Fan Performance Curve or Fitted Parabola? How to Tell in AHU Software

Every AHU selection tool draws a fan curve, and it is the most reassuring picture on the screen: a smooth arc, your operating point sitting on it, efficiency and absorbed power to two decimals. The picture tells you nothing about where the numbers came from. A parabola fitted through your own duty point looks exactly like a line read from the manufacturer's selection engine — and diverges from it by tens of percent at the ends.

What a fitted curve looks like from the inside

The cheapest possible fan curve is psf = a − b·Q²: a parabola anchored on the duty point you asked for, zeroed at some arbitrary multiple of your flow, with efficiency drawn as a triangle peaking near the middle and shaft power derived from those two. It costs nothing to compute, it always succeeds, and it is smooth everywhere.

That last property is the tell. Real fan lines are not parabolas: backward-curved impellers change shape approaching free delivery, forward-curved wheels have a saddle, and a fitted arc has no stall region and no surge limit.

We measured this on our own code in August 2026. The numbers are worth quoting because they are typical of the method rather than of any one implementation — the same fan's published line, at the same speed:

Point	Fitted parabola	Supplier's published line	Error
500 m³/h (near shut-off)	~597 Pa	~456 Pa	+31 %
Duty point, 4 250 m³/h	matched by construction	—	0 %
7 440 m³/h (near free delivery)	~47 Pa	~112 Pa	−58 %
Absorbed power at duty	648 W	815 W	−20 %
Static efficiency at duty	75.0 %	59.7 %	+15 points

The fit is exact at the anchor and wrong everywhere else — including on the two quantities that decide compliance. A 20 % power error at the duty point is not a rounding issue; it moves SFP and it moves the ECP-05-2026 class letter.

The trap that makes a fake curve look real

There is a subtler failure than an honest fit, and it catches careful implementations. Some vendor selection libraries hold the requested static pressure as session state. Sweep the flow alone — 2 500, 3 000, 3 500 m³/h — and the engine dutifully answers, but it is solving the speed required to hold your original pressure, not the fan's line at constant speed. Pressure comes back identical at every flow while the reported rpm climbs. Plot that and you get a flat line, not a curve.

The correct call pins both coordinates on each point read from the engine's own chart data. Pinned that way, speed holds to under 1 % across the line and every point reports valid. Vary flow alone and you are reading your own input back.

Four questions that separate the two

1. Ask for a curve on an article number that does not exist. A tool reading the supplier's engine returns nothing, or an error. A fitted tool returns a complete, confident curve — the fit does not need the fan to exist.
2. Check the free-delivery end. A published line runs down to zero static pressure. Software that filters out "empty" values often drops that point silently, because a measured 0 Pa and a missing value look the same to sloppy code. A curve that stops short of the axis has been trimmed.
3. Change the duty and re-plot. If the whole curve translates with your operating point, it is anchored on you. The vendor's line for a given fan at a given speed does not move because you changed your mind about the duty.
4. Ask what the curve is a curve of. At constant speed? At constant pressure? Scaled from one catalogue measurement by the vendor's own affinity routine? All three are legitimate and they are different objects. A tool that cannot say which one it drew did not read one.

What "no curve" should look like

When the engine returns no chart block or an error, the honest answer is a refusal carrying the engine's own sentence — not a synthesised arc. This matters most on EC plug fans with hub motors, where the impeller sits on the motor's rotor and there is no shaft to measure. Those machines publish mains input power and system static efficiency; shaft power and shaft efficiency come back as zero because they do not exist. A tool that treats those zeros as measurements prints 0.00 % efficiency; one that back-calculates a shaft power from a chosen motor efficiency is inventing the number ErP (EU) 1253/2014 is assessed on.

Two housekeeping points decide whether a curve is comparable at all. Test conditions: ISO 5801 / AMCA 210 installation type, quoted with the reference air state of 1.20 kg/m³ at 20 °C — density scales the pressure and, in most engines, silently decides which fans are offered. And system effect: AMCA 201 defines factors read off curves per outlet geometry and inlet condition, not a flat percentage. A tool adding a blanket 5 % is applying a habit, not a standard. See also our note on certified manufacturer data versus curve fits.

AirSelect3D reads the operating-point curve from the supplier's own chart engine, labels every curve with its basis, and refuses rather than fabricates when the engine has nothing to give.

Plot a real fan curve against your duty →