A Day in the Life of a Wind Tunnel

We’re in the thick of grand tour season, where new models of road and TT bikes tend to find their ways onto the field of play. We saw that with a new, stealth Cannondale Slice in the Giro, and we saw it again last week, when certain riders on the Garmin Cervelo team were situated aboard the new Cervelo S5 road race bike.

I was the lone industry observer invited to the wind tunnel testing of the finished S5 product. This occurred over the past Spring, at the San Diego Low Speed Wind Tunnel (LSWT). I thought I’d describe a bit how this process occurs, and what I witnessed.

San Diego Low Speed Wind Tunnel
As many Slowtwitchers know, this wind tunnel was a part of the Convair facility, formally Consolidated Aircraft, which designed and built the B-24 Liberator (the heavy WWII bomber that still holds the distinction as the most mass produced airplane in U.S. military history). The wind tunnel was built in 1947, at the dawn of the age of jet airplanes. That said, this is a low speed tunnel, "low speed" being relative. We're talking about things tested up to 300mph.

Only twenty percent of the tunnel time at the San Diego LSWT is occupied by sports related activities. The rest of the time they test for aerospace customers, which for this tunnel include Cessna, Boeing, Gulfstream, Northrop Grumman, General Atomics, Raytheon, and Lockheed Martin. Aircraft programs using this tunnel for testing include the F-106, B-58, F-111, F-16, Global Hawk UAV, Tomahawk Cruise Missile, and Advanced Cruise Missile.

This is one of the reasons wind tunnel time is scarce and hard to come by for many in the bike industry. When a 3-to-5 week aerospace project needs to extend its testing, the 1-to-5 day bike project gets cut.

The people who test
This is a collaborative effort. You need people who know the tunnel, as well as people who know bikes. David Sanford is the LSWT Aerotest Engineering Manager. He runs things.

On Cervelo’s side is Dave Kennedy, a consultant to Cervelo who cut his chops at Buffalo Forge—the world’s largest industrial fan company, and a pioneer in the development of moving large masses of air. He was responsible for aerodynamic, mechanical, and sound design at Buffalo Forge’s large equipment test lab.

Dave worked with Steve Hed and John Cobb at the Texas A&M tunnel, on getting the tunnel to have better data resolution so they could make further refinements to their designs and test subjects. He worked on Project 96 with Doug Milliken, Jim Martin, Ed Burke and Chet Kyle.

In 1995 Dave met Phil White and Gerard Vroomen while they were volunteering to help on some of the wind tunnel refinements at A&M. Dave has been helping Cervélo ever since.

Damon Rinard is the senior Advanced R&D Engineer at Vroomen White Design; as well as Race Engineer for Garmin-Cervélo Pro Cycling Team. He got involved in a Human Powered Vehicle design competition and built a carbon bike in his garage (yes the one on Sheldon Brown’s website). He and Dave Kennedy share a background in HPV design. Damon moved to Wisconsin and and worked for Trek, involved in projects like the Bontrager tandem wheels, radial ridge rim project, Bontrager carbon fiber hubs, the TTX, 2008 Madone, and Levi Leipheimer’s aerobars. He brought computational fluid dynamics (CFD) to Trek and was involved in all of the wind tunnel tests Trek did in the last five years of his career there. Phil and Gerard recruited him from Trek.

Ivan Sidorovich is a CFD specialist for Vroomen White Design. Ivan has a Masters in Engineering and started his career working on shock and vibration, however, his love was CFD and aerodynamics. When Damon Rinard came on board at VWD, he suggested they bring a CFD capability in-house. Ivan’s responsibilities have allowed him to improve the CFD modeling that VWD has done to improve correlation of the wind tunnel with the theoretical computer models they use for simulation. Ivan’s work has allowed Cervélo to peel away the next layer of the onion, so to speak, and look for even more intricate areas that leave room for improvement.

Sweat the details
Okay, let's get this party started. But, we're not ready to test yet! Precision is king in the tunnel, so, there's some work to do before anybody or any bike gets test.

First, pump up the tires, on all the wheels, to 120psi in Cervelo’s case. Then mount the wheels to the symmetrical bike.

What is this? It’s a special bike that Cervélo built to check the accuracy of the tunnel left to right. It's pictured above. Cervélo found this was a great tool, and donated it to the San Diego LSWT. It is used by the tunnel to set up the wind tunnel before all manufacturers using the tunnel install their test bikes. Most bikes are pretty symmetrical, but this one is 100 percent symmetrical: no derailleur hangers on the right, or divots for chain clearance etc. All non-symmetrical features have all been removed. The symmetrical bike is mounted on the wind tunnel’s balance to check tunnel alignment left-to-right.

Cervelo also brings a “calibration bike” to the tunnel. Cervélo uses a first generation P3C with a “calibrated’ frame tube over the fork steer-tube extension. Cervélo has found that to get consistent repeatability, they remove the handlebars and place a tube approximately 10” in length into the air above the headset top cap to simulate the drag that would be produced with handlebars in place. This vertical tube also allows them to cancel the different length of various steer tubes when comparing forks. This bike has actually been around longer than the symmetrical bike, and it’s handled with extreme care. Using this same bike at every tunnel session makes it possible to compare new data to old data produced years ago. Old data is still good data and worth learning from.



The calibration bike tests the tunnel’s calibration. If the numbers are the same as the last calibration bike test run, and also in line with historical wint tunnel data values for this bike, the tunnel is ready. If not, something needs to be examined and corrected before further testing can commence.

Next, a run is performed with nothing on the balance other than the struts, so the drag of the struts (thin vertical airfoil shaped plates that hold the bike in place) can be subtracted out using simple math. Tunnel engineers will tare the balance accordingly.

Now, and only now, are you ready to test. The typical testing day is 10 hours. The very last test is a re-run of the calibration bike, to prove that the wind tunnel is providing the same data as it was at the beginning of the day.

The Balance
Let’s talk about this for a moment. It’s the part of the tunnel that measures the drag on the bike. Think of it as a really expensive and really accurate scale capable of measuring the force applied to the model in 6 degrees of freedom. The balance used at the San Diego LSWT is very special. It is accurate to +/- 5 grams. The balance can hold the model in the middle (vertically) of the test section, and remove the boundary layer using a splitter plate. The computer collects data from the balance at a rate of 10 samples per second on all 6 degrees of freedom.

Six degrees of freedom (DOF) refers to the motion of a rigid body in three dimensional space, namely the ability to move forward/backward, up/down, left/right combined with rotation about three perpendicular axes (pitch, yaw, roll). As the movement along each of the three axes is independent of each other and independent of the rotation about any of these axes, the motion indeed has six degrees of freedom.

More power to him!
The wind velocity used by Cervélo is usually 30mph. According to Cervelo’s engineers, this speed gives them the best combination of accuracy to real life conditions and the resolution to see differences in drag with different variables on the model. Flow characteristics are consistent at speeds above and below 30mph, so drag data taken at 30mph is easily and accurately scalable to answer questions about racing at slower and faster speeds.

Thirty miles per hour has become the de-facto industry standard for wind tunnel test speed, a standard established by Steve Hed and John Cobb in their voluminous testing more than two decades ago. Cobb was testing bikes at 30mph in the tunnel as far back as 1985.

The operator in the control room increases the speed of the prop manually until the instrumentation in the test section—the part of the tunnel where the balance is located—is measuring the appropriate speed for the test. So the wind speed at the test is measured, not driven.

After every run the tunnel is stopped, and the drag data recorded at 0 wind velocity. If the data shows the same numbers as at the beginning of the sweep at 0, everything is okay. No one touches the balance or the model until the 0 has been checked or else the data from the sweep would be wasted!

Bike set-up
All bikes tested should be set up with the same fit coordinates. All bikes tested should have chains in same position—in the case with Cervelo's test, the 53t ring and second cog; and housings installed with solid internal wire to keep stiff for repeatable data. All components on bikes need to be identical: same derailleurs, wheels, tires, tire pressure, same cranks (except where a proprietary crank is required), chain position on cogs, crank orientation, same saddle, with the saddle level.

The same beta sweep is run for all tests, with the same mannequin—DZ himself and his position on the bike is 100% repeatable. The same wheels are used, with same tires in same orientation; same tires and tire orientation. Stock bars are used, or the fastest ones that fit as necessary to match position. Bikes are tested with and without bottles (Note: same bottle for all tests).

Beta Sweep
The drag is taken at zero yaw, that is, with the bike facing into a direct headwind. Then, a turret on which the balance sits is turned according to a “beta schedule”. This is the series of angles the model will be rotated to during a run for collecting data. Cervélo was using a self-defined protocol named “Beta Schedule 5” that tested angles of 20, 15, 12.5, 10, 7.5, 5, 0 -5, -7.5 -10, -12.5, -15, -20, 0. According to Cervélo, in areas where the data shows the most variation more yaw angles are tested, whereas in areas where data is—from experience—less variant, a larger gap between angles is sufficient.

During each step in the beta sweep the tunnel operator will wait until the data is stable and then hit a “capture” button. The data is captured at a rate of 15 samples per second for 10 seconds and when it’s complete the balance automatically moves to the next beta angle on the beta schedule. At this point the tunnel operator may need to adjust the fan motor power output to maintain a consistent wind velocity in the test section as the “apparent” size of the model may block the air, and increase the wind speed. Once wind speed is normal, and the drag numbers stabilize, the operator hits the capture button. The beta schedule that Cervélo was using at this test took approximately 20 minutes per run.

Photo documentation
Pictures are taken of the setup every time, and what is unique about the current run’s setup. There is a “run number board” that has the test session, and what test number of the day on a placard that is placed in the photos. The tunnel crew—Dave Sanford in this case—takes photo documentation, as does Cervélo’s crew, and any other photographers in the audience—me, for example.

Are all tunnels created equal?
In a word, no. What makes one tunnel superior to others? Things like the professionalism and experience of the staff. And the accuracy of the balance. In the case of the tunnel in San Diego, as noted, +/- 5 grams in all 6 degrees of freedom. This for dynamic objects. For static objects, better accuracy yet.

The construction of the tunnel itself is important. This tunnel features low turbulence, and this cuts down the variance of the data. A big “test section” cuts down on wall interference, and a very large contraction ratio—the size of the tunnel from the fan to the test section—of 6.5 to 1 stablizes air and produces even air flow.

In addition to standard instrumentation, this tunnel also has a special anemometer, pitot tube, and barometer that measures dynamic pressure. A custom designed splitter plate goes completely across the tunnel creating a minimum boundary layer condition—verified by tuft testing. Cervélo aided in the design and construction of the splitter plate.

This tunnel—unlike many others—is exceptionally quiet. Acoustical vibration is therefore not an issue here, and thus does not affect measurements.

Personal impressions
I own a bike shop, Moment Cyclesport, a mile away from the San Diego Wind Tunnel. Before my foray into the bike industry, I put my mechanical engineering background to work in the high performance vehicle design business, for Shelby American, and Ford GT, with a stint in between working for SPARTA Inc., a local defense contractor. So, wind tunnels and high performance modeling is not new to me.

How would I rate Cervelo’s approach to wind tunnel testing against what I’ve experienced in the past? This company has a genuine interest in getting the best data and most “fair” comparison from these tests. Good data can be used to make better bikes. Bad data is worthless. Gaming the data is very difficult and very expensive.

I asked the Cervélo engineers, "Why not release the entire data set if you’re telling the world how fair you are?" Their answer: This data, by design, doesn’t answer every question—the scope of the testing is narrow; confined to how bikes compare one to another when set up in what is considerred the "best" position—nevertheless the public will likely try to use the data for everything. Also, to be sure, Cervélo doesn’t want to make it too easy for the competition to catch up.

That said, integrity is part of the company. The data is the data. Sometimes things go well and sometimes they go poorly—it’s a learning experience for all, including the tunnel and the company renting time in it.

Therefore, it's rare-to-nonexistent for a company like Cervélo to give me free reign over their data. I had access to everything—all the data—and could ask for any data if I thought there was something being misrepresented or hidden from me.

I enjoyed this access for two very long testing days. I did my best to describe the mechanics of what this particular test looked like on these particular days. It may be illustrative to compare protocols and processes with other tunnels and other bike companies.

I'd like to extend my thanks to Cervélo for drawing back the curtain and giving Slowtwitch.com a chance to see their process and write about it uncensored, unembargoed, and unvarnished.