---
title: "Choosing a VFD: what actually decides the size"
description: "Drives are sized on current, not kW — and the current on the box is before derating. Overload class, control mode, cable length and harmonics."
date: "2026-05-03"
author: "Divakar B"
source: "https://energycalchq.com/blog/choosing-a-vfd"
license: "© 2026 Divakar B. Quote with attribution to https://energycalchq.com/blog/choosing-a-vfd"
---

A variable frequency drive is usually specified by matching kilowatts: 22 kW
motor, 22 kW drive, done. That works often enough to be a habit and fails in a
way that is expensive to discover, because the drive runs fine on the bench and
trips on overload in August.

Drives are rated in current. The kilowatt figure on the box is a convenience
label that assumes a standard four-pole motor at 400 V, 40 °C, sea level and the
factory switching frequency. Change any of those and the usable current falls.

## Size on current, then derate

![Bar chart of a 45 A drive falling to 32.5 A after ambient, altitude and switching derating, below the 38.2 A motor it was bought for](/blog/vfd-derating-stack.svg "Each factor applies to what the previous one left. The drive that matched on kW is six amps short of the motor.")

Start with the motor's **full load current** from its nameplate — not from a
chart, and not from its kW rating. Then apply what your installation does to the
drive:

| Condition | Typical factor |
|---|---|
| Ambient above 40 °C | −2 % per °C, to about 50 °C |
| Altitude above 1000 m | −1 % per 100 m |
| Switching frequency raised from 4 to 8 kHz | −15 to −25 % |
| Switching frequency raised to 12 kHz | −30 to −40 % |

The switching frequency one surprises people. Raising it makes the motor quieter
and the current waveform smoother, and it is often turned up during
commissioning by somebody solving an audible-noise complaint. It also increases
switching losses in the IGBTs, which is paid for in continuous current. A drive
set to 8 kHz can be a fifth smaller than the one you ordered.

Work the motor current out with the [three-phase current
calculator](/tools/three-phase-current) if you have kW but no nameplate — and
remember the nameplate wins whenever the two disagree.

## Overload class is not a detail

Every drive has two ratings, and the catalogue leads with the flattering one.

| Rating | Overload | For |
|---|---|---|
| Normal duty / variable torque | 110 % for 60 s | Centrifugal fans and pumps |
| Heavy duty / constant torque | 150 % for 60 s | Conveyors, mixers, compressors, crushers |

The same physical drive is often sold as, say, 22 kW normal duty and 18.5 kW
heavy duty. Specify from the catalogue's headline number for a conveyor and you
have bought a drive one frame too small.

The load's torque curve decides which column you are in:

- **Torque rising with speed²** — centrifugal fans and pumps. Almost no torque
  at standstill, so normal duty is genuinely enough.
- **Constant torque** — conveyors, positive displacement pumps, extruders. Full
  torque from zero. Heavy duty.
- **High breakaway** — crushers, mixers starting loaded, anything that may start
  against a jam. Heavy duty, and consider a frame above it.

## Control mode

Three levels, and paying for the top one when you need the bottom one is
wasteful; the reverse does not work at all.

**V/f (scalar)** holds a fixed volts-per-hertz ratio. Simple, robust, and the
only mode that will drive several motors from one drive. Torque falls away below
about 10 Hz, so it suits fans and pumps and little else.

**Sensorless vector** estimates rotor position from the current waveform and
gives close to full torque down to 1–2 Hz. This is what a conveyor that must
start loaded needs, and it is the sensible default for anything that is not a
fan. One drive, one motor.

**Closed-loop vector** adds an encoder for full torque at zero speed and precise
speed holding. Hoists, positioning, winders. It brings an encoder, its cable and
its failure modes with it — do not specify it unless the application genuinely
needs standstill torque.

## Four things that damage motors, not drives

A drive's output is a train of steep-edged pulses, not a sine wave, and that has
consequences the datasheet does not lead with.

**Reflected wave on long cables.** Voltage doubling at the motor terminals gets
worse with cable length and faster IGBT edges. Beyond roughly 50 m, fit an
output reactor or a dV/dt filter; beyond 100 m, a sine filter is often the
honest answer. Ignoring it punctures winding insulation over months.

**Bearing currents.** Common-mode voltage discharges through the bearing races
and pits them. Above about 100 kW — sooner on 690 V systems — specify an
insulated non-drive-end bearing or a shaft grounding ring.

**Cooling at low speed.** A standard motor is cooled by a fan on its own shaft.
Run it at 20 Hz for hours and it is producing heat with almost no airflow.
Constant-torque duty at low speed needs a force-ventilated motor or a derated
one.

**Insulation class.** An inverter-duty motor has reinforced winding insulation
for exactly these reasons. Standard motors tolerate drives; inverter-duty motors
survive them.

## Harmonics, and the reactor you should have specified

A six-pulse drive — which is nearly all of them below a few hundred kW — draws
current in pulses, not sinusoidally. Typical current distortion is 35–45 % with
nothing fitted.

| Mitigation | Rough current THD |
|---|---|
| Nothing | 35–45 % |
| 3 % line reactor or DC choke | 28–35 % |
| 5 % line reactor | 25–30 % |
| 12-pulse | 10–12 % |
| Active front end | Under 5 % |

A line reactor is cheap, passive and has no failure mode worth worrying about.
Specify one on every drive above a few kW unless there is a reason not to — it
also protects the drive's input rectifier from supply transients, which is worth
the cost on its own.

The reason to care beyond IEEE 519 compliance: harmonic current raises RMS
current in every cable and transformer upstream, and it will resonate with any
capacitor bank on the same supply. The
[harmonic loss calculator](/tools/thd-losses) puts a number on both, and
[APFC panels](/blog/apfc-panel-step-sizing) covers the detuned reactors that
keep a correction bank alive on a drive-heavy site.

## Protection, and the two things that are different

**A Type B residual current device**, not Type AC. A drive's rectifier can
produce leakage with a DC component, and DC leakage saturates a Type AC device
so it stops detecting the AC leakage it was fitted for — silently. The
distinction is set out in [MCB, MCCB, ACB and
RCCB](/blog/mcb-mccb-acb-rccb-difference).

**Short-circuit protection is the drive's, not the motor's.** The drive protects
the motor electronically; the upstream device protects the cable and the drive.
Many manufacturers specify semiconductor fuses for the input, and a plain MCCB
will not protect the rectifier. Follow the drive manual's table rather than
sizing it like a DOL circuit — though the cable itself is still sized the
ordinary way with the [cable sizing calculator](/tools/cable-size).

## What else to settle before ordering

- **Panel heat.** A drive dissipates roughly 3 % of its rating. Three 22 kW
  drives put about 2 kW into an enclosure, which is a cooling calculation, not a
  rounding error — and it derates everything else inside, as in
  [IP ratings for panels](/blog/ip-ratings-for-panels).
- **EMC filter category.** C1 for residential, C2 for most industrial, C3 for
  restricted-access installations. Built-in filters raise earth leakage, which
  interacts with your RCD choice.
- **Braking.** An overhauling load — a hoist, a high-inertia fan coasting down —
  pushes energy back. That needs a braking resistor sized for the duty cycle, or
  a regenerative front end.
- **A bypass contactor** on critical loads, so a failed drive does not stop
  production while it is replaced.
- **Communications.** Most drives speak Modbus RTU over RS485 as standard. If
  they will share a bus with your metering, budget the poll cycle with the
  [RS485 calculator](/tools/rs485-cable) — drives answer slowly.

## The order that avoids the expensive mistake

1. Motor **full load current** from the nameplate.
2. Load torque curve → **normal or heavy duty**.
3. Apply **ambient, altitude and switching frequency** derating.
4. Confirm the derated continuous current still exceeds the motor's FLC.
5. Control mode from the application, not the price list.
6. Cable length → output filtering.
7. Line reactor, essentially always.
8. Type B RCD, and the manufacturer's input protection.

Steps 1 and 3 are the ones people skip, and together they are the reason a drive
that matched on kilowatts trips on a hot afternoon in its second summer.