Gyro/Accelerometer Board Data Sheet

Approximate size: 62mm x 50mm x 26mm
                  2.5" x  2"  x  1"

Approximate weight:  3 ounces = 90 gm

Input voltage:  5V DC
Input current:  about 50ma

Signal outputs: 0 - 5V DC;  "0" = 2.5V

The gyro/accelerometer board is a stand-alone analog board requiring
only 5V DC input voltage and about 50 ma of current.

It provides 7 channels of analog output from a 20 pin
connector as follows:

Channel 0 (pins 1,2)   Accelerometer x
Channel 1 (pins 3,4)   Accelerometer y
Channel 2 (pins 5,6)   Accelerometer z
Channel 3 (pins 7,8)   unused
Channel 4 (pins 9,10)  temperature
Channel 5 (pins 11,12) gyro x
Channel 6 (pins 13,14) gyro y
Channel 7 (pins 15,16) gyro z

Pins 17,19 ground
Pins 18,20 5V (DC) input to board

   The accelerometers are Analog Devices ADXL202AE 2-axis accelerometers.  
There are 2 of them, but one X axis is not used. These
accelerometers read +/- 2 gravities full scale.  Using our A/D board 
(which is only 8 bit) we set the center at about 127 - then 2 gravities is 
about 60 counts. This gives a range of 127+120 (+2g) to 127-120 (-2g).

   The gyros are Murata ENC-03J Gyrostar piezoelectric vibrating gyros.
The gyros have a maximum reading of 300 degrees per second full scale. 
Since they are solid state devices, they will require calibration.  Their
performance varies by about 10% - 20% over a single production lot according
to the manufacturer.  I am currently working on calibrating the board I have.
The gyro data will need to be sampled periodically and integrated over time
to give you an absolute heading or roll or pitch.  Otherwise of course you
just get the instantaneous yaw, pitch & roll rates.
        

Factors effecting measurement accuracy  

Gyro bias

All gyros have an output signal bias which is the observed signal when no input is present. In this case you can measure the gyro signal when the board is stationary. I would be careful to orient the board such that the gyro being studied is horizontal in the east-west direction. The rate of the rotation of the Earth is 0.0041666 degrees per second but cannot be measured by these gyros.

Procedure 1A: Collect output signals over a period of several minutes. Find the average signal for each second and record the average temperature reading for each second (obviously at the same time). This should allow you to find the bias and perhaps how the bias changes with temperature. This will require turning off the board and letting it cool down to room temperature between each test.

Gyro misalignment

The gyros are supposed to be orthogonal - but they are not exactly correct. This means that rotation about one axis will cause signals to be measured on other axes too.

Procedure 2A: Mount the board on a motorized turntable which can be set to rotate at constant (slow) rates - say 2 to 4 degrees per second. Measure the output signals from all three gyros. Clearly two should theoretically show zero. Find the average signal for each second and record the average temperature reading for each second (obviously at the same time). By comparing the measured output signals with the actual expected values (2 - 4 degrees per second) you can find the misalignment geometrically using sines and cosines of the unknown angles (after subtracting out the gyro biases).

Accelerometer bias

All accelerometers have an output signal bias which is the observed signal when no input is present. Of course in this case no input signal is not easily obtained since we are constantly subjected to one gee of acceleration due to gravity. Thus in the case of the accelerometers, the bias would be the difference between the observed output signal and one gee. Obviously the accelerometer must be oriented with the axis being studied vertical.

Procedure: Same as 1A above - except that you expect the output signal to be one gee. Repeat for each axis. Actually the expected value can be calculated from the following expression:

At a latitude of L, the acceleration due to gravity at sea level is approximately g= 9.780327 [ 1 + .0053024*sin2(L) - .0000058*sin2(2L) ] meters per second per second. G310 is at approximately 39 degrees 29.01 minutes N latitude, giving an angle L of approximately .689117 radians, and an acceleration due to gravity (at sea level) of approximately 9.81012397 meters per sec2

The above paragraph be extracted from the following website: http://www.rose-hulman.edu/~rickert/Classes/ma112/gravity.html You might look at this paper too: http://www.imar-navigation.de/beispiele/decision_assistant.pdf

Accelerometer misalignment

The accelerometers are supposed to be orthogonal - but they are not exactly correct. This means that acceleration along one axis will cause signals to be measured on other axes too.

Procedure 4A: Repeat the experiment of (3) but this time collect data on all axes and again average over time and track the temperature. Also of course subtract off the accelerometer biases. By comparing the measured output signals with the actual expected values (one gee) you can find the misalignment geometrically using sines and cosines of the unknown angles (after subtracting out the accelerometer biases).

Non-linearity

The devices do not produce a perfectly linear response over the entire range of sensitivity. This is usually about 1/2% of full scale. 

Noise

The faster you sample a signal the more important noise will become as a factor. If you look at the data sheet for the accelerometers, you will see that the noise when sampled at 100Hz is about 10 mg; at 200 Hz it is 14 mg and at 500 Hz it is 23 mg. I project the error to be at least 40 mg at 1000 Hz.

Suggestion: Try changing your sampling rate to 60 Hz (which seems to be the rate indicated by the data sheet) and see if you get better results.

Gyro level arms

This refers to the linear displacement of the sensors from the actual center of rotation. This is a difficult calculation but basically you need to know the orientation of the board (i.e. - roll, pitch, & yaw angles) with respect to the origin of the reference frame in which the rotation actually takes place. Then the measured rotations can be converted into actual rotations based on the lever arms and angles.

Accelerometer corrections

Obviously the orientation of the board will determine the actual direction the accelerometers are pointing. And the measurements must be corrected to line up with a horizontal Earth's surface.

 General Information: crwillis@androidworld.com