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