- Garmin speed sensor. I realised after posting the chart this morning that my protocol has a slight flaw, because I didn't change the wheel circumference when I dropped the tyre pressure. Today, I measured the circumference changes, and there is almost a 2% reduction, going from 45psi to 15psi, so I need to correct the measured speeds and see how much that changes the CRR results. A slower speed for the lower pressures should give a higher CRR values for the lower pressures I think. I'll re-post the corrected chart once I do it.

Will increase it, its trivial to do, even if just for one user :)

Cheers
Mark

This is an example do it in Anadia velodrome. Wooden surface. There are many options for Crr and Cda values.

Any link where see specific protocol for indoor velodrome for calculate Cda and Crr values?

Is require to know when and where start the test?

My question is about if it´s possible to determine in stable test Cda and Crr or is necessary change velocity.
And with it we can get lot of combination numbers for Crr and Cda.

My question is about if it´s possible to determine in stable test Cda and Crr or is necessary change velocity.
And with it we can get lot of combination numbers for Crr and Cda.

Estimating CdA and Crr from the same data set is possible but it's pretty demanding of good data and good experimental technique. The short answer is it's necessary to change velocity. The fuller answer will require a little dip into statistical estimation. I'll try to keep it short and understandable.

In order to estimate two unknowns, you need to have at least two equations. If all your data are taken at a "stable test" at a single speed or power, you'll only have one equation so the system is undetermined. What's also true is that if the data vary but only by a little bit, you you can solve for estimates of CdA and Crr but the data are *nearly* collinear so although you were able to do the estimation, they're estimated poorly.

So, if you're trying to estimate both CdA and Crr from the same data set, you have to have a pretty wide range of speed and power--and it's good to have a lot of equations so you can check the precision of the estimates (there's a bunch of statistical mumbo-jumbo embedded in the last part of that sentence). A secondary issue is that you'd like the speed and power to be pretty accurate and precise, and a third issue is that you don't want there to be wild swings in speed and power because some power meters are a little sensitive and laggy in responding to big changes in power.

All together, if you want to separate CdA and Crr, you need accurate data and a wide range in speed and power.

Andy Shen once suggested a monotonically increasing speed protocol for velodrome testing: start slow and increase speed a little bit each lap. There are other ways to ensure that you get a pretty wide range in speed and power, but this will work. Find the top power you can sustain for a couple of laps, then start at perhaps half of that power, then each lap increase speed smoothly so that after 10 laps or so you're at your max that you can sustain without ruining your position. If you can, do a couple of extra laps slowing down in speed and power. A protocol like this ensures both a wide range in speed and power, and also keeps sudden changes in them relatively small.

In theory yes, under the assumption that both CdA and Crr are not dependent on any other variable and especially not dependent on speed.

But:

Both assumptions are questionable.

The normal force varies with speed on a track due to the centrifugal force in the turns


It is difficult to get a wide range in speed on a track without altering the position. One needs a minimum speed (> ~25 km/h) in the turns in order to stay on the track. To go over 50 km/h for some time is not given to everyone.

Overall one has to be very lucky to get such good data to separate CdA and Crr. I never had this luck.

Estimating CdA and Crr from the same data set is possible but it's pretty demanding of good data and good experimental technique. The short answer is it's necessary to change velocity. The fuller answer will require a little dip into statistical estimation. I'll try to keep it short and understandable.

In order to estimate two unknowns, you need to have at least two equations. If all your data are taken at a "stable test" at a single speed or power, you'll only have one equation so the system is undetermined. What's also true is that if the data vary but only by a little bit, you you can solve for estimates of CdA and Crr but the data are *nearly* collinear so although you were able to do the estimation, they're estimated poorly.

So, if you're trying to estimate both CdA and Crr from the same data set, you have to have a pretty wide range of speed and power--and it's good to have a lot of equations so you can check the precision of the estimates (there's a bunch of statistical mumbo-jumbo embedded in the last part of that sentence). A secondary issue is that you'd like the speed and power to be pretty accurate and precise, and a third issue is that you don't want there to be wild swings in speed and power because some power meters are a little sensitive and laggy in responding to big changes in power.

All together, if you want to separate CdA and Crr, you need accurate data and a wide range in speed and power.

Andy Shen once suggested a monotonically increasing speed protocol for velodrome testing: start slow and increase speed a little bit each lap. There are other ways to ensure that you get a pretty wide range in speed and power, but this will work. Find the top power you can sustain for a couple of laps, then start at perhaps half of that power, then each lap increase speed smoothly so that after 10 laps or so you're at your max that you can sustain without ruining your position. If you can, do a couple of extra laps slowing down in speed and power. A protocol like this ensures both a wide range in speed and power, and also keeps sudden changes in them relatively small.

My doubt about this is if Cda is velocity dependent, and if we are doing many test at different velocities for estimate one value of Cda and one for Crr, could be that only Crr value is correct? Or neither?

I understand that we need 1 Crr but many Cda´s for each velocity.

It is often stated, that because power to overcome aero drag scales with the cube of speed and power to overcome rolling resistance scales linearly with speed it is possible to separate Cda and Crr. I was never successful to separate them with actual power data from an aero test on an indoor track.

Based on outdoor testing I'm pretty sure Crr increases with speed, making the relationship with power non-linear. At least that is the only way I could get curves to match.

Cda is dependent for velocity only for Reynolds number?

For regression method if Cda is not accuracy (for Reynolds number), can report Crr not accuracy, is this correct? Because one variable depend on the other.

If this were true, how would you explain not seeing it on indoor roller tests ?

The essential piece is: over the speeds at which we usually ride, CdA is not dependent on speed. If it were, we wouldn't be able to predict pwer for speed based on wind tunnel tests that are done at a single speed.

If this were true, how would you explain not seeing it on indoor roller tests ?

When Dan Bigham did skinsuit tests in the velodrome he'd typically test at a range of speeds and showed CdA dropping with speed (typically, not always). https://www.cyclingweekly.com/...s-for-cycling-360975