Bike physics

I just loved reading through John Swanson's paper and results on
http://www.bikephysics.com/ . So much so, I recreated his experiments
at home, using my PC microphone input as a data acquisition device.
The resulting wav files I ran through an Octave script (free open
source mathematical analysis tool), and collated the results in a
spreadsheet.

It looks like my bearings need repacking at the least. The bearing
drag is substantially higher than those John measured. I replaced the
bearings in my front hub last year with double sealed cartridge
bearings from the local bearing supplier. The bearings I removed that
came from the factory were single sealed only on the outer. At least
I can make meaningful comparisons before and after, and compare other
peoples wheels, if they dare test their magic ceramic balls.

James.

"James" wrote in message
...
I just loved reading through John Swanson's paper and results on
http://www.bikephysics.com/ . So much so, I recreated his experiments
at home, using my PC microphone input as a data acquisition device.
The resulting wav files I ran through an Octave script (free open
source mathematical analysis tool), and collated the results in a
spreadsheet.

It looks like my bearings need repacking at the least. The bearing
drag is substantially higher than those John measured. I replaced the
bearings in my front hub last year with double sealed cartridge
bearings from the local bearing supplier. The bearings I removed that
came from the factory were single sealed only on the outer. At least
I can make meaningful comparisons before and after, and compare other
peoples wheels, if they dare test their magic ceramic balls.

Let's see if I got this straight. The test methodology is to unmount the
tire, spin the front wheel, and record the whoosh of the spokes as they pass
by the mic; and then repeat with a rim strip applied? It sounds too easy. It
would be even more too easy if you packaged an application to do the
recording and run the calcs.

On Aug 2, 11:48*pm, "MikeWhy" wrote:
"James" wrote in message

...

I just loved reading through John Swanson's paper and results on
http://www.bikephysics.com/. *So much so, I recreated his experiments
at home, using my PC microphone input as a data acquisition device.
The resulting wav files I ran through an Octave script (free open
source mathematical analysis tool), and collated the results in a
spreadsheet.

It looks like my bearings need repacking at the least. *The bearing
drag is substantially higher than those John measured. *I replaced the
bearings in my front hub last year with double sealed cartridge
bearings from the local bearing supplier. *The bearings I removed that
came from the factory were single sealed only on the outer. *At least
I can make meaningful comparisons before and after, and compare other
peoples wheels, if they dare test their magic ceramic balls.

Let's see if I got this straight. The test methodology is to unmount the
tire, spin the front wheel, and record the whoosh of the spokes as they pass
by the mic; and then repeat with a rim strip applied? It sounds too easy. It
would be even more too easy if you packaged an application to do the
recording and run the calcs.

Pretty close. I connected the contacts of the computer sensor direct
to the microphone input. That is I made a cable with a 3.5mm stereo
jack connected to the bike computer sensor (with the bike computer
disconnected). This differs from John's experiment only in that I
used my PC soundcard microphone input as a data acquisition device.

John's setup has a problem in that the data acquisition device he used
has a very low sample rate, 240Hz. As a result he gets very poor
accuracy in the measurement of rotation times. The PC microphone
input can sample typically at 44100Hz for CD quality recordings.

I used some copper wire instead of rim tape to add a known mass
(actually moment of inertia) to the rim. Didn't have spare rim tapes
available.

I really don't have time to package the whole shebang up,
unfortunately, at least not any time soon. I'm happy to share the
octave script and spreadsheet if anyone is interested though. (Octave
and Openoffice are free opensource applications that run under Linux
and Windows, maybe even Mac.)

Cheers,
James.

On Aug 2, 2:48*pm, "MikeWhy" wrote:
"James" wrote in message

...

I just loved reading through John Swanson's paper and results on
http://www.bikephysics.com/. *So much so, I recreated his experiments
at home, using my PC microphone input as a data acquisition device.
The resulting wav files I ran through an Octave script (free open
source mathematical analysis tool), and collated the results in a
spreadsheet.

It looks like my bearings need repacking at the least. *The bearing
drag is substantially higher than those John measured. *I replaced the
bearings in my front hub last year with double sealed cartridge
bearings from the local bearing supplier. *The bearings I removed that
came from the factory were single sealed only on the outer. *At least
I can make meaningful comparisons before and after, and compare other
peoples wheels, if they dare test their magic ceramic balls.

Let's see if I got this straight. The test methodology is to unmount the
tire, spin the front wheel, and record the whoosh of the spokes as they pass
by the mic; and then repeat with a rim strip applied? It sounds too easy. It
would be even more too easy if you packaged an application to do the
recording and run the calcs.

It's that mambo magic. You gotta have faith man. -- AJ

On Aug 2, 6:48*am, "MikeWhy" wrote:

Let's see if I got this straight. The test methodology is to unmount the
tire, spin the front wheel, and record the whoosh of the spokes as they pass
by the mic; and then repeat with a rim strip applied? It sounds too easy.

And probably is. All of the measurements are done with an unloaded
wheel. That should be ok for comparing the aerodynamic effects of two
wheels (although the actual power numbers will be off since the wheel
being tested isn't moving forward). But I'd be surprised if the
bearing drag figures are really meaningful. When unloaded these would
probably be dominated by the drag of the seals and any amount of
preload. I could easily see a situation where one wheel has less drag
than another when both are unloaded but has more drag when equal loads
are applied.

On Aug 3, 12:30*pm, Peter Rathmann wrote:
On Aug 2, 6:48*am, "MikeWhy" wrote:

Let's see if I got this straight. The test methodology is to unmount the
tire, spin the front wheel, and record the whoosh of the spokes as they pass
by the mic; and then repeat with a rim strip applied? It sounds too easy.

And probably is. *All of the measurements are done with an unloaded
wheel. *That should be ok for comparing the aerodynamic effects of two
wheels (although the actual power numbers will be off since the wheel
being tested isn't moving forward). *But I'd be surprised if the
bearing drag figures are really meaningful. *When unloaded these would
probably be dominated by the drag of the seals and any amount of
preload. *I could easily see a situation where one wheel has less drag
than another when both are unloaded but has more drag when equal loads
are applied.

Actually, according to the bearing loss calculator on the SKF website,
bearing losses due to load accounted for about 1/5 to 1/4 of the total
bearing loss. The rest was the bearing seal friction.

I imagine if the bearings were horribly degraded and/or poorly
lubricated then the bearing losses due to load would increase rapidly,
however this not being the case, load losses are typically small by
comparison to seal friction.

Regards,
James.

On Aug 3, 1:12*pm, James wrote:
On Aug 3, 12:30*pm, Peter Rathmann wrote:

On Aug 2, 6:48*am, "MikeWhy" wrote:

Let's see if I got this straight. The test methodology is to unmount the
tire, spin the front wheel, and record the whoosh of the spokes as they pass
by the mic; and then repeat with a rim strip applied? It sounds too easy.

And probably is. *All of the measurements are done with an unloaded
wheel. *That should be ok for comparing the aerodynamic effects of two
wheels (although the actual power numbers will be off since the wheel
being tested isn't moving forward). *But I'd be surprised if the
bearing drag figures are really meaningful. *When unloaded these would
probably be dominated by the drag of the seals and any amount of
preload. *I could easily see a situation where one wheel has less drag
than another when both are unloaded but has more drag when equal loads
are applied.

Actually, according to the bearing loss calculator on the SKF website,
bearing losses due to load accounted for about 1/5 to 1/4 of the total
bearing loss. *The rest was the bearing seal friction.

I should add that these numbers were for some assumed loads I felt
were typical of those experienced by bearings in normal use. I think
it was 250N (approx 25 kg). This is per bearing of course. 50kg on a
wheel is not unreasonable, but depends a lot on the rider and position
on the bike of course.

This was for bearings 61901-2RS1. Double sealed cartridge bearings
compatible with my Mavic hubs (although the factory ones only have one
seal on the outside surface).

What is interesting from the SKF calculator, is that as speed
increases the sliding friction moment decreases, while the rolling
friction moment increases and the frictional moment of seals remains
constant.

The formulas used by SKF are full of exponents and factors I have no
intention of investigating.

JS.

On Aug 2, 8:12*pm, James wrote:
On Aug 3, 12:30*pm, Peter Rathmann wrote:

On Aug 2, 6:48*am, "MikeWhy" wrote:

Let's see if I got this straight. The test methodology is to unmount the
tire, spin the front wheel, and record the whoosh of the spokes as they pass
by the mic; and then repeat with a rim strip applied? It sounds too easy.

And probably is. *All of the measurements are done with an unloaded
wheel. *That should be ok for comparing the aerodynamic effects of two
wheels (although the actual power numbers will be off since the wheel
being tested isn't moving forward). *But I'd be surprised if the
bearing drag figures are really meaningful. *When unloaded these would
probably be dominated by the drag of the seals and any amount of
preload. *I could easily see a situation where one wheel has less drag
than another when both are unloaded but has more drag when equal loads
are applied.

Actually, according to the bearing loss calculator on the SKF website,
bearing losses due to load accounted for about 1/5 to 1/4 of the total
bearing loss. *The rest was the bearing seal friction.

I imagine if the bearings were horribly degraded and/or poorly
lubricated then the bearing losses due to load would increase rapidly,
however this not being the case, load losses are typically small by
comparison to seal friction.

Most of my hubs have no bearing seals, but the bearings are probably
not as precisely machined as the sealed cartridge type. Still
wouldn't be surprised therefore if comparisons based on a zero-load
test failed to accurately represent performance when loaded.

On Aug 2, 10:07*pm, Peter Rathmann wrote:
On Aug 2, 8:12*pm, James wrote:





On Aug 3, 12:30*pm, Peter Rathmann wrote:

On Aug 2, 6:48*am, "MikeWhy" wrote:

Let's see if I got this straight. The test methodology is to unmount the
tire, spin the front wheel, and record the whoosh of the spokes as they pass
by the mic; and then repeat with a rim strip applied? It sounds too easy.

And probably is. *All of the measurements are done with an unloaded
wheel. *That should be ok for comparing the aerodynamic effects of two
wheels (although the actual power numbers will be off since the wheel
being tested isn't moving forward). *But I'd be surprised if the
bearing drag figures are really meaningful. *When unloaded these would
probably be dominated by the drag of the seals and any amount of
preload. *I could easily see a situation where one wheel has less drag
than another when both are unloaded but has more drag when equal loads
are applied.

Actually, according to the bearing loss calculator on the SKF website,
bearing losses due to load accounted for about 1/5 to 1/4 of the total
bearing loss. *The rest was the bearing seal friction.

I imagine if the bearings were horribly degraded and/or poorly
lubricated then the bearing losses due to load would increase rapidly,
however this not being the case, load losses are typically small by
comparison to seal friction.

Most of my hubs have no bearing seals, but the bearings are probably
not as precisely machined as the sealed cartridge type. *Still
wouldn't be surprised therefore if comparisons based on a zero-load
test failed to accurately represent performance when loaded.- Hide quoted text -

- Show quoted text -

WHY ?

In article
,
James wrote:


[...]


Third, he numerically differentiates his data twice.
Those skilled in the art would no more do this than
sell their mother down the river; probably less likely.

Possibly not ideal given the number of samples per rev, which could be
addressed by using multiple magnets evenly spaced, but in the absence
of anything better.. At least it's not as bad as numerical
integration.

Au contraire. You speak knowingly, yet the mathematics
contradicts your assertion. Numerical integration
smoothes, numerical differentiation introduces noise.

Numerical differentiation of data is not undertaken
lightly. The error analysis needs to be done. ND
amplifies noise, and real world data contains noise;
unknown, unexpected, perverse noise.

--
Michael Press