'Reference' accelerometer for force plates

[Tim Bhatnagar]

Hi,

We are going to be moving our gait lab into a new building and are currently working on completing the design documents. In our current lab, we were fortunate enough to have our force-plate system mounted on an independent concrete pillar, such that any vibrations from our current building would have a minimal effect on the forceplate measurements.

Unfortunately, in the new building we do not have an independent support pillar, and we are not on the ground floor. We have had a vibration analysis done on the floor, and although there were no 'large significant' vibrations measured, I would like to use a reference accelerometer - mounted to the same mounting plates as the force plates - in order to measure and filter out any 'noise' from the building (or the parkade below us :S ).

1) Have any other gait labs done this?
2) Any recommendations for accelerometers? (I'm asking here before I go to AMTI!)

Thanks,
Tim Bhatnagar

[Ton van den Bogert]

Re: 'Reference' accelerometer for force plates

Tim,

We have done something similar for an instrumented treadmill using multiple accelerometers. This was more for active movements than for vibrations, but the principle is the same. If you assume there is only vertical vibration, you can do it with one accelerometer.

The force error caused by vibration is proportional to the accelerometer signal. You need a calibration experiment to determine this relationship. No external force applied to the treadmill, and record vertical force and vertical acceleration while there is vibration. Have someone jump on the floor next to the force plate. Then do linear regression to predict force from acceleration: F = A + B*a. During a human subject test, you record acceleration a, compute A + B*a, and subtract this from the recorded force to compensate for the vibration. You don't need to calibrate the accelerometer and you don't need to subtract gravity, this calibration will do that for you.

The error is caused by the inertial effect of the moving force plate mass that is located between the foot and the load cells. From this you can estimate what the resolution of the accelerometer needs to be. To get 1 N accuracy, your accelerometer needs to have an accuracy of 1 N/ mass (kg) and that would be in m/s^2.

For the instrumented treadmill that was challenging because the mass was about 150 kg, so 1 N precision requires accelerometer resolution of 0.0067 m/s^2 . We used triaxial accelerometers (4030 2G range, Measurement Specialties, $170 at digikey.com). For vertical vibration, you could use a single-axis accelerometer and the sensitivity does not have to be as good for a force plate because the mass will be smaller. I would definitely use an accelerometer with analog output and record it as an extra signal with the force plate analog signals. That way, you are sure that all signals are measured at the same time.

We have recently submitted a manuscript on this, to do this for all six axes of the force plate (Fxyz, Mxyz). For one axis, it is quite simple.

I wonder why force plates don't have this already built in. It does not add much to the cost, and makes the measurements much less sensitive to floor vibration.

Ton van den Bogert

[Tim Bhatnagar]

Re: 'Reference' accelerometer for force plates

Hi Ton,

Thanks very much for the input. The method for using the accelerometer to compute the true force signal (sans systemic vibrations) is particularly appreciated.

The analog output characteristics you mentioned were something we had in mind, particularly so that the signals from the accel. and the FPs would be synced. FYI (you likely already know) - We *did* also note that since our A->D board has a finite time delay between signal channels being processed, there will be a small time offset between the accel. signal and the FP signals (we plan on using 5 force plates, with 6 channels each), so we are planning to, at the very least, characterize that time difference.

I also agree that it would be advantageous for FPs to come with on-board reference accelerometers - perhaps its something that we'll see in an upcoming generation of plates.

I'll be looking for the manuscript you mentioned - great to see assessment methodologies being published!

Thanks Ton,
Tim

[Ton van den Bogert]

Re: 'Reference' accelerometer for force plates

In our setup, the accelerometers and the force plates have analog outputs, and they all go into the same AD board. So there should not be a synchronization problem. But now I am starting to question that assumption...

If there is a time delay between channels on the same AD board, it should be small enough that it does not matter, when considering the bandwidth of the signals.

I am curious what AD board you use and how large the time delay between channels is.

Ton