise gyroZ deg/s RMS 0.222 0.080 0.667gy g/AllVar gyroX deg/s/√Hz 0.047 0.020 0.092AllVar gyroY deg/s/√Hz 0.054 0.020 0.040AllVar gyroZ deg/s/√Hz 0.036 0.017 0.116

Zero Rate Bias gyroX deg/s ±2.63 ±1.744 ±2.139Zero Rate Bias gyroY deg/s ±2.593 ±0.880 ±0.671Zero Rate Bias gyro Z deg/s ±3.010 ±1.582 ±1.203

Z R t Bi T C f X d / /d C ±0 002 ±0 021 ±0 032

More data on:Noise and Allan VarianceZeroRateBiasTempCoef gyroX deg/s/degC ±0.002 ±0.021 ±0.032

ZeroRateBiasTempCoef gyroY deg/s/degC ±0.005 ±0.023 ±0.014ZeroRateBiasTempCoef gyroZ deg/s/degC ±0.001 ±0.106 ±0.023

Sensitivity gyroX deg/s/LSB 0.061 0.060 0.063Sensitivity gyroY deg/s/LSB 0.061 0.064 0.063Sensitivity gyroZ deg/s/LSB 0.061 0.062 0.070

SensitivityTempCoef gyroX %/degC ±0.010 ±0.016 ±0.023

Noise and Allan VarianceZero Input Bias SensitivityNon-Linearity Error

SensitivityTempCoef gyroY %/degC ±0.003 ±0.010 ±0.027SensitivityTempCoef gyroZ %/degC ±0.001 ±0.003 ±0.008

Integral NLE gyroX % FSR ±0.04 0.049 0.337Integral NLE gyroY % FSR ±0.049 0.062 0.151Integral NLE gyroZ % FSR ±0.094 0.017 0.118CrossAxis gryoXY % ±0.075 ±0.522 ±1.235CrossAxis gyroXZ % ±1 781 ±1 162 ±0 452

Cross-Axis SensitivityPerformance over temperatureavailable from ACUTRONIC


CrossAxis gyroXZ % ±1.781 ±1.162 ±0.452CrossAxis gyroYZ % ±0.649 ±3.719 ±0.627CrossAxis gyroYX % ±0.077 ±0.775 ±1.041CrossAxis gyroZX % ±1.708 ±0.714 ±0.701CrossAxis gyroZY % ±0.427 ±2.051 ±0.105

Conclusion – Testing Approaches

Integrator of MEMS inertial sensors should select testing parameters based on expected application of the device containing MEMS inertial sensors

IEEE 2700 Standard should be used to define parameters that are to be tested

Test methodology needs to be defined, Acutronic can help

Based on test results, developers and integrators can make sensor selection b d ifi d f th t li ti (i l i l bi hi hbased on specific needs for that application (i.e. low noise, low bias, high sensitivity, etc.)

This type of testing provides additional differentiation for the MEMS sensors


This type of testing provides additional differentiation for the MEMS sensors selection process (in addition to cost, size and power requirements)

Conclusion – Testing Approaches

For example, in the data set one type of sensor (InvenSense MPU3050) has significantly lower Noise RMS values than the other two sensors– If RMS Noise is a very important parameter for the given application this

sensor would have been the choice for that application

Testing also allows comparison of actual live sensors data to manufacturer’s bli h d ifi tipublished specifications

– It provides valuable measurement of “quality” of performance

With th t i i d ACUTRONIC h t l b (iTS L b) t id t ti With that in mind ACUTRONIC has set up a lab (iTS Lab) to provide testing services to encourage wider application of the IEEE 2700 Standard– iTS Lab enables such testing for all types of inertial MEMS sensors


www.acutronic.com

Thank you for your attention!For more information contact:For more information contact:Dino [email protected] 926 1192412-926-1192


testing approaches for characterization and selection of ...meptec.org/resources/9 -...

Documents