Maybe also measure tarmac temperature. You will also have some kind of indication on temperature changes on your head unit file. I would think tyre temperature tracks decently close to air after riding a short while.
Temperature does indeed affect Crr.
I’m thinking more and more that I’d be best of doing this at some ungodly hour in the morning. Like starting at 3 or 4 AM. You can get really consistent air temps (often unchanged hourly forecast), really low wind (under 3 mph reported), and perfectly consistent sun (zero).
No cars or pedestrians is a bonus. I’m sure so can find places to mount lights.
It’s a suburban street lined with houses, so anyone who’s up might reasonably wonder what in god’s name someone is up to riding a weird looking bike in the middle of the night, wearing a skin suit, in an aero tuck, rolling back and forth on their half mile street at 10 mph.
But all in the name of science, right?
Hmm 23mm works perfectly on an outdoor 333m concrete track for nationals. If you want to sell one, I know a guy looking for a fast tire.
Attached a plot as incentive to ride the same line every time to the best of your ability. This is not some extreme example with chunks of asphalt missing by the way. In a blind test, you would need to concentrate to feel the difference, yet it clearly shows in the FFT of the acceleration signal.
Attached a plot as incentive to ride the same line every time to the best of your ability. This is not some extreme example with chunks of asphalt missing by the way. In a blind test, you would need to concentrate to feel the difference, yet it clearly shows in the FFT of the acceleration signal.
Interesting. I’m assuming the “Raw accelerometer output†is voltage (mv?) with a linear correlation to Gs.
To me this chart raises the question: how do these vibration differences correlate to differences in average CRR over the courses? Do these 2 paths reflect a 0.1% difference in total tire-ground losses or a 10% difference?
I’m thinking the magnitude of CRR difference between those courses depends on whether or not the bike was in a tire pressure / surface roughness zone where impedance / vibration losses are significant compared with losses from hysteresis.
The attached chart (from https://silca.cc/…=e25396f57&_ss=r ) shows how impedance losses are insignificant, until psi rises to a certain point; and suddenly they matter a lot.
So maybe the takeaway here is that the test should be run using the a lower pressure to minimize the magnitude impedance losses, and thus minimize the difference in impedance losses between runs.
I had been planning to run ~88 psi, as that’s what I usually race at. But maybe I’ll go with the value recommended by the HED chart (here: https://www.slowtwitch.com/...e_Pressure_7410.html ), which I know has been criticized for being too low for optimal speed compared with other guides, but I’m sure they had their reasons (comfort, grip, etc.). I’m 155 lb so their chart has a 23mm tire at 72 psi.
That should do the trick as far as controlling differences impedance losses.
Also, a more straightforward control, I’ll be marking waypoints using chaulk to try and get each run as consistent as possible.
Keep the input coming!
I did some runs today with both Corsa Speed 2 (CS2) and Veloflex Record (VR). Early morning with no wind. The goal was to get a CdA estimate to use for CRR tests, and do some low-speed rolldown tests (5-13 mph) with set start-stop points (consistent elevation drop) to find CRR values using VE.
About 1/3 of the CRR results (9 tests total) are not consistent with the rest, including all 3 tests of the Veloflex. Perhaps there were some fluke wind gusts. I got consistent CS2 results both before and after those.
However the other 2/3 of the CRR results are extremely consistent, and I think the low speed rolldown test is extremely promising to detect small CRR differences.
I’ve got some additional work to do on the data, and I may need to re-run the tests, but progress is being made. And I don’t want to jump the gun, but even with the variability in results, the Veloflex is looking solid. Though not as spectacular as the Aerocoach results suggest.
If you are really only interested in Crr then you could do all the testing on a pair of rollers. The simplest test is to put the different tires on similar front wheels and ride at a constant speed on the rollers. You can then compare the average wattage for each tire, after you let each tire warm up. The nice thing about rollers is that they exaggerate the wattage difference between tires (I think the difference is a factor of 3ish), so it is pretty easy to pick the tire with the lowest Crr.
While roller testing is great, it doesn’t give the complete picture. My hunch is that is a big part of why we’re seeing discrepancies in the Veloflex Record results.
Actually, as long as your tires are not over-inflated for the surface, roller testing gives the BEST results when trying to evaluate tires for TT/Tri purposes. Rollers aren’t the source of the VF Record result discrepancies…more likely is product changes.
I did some runs today with both Corsa Speed 2 (CS2) and Veloflex Record (VR). Early morning with no wind. The goal was to get a CdA estimate to use for CRR tests, and do some low-speed rolldown tests (5-13 mph) with set start-stop points (consistent elevation drop) to find CRR values using VE.
About 1/3 of the CRR results (9 tests total) are not consistent with the rest, including all 3 tests of the Veloflex. Perhaps there were some fluke wind gusts. I got consistent CS2 results both before and after those.
However the other 2/3 of the CRR results are extremely consistent, and I think the low speed rolldown test is extremely promising to detect small CRR differences.
I’ve got some additional work to do on the data, and I may need to re-run the tests, but progress is being made. And I don’t want to jump the gun, but even with the variability in results, the Veloflex is looking solid. Though not as spectacular as the Aerocoach results suggest.
Nice work. I am trying to decide on running the Power TT again this year or switch to the Records. They are just so thin but so tempting. I ran them yesterday as a mock TT test but I know my roads are clean here.
Thanks! The Power TT is actually another one I want to measure, as Aerocoach shows the 23mm (25.6w) as far slower than the 25mm (22.2). This large discrepancy between 23mm and 25mm tires isn’t present in any other tire Xav tested.
I actually bought a pair of 23mm TTs based on BRR results for the 25mm version, but went to CS2 after seeing Xav’s results.
I’m having trouble finding the time & wind-free days to get the VFR vs. CS2 test done. But when I finally do I’ll aim to do a 23mm CS2 vs TT next!
I have a lot of faith in the rolldown test method with VE that I’ve worked out. The 3x CS2 runs 80 psi gave CRR results of .300 / .300 / .301 when I did VE, no tweaking needed. And one of the runs was done 45 minutes after the first 2, with testing the Records in between. The Record runs weren’t as consistent but both beat the CS2.
I’m still determined to get this done just need another chance between life and the weather.
low-speed rolldown tests (5-13 mph) with set start-stop points (consistent elevation drop) to find CRR values using VE.
I vary the “entry” speed at the start point (and thus, the stop point).
low-speed rolldown tests (5-13 mph) with set start-stop points (consistent elevation drop) to find CRR values using VE.
I vary the “entry” speed at the start point (and thus, the stop point).
Interesting, for what purpose?
I have been working hard to get entry speeds consistent, in an effort to reduce the impact of error in the CdA estimate on the CRR delta between tires.
low-speed rolldown tests (5-13 mph) with set start-stop points (consistent elevation drop) to find CRR values using VE.
I vary the “entry” speed at the start point (and thus, the stop point).
Also I don’t want end the run at a “stop point†of zero velocity. I want to end the run at the start point, just with less velocity (of course), so I can be sure of zero net elevation change and use that as a known value in VE.
Thanks! The Power TT is actually another one I want to measure, as Aerocoach shows the 23mm (25.6w) as far slower than the 25mm (22.2). This large discrepancy between 23mm and 25mm tires isn’t present in any other tire Xav tested.
I actually bought a pair of 23mm TTs based on BRR results for the 25mm version, but went to CS2 after seeing Xav’s results.
I’m having trouble finding the time & wind-free days to get the VFR vs. CS2 test done. But when I finally do I’ll aim to do a 23mm CS2 vs TT next!
I have a lot of faith in the rolldown test method with VE that I’ve worked out. The 3x CS2 runs 80 psi gave CRR results of .300 / .300 / .301 when I did VE, no tweaking needed. And one of the runs was done 45 minutes after the first 2, with testing the Records in between. The Record runs weren’t as consistent but both beat the CS2.
I’m still determined to get this done just need another chance between life and the weather…
Do you have a protocol for warming up the tires before each run? I see a pretty big change in Crr on the rollers until everything gets up to a constant temperature, which I assume is the equilibrium point between international friction and heat loss through the roller and surrounding air. I definitively need to get out my new IR thermometer next time I do testing to see how much temp and pressure change from cold to whatever the equilibrium temp is.
I vary the “entry” speed at the start point (and thus, the stop point).
Interesting, for what purpose?
It lets you double-check the estimated Crr and CdA.
I have been working hard to get entry speeds consistent, in an effort to reduce the impact of error in the CdA estimate on the CRR delta between tires.
I just estimate both CdA and Crr simultaneously from the data. There are two unknowns, so you need (at least) two equations to solve for the two unknowns. Varying the entry speed gives you the two equations.
Thanks! The Power TT is actually another one I want to measure, as Aerocoach shows the 23mm (25.6w) as far slower than the 25mm (22.2). This large discrepancy between 23mm and 25mm tires isn’t present in any other tire Xav tested.
I actually bought a pair of 23mm TTs based on BRR results for the 25mm version, but went to CS2 after seeing Xav’s results.
I’m having trouble finding the time & wind-free days to get the VFR vs. CS2 test done. But when I finally do I’ll aim to do a 23mm CS2 vs TT next!
I have a lot of faith in the rolldown test method with VE that I’ve worked out. The 3x CS2 runs 80 psi gave CRR results of .300 / .300 / .301 when I did VE, no tweaking needed. And one of the runs was done 45 minutes after the first 2, with testing the Records in between. The Record runs weren’t as consistent but both beat the CS2.
I’m still determined to get this done just need another chance between life and the weather…
Do you have a protocol for warming up the tires before each run? I see a pretty big change in Crr on the rollers until everything gets up to a constant temperature, which I assume is the equilibrium point between international friction and heat loss through the roller and surrounding air. I definitively need to get out my new IR thermometer next time I do testing to see how much temp and pressure change from cold to whatever the equilibrium temp is.
To account for tire temperature, I used a SEEK Compact phone-mounted IR camera to measure tire temps before and after the runs.
https://www.thermal.com/compact-series.html
I did play around with this a bit. It was early morning and overcast, no real sun loading. I found that riding on the tires did not increase tire temps whatsoever, taken immediately upon a brake-less stop (using a hill to stop the bike).
The one thing that did drive tire temps up was braking. I didn’t do any high speed or repeated braking that day, but the light braking I did would drive the brake surface & tire temps up 3-8 degrees. Tires were almost exactly the same temp as the braking surface - I guess the heat conducts very quickly from the wheel compared with the low thermal mass of the tire.
The wheel-tire system then cooled to ambient after a few minutes.
Once I’d established that tire temps immediately post-run we’re exactly the same as pre-run (assuming no braking) then I’d just record tire temps before the run, ensure they weren’t elevated, and avoid braking. The roll downs are low speed (to maximize the ratio of tolling resistance to air drag) so this wasn’t actually too hard to do.
I vary the “entry” speed at the start point (and thus, the stop point).
Interesting, for what purpose?
It lets you double-check the estimated Crr and CdA.
I have been working hard to get entry speeds consistent, in an effort to reduce the impact of error in the CdA estimate on the CRR delta between tires.
I just estimate both CdA and Crr simultaneously from the data. There are two unknowns, so you need (at least) two equations to solve for the two unknowns. Varying the entry speed gives you the two equations.
Got it. I remember this from your original VE paper and have tried it once before, though not too carefully.
Have you done direct CRR comparisons using this method? I’m curious how much precision you can reliably get for a delta CRR.
I think I actually am basically doing what you describe. I did a high speed run first to establish CdA using an estimated CRR. For the low speed run, instead of a standard VE test with the “zero elevation change low speed rolldownâ€.
The final CRR and CdA combination does have to work for both runs. Although worth saying… I was initially worried about relying on high- and low- speed runs for precision work because CRR and Cd vary with speed, and I’m doing much of the low speed runs under 10mph, where I think this is probably significant and hard to control. That’s l part of what led me to low speed testing, where CdA errors are less of an issue, and rolldown testing, where speed profiles between low speed runs are very similar, further minimizing the impact of errors in absolute CdA / CRR values on “delta CRRâ€.
Ultimately it’s actually not that hard to validate a delta CRR method. You can do runs with the same tire at different pressures, where you have a small but roughly known CRR change (say, based on BRR’s data at 60-80-100 psi). If the deltas you measure lines up roughly with what was determined expect from 3rd party CRR tests at different pressures, then I’ll consider it validated.
I had some luck with this validation with the CS2s on the first day. I’ll ultimately post the validation data along with final results when I can get another full round of good tests done.
No ETA on all this unfortunately.
Have you done direct CRR comparisons using this method? I’m curious how much precision you can reliably get for a delta CRR.
Only for proof of concept. I usually do powered runs, not coastdowns. Coastdowns take at least twice as long (cuz you have to go back up to the start) but they do have the advantage that you don’t need to worry about the accuracy of the power meter – so if I had a single-sided power meter I might do coastdowns. However, since you’re relying on the speed being accurate, you absolutely need to measure loaded rollout carefully and use a wheel sensor rather than relying on something like GPS speed.
Have you done direct CRR comparisons using this method? I’m curious how much precision you can reliably get for a delta CRR.
Only for proof of concept. I usually do powered runs, not coastdowns. Coastdowns take at least twice as long (cuz you have to go back up to the start) but they do have the advantage that you don’t need to worry about the accuracy of the power meter – so if I had a single-sided power meter I might do coastdowns. However, since you’re relying on the speed being accurate, you absolutely need to measure loaded rollout carefully and use a wheel sensor rather than relying on something like GPS speed.
Wheel speed sensor for sure 
My original concept for the low speed run was powered. But with extremely low speeds (~6-12 mph range where RR losses are comparable to aero losses in terms of scale) I ran into some issues…
First I suspect that power meters lose accuracy at low power levels, since they are strain gage based (I used Garmin Vector 3). To stay in the 6-12 mph range, steady state power would be solidly under 20 watts (when using racing tires, skin suit, and aero tuck).
I also doubt I could pedal at that low power without gaps, that would further degrade the data.
One workaround I tried was using an “impulse†method where I pedaled for 7 second intervals every time I hit 6 mph to boost my speed to 12 mph, but I got messy results when I plotted VE. I assume this is an issue with data collection & recording with short bursts and a lot of on/off.
So to get good data collection at 6-12 mph, the only way I could think of was rolldowns.
The rolldown tests I did first were from point A to B where I guessed the elevation change.
The next set will be from point A (starting around 12-15 mph) back to point A (ending at 4-6 mph). Elev change will then be known (zero), not estimated. I have a loop picked out but could also use an out-and-back that turns around on a slope.
It’s good that you’re trying different protocols, measuring ambient temp, and refining your technique. Crr is small so you need good, repeatable, technique to get reliable results. When you have the right technique to estimate Crr, that carries over to estimating CdA because CdA is, relatively speaking, an easier nut to crack.
… I was initially worried about relying on high- and low- speed runs for precision work because CRR and Cd vary with speed…
…Ultimately it’s actually not that hard to validate a delta CRR method. You can do runs with the same tire at different pressures, where you have a small but roughly known CRR change (say, based on BRR’s data at 60-80-100 psi). If the deltas you measure lines up roughly with what was determined expect from 3rd party CRR tests at different pressures, then I’ll consider it validated.
Crr increases quite a lot with speed on the roads where I tested… with the caveat that it could have been a mysterious “something else”. Temperature effects Crr too.
BRR’s data for Crr vs pressure would only be valid for the apparatus he was using. This would certainly not be valid on a road where vibration losses are a factor.
What you really want to do is compare the overall resistance of one tire vs another, at race speed and expected race conditions. If you are testing without airspeed measurement, then very low wind is your best bet, although you will miss aero differences. I suspect on most wheels the VFR will be aerodynamically better at yaw, because it is smaller.
Question is: What is the precision of your testing relative to the small difference in Crr you expect? Doing a bunch of test runs where you are trying to keep all variables the same (or compensate for the changes), will show you if you have a reasonable chance of picking the winning tire with any certainty. No tire swapping is necessary to do that.
Also, note that you need to nail the weighted rollout for each tire.