The NASA Ames low speed open wind tunnel is a cavern: A vast 120 foot wide yawning gulf, its ceiling studded with lights, it makes humans look puny & insectile. In its centre, a rotating disc holds full size aircraft or trucks. Shadow swallows the rear as the cylcopian duct disappears into a bank of colossal fans, relentlessly grasping. Up front, a huge grid shapes and straightens the air as it builds past gale force to 115 miles per hour. Supplying this colossal energy takes 104 megawatts, a small city’s worth of power.  It’s the largest wind tunnel in the world.

There are lots of wind tunnels in the world, from giants like this and the S1MA transonic wind tunnel in France (powered by its own hydroelectric reservoir) to the frigid ETW tunnel in Germany that uses compressed cryogenic nitrogen, and even little lab-scale tunnels in universities scattered across the globe. They range from low speed tunnels for modelling cars and city centres, to hypersonic tunnels that can recreate the aerodynamics of capsules re-entering the Earth’s atmosphere. They can use air, nitrogen, water or even oil. And their uses branch across industry, academia and sport…

There are lots of handicaps in sport, designed to make things a bit fairer & more exciting, but one of the weirder ones is in Formula 1, which punishes winning teams by… restricting wind tunnel time. But what is a wind tunnel and why do we benefit so much from its breezy embrace?

On a much more humble scale, it’s been many years since I personally last used a tunnel, and I remember the thrill of it, because it’s a little like discovering a truth in the universe, which will be yielded to you with a stiff breeze and lots of drama. As the fans rev up, the air is pulled through an austere grid and into the test chamber, and you realise that you’re about to discover something new. Perhaps whatever you’ve got in there will perform as you expected… and perhaps it won’t. The hum of motors and the stabilization of brutal winds behind a friendly little window are about to let you know, and they don’t care about your feelings. No pressure.

There’s also a weird sense of foreboding about the things: I can’t actually remember any movies that depicted a wind tunnel death, but it has a terrible potential about it, with the dark tunnel disappearing into gloom, the fans and the terrible suction that builds and builds. The only actual deaths that I know of are in skydiving tunnels, where free-fallers can practice their moves, but never in a research tunnel, despite the ominous look of that terrible, grasping maw. 

These things are used by serious professionals after all.

Wind tunnels are used in all sorts of industries, and while the advent of high speed computational simulation has chipped away at their niche, they remain critical for many, many things. You probably wouldn’t volunteer to test fly a plane that had only been simulated on a computer, after all. But how do they work, and what subtleties guide their use?

Let’s turn up the fans…

A wind tunnel can be continuous flow, or it can work from a pressurised reservoir. The former uses fan arrays and a series of convergent and divergent ducts, along with flow-straightening grids, to create the flow environment. The latter, positive reservoir based tunnels, are often used to model high supersonic or hypersonic environments, where a high-pressure reservoir is quickly emptied & accelerated through the test section.

Most wind tunnels are fan-driven, and use a fan array to “pull” air through the test volume. Such continuous flow tunnels are separated into ‘open’ and ‘closed’ flow: Basically, an open flow pulls from the atmosphere, gives the air a single run through the test volume and spits it back out into the atmosphere again at the end. The big 120 foot wide NASA Ames tunnel is one of these, and they’re usually aimed at low speed, subsonic aerodynamics. A closed section tunnel by contrast is, well you can probably guess; it’s a closed loop. The fan pulls the air into a circuit, where vanes redirect it and prevent odd velocity profiles around the corners. A closed section is useful for liquids, oils, and in case you don’t have a convenient path to the outside world and don’t fancy knocking down a wall. They’re also useful if you want to study transonic flows, as being closed-off to the world means that you can play a number of tricks with pressure and temperature, which we’ll explore later.

In all cases, the tunnel needs a number of key features: Convergent ducts (which get narrower in cross-section as they progress), diffusers (generally the opposite), and flow straightening grids. The reason for all of this is that we can’t just shove any old air into the test section.

No, it needs to be good air! Clean, smooth flow, unblemished by turbulence, hitting the test piece with all the blissful naïveté of an infant. 

So how do we create this perfect airflow? The key is first to understand the difference between a laminar flow and a turbulent one. A laminar airflow has perfectly smooth flow lines and behaves as most of us imagine all air behaves. The still air we walk through is a laminar flow when it hits us; it hasn’t yet garnered enough energy to be choppy and turbulent, and this is what we want hitting our test piece. 

By contrast, turbulent air is flow where the kinetic shear effects in the flow have overpowered the viscous stickiness that’s trying to hold it together, and so the flow deteriorates into a cluster of chaotic eddies, splitting, breeding and propagating fractally almost to the quantum scale. Whether a flow will tend towards laminar or turbulent is dependant on our first scaling factor, called the Reynolds Number, which we will revisit a bit later. 

Now if you run air next to a long enough surface at a high enough speed, you are guaranteed to get turbulence, which is not what we want if we want our test chamber to recreate something moving through clean, everyday still air. And since our wind tunnel, particularly in a closed design, is just an endless progression of surfaces and slip planes, that means that we need to laminarize our flow before it hits whatever it is that we’re testing inside the tunnel.

Laminarizing a flow requires two things: A favourable pressure gradient, which can be provided by an accelerating duct, and a fine grid, which stops turbulence physically by shoving the air into a series of tubes. Put these things together and your test section can run on lovely, clean laminar flow, just as mother used to make…

…Well, as long as you don’t put your test section too close to the tunnel walls anyway. Remember that they can generate turbulence too. Oh well, nothing’s perfect, and sometimes we just have to be good enough. Onwards we go.

One of the primary advantages of wind tunnels is that they bridge a gap between cheap iterative scalability (computational fluid dynamics simulations) and real flight tests. Flight tests, after all, are dangerous, slow and very, very expensive. Computational Fluid Dynamics (CFD), is a monster that has grown to devour much of conventional fluid dynamics, but it’s not omnipotent: There remain many things that the procrastination of computers just can’t do well: Acoustics, transonic shock behaviour, interaction with roads or water and so-on. Even manufacturing intolerances.

Remember: Small things like rivet heads and manufactured panel gaps have real aerodynamic effects, and it’s hard to replicate this on a computer. Moreover there are specific flow scenarios that are difficult to model computationally even with a perfect model of the geometry. 

For example, maybe what you’re interested in is the breakdown of flow over the surface of an aerofoil, such as on a fan working at extremes, or the bluff backside of a racing car, or a fighter plane under extreme manoeuvre. Likewise there are more complex interactions, such as aeroelasticity (when structural vibration and aerodynamic interactions feed off each other to brutal effect), or unsteady compressive boundary layer transitions (where a boundary layer transitions to turbulence in a supersonic flow, producing a shock wave which interacts with the boundary layer which interacts with the shock wave again…). Well, you get the point: When reality gets complex, computers don’t always keep up. 

And then you might need some tunnel time!

When you use a wind tunnel, you need to get some data out of it, and a variety of methods exist to pull meaning from air: You can stud your wind tunnel model with pitot tubes, little pencil lead-thin metal tubes that measure static and dynamic pressure… or you can mount the model on a ‘sting’, with strain gauge sensors to give you information about the force the airflow is putting on it, or you can fire criss-crossing phased lasers into the air around it and use Laser Doppler Anemometry to measure particle speed in the tunnel, or, or, or…

You can even use fluorescent paint. 

Yup. Paint your little model plane, rocket or F1 car with a fluorescent paint that quenches in the presence of oxygen and you can illuminate it with UV light and use a camera to measure pressure variation from glow-in-the-dark paint. 

Or add a road: For racecar aerodynamics, it’s important to replicate the fact that they’re going to be on a road with wheels spinning, so the model needs to be mounted on a rolling platform in the wind tunnel, moving at the same speed as the free air flow. Stings or pressure sensors can then measure the force transmitted through the wheel in a variety of scenarios, to give a measure of downforce and grip. 

In all cases, wind tunnels are a crucial intermediate step in-between computer aided design and actual physical testing where a real person is at risk. Sometimes it’s just easier, cheaper and safer to use a small wind tunnel test instead of a supercomputer cluster or shoving some poor sod into an experimental race car. 

Which begs the question: Can we simply do all of our wind tunnel testing on nice, cheap little tunnels?

The gargantuan NASA Ames tunnel is one end of an extreme, but you don’t necessarily need to use a tunnel this big. 

In all cases, the important thing is that, whatever you’re testing, it needs to replicate reality, or else what’s the point? Replicating reality in miniature is harder than you’d expect, and to solve it we need to understand something called scaling parameters. 

Before we get there, we can consider lift or drag: The equation for lift over a lifting body is shown below, and you can see that for a given geometry it depends on both the wing area, density, air velocity squared and the lift coefficient of that shape. Now, if we miniaturize this shape, all of the other factors in that equation can be accounted for, just so long as we keep the lift coefficient constant… and that depends on the overall flow characteristics being kept the same. To do this we need to match scaling parameters.

The first of these is the Reynolds number, which we briefly mentioned earlier. This governs the transition of the flow from laminar to turbulent and where this will occur. The Reynolds number of a flow has a colossal effect on macro-scale flow characteristics, which you can see if you compare the flow over a bluff bodies at various values of Re. 

Obviously, unless your Re is really, really big (or vanishingly small), then miniaturization needs to match this parameter in order for you to be able to take realistic data from a wind tunnel experiment. But how do you do this?

The Reynolds number depends on the characteristic length L of your testpiece, which has now shrunk, so we need to make it up some other way, typically by either increasing flow velocity, increasing density or reducing viscosity. The last two are tricky to manage, but if you can manipulate the temperature in your wind tunnel then you have a means to do this. If you get really stuck you could even supplement air for water, oil or some other fluid, and just so long as you match up the Reynolds number you’ll get flow behaviour that looks the same, even if you’re running in water instead of air. Strange but true. 

Our second scaling parameter is a little more intuitive: It’s the Mach number, which is basically how fast you’re going relative to the local speed of sound. This is completely unimportant if you’re well below Mach 1, but as you get closer to Mach 1 (the speed of sound) it becomes a bigger and bigger influence on how the flow over your jet fighter, engine, rocket or land speed record car behaves. Once you’re past the speed of sound you’re into a regime called ‘compressible flow’, where you’re moving so fast that air can’t get out of the way in the normal fashion and so forms vicious shockwaves, with inevitably huge effects on behaviour.

You can even be in the transonic flight regime, which is right on the cusp of Mach 1, and this is trickier still for wind tunnel testing, for reasons we’ll get into later. Solving it may demand either Going Big or Going Ice-Cold with a cryogenic tunnel. 

There are also a couple more scaling parameters we haven’t mentioned yet, but they’re mainly a factor for ships and marine gear, which we’ll cover in the water tunnels and tow tank section.

Speaking of which… what about the freaks?

Not all wind tunnels are tubes full of air.

For example, you might want to test a sailboat, which is a little bit complicated: A sailboat typically moves at an angle to the wind, not directly in-line with it, which would be OK if the wind gave the same airspeed at all altitudes but, troublesome zephyr that it is, it doesn’t.

Near the sea, the same frictional influence that causes us to worry about Reynolds number and turbulence up above causes the wind to slow down when near a boundary. Higher up, it moves faster. This means that for a sailboat, which not only sails at an angle to the air but also has to cope with this boundary layer effect over the length of its sails, the sail sees ‘twisty’ air. 

Yep. From the point of view of the sea the wind is just the wind, no twisting there. From the point of view of the boat sailing at an angle however, the wind twists as it nears the base of the sail, pulling it in subtly different directions. Accommodating this in a wind tunnel, since you can’t make the ship move, means physically twisting the airflow (the ‘wind’ in the sails) with purpose-build vanes. Wrap your head around that one!

Then there’s racecar wind tunnels, which need to simulate a rolling road at the exact speed of the air, and measure the load being applied through each tyre as well, and even cornering loads, which takes some thinking about. 

But to get to the really… odd tunnels, you need to be dealing with transonic flow, which is a particularly cool thing to deal with.

No, I said that’s wrong: It’s a particularly cold thing to deal with. Ice cold.

In Cologne, Germany, lives a strange beast, and Nitrogen runs in its veins. It's called the ETW, or European Transonic Wind tunnel, and it exists to bridge a very frustrating experimental gap in aerospace design development.

Modern airliners are big beasts, no question. The biggest, such as the Jurassic monstrosity that is the Airbus A380 ‘super jumbo’, can mass over five hundred tons at takeoff and cut the air with a wing eighty metres from tip to tip, which isn't far from the length of a standard football pitch. Such beasts are sold on extremely thin fuel efficiency margins, where tiny performance increments are hugely important, and they cruise at a transonic Mach number range of between 0.78 and 0.85, just a hair under the speed of sound.

This is all highly inconvenient!

It's inconvenient because the experimental aerodynamicist that needs to make sense of this tangle of requirements needs to reconcile two scaling parameters that don't play well together; Reynolds number and Mach, and both of them are important for an airliner at cruise conditions. Reynolds defines the point of turbulent transition on the wing (and it will transition! Airliners are too big and fast for it not to), and Mach defines whether a weak shockwave will form around the wing's quarter chord. These two factors each play off each other like children on a seesaw and have a huge effect on the behaviour of the wing and fuel burn lost to drag.

There are two ways out of this: You could opt to Go Big! and use something like the S1MA, a giant French transonic wind tunnel powered by its own hydropower reservoir and capable of hitting Mach 1 in a test section big enough to take models eight metres wide. 

That's a lot of power right there! But it still leaves you lacking in Reynolds number, even though it gets you to within an order of magnitude. What if you want to get everything exactly flight-perfect?

You could use the ETW. This borrows from a technique pioneered by NASA in the 70s and eschews normal air for pure nitrogen, which is pressurised and cryogenically chilled. This is done to both increase density and reduce viscosity, which hugely increases the Reynolds number and lets you model turbulence effectively on small scale models. Mach number also increases a little with the plunging temperature, but not as much so it's possible to get both scaling parameters to meet in the middle at flight-equivalent conditions, even for something as challenging as a scaled model of a big airliner. The ETW does this by allowing the pure, recirculated nitrogen to be chilled down to as low as -160 degrees Celsius, with internal pressures than can be ramped up to 4.5 times sea level atmosphere.

That's an awful lot of trouble to go to, just to get the aerodynamics right, but it's worth doing if you're designing a new passenger plane.

It might also be easier than going for a gigantic conventional tunnel powered by a hydroelectric dam, to be fair, although Onera's gargantuan S1MA has its own charms. For example, a big tunnel is better if what you want is to test structural deflection, or acoustics.

As ever, it's all down to what you're trying to accomplish. Horses for courses and all that.

Okay, yes I know. “Jordan you dope, there's no wind in these. It's water, d'oi!” And you'd be right, but these niche installations tell us a few extra things about scaling parameters and so are a fitting last stop on our Tour O'Tunnels.

For one, while some small props or rudders might be Reynolds sensitive, the fact that they're in water tends to make most nautical applications experience Reynolds numbers so cartoonishly huge as to be almost ignorable, even with aggressive model scaling.

By contrast, something called cavitation can become incredibly important. Cavitation is the pressure-induced boiling of water when exposed to a sufficiently low static pressure, such as over the surface of a propellor. This creates a cloud of tiny steam bubbles, popping into existence for the briefest of instances before violently imploding. On water based applications, this can be a Very Big Deal.

So as you miniaturize, you might need to scale to the cavitation index. OK, fine. No problem, right? After all, you don't have to worry about Reynolds or Mach (ha! Just try to get near the speed of sound in water), so no biggie.

Well… almost.

Sometimes you need to test a powered ship model, and you might do that in a tow tank, half air and half water. In this, the behaviour of the ship (or rig, or platform or whatever) is defined by its wake, and the influence of waves, which in turn are governed by gravity.

Which, of course, you can't change, making it very difficult to match both this and the cavitation index to real-world values.

One of the very few places on Earth to try to square that circle is MARIN, in the Netherlands, that runs a 260 metre long tow tank in a sealed environment that can be depressurized down to 2.5% of normal atmospheric pressure. This allows the cavitation index to be scaled to the mini-size models and propellors you need to develop new ships, as it reduces the pressure drop needed before the water boils and cavitates. That's a lot of trouble to go to in order to get a ship right, but as we've seen, there are wind tunnels in this world that will twist a flow for a sailboat, or embrace the savagery of sub-arctic temperatures to get everything just right. It's as if Goldilocks ran a physical research institute.

The things people will do to make an experiment work.

What is the wind tunnel but a connection between two islands surrounded by sea, metaphorically at least: One is the land of simulation and iterative design, the engineer's playground where the trees are stick-&-cloud abstractions and the faceted ground is perfectly uniform. Everything is perfect on Simulation Island, but nothing stays the same.

The other island is reality. Here the trees are creaking, twisting fractal monstrosities and the ground is littered with the twisted wreckage of broken dreams. It is here that engineers come to be humbled.

The wind tunnel connects the two islands. To enter is to commit yourself to learning what your biases are, and how they are wrong. You may emerge, in red-faced enlightenment, or you may remain lost forever under the sea of Maybe. 

We can create wonders in the digital world. You might be creating one now, but be wary.

Eventually, we all need to raise our heads and look sternly into the breeze.

Turn up the fans…