A Solid State Tilt Meter for Current Meter Attitude Determination

Albert J. Williams 3'', Applied Ocean Physics and Engineering

Woods Hole Oceanographic Institution
Woods Hole, MA 02543 USA
awilliamsbwhoi.edu

-
Abstract Earth coordinate determination of current components of the generally dipping earth's magnetic
requires measurement of beading and tilt of the sensor to field. Solid state arrays of magnetic sensors, such as a
rotate the instrument frame measurement of velocity into three-axis magnetometer composed of orthogonal
earth frame coordinates. Gimbaled magnetometer coils magneto-inductive sensors, entirely remove the need for
allow the magnetic heading to he determined hut solid state
three-axis magnetometers require tilt i n addition to the the bearing or pivot. But this requires an independent
earth's magnetic vector to resolve heading. A measure of measurement of vertical to resolve the magnetic vector
tilt is also desirable if the current meter cannot he assured into a horizontal direction. Such a vertical sensor may
of a vertical orientation. Two-axis MEMS accelerometers be a sensor of direction only but can also be a two-
by Analog Devices, Inc. are capable of providing this tilt component accelerometer and avoid the bearing or pivot
measurement but require temperature correction to of a mechanical sensor. A two-axis accelerometer was
remove a significant error at zero tilt. The determination selected to provide vertical orientation for a three-axis
and application of this temperature coefficient to correct magnetometer compass in the Modular Acoustic
the error has enabled a versatile attitude sensor to be Velocity Sensor (MAVS) current meter [I ,2].
incorporated in the Modular Acoustic Velocity Sensor
(MAVS) for conversion of flow measurements by the sensor
into earth-coordinate currents. Residual tilt error is less INSTRUMENTATION
than 0.3 degrees from 5'C to 28'C. Absolute accuracy of
the sensor is better than 1' between -50' and +30' and is A. ADXL202 Two-Axis Accelerometer
better than 3- between -7W and +70'. Because the The Analog Devices, Inc. ADXL202, is a MEMS
accelerometer works equally well inverted, the current integrated circuit that measures two components of
meter can he mounted upside down to place the velocity acceleration in the plane of the device with a range of
sensor above instead of helow the housing without opening
the case. A second socket for the accelerometer at right plus and minus 2g or +/- 20 m/s2[3]. The output is a
angles to the first permits a horizontal mounting of the duty-cycle modulated square wave. The ratio of the 'on'
current meter with only minor internal rearrangement of a cycle to the full cycle can be used as a measure of the
part. component of acceleration along the axis being
measured. Altematively, the average voltage produced
INTRODUCTION by the fraction of the time the cycle is on can be
Attitude sensing in the context of current measured. An integrator or a simple resistance/capacitor
measurement means the determination of vertical and the filter can he used to average a number of cycles of the
determination of heading, two vector directions square wave before digitizing the voltage from the
orthogonal to one another. There is no attempt to device. This simple circuit provides a sensitive and
measure linear accelerations or rotational motion but accurate measure of the sine of the tilt angle of the
only steady, fixed orientation (motion with accelerations device. The two channels correspond to two orthogonal
less than 0.01 g can be neglected as can oscillatory components of tilt: pitch and roll.
motions with zero mean displacement). The direction of Table I shows the results of tests made to select the
vertical is referenced to gravity. The heading is appropriate number of digitizations required to reduce
referenced to the projection of the earth's magnetic field the variability below an acceptable threshold. With too
on the horizontal plane, itself defined by local gravity. few measurements, the electronic noise and the hit noise
Attitude of the sensor is very important for current of the digitizer (12 bit AID in the Tattletale 8 single
measurement to calculate the transport of water board computer that is the controller for MAVS) give an
accurately. In fact, the direction of motion is at least as unacceptable uncertainty in tilt angle. Too many
important as the speed of motion in open ocean transport measurements increase the time required obtaining an
measurements. One of the more limiting characteristics angle and the power consumed in its determination is
of mechanical current meters has been the vane and vane excessive for the purposes of the current measurement.
follower and the magnetic card compass required to Based on a statistical test of standard deviation and of the
determine the flow direction. In acoustic current meters, range of measurements with a single digitization, IO
there are generally no vanes but a compass is still present digitizations, and 100 digitizations, an optimum of five
whose output must be related to the instrument frame for digitizations was selected. Five measurements takes
determination of flow direction and the compass may ahout 25 ms. This ensured that variations from noise and
have mechanical parts subject to friction or to stickiness digitization granularity were less than 0.5Oof tilt.
that compromise the determination of heading. Flux gate Using the 5-digitization measurement of angle, the
sensors partly avoid the bearing problem of card or linearity of the tilt measurement was checked and is
needle compasses but they too must have at least a shown in Fig. 1. Tilt was set with a machinist protractor
pendulous component to resolve the horizontal and bubble level with an expected error of about 0.3".

0-7803-8669-8/04/$20.00 02004 IEEE. - 1395 -

Since the output of the accelerometer is proportional to 8.ADXL202 Temperature Coefficient
the sine of the tilt angle, this is converted to an angle by The stability of the ADXL202 over voltage
forming the arcsine of the reading, suitably scaled. This fluctuations is acceptable if the A D converter reference
presents two issues: first, the scale factor must be voltage is used to power the ADXL202. This is true
determined in calibrations, and second the absolute value since the duty cycle modulation is insensitive to the
of the argument of the arcsine must not exceed 1.000. In voltage driving the circuit. And the amplitude of the
the example shown, as well as in the standard calibration average of the modulated square wave is proportional to
procedure, the tilt is set first to 0" and the offset noted. the supply voltage. But there is a characteristic of the
Then the tilt is set to about 45' and the reading is ADXL202 that is not acceptable without correction, the
compared to that at 0". This difference is divided by the temperature coefficient of the zero offset. In the
sine of the tilt angle and is the scale factor for this technical specification sheet this offset is graphically
sensor. For all subsequent tilt measurements, the reading presented as an uncertainty in the zero g offset as a
of the sensor is corrected for zero tilt offset (the offset function of temperature away from 25°C that is as great
value is subtracted from the reading) and that result is as 0.lg at the freezing point of water, a temperature that
divided by the scale factor. This is the argument for the is likely to be encountered in an oceanographic
arcsine that gives the tilt angle. instrument. This would result in an error of more than 5'
tilt. Such an error would be unacceptable. A simple
TABLE I calibration run with a heat gun on an A D Z 2 0 2 clamped
SAMPLE AVERAGING FOR TILT
in a horizontal orientation shows the uncompensated
temperature offset. An example can be seen in Fig. 2.
Pitch for Single Measurement STD 0.289' Range 1.4'
Note that the temperature offset of the roll axis is
unrelated to that of the pitch axis. To make this
Roll for Single Measurement STD 0.447O Range 1.E0 measurement, a copper bar with a thermistor mounted to
it was clamped to the tilt sensor and the bar was heated
Pitch for 10 Measurements STD 0.1 11' Range 0.5' with a heat gun over a period of a minute and then
allowed to cool down. The heat was applied to the end
Roll for 10 Measurements STD 0.134O Range 0.5' of the bar closer to the tilt sensor than the thermistor and
the temperature gradient along the bar produced the
Pitch for 100 Measurements STD 0.0509 Range 0.3' hysteresis in tilt that is visible at the high temperature
end of the plot.
Roll for 100 Measurements STD 0.097O Range 0.2' Temperature Uncompensated Tilt Sensor 1043

10 0
Tlltvs Pitch Readlng I I

--
60
60
B 60
50
40
Pe 40 I
% 30
3 20
c 00
E 20
-7 n
10 0 20 40 60 00
C.IP.I"E
0
0.0 10.0 20.0 30.0 40.0 50.0 60.0 Fig. 2. Tilt appears to increase as much 8' with an increase
Tilt degrees in temperature from 25% lo 60DCfrom an uncompensated
tiit sensor. Heal was applied with a heat gun while the
Fig. 1. Applied tilt angle and measured pitch derived sensor was clamped in a horizontal orientation. Hysteresis
from 5 digitizations of the ADXL202 duty cycle output. in the tilt measurement, when the heating stopped, results
The scale factor was chosen for a best fit. from the separation of the thermistor from the tilt sensor
causing a temperature tracking error during rapid heating.
The second concern is that the calibration procedure
may give a reasonable fit to the tilt over the range +/-70" C. Temoerature Comoensation of the ADXL202
but when the tilt approaches 90" the argument Temperature changes are relatively easy to
approaches 1.000 and may exceed it because of noise or implement in the positive direction from room
slight error in the scale factor or even from instantaneous temperature to 40°C or higher but are not particularly
accelerations due to instrument motion. This will cause relevant to oceanic conditions. Likewise, temperature
an error in the arcsine routine. Either this condition must excursions below freezing are easy to implement witb
be computationally trapped before the arcsine routine is circuit cooler (a Freon snrav that boils about -50°C) but
I ,
imp1emented Or the result Of such an error be when the temoerature rises to between freezing and ioom
~~~ ~~~
~1 I

acceptable. temperature, condensation of water vapor onto the circuit

makes measurements ineffective. Fig. 3 shows an
example of the lower and the higher temperature range

- 1396 -
where the output is the sum of five digitizations in
millivolts.
nlt Tempco l o o i

-50 -30 -10 10 30 50

CeIsIUs

Fig. 3. Temperature coeffldent measurements made Fig. 4. The thermoelectric cooler with temperature
with heat gun and circuit cooler. The region of oceanic calibrating board contains six ADXL202 tilt sensors
interest is inaccessible due to condensation of moisture under a copper bar held in dose contact with clamps. A
on the circuit. thermistor is inserted into a hole in the bar and coupled
to it with heat sink arease. The machinist vise is set to
hold the tilt sensorsiorizontal during the calibration. For
To obtain measurements in this missing range, a descending temperature. the cooling fins sequester
thermoelectric cooler was employed, which in the cool- moisture so the circuits remain dry.
down mode collected moisture on the cooling fms inside
the chamber, keeping it off the circuit, and permitted Temperature Compensated Tin Sensor 1043
coefficients to he determined between 24°C and 6”C,
more representative of the oceanic range. Fig. 4 shows a
machinists vise in a thermoelectric cooler holding a
circuit board with six ADXL202 sensors and a thermistor
clamped under a copper bar. The cooler is closed at an
elevated temperature to ensure that all the surfaces are
dry. Then measurements of millivolt oulput readings are
made for each of the six sensors, for both pitch and roll
axes, and the thermoelectric cooler is run ovemighi to
lower the temperature below 6°C after which the
readings are made again. From these two readings the 0 10 20 30 40 50 60
temperature coefficients (tempo readings) are C.I*i”*
calculated. Fig. 5. Tilt was zeroed at 48% and the temperature was
For each measurement taken in the MAVS current reduced to 5 T in the thermoelectric cooler. Between
meter, this temperature coefficient is used to reference 24°C and 5% the pitch and roll readings were within 0.3O
the millivolt reading kom the ADXL202 sensor to what tilt angle of the coemcientdetermining settings at 16%
it would be at 0°C. The zero tilt offset referenced to 0°C and 5.5%.
is subtracted Som the measurement, also referenced to
O T , to obtain a reading. Then the scale factor is applied Temperature CompensatedTilt Sensor 1043
and the arcsine run to yield the tilt. Fig. 5 shows the
temperature compensated tilt measurement in a .-
descending temperature run that was made inside the
-
45 8
thermoelectric cooler. The temperature compensation 45 6
I454
coefficients were determined from the measurements
made at 16°C and 5.5T and at these temperatures the
offset in tilt in Fig. 5 is the same. Because the zero was
set at 48% there is an offset at the coeffcient- I

44 4
determining temperatures of -1.7”C that would normally 44 2
be absent if the zero-tilt offset had been obtained at room 44 0
temperature. 5 15 25 35 45
A final question remained about ihe temperature Cdsius
coefficient effect on scale. To test this, the machinist Fig, 6 , There is negligible ewed of temperature
level was set to 45’ tilt and a temperature run was made coefficient on the scale factor. This tilted temperature
in the thermoelectric cooler. This is illustrated in Fig. 6. run shows less than 0 . 3 O of tilt uncompensated between
3 0 T and 8°C.

No single temperature coefficient will suffice for all

ADXL202 devices. As Fig. 7 shows, the distribution of

- 1397 -
temperature coefficients for twenty units that have been orientation of the current meter. The package is small,
calibrated extend from 10 mVPC to -2 mVPC. The pitch low power, fast, accurate, and easy to interface to the
coefficients are consistently more positive than the roll MAVS controller. The process of obtaining the
coefficients but each axis of each device must go through temperature coefficients of each sensor is not overly
a temperature cycle for determination of the temperature complex and the scale factor is similarly easily
coefficients. Only a single coefficient is determined determined. However, these four coeficients must be
however, a linear fit to two points on the curve, at about determined for each lilt sensor.
room temperature and as cold as the thermoelectric
cooler can take it, typically 24°C and 5.5"C.
Temperalure Coafficienta forTilt Sensors

9
8
7

k6
s 5
$4
y 3
2
1
0
11109 8 7 6 5 4 3 2 10-1-2-3-4
mvldegree C
Fig. 1. Temperature coefficients for twenty
ADXL202 devices show a spread of values. No Fig. 8 . Altitude module with the three-axis magnetometer
single coefficient can be used and pitch and roll compass and the two-axis tilt sensor. There is a right
for each device must be obtained. angle DIP socket to mount the ADXL202 circuit into if the
MAVS current meter is to be deployed horizontally.
TI111005 Invened. Corrected
D. Cornpasflilt Attitude Sensor in MAVS
The combination of the three-axis magnetometer 2
compass with the two-axis accelerometer in an attitude 1.5
sensor for M A W is illustrated in Fig. 8. The calibrated
ADXL202 on a DIP adapter is plugged into a 14 pin DIP
socket on the board. While there may be as much as a 2"
-
I
5 1

0.5
variation in alignment each time this is plugged into the
socket, the calibration of level made before a deployment % o
determines the correct offset. There is a right angle DIP -0.5
socket also on the board to allow the tilt sensor to be -1
moved for applications where the current meter is to be 5 15 25 35 45
mounted horizontally. This keeps the normal reading of CSlSiUI
the tilt sensor near 0" where it is sensitive and well away
Fig. 9. Temperature compensation is effectivewith t h e
from the arcsine error condition of 90". The same sensor inverted as can be seen with the temperature run
temperature coefficients and scale factors that were of tilt sensor 1005.
determined during calibration are appropriate for the new
orientation and no change need be made except to ACKNOWLEDGEMENT
indicate in the MAVS setup menu that the orientation of This work was made possible by the Woods Hole
the current meter is now horizontal. The saving of a Oceanographic Institution and Nobska Development,
second calibration was sufficient motivation for moving Inc. WHO1 Contribution number 11222
the sensor rather than installing two sensors and allowing
software to select the Orientation. It is not necessary to REFERENCES
even move the tilt sensor to the other socket if the [l] Thwaites, F.T., and A.J. Williams 3rd, 1996.
current meter is inverted. The orientation change need "Development of a Modular Acoustic Velocity
only be entered in the setup menu of MAVS. Fig. 9
Sensor", Oceans 96, IEEE Catalog Number
shows the temperature compensation behavior of the 96CH35967, pp. 607-612.
inverted sensor. [2] Williams, A.J.3", and F.T. Thwaites, 1998. Earth
coordinate 3-D currents from a Modular Acoustic
SUMMARY
Velocity Sensor, Oceans 98, IEEE Catalog Number
A solid state tilt meter, the ADXL202 two-axis 9 8 ~ ~ 3 6 2 5pp.
9 , 244-247.
accelerometer, can resolve the earth's magnetic field [3] Analog Devices, Inc., 1999. "ADXL202iADXL210
vector measured by a three-axis magnetometer onto the Technical Specification", One Technology Way,
horizontal plane to obtain a magnetic beading. The two- P.O. Box 9106, Norwood, MA 02062-9106, U.S.A.
axis accelerometer can also remove residual tilt in the

1398 -