Help Needed with Tracker Physics Program Measuring Acceleration

[bobfrancis1980]

TL;DR

When I look at the free-fall Vy (velocity y-axis) results it is a smooth upward line but Ay (acceleration y-axis) it is erratic and all over the place.

I was wondering if anyone here had experience with the physics program Tracker?

I used it to analyze a free-fall video I made in an effort to learn how the program works. The video is 240fps and 720p. The program reports a smooth upward line for velocity. After the velocity plot is done I double click on it to open the Data Tool. I click the Analyze button and it gives parameter A (the overall acceleration rate) as 10.72 so semi close to gravity.

But if I open the Ay plot for acceleration of the y-axis its a jagged up and down mess with acceleration rates frequently hitting 30m/s2 to -5m/s2

I have not figured out how to smooth out this acceleration. I tried Finite Difference and Bounce Detection

I was originally using accelerometers to measure the acceleration rate as the free-fall objects fell and their acceleration rates were much smoother, consistent, and in hindsight, accurate, but got chewed out by some physicists on reddit that the best method is a video camera and frame analyzation software which is why I am trying Tracker.

I also used hand dropping as my release method but have since converted to an Arduino controlled pair of solenoids that hold a piece of fishing line attached to the back of the free fall object. When I turn off the solenoids with my cellphone over bluetooth it initiates the drop.

Any help with getting the acceleration plot fixed would be appreciated.

Thank you,
Robert

 

[Gordianus]

I find too difficult to read data from the screen capture. Can you upload it again?

 

[Ibix]

What kind of accelerometers were you using?

My first observation is that the lighting in the video images is quite dim. The software is presumably tracking the position of the object by frame differencing, and if the light is low you probably have a lot of noise which may be messing up its position estimates. Can you get more light on the scene? Do the drop outdoors if the weather's not too bad where you are? Or use a bigger camera aperture?

 

[robphy]

The resolution of your screencapture is too poor to read.
On your "jagged" graph, note that there may be a power-of-ten multiplier, which may suggest that the graph is actually approximately constant.
But without being able to read the text on the screencapture, I can't say more.

Did Tracker autotrack this?
If so, did you zoom in to see how well it autotracked it?

 

[Ibix]

The resolution of your screencapture is too poor to read.

OP: note that this is probably the fault of the forum software and its aggressive compression settings. Try cropping the screenshot to contain just one of the graphs and posting each one separately.

 

[Gordianus]

I have no experience with Tracker but I visited their page and I think the program captures position versus time and speed and acceleration come from noisy differentiation. This Is just a wild Guess.

 

[bobfrancis1980]

Here are the plots:

Bounce Detection Algorithm Velocity of Y-Axis
Bounce Detection Algorithm Acceleration of Y-axis
Finite Difference Algorithm Velocity of Y-axis
Finite Difference Algorithm Acceleration of Y-axis

Even Bounce Detection Ay oscillates between 20m/s2 and 0m/s2 for much of the duration.

I did use Autotracker. I have so many tests to run I can't possibly tag them all by hand.

The accelerometers I used were the BMI270 built into the Arduino Nano 33 BLE Rev2 and the BNO055 built into a board by Adafruit.

I'll try adding light, its already one on the brightest spots in the house

 

[Gordianus]

From the first graph I see the speed changes from 1 m/s to 6 m/s in about 0.5 s. From this data I compute an average acceleration of 10m/s2. Not too bad at all. Finite differences from experimental data are frequently too noisy.

 

[marcusl]

Use a smoothing process instead of a differencing one. Fit your data to a 2nd order polynomial and read off the acceleration from the fit parameters.

 

[robphy]

I did use Autotracker. I have so many tests to run I can't possibly tag them all by hand.

I'll bet that the AutoTracker got confused at the times where the velocity graph had a kink.
It's fair to manually correct the tracking.... Zoom in and drag the tracked-point to where it should be.

I guess this is your question
https://www.compadre.org/osp/bulletinboard/TDetails.cfm?ViewType=2&TID=6027&CID=147991&#PID147992

By the way, you can tell Tracker what your frame rate is... no need to process it elsewhere.
It's possible that if you process the raw video in another application,
that application might create artifacts.

While probably not responsible for the kinks in the velocity graph,
here are suggestions for better results.

• Use a larger feature to set the length-scale... the little segment is quite small.
• Does your y-axis pass through the tracked points of your dropped-object?
  You can adjust the origin and orientation of those axes.
• Are you tracking the center of mass? It looks like it's tracking the top of an irregular object.
• You could try other finite differencing schemes:
  https://web.media.mit.edu/~crtaylor/calculator.html

 

[A.T.]

I was originally using accelerometers to measure the acceleration rate as the free-fall objects fell and their acceleration rates were much smoother, consistent, and in hindsight, accurate,

An accelerometer in actual free fall measures zero. In the atmosphere it measures the deviation from free fall due to drag.