# Getting Better Motion Sensor Data

The Science WorkshopTM motion sensor is a very handy way to determine your position, velocity, and acceleration during a motion. (How does it do that, anyway?) However, I'm sure that you've noticed that the kinematics graphs produced by the motion sensor are sometimes (ok, often - well, generally) pretty rough - especially the acceleration vs. time graphs. This is because slight irregularities in the round-trip times show up as irregularities in the position data, which are magnified in the velocity calculation, then magnified again in the acceleration calculation.

What can you do? First, you should make sure that the motion detector is not getting reflections from some unwanted object. Try clearing the area, re-aiming the detector, and adjusting the beam (near/far). This probably won't remove all of the "jitters" in the data, but it can help a lot. Beyond that, you can either work around it, or try to fix it.

### Curve Fitting:

 To create acceleration vs. time graph shown above: Zoom in () on the region you want to examine. Highlight the region in which you want the curve fit using the mouse. Select a curve fit for the data.
To work around the problem and ignore the "noise" in the data, you could try a curve fit. For instance, suppose your acceleration vs. time graph looks linear, but really noisy. Try this:

1. Go to the graph options dialog (), and turn off "Connected Points". ("Connected Points" are usually a bad idea, anyway!)
2. Try a curve fit - remember that you can restrict the curve fit to a particular region of the data.

### Smoothing the Data:

 The Experiment Calculator window shown above displays a smoothing function applied to motion sensor position data. The "smooth" function is found in the "Special" submenu of the "f(x)" menu, and the "@1.x" is found in the "INPUT" menu.
"Smoothing" is a statistical process that tends to do just that - smooth out the "jumps and jiggles" in a data set. It does so by averaging the current value with a certain number of adjacent points.

Well, doesn't that change the data? Yes, it does. Isn't that somehow underhanded and unsavory, if not downright illegal? Well, no, it isn't. If the "jumps and jiggles" are truly random noise (just about the same number of high and low jumps, etc.), then there is nothing wrong with trying to remove it. Just be sure to:

• keep the original data, and
• properly label the smoothed data.

Science WorkshopTM makes it easy to smooth a data set. In the "Special" submenu of the "f(x)" menu in the Experiment Calculator window you will find the "smooth(n, x)" function. Here, "x" is the data set to smooth, and "n" is the number of points to use in the smoothing operation. How many points should you use? You need to play around with it. Too few, and the data isn't really affected by the smoothing operation, and too many can remove a great deal of the information content. It's a judgment call.

 This graph uses the same data as the graph above, but the original round-trip time data (only) was moderately smoothed (n = 5).