The Perceived Problem of Wind Tunnels


I was looking at some wind tunnel test footage and noticed the bikes are on rollers of some kind. This led me to think of a possible issue with wind tunnel testing.

When on the road, the bottom of your wheels are stationary to the ground while the top are moving 2x your relative speed. Would this fact alter the way a wind tunnel tests the drag of a given wheel since the bottom of the wheel moves? Or does the speed the wheels rotate match the wind speed?


If the top and bottom of your wheels and tires are moving at different angular velocities something has gone horribly wrong. The entire wheel stays at the same angular velocity because the rim is hopefully holding it together. If the bottom of the wheel was stationary to the ground and didn’t move when you’re riding you wouldn’t go anywhere. So the rollers actually simulate how a wheel actually moves when we’re riding.

The aerodynamic drag is calculated based on the amount of force the wind applies to the bike/rider/etc using a scale like apparatus. The rollers aren’t used for any kind of drag calculations. The wheel speed doesn’t need to match the wind speed it could be slower or higher.


The picture I was trying to create was the relative motion of the tires to the ground. The actual contact point is stationary to the ground unless you’re slipping. So my thought was that, assuming zero ambient wind, the bottom transverse speed of the wheel is zero and the top is 2x rider speed. The idea I had was that this would have a different effect in the real world vs wind tunnel if the rider speed (on rollers) did not match the wind speed.

Let’s say the wind tunnel had no rollers and the wheels did not spin, that would mean the wheel would be subjected to an even pressure profile rather than a sloped one.

I hope I didn’t make this more confusing. This is just a random thought experiment I had and thought I’d share.


This is interesting, if you think about it they could use a combination of roller speed and wind tunnel speed to simulate different wind conditions. Basically the delta between roller speed and wind tunnel speed would be wind.


The link recently shared in the “Aero Wheel Upgrade” thread mentioned this too but I haven’t had time to read up on how they tested it:


The only thing with that is the rider would experience a different wind speed than the wheels. And if you’re a wheel manufacturer, I would imaging you would want the roller speed to match the wind speed such that the relative wind speed at the bottom of the wheel is zero. In your example, yes you could simulate head winds now that I think about it.


Ok gotcha. So relative to the wind is what you mean when you said the wheel is stationary near the ground and twice the velocity of the wind near the fork. Assuming wind speed and roller speed are the same. That makes more sense thanks for clarifying.


Yes exactly, so since drag force increases by velocity squared, I would imagine the top of the wheel sees more drag than the bottom (which would be about zero with no additional wind). I just don’t know if you could use this information for a better wheel profile since every part of the wheel sees every direction.


Awesome read! They definitely address this in the paper and point out many other things. It’s interesting that, even though enve claims to optimize for 25c tires, the 23c ones were better. I’d like to see aero drag vs rolling resistance now. I still would imagine for a typical road surface, 25c is the way to go.


Flo Cycling (criticized in that link) did an analysis looking at a bunch of different tires on their own wheels, adding up the losses from aerodynamic drag and rolling resistance. The differences aren’t huge, but they did find (for example) that on their wheels a 25c Conti GP4000SII tire is essentially just as fast (ever-so-slightly faster, but within the realm of sampling error) as a 23 Conti GP4000SII.

But the 25c will be more comfortable, and less likely to flat.


This is more of a general comment on wind tunnel testing for cycling but…

the thing is, there is no universal wind tunnel standard, so any manufacturer can change their testing procedure to whatever suits them best no matter how “realistic” or practical it is, then claim it’s “the fastest” without ever having to disclose exactly how they reached their claim of “saving x watts at x speed” which is just hot air without any real context. I think it’s shocking that some component manufacturers and even independent reviewers do their “testing” without a rider/dummy in the wind tunnel at all.

What’s great about the Hambini article is that not only does it highlight this but also shows just how far behind the cycling industry is in aerodynamics. I think anyone who has even the slightest experience with physics knows that the difference between steady state and transient responses is worlds apart, it’s like taking Newtonian physics and applying it at quantum levels. FLO seems to take this to the extreme which makes it the perfect target to pick on–a couple dudes straight out of college design a rim to be theoretically “the fastest” at steady state and with minimal yaw, then throw a bunch of marketing at it to great success. The result is that they perform quite excellent in those specific conditions, but mix in a little oscillation and some ramping yaw angles like you would experience on the open road and the design becomes, very, very poor at handling reattachment.

To be fair though, apparently it takes wind tunnels on the level of Boeing or whatever other aerospace equivalent to even be able to properly test for things like transient analysis, which is obviously not on the table for companies in such a niche market like cycling.


I find it hard to believe that it’s worth trying to optimise anything for an effective 15 degree yaw angle (or greater) for a decent time trialist. At 50km/h forwards, to get a yaw angle of 15 degrees means you have to have a 15km/h wind, dead sideways, and 1m from the ground or less. While wind speed as defined by the meteorologist might hit these speeds regularly, they are measuring 10 metres above the ground, not at <1m.
The author states:
“the crosswind velocity on a bike often exceeds the forward velocity”
If the wind is at 50kph 1m from the ground, I’m not riding my bike that day, and most TTs would be cancelled on safety grounds as the meteorological wind speed is likely to be twice this!

Variability is interesting and something that is difficult to model as well as to test in a wind tunnel (in a repeatable way), but I don’t think their absolute numbers apply to many cycling events.


I agree. The ability to measure a transient analysis consistently appears to be extremely difficult. However, in my anecdotal experience, riding Reynolds Strike SLG disc to ENVE SES 5.6 Disc wheels made a huge difference in crosswind stability. That could come down to how the front wheels are different depths, but the Reynolds were quite jittery and made descenting sketchy whereas now, on the Enve wheels, it’s like a gentle nudge that’s predictable. They seem to know what they’re doing. I’m sure that info is basically secret at this point, though.


In real life it’s pretty rare and to have a constant crosswind at a single angle throughout the entire ride. Even at times where there seems to be constant crosswind, you’ll have random gusts of wind and all types of swirl coming at you, all of which effects the dynamic nature of aerodynamics where wind deflects off the rim and how well it reattaches to the trailing end severely affects speed. On another forum, Hambini mentioned that their yaw angles only look higher than “normal” (angles used by Mavic, FLO, etc) because it’s translated to show effective yaw rather than the straight up, absolute yaw–even the smoothest bike rider in the world can add up to 6 degrees of yaw in a perfectly windless and straight environment simply by pedaling normally and other manufacturers/protocols would consider this situation “zero degrees.”

That said, if you look at the “time spent at yaw angle” graphs, the majority of time spent for the transient testing is still ~15 degrees. The distribution from 0-15 is just more steady instead of sharp diving like in steady-state. This suggests that it’s important to be fast throughout 0-15 and not biased towards sub-5 degrees.

I think the added yaw bias of “rocking the bike” not being accounted for all in most wind tunnel yaw distributions is one of the biggest takeaways from the study.


I agree wholeheartedly with this. Even more so for the back wheel where the turbulence introduced by the riders legs has got to be massive. However, for a front wheel, in a TT, I reckon it’s closer to 0 degrees for more of the time. Sadly I have no data. Soon a bunch of people are going to be measuring this stuff with the various systems that are coming onto the market (And I understood that this was what Flo did to get their numbers (basically mounting a weather station out in front of someone’s bike). I guess we’ll get a lot of studies that we can poke through then (and when the price comes down enough be able to do our own). :slight_smile:


I definitely remember seeing a new product launch this year similar to a pitot tube that mounts under a Garmin or similar that measures real time wind, basically providing cda.


Sadly I have no data. Soon a bunch of people are going to be measuring this stuff with the various systems that are coming onto the market

Yup, I agree! If everyone would send me their equipment in question and the measuring equipment, I will volunteer myself to ride loops back and forth to Hawi every other weekend and record the data…for science! :grimacing:

However, for a front wheel, in a TT, I reckon it’s closer to 0 degrees for more of the time.

I think the most interesting data as of late is the stuff the Hambini team published. It seems like a good attempt to address the yaw angles seen in real life. Basically hundreds of rides with three pressure transducers mounted on the handlebars to get a measurement of yaw angles encountered. From that data of real rides, they designed their test to try and recreate the oscillating nature of yaw angles in their more advanced wind tunnel. From there, they showed that wheels like FLO which were optimized for a steady state and yaw angles mostly between 0-5 degrees failed when they did the oscillating stuff in their fancy aerospace wind tunnel.

I feel like it was a big blow to the FLO thesis and I have yet to see a good rebuttal. Has anyone else seen any updates on the debate?


Only responses I’ve seen by FLO on slowtwitch was that they “ensured” everyone that their testing procedure was focused on outdoor, “real world” scenarios and questioned the procedure such as the 23mm tire width (contis) used… hambini came back with results showing that 25mm contis were even slower.

Seems like they’ve stopped putting a hand in the public debate and, from what hambini is hinting lately, is one of the companies that have fired legal warning shots at them to take their result off the website… I’m no expert in law but I would think that they have no legal ground to stand on for an independent review like this if it comes down to an actual judiciary.

Disclaimer: I’ve actually owned FLO45s for up to a year and while I can’t speak to how fast/aero they were, I was very pleased by the quality and build of them to go with an excellent pricepoint (even the hubs seemed like a step up from your typical Novatec-rebrand). That said, reading all this, seeing FLO have no scientific response, then presumably just shutting down and playing the lawyer game instead brings them quite a notch down in credibility.


Thanks for the update! Yea, I’m all about the science of it and this seems like a blow against their credibility. I can understand their dilemma, but it’s usually the response and coverup that gets you in deeper trouble. I was about to get a set of deep FLOs as a set of racing wheels but due to the Hambini debacle, I’m a little less inclined to. Although, bang for the buck, a deep set of FLO aluminums might still be worth it cause it’s so cheap. We’ll see…


DC Rainmaker lists several manufacturers promising consumer units coming this year in his keynote (don’t have time to find it for you, sorry) here:
They include :

  • Alphamantis (currently used by track based people like Aerocoach),
  • SwissSide
  • Aerolab
  • RedIsFaster
    and a number of others. I have no idea which (if any) are good. :slight_smile: