A waveform held inside a limit band

A compliance claim that leaves out the window, the percentile or the place cannot be tested. This page names the three for each limit. A measurement campaign at the site shows whether it meets them.

Power quality standards do three things. They define the quantities, prescribe how to measure them and set limits at a named point. The window is the period each measurement is averaged over. The percentile says how many of those averaged values must sit under the limit. For the supply standards on this page, a facility is compliant when an instrument at the named point reports that percentile of windowed values under the limit. Equipment standards work the other way: they say what a motor or a tool must tolerate.

This page names the standard behind each limit, what it measures and where. It also shows what a corrector beside the panel contributes under each limit. The corrector supplies the loads’ distorted current itself.

What a limit is

The controller works on the current side of each limit. We designed it to help a site meet IEEE 519-2022 at the point of common coupling. The controller does not provide ride-through 6.

  • Utility connection. Harmonics at the point of common coupling. IEEE 519 reads daily 99th percentile values over 3 seconds and weekly 95th percentile values over 10 minutes 1.
  • Motors. Unbalance at the motor terminals. NEMA MG 1 rates a motor for 1% unbalance and derates it to 75% of rating at 5% 3,9. Derating means the motor is allowed less output than its nameplate says.
  • Data centers. Large computational loads. Ride-through means the equipment keeps running through a brief drop in voltage. ERCOT now requires a large computational load to ride through voltage at 50% of nominal for half a second 17.
  • At the panel. What the controller changes. The current distortion and current unbalance measured upstream of the panel fall. Harmonic orders above the 23rd remain 6.

Each standard reads its limit at its own place. IEEE 519 evaluates harmonics at the point of common coupling (PCC), where the plant meets the utility 1. EN 50160 evaluates supply voltage, harmonics and unbalance at the customer’s terminals, as 95% of the 10-minute values in a week 2. NEMA MG 1 evaluates unbalance at the motor terminals with no window at all, because it rates the motor and not the supply 3.

A limit names three thingsA three by three table setting IEEE 519, EN 50160 and NEMA MG 1 against the window, the percentile and the place each one names.A limit names three thingsThe three standards this page opens with, side by sideWindowPercentilePlaceIEEE 5193 s and 10 mindaily 99th, weekly95thpoint of commoncouplingEN 5016010 min95% of a weekthe customer’sterminalsNEMA MG 1No window at all. It rates the motor, notthe supply.the motor terminalsA limit names three thingsWindow, percentile, placeIEEE 519Window3 s and 10 minPercentiledaily 99th, weekly 95thPlacepoint of common couplingEN 50160Window10 minPercentile95% of a weekPlacethe customer’s terminalsNEMA MG 1WindowNo window at all. It rates themotor, not the supply.Placethe motor terminals
NEMA MG 1 names no window at all, because it rates the motor and not the supply.

A Class A instrument is one that measures the way IEC 61000-4-30 prescribes, so its numbers can be compared between sites 5. A corrector that acts inside a cycle changes the instrument’s shortest readings as well as its 3-second and 10-minute values. A device that acts in seconds changes only the 3-second and 10-minute values.

The limits

IEEE 519-2022 is North America’s harmonic standard. Below 1 kV it limits voltage to 5% per harmonic and 8% total, and current by total demand distortion from 5% to 20% with the short-circuit ratio 1. The 2022 edition is titled a standard rather than a recommended practice, and its values are stated as limits 7.

An illustrative trace, not a capture. It draws three harmonics at 20%, 14% and 9% of the fundamental. The readout combines the three: square each harmonic’s share, add them and take the square root. Correction runs through the 23rd order, so higher orders remain 6. IEEE 519-2022 limits voltage distortion below 1 kV to 8% total and current by total demand distortion through the 50th order 1.

ANSI C84.1 sets service voltage at plus or minus 5% in Range A and asks utilities to hold unbalance to 3% at the meter 8. NEMA MG 1 rates motors for 1% unbalance and derates above it 3,9.

EN 50160 holds supply voltage within plus or minus 10%, total harmonic distortion (THD) under 8% and unbalance under 2% for 95% of a week 2. IEC 61000-2-2 sets compatibility levels in the 2 to 150 kHz band that switching equipment fills 10. MIL-STD-1399 holds shipboard 60 Hz power to 5% THD and 3% unbalance 11.

Every limit names a window, a percentile and a placeEvery limit names a window, a percentile and a place.Every limit names a window, a percentile and a placeIEEE 519-2022, Table 2Current distortion limit, TDD, as a percentage ofmaximum demand currentI_sc / I_L at the PCCTDD limit<205%20 to 508%50 to 10012%100 to 1,00015%>1,00020%IEEE 519-2022, Table 1Voltage distortion limit as a percentage of the nominal busvoltageBus voltageLimit≤ 1 kV8.0%Total THD5.0%Single harmonic1 to 69 kV5.0%3.0%69 to 161 kV2.5%1.5%> 161 kV1.5%1.0%Motor derating against voltage unbalanceDerating in % of rated horsepower. Reproduced from a 2005 motor-protection bulletin, not from the NEMA text.75%85%95%01234598%95%88%82%75%NEMA MG 1:rated loadto 1%Voltage unbalance at motor terminals (%)IEEE 519 values were read from the 2014 text with vendor confirmation for 2022. The unbalance derating is reproducedfrom a 2005 motor-protection bulletin’s reproduction of the NEMA curve, not from the NEMA text. The 1% line is NEMAMG 1.Every limit names a window, a percentile and a placeIEEE 519-2022, Table 2Current distortion limit, TDD, as a percentage of maximum demand currentI_sc / I_L at the PCCTDD limit<205%20 to 508%50 to 10012%100 to 1,00015%>1,00020%IEEE 519-2022, Table 1Voltage distortion limit as a percentage of the nominal bus voltageBus voltageLimit≤ 1 kV8.0%Total THD5.0%Single harmonic1 to 69 kV5.0%3.0%69 to 161 kV2.5%1.5%> 161 kV1.5%1.0%Motor derating against voltage unbalanceDerating in % of rated horsepower. Reproduced from a 2005 motor-protectionbulletin, not from the NEMA text.75%85%95%01234598%95%88%82%75%NEMA MG 1: ratedload to 1%Voltage unbalance at motor terminals (%)IEEE 519 values were read from the 2014 text with vendor confirmation for2022. The unbalance derating is reproduced from a 2005 motor-protectionbulletin’s reproduction of the NEMA curve, not from the NEMA text. The 1%line is NEMA MG 1.
The current and voltage panels are IEEE 519-2022 Tables 2 and 1 1. The third panel is the NEMA MG 1 derating curve as a 2005 motor-protection bulletin reproduces it, with the NEMA 1% line 9,3.

What the equipment must tolerate

Compliance has a second side: what the equipment must survive. The ITI curve gives 120 V equipment 20 ms of dropout tolerance, half a second at 70% voltage and 10 s at 80% 13. SEMI F47 sets 50% for 200 ms, 70% for half a second and 80% for one second 14. IEC 61000-4-11 Class 3 tests dips to 0% for a cycle and 70% for 30 cycles 15. IEEE 1668 adds three-phase points 16.

ERCOT’s ride-through rule is NOGRR282 17. A FERC order of 2026-07-16 (Docket RD26-7-000) directs NERC to file reliability standards for computational load integration by the end of 2026 18.

How the industry has dealt with it

Compliance is a measurement campaign. The commissioning engineer installs a Class A analyzer at the PCC for at least a week. The engineer computes the 3-second and 10-minute percentiles, then compares them with the IEEE 519 tables for the site’s short-circuit ratio 1,5. No published figure says what a missed limit costs a facility.

Mitigation follows the report: a capacitor bank for displacement power factor, a tuned or active filter for orders over the limit and a K-rated transformer for the heat. Several standards the industry still cites by name have left active status, including the IEEE color books and the C62.41 surge standards 19. A campaign report has to name the edition it was measured against.

How Software-Defined Electricity handles it

We designed the controller to meet IEEE 519-2022 at the PCC. It carries a CE conformity assessment under IEC 62477-1 and is installed with a UL field evaluation 6. Those facts cover the product. The measurement campaign is what shows whether the site meets its limits, and we recommend that a third party run it, so the number that says the site complies is theirs.

The controller contributes on the current side. It corrects the current, not the voltage. It supplies harmonic current through the 23rd order, reactive current and unbalanced current locally, in under 5 microseconds. The total demand distortion and the current unbalance measured upstream of the panel fall. The displacement power factor rises toward 1.0 6.

Where the PCC aggregates many panels, the measured voltage distortion moves by one panel’s share: the current that panel no longer draws, times the plant’s upstream impedance. The distortion that arrives from the utility side stays 4. The SEMI F47 and IEC 61000-4-11 test points all lie below 85% of nominal. That is outside the controller’s stated band of plus or minus 15% 6.

Where it matters

A data center now answers to ERCOT and FERC as well as IEEE 519. A defense facility and a ship answer to MIL-STD-1399 at the switchboard. An industrial plant answers to IEEE 519 at the PCC and to NEMA MG 1 at every motor.

The places map to the three directions this site calls Opportunity Pillars. The PCC and the motor terminals are the Consumption side. The meter and the customer’s terminals are the Delivery side. Ride-through for computational loads is where Delivery meets Generation.

What to measure

Name the point, the window, the percentile and the edition on the report.

From the measurement point to the report lineThe three standards the page opens with, each with its place, window and percentile copied from the page’s own sentences. NEMA MG 1 names no window because it rates the motor and not the supply.Each limit is read at its own place, window and percentilePlaceStandardWindowPercentilePoint of common couplingIEEE 5193 s and 10 min99th daily, 95th weeklyCustomer’s terminalsEN 5016010 min95% of a weekMotor terminalsNEMA MG 1no window: it rates the motorA report that names all four can be checked. One that names only the number cannot.Each limit is read at its own place, window andpercentilePoint of common couplingStandard: IEEE 519Window: 3 s and 10 minPercentile: 99th daily, 95th weeklyCustomer’s terminalsStandard: EN 50160Window: 10 minPercentile: 95% of a weekMotor terminalsStandard: NEMA MG 1Window: no window: it rates the motor(it rates the motor, not the supply)A report that names all four can be checked. One that names only thenumber cannot.
The three standards this page opens with, each with the place, window and percentile the page gives for it. A report that names all four can be checked.

Keep the corrector within its rating through the peak hour. The weekly 95th percentile is set by the worst 5% of the week’s 10-minute values, and those fall in the peak hours. A device that stays within rating through the peak lowers those values and so lowers the percentile. A device that runs out of rating during the peak improves only the off-peak values, and those were already under the 95th percentile, so the percentile stays where the peak left it 4.

Start by logging one week at the PCC before claiming a number.

Sources
  1. IEEE 519-2022, clauses 3 and 5, Tables 1 and 2.
  2. EN 50160:2022+A1:2025.
  3. NEMA MG 1-2024, clauses 12.45, 14.36.2 and 30.1.2.
  4. Rule review for this series: the averaging-window rule, the two-share rule, window and rating. First-party derivation.
  5. IEC 61000-4-30:2025, clauses 4.4, 4.5 and 5.4.2.
  6. 3DFS, VectorQ Technical Overview (current revision), Technical Specifications table and p. 14. First-party.
  7. Trade summaries of the IEEE 519-2022 changes.
  8. ANSI C84.1-2020 (R2025), Table 1 and Annex C.
  9. A leading maker’s fuse and motor protection guide, Motor Protection, Voltage Unbalance and Single-Phasing (2005), derating table.
  10. IEC 61000-2-2:2002+AMD1:2017+AMD2:2018.
  11. MIL-STD-1399 Section 300 Part 1 (2018), Table II.
  12. 3DFS, electrical network analysis of server racks at a data center, p. 25. First-party.
  13. Information Technology Industry Council, ITI (CBEMA) Curve Application Note (2000).
  14. SEMI F47-0706 (Reapproved 0812), Table 1.
  15. IEC 61000-4-11:2020, Table 1, Class 3.
  16. IEEE 1668-2017, three-phase points as reproduced in the EPRI seminar of 2018.
  17. ERCOT, NOGRR282, Large Computational Load Ride-Through Requirements, effective 2026-08-01.
  18. POWER magazine, FERC orders mandatory NERC reliability standards for data center and other computational loads (2026-07-16).
  19. IEEE standards status pages: IEEE 141, 241, 446 and 1100 Inactive-Reserved since 2021-03-25. IEEE C62.41.1 and C62.41.2 Inactive-Reserved.

Version 5.1