Vibration & Rollers

When studying rolling resistance while designing the FLO All Sport and Gravel wheel lines, we used rollers as well as on-road testing to collect data. The rollers allowed us to apply temperature compensation to the on-road data. Rollers produce a clean lab-like environment. We wanted to test the vibration produced while on rollers to gain more insight into measuring impedance breakpoints.

Why did we want to look at vibration data on rollers? The main reason was to remove as much noise from the road as possible to see what we would find. The data we’ve been getting from the sensors on road is complex. If we eliminate a lot of the broad spectrum noise from the road discussed in Natural Frequency: Does It Matter In Cycling?, what would we see in the FFT plot?

To make the data comparable, we ran the same pressure range of 60-120 psi in 5 psi increments that we did in our on-road testing.

We decided to mount the sensors in two locations. The first was on the front fork, and the second was on the wheel. This matched the locations for our on-road testing.

To date, we have been looking at excitation (vibration) patterns in the data based on the natural frequency of each wheel. What we’ve not been able to see is much detail in the FFT plot due to all the broad-spectrum noise. Using the rollers, we were able to remove the broad-spectrum noise from the road to produce a very clean signal. Below is a graph of the road data at 85psi, followed by a graph of the roller data at 85 psi. We used the resultant data for each chart.

The roller data is creating clean and cyclical spikes on the FFT plot. It’s interesting to point out that the excitation in the wheel’s natural frequency range still exists.

You will also see a series of cyclical spikes in the plot. What are these coming from? This took some time to determine, but here is what we believe is happening.

FFT is a frequency plot. On the roller graph above, you see that the cyclical spikes happen roughly every 3 Hz. The test was conducted at roughly 15 mph, which means that the wheel rotates roughly three times per second at that speed. In vibration, you can get a harmonic effect on multiples of the input frequency, which means that even though the wheel is spinning at 3 Hz, you can see the signal at any multiple of three. We believe this is why we are seeing the clean spikes on the rollers when the broad-spectrum road noise is removed.

You may think that we would be able to measure an impedance break point on rollers based on this data, but there is a key difference in the road vs. roller data—the magnitude of the vibration data. Below are four pairs of charts. The first two pairs are the road data at 60 psi and 120 psi for the fork and wheel, and the third and fourth pairs are the roller data at 60 psi and 120 psi. If you look at the y-axis for the wheel sensors in the road and roller data, you will notice that the energy does not change on the rollers when comparing 60 psi to 120 psi, but that it does change when looking at the data on the road. This is strong evidence that we will be able to determine a break point by measuring vibration on the road.

The rollers have taught me a lot. First, excitation is happening on both the rollers and the road, but our acceleration energy is only changing on the road. I’ve also learned that pure rotation creates excitement in the wheel. That honestly surprised me. I did not expect to see excitation the natural frequency range on the rollers like we did. This means that wheel speed will play into the equation somehow. It may be that the faster we go the more resolution we see on the FFT plot, but it may also produce more energy on the excitation at the natural frequency range.

We have more testing to do. We will be back shortly with more on this project. Stay tuned for more blogs next week that are a bit lighter reading.