Short answer: A harmonic filter is selected when a measurement indicates a problem: if the total demand distortion (TDD) at the point of common coupling exceeds the limit, if the total harmonic distortion (THD) in the voltage exceeds or approaches the EN 50160 limit, if capacitors fail due to resonance, or if the neutral and transformer overheat. The next steps are reduction at the source, detuned compensation if capacitors are present, and, if necessary, a tuned passive or active filter. “No filter required” is also a valid measurement result.

Filter proposals often come after a capacitor has blown. However, a filter is a design decision based on the grid impedance, harmonic spectrum, and load profile; if chosen incorrectly, it can create a new resonance or push the facility into capacitive territory at night.

Service note: AES Innovation performs the measurements and evaluations described on this page with its own engineers. Energy analysis, harmonic measurement, and compensation evaluation are outside the scope of our TÜRKAK accreditation; our accreditation scope includes periodic inspections of grounding, leakage current, switchboards, and thermography in low-voltage installations.

In what situations is a filter justified?

The solution to the problem is a measurable one, justifying the filter investment. We cite four reasons in our assessments:

  • The current TDD at the point of common coupling (PCC) is above the limit. For systems up to 69 kV, IEEE 519-2014 sets this limit between 5% and 20% depending on the ratio of short-circuit current to maximum demand current (Isc/IL); the weekly 95th percentile of 10-minute values is used (IEEE 519-2014, clause 5.2 and Table 2).
  • The voltage THD is high. For low voltage, EN 50160 sets limits for individual harmonics that apply to 95% of the weekly 10-minute values and, in addition, requires the voltage THD (up to the 40th order) not to exceed 8% (EN 50160:2022+A1:2025, clause 4.2.5).
  • There are signs of resonance: recurring capacitor and fuse failures, voltage distortion that increases when the stage is activated, and capacitor current above the rated value.
  • The neutral conductor or transformer is heating up for no apparent reason, and measurements indicate this is due to harmonic currents.

The limitations in IEEE 519 apply to the common connection point, not to a single device or feeder within the facility (IEEE 519-2014, clause 5); in industrial installations fed by their own transformer, this point is usually the MV (medium voltage) side of the transformer (IEEE 519-2014, clause 1.2). High current THD in a single drive board is therefore not a justification for filtering alone. If none of the four conditions are met, the conclusion is ’no filtering required“; continuing monitoring is sufficient.

Relating to: What are Harmonics? THD, TDD, and Limit Values

Order of decision: from source to filter

Even if the measurement shows problems, the first step shouldn't be a central filter; in practice, the following sequence is usually followed:

  • Measurement and diagnosis.
  • Reduction at the source: AC line reactor or DC choke coil at the drive input, 12 or 18 pulse rectifier, active front-end (AFE) drive.
  • If a capacitor is present, then reactor-based (detuned) compensation is used.
  • A passive filter tuned to specific harmonics.
  • Active harmonic filter or hybrid solution.

Reducing harmonics at the source is particularly inexpensive when purchasing a new drive. The characteristic harmonics of a rectifier are found using the relationship h = m × (number of pulses) ± 1; in a six-pulse circuit, these are 5, 7, 11, 13, and so on (IEC 61642:1997, clause 1.3.3). Using the same relationship, in a 12-pulse setup, the first harmonics shift to 11 and 13, and in an 18-pulse setup, to 17 and 19; in ideal operation, the dominant 5th and 7th harmonics disappear, but in reality, some remain due to imbalances (IEC 61642:1997, clause 1.3.4). Line reactors and DC choke coils reduce the amplitude but do not change the magnitudes.

Capacitor bank and resonance

Compensation capacitors form a resonant circuit in parallel with the transformer and overhead grid inductance; near the resonance frequency, a small harmonic current transforms into a large voltage distortion (IEC 61642:1997, clause 2.1). The primary guideline on this issue is the IEC 61642, an older text with its first edition from 1997; however, IEC 61921:2017 refers to it for harmonics (IEC 61921:2017, clause 6.3.1.1). In the standard's calculation example, with a 1,000 kVA transformer and a 160 kVAr capacitor, the capacitor current reaches 1.45 times its rated value, while the usual limit is 1.3 times; the voltage distortion at the load busbar reaches approximately 8% due to a single harmonic (IEC 61642:1997, clause 2.4).

There is a known approximation for the degree of resonance: h ≈ √(Sk / Qc). Sk is the short-circuit power at the point where the capacitor is connected, and Qc is the capacitor power in the circuit (IEC 61921:2017, Appendix B.2). Hypothetical example: A 1,600 kVA transformer with a short-circuit voltage (uk) of 6% will provide approximately 1,600 / 0.06 ≈ 26.7 MVA of short-circuit power if the upper grid is ignored. In an automated switchboard, the degree of resonance shifts as the stages are activated:

The capacitor in the circuitApproximate resonance levelFrequency (50 Hz)
200 kVAr11,5≈ 577 Hz
400 kVAr8,2≈ 408 Hz
600 kVAr6,7≈ 333 Hz
800 kVAr5,8≈ 289 Hz

As capacitor power increases, resonance approaches the dominant 7th and 5th harmonics of six-pulse drives; it decreases slightly when the upper grid impedance is added (approximately 5.5 at 800 kVAr with a 250 MVA assumption on the MV side). In the calculation, the total of all capacitors on the LV (low voltage) side of the same transformer is taken (IEC 61642:1997, clause 3.2). This approach is a risk assessment, not a design calculation.

Reactor-based (detuned) compensation

The most common way to prevent resonance is to connect a reactor in series with each capacitor stage. The stage is tuned to a frequency below the lowest significant harmonic; above this, it behaves inductively and cannot resonate with the grid at harmonic frequencies (IEC 61642:1997, clauses 2.5 and 3.3). The reactor is defined by the ratio of inductive reactance to capacitive reactance (p = XL / XC); the series resonance frequency becomes fr = f1 / √p, where f1 is the grid frequency (IEC 61642:1997, clause 2.5). Three common values for a 50 Hz grid are:

Reactor ratio pSeries resonance frequencyHarmonic degree
5.67%≈ 210 Hz4,2
7%≈ 189 Hz3,78
14%≈ 134 Hz2,67

The standard assumes that the first harmonic with significant amplitude is the 5th harmonic in most networks, and in this case, it recommends keeping the tuning below 5·f1, i.e., choosing p above 4%. If the phase-to-phase 3rd harmonic voltage is strong (examples given in the standard are single-phase rectifiers and over-excited transformers), the tuning should be below 3·f1, i.e., p > 11% (IEC 61642:1997, clause 2.5). In practice, 7% is common in industrial installations where the 5th harmonic is dominant: 189 Hz allows for ample margin compared to 250 Hz, tolerates upward shift in tuning as the capacitor ages, and does not strain the tap because it draws less 5th harmonic. In installations where the phase-to-phase 3rd harmonic is prominent, 14% is preferred, which satisfies this condition. In the standard example, adding a 7% reactor reduces the voltage distortion at the load bus from 9.8% to 3.8% (IEC 61642:1997, clauses 2.4 and 2.5.2); the improvement comes from the elimination of resonance, not from the elimination of harmonics.

A series reactor increases the capacitor voltage in the base component to approximately U / (1 − p): approximately 430 V at 7% and approximately 465 V at 14% on a 400 V mains. Therefore, the capacitor's rated voltage is chosen above the mains voltage (IEC 61921:2017, clause 6.2.1). It is also generally not possible to add a reactor to existing reactorless capacitors; it is likely that the existing capacitors were not selected to withstand the additional voltage and current that the reactor would introduce. Reactor and reactorless stages are not mixed in the same system (IEC 61642:1997, clause 3.3).

In new internal electrical installation projects, this calculation is mandatory: the regulation requires that, in addition to the compensation calculation in the final project, the total load current calculation for harmonic loads, the resonance frequency calculation, and the determination of the reactor/capacitor ratio (p in practice) must be included (Regulation on the Preparation of Electrical Internal Installation Projects, Article 10/c-4). In existing installations, the same questions are answered by measurement.

Relating to: Compensation Systems and Reactive Power Control

Adjustable passive filter

The tuning frequency of a tuned filter differs from the harmonic to be filtered by a maximum of 10%; in reactor compensation, the tuning is more than 10% below the lowest significant harmonic (IEC 61642:1997, clauses 1.3.10 and 1.3.11). Since the filter's impedance is lower than that of the grid at the target harmonic, most of the current of that harmonic flows to the filter; at the fundamental frequency, the filter provides reactive power like a capacitor. The branches are usually tuned to degrees 5, 7, 11, 13, slightly below the harmonic (IEC 61642:1997, clause 3.4).

In large, continuously operating six-pulse loads, it performs two functions at a low cost per kVAr. Its disadvantages also stem from this characteristic: because its capacitive power remains constant while it is in operation, the system may switch to the capacitive region at night or on weekends. The filter also acts as a path for harmonics from the MV grid; transformer impedance, source spectrum, and harmonic voltages on the MV side must be known during design (IEC 61642:1997, clause 3.4). Tolerances, temperature, and aging shift the adjustment; as capacitance decreases, the adjustment frequency approaches the harmonic, and the branch draws more current. The standard requires control of these effects, finding a tolerance of ±3% in reactor inductance acceptable for most filters (IEC 61642:1997, clauses 3.5 and 3.5.2).

Active harmonic filter

An active harmonic filter is a power electronics-based inverter connected in parallel to the load. It measures the current using current transformers on the load or grid side, calculates the harmonic components, and feeds their opposites back to the grid; the source side sees a current close to a sine wave. Because it continuously monitors load variations, it is more flexible than a passive filter in variable loads, lines with numerous small drives, and buildings with a high computing load using UPS (uninterruptible power supply).

It also has limitations. The cost per ampere is high, and it has its own losses and cooling requirements. It compensates for incorrect current if the location or orientation of the current transformers is wrong. It switches to limiting mode when the rated current is reached, and the excess remains in the grid. Most importantly, it does not change the grid impedance: it does not eliminate the resonance created by a reactorless bank at another point; that bank is corrected first. IEC 61642 covers only passive filters and shunt capacitors (IEC 61642:1997, clause 1.1); there is no guidance for active filters in that guideline, the selection is based on measurement data.

In a hybrid solution with reactive power reactor compensation, the remaining harmonic is absorbed by a smaller active filter; this is generally more economical than a single large active filter.

Shunt reactor: not a harmonic filter.

In installations that supply capacitive reactive energy at low load because of LED drivers, UPS input filters and long cables, a shunt reactor offsets this surplus by drawing inductive reactive power; it does not filter harmonic currents. However, it should be considered if a capacitive ratio problem is observed in the bill or measurement. Beforehand, it is checked whether the compensation relay is disengaging overnight; sometimes the problem is simply the relay setting.

Comparison of solutions

SolutionWhenAdvantageAttention
No filter, no tracking.Values are below the limit, no signs of resonance.No investment required.The measurement is repeated when a new load is added.
Reduction at the sourceNew drive purchases; a few large sourcesIt solves the problem on the spot.Limited for existing drives; AFE is expensive
Reactor compensationCompensation is necessary; there are harmonics in the grid.It prevents resonance and counteracts reactive power.It is not an adjustable filter; p is selected according to the spectrum.
Adjustable passive filterLarge, continuous, spectrum-defined charge.Low cost per kVArCapacitive persistence at low load, tuning drift.
Active harmonic filterVariable, distributed loads, buildingsIt tracks the changing spectrum, it doesn't add constant capacitive power.Cost, losses, current transformer placement.
HybridReactive power and variable harmonics together.Balance between cost and performance.Control coordination is required.
Shunt reactorCapacitive ratio problem at low loadIt balances capacitive energy.Harmonics do not resolve

Data required before applying the filter.

Things to have ready before requesting a quote:

  • Power quality records for at least one week at the point of common coupling: voltage and current harmonics, 10-minute values and weekly percentages (IEEE 519-2014, clause 4.4).
  • Transformer power (kVA) and short-circuit voltage (uk); medium voltage (MV) short-circuit power from the distribution company, if available.
  • The breakdown and fault history of the existing compensation panel; in panels conforming to IEC 61921, the reactor value, ratio, or adjustment frequency is given on the label or in the instruction document (IEC 61921:2017, clause 4).
  • Harmonic spectrum for common connection points and large feeders (drive boards, UPS).
  • Night, weekend, and full load profile; average of maximum demand flows for TDD over the last twelve months (IEEE 519-2014, clause 5.2).
  • Operation with generator, emergency procedures and planned load increases (IEC 61642:1997, clause 3.1).
  • Inductive and capacitive reactive power ratios in the latest bills.

Inquire whether the reactor stage power in the proposals is the net value including the reactors; IEC 61921 defines the rated power including the reactors (IEC 61921:2017, clause 3.5).

Frequently Asked Questions

Active filter or passive filter?

Spectrum and load profile determine the solution. For continuously operating, large loads with pronounced 5th and 7th harmonics and reactive power requirements, reactor compensation or tunable passive filters are often sufficient and economical. For variable, distributed, or building-type loads, active filters may be more suitable. The decision should be based on measurement, not a catalog.

Does reactor-based compensation eliminate harmonics?

No. Its primary function is to prevent the bank from resonating with the grid and amplifying harmonics. It also absorbs some harmonic currents above the tuning frequency (IEC 61642:1997, clause 3.3), so some improvement is seen; however, it is not a filter designed for a specific TDD (Total Distribution Demand) target. If the limit is still exceeded, the filter is a separate decision.

How is harmonic filter size determined?

In active filters, the criterion is the amperage of the harmonic current to be suppressed at the connection point; a margin is added to the highest measured values for the planned load increase, transformer power is not the criterion. In passive filters and reactor compensation, the required kVAr and the harmonic current that the branches will carry are calculated together; the components are selected according to the worst-case operating conditions, tolerances, and harmonics from the grid (IEC 61642:1997, clause 3.5).

Is there a risk of reactive power penalty if I install a filter?

It's possible. An adjustable passive filter provides constant capacitive power; when the load decreases, the facility can switch to the capacitive zone. The Electricity Grid Regulation limits the ratio of inductive reactive energy to active energy to 20% and the capacitive ratio to 15% for consumers directly connected to transmission and distribution license holders (Electricity Grid Regulation, Article 14). The rate and cost applied in distribution-connected facilities are determined in the agreement with the distribution company and in the tariff decisions of the Energy Market Regulatory Authority (EPDK); in practice, the same 20%/15% ratios are common, and your bill and agreement are binding. Therefore, the filter design should also be tested under a low-load scenario.

Sources

  • Regulation on the Preparation of Electrical Interior Installation Projects, Compensation facility content in the final project (Article 10) — Official Gazette, 3.12.2003, No. 25305
  • Electricity Grid Regulations, Compensation of reactive energy (Article 14) — Official Gazette, May 28, 2014, No. 29013
  • IEC 61642:1997 (first edition), Industrial ac networks affected by harmonics – Application of filters and shunt capacitors
  • IEC 61921:2017, Power capacitors – Low-voltage power correction factor banks
  • IEEE Std 519-2014, IEEE Recommended Practice and Requirements for Harmonic Control in Electric Power Systems
  • EN 50160:2022+A1:2025, Voltage characteristics of electricity supplied by general electricity grids.

Relating to: What is Energy Analysis and How is it Performed?

Relating to: Reactive Energy Penalty and Compensation

Relating to: How many kVAr capacitors are needed for compensation?

Relating to: Power Quality Measurements

You can contact us to schedule measurements at your facility before investing in a filter or reactor-based compensation system. For a sound decision, we need transformer and compensation panel information with at least one week of recordings at the common connection point.