Accelerometer + Tilt compensation

Higher sample rate does not improve accuracy. You need to phase correct the GPS signal with decoded gold code. First, you need to have the key.

Have you sampled it without motion? Did you get any X and Y components in that case?

Or the sensor is tilted.

You have to do the inner product on each component.

All the components together is a vector. The components are scaler.

If the desired information is not navigational, what is it?

If you can afford two sensors, you can get that as well. Even though gyros do the rotation sensing thing best.

That can simplify things much. Simply place the vessel in a reference cradle that is aligned with a bubble level. Read the sensor output. It gets real easy after that. Dual sensors will get yaw for you as well. And a manual tip of the cradle will really help the calibration / offset calculation.

I assume inner product =3D dot product. From what I can tell by searching, it looks like they are..

That's ok, I think this is getting over-complicated. All I am looking for is the forces experienced onto the accelerometer, after being compensated for the average vector (gravity). This first and foremost includes roll/pitch, and maybe Yaw if I can get some other method to work. If there are external forces that will influence this OTHER than gravity, so be it... so for this application, we don't care about external force 'O'.

I think I mentioned this somehwere before... I just need the vectors out of the accelerometer to be compensated for gravity and mounting tilt. It's more to see what is going on relatively in the boat as opposed to tracking where it's going..

JosephKK wrote in news: snipped-for-privacy@4ax.com:

How about PLACING the whole assembly in a float? I think if you can center the CG in three dimensions, you're all set.

What you're asking to do is not possible in an unconstrained case (and a canoe/kayak in whitewater is pretty unconstrained) unless you include gyros and fully track all six degrees of freedom.

On the other hand, if you would accept knowing the total instantaneous acceleration ("g force") at any given time, then all you need to do [*] is a one-time calculation of the misalignments, bias, and scale factors and accept that the gravitational acceleration is just another component acting on the vehicle.

nter

hah, ya, if I could do something like that, then it would open up a lot more avenues to explore.. unfortunately, I am restricted to a simple PCB in a box.

No, we are just confusing each-other over time :) In calm conditions, zero-speed, I am trying to measure pitch and roll. I am trying to do this so I can know how to de-rotate all my future readings which will be in calm conditions, but non-zero and non-constant velocity. Knowing the (X,Y,Z) forces on the boat is of particular interest in my application (when the boat is moving), so I just want to make sure that my readings are compensated for errors in mounting the sensor as much as possible. Things like waves and what not are out of my control, and are not of concern to this application.

jdhar wrote in news:ef127ec5-bc55-45f7-a332- snipped-for-privacy@59g2000hsb.googlegroups.com:

and

Still sounds to me like high-pass filtering is what you need to do. Everything you seem to be saying seems to bring this home. This is how accelerometers are dealt with, unless there's a reason to do otherwise, but I still haven't seen an "I need low frequency content because...." statement.