Fan Array Wind Generators vs Closed-Circuit Wind Tunnels: Choosing a UAV Test Facility
Fan array wind generators and closed-circuit wind tunnels solve different problems. Compare flow quality, free flight, cost and control for UAV testing.
A drone programme reaches a point where somebody asks for a wind test facility, and two very different pieces of hardware get proposed under the same budget line: a fan array wind generator — a modular wall of individually controlled fans — or a conventional closed-circuit wind tunnel.
They are not competing versions of the same machine. They answer different questions. A fan array answers "does the aircraft cope?". A tunnel answers "what are the numbers, to what accuracy?".
Below is an honest comparison, including the cases where we would tell a client not to buy a tunnel.
1. How a fan array actually works
A fan array is built from modules roughly 25 × 25 cm, each containing a small grid of "wind pixels" — commonly pairs of counter-rotating fans that can be commanded individually. Modules stack into a wall of essentially arbitrary size, and a hierarchical controller drives them per module or per pixel.
That architecture buys something a classic tunnel cannot offer: the flow field can be shaped in space and time. A gradient across the outlet, a rotating vortex, a step change in direction, a programmed shear layer — all become software.
Published figures give a sense of the class. Each wind pixel — a pair of counter-rotating fans — is rated to push flow up to 16 m/s, and a module packs nine of them into roughly 25 × 25 cm; wall size and total power scale with however many modules the outlet needs. Commercial wind walls are quoted up to 20 m/s, or up to about 58 m/s when a convergent is fitted. Turbulence intensity sits around 5%, and can be brought below 1% with an optional flow filter.
Those four numbers — area, speed, power, turbulence — are the whole comparison in miniature.
2. What a fan array is genuinely better at
- Free flight. The aircraft flies with its own autopilot in the loop, unsupported, in front of the wall. No sting, no balance, no support interference — and you are testing the system, which is where most urban failures live.
- Programmable disturbance. Reproducing a specific event — a building wake, a rotor-downwash crossing, a sudden 40° direction change — and repeating it identically for two firmware builds is exactly what the array was designed for.
- Scalability. Need a bigger flow field? Add modules. A tunnel's test section is fixed in concrete on day one.
- Access and turnaround. Swapping payloads, propellers or battery packs between runs takes minutes, and the vehicle is reachable from every side.
- Modest civil works. It is a machine in a hall, not a building wrapped around a duct.
For control-law development, autopilot tuning, gust-response demonstrations and customer-facing trials, this is often the correct purchase.
3. Where a fan array stops
The limits are physical, not commercial, and honest suppliers state them.
- Flow quality. Around 5% turbulence intensity is fine for robustness testing and useless for a clean drag polar. Filters help, at the cost of speed and power. Certification-grade aerodynamic coefficients are not produced in that environment.
- Open jet behaviour. The flow leaves the wall and immediately begins entraining room air, spreading and decaying. The usable, uniform core is significantly smaller than the outlet area, and it shrinks with distance.
- Speed ceiling per unit area. Power scales with the cube of velocity. Getting a large outlet to high speed is where the array's electrical bill overtakes a tunnel's — an open jet accelerates room air from standstill every second it runs.
- No environment control. Temperature, humidity and air density are whatever the hall is doing. Repeating a measurement in July that you took in January is not straightforward.
- Forces are inferred, not measured. Without a balance, aerodynamic coefficients come from onboard telemetry and models rather than direct force measurement.
One case makes the speed ceiling concrete rather than abstract:
[Article not found: high-speed-fpv-interceptor-drone-testing]now fly at 300–400 km/h, a segment growing fast enough that we get asked about it directly. No fan array on the market reaches that speed — the commercial ceiling sits around 58 m/s (roughly 209 km/h) even with a convergent fitted, a little over half of what an interceptor needs. For this speed class a closed-circuit tunnel is not the premium option; it is the only one that physically works.
4. What a closed-circuit tunnel does differently
A closed loop recirculates the same air through corner turning vanes, a settling chamber with honeycomb and screens, and a contraction that both accelerates the flow and cuts turbulence. That geometry buys three things:
- Flow quality by construction. Low-turbulence conditions and a uniform velocity profile across the section — the baseline required for repeatable coefficients and for studying a disturbance you deliberately introduce.
- Environment control. Because the air is captive, temperature can be held. That is what makes summer and winter data comparable, and it is a precondition for climatic and icing modules.
- Running cost. The fan reuses kinetic energy already in the loop instead of accelerating still air. At the same size and speed that is worth roughly 60–70% of the electricity bill versus an open-circuit arrangement — the arithmetic is in
.
The price is rigidity: the section size is decided at design time, the model is usually supported, blockage limits apply, and the building is part of the machine.
5. Task-to-facility mapping
| What you need to do | Fan array | Closed-circuit tunnel |
|---|---|---|
| Tune an autopilot against gusts | Yes — free flight, programmable disturbance | Possible, but constrained |
| Demonstrate robustness to a customer | Yes — visual, fast turnaround | Less compelling |
| Reproduce a specific building wake | Yes — spatial and temporal control | Only with model-scale urban simulation |
| Measure drag polar and rotor efficiency | No | Yes — balance, low turbulence |
| Produce certification-grade coefficients | No | Yes |
| Test icing, cold soak, hot-and-high | No | Yes, with a climatic module |
| Acoustic measurements | Hard — array noise dominates | Yes, in an anechoic or treated section |
| Scale to a larger airframe later | Yes — add modules | Only if designed in from the start |
| Run 8 hours a day for years | Costly at speed | Yes — this is the efficient case |
6. The hybrid answer, and how to phase it
Most serious programmes end up with both, and the sequencing usually matters more than the choice.
A pragmatic path: start with an array while the control laws are immature and the airframe is still changing — that is when you need many fast, cheap, forgiving iterations. Add a tunnel when the questions turn into numbers: efficiency, endurance, rotor interaction, noise, certification evidence, environmental envelope.
The mistake is buying the tunnel first, at a size fixed to an airframe that will not exist in eighteen months.
There is also a middle path we build regularly: a closed-loop tunnel with an open-jet test section, which recovers much of the access and free-flight character of an array while keeping the loop's flow quality and thermal control. It costs more than a wind wall and less than the wrong facility.
7. Five mistakes we see
- Comparing peak speeds. 20 m/s from a wall and 20 m/s in a tunnel are not the same 20 m/s. Compare the uniform core, the turbulence intensity and the temperature stability.
- Ignoring the hall. An open jet interacts with the room. In a small hall, recirculation quietly corrupts every run.
- Forgetting blockage. In a tunnel, an oversized model invalidates the data long before it looks tight.
- Assuming acoustics comes free. Fan arrays are loud by design; propeller noise work needs a treated section.
- Sizing for today's airframe. Both facilities outlive the vehicle that justified them. Specify for the family, not the prototype.
8. Deciding with your own test matrix
The honest summary: if your open questions are about behaviour, buy the array. If they are about numbers, buy the tunnel. If they are about certification and product economics, you will eventually need both — plan the order deliberately instead of discovering it.
We design and build custom closed-loop tunnels, including open-jet sections for free-flight work, and we will say plainly when a project does not need one yet. The starting point is the same either way: what has to be measured, to what uncertainty, at what speed, on what airframe. Related reading:
[Article not found: evtol-drone-crosswind-gust-wind-tunnel-testing]and our
[Article not found: horizontal-vertical-inclined-wind-tunnels-comparison].
Send us your test matrix and we will tell you which parts of it need a tunnel — and which do not.
Sources and notes
- Fan array architecture (modules of about 25 × 25 cm, nine wind pixels of counter-rotating fans, individually controllable, hierarchical control): How a WindShaper Works, Tyto Robotics.
- Per-pixel rated speed of 16 m/s and module architecture: Custom Wind Tunnel Datasheet, Tyto Robotics/WindShape.
- Active per-cell control and turbulence characterisation of fan-array wind generators, as a general reference on the technology class: Turbulence enhancement of a fan array wind generator, arXiv and Aerodynamic Characterization of a Fan Array Wind Generator, arXiv.
- Quoted commercial limits of up to 20 m/s, up to about 58 m/s with a convergent, turbulence intensity around 5% and below 1% with a flow filter: How WindShaper fan arrays enhance drone testing, Dantec Dynamics.
- Real-scale atmospheric wind and turbulence replication with fan arrays for UAV and AAM validation: AIAA SciTech 2023-0812.
- Energy comparison between closed and open circuits (60–70%) is TunnelTech project data, consistent with our published article on circuit selection.
- Figures for specific commercial products are cited from their suppliers and should be re-checked against the current datasheet before use in a proposal.
Ζητήστε διαβούλευση
Talk to our engineering team about your UAV test facility