Article

The Art and Science of Home-based Small Wind Tunnel Testing

August 24, 2025

Velocity vectors computed from smoke flowing over a model car in a small wind tunnel
Computing velocity vectors

There is something hypnotic about smoke curling around a car the size of your palm. I set up a small smoke wind tunnel at home, ran it over a die-cast model with its wheels spinning, and watched textbook aerodynamics unfold on the kitchen table. Then I turned the footage into numbers: optical flow to extract the velocity vectors, and SINDy to find the equations behind them.

To be clear from the outset, this is a demonstration rather than a laboratory experiment. There was no camera stabilisation, no masking and no controlled environment. But the physics on display is striking, and the full video is below.

The kit

The setup is modest. A small smoke-visualisation wind tunnel from Fun-Tech-Lab provides the airflow; a 1:64 scale car with rotating wheels is the test subject; and a Sony A7III, focused by hand, records the result. The footage is graded in DaVinci Resolve, the same software used to edit Oppenheimer and Top Gun: Maverick. A ruler in the frame provides the calibration from pixels to physical units, and the optical flow and SINDy analysis run in Turbulence Realm.

Small tunnels like this one earn their place despite their imperfections. They are quick to set up, cheap to fail with and immediately visual: smoke shows separation, reattachment, shear layers and wakes directly, with no post-processing at all. That makes them superb for building physical intuition. What they do not give you are absolute numbers; their value is qualitative.

Exploring the fundamentals of aerodynamics no longer requires a multi-million-dollar facility.

From smoke to equations

The process moves in five steps. First, the car goes into the tunnel and the smoke goes on, the spinning wheels adding a touch of realism. Second, the camera captures the patterns as they develop around the body; the A7III’s low-light performance helps, though serious research would call for a high-speed monochrome camera and matching lenses.

The wind-tunnel footage being colour graded in DaVinci Resolve
DaVinci Resolve during editing

Third, the footage is colour graded to sharpen the contrast between smoke and background. Fourth, optical flow extracts the velocity field, the direction and strength of the air’s movement at every point around the car. Fifth, SINDy treats that field as a dynamical system, identifies the equations that govern it, and uses them to predict how the air will behave. A practical tip: start with about ten seconds of footage (300 frames at 30 frames per second), depending on how much memory you have.

Comparison of measured and SINDy-predicted flow around the model car
Actual vs. SINDy-predicted flow
Velocity magnitude predicted by the SINDy model
SINDy-predicted velocity magnitude

What the smoke revealed

The smoke behaved as theory says it should. Vortices formed where they were expected around the body, and the rotating wheels left their own distinct signature on the surrounding flow. SINDy’s predictions matched what the eye could see, with errors, as any model trained on a hand-held demonstration will have. And the experience underlined how much a small-scale test can say about the behaviour of a full-size vehicle.

You do not need everything I used to try this yourself. A basic tunnel and a decent camera are enough to start, and the physics is fascinating at any level.

Final thoughts

Experiments like this are a reminder of how far the gap between professional research and the curious amateur has narrowed. Whether you are a seasoned engineer or simply someone who likes to tinker, there has never been a better time to go looking for the invisible forces that shape our world.