The Latest

FEATURED BUILD - FSW3 with PowerTap.

Our last trip to the wind tunnel proved that the Kinlin rims in our FSW3 wheels are every bit as fast as those 40+mm carbons you use on race day. So now that your everyday alloys can also be your game day wheels, there's no better time to add a PowerTap. Especially since we've added tires (installed) and knocked $135 off the price.



Subscribe to our emails:

Our emails include resources, tips and insights to help make you a better cyclist and a more informed buyer, whether you buy from us or not. If you like, sign up below and we'll send you the next one.

Site Search
« Meaningful Differentiation | Main | Wind Tunnel Testing the Al33, XR31T(FSW3), and other alloys »

Wind Tunnel Testing the Al33, XR31T(FSW3), and others, Part 2

Yesterday's was a quick post to show the basic results. Today's post will explain the Angle of Attack distribution that we've used, and fill in some of the finer details on the test. 

The test occurred at A2 Wind Tunnel in North Carolina. A2, along with the San Diego Low Speed Wind Tunnel, is one of the two default standard, publicly accessible wind tunnels in the US. If you want credible aerodynamics data, A2 is a an outsanding place to get it. No one from November was at A2, we Skyped in during the test. The test is a very standard one - 30mph test wind speed, temperature and pressure are normalized. We used a 23c sized (Continental's chosen descriptor) Continental GP4000sII tire, inflated to the wind tunnel standard 100psi. The alloy wheels used a tube with a 48mm valve stem while the Zipp used a 60mm valve stem, so valve stem protrusion was normalized as closely as possible. Other details of the builds have been detailed before

The standard test sweep is to go from 0* to 20* angle of attack, or yaw angle, in 2.5* increments. This is enough resolution to give an accurate representation of how the wheels perform through any statistically significant wind situation you will encounter. Wheels were tested alone, front wheel only. There are already some rattles of "this is irrelevant because it doesn't account for frame and fork" pushback on this. Simply, testing wheels standalone has decisively proven to have outstanding transfer to their performance in a bike system, and it's impossible to test with the range of bikes/forks/situations to satisfy everyone. The validity of this test's scope is established legislation, which you are free to relitigate as you wish, but it's not something we'll engage in arguing. 

Al33 in the tunnel at A2

Wind speed is 30mph. This is the standard test speed as it's been established to give the cleanest data. You can scale with software to produce results for more or less air speed, but the shape of the curves doesn't change - a wheel that's a laggard at 30mph doesn't become a star at 20, it stays a laggard. But since the effects of air resistance increase so quickly with air speed, the differences between wheels get compressed at lower air speeds.If you want to do the quick and dirty calculations on watts versus grams of drag versus time in the mythical 40k TT, here's the teacher's edition: using the 30mph parameter, 10g of drag roughly equals 1 watt, and 1 watt roughly equals 3 seconds in the mythical 40k. Not good enough for real science, but good enough to become a hyper-aware wheel consumer. 

Something we included in yesterday's chart that we've previously omitted are the blue bars showing the amount of time you're likely to spend encountering any given wind angle. How we arrived at this distribution needs explaining.

It had been an established convention that 10* was THE heavyweight angle, likely because Zipp's collateral always placed such heavy importance on 10* in particular, and the 10* to 20* range in general. When they give a time savings figure, it is computed at 10* as the only angle. However, the world has long since moved past "because I said so" as an acceptable premise, and Trek and Flo have both gone to the trouble of doing actual data collection in real world situations, and come up with distributions that show that lower angles are actually vastly more prevalent. In fact 10* is shown to be largely irrelevant.

We've linked a few hours of quality technical reading there, but the abstract is this: at real world riding speeds and in real world conditions, this is what you see. The Flo data is very plainly presented in percentages, and you'll notice that if you add the percentages up they don't equal 100% - that's because 0 to 20* doesn't encompass every situation encountered. You have a bunch of small data points going out from >20*. The Trek data is harder to break down, but break it down we did. 

The frequencies we show are a straight average of those three data sets - Flo's gathered data, Trek's data from the Ironman AZ course, and Trek's data from the quite windy Ironman Hawaii course. Anyone can argue with our methodology on this, but we think that this is the most robust, relevant, defensible distribution available. And for what it's worth, Hawaii distributed a bit differently, being very exposed and windy, but the Flo and Arizona data matched very well, and Hawaii really wasn't that different.

We are not asking you to believe anything on faith. Every bit of what's been done and how it's been done are available to you here and in the links. As none of these rims is our own or available exclusively through us, we derive no benefit from the data leaning one way or another, or shading the data in any way. This entire exercise to to provide the wheel market with good information to become educated about wheel aerodyamics. 

More in subsequent posts, but I've run long as is, so that's it for now. 

PrintView Printer Friendly Version

EmailEmail Article to Friend

Reader Comments (24)

Great! - loving the results and the data; thanks for doing the tests!!!!

It just occurred to me (and maybe this is already being done somewhere?) that - if one is happy with accepting the ‘average’ wind angle frequencies provided - you can now use all your fine-grained data to produce one number to compare them all: the 'expected utility' of each wheel (disutility?). So, for each angle, get dragWatts x frequency_of_that_angle, then sum up all those numbers for a particular wheel.

Conceptually it's a weighted average of the drags. Looking at the graph, those far out angle account for out 20% of the time, and that's where the drags are really different, so that can compensate for some relatively poor performance at the more common angles. (Most of your curves are nearly laminar so I’m not sure this will produce any surprises, but would be a fun exercise!)

February 10, 2017 | Unregistered CommenterErik

Glad you're enjoying it.

On the one number thing, we can and we will, but you could actually do it for yourself if you don't want to wait a couple of days. All the info is right there.

February 10, 2017 | Unregistered Commenterdave

Oh, you'll see a benefit!! Your integrity alone deserves much more than a second look...

February 10, 2017 | Unregistered CommenterScott Booth

Dave, looks like the test went well and the data is definitely interesting. I'm surprised the 303 isn't further ahead at higher yaw but I guess it is only a 45 mm deep rim which isn't that much more than the Al33.

Erik - I replicated Flo's method for producing the one number that they call the 'Net Drag Reduction Value' in an excel spreadsheet using their yaw weightings. It needs a little more information than has been presented here so far because you need the actual drag numbers at each yaw sweep increment. I guess you could read the values off the plot but it would reduce the error to have the values used to generate the plot.

From that data it's pretty straightforward. You calculate the slope of the drag curve between each sweep increment and interpolate it to the weighted yaw bins (Flo gives their weights in 1 degree increments vs. 2.5 degrees for the tunnel sweeps). Then weight that interpolated drag by the percent time spent at the given yaw angle to get the drag contribution of that yaw angle, then sum all the yaw angles to get your total drag.

February 10, 2017 | Unregistered CommenterBen

Well, most wheel companies for obvious reasons will cherry pick their data to give them the best spin possible. That's marketing 101.

IMHO collecting wind data over a few actual Ironman courses is only just slightly less useless, as the data is still very regionally and even course dependant. A rider in Amarillo, TX is going to have vastly more crosswinds to deal with than someone in San Diego, CA. And us peeps in the great-plain states usually don't get the benefit of tree cover to shield ground level winds as east coasters do.

In short, the real answer on what practical yaw angles you'll experience riding is extremely complex in order to account for regional, seasonal, orographic and ground cover variables.... but it's really much easier to just drink the "deeper carbon is better" Kool-aid the big company wheel makers spew.

February 10, 2017 | Unregistered CommenterCOOSKULL

The only real honest to god answer to that concern is that no test is valid for everyone. Some guy on reddit is sticking his finger in my chest because no test is valid without it being specific to a particular frame/fork. The reasons the wind angle distributions we're using are so valid to me are that they a) show their work and b) cover a variety of conditions. Even in windy and exposed Kona, shallower angles prevail, which is also intuitive. Compare those to the old "take our word for it - 10* is where it all matters."

Anyone who makes any aerodynamics claim HAS to show evidence to support it, otherwise just please shut up, right? On the other hand if showing evidence requires a $1,000,000 ante, who's going to do it? You have to set parameters. If you ride where it's really really windy and you know it, you're probably on the wheels that won't put you into the hospital. If you live in Kitzbuhel (I just really wanted to say Kitzbuhel), you don't ride carbon wheels of any sort. If you're a tri racer somewhere on the Gulf Coast, go DEEP!

As many times as I say it, no one's really going to believe it, but this test was HONESTLY designed with the hope of showing that the Al33 had credible aerodynamics claims. That's why it was done. It was as transparent as it possibly could have been, very prominently including us pre-announcing what the wheel line up would be. As I said on another forum, we burned the lifeboats on that one. If the Al33 had sucked, that would have been shown.

So really all you can do is just be transparent and use a good methodology and empower and enable people to make good decisions for themselves. I know it's not credible for a person who sits in my chair to say that, but it's true. Mike and I started this company in order that we would be able to do jobs that allowed us to do that, if that's Cub Scout enough for you.

I think we did the best job of this that could be done, hopefully it provides value.

February 10, 2017 | Unregistered Commenterdave

For the record, my post wasn't meant to needle you or the Al33 rim in particular.

I am just skeptical that just because one has wind data from 2 Ironman courses that those results can be extrapolated to represent the majority of the riding population. But then again I haven't looked at the Kona data so maybe it does put a typical upper limit on yaw angles.

But yes, the buyer should be able to objectively evaluate what kind of wind riding conditions they typically face.


February 10, 2017 | Unregistered CommenterCOOSKULL

Oh I'm completely with you! Don't sweat and don't take mine as antagonistic to you.

Kona data has points out to like 60*, which anyone can go see (again, we love that anyone can go look at what we used and judge it for his/her own).

You're better off knowing something than nothing, right? And even better if you can see why you're being told something. As the contracted project managers of this test, we tried to solve for "what is going to answer the customer concerns and questions that will arise, as credibly as possible." That's the yaw angle for which we hoped to optimize.

February 10, 2017 | Unregistered Commenterdave


As usual, some really good information here. I was really surprised by the Flo30 data considering that they based their whole design off of the data they collected that showed riders spend the vast majority of their time at shallower wind angles. Can you confirm that it was the full carbon Flo30 and not the model with the aluminum brake track? I assumed it was the full carbon version but didn't see it stated definitively.

I discovered your website and blog about 9 months ago and poured through most of it. I remember reading your decision to discontinue the Rail34 was in large part due to the fact that there really wasn't a significant aero gain in shallower carbon rim designs over aluminum options at the time. Especially when you factor in most amateur riders don't spend much, if any time riding at the standard wind tunnel test of 30mph. The AI33 wasn't an option at that time and I don't recall if the Kinlin XR31T was a factor in the decision to discontinue the Rail34.

I think this test probably only confirms your decision to discontinue the Rail34. I guess my question is what kind of rider really benefits from the AI33 or XR31T? How fast does a group ride or crit race have to be before you would say that these two options are a worthy investment over a set of some of your other shallower, aluminum offerings? And at whatever point that is, would that type of rider be even better off with a set of deeper carbon rims similar to the Rail52?


February 11, 2017 | Unregistered CommenterMatt Weaver

Hi Matt,

Good questions.

First, the FLO30 is their alloy clincher wheel. It was not part of their redesign. It is still a current product, though. 45mm is the shallowest carbon wheel they do. The reason we chose the FLO30 is that it is a 30mm wheel advertised as having and popularly purchased for aerodynamic benefit. Had we not included it, it would have been fishy.

The Rail34 decision pre-dates our working with the XR31T by a bit. The XC279 (slightly shallower and narrower Kinlin rim) was one of the rims that initiated the Rail34 decision, and honestly the whole line of argument/inquiry that we've been on for 2.5 years now. We intuitively felt confident that XR31 would be an aero improvement over XC279 but stated it just that way - "we're confident that is is faster." We're very exact about what claims can be made and what can't, and we're proud of how we approach that especially in context of the standard bike industry hype.

Our general take has been that unless you're really chasing top level TT performances, wheel aerodynamics don't make a substantive difference to the average rider or racer. This test galvanized that philosophy and backed it with more and harder evidence.

The FSW3 wheel set is the least expensive set of wheels we sell - barring a custom build of same hubs and rims with round spokes, which is $40 less - so there's no investment to be made there versus any other choice. I mean, literally, it's 1/4 the price of much of what's billed at as aero wonder wheels, it's hand built by us, it asks you to make zero compromise in terms of spending money or being able to brake or whatever else. Our enthusiasm for it rose exponentially we were able to source them in a unique to us matte finish, so at this point we don't think it asks for an aesthetic compromise either.

The Al33 is the best looking alloy rim we've ever seen, maybe the best looking rim we've seen. You've got carbon that gives you the "I've got carbon" look and introduces brake compromises, for which you get to pay a lot extra. Any carbon within shouting distance of Al33's price is going to come with zero confirmation of aero performance, braking parameters, anything. So to me that decision looks like "hmm, for $715 I can get Al33 with very good hubs, great spokes, hand built. For the same dough I can get some random carbon off of an anonymous internet seller. For $1200-ish I can switch to carbon wheels from a company I know about, with similar hubs and spokes to the $715 build. And for $1600 to $3500 I can get expensive carbon." If you gave me that decision and $3500 and said "keep what do don't spend" I'd buy the Al33 without a second though. If you gave me $3500 and said "give me back what you don't spend" I'd still get the Al33 without a second thought.

Without poring through all of our other builds, the Al33 is a bit more than an Easton build. I love the Easton rims, would happily use them. The price difference isn't significant to me - get the ones you prefer. Same exact story with HED builds.

You your last question, Zipp's own collateral gives an 8 or so seconds difference in the mythical 40k TT to a 404 over a 303, and that's for a Firecrest 404 versus a pre-Firecrest 303. Given the compromises involved, I don't think anyone legitimately benefits from a "middle ground" carbon clincher.


February 11, 2017 | Unregistered Commenterdave

Dave, did you guys get any time to do 25c tire testing? I'd speculate that the uber wide rims (>24mm wide) might get a slight performance benefit with the wider tires.

February 12, 2017 | Unregistered CommenterCOOSKULL

The one number to rule them all*:
Zipp, Kinlin, Al33 have weighted average of 20.6, 20.7, and 20.7 watts of drag, respectively; i.e., a tie.
Flo and Hed have weighted average of 22.7 and 22.7 watts of drag, respectively; a tie, but worse than Zipp/Kinlin/Al33 category.

*I had to manually extract numbers from the graph, so possible my estimations are off in places, and/or I made a typo.

February 12, 2017 | Unregistered CommenterErik

My 'one number' was a LOTR reference, not meant to imply a one-size-fits-all here. As you've said, Dave, this depends obviously on the frequencies of time spend in different wind angles; the 'one number' was based on the distribution in your graph. It would be easy to plug in different numbers to see different outcomes! If two people agree on conditions though, this number is useful, but can only be derived after the careful testing that you guys did (thanks!).

February 12, 2017 | Unregistered CommenterErik

Erik - You're stepping on our lines! And yes you are exceedingly close to what we came up with.

COOSKULL - We got a good run with the 303 and a 25, and a run with the 25 on the Al33. The error margin on the Al33 run was high, and the run gave some crazy low drag values out at high yaw. Figuring that out with A2, which perhaps means a rerun if it's warranted, is one of tomorrow's priorities. Generally speaking, going to the wider tire coast you a couple of watts and causes an earlier stall - very closely following what we documented here: Also bear in mind that a "25mm" GP4000sII is over 28mm on every rim in this test - they're huge tires.

February 12, 2017 | Unregistered Commenterdave

Sorry! :) I was working on some graphs for work today and I couldn't resist. But, looking forward to your vetted numbers!

February 12, 2017 | Unregistered CommenterErik

Hello, this is really cool stuff. As a guy who doesn't want to spend $3500 on a set of wheels, it is interesting to know that a 404 firecrest may only save ~20sec or less over a TT assuming the time is spent at those shallower wind angles over one of these rims. That is a lot of money, and that is also assuming I am at 30mph, which few riders can really maintain for a long time. I assume at 25mph it is even less.

A couple questions-

It appears the xr31T has the lowest drag up until about 6 degrees or so, and then still holds its own against the others for quite a while. How do you like it verses the new AL33 rim? You haven't listed a price for rim only on the AL33(not that i could find), but the XR31T is quite affordable. I wasn't sure if looks aside, if you would just go with the 31T.

Were you able to compare these results to the test you did in 2014 with the pacenti SL23 and XC279? I have built with both of those rims and like them quite a bit. Just wondering if the benefits of the XR31T or AL33 would be enough to justify building a set of those. :)


February 13, 2017 | Unregistered CommenterBrian

Hi Brian -

The XR31T rims are about as nice as the XC279 fit and finish wise. Our XR31Ts have a matte finish which makes them about a million times more attractive to our eyes (and we're in the middle of merchandizing the components section, where they'll be available). There just plain aren't going to be enough Al33 rims in the first order to sell rims only, unless some more become available. We are a couple of sets away from sold out of the first run.

As to comparisons between the 2014 tests and this one, I mean, what I'd say is that we've established that these differences are in fact pretty small. Carbon imposes a LOT of cost and variables and use case liabilities for what now seems conclusively to be small or no benefit. Amazing what you find when you go looking. As to the differences between XC279 and SL23 and XR31 and Al33, taking some logical steps I'll say that this class of 2016 is better than the class of 2014, but not that you're going to notice a big difference in aerodynamics. A couple of watts? Maybe.

February 13, 2017 | Unregistered Commenterdave

Very cool. Thanks. I would be interested possibly in building with one of those AL33, but it sounds like in the meantime, an XR31T is pretty close. I look forward to seeing more info on these rims from you in the future!

February 14, 2017 | Unregistered CommenterBrian

Brian - The XR31T rims and a bunch of others are now up in the Parts and Pieces section of the store - Dave

February 15, 2017 | Unregistered Commenterdave

Hi Dave,

Thanks for sharing a great test !

I'm wondering if you ended up finding out what happened with the GP4000S II 700x25 test on AL33 and if there is a chance you will compare 303 to AL33 with that tire at some point (don't know if A2 did the test again or if it was just cancelled) ?

I'm impressed with the results you got on both Kinlin and AL33 rims relative to 303 benchmark and I'm curious to see if the difference between the wheels is maintained when going up one tire size.

Again, awesome work ! Really enjoy reading here !

March 1, 2017 | Unregistered CommenterPeter

PostPost a New Comment

Enter your information below to add a new comment.
Author Email (optional):
Author URL (optional):
Some HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>