Accelerometer ADXL322 vs DE-ACCM2G2

[Imignant]

I have been comparing the ADXL322 from sparkfun, http://www.sparkfun.com/commerce/product_info.php?products_id=849
to the DE-ACCM2G2 from Dimension Engineering, http://www.dimensionengineering.com/DE-ACCM2G2.htm
It seems form the searches I had done that they have the same Accelerometer chip just on different boards.  Is this correct?
From what I can tell the DE-ACCM2G2 seems like it is a more complete package.  I want to use this to keep the gripper in my arm parallel with the ground.
Does anyone have experience with these two, or can tell me the difference.
I'm going to be using a Arduino Duemilinova or a roboduino.
Thanks

[dunk]

I want to use this to keep the gripper in my arm parallel with the ground.

you know an accelerometers readings will be deflected by acceleration as well as gravity right?
to tell the arms orientation it will have to be moving at constant velocity (or stationary).


dunk.

[Imignant]

I'm kinda following you, but not all the way
So you're telling me that as the rest of the arm moves it is going to effect the output of the accelerometer.  If this is so, what would be the best way to accomplish what I'm trying to do (without complex math )
Is there some kind of mercury switch or maybe a gyro?
Thanks

[random robots]

id bet a gyro would work. it detects a change in movement, and you program it to react depending on what it direction it detected a movement and react by moving in the opposite direction. a tilt sensor could also work by detecting if it is off center then change it to be on center, depending on which side was high.

Patrick

[dunk]

there is no sensor that does what you want without complex maths or some other sort of technical issue.

any sort of tilt switch is going to suffer from the same issues as an accelerometer.

a rate gyro can tell you the rate at which the sensor is rotating and is immune to acceleration but has no reference point so any inaccuracies in the readings will cause the gyro to "drift".

read up on IMUs (Inertial Measurement Units).
most IMUs combine the output from an accelerometer with a rate gyro and lots of complex maths.

other solutions for IMUs involve magnetometers (ie. 3D compass) but these suffer from being deflected by power in wires and proximity to ferrous objects.


if you want to use an accelerometer on it's own you will need to do something like this:
1. move the arm into position and wait for it to stop moving.
2. take accelerometer reading.
3. move gripper parallel with the ground.


if i was trying to do this i'd track the angle of the joints and work it out with basic trigonometry.
KISS.
for this to work you'd either need to be using servos as actuators of put encoders of some description on the joints.


dunk.

[Imignant]

All of my joints are gear driven so all joints are made with large 5k POTs to position the servos, is there a way I can use those for sensor input regarding the wrist?

[dunk]

so you are using servos?
am i reading that right?

with servos you already know the position you have told the servo to move to so you don't need a sensor.
or am i missing something?


dunk.

[Imignant]

Yeah, I'm using servos.  The only thing I think your missing is that the rest of the arm will be remote control for now and it will be mounted on a track chassis.  I have never used a microcontroller.  This project is kinda my stepping stone from RC to robot. Eventually I want to make it a line follower or something but for now I would rather not have to control the wrist through RC.

[dunk]

ah. got you.
now i understand.
so although you are using servos you are controlling them with conventional RC equipment.

you know you will need to use a microcontroller and do some programming with either of the accelerometers you linked to?

so it sounds like a simple problem right?
keep something pointing the right direction with no frame of reference.
it's the same problem faced by self leveling suspension systems in cars and autopilots on UAVs.
there has been a *lot* of research into the problem and it is still only solvable with a lot of work and processing power.

question:
does it matter if your gripper goes a little crazy while the arm is moving but returns to level when the arm stops?
if this is ok then you could just use the accelerometer.

question2:
have you seen the $50 robot tutorial yet?
http://www.societyofrobots.com/step_by_step_robot.shtml
it is designed as an introduction to microcontrollers and shows you what is involved with controlling servos.
with a bit of work you could implement control of all the servos in your arm from a microcontroller and control them with a joystick using only what you learn from the $50 robot tutorial.
(no radio control though.)


dunk.

[Imignant]

Yeah, I read the $50 robot as well as the line follower one, which I would like to use eventually.
So you say I need a point of reference.  How about the Chassis, if I had a second accelerometer on the chassis could I not tell the wrist servo to keep the same position as that? maybe? not?
Thanks