Sie sind auf Seite 1von 8

Using Low-Cost ME Accelero

Strapd
Bradford S. Davis U S Army Research Laboratory, Weapons and Materials Dir., AMSRL-WM-BA, .. Aberdeen Proving Ground, MD 21005-5066
Abstract:

Low-cost, micromachine4 inertial sensors have been steadily emerging into the commercial marketplace. Some of these commercial off-theshelf (COTS) sensors were evaluated for their insertion potential into military applications such as test and evaluation (T&E) and smart munition guidance. Performance requirements to provide time, space, position, and attitude information (TSPAI) and allow for navigation capabilities are fast approaching those needed, especially when integrated with a global positioning system (GPS) receiver. Artillery projectiles and rockets, instrumented with commercially available low-cost microelectromechanical systems (MEMS) accelerometers, spin sensors, and custom telemetry units, have been flight tested with good success. Test results suggest that many of these sensors are rugged enough for launch survivability and subsequent operation. Analysis of the accelerometer flight data shows good agreement with ground-based radar and other truth measurements. Ground tests of a MEMS gyroscope on a ballistic flight simulator have also been conducted. Tbe results show promise for directly measuring a projectiles pitching and yawing behavior. This paper will present some of the test results and describe the challenges for using these devices as strapdown inertial measurement units (IMUs) on rolling projectiles. INTRODUCTION During the past several years, the Advanced Munitions Concepts Branch (AMCB), U.S. Army Research Laboratory (ARL), has been assessing low-cost, inertial devices for potential military applications. The low cost, small size, and low power of MFMS make them excellent candidates as IMUs for a variety of Department of Defense @OD) and Army applications. A major impetus for the early efforts was geared toward the Armys Subminiature Telemetry and Sensor Systems (HSTSS) program [l]. For this application, the sensors must survive harsh launch environments and support airbome telemetry measurements. Projectile types of interest range

from rockets that are launched at hundreds of gs to tank munitions that are launched at tens of thousands of gs. Other Army programs are interested in low-cost guidance, navigation, and control (GNC) for competent munitions [2]. For these applications, fuze retrofits will house the inertial sensors and maneuverable mechanisms or thrusters allowing for course correction. This technology will provide affordable precision strike without expensive seeker systems. Over the past few years, a number of ground and flight tests have been conducted to verify that these sensors could survive high-g launch and operate on a rolling projectile. Extensive ground tests have consisted of environmental, shock table, air gun, and flight simulator testing of accelerometers, gyroscopes, and spin sensors. Artillery projectiles and rockets were instrumented with accelerometers t measure launch and flight accelerations o and augmented with telemetry systems to transmit the data.

SENSOR DESCRIPTIONS
Accelerometers
Today, there are about ten manufacturers making MEMS accelerometers, primarily for the automotive air bag market. Analog Devices (AD) was one of the first to manufacture such a lowcost surface-micromachined directcurrent @C) accelerometer. The device comes complete with all conditioning circuitry built onto the same chip. Their accelerometer line and measurement mges have been expanding from the ADXL05 and ADXL50 in a 10-pin To100 metal can (Figure 1, top) to the newer ADXL150 axl ADXL181 in a 14-pin ceramic package (Figure 1, bottom). They currently range from 5 to 500 gs. Performance specifications are close to those required for T&E applications. prices for these accelerometers are between $20 and $30. AMCB obtained a small quantity of accelerometers to demonstrate high-g launch survivability and subsequent operation in low-g flight environments typical of artillery projectiles. MEMS accelerometers from

594
0-7803-4330-1/98/ $10.00 0 1998 IEEE

Motorola and Endevco have also been ground and flight tested with much success. Still others have been surveyed for their performance specifications, but not yet tested.

strapdown projectile applications. Figure 2 shows the prototype gyroscope made up of a 2.25-in printed circuit board (PCB) enclosed in a plastic housing.

Figure 2. DraperBoeing prototype MEMS gyroscope. Figure 1. ADXL family of accelerometers.

Roll Rate Sensors


On rolling projectiles, the influence of spin accelerometers and gyroscopes is dramatic as will evidenced in other sections of this paper. A measure of projectiles spin rate will be required to compensate these spin induced errors. on be the for

Gyroscopes
MEMS rate sensor development is trailing that of the accelerometers predominantly because they are electronically more complex. There are a few companies with viable prototypes. Research and development, fueled by the auto industry, is directed at active handling and platform stabilization systems. Auto specifications for the gyros bias stability is 1 /sec. That is close to those needed for rocket/missile IMUs but far away from projectiles with a long flight time. Defense Advanced Research Projects Agency (DARPA) MEMS is supporting the production of such a micromachined silicon inertial rate sensor. The MEMS gyroscope is being developed through an alliance between Boeing North American and Draper Laboratory, as reported by Connelly and Brand [3]. The projects intent is to transition the MEMS technology from a research and development environment into a COTS product. AMCB is a DOD advisory member to identify potential military applications. It is anticipated that the military version will use the COTS technology as a building block. The Technology Reinvestment Program will identify the necessary modifications to the packaging and performance through testing of the automotive quality gyros. The fiist batch of gyro prototypes was produced in early 1997. T e h t prototypes have undergone extensive qualification testing a Boeing. AMCB was given a sample to test in May 1997. The intent was to verify Boeings preliminary performance specifications and perform additional specialized tests designed to check the gyroscopes capabilities as an IMU for

The SCSASO is a miniature, low-cost, magnetic angular rate sensor that is capable of providing one count per revolution of a projectile spinning in space. Under an engineering services contract, sensors were supplied to d AMCB for gyroscopic, high-g shock, performance, a flight testing with good results [4]. The sensor bandwidth is set for 35 - 1,000 Hz and draws very little current. Sensor Applications designed the sensor using giant magnetoresistance ratio (GMR) materials. The sensor is capable of providing both a digital and an analog output, referenced to half of the supply voltage. The expected cost of the sensor will be around $50. Magnetic spin sensors with ranges down to 0 Hz are also under development.

Figure 3. Magnetic field sensor. HIGH-G SHOCK TESTING Prior to actual flight testing, components are first subjected to a laboratory simulation of launch accelerations on a

595

shock table and then graduate to an air gun to assess survivability limitations. Over the past few years, a number of MEMS accelerometers have been shock tested. Due to the their small size and low weight, results have shown durability. Unpowered ADXL05 and ADXLSO accelerometers weze tested on AMCBs IMPAC shock table during tests performed by Davis [5]. Accelerometers were oriented in both the axial and transverse orientations within an aluminum fixture and subjected to shocks up to 35 kgs with <l-ms duration. The shock table is pictured in Figure 4. The shock amplitude was increased with each test until failure occurred Minimum shock with a failure occrrrred near 20 kgs. Even with limited sample size, it was clearly evident that these accelerometers were consistently able to survive extreme artillery-level gun launch accelerations. Follow-on air gun tests by Davis et al. [6] also showed further resilience to shock. A limited number of units were shocked up to 95 kgs with l-ms duration. Failure occurred at a minimum shock of 50 kgs. When failure did occur, the cause was by stiction or a detached beam. More recently, ADXL150 and 181 accelerometers have been shock-table and air-gun tested with more elaborate test setups. Accelerometers were powered and their output was recorded. Figure 5 shows the raw data for a typical accelerometer before, during, and after shock-table testing. The initial 0-g bias level was 2.5 V. The accelerometer was clipped during the shock but then recovered. It rang during vibrational movement of the table until it settled. Scale factor errors and 0-g bias shifts were measured and found to be minimal. This agreed with airgun tests of unpowered ADXL150s and ADXL18ls performed by Davis et al. [7].

10 Time (msec)

15

20

Figure 5. Accelerometer output from shock-table test. FLIGHT TEST DESCRIPTION The AMCB electronics team routinely builds customized instrumentation/telemetry packages for the T&E community and DOD contractors. Over the past three years, a number of separate flight tests were conducted to demonstrate launch survivability and operation on both rocket and artillery bodies. Figure 6 shows the typical hadware consisting of accelerometers, telemetry components, associated FTBs, and a power supply housed in an artillery fuze body. The flight tests have included axially and radially oriented micromachined accelerometers in single and multiple combinations. Yawsonde sensors were also included as a truth measurement. They provided a detailed knowledge of the yaw and spin history of projectile flights [SI. The fuzeanfigured instrumentation packages were flown on 155-mm and 105-mm artillery projectiles. Power was supplied by nickel cadmium (nicad) batteries so that the telemetry transmitted data immediately out of the gun. The raw sensor data were signal conditioned, passed through a voltage controlled oscillator (VCO), mixed, and transmitted to a ground receiving station. The data were recoved by the ground station instrumentation (consisting of an telemetry receiver, and discriminator), processed by a data acquisition system, and reduced with graphing and data analysis software.

Figure 4. Shock table apparatus.

596

Figure 6. Flight instrumentation package assembly. CALIBRATION AND COMPENSATION

Scale Factor and 0-g Bias Level Determination


Before each firing, the low-g ranged accelerometers were statically calibrated at the gun site. A +l-g calibration test was performed by placing the package nose up and nose down. The output was used to obtain zero-g bias level and the scale factor calibration. The calibration was performed as close in time to the launch as possible. This would ensure minimal 0-g bias shift due to thermal heating. For the high-g ranged accelerometers, like the ADXL181, laboratory centrifuge tests were necessary to determine the ee scale factor. After the flight instrumentation packages w r assembled, they were screwed onto the spin fixture shown in Figure 8. The accelerometer was offset from the spin axis by a known radius. By varying the spin rate ancl recording the accelerometer output, the scale factor could be easily calculated. The misplacement radial offset of the accelerometer within the fuze body package was also determined from similar tests.

Error Sources
Knowing the sensors exact location and orientation relative to the projectile's center of gravity and spin axis is critical to making a quality measurement. Angular misalignment and cross-axis sensitivity of the sensor are two such emlr sources that cause acceleration or rate orthogonal to the measurement of interest to migrate into the measurement causing scale factor and/or bias errors. Sensitivity to changes in temperature can also cause bias and scale factor errors. Radial offset of the sensor from the projectile's axis of rotation can introduce errors under spin. The radial offset can be a combination of mechanical misplacement (i.e., human placement and/or machining tolerances of the ogive/fuze thread interface) and unbalance (i.e., static and/or dynamic) of the projectile (see Figure 7). Unbalance causes the projectile to wobble around the mass center of gravity (CG) instead of the geometrical CG. Rotation introcluces centrifugal acceleration that is proportional to the radial offset multiplied by the angular velocity squared, causing a bias error. This acceleration is measured directly by an accelerometer with its sense axis oriented perpendicular to the projectile's spin axis. An axially oriented accelerometer will sense a component of this centrifugal acceleration through cross-axis sensitivity and/or misalignment. Compensation of the these errors can be made if their values can be determined through direct measurements or approximated through simulation.
(EffectsExaggerated)
Offset Radius , Maw Center of Gravity

Figure 8. Centrifuge spin fixture.

Misplaeehent

Figure 7. Representation of radial offset.

597

Radially Oriented Accelerometer Data


Radially oriented ADXL50 and ADXL181 accelercnneers were flight tested on board artillery projectiles. Data was available at muzzle exit through ground impact. Figure 9 shows the radial acceleration from an ADZ181 in flight. The DC shape is predominantly a measure of the centrifugal acceleration as seen by the centrifugal force calculation overlaid on the accelerometers output. After removal of the centrifugal force, the remaining oscillating part corresponded to the projectiles yawing behavior modulated near the spin frequency. The large radial acceleration amplitude near the beginning is due to tipoff. This amplitude and frequency information was analyzed with a Fast Fourier Transform m) compared to the and yawsonde data to determine exact contributors.

the three were observed but were accounted for in the analysis. Figure 12 shows a 1-s enlargement of the data, at 10 s into the flight, revealing oscillatory behavior. An FFT of this data revealed that there were three dominant frequencies present. These frequencies occurred at combinations of the spin, fast-mode, and slow-mode frequencies. The amplitudes and frequencies were c u m p a to the projectiles yawing behavior as measured by the yawsonde to obtain the exact combinations of each. The results were verified by the 6-DOF equations of motion as seen in Figure 13.

::: 1
5 -1.0
-1.5

Projectiles Axial Acceleratibn

Projectiles Radial Acceleration


-T
h

3 -2.0

a J

u l

-on W

-50:.

2 -2.5
-3.0 -3.5 -4.0 L 0
.
i

Accelerometer Output

10

Time (sec)

15

20

25

l 30

.
:

Figure 10. Axially oriented accelerometer output.

0 5 10 15 20 25 30 35 40 45 50 Time (sec) Figure 9. Radially oriented accelerometer output.


n

Axially Oriented AccelerometerData


A number of artillery projectiles and rockets weze instrumented with axially oriented accelerometers [9,10]. Figure 10 shows test data from an axially oriented ADXL05 on board an artillery projectile. The acceleration curve was compensated for spin-induced effects using the lab calibration information and the on-board magnetic spin sensor data. A 6-degnxs-of-fmdom O F ) trajectory simulation of the accelerometers output while on board the artillery projectile was also performed. The accelerometers cross-axis sensitivity, misalignment, and the location away from the projectiles CG and spin axis were measured from laboratory measurements or modeled from the manufacturers specifications. In Figure 11, the fit of the spincompensated acceleration output was compared to a fit of Weibel radar data that had been corrected for gravity a d slant range effects and to a fit of the simulated data. Very good agreement was observed. Small differences between

e -0.5 :
0

$ 3 -1.0

-ln5 -2.0 0

3
5
10 15 20 25 30 Time (sec)

Figure 11. Fits of axially oriented accelerometer output vs. radar vs. simulation.

598

gyro's scale factor and bias errors performed well within the automobile specifications when temperature remained constant and large rotation rates were absent. Undex rotation, the scale factor was attenuated and bias offsets were present. Figure 15 shows the flight simulator input profile for a coning rate of 2 "/s (I" angle of attack, l-Hz frequency) on the inner and outer axes and an overturning rate of 1 "/s superimposed on the inner axis. The raw gyroscope output is presented in Figure 13. The gyro did an excellent job of tracking the amplitude and fquency of the input rate. In the next test, the conditions were repeated and a 30-Hz roll rate was then added to the roll axis. The raw gyroscope output is presented in Figure 16. Figure 17 shows a 2-s enlargement. Frequency modulation occurred at the roll rate plus the coning rate as expected. Spin-induced bias and scale factor attenuation were present as seen by a voltage shift and a peak-to-peak voltage decrease. On a spinning projectile, these effects could only be compensated for with a prior spin calibration and knowledge of the spin rate in flight.

-1.0 . . 10.0

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

10.2

10.4 10.6 Time (sec)

10.8

11.0

Figure 12. Enlargement of accelerometer output.

Accelerometer Simulation

20

40

sb

80

160

12brIZ

Frequency (Hz) Figure 13. FFT of axially oriented accelerometer. Similar comparisons to radar data were made with the accelerometer data from rocket flight tests [ll]. More recently, a flight test was conducted to compare Motorola's MMAS40GlOD, Endevco's 7592-30, and AD'S ADXL150 on the same projectile [7]. Test results showed that all accelerometen survived a 17,OOO-g launch on board a 105m m projectile and transmitted meaningful data. Acceleration curves were ampafed to ground-based mdar data. R I G H T SIMULATOR TEST RESULTS

AMCB's 3-DOF projectile flight simulator, pictured in Figure 14, was used to test the Boeing/Draper prototype MEMS gyroscope [12]. Tests included scale factor, zerorate bias, and cross-axis sensitivity, performed at ambient conditions. Performance was also tested under simulated projectile flight conditions. In all, tests concluded that the

Figure 14. 3.-DOFprojectile flight simulator.

599

3.5 1
h

M crt

3i 95
a" d
.r(

3.4

cI .

$ 85
4

.3" a"
v1

90

z 3.3 2 B

'
a

3.2

3.1 3.01 . 4.0


. . . . . . . . . . . . . . . . . . . . . . . .

-;p
3 2.6

10 15 20 Time (sec) Figure 15. Flight simu1ator:motion profile.

4.8 5.2 5.6 6.0 Time (sec) Figure 18. Enlargement of gyro output. 4.4
INTEGRATED GPS/IMU

0 2.4

2.3

10 Time (sec)

15

20

Figure 16. Gyro output to coning and overtuming.

Rockwell Collins was contracted by AMCB to develop a toplevel HSTSS design of a miniature rugged integrated GPS/IMU for the application of short-range, slow rolling rockets and missiles. AMCB supplied Rockwell Collins with specific performance, space, and power requirements. Rockwell performed an extensive survey of the market to identify those MEMS that have high potential for integration into an IMU that satisfy HSTSS requirements. The Study results, described by Minor and Rowe [131, indicated that the current and near-term MEMS technology could be implemented to satisfy the HSTSS performance goals, for supplying an accurate real-time navigation solution. A similar study for highly rolling artillery projectiles will be conducted in the future.

3.6 -j

CONCLUSIONS
Ground tests have shown that low-cost COTS MEMS devices could survive even artillery-level shock environments and still operate without serious degradation. This was demonstrated with MEMS accelerometers during many instrumented flight tests. Data were mluced and acceptably compared to ground-based radar, yawsonde truth data, and simulation for T&E applications. A study has also shown that COTS MEMS could be implemented into an integrated GPS/IMU to supply TSPAI for rocket and missiles. Compensation, especially of the spin-induced effects on rolling accelerometers and gyroscopes, will be required for a quality inertial measurement. Therefore, proper calibration and inclusion of an on-board spin measurement will be essential. For these flight tests, the spin measurement was

11

I I I '

2.9

10 15 20 Time (sec) Figure 17. Gyro output to coning, overtuming, and spin. 0 5

600

supplied by a magnetic spin sensor and compensation was done during post processing of the data. AMCBs test results have been shared with the respective MEMS fabricators. The designers have benefited from the results by modifying manufacturing processes to improve sensor performance. Modifications to some of the COTS performance parameters may also be needed to tailor the devices to specific applications. This type of relationship is critical for the successful implementation of MEMS sensors into DOD applications. In FY98, AMCB will continue to test new MEMS COTS devices and those leveraged from the DARPA MEMS program.

[5] B.S. Davis, Miamlecuomechanical Accelerometer Shock Testing and Assessment for Use as Part of a Projectile Inertial Measurement Unic U.S. Army Research Laboratory, Aberdeen Proving Ground, MD, ARL-MR287, February 1996. B.S. Davis, L.W. Burke, E. Irwin, C. Myers, a d n C. Mitchell, High G Air Gun Testing of Hardened Subminiature Telemetry and Sensor System (HSTSS) Devices, U.S. A m y Research Laboratory, Aberdeen Proving Ground, MID, ARL-MR-306, June 1996.

[a

ACKNOWLEDGMENTS
The author would like to thank the AMCB electronics team (L. Burke, E. Ferguson, D. Hepner, E. Irwin, C. Mitchell, P. Muller and C. Myers) for their support with the electronics, hardware fabrication, telemetry, and data acquisition during the ground and flight tests. T. Harkins, M. Hollis, and G. Brown are also thanked for their assistance with the prior work that was included in this paper. Team leader F. Brandon and Branch Chief W. DAmico kindly reviewed this paper. The Office of the Secretary of Defense (OSD) is appreciated for sponsoring the Commercial Technology Insertion Program (CTIP) funded the evaluation and qualification that of some of these MEMS sensors. HSTSS, also sponsored by OSD, is recognized for providing funding. REFERENCES
[l] W.P. DAmico, L.W. Burke, R.J. Faulstich, and A. Hooper, The Hankned Subminiature Telemetry and Sensor System Technology Demonstration Phase, U.S. Army Research Laboratory, Aberdeen Proving Ground, MD, ARL-TR-1206, October 1996.

[A B.S. Davis, T.G. Brown, C.R. Myers, and M.S.L. Hollis, Ground and Flight Testing of Microelectromechanical Systems (MEMS) Sensors for the Commercial Technology Insertion Program (CTIP), U.S. Army Research Laboratory, Aberdeen Proving Ground, MD, MR-384, January 1998.
[8] W.H. Clay, A Precision Yawsonde Calibration Technique, U.S. Army Ballistic Research Laboratory, Aberdeen Proving Ground, MD, MR-2263, January 1973.

[9] B.S. Davis, T.E. Harkins, and L.W. Burke, Flight Test Results of Miniature, Low-Cost, Spin, Accelerometer, and Yaw Sensors, AIAA 97-0422, Reno, NV, January 610, 1997. [lo] T.G. Brown, D.J. Hepner, F.J. Brandon, M.S Hollis, E.M. Ferguson, W.P. DAmico, andL.W. Burke Flight Testing of a 2.75 Rocket: M257 Illuminatinl Warheal with MK66, mod.2 Rocket Motor, U.S. h) Research Laboratory, Aberdeen Proving Ground, MD, MR, 297, March 1996. [ll] T.E. Harkins, B.S. Davis, and D.J. Hepner, Usinj Accelerometers for Velocity and Position Estimation Results from 2.75-Inch Rocket Flight Tests, U.S. Arm! Research Laboratory, Aberdeen Proving Ground, MD ARL-MR-305, June 1996. [12] B.S. Davis, Flight Simulator Performance Tests a the BoeinglDraper Microelectromechanical System (MEMS) Gyroscope, U.S. Army Research Laborator) Aberdeen Proving Ground, MD, MR-388, March 1998. [13] R.R. Minor and D.W. Rowe, Utilization c GPSMEMS-IMU for Measurement of Dynamics fc Range Testing of Missiles and Rockets, presented at IEE Conference on PLANS, Palm Springs, CA, April 20-2: 1998.

[2] W.P. DAmico, Low-Cost Competent Munitions (LCCM) Self-Correction Devices - An Initial Study and Status, U.S. Army Research Laboratory, Aberdeen Proving Ground, MD, ARL-TR-1178, August 1996. [3] J.H. Connelly and G.N. Brand, Advances in Microelectromechanical Systems for Guidance, Navigation, and Control, AIAA Report 297, August 1997.
[4] B.S. Davis, M. Clymer, and G. Graves, Development of a Projectile Spin Counter and Orientation Sensor, SPIE Infrared Devices and Sensors Proceedings, vol. 2742, April 1996.

601

Das könnte Ihnen auch gefallen