A busbar is electrically a single point — every incomer pours current in, every feeder draws current out, and Kirchhoff's current law says those must sum to zero. That is the entire idea behind a bus differential. Put a CT of the same ratio on every circuit connected to the bus (incomers and feeders), wire them all in parallel into one zone, and add up their secondary currents.

For a fault outside the bus — anywhere down a feeder — whatever flows in through the incomers flows straight back out through that feeder, so the paralleled CT currents cancel and the sum is zero. For a fault on the bus itself, every source now feeds the fault and nothing leaves, so the sum jumps to the full fault current. A single measurement — the spill in that paralleled connection — tells the relay “through” from “internal” with no time delay and no coordination with anything else. That is why bus protection is fast and unit (it protects a strictly bounded zone).

In theory a through fault gives exactly zero spill. In reality the bus carries enormous through-fault currents — 40,000 A primary here — and the CT on the faulted feeder sees that entire current while the others share it. That one CT can saturate, collapse its output, and stop reproducing its share. Now the paralleled currents don't cancel: a false spill appears that looks exactly like an internal fault.

This is the central tension of every bus scheme. Make it sensitive enough to catch a small internal fault and it becomes prone to mis-tripping on the spill from a saturated CT during a huge external fault — the worst possible moment to trip the bus falsely. Every busbar design is, at heart, an answer to: how do I stay stable on a through fault yet still operate on an internal one?

The classic, brutally robust answer is the high-impedance scheme. The trick is to ask not “how much spill current?” but “how much voltage does that spill develop?” The relay is a high-impedance voltage element in series with an external stabilising resistor, R_stab, across the paralleled CTs.

Consider the worst case: a through fault where the faulted CT fully saturates. A saturated CT is just its own winding resistance — a low-impedance shunt. So the spill current takes the easy path back through that dead CT and the leads, and the most voltage it can develop across the relay is only $I \cdot (R_{CT} + 2R_{lead})$. Set the relay pickup voltage V_s just above that floor and the scheme physically cannot operate on a through fault, no matter how badly the CT saturates. Here that floor works out to about 160 V — the bar V_s must clear.

For an internal fault the situation inverts: every CT drives current into the high impedance, the voltage shoots past the CT knee-point and the element operates instantly. The same R_stab that limits the through-fault spill current is what forces the internal-fault voltage up.

Stability is not free. Every one of the 6 CTs paralleled into the zone is energised to V_s at the moment of operation, and each draws its own magnetising current — current that flows uselessly into the CT core instead of into the relay. The minimum primary current the scheme can detect is therefore $\text{CTR}\times(I_{relay} + N\cdot I_{mag})$, and that N·I_mag term grows with the number of circuits. A bus with many feeders is inherently less sensitive than a small one — the magnetising tax scales with the bus.

Here the scheme operates at about 280 A primary — 70% of the smallest circuit's rating. Pushing V_s higher for stability raises the magnetising draw and erodes sensitivity; this is the stability-versus-sensitivity trade made numeric.

An internal fault tries to drive the entire fault current into a high impedance, so the voltage across the relay can spike to kilovolts — here a theoretical peak near 11.8 kV — enough to flash over the CT and relay insulation before the breaker even clears. The fix is a non-linear resistor, the metrosil (varistor), connected across the element: at normal voltage it does nothing, but above a threshold its resistance collapses and it clamps the peak below the insulation withstand. On this bus the peak is high enough that a metrosil is genuinely required.

The second net is the check zone: a second, completely independent set of CT cores covering the same whole bus, feeding its own high-impedance element. The bus is only tripped when the discriminating (main) zone AND the check zone both pick up. A single CT failure, a wiring error, or a shorted lead can fool one zone — but it is vanishingly unlikely to fool two independent ones simultaneously. The check zone is the security that lets you trust an instantaneous, un-delayed trip of an entire substation bus.

Everything above is the high-impedance scheme. The modern alternative is a low-impedance, biased (percentage-slope) numerical bus differential, and the contrast is worth understanding because most new buses are built on it.

In short: high-impedance wins on simplicity and proven through-fault stability, but it demands matched CT ratios, dedicated CT cores, a physical resistor and metrosil, and it is awkward on a bus whose topology changes. Low-impedance numerical wins on flexibility — it compensates CT ratios in software, watches each CT for the onset of saturation and rides through it with a restraining bias, and follows isolator/breaker status to dynamically reassign circuits between zones as a sectionalised or double bus is reconfigured. The price is a more complex relay whose correct behaviour depends on the integrity of that status wiring and its settings.