The promise
Every variable speed drive proposal is built on one relationship. Slow a centrifugal pump down and flow falls in proportion to speed, head falls with the square of speed, and power falls with the cube of speed.
That third one is the whole business case. Drop to 80% speed and power falls to roughly 51%. Drop to 50% and you are at about 12.5%, one-eighth of the energy for half the flow.
Against throttling a valve, where you burn the energy you just paid to create, this looks unanswerable. Often it is. But a significant number of installed drives never deliver anything close to their projected savings, and the reason is nearly always the same, knowable before the order was placed, and simply not asked about.
State the laws properly first
The affinity laws are frequently written with speed and diameter combined, which causes real errors. They apply to one variable at a time.
For a speed change at fixed diameter: Q2 = Q1 x (N2/N1), H2 = H1 x (N2/N1) squared, P2 = P1 x (N2/N1) cubed.
For a diameter change at fixed speed, the same exponents apply to the diameter ratio, with the caveat that impeller trimming is an approximation that degrades beyond roughly 10 to 15% reduction, because trimming changes the blade geometry's relationship to the casing in a way that a speed change does not.
What the laws do not say is that your system will behave this way. They describe the pump alone. You can work any of this through with our affinity laws calculator.
The number that decides everything: static head
A system curve has two parts.
Friction head is the energy lost pushing liquid through pipe, fittings and equipment. It rises roughly with the square of flow.
Static head is the vertical lift plus any pressure you must overcome at the destination: a tank elevation, a vessel pressure, a boiler drum. It does not change with flow at all. At zero flow it is still there.
That distinction decides your savings.
A pure friction system - circulation loops, cooling water, filter feed, anything returning to roughly the level it started from - has a system curve rising from the origin. Reduce speed and the new operating point tracks down close to the pump's own affinity curve. Here the cube law substantially holds, and a drive is often excellent.
A high static head system - transferring to an elevated tank, feeding a pressurised vessel, pumping up a hill - has a system curve starting high on the head axis and rising gently. Now the picture changes completely.
Because the pump's head falls with the square of speed while the static head stays put, most of the head reduction goes into no longer overcoming the static lift. You lose flow rapidly for relatively little power saving. And there is a minimum speed below which the pump develops less head than the static head and delivers nothing at all: it runs, it consumes power, and no liquid moves.
A rough rule: if static head is more than about half your total head, treat cube-law savings claims with suspicion until someone has plotted the actual curves. If static head is under about 20% of the total, the savings are usually real.
A worked comparison
Two systems, both 50 m total head at 100 cu m/h.
System A has 5 m static and 45 m friction. Slow to 80% speed: flow drops to roughly 82 cu m/h, and power to somewhere near 55% of the original. Close to the theoretical 51%, and a strong result.
System B has 40 m static and 10 m friction. Slow to 80% speed: pump head falls to about 32 m, which is below the 40 m static head. Flow stops entirely. The usable speed range is perhaps 90 to 100%, and the achievable saving is a fraction of what the brochure implied.
Same pump, same drive, same cube law. Completely different outcomes, decided by a number that costs nothing to establish beforehand. Our head converter and pressure converter are useful here if your figures arrive in mixed units, which they usually do.
The four other limits worth knowing
Motor cooling. A standard TEFC motor is cooled by a fan on its own shaft. Slow the motor and you slow the fan, so cooling falls faster than the load does. A motor running continuously at 30 to 40% speed under significant load may overheat even though it is nowhere near full-load current. Continuous low-speed operation needs either an inverter-duty motor, separate forced ventilation, or a documented minimum speed.
Minimum flow. Slowing the pump reduces flow, and reduced flow eventually takes you into the region where internal recirculation begins, with the radial loading, vibration and seal consequences described in our companion article on mechanical seals. The pump minimum continuous stable flow does not disappear because a drive is fitted. It is a hard floor, and the drive's minimum speed setting must respect it.
NPSH required changes with speed. Reducing speed reduces NPSHr, which is genuinely helpful. But if you later use the drive to run above nameplate speed, which is technically possible and occasionally proposed, NPSHr rises steeply and the suction margin you designed around may disappear. Check it before over-speeding anything.
Electrical effects. Drives generate harmonics that can affect other equipment on the same supply, and long motor cables can produce voltage reflections that stress winding insulation. Neither is a reason to avoid drives; both are reasons to involve someone electrical early rather than treating the drive as a mechanical purchase.
When something else is the better answer
A drive is not the only way to reduce a pump's output, and it is not always the cheapest.
Trim the impeller if the duty is genuinely fixed and the pump is simply oversized. It costs a fraction of a drive, adds no electrical complexity, and the energy saving is permanent. The trade-off is that it is not reversible and you need confidence the duty will not change. Our impeller diameter check calculator will tell you whether a proposed trim stays within sensible limits.
Consider two smaller pumps where demand varies widely. Running one pump near its BEP often beats running one large pump at 45% speed, and you gain redundancy.
Fit a drive when flow demand genuinely varies over time, when the system is friction-dominated, and when you want control as well as savings: soft starting, process stability, and the ability to retune as the plant changes.
The question to ask first
Before anyone quotes you a payback period, ask: what is the static head, and what is the minimum speed at which this pump still delivers flow?
If the answer is a shrug, the payback figure is a guess. If the answer is a number, you can calculate the real saving yourself in about ten minutes, and the pump power calculator will do the arithmetic once you have the duty points.
The cube law is real. It just describes the pump, and you are buying a result from the system.
