Re: Suspension losses

On 1/2/2025 11:06 AM, AMuzi wrote:

On 1/2/2025 9:42 AM, bp@www.zefox.net wrote:

AMuzi <am@yellowjersey.org> wrote:

>
I don't have a coherent argument either way but a rumble
strip test introduces a repeatable experience so that
various data may be compared.  Each rider on a dirt or
gravel path, and each ride experience by any given rider, is
an unique set of impedimenta such that data cannot be as
readily compared.

>
AMuzi <am@yellowjersey.org> wrote:

>
I don't have a coherent argument either way but a rumble
strip test introduces a repeatable experience so that
various data may be compared.  Each rider on a dirt or
gravel path, and each ride experience by any given rider, is
an unique set of impedimenta such that data cannot be as
readily compared.

>
But one can observe that in the case of smooth pavement,
suspension losses vanish, while hysteresis losses persist.
>
In the end a bike is an overdamped resonator excited by the
pavement and damped by hysteresis, separately in the tire and
suspenesion. In that limit, suspension would be faster if used
with very hard tires on very smooth surfaces. In the limit of
hard tires and no suspension, the dissipative element becomes
the rider whose elastic properties are apt to be poor, perhaps
accounting for the apparent slowness of solid tires.
>
Use of a rumble strip for testing is equivalent to selecting
a particular excitation spectrum. Choice of spectrum will affect
dissipation depending on internal resonances of the bike/rider
system. A real road likely corresponds to a 1/f spectrum, but
a rumble strip will likely be something else. How much difference
that makes isn't clear but it could be estimated using a mechanical
analogy equivalent circuit of the kind used to model loudspeakers.

This is a great analysis and reveals a highly problematic aspect of the "rumble strip" test. As Bob notes, it's essentially limiting the noise input into the system to a somewhat narrow spectral component (though the 1/f assumption for real-world is way to broad)
The idea of using the rumble strip test seems adequate at first, but is prone to misleading results. Since the rumble strip sets up a regular frequency component, there's a possibility that a resonance or cancellation effect can occur which can dramatically skew the results.
In the world of environmental testing, physical vibration analysis is typically broken up into three different stimuli - swept frequency, noise*, and environmental specific (usually a combination of noise with higher energy components around certain frequencies).
It's nearly impossible to simulate all the possible real-world conditions, which is why the testing regimen includes a sweep - the intent being to see any resonances. I've personally witnessed an electronic assembly quite nearly disintegrate with the right frequency and energy input. The task then was to redesign the piece such that the resonance was damped.
It's easy to see how this can translate to the rumble strip test. Under the right conditions, one might actually see a speed _increase_ as a result of a sympathetic resonance.

>
A pair of series RLC circuits (one for the road-tire interface
and a second for the suspension-rider interface) would be a good
start. I'm not skilled enough to do the calculations, but others
on this group likely are. The hardest part is apt to be finding
an equivalent circuit for the rider, who isn't a rigid mass but
rather a dissipative blob....8-)

In the old days, we had to do reiterative tests on massive vibration tables. These days, FEA software removes the vast amount of guesswork. The last few times I've had to conduct these tests I only had to do one test twice, and the problem turned out to be an assembly specification error rather than inherent design.

Thanks for reading,
>
bob prohaska
>

 Clever.
 I take from that, you think the actual impact/height change/velocity change etc from various irregular surfaces can be quantified for any given random gravel (or road) experience and used to compare efficiency for other iterations.

"Real-world" would simulate a more stochastic environment with larger "impact" events rather than a regular "sinusoidal" spectrum like a rumble strip. Currently, for example, we use this for our truck-mounted electronics:
https://cvgstrategy.com/wp-content/uploads/2019/08/MIL-STD-810H-Method-514.8-Vibration.pdf
Refer to page 514.8C-5 (Page 58 in the PDF).

 I hadn't thought of that, but if that's true then the rumble strip test isn't necessary for comparison. Which assumes sensors have adequate sensitivity across whatever range and that software for that data truly derives actual impedimenta values.

Even low-cost accelerometers have incredible accuracy, sensitivity, and repeatability across spectrum they're designed to operate these days. We have two 3-axis units accurate to .01G that we paid like $25 each for - coupled to a mid-range oscilloscope they give more than adequate results for our "warm fuzzy" testing before we send of to a testing lab for 3rd party analysis.
*"Noise" being a broad term meaning quasi-random frequency and amplitude components within limits.
--
Add xx to reply