Differential comparison between two accelerometers - Has anyone done this as a

Beau Schwabe

I was talking to a friend of mine in our regular Saturday robotics meeting get together and for a project that he is working on, it dawned on me that if you 'locked' two accelerometers together on the same plane, i.e., at opposite ends of a ruler with the SAME orientation (they wouldn't need to be as far apart as opposite ends, but you get the idea) ... you could normalize any effects due to acceleration, because the sensors would both experience the same forces, thus making acceleration a common mode variable. What you would see if you looked at the differences between the two is the change in angular rate... in a sense a Gyro.

Currently my accelerometer is tied up in something else, and I only have one at the moment, but if someone has done such an experiment I would be very interested in their results.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe

IC Layout Engineer
Parallax, Inc.

Timmoore

I remember reading a discussion of the same idea, I think it was on diydrones.com.

rjo_

Beau


I have the RadioShack //Parallax accelerometer... cant remember the manufacturer and the chip is buried beneath a mountain of stuff. It is a MEMS chip, which depends upon conductivity changes around a little bubble of gas... which changes significantly with temperature as well as with acceleration.

Other than the temperature sensitivity... what a nice little unit!!! I have been trying to get back to it and don't when that will happen...

ANYWAY... if u would place two of these units next to each other in opposite directions... a change in the delta should be the change due to temperature. If you would then feed that value into your calculations from the wings... I can't see how you could be wrong... and it would sell 4 of these units instead of one... which puts this right next to global warming in terms of importance[noparse]:)[/noparse]

Rich

Phil Pilgrim (PhiPi)

Beau,

I've wondered the same thing for a couple years now but never had the time to try it. I can see no reason at all why it wouldn't work. I think the physics is sound.

-Phil

John Abshier

I cannot find it, but there was a lengthy thread about using multiple accelerometes go get both acceleration and rotation.

John Abshier

rjo_

on rethinking it... there might be a little problem with my idea[noparse]:)[/noparse] Probably best to deal with the temp independently[noparse]:)[/noparse]

Timmoore

this thread http://diydrones.com/forum/topics/limitation-of-triaxis ?

Beau Schwabe

@rjo_ - Using that method to measure temperature might work, but it could probably also be done by inverting the signal in software rather than the unit.


@Phil Pilgrim (PhiPi) - Thanks Phil ... just making sure I didn't hit my head on something in a moment of clarity.


@John Abshier - "...get both acceleration and rotation" - that's what I was sort of going for. My understanding is that you typically take the readings from a Gyro and an Accelerometer both to determine an accurate heading.

Method #1: accelerometer + gyro

Given:
acceleration is the change in rate over time

Gyro = typically deg per second (Angular rate)
Accelerometer = inertial acceleration minus the local gravitational acceleration

So if you know the time frame that you are sampling the accelerometer at, you
should be able to isolate the gravatational vector by applying the Angular rate
data from the Gyro,  leaving you with just acceleration from the accelerometer.
And from the Gyro you can calculate acceleration by applying the same time frame
that you sampled the accelerometer with.  Leaving two acceleration variables to
cross correlate with one another.

Method #2: using 2 accelerometers

Note: this only works if both accelerometers are oriented in the exact same position
and fixed to a common rigid surface.

accelerometer1 = inertial acceleration minus the local gravitational acceleration
accelerometer2 = inertial acceleration minus the local gravitational acceleration

If you take the difference between accelerometer1 and accelerometer2 you create a
common mode situation between both accelerometers that cancel out any influences caused
by the local gravitational acceleration and inertial acceleration.  Only if you rotate
the accelerometers relative to one another about the 'common rigid surface' will you
see rotational acceleration.  Applying this value with time, and you will see the value
in terms of angular rate.  Which could be re-applied back into any or both of the
accelerometer values (similar to method #1 above) to get comparable values returned.

So, it seems without adequate testing that there are components of each method that can be derived into similar if not the same results. One using a gyro/accelerometer combo, and the other using two accelerometers. Method #2 might be a little easier to implement, but I'm not sure how accurate one is over the other.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe

IC Layout Engineer
Parallax, Inc.

dre

Beau, Would the following paper be of interest:

On the 'Memsic' web-site find the Application Note:

AN-00MX-012.pdf

"Inclination Sensing of Moving Vehicle"

Sorry that I could not link this to you directly but in searching for the link I could not find this application. Maybe you will have better luck. If not I can send you a copy.

cheers, David

SRLM

@dre
www.memsic.com/data/pdfs/an-00mx-012.pdf

The system in the app note uses a single dual axis accelerometer to measure rotation in a single axis, but it appears that if you add four more acceleration axes then you can a total of three rotation axes. In that case then, it shouldn't matter what orientation that the two triple axis accelerometers are mounted relative to one another, since it appears that you could calibrate it out. For convenience though it might be better to align them.

Post Edited (SRLM) : 10/2/2009 11:03:41 PM GMT

Carl Hayes

Beau, if there is no rotation but only translation, the accelerometers ought to give the same output.

If there is rotation, it's more complex.· Consider, for example, the effect of changing the axis of rotation.

If the axis of rotation is not parallel to·a line passing through both accelerometers, then the centripetal accelerations experienced by the two accelerometers will be unequal both in direction and in magnitude.· That may make your world a much messier place.

Sounds like fun.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
· -- Carl, nn5i@arrl.net

John Abshier

Finally found the thread.

http://forums.parallax.com/forums/default.aspx?f=25&p=1&m=326074

John Abshier

Post Edited (John Abshier) : 10/4/2009 10:12:30 PM GMT

Philldapill

Ha! I've thought about the very same thing. I'd like to build a quadrocopter one of these days, but all the gyro's I've found thus far have been too expensive for my taste. However, I think there will be problems unless you use some major filtering... Chances are, there will be a good bit of uneven vibration in the setup. Each accelerometer will experience different amounts of vibration, due to the nature of their distributed locations. Hypothetically, it SHOULD work... Would this be a Kailman filter application?

BTW, ST Micro just came out with some MEMs gyro's back in June I believe... I think the price for a dual axis is something like ~$6. VERY cheap. In fact, I think that's cheaper than two of the accelerometers they make.

Still, an interesting exercise.

Beau Schwabe

Thanks for the link John... not all that long ago either. Still it might be worth putting something like that together ... I would place the sensors much closer to one another than how it was described in the PDF.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe

IC Layout Engineer
Parallax, Inc.