Trans. JSASS Aerospace Tech.

Japan
Vol. 17, No. 4, pp. 421-426, 2019
DOI: 10.2322/tastj.17.421

The Study on Performance of MEMS IMU for Launch Vehicle

under High Vibration Environment

By Eri SHIMANE,1) Shuichi MATSUMOTO, 1) Takafumi MORIGUCHI, 2) Yuzo IWAI,2) and Ryohei UCHINO 2)
1)
Research and Development Directorate, JAXA, Tsukuba, Japan
2)
Sumitomo Precision Products Co., Ltd., Amagasaki, Japan
(Received June 30th, 2017)

Along with the improved accuracy of MEMS gyroscopes and accelerometers in recent years, JAXA has been studying
a navigation-grade Inertial Measurement Unit (IMU) using MEMS gyroscopes and accelerometers for launch vehicles. One
issue regarding application of the MEMS IMU for launch vehicles is how to maintain measurement accuracy under high
vibration environment during the launch phase. We thus developed a trial model of high accuracy MEMS IMU and
evaluated its measurement accuracy under a high vibration environment by conducting random vibration tests. This paper
presents the issues of the MEMS IMU for launch vehicles, trial model of the MEMS IMU, and evaluation results of the
MEMS IMU under a high vibration environment.

Key Words: MEMS, IMU, Gyro, Navigation, Rocket

Nomenclature environmental resistance. Such requirements are why existing

IMUs for launch vehicles are large in size and costly.
Ab : acceleration bias The gyroscopes and accelerometers developed for
Abcorr : corrected acceleration bias commercial use in car navigation, the controllers of video
aM : averaged acceleration in major cycle games, and smartphones, however, have become smaller and
am : measured acceleration in minor cycle less expensive. gyroscopes and accelerometer using Micro
N dif : number of major cycles using bias Electronic-Mechanical Systems (MEMS) technology are used
estimation at each major cycle for such applications. MEMS devices can be mass-produced
N init : number of major cycles using in initial bias
at low cost, giving the use of the same production process
estimation
nminor with semiconductors. Moreover, the accuracy of MEMS
: number of minor cycles in major cycle
gyroscopes and accelerometers has been improved in recent
Rb : rate bias
years.
R bcorr : corrected rate bias
We have thus been studying way to improve MEMS
Tmajor : duration time of major cycle
gyroscope accuracy and apply the IMU using MEMS inertial
tfnovib , ts novib : end and start time to estimate preclusive sensors, which we called the MEMS IMU, for launch
bias error caused by vibration machine
excitation vehicles.
Ab , R b : preclusive acceleration and rate bias error We also developed a trial model of the MEMS IMU and
caused by vibration machine excitation evaluated its performance under the flight conditions of
 : integrated angle launch vehicles, such as a high vibration environment and
v : integrated velocity radiation environment.
M : averaged angular rate in major cycle
m : measured angular rate in minor cycle 2. Issues of MEMS IMU Application for Launch Vehicles
Subscripts
i : in i-th minor cycle Since launch vehicles fly in a special environment that
init : in initial bias estimation entails high vibration, a wide temperature range, and a severe
j, k : in j-th and k-th major cycle radiation environment, the IMU for a launch vehicle needs to
be resistant to such a severe environment and must maintain
1. Introduction highly accurate inertial measurements. The technical issues
for a launch vehicle’s IMU are as follows:
The Inertial Measurement Unit (IMU) equipped with
(1) Highly accurate inertial measurements for launch vehicle
gyroscopes and accelerometers is one of most important
guidance
components for the guidance, navigation and control system
(2) Maintaining accuracy for a wide temperature range
of launch vehicles. The inertial sensors that obtained
(3) Maintaining accuracy for a high vibration environment
information about the angular rate and acceleration of launch
(4) Ensuring normal operation in a severe radiation
vehicles require high accuracy, high reliability, and
environment

Copyright© 2019 by the Japan Society for Aeronautical and Space Sciences and ISTS. All rights reserved.
J-STAGE Advance published date: January 31st, 2019

421
 Trans. JSASS Aerospace Tech. Japan Vol. 17, No. 4 (2019)

Highly accurate inertial measurements of IMU is necessary IMU whose model number is HGM-02A. HGM-02A uses the
for launch vehicle to perform launch guidance for orbit same type of MEMS gyroscope as MARS and doesn’t have
insertion. Typical requirements for the inertial measurements the digital temperature correction circuit used on MARS.
for launch vehicle 1) is as follows: Figure 2 shows its appearance and Table 1 lists its
(a) Gyro bias error is less than 0.1 deg/h (3σ) specifications. The Allan variances of MEMS gyroscope and
(b) Accelerometer bias error is less than 130 μG (3σ) accelerometer measurements under a static state on
HGM-02A are shown in Figs. 3 and 4. We can recognize that
Given the recent improvements made in MEMS gyroscope the bias instability of the MEMS gyroscope measurement is
and accelerometer performances, the MEMS IMU that we less than 0.1 deg/h and that the bias instability of the MEMS
have been developing using such highly accurate MEMS accelerometer measurement is less than 0.1 mG. HGM-02A
inertial sensors can achieve the gyro and accelerometer bias samples and quantizes gyroscope and accelerometer
accuracy in ideal static conditions and will soon be available measurements at 500 Hz, averages the quantized data, and
for the launch guidance of small rockets with a short mission outputs the data at 50 Hz. The bandwidths of rate and
time. acceleration output data are set to about 20 to 50 Hz.
The second issue of accuracy for wide a temperature range One of the main issues of the MEMS IMU relative to the
can be solved by using the temperature correction technique. flight conditions of launch vehicles is the high vibration
To solve this issue, we have developed a high accuracy environment. We conducted random vibration tests using the
MEMS angular rate sensor called MARS.2,3) The high MEMS IMU trial model to determine whether it can be resist
accuracy of MARS is achieved by digital temperature the vibration environment of the H-IIA rocket and maintain its
correction, and the latest model of MARS, MARS III, performance.
achieved low bias instability of 0.1 deg/h.4) Figure 1 shows the
Allan variance of the MARS III breadboard model with a full
temperature range. This result is similar to the Fiber Optical
Gyroscopes (FOG) in terms of performance, and we evaluated
that the MEMS gyroscope potentially has sufficiently high
accuracy to guide and navigate launch vehicles.
The third issue of accuracy for high vibration is severe
because MEMS gyroscopes and accelerometers have a
mechanical structure, and an anti-vibration mount cannot be
used for inertial measurement. Thus, we evaluated the degree
of deteriorated accuracy of the MEMS IMU under a high Fig. 2. Trial model of the MEMS IMU (HGM-02A).
vibration environment by conducting random vibration tests,
which is a main topic of this paper. The fourth issue of the Table 1. Specifications of the MEMS IMU (HGM-02A).
radiation environment is another serious problem to be solved
Items Specifications
regarding the MEMS IMU for launch vehicles, and is now
Size 74(W)×74(L)×60(H) [mm]
being studied. This paper thus focuses on the issue of
Weight 500 [g]
maintaining accuracy in a high vibration environment.
Angle rate measurement range ±100 deg/s
Acceleration measurement range ±2 G

Fig. 1. Allan variance of the MARS III breadboard model.4)

3. Trial Model of High Accuracy MEMS IMU

Fig. 3. Allan variance of MEMS gyroscope of HGM-02A.

To evaluate accuracy and environmental resistance in the

IMU configuration, we developed a trial model of the MEMS

422
 Trans. JSASS Aerospace Tech. Japan Vol. 17, No. 4 (2019)

Overall Acceleration RMS: 79 m/s2 (rms)

Fig. 4. Allan variance of MEMS accelerometer of HGM-02A. Fig. 6. Random vibration level profile for H-IIA onboard components.

4. Vibration Test Results and Data Analyses 4.2. Data acquisition results of vibration tests
In the vibration tests, the angular rate and acceleration
4.1. Vibration test setup and conditions measurements of the MEMS IMU were obtained with a
To evaluate the angular rate and acceleration measurement sampling rate of 50 Hz. Figure 5 shows one of the vibration
accuracy of the MEMS IMU in a high vibration environment, test results using the H-IIA vibration level profile shown in
we conducted a vibration test using HGM-02A the trial model Fig. 6. About 50 seconds before the start of vibration, the
of MEMS IMU at the Tsukuba Space Center in February 2017. vibration machine table was excited by electromagnetic force
Figure 5 shows a photo of the vibration test setup. As this for vibration motion, which is indicated as a symbol, ‘ △ ’ , in
vibration test was conducted as a part of the vibration test Fig. 7. Vibration motion starts from a level of -24 dB for the
campaign for commercial-off-the-shelf (COTS) navigation target vibration level and its vibration level is increased in
sensors, COTS global navigation satellite system (GNSS) steps of 6 dB until reaching the target vibration level. As
modules were tested with the MEMS IMU. The MEMS IMU Fig. 7 shows the test results of z-axis vibration, the increase in
was set at the center of the vibration test table as shown in vibration level can be seen in the z-axis acceleration data
Fig. 5. Although many test patterns and vibration levels were shown at the bottom right in Fig. 7. The vibration start time is
used in this vibration test campaign, we employed a random indicated by symbol, ‘ ▲ ’ , and the duration of the target
Overallvibration levelRMS:
Acceleration profile
79 for
m/s2H-IIA
(rms) onboard components as a vibration level is indicated by symbol, ‘ ◆ ’ .
reference in this paper. Figure 6 shows the random vibration From Fig. 7, we can recognize that the measurement data
level profile; its overall acceleration root-mean-square (RMS) sampled at 50 Hz contains much noise, thus making it difficult
is 79 m/s2 (rms). This paper presents and discusses the to evaluate small bias errors of the high accuracy MEMS IMU
vibration test results of z-axis vibration, which is from raw measurement data.
perpendicular to the vibration test table.

MEMS IMU

Vibration
Testing Machine Fig. 7. Angular rate and acceleration raw measurement data of the
Fig. 5. Vibration test setup. MEMS IMU (sampling rate of 50 Hz).

423
 Trans. JSASS Aerospace Tech. Japan Vol. 17, No. 4 (2019)

4.3. Data analysis methods and results of vibration tests (8). Ndif in Eqs. (7) and (8) denotes the number of major
To evaluate small bias errors on the high accuracy MEMS cycles using bias estimation at each major cycle and is set to
IMU, the noise in raw measurement data must be reduced. An 10, indicating an averaging time of 10 seconds in this paper.
averaging technique is used to reduce the noise in the data;
moreover, data integration technique is employed to extract k  k  N dif
small bias errors. In this paper, raw measurement data of 50 R bk  (7)
N dif  Tmajor
Hz sampling is called minor cycle data, and the averaged data
of measurement data for 1 second is called major cycle data, vk  vk  N dif
Abk  (8)
which means that the minor cycle is 50 Hz and the major N dif  Tmajor
cycle is 1 Hz. The averaged angular rate (M) and averaged
acceleration (aM), are obtained in Eqs. (1) and (2) below.

nminor
M j  n 1 minor

i 1
m ji (1)

nminor
a M j  nminor
1
a
i 1
m ji (2)

Although averaging is very useful for reducing noise in

measured data, it is not enough to extract small bias errors on
the high-performance MEMS IMU. Thus as the next step, we
employed a data integration technique for each axis.
Integrated angle () , and integrated velocity () are defined
in Eqs. (3) and (5). Because MEMS gyroscopes and
accelerometers have initial bias errors that are to be canceled
by initial alignment operation, the integrated angles and
accelerations are integrated after subtracting initial bias errors
Fig. 8. Integrated angles and velocities of the MEMS IMU
as in Eqs. (3) and (5). The initial bias errors are estimated by corresponding to measured data in Fig. 7.
averaging more than 100 seconds before vibration operation
starts as shown in Eqs. (4) and (6). More than 100 major
cycles are used in initial bias estimation (Ninit).

k
k  Tmajor   ( M j  R binit ) (3)
j 1
N init
R binit  1
N init 
j 1
Mj (4)

k
vk  Tmajor   (a M j  Abinit ) (5)
j 1
N init
Abinit  1
N init a
j 1
Mj (6)

Figure 8 shows the integrated angles and velocities

corresponding to the measured data in Fig. 7. The small slopes
Fig. 9. Estimated angular rate biases and acceleration biases
starting from the initial flat zero in Fig. 8 are caused by small corresponding to measured data in Fig. 7.
bias error in the angular rates and accelerations of the MEMS
IMU. From Fig. 8 we can recognize small bias errors in the Figure 9 shows the estimated angular rate biases and
integrated angles of all axes and the integrated velocities of acceleration biases corresponding to the measured data in
the x-axis and y-axis and relatively large bias error in the Fig. 7. From Fig. 9, we can clearly recognize small bias errors
integrated velocity of the z-axis. We can also see that small in the integrated angles of all axes and integrated velocities of
bias errors arise not from the start of vibration, but from the the x-axis and y-axis, and relatively large bias error in the
start of vibration machine excitation. integrated velocity of the z-axis. We also find that small bias
To quantize the bias errors during vibration motion, errors arise not from the start of vibration, but from the start of
averaged bias errors are calculated using the time difference vibration machine excitation. We assume that such small bias
of integrated angles and velocities as shown in Eqs. (7) and errors after the start of vibration machine excitation are due to

424
 Trans. JSASS Aerospace Tech. Japan Vol. 17, No. 4 (2019)

electromagnetic sensitivity of the MEMS IMU and

electromagnetic radiation from the vibration testing machine
excited by electromagnetic force for vibration motion.
Separately from this paper, we will confirm the cause of these
bias errors after the start of vibration machine excitation. In
this paper, these bias errors after the start of vibration machine
excitation are treated as fixed bias errors to be excluded for
the evaluation of MEMS IMU accuracy under a high vibration
environment. Although the bias errors caused by
electromagnetic radiation from the vibration testing machine
could become larger as the vibration level went up, it is H-IIA vibration level is
0 dB (79 m/s2 (rms))
difficult to distinguish between the increase of bias errors due
to vibration itself and the increase of bias errors caused by
electromagnetic radiation. Thus this paper evaluates the bias
errors of IMU during vibration as the overall bias errors that
exclude the fixed bias errors caused by vibration machine Fig. 10. Estimated angular rate bias errors caused by vibration motion.
excitation and include the increase of bias errors caused by
electromagnetic radiation during vibration. 5. Accuracy Evaluation under High Vibration Environment
To estimate preclusive bias errors after the start of vibration
machine excitation, we extracted the period when the 5.1. Angular rate accuracy evaluation under high vibration
vibration machine was excited without starting vibration As shown in the z-axis vibration test data evaluation results
motion. The preclusive bias errors after the start of vibration described in Section 4.4, the maximum angular rate bias error
machine excitation are estimated by averaging bias errors caused by the H-IIA level vibration is about 3 deg/h, which is
during the extracted period as shown in Eqs. (10) and (12). larger than that of the H-IIA IMU. However we believe that
And angular rate bias errors and acceleration bias errors the MEMS gyroscope of the MEMS IMU trial model can be
caused by vibration motion are estimated by subtracting the applied to a small rocket having a short mission time and not
preclusive bias error as shown in Eqs. (9) and (11) below. very high orbital insertion accuracy, and to an onboard flight
path monitoring system for rocket flight safety using a global
R bcorr k  R bk  R b (9) positioning system (GPS) receiver.
 tf novib   tsnovib 5.2. Acceleration accuracy evaluation under high vibration
R b  (10) As shown in the bottom-right plot in Fig. 9, the z-axis
tf novib  ts novib acceleration bias is very large and we consider it difficult to
Abcorr k  Abk  Ab (11) use the acceleration outputs of the MEMS IMU trial model
under a high vibration environment as inertial measurements
vtf  vtsnovib
Ab  novib (12) for launch vehicle navigation. We figured out the root cause
tf novib  ts novib of the large z-axis bias error as being resonance between the
MEMS accelerometer and printed wiring board (PWB). As
4.4. Angular rate bias errors caused by vibration motion MEMS accelerometers are attached directly on a PWB, the
The same data analyses described in Section 4.3 were MEMS accelerometer resonated with the PWB at the PWB’s
conducted for all valid z-axis vibration test data using the natural frequency. To cope with this problem, we changed the
H-IIA vibration level profile and its level-down profiles. design so that MEMS accelerometers are installed on a sensor
Figure 8 shows the estimated angular bias errors caused by mount.
vibration motion. The H-IIA vibration level whose overall
acceleration RMS is 79 m/s2 (rms) corresponds to ‘0 dB’ in 6. Conclusion
Fig. 10. From Fig. 10, we can recognize that angular rate bias
errors caused by random vibration motion are proportional to We have tested the MEMS IMU trial model in the
the vibration level (i.e. acceleration spectral density). Using vibration environment of the H-IIA vibration level. From the
the estimated angular rate bias errors of the z-axis shown in vibration tests and evaluation of its data, we extracted
Fig. 10, vibration rectification error (VRE) is estimated to be technical issues to be solved in MEMS IMU system design
about 0.5 deg/h/Grms. and recognized the accuracy of the MEMS IMU under a high
vibration environment. We conclude that the design of the
MEMS IMU trial model, HGM-02A, and the modified
MEMS accelerometer mount design can be applied to a small
rocket having a short mission time and not very high orbital
insertion accuracy. We will continue to study the accuracy
and environmental resistance of the MEMS IMU toward
application of the MEMS IMU for launch vehicles.

425
 Trans. JSASS Aerospace Tech. Japan Vol. 17, No. 4 (2019)

References

1) Matsumoto, S., Komatsu, M., Shimane, E., Kobayashi, Y., and

Yamazaki, S.: Effectiveness and Limitation of IMU-GPS
Integrated Navigation and GPS Attitude Determination for
Launch Vehicle Guidance, Proceedings of 10th International ESA
Conference on Guidance, Navigation, and Control System,
Salzburg, 2017.
2) Sasada, T., Shimane, E., Nishida, H., Moriguchi, T., Uchino, R., 
and Magosaki, F.: An Interim Report on the Development of
High-Accuracy MEMS Gyros for Space Applications, Proceedings
of 57th Space Sciences and Technology Conference, Yonago,
2013, 2G02 (in Japanese).
3) Sasada, T., Shimane, E., Nishida, H., and Moriguchi, T.:
Development of High-Accuracy MEMS Gyros for Space
Applications, Proceedings of 9th ESA Round Table on Micro and
Nano Technologies for Space Applications, Lausanne, 2014.
4) Sasada, T. and Nishida, H.: Adaptation of Navigation-grade
MEMS Gyros for Future Launch System, Proceedings of 30th
International Symposium on Space Technology and Science,
2015-g-15, Kobe, 2015.

426