Most power-quality conversations are about which technology to install — a capacitor bank, a passive filter, an active filter, a static VAR compensator. That is the right debate to have, and we have written about it elsewhere. But it quietly assumes the question of where the device goes has only one answer: at the main incomer, next to the utility meter, where the facility connects to the grid.

It is an understandable instinct. The meter is where the bill is calculated, where the utility's power-factor penalty is assessed, and where a compliance engineer measures distortion against IEEE 519. Fix the numbers at that point, the reasoning goes, and the problem is solved. The meter is the scoreboard, so correct at the scoreboard.

The instinct is wrong, and it is wrong for a reason rooted in basic circuit physics rather than in any particular product. A shunt correction device — whatever its technology — only cleans the current flowing upstream of the point where it connects. Everything downstream of that point continues to carry the full, uncorrected current. Place the device at the meter, and "downstream" is your entire internal distribution network: every metre of cable, every distribution board, every step-down transformer, and every motor in the building. That is where most of the copper is, most of the heat is, and most of the wasted money is.

The core mechanism
A correction device cleans the current on the conductors between itself and the source. It does nothing for the conductors between itself and the loads. Connect it at the utility meter and you have cleaned the shortest, coolest, best-sized run in the whole facility — the few metres of service entrance — while leaving the long, hot, distributed internal network exactly as it was.

1. The meter is a boundary, not the network

The point where a facility connects to the utility has a formal name: the point of common coupling (PCC). It is the boundary between two domains of responsibility. On the utility side, the grid operator cares about what the facility injects back into the shared network — harmonic currents, reactive demand, flicker — because those affect every other customer on the same feeder. IEEE 519-2022 sets its distortion limits precisely at this boundary, for exactly this reason: it is a standard about being a good neighbour on a shared grid.

This matters, because it explains why "correct at the meter" feels authoritative. The standards point there. The penalty is assessed there. The compliance measurement is taken there. But the PCC is a boundary of billing and responsibility, not a representation of where your energy is actually being lost. The losses do not occur at a boundary. They occur along conductors — distributed across the hundreds or thousands of metres of cable that fan out from the incomer to every load in the building.

Correcting at the PCC satisfies the utility and cleans the bill. It does not, and cannot, clean the network on your side of the line.

2. Where the losses actually happen

The energy wasted by poor power quality is overwhelmingly I²R loss — resistive heating in conductors, windings, and connections, proportional to the square of the current flowing through them. Reactive current and harmonic current both inflate the RMS current a conductor carries without delivering any useful work, and the conductor heats up for the privilege.

The critical word is distributed. Those losses are not concentrated at one point you can stand next to. They are spread along every conductor that the excess current traverses, in proportion to that conductor's resistance and the square of the current it carries. A typical industrial site has a loss profile that looks roughly like this:

The geometry is the whole point. The service entrance — the only part a meter-point device cleans — is the shortest, coolest, lowest-resistance run in the facility. The internal network it leaves untouched is where the conductors are longest, the temperatures highest, and the cumulative resistance greatest. A device at the meter cleans the part of the network that was already losing the least.

A useful analogy
Correcting power quality at the meter is like filtering the water at the point where the mains enters the building — after the water has already passed through every pipe, tank, and fixture in the house. The water leaving your property is clean. Every pipe inside it carried the unfiltered water the whole way. To protect the pipes, you filter at the source of the contamination, not at the exit.

3. The single-point correction trap

Picture a facility that has done everything "right" by the conventional playbook. A high-quality active filter sits at the main switchboard, immediately downstream of the meter. The compliance report is clean: voltage distortion at the PCC is under 5%, current distortion is within the IEEE 519 envelope, and displacement power factor reads 0.99. On paper, this is a solved problem.

Now follow a single harmonic current produced by a variable-speed drive deep in the plant. It is born at the drive's input rectifier. It flows up the final circuit to the motor control centre. It flows up the sub-main to the distribution board. It flows up the main feeder to the switchboard — and only there, at the switchboard, does the active filter finally cancel it so that nothing distorted continues up to the meter.

Every conductor that harmonic current travelled on the way to the filter carried it in full. The drive's final circuit, the MCC busbar, the sub-main, the main feeder, and any transformer in that path were all heated by it. The filter cancelled the current at the last possible moment before the boundary — protecting the utility, satisfying the standard, and doing essentially nothing for the kilometres of internal cable and the dozens of loads that the current passed through to reach it.

This is the single-point correction trap: a measurement at the meter cannot distinguish between a network that is genuinely clean and a network that is filthy right up to the point where the filter sits. Both produce an identical, compliant reading at the PCC. The bill looks the same. The internal network does not behave the same at all.

4. Centralised correction versus correction at the load

There are, broadly, two philosophies for where correction belongs.

Centralised correction

A single large device at the main switchboard or incomer, sized for the whole facility. Its appeal is commercial and logistical: one unit to buy, one location to install, one point to maintain, one compliance measurement to take. Its limitation is the one this article is about — it cleans only the current upstream of itself, leaving the internal network downstream uncorrected. It is the right architecture when the dominant concern genuinely is the boundary: meeting an IEEE 519 connection condition, removing a utility power-factor penalty, or qualifying for grid connection.

Correction at the load

Correction placed at or close to the loads that create the distortion — at the motor control centre, the drive panel, or the individual machine. Here the dirty current is cancelled before it ever enters the internal distribution network. Every conductor upstream of the load, all the way back to the meter, then carries clean current and runs cooler. The whole network benefits, not just the service entrance.

The trade-off is real and worth stating plainly: load-level correction means more correction points and a more considered design than dropping a single box at the incomer. The engineering question is not "which philosophy is purer" but "where is the distortion produced, and how much of the network sits between that source and the meter?" In a modern plant, where dozens of non-linear loads are scattered across the site, the answer is usually "a great deal" — which is exactly why the meter is the wrong place to make a stand.

The principle in one line
Correct as close as practical to the source of the distortion. The closer the correction is to where the dirty current is born, the more of the network is shielded from it — and the more of the distributed I²R loss you actually recover.

5. The grid-capacity dimension

Wasted energy is the loss everyone thinks of first, but meter-point correction misses a second, less obvious benefit entirely: recovered network capacity.

A cable, a busbar, and a transformer are all rated by the RMS current they can carry before overheating. Reactive and harmonic current consume that rating without doing useful work. A feeder running at 600 A of which 90 A is reactive-and-harmonic "passenger" current has only 510 A of its rating left for real load. Cancel that passenger current and the feeder can carry more productive load within the same thermal limit — capacity unlocked without pulling a single new cable. We explore this in detail in a dedicated article on grid capacity.

Where you correct decides which conductors get that capacity back. Correct at the meter, and only the service entrance — already the least-loaded run — is relieved. The internal feeders and transformers, which are usually the parts running closest to their thermal limit and the parts you would actually like to free up before adding a new production line, see no relief at all. Correct at the load, and every conductor between the load and the meter is unburdened, releasing capacity exactly where a growing facility tends to run out of it first.

For a site contemplating electrification, a new line, or an EV-charging build-out, this is frequently the more valuable half of the case. The energy saving shows up on the bill; the capacity headroom is what lets the expansion happen at all without a costly upgrade to the internal infrastructure.

6. What correcting in the right place requires

Correcting where the distortion is produced — rather than only where it is billed — is a system-design problem, not a single-box purchase. It needs three things working together:

HarmoniQ's three-component architecture is built around this principle of place, not only technology. The Filter (active harmonic cancellation), Alpha (impedance matching and line conditioning), and Booster (solid-state true power factor correction) are deployed against the actual topology of the site — positioned to shield the internal network, not merely to produce a compliant reading at the boundary. The components and the physics behind them are detailed in the product documentation.

7. How you would even know

The reason meter-point correction persists as the default is that the meter is the one place almost every facility already measures. A single revenue meter at the incomer is universal; instrumentation distributed through the internal network is rare. So the only data most operators have describes the boundary — and a boundary measurement is structurally incapable of revealing a network that is dirty right up to the correction point.

Seeing the real picture requires measuring inside the network, not just at its edge: a power analyser placed at the motor control centres, the sub-distribution boards, and the larger individual drives, not solely at the main incomer. That is what reveals where the distortion is actually produced, which conductors are carrying the passenger current, and how much loss and capacity a correction point would recover at each candidate location. Without it, "correct at the meter" is not a decision — it is the absence of one, made by default because the meter was the only thing anyone looked at.

The rule of thumb
If a facility's distortion is produced by loads scattered across the site — which describes essentially every modern industrial plant — then a single correction device at the meter leaves the majority of the internal network carrying dirty current. The bill and the compliance report can look excellent while the cables, transformers, and motors inside the fence go on wasting energy and surrendering capacity. The technology was never the limiting factor. The location was.

Summary

The power-quality debate is usually framed as a choice of technology. It is at least as much a choice of place. A shunt correction device cleans only the current upstream of where it connects, and the losses it is meant to recover are distributed along every conductor the dirty current travels. Connect the device at the utility meter and you clean the shortest, coolest run in the building — the service entrance — while the long, hot, distributed internal network keeps carrying exactly the current it always did.

The meter is a boundary of billing and regulatory responsibility, and correcting there has a legitimate job: satisfying the utility and removing a penalty. But it is not where your energy is lost and it is not where your capacity is consumed. Those happen inside the network, on the conductors between the loads and the meter — which is precisely the region a meter-point device cannot reach.

Correct as close as practical to where the distortion is produced, and the whole network runs cleaner, cooler, and with more headroom — not just the few metres the utility happens to measure. Getting there starts with measuring inside the network, not only at its edge, so the correction is placed where the physics says it will actually pay.