Speed and position from force plate

[Ross Miller]

Re: Speed and position from force plate

A student in our lab was performing this calculation recently (for push-ups, not gait, but the same principles apply) and we noticed the same thing concerning mass that Phil and Allan mentioned. The mass measured on the force plates vs. the mass needed for the calculated CoM velocity over a single repetition to be ~ periodic were often quite different, sometimes by up to 4 kg or so. There are several potential sources of this error but it surprised me that it was so big.

Ross

[Allan Carman]

Re: Speed and position from force plate

DR Phil Riches is correct “the calculations high highly sensitive to the assumed mass of the body and initial vertical velocity of the c.o.m.”

Expanding on my first post; Stepping onto, pushing off or landing on a portable force plate will cause an offset to be introduced into the force data, where the channels are no longer zeroed. Hence, during a few seconds of quiet standing immediately pre and post jump there will be differences between the known body weight and mean body weight over these periods. These offsets pre and post jump can be used to correct force data or in effect software re-zero the vertical and horizontal forces of both pre and post jump. A small offset in force data will produce an offset in the acceleration data (not zero over the periods of quiet standing). This offset will cause an accumulative error in velocity data with integration and a greater error in position data (either in a positive or negative direction). The drift in velocity and displacement data is evident in your results, where the further you progress in time the more varied and scattered the curves become. Pinning the velocity and displacement data to zero at -0.5 seconds prior to the jump does not fixed the errors in force and acceleration.

I think you need to go back and correct the force data.

Cheers
Allan

[David Pagnon]

Re: Speed and position from force plate

The problem is that it's what I actually did, I started the integration just at the beginning of the counter movement...

How do softwares manage to calculate the take off angle? Is there a trick, or is it just not perfectly accurate?

[Dr Phil Riches]

Re: Speed and position from force plate

Hi
In my experience, the calculations high highly sensitive to the assumed mass of the body and initial vertical velocity of the c.o.m. assumed at the start of the integration period. Small changes in these two parameters (within sensible bounds) will give a whole raft of different solutions. Due to the integration, errors are cumulative getting worse with time, so try to start your integration as close to the start of the movement as possible.
Good luck
Phil

[David Pagnon]

Re: Speed and position from force plate

Hello,

Thanks for your answer. I haven't seen any drift during wuiet start ce. However The force plate is portable yes, and despite it is screwed to a sort of rail it probably explains some of the drift after integration.

To give a little more details, there are actually two plates, one for taking-off and one for landing. The experiment is a kind of long jump from one plate to another, and we move the landing plate along the rail to get different lengths. That's probably why Vx and X are worse than Vz and Z. Then I concatenate the signals of the two plates to integrate. Here are pictures if you want to have a clearer idea.

https://photos.app.goo.gl/5hMP6JfL8EJpwJhx1

While the Z forces go back to zero after landing, the X are perfectible. You're right, I should re-zerl the landing plate. I'm not sure it would totally fix Vx though, what do you think? Maybe I should filter the forces too. Is there any custom to follow, or would a low-pass Butterworth filter 4th order 6Hz seem okay?

I actually don't care much about position; however there are two variables I'd like to consider:
- the take-of angle (as arctan(Vz/Vx) I would think). Would it make sense to compute it as it is? Or should I give up on it? Should I apply a correction, and if so what?
- the point where Vz=0, which gives me the limit between excentric and landing phases, both for take off and landing phases. Vz seems to be relatively relevant, but now I'm not sure of anything anymore.

Thanks again!
David

[Allan Carman]

Re: Speed and position from force plate

Hello David

My first question is do you have a portable force plate (such Kislter Quatro or AMTI Accupower) that is not bolted to the floor?

The zero vertical and horizontal force levels in portable plates drift following any type of jump as the plate is not secure and therefore they need to be re-zeroed prior to every jump (this is not good for multiple jumps). This shift in the zero force level is most evident when integrating acceleration to get displacement data. Where following the subject stepping onto the plate you may see a small linear drift in displacement during quiet standing prior to the jump. However following the impact of landing you can see very large linear drifts in vertical and horizontal displacement of quiet standing post jump (either negative or positive). What I do is have the subject stand quietly pre and post jump and auto correct vertical force data both pre and post jump to known body weight and re-zero horizontal forces to zero. This helps to get reasonable displacement data pre and post jump, although not always perfect.

On predicting jump height from force data you can use time in the air, impulse relative to body weight or displacement derived from acceleration. I have recommended using impulse as the most reliable, followed by time in the air and lastly displacement data.

The message is that for a portable force plate zeroing a plate prior to stepping onto the plate is not sufficient. As the act of stepping onto the force plate can alter vertical and horizontal force zero levels, however it is impact where large changes in force zero levels can take place. This may not be evident in the force data but will be if you integrate acceleration to get displacement data.

Cheers
Allan

[David Pagnon]

Hello everyone,

This is my first post here, so if I'm doing anything wrong please let me know!

I've got a signal on a force plate, the subjects are jumping then landing. I get the z and x components of the force along the time.

I'd like to integrate to get the speed (+ the take off angle), and then the position. The theory is simple : F-mg=ma. I'm doing a trapezoidal integration.

My problem here, is that speed and position should go back to zero at the end, and it's not. If I manage to share a picture you'll see that Vz is not too bad, but Z is terrible. That's even worse with Vx, and X is total non sense.

Do you have any idea how to prevent that drift? I would tend to think it's a common issue in movement sciences but I can't find anything about it. How do biomechanics software do?

Thanks in advance!