For Programmers: Free Programming Magazines  


Home > Archive > Matlab > July 2006 > Re: Accelerometer









You are viewing an archived Text-only version of the thread. To view this thread in it's original format and/or if you want to reply to this thread please [click here]

 

Author Re: Accelerometer
G. Grewal

2006-07-28, 7:04 pm

Paul Selvadurai wrote:
>
>
> Thanks everyone for your input.
>
> Paul Selvadurai


Hi,

I am working on a very similar problem and was wondering if someone
has any experience with this topic:

Assuming that the accelerometer is not ideal and suffers from:
- thermal noise
- electrical offset
- drift
- quantization error

Then the displacement calculation will have an error that rapidly
increases with time. In my case, I am trying to calculate
displacement continously for 5 minutes. At the end of the 5 minutes,
I will have a significant error in displacement.

The error can be reduced by doing the following to the accelerometer
data:

1. Remove mean
2. Remove linear trend
3. Apply Lowpass Filter
4. Reset the Integration after a fixed interval of time.

The problem with this method is the resetting of the integration,
which causes discontinuities in the displacement.

Does anyone know of a robust technique/MATLAB functions for obtaining
displacement from noisy accelerometers?

Regards,
G. Grewal
Scott Seidman

2006-07-28, 7:04 pm

"G. Grewal" <gogi51@hotmail.com> wrote in news:ef36466.4
@webcrossing.raydaftYaTP:

> Does anyone know of a robust technique/MATLAB functions for obtaining
> displacement from noisy accelerometers?
>
> Regards,
> G. Grewal
>


"Noise" will not impact a twice-integrated output at all. The offset and
drift, though, will blow up the two integrations.

The robust way to deal with an accelerometer is to HIGH PASS filter its
output prior to integration, not low pass.

--
Scott
Reverse name to reply
G. Grewal

2006-07-31, 7:17 pm

Scott Seidman wrote:
>
>
> "G. Grewal" <gogi51@hotmail.com> wrote in news:ef36466.4
> @webcrossing.raydaftYaTP:
>
> obtaining
>
> "Noise" will not impact a twice-integrated output at all. The
> offset and
> drift, though, will blow up the two integrations.
>
> The robust way to deal with an accelerometer is to HIGH PASS filter
> its
> output prior to integration, not low pass.
>
> --
> Scott


Scott,

Thanks for your reply.

You are right, applying a High pass filter on the data will reduce
error significantly.

Gagan

> Reverse name to reply
>

Scott Seidman

2006-07-31, 7:17 pm

"G. Grewal" <gogi51@hotmail.com> wrote in news:ef36466.9
@webcrossing.raydaftYaTP:

> Scott Seidman wrote:
>
> Scott,
>
> Thanks for your reply.
>
> You are right, applying a High pass filter on the data will reduce
> error significantly.
>
> Gagan
>
>


Depends on why the DC-component arises in the first place. The more
perpendicular your accelerometer is to gravity, or at the very least is
at a constant position with respect to gravity, the happier you should be
with your results. If your accelerometer BECOMES misaligned with gravity
as a result of the earthquake you're measuring, then assuming the signal
has no DC gravitational component will cause errors, but they are
generally small (cosine of the angle of the change)

--
Scott
Reverse name to reply
Sponsored Links







Also available: Server administration forum archive | Web Design forum archive | Software forum archive | Hardware reviews archive

Copyright 2008 codecomments.com