Low-speed wind tunnels and wall corrections

Why it matters

Most of the airfoil and wing data used in design — lift curves, drag polars, stall behaviour — come from low-speed wind tunnels. A tunnel is not free air: the walls squeeze the flow round the model and constrain its downwash, so raw readings are biased by a few percent or more. Knowing how a tunnel works and how to correct its data is what turns measurements into numbers you can design with.

Key ideas

Layout of a low-speed tunnel.

• Open-circuit (Eiffel) tunnels draw air from the room and exhaust it back; simple and cheap but sensitive to room disturbances. Closed-return (Göttingen) tunnels recirculate the air; quieter, more uniform flow and lower power, at higher cost.
• Closed (solid-walled) test sections are most common; open-jet sections give easy access but different wall effects.
• Components, upstream to downstream: settling chamber with honeycomb (straightens the flow) and fine screens (reduce turbulence); contraction (area ratio typically 6–10) that accelerates the flow and further reduces relative turbulence; test section; diffuser (small angle, about 3° half-angle, to recover pressure without separating); fan and, in a closed return, corner turning vanes.

Setting and measuring speed. Continuity and Bernoulli between the settling chamber (area A₁) and the test section (A₂) give the test-section speed from the measured static-pressure drop across the contraction. A Pitot-static probe gives q∞ directly. Forces are measured with an external or internal strain-gauge balance, pressures with surface taps, drag of 2-D models often with a wake rake (momentum deficit).

Similarity. Model tests must match the full-scale Reynolds number as closely as possible (and Mach number when it matters). Low-speed tunnels usually fall short in Re, so boundary-layer trips are used to force transition and c_l,max and c_d need scale corrections.

Wall (boundary) corrections. The walls change the flow in several ways:

• Solid blockage: the model's volume narrows the effective passage, so the local speed past the model is higher than the measured upstream speed. Correction factor ε_sb depends on model volume and tunnel area.
• Wake blockage: the low-speed wake also reduces the effective area, raising the speed outside it: ε_wb ∝ (model size/tunnel size)·c_du. Total blockage ε_b = ε_sb + ε_wb is applied as V = V_u(1 + ε_b), so q = q_u(1 + ε_b)² — corrected coefficients are smaller than uncorrected ones.
• Streamline curvature (2-D tests): the floor and ceiling straighten the flow, making the model act as if it had extra camber; this raises c_l and the nose-down moment. The parameter is σ = (π²/48)(c/h)².
• Lift interference / downwash correction (3-D tests): walls of a closed tunnel prevent the full downwash of the trailing vortices (they act like image vortices giving upwash), so the measured α is too low and induced drag too small. Corrections: Δα = δ(S/C)C_L and ΔC_D = δ(S/C)C_L², where δ depends on tunnel shape and model span (from charts) and C is the test-section area.
• Horizontal buoyancy: the static pressure falls along a closed test section as the wall boundary layers grow, producing a spurious drag force on the model.

Typical good practice: model frontal area under about 5–7.5 % of the test-section area, wing span under about 0.8 of the tunnel width.

Reducing wall interference at source: slotted or perforated walls, adaptive (streamlined) walls, or small models; residual effects are corrected by formulas or by CFD of the model in the tunnel.

Formulas

• A₁V₁ = A₂V₂ and p₁ − p₂ = ½ρ(V₂² − V₁²) ⇒ V₂ = √{2(p₁ − p₂)/[ρ(1 − (A₂/A₁)²)]} — tunnel speed from contraction pressure drop.
• Re = ρVc/μ.
• V = V_u(1 + ε_b), q = q_u(1 + ε_b)², ε_b = ε_sb + ε_wb — blockage; subscript u = uncorrected.
• 2-D: σ = (π²/48)(c/h)², h test-section height (m); ε_wb = (c/2h)c_du; ε_sb from model shape factor in your data book.
• 2-D corrected values (Barlow, Rae and Pope): c_l = c_lu(1 − σ − 2ε_b); α = α_u + (57.3σ/2π)(c_lu + 4c_m,c/4u) (degrees); c_d = c_du(1 − 3ε_sb − 2ε_wb); c_m,c/4 = c_m,c/4u(1 − 2ε_b) + σc_l/4.
• 3-D closed tunnel: α = α_u + 57.3·δ(S/C)C_L (degrees) and C_D = C_Du + δ(S/C)C_L².

Worked examples

Example 1 (standard). A tunnel has a contraction area ratio A₁/A₂ = 9. The static-pressure drop across the contraction is 1500 Pa in air of ρ = 1.225 kg/m³ and μ = 1.79 × 10⁻⁵ Pa·s. Find the test-section speed and the Reynolds number per metre.

1. (A₂/A₁)² = (1/9)² = 0.01235.
2. V₂ = √{2 × 1500/[1.225 × (1 − 0.01235)]} = √(3000/1.2099) = √2479.6 = 49.8 m/s.
3. Re per metre = ρV/μ = 1.225 × 49.8/1.79 × 10⁻⁵ = 3.41 × 10⁶ per m.

Answer: V ≈ 49.8 m/s; Re ≈ 3.4 × 10⁶ per metre of model length.

Example 2 (GATE level). A 2-D airfoil of chord 0.3 m spans a closed test section of height 1.2 m. Raw data: α_u = 6°, c_lu = 0.80, c_du = 0.012, c_m,c/4u = −0.050. Solid blockage from the data book is ε_sb = 0.0065. Apply the 2-D corrections.

1. σ = (π²/48)(0.3/1.2)² = 0.2056 × 0.0625 = 0.01285.
2. ε_wb = (c/2h)c_du = (0.3/2.4) × 0.012 = 0.0015; ε_b = 0.0065 + 0.0015 = 0.0080.
3. c_l = 0.80 × (1 − 0.01285 − 0.016) = 0.80 × 0.9711 = 0.777.
4. α = 6 + (57.3 × 0.01285/2π)(0.80 + 4 × (−0.05)) = 6 + 0.1172 × 0.60 = 6.07°.
5. c_d = 0.012 × (1 − 0.0195 − 0.003) = 0.0117.
6. c_m,c/4 = −0.05 × (1 − 0.016) + 0.01285 × 0.777/4 = −0.0492 + 0.0025 = −0.0467.

Answer: corrected α = 6.07°, c_l = 0.777, c_d = 0.0117, c_m,c/4 = −0.047. The raw lift was about 3 % too high.

Common mistakes

• Correcting in the wrong direction: blockage raises the true speed at the model, so corrected coefficients are lower than raw ones.
• Forgetting the (A₂/A₁)² term when deriving speed from contraction pressure drop.
• Treating wall corrections as a single additive "factor"; each effect has its own formula and sign.
• Using 3-D downwash corrections on a 2-D (wall-to-wall) model, or vice versa.
• Ignoring the Reynolds-number mismatch between model and full scale.
• Choosing a model so large that corrections exceed a few percent — they then become unreliable.

For GATE AE

Expect: test-section speed from contraction pressure difference; Reynolds number of a model test; blockage ratio; 2-D streamline-curvature parameter σ and corrected c_l or α with given correction factors; 3-D Δα and ΔC_D from δ, S and C; and MCQs on tunnel components (honeycomb, screens, contraction, diffuser angle) and the physical origin of each wall effect.

Quick check

1. What is the purpose of the honeycomb and screens?
2. Blockage ratio of a 0.5 m² model in a 10 m² test section?
3. Do blockage corrections increase or decrease measured coefficients?
4. Why does a closed tunnel under-predict induced drag?

Answers: 1. The honeycomb straightens the flow; screens reduce turbulence and even out the velocity profile. 2. 5 %. 3. Decrease them (true q is higher than measured). 4. The walls suppress part of the trailing-vortex downwash (upwash from image vortices), so α_i and C_D,i appear smaller.