---
title: "APFC panels: sizing the steps, and avoiding resonance"
description: "How an automatic power factor panel decides what to switch, why the smallest step matters more than the total kVAr, and when you must fit detuned reactors."
date: "2026-05-16"
author: "Divakar B"
source: "https://energycalchq.com/blog/apfc-panel-step-sizing"
license: "© 2026 Divakar B. Quote with attribution to https://energycalchq.com/blog/apfc-panel-step-sizing"
---

An automatic power factor correction panel is a bank of capacitors, a set of
contactors, and a controller that decides how many to switch in. The concept is
simple enough that it gets specified by total kVAr and nothing else — and that
is how plants end up with a correctly sized bank that never reaches its target,
or worse, one that makes the harmonic situation dramatically worse.

This is what actually needs deciding: the step arrangement, the switching
device, and whether you need detuned reactors.

## Why fixed compensation is not enough

A fixed capacitor bank supplies the same kVAr regardless of what the plant is
doing. Size it for full production and it overcorrects badly at night, pushing
the power factor leading — penalised on many tariffs, and a cause of voltage
rise at light load. Size it for the night load and it does almost nothing during
the day.

Fixed compensation is right in exactly one case: a base load that genuinely
never goes away. A transformer's magnetising current, or a large motor that runs
continuously, can be fixed-compensated at the load. Everything that varies needs
switching.

![Chart of reactive demand over a day with a five-step capacitor bank following it](/blog/apfc-steps-following-load.svg "The controller always switches to the step below demand. The gap it cannot close is one step wide — which is what makes the smallest step the important number.")

## The smallest step decides your result

Everyone specifies the total kVAr. Work it out from your load and target with
the [power conversion calculator](/tools/power-conversion) and it is
straightforward. The number that decides whether you *reach* the target is the
smallest step.

The controller can only switch whole steps, and it will only switch to a level
at or below the reactive demand — going above means overcorrecting. So the
residual reactive power is always somewhere between zero and one step. A 100 kVAr
bank in one step is nearly useless; the same 100 kVAr in eight steps of 12.5
tracks the load closely all day.

Two common arrangements:

| Scheme | Example | Steps available | Notes |
|---|---|---|---|
| Equal | 8 × 12.5 kVAr | 12.5 kVAr resolution | Simple, even contactor wear, most common |
| Binary | 12.5 : 12.5 : 25 : 50 | 12.5 kVAr resolution from fewer contactors | Fewer devices, but the large steps switch often and wear unevenly |

Equal steps are the usual choice in India, and the controller can rotate which
step it switches first to even out contactor wear. Binary schemes save
contactors at the cost of switching the big ones more.

A practical starting point: **make the smallest step about 5–10 % of the total
bank**, and never larger than the smallest meaningful load change on the plant.

## The C/k setting, and the commissioning mistake

Every APFC controller needs to know how much current one step represents, seen
through the incomer CT. That is the C/k setting:

```
C/k = (step kVAr × 1000) / (√3 × line voltage × CT ratio)
```

Set it too high and the controller cannot see the effect of its own smallest
step, so it hunts — switching in and out repeatedly, wearing out contactors and
capacitors. Set it too low and it becomes over-sensitive and does the same.

Hunting is the single most common APFC complaint, and it is almost always C/k,
a CT on the wrong circuit, or a CT with reversed polarity. Before replacing
anything, check that the CT sees **all** the load *and* the capacitors — a CT
positioned so it does not see the capacitor current cannot possibly control it.

## Switching devices

**Capacitor duty contactors** are not ordinary contactors. Energising a
capacitor draws a huge inrush — 30 to 200 times rated current for a few
milliseconds — so capacitor duty contactors have early-make auxiliary contacts
with damping resistors that pre-charge the capacitor before the main contacts
close. Fitting an ordinary AC-3 contactor instead welds the contacts within
months.

**Thyristor switching** replaces contactors where the load changes fast — spot
welding, lifts, crushers, presses. A contactor-based panel needs a discharge
delay of 30 to 60 seconds before a step can be re-energised; a thyristor module
switches at the zero crossing in milliseconds, with no inrush and no wear. It
costs several times more and needs cooling, and it is the right answer whenever
the load changes faster than the bank can follow.

## Harmonics: where APFC panels go badly wrong

This is the part that turns a saving into a failure.

Capacitors are a low impedance to high frequencies. Put a capacitor bank on a
supply that also has inductance — every transformer is an inductance — and there
is a frequency at which the two resonate. If a harmonic current the plant
generates happens to sit near that frequency, the resonant circuit amplifies it.

Symptoms: capacitors running hot, fuses blowing for no visible reason,
capacitors failing repeatedly, nuisance tripping, and a measured harmonic
distortion that is worse *after* the correction was installed than before.

The cure is a **detuned reactor** in series with each capacitor step. The reactor
shifts the resonant frequency below the lowest significant harmonic, so the
combination is inductive at all harmonic frequencies and can no longer amplify
anything.

| Detuning | Tuned near | Use when |
|---|---|---|
| 7 % | 189 Hz | The usual choice — VFDs, rectifiers, 5th harmonic dominant |
| 14 % | 134 Hz | Significant 3rd harmonic — heavy single-phase and LED load |
| None | — | Only where non-linear load is genuinely negligible |

A working rule: **if non-linear load exceeds about 20 % of the plant, specify
detuned reactors.** In 2026 that describes most plants. Reactors add cost and
they raise the voltage across the capacitors, so the capacitors must be rated
for it — a 440 V capacitor in a 7 % detuned step on a 415 V system is normal,
and fitting standard 415 V capacitors there will fail them.

If distortion is severe, correction is the wrong tool entirely and you need an
active harmonic filter. Measure before you decide.

## Things that make a panel last

- **Discharge resistors** on every capacitor, sized to bring the terminals below
  50 V within a minute. Without them a switched-out capacitor stays charged and
  the next switching operation happens onto a live capacitor.
- **Ventilation.** Capacitor life halves for roughly every 7–10 °C above rated
  temperature. An APFC panel in a hot corner with no ventilation is a panel that
  will be rebuilt in three years.
- **HRC fuses** on each step, rated for capacitor duty.
- **Annual capacitance check.** Capacitors lose capacitance as they age. A step
  that has degraded 20 % is still switching, still drawing current, and no
  longer delivering what the controller thinks it is. Measuring each step
  annually is a ten-minute job that explains a drifting power factor.

## A specification that will not embarrass you

1. Total kVAr from the measured load and a target of 0.95–0.98, not unity.
2. Smallest step at 5–10 % of the total.
3. Capacitor duty contactors, or thyristor modules if the load cycles fast.
4. 7 % detuned reactors unless you have measured the harmonics and can justify
   leaving them out.
5. Capacitors rated for the voltage *with* the reactors fitted.
6. Discharge resistors, HRC fuses, and forced ventilation.
7. The controller CT on the incomer, seeing both load and capacitors, polarity
   verified at commissioning.

Work out the kVAr first with the [power factor
calculator](/tools/power-conversion), and read [what the penalty actually
costs](/blog/power-factor-penalty-what-it-costs) if you still need to justify
the spend — the demand-charge saving is usually larger than the penalty it
removes.