suspension FFT

[gregpe]

Hi
Motec forum - Really great idea:).
I have a question how to interpret suspension FFT chanell?
Greg

[JamieA]

Here is some info on FFT's lifted from i2

FFT - The FFT performs a Fast Fourier Transform on a number of channels.

A FFT shows the density of frequency components for the selected data.

The maximum frequency that an FFT can represent is equal to the channel sample rate divided by two.

FFTs are often used to analyse Suspension position data.

Windowing
When an FFT is performed it is performed on a segment of data which has an abrupt start and end which can cause errors in the FFT.
To reduce this affect a windowing function can be applied to the data which reduces the amplitude of the data to zero (or near zero) at the start and end of the data – this is called windowing.A number of common windowing functions have been provided – in practice either windowing function should give satisfactory results.

FFT Points
The number of FFT points sets the number of points used in the FFT calculation, if the number of points selected by the zoom level is greater than the FFT Points setting then multiple FFTs are performed and the results are averaged.

The data is zero padded to ensure the total number of points is a multiple of FFT Points.

Note that the number of FFT points directly affects the number of frequency bins (The number of frequency bins = half the number of FFT points)

Smoothing
The smoothing function performs an averaging function on the frequency bins. The number of bins that are averaged is the smoothing percentage multiplied by the number of frequency bins (which is half the number of sample points).

Note that when smoothing is applied some of the first and last frequency bins will be lost.

MATLAB
The i2 FFT has been tested against the Matlab equivalent.
To get equivalent results in Matlab use the Welch method which is an extension of the Bartlet method used by i2.

Example:
PSD of channel x at rate r with FFT points set to 1024.
pwelch(x, 1024, 0, 1024, r)
Note – ensure all axis settings are equivalent when comparing results.


Others with more info??

[Holmz]

Good synopsis.

A Hanning window is generally optimal.

If you ran the car at various speeds you should potentially see the frequencies where the system resonates change
assuming the surface is the same.
The car is pretty heavy and the springs in the suspension pretty stiff so the resonant frequency will be in the Hz range - (< Hz for a boat of a car and maybe a few Hz for a stiff car).

The suspension is light so the wheels will generally have a resonant frequencies that are independent of speed.
If you have some strong high frequency components - that indicates issues with the unsprung side.

If you took the dampers out you would see these components really stick up.
If you are not seeing them then the suspension is doing its job to prevent the fundamental frequency(s) from getting excited.
A washboard road would provide a good surface to pump the system.
I believe that the transition you generally feel at 5-20km/h is the chassis fundamental frequency.
You might find the unsprung resonance if you go fast enough and the dampers are backed off.