Buyer's guide

Soft starter or variable frequency drive: which one the duty actually needs

29 September 2026 11 minute read

If the process needs the motor to run at less than full speed at any point, a drive is already the answer — nothing else holds a reduced running speed. If the motor only ever needs to be at full speed or stopped, and the start and stop just need softening, the cheaper device is usually the right one. The rest of this article works through why, factor by factor, so the choice can be checked against the actual duty rather than a habit.

How a soft starter controls the motor

A soft starter controls voltage only; it does not touch supply frequency. Each line into the motor carries a pair of anti-parallel thyristors, and the control stage fires them later or earlier within each half-cycle of the AC waveform. Firing later lets less of the half-cycle through; firing earlier lets more through. That firing angle is what sets how much of the incoming voltage actually reaches the motor at any instant.

During a start, the firing angle is delayed progressively less on each successive half-cycle, so the average voltage delivered to the motor climbs from whatever starting level was set up to full line voltage by the end of the ramp time. Frequency never moves — only the amount of the fixed-frequency waveform that gets through changes. That single fact is the root of most of the differences covered further down: a soft starter reshapes an AC sine wave, it does not synthesise a new one.

How a drive controls the motor

A variable frequency drive works differently at the front end. Rather than gating the incoming waveform the way a soft starter does, it builds its own output internally, which lets it set frequency and voltage together — coordinating a lower frequency with a lower voltage at start, for instance — instead of being limited to adjusting voltage alone.

Building its own output this way is also the source of most of the drive-side complications discussed later: a device that shapes its own output waveform keeps an active switching stage in the circuit for as long as it runs, and that ongoing switching is what the harmonic and cable-length discussion further down comes back to. A soft starter, gating an existing AC waveform rather than building a new one, has no equivalent stage once it stops gating.

Starting torque and current — the square-law relationship

Induction-motor torque at a fixed frequency varies with the square of the applied voltage, while starting current varies roughly in direct proportion to voltage. That mismatch matters at start-up: cutting voltage during a soft start costs torque much faster than it costs current, because the current falls in a straight line with voltage while torque falls with voltage squared.

A drive avoids that particular penalty because it does not have to hold frequency fixed while cutting voltage. It can lower frequency together with voltage, keeping the volts-per-hertz ratio — and with it, motor flux and available torque — close to the rated value throughout the ramp. That is the practical difference behind "a drive starts smoothly at full torque, a soft starter trades torque for a gentler current profile."

None of this removes the soft starter's core purpose, though. Even where no ongoing speed control is wanted, ramping the voltage rather than slamming it on in one step is what cuts the inrush that a direct-on-line start would otherwise deliver instantly.

Worked example: does a 60% voltage-reduction start actually start the load?

Take a motor with a full-load current (FLC) of 50 A. Started direct-on-line, its locked-rotor current runs 5 to 8 times FLC; taking the mid-point of that range, 6.5 × 50 A = 325 A. Assume, as a stated modelling input rather than a researched figure, that this motor's DOL locked-rotor torque is 150% of full-load torque (FLT).

Set the soft starter to a 60% voltage-reduction start point (that is, 60% of line voltage applied at the start of the ramp):

• Current scales with voltage: 0.6 × 325 A ≈ 195 A.
• Torque scales with voltage squared: 0.6² × 150% FLT = 0.36 × 150% FLT ≈ 54% FLT.

Now check that 54% FLT against two different loads:

• A loaded conveyor with a 60% FLT breakaway-torque requirement: 54% is short of 60%. The motor will not break the load away — it stalls at this setting.
• A centrifugal pump with a 40% FLT breakaway-torque requirement: 54% clears 40% comfortably, and the pump starts.

The same 60% voltage-reduction setting starts one load and stalls the other. That is the practical reason a soft starter's start setting has to be checked against the specific machine's breakaway torque rather than picked from a generic table.

Is "current limit" a separate device?

No. Current limit is a starting mode set within the soft starter's own control, using the same firing-angle mechanism as the voltage ramp — the control simply caps current during the start instead of following a fixed voltage-versus-time profile. It is a setting, not a separate piece of equipment, and it still cuts inrush for the same reason a voltage ramp does: the full line voltage is never applied in one step.

Motor starter types — where direct-on-line fits

Direct-on-line — a single step, not a ramp

Direct-on-line (DOL) starting applies full voltage to the motor in a single step, with no ramp at all. Whether that is acceptable comes back to the same torque-and-current arithmetic worked through above: DOL delivers full locked-rotor current and full locked-rotor torque at once, so it only suits a duty where the supply can absorb that inrush and the mechanical system does not need it softened. Short-circuit protection ahead of the starter still has to be sized for that inrush; the breaker sized for short-circuit protection ahead of the starter is a separate design decision from the starter itself, but the two are chosen together.

Energy saving — the cubic power law, and where it is actually captured

The affinity-law relationship

For a centrifugal pump or fan, power drawn varies roughly with the cube of shaft speed. A modest speed trim therefore cuts power by a much larger fraction: running at 80% of full speed draws roughly 0.8³ ≈ 51% of full-speed power — around half, from a 20% speed reduction. That relationship is specific to variable-torque, variable-speed loads. A load that must run at essentially constant speed and torque gets no equivalent running saving from speed control, because there is no lower-power operating point to move to.

Capturing it means replacing throttle control, not swapping the motor

Where flow is currently trimmed by a mechanical damper or valve while the motor runs at full speed, the saving is captured by replacing that throttle with drive-based speed control, not by anything to do with the motor itself. A soft starter cannot deliver this saving under any circumstance: its job ends once the ramp finishes and the bypass contactor closes, at which point the motor is running at full speed exactly as it would after a DOL start.

Harmonics injected into the supply, and what mitigation each needs

A drive's rectifier front end draws non-sinusoidal current from the AC supply, injecting harmonic currents upstream into the plant's power system. A soft starter's semiconductors gate an existing waveform rather than rectifying to a DC bus, so they are not this type of harmonic source.

Where a drive's harmonic injection needs to be assessed or limited, several reference documents apply depending on jurisdiction and connection point. IEEE 519 sets recommended limits on harmonic voltage and current distortion at the point of common coupling between a facility and the utility, and is the reference most commonly used to judge whether a drive installation needs mitigation. IEC 61800-3 sets the electromagnetic compatibility requirements, including conducted emission limits, for adjustable-speed power drive systems, and is the IEC-side analogue. In the UK, Engineering Recommendation G5 sets planning levels for harmonic voltage distortion at the point where non-linear equipment — drives included — connects to a distribution network, and is the network-code document a project may need to satisfy alongside or instead of IEEE 519.

Where an installation sits close to or beyond the applicable IEEE 519 or IEC 61800-3 limit, mitigation is available; the specific technique is a detailed-design decision that trades performance against cost and complexity, made once the installation's distance from the limit has been established. Neither standard prescribes a single required technique — both set the limit to be met, not the method of meeting it.

Cable length limits and motor insulation stress on a drive output

A drive's output stage keeps switching throughout the run, and its fast-edged, pulse-width-modulated output travels along the motor cable as a series of pulses. At the impedance mismatch where the cable meets the motor windings, those pulses can reflect, and where the incoming and reflected pulses add, the voltage seen at the motor terminals can substantially exceed the drive's own output voltage. This effect worsens as cable length increases, because it is the cable's own inductance and capacitance that sets up the travelling-wave behaviour behind it — short runs see much less of the effect than long ones. That is why drives normally carry a cable-length limit for a given switching frequency, or need added output filtering beyond it.

NEMA MG 1 Part 31 sets insulation withstand requirements for motors intended to be fed from PWM inverters, including limits on peak voltage and minimum rise time at the motor terminals. For a 460 V motor, the referenced requirement is capability to withstand roughly 1,431 V peak at a rise time around 0.1 microsecond. A motor not built to that standard, run on a drive with a long unfiltered cable, carries raised risk of premature winding-insulation failure from the repeated voltage-stress pulses — which is why inverter-duty (or equivalently rated) motors are specified for drive service.

The two halves of NEMA MG 1 treat this differently, and the difference is the specification point. Part 1 classifies a motor's insulation system by thermal-endurance class alone — how hot the winding may run, and for how long. It sets no provision for peak voltage, voltage rise time or waveform, because nothing in a directly-supplied motor's service imposes them. Part 31 adds exactly those provisions for inverter-fed motors. A soft starter once bypassed, and a DOL contactor throughout, deliver an unswitched waveform, so neither raises the question Part 31 exists to answer.

Heat and panel space — bypass contactor against continuously switching output

Once a soft starter has ramped a motor to full voltage, a bypass contactor closes and carries the running current through a set of contacts instead of through the thyristors. From that point until the motor stops, the semiconductor stage is out of the current path entirely — it is only active during the start and stop transients.

A drive's output stage has no equivalent bypass. It keeps switching for the entire run, typically at frequencies in the low-kilohertz range, and that continuous switching is both the source of the drive's characteristic output waveform and a continuous loss path that the enclosure must dissipate for as long as the motor runs. The soft starter side of that contrast has a citable figure behind it: left unbypassed, the thyristor stage dissipates roughly 1.5 W for every amp of running load, on each phase — and that is exactly the loss a bypass contactor takes out of the enclosure once the ramp finishes. A drive's output stage carries an equivalent-in-kind loss for as long as the motor runs, with no bypass to remove it.

The comparison table

The table below draws only on the factors already established above — nothing new is introduced here, it is a synthesis laid out for comparison.

Questions buyers ask

Does a soft starter save energy the way a drive does?

No, not while running. It only smooths the start and stop; once its bypass contactor closes the motor runs at full speed and draws full-speed power. A drive saves energy only on variable-torque loads such as pumps and fans, where the cubic power-speed relationship applies.

Can a soft starter replace a VFD if the process might need speed control later?

No. A soft starter has no mechanism to hold a reduced running speed — its function ends at the bypass. If any part of normal operation needs less than full speed, the process needs a drive from the start.

Why does a drive need a longer cable-length allowance than a soft starter?

Because its output keeps switching at kilohertz frequencies throughout the run, and those fast edges can reflect at the cable-to-motor impedance mismatch, raising terminal voltage as cable length grows. A soft starter's bypassed run carries no such switching.

Is a plain direct-on-line starter ever still the right choice?

Yes, for small motors and low-inertia loads where the supply can absorb the inrush and the mechanical system tolerates a full-voltage step start. It remains the cheapest option when neither a soft ramp nor speed control is needed.

Does a soft starter need an inverter-duty motor the way a drive does?

No. Once bypassed, a soft starter delivers the same unswitched supply as a DOL contactor. The voltage-stress provisions that NEMA MG 1 Part 31 adds for inverter-fed motors — peak voltage and rise time — answer a question a bypassed soft starter never raises. Part 1 classifies insulation by thermal-endurance class alone and carries no such provision.