XFOIL asks you for Ncrit. Here's what a wrong answer costs.

Ncrit isn't a property of the airfoil. It's a property of your environment — how quiet the wind tunnel is, how clean the incoming air is. Yet it hides behind a default of 9 that most people never change.

NACA 0008 Re 1.0×10⁶ α 0.00° Ncrit 9.00
vs. XFOIL's default Ncrit = 9
Ncrit 9.00
Small marks: the five test environments below. Bright mark: XFOIL's default.
Angle of attack α 0.00°

Airfoil

Where are you testing?

Laminar Turbulent Where transition would be at the default Ncrit = 9

What the N factor actually is

Small disturbances in the boundary layer get amplified as they travel downstream. N is the natural logarithm of that amplification: N = 9 means a disturbance has grown about 8,100 times. The eN method assumes transition happens once disturbances have grown by some critical factor.

The catch is that the critical factor depends on how large the disturbances were to begin with, and that depends on how clean the incoming flow is. This is the receptivity problem: there is still no theory that lets you compute how outside disturbances enter the boundary layer. So Ncrit has to be calibrated. It cannot be derived.

Thin symmetric sections are hit hardest

Switch between the two airfoils. Moving Ncrit from 1 to 13.5 shifts transition on the NACA 0008 by about 69% of the chord and changes drag by a factor of 2.24. On the NACA 2412 the figures are 38% and 1.86. The thin symmetric section is roughly twice as sensitive to this parameter as the cambered one. Thin, low-Reynolds-number sections are exactly what model aircraft and small drones use, and exactly where people most often leave the default in place.

The two surfaces swap order near N ≈ 8

On the NACA 2412 at α = 0°, the upper surface transitions further aft than the lower surface when Ncrit is low. Near N ≈ 8 the two coincide, and above that the order reverses. At low N, transition happens far forward, where both surfaces are still in a favorable pressure gradient. At high N, transition is pushed aft, the adverse pressure gradient on the upper surface takes over and pulls its transition point back, while the lower surface, with a much gentler pressure recovery at zero incidence, stays laminar longer. A symmetric section can never show this, because its two surfaces are identical.

Why the curve flattens at high N

Raising Ncrit moves transition aft, but not past the point where the laminar layer separates. Once an adverse pressure gradient separates the laminar flow first, no value of N changes anything. This is a built-in limit of the eN method, not a numerical error. It is also why high-Ncrit cases get harder to converge: the gaps in the curve are cases XFOIL could not solve, not values of zero.

When this tool will mislead you

Mack's correlation, N = −8.43 − 2.4 ln(Tu), holds only for freestream turbulence between roughly 0.08% and 3%. Below that range the fit has no data behind it. Above it, the boundary layer goes turbulent through bypass transition and the eN approach stops applying altogether; no value of N describes it. When an environment falls outside the range, the panel says so instead of handing you a number.

Hollow dots on the curve are computed data points. The lines between them are linear interpolation.