I'm trying to collect data from an older piece of equipment, a force plate
with piezoelectric load cells and charge amps. I'm collecting and logging
the data through a multi-channel A/D board and computer. Due to the
advanced age (>20 years) of the charge amps, I'm getting a lot of noise in
what should be a quiescent signal. The without digging into the charge
amps (something I *really* want to avoid!), there are two ways to deal with
the noise - a discrete component low-pass filter, or some sort of numerical
manipulation of the data. The latter is prefered beacuse it's a lot easier
to change the coeffecient of a variable in an algorithm than purchase and
install a discrete component in a circuit.
In the past, one method we have used to smooth data is a rolling average.
Normally this was done on relatively low frequency data (the data of
interest occured at a low frequency respective to the sampling rate), and
the frequency spectrum had never been of interest. In the current
experiment, I am sampling at 100Hz with an expected maximum frequency of
motion of 10Hz, a 10:1 ratio. Additionally, this time we *are* interested
in the frequency spectrum of the signal (the rate of postural sway in the
subjects - wich will be below 1Hz). I'm using a Fast Fourier Transform to
extract the frequency spectrum data.
My questions are this: how will using a rolling average to smooth the data
affect the frequency response of the system? How will changing the window
size of the rolling average change the frequency response?
Guidance or references on these questions would be greatly appreciated.
Dan Major
Univ. of Okla. School of Industrial Engineering
major@ou.edu htp://www.ecn.ou.edu/~major
---------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
---------------------------------------------------------------
with piezoelectric load cells and charge amps. I'm collecting and logging
the data through a multi-channel A/D board and computer. Due to the
advanced age (>20 years) of the charge amps, I'm getting a lot of noise in
what should be a quiescent signal. The without digging into the charge
amps (something I *really* want to avoid!), there are two ways to deal with
the noise - a discrete component low-pass filter, or some sort of numerical
manipulation of the data. The latter is prefered beacuse it's a lot easier
to change the coeffecient of a variable in an algorithm than purchase and
install a discrete component in a circuit.
In the past, one method we have used to smooth data is a rolling average.
Normally this was done on relatively low frequency data (the data of
interest occured at a low frequency respective to the sampling rate), and
the frequency spectrum had never been of interest. In the current
experiment, I am sampling at 100Hz with an expected maximum frequency of
motion of 10Hz, a 10:1 ratio. Additionally, this time we *are* interested
in the frequency spectrum of the signal (the rate of postural sway in the
subjects - wich will be below 1Hz). I'm using a Fast Fourier Transform to
extract the frequency spectrum data.
My questions are this: how will using a rolling average to smooth the data
affect the frequency response of the system? How will changing the window
size of the rolling average change the frequency response?
Guidance or references on these questions would be greatly appreciated.
Dan Major
Univ. of Okla. School of Industrial Engineering
major@ou.edu htp://www.ecn.ou.edu/~major
---------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
---------------------------------------------------------------