accelerometers, 2 or 3 axis

rwgast_logicdesign

ok so i ended up with 2 of the parallax memsic 2 axis accelerometers today.

I looked at the store and noticed that the memsic 2 axis is the same price as a 3 axis. does anyone know why this is?

What exactly is the 3rd axis for? the 2 axis has x,y and tilt, so can a 3 axis do x,y and then z for height movements?

i know an accelerometer can measure accelration, force, and vibration but can it be used to track distance and speed on the axis? like can i use these things to figure out my bots gone 20ft at 2 miles an hour? is it a suitable replacement for an optical mouse sensor?

i also have a 3 axis out of a wii chuck, so im just really tryin to find out in what situation i would use which accelerometer. and im also curious if theres any benefit to using 2 2axis over a 3axis becuase you get a 4 axis.

SRLM

An accelerometer measures acceleration, nothing more. Vibration is just oscillating acceleration over a period of time, and force is mass multiplied by acceleration. Speed is the integration of acceleration over time, and position is the integration of speed over time. It's here that high school mathematics becomes very useful.

Nowadays, nearly all low end accelerometers are 3 axis. There isn't much reason to keep them 2 axis anymore when they're all being put in cell phones.

Mathematically, there is no benefit to using two dual axis accelerometers over one triple axis accelerometer. In practice you could use the redundant axis to provide noise reduction.

rwgast_logicdesign

ok so i looked around and found out you can use an accelerometer to measure distance and speed, the problem is its to noise to be accurate. So how exactly would you go about cleaning us the signal with extra axis, like u suggest?

also i agree a 3 axis is cheap but if you have two 30 dollar 2 axis accelerometer why throw them a way. im still failing to see the need of a 3 axis in most applications that arent flying. what would you use a z for on a robot traveling on the ground

Circuitsoft

The only real way to clean up integration error is to use something like GPS to give you absolute position that'll have position noise that can be cleaned up by accelerometer-based predictions. The usual method is to use something called a Kalmann filter. You can find examples on these forums if you look hard enough.

Mark_T

Whenever you have integration in the real world (outside maths) there is drift - the errors and noise in the signal being integrated become a random drift in the output. Doing two integrations from a MEMS device taking acceleration signal to displacement means drift rapidly becomes enormous - so unless using short timescales you have to correct for drift using a separate source of position information (which only needs to be accurate in the longer term).

MEMS accelerometers are not very good - noise is high and linearity errors are in around-a-percent range. "Real" accelerometers used in inertial guidance systems are many orders of magnitude better in performance (and have to be).

Remember most of the time these MEMS devices are only being used to measure the _angle_ of the accelation vector due to the earth's gravity.

Duane Degn

I'm doubtful a second memsic device can be used for a third axis.

While some accelerometers don't care about being upside down or not, based on the memsic documentation, I think the memsic would provide more accurate readings when mounted right side up. I think mounting a memsic accelerometer at 90 degrees from level (in order to gain a third axis) will greatly reduce its accuracy.

After lookng over the memsic documentation, I'm guessing its advantage over a 3 axis version is it may be easier for a Basic Stamp to use the pulses the memsic module uses to communicate than having a BS try to read from a different module with some other protocol.

rwgast_logicdesign

Ok well like I said I have some wii accelerometers and what not from nun chucks which are 3 axis and 6gs i think. But I just got these randomly and looking at the quick start they seem easy to use,and there alot smaller than the wii nunchuck board, even sawed up. These would be fine in a wheeled robot to find its acceleration, figure out if its going up or down a hill, and get an overall idea of if the terrain is bumpy or smooth correct? Is there any uses im missing?