Centrifugal fan and pump shaft power scales with impeller speed cubed, so a 20% speed reduction drops power to about 49% of full-speed draw and a 10% cut still removes 27% of input power, per the fan affinity-law summary published 2025-05-07 [S1].
The same cubic rule applied to pumps shows a 20% slow-down on a 20 HP centrifugal dropping shaft load to 10.2 HP, illustrating why variable speed drives (VFDs) are the default retrofit for variable-flow centrifugal loads rather than valve or damper throttling [S5].
The three affinity relationships, with the numbers engineers actually use
Flow scales linearly with speed (Q₂/Q₁ = N₂/N₁), head or pressure scales with the square (H₂/H₁ = (N₂/N₁)²), and shaft power scales with the cube (P₂/P₁ = (N₂/N₁)³) [S1][S5]. A halving of speed therefore cuts flow by 50%, head by 75%, and power by 87.5% on a centrifugal machine whose impeller geometry and system curve stay fixed, as called out in ANSI/HI 9.6.7 and ASHRAE Fundamentals [S3]. The power relationship is not an independent empirical fit; it falls directly out of power being proportional to flow times head, so the cube term is mathematically forced by the first two [S3]. The table below lines up the working numbers every process engineer should keep in their head before opening a VFD datasheet, all derived from the (N₂/N₁)³ formula [S1][S3][S4][S5].
The 80% / 51% cell is the headline number a controls engineer quotes in a payback meeting, because a 20% turn-down is the realistic operating range for most HVAC and process fans most of the time, not a best case [S3].
Throttling valve, inlet vanes, and VFD on a friction-dominated curve
A throttling valve or outlet damper does not change the affinity law for the fan or pump itself; it forces the machine to push the same flow against an artificially steepened system curve, so the impeller still spins near full speed and roughly 90-95% of full power is still drawn at 70% flow [S3]. Variable-inlet-vane control on fans improves on a damper, recovering some of the loss down to about 65-75% of full power at 70% flow, but only VFD speed reduction reaches the 35-45% band the affinity law predicts [S3]. For a constant-torque load that would not be a good trade, but fans and pumps are quadratic-to-cubic torque loads, and that load shape is exactly what a variable-speed-drive is sized against, with V/f control curves and short-term overload ratings that fit this load profile cleanly [S3]. Practical VFD retrofit cost on a centrifugal pump cluster runs roughly $8,000-$15,000 per drive, against $500-$2,000 for an impeller trim that gives a permanent turn-down but kills flexibility [S5]. The selection rule is direct: trim the impeller when operating point is fixed, install a VFD when demand moves daily or seasonally, and expect a 6-18 month payback on the VFD path on high-runtime variable loads [S5].
When the cube law overstates the savings
Affinity laws are most accurate for speed changes under about 25% from the rated point; beyond that, hydraulic or aerodynamic efficiency shifts and the predictions start drifting from the manufacturer curves, so any major turn-down should be verified on those curves or with field testing [S5]. Motor efficiency itself moves opposite to speed, because induction motors typically get more efficient at higher RPM, which slightly offsets the cube-law savings at very low speeds and is one reason VFD-driven systems do not quite hit the textbook 87.5% number at 50% speed [S4]. Static head (lift, elevation, or a closed-loop minimum pressure) is the other killer of the cube law: where the system curve has a significant static component, power falls more slowly than (N/N₁)³ because the impeller is still pushing that constant head even at low flow, and a VFD can actually over-pressurise the loop if the minimum speed is set too low [S3]. A minimum-speed floor of 30-40% on fans and 40-50% on pumps is a common industrial starting point, paired with a PID loop on the controlled variable (duct static, differential pressure, or flow) to ride the curve [S3].
Real load shapes where the cube law actually pays back
Data-hall CRAH and computer-room-air-handler fans are the cleanest case because the load tracks IT heat output, which itself follows utilisation, and most rooms sit well below 100% load for most of their life; a 10% speed cut on the supply fan removes 27% of fan power per the affinity-law summary [S1]. HVAC supply and return fans on commercial buildings are the second cleanest case, with a 20% turn-down cutting fan power by roughly 50% under the (0.8)³ relationship [S4]. Industrial chilled-water pumps, cooling-tower fans, and boiler combustion-air fans all have the quadratic-to-cubic torque shape that lets a VFD ride the affinity curve, and a related control-and-sourcing walkthrough of plant air-handling and construction machinery and equipment shows the same motor-control logic applied to dust-collection and tunnel-ventilation fans on infrastructure jobs. For pump-side systems, the same physical rule, plus the Cv derating and line-size matching rules covered in butterfly-valve sizing for a 12 inch water main, is what determines whether a VFD or a control valve should be the primary flow regulator.
Standards, sourcing, and what to pin in the spec
The pump-side affinity relationships are formalised in ANSI/HI 9.6.7, with ASHRAE Fundamentals as the common companion reference for the fan side [S3]. For the drive itself, energy management clauses in the project spec should call out V/f or sensorless vector control, a 110-150% short-time overload rating for 60 s, and a minimum switching frequency compatible with the motor insulation class, because the drive is the component that actually delivers the cube-law energy result to the energy meter downstream. For facilities where lighting and fans share a load shed scheme, the same VFD bus can be coordinated with lamps and light fittings controls to flatten demand peaks; the affinity-law savings are a one-line entry in the energy model and they stack with the lighting and lighting equipment and electric lamps reductions rather than competing with them. The next trackable signal to watch is the manufacturer's published part-load efficiency curve on the actual motor-VFD pair being bid, because that is the only document that converts the textbook 51% number into a real kWh figure the utility bill will recognise [S3].