Shared Neutrals Across Transformers: Why Phase Balancing Is Not Enough

While collecting data to design a new feeder for a service building (read the earlier main feeder diagnosis), the maintenance technician told me that no matter how he redistributed the three-phase loads, he could not balance them. Phase A was always higher than the others. Stranger still, on days off when nobody was working in the building, phase A remained 30-40 A higher than the other two phases. I asked to trace the entire upstream system and found what had made phase balancing impossible from the start. In this article, I cover the tracing process, the measurement traps, and the repair sequence that must be followed in order.

The symptom: phase A stays 30-40 A higher when there should be no load

There were two initial symptoms. First, the loads would not balance: no matter how circuits were moved between phases in the panel, phase A remained higher than the others. Second, on days off when every department was closed, phase A still remained 30-40 A above the other two phases.

The second symptom was the main clue. Loads that disappeared on days off belonged to the building, while loads that did not disappear on days off suggested they were not building loads. With that hypothesis in mind, I traced the actual cable route and found several houses connected to the building's main feeder, even though the houses already had their own supply line from another transformer.

The actual configuration: two systems that should be separate but are not

The intended configuration was for each of the two transformers to supply its own system. The new transformer supplied the service building, while the old transformer supplied the flats and houses. But tracing the actual installation revealed two points that joined the two systems into one.

This comes down to two points. First, the neutrals of the two transformers were interconnected. Second, several houses shared the building's main feeder neutral, and some went further: both their phase and neutral were connected directly to the building's main feeder. Their own supply line was running right overhead; it simply was not being used.

Why interconnected neutrals prevent phase balancing

In a balanced three-phase, four-wire system, neutral current approaches zero because the three phase currents cancel out as vectors. The greater the imbalance, the higher the neutral current. The approximation formula is as follows.

import math

def neutral_current(ia, ib, ic):
    """กระแสนิวทรัลของระบบ 3 เฟส 4 สาย กรณีโหลดไม่สมดุล (ประมาณ)"""
    return math.sqrt(ia*ia + ib*ib + ic*ic - ia*ib - ib*ic - ic*ia)

print(round(neutral_current(100, 100, 100), 1))   # สมดุล  -> 0.0
print(round(neutral_current(140, 100, 100), 1))   # A เกิน 40 A
print(round(neutral_current(292, 100, 103), 1))   # ถ้าเฟส A ต่างเกือบ 200 A

ผลที่ได้จากสคริปต์ข้างบน

  100 / 100 / 100   ->    0.0 A     สมดุล นิวทรัลแทบไม่มีกระแส
  140 / 100 / 100   ->   40.0 A     ต่างกัน 40 A นิวทรัลรับ 40 A
  292 / 100 / 103   ->  190.5 A     ต่างเกือบ 200 A นิวทรัลรับหนักมาก

สูตรนี้ใช้กับโหลดที่เป็นความต้านทานล้วนและคลื่นไม่เพี้ยน
ถ้าเป็นโหลดสวิตชิ่งเยอะ ต้องคิดฮาร์มอนิกลำดับที่สามที่บวกกันในนิวทรัลอีกชั้น

When the neutrals of two systems are interconnected, this does more than increase neutral current. It allows current to flow across systems. Current serving house loads supplied by the old transformer can return through the interconnected neutrals and appear on the new transformer side. As a result, what the technician measures at the building's main panel no longer represents building loads alone.

The measurement trap: meter readings of 2928 / 0996 / 1032

A power meter was already installed on the panel at the site. Its display showed the three phases as L1 2928, L2 0996, L3 1032, with the unit shown as amps. After taking a photo, I almost put those readings straight into the documentation. Fortunately, I stopped to think first, because this set of numbers could be interpreted in several ways, each leading to an entirely different conclusion.

I use three checks to verify this. First, I open the meter's settings menu to check the CT primary setting, then inspect the actual CT fitted around the conductor to see whether its ratio matches. Second, I measure the same conductor at the same time with a clamp meter and compare the readings. Third, I compare the total power calculated by the meter with the displayed current. If the ratio is wrong, those two sets of figures will contradict each other.

เช็คความสมเหตุสมผลของค่าที่อ่านได้

  P (kW) ที่มิเตอร์แสดง  เทียบกับ  sqrt(3) x V x I x pf

  ถ้ามิเตอร์บอก I รวมระดับ 290 A ที่ 380 V  ->  ควรเห็น P ราว 170-190 kW
  ถ้าจอบอก P อยู่แถว 60-70 kW               ->  ค่ากระแสที่อ่านผิดอัตราส่วนแน่นอน

หลักการ: อย่าเชื่อค่าตัวเดียว ให้หาค่าที่สองที่คำนวณกลับมาหากันได้เสมอ

The repair sequence that must be followed in order

Once the configuration is understood, the solution is straightforward, but the order matters greatly. Changing the sequence creates an immediate hazard.

1. Coordinate with the electricity utility to shut off power to all houses before starting work. Never work on energized conductors.
2. Separate the house neutrals from the building's main feeder one house at a time, covering every house along the cable route. Any house taking both phase and neutral from the building's supply must have both moved back to the house supply line.
3. Only after every house has been dealt with should the utility separate the neutrals at the transformers, connecting the new 120 ตร.มม. neutral conductor prepared for the job. This size is sufficient for the entire building's neutral at present.
4. Measure each phase current again throughout the system. The readings now belong entirely to the building, so load redistribution can begin to balance the phases.
5. Follow with connection maintenance: clean the terminals, retighten them, and rewrap every joint with self-amalgamating tape, starting with the most severely damaged sections.

Side effects the project owner must be told about in advance

I consider this just as important as the work itself. When a system that has been interconnected for a long time is corrected, loads that had been drawing through the larger cable return to the cable that should have carried them in the first place, and that cable may not be large enough.

In this case, the cable running from the generator room to the houses was a small aluminium cable. While half the houses were tapped into the building's supply, this cable appeared adequate. Once every house was moved back, the entire load would fall on it. Voltage drops that had not occurred before would then appear at the houses.

The solution must include installing a larger cable on the upstream section of the house supply, alongside separating the systems. This must go into the documentation and be explained to the project owner from day one. Otherwise, when the work is finished and someone calls to complain about low voltage, the people doing the work will get the blame despite having followed every step correctly.

Connection maintenance buys time; it is not a fix

The clearest evidence from the site came on the first day of repairs: merely adjusting the terminals and retightening the wires at a few connections restored operation. That confirmed that accumulated heat had indeed loosened the connections. But this system's main feeder has a splice at every pole, meaning the entire route needs to be checked and retightened. As long as the overload remains, those connections will keep loosening again.

Summary: priorities for the whole case

I will set out the priorities clearly once more. The main cause of the melting cable is a single conductor carrying a load far above its rating. The interconnected neutrals and unauthorized house taps are aggravating factors that also prevent meaningful measurements. Both problems must be addressed, and fixing one cannot substitute for fixing the other.

Once the neutrals are separated and the phases balanced, unrelated loads are gone and the system can be measured meaningfully. But the existing main feeder still has to carry all five zones, including the heaviest loads. The new feeder therefore remains necessary. I covered cable and breaker sizing for the new feeder in the article on current ratings and installation methods.

There is one thing I want readers to remember from this case: before trusting a measurement, establish whether it actually belongs to your system. And before putting a display reading into a report, establish the configured current transformer ratio. These two checks can prevent the most expensive mistakes. If I have left anything out of the explanation, or there are other system configurations unlike this one, I will come back and add corrections later.