Why one arrester is not enough: the relief wave
A stroke on the middle phase (S) of the T20–T21 span sends two surges, one towards each tower. When an arrester's gap sparks over, it clamps the conductor and launches a relief wave: a wave of opposite sign that travels back along the line and lowers the voltage wherever it passes. The far tower is saved only if that wave arrives before its insulator flashes. Compare one TLA with TLAs at both towers.
Arrester on S at T20 only
Voltage along the line
T19 → T22Lattice diagram
distance → · time ↓Voltage across the S insulators
T20 and T21Arresters on S at T20 and T21
Voltage along the line
T19 → T22Lattice diagram
distance → · time ↓Voltage across the S insulators
T20 and T21How far does a single TLA protect the other tower?
Each point is a full simulation with one TLA at the selected tower. It compares when the far insulator would flash with when the relief wave gets there.
The timing argument (paper, eq. 9)
Let the strike be a distance x from the protected tower. The surge reaches the far tower at (L − x)/c. The relief wave sets off from the protected tower once its gap sparks, at about x/c + τg, and reaches the far tower at x/c + τg + L/c. The far insulator flashes τf after the surge arrives. Protection therefore needs
The span length cancels out. A single TLA covers only strikes close to it, whatever the span length. With τf ≈ 0.7–0.9 µs and τg ≈ 0.3 µs, xp is about 60–90 m. The paper's T20–T21 exposure lies over the valley, 160–320 m from the towers, so it needs TLAs at both ends.
Model behind the animation
Line
Stroke
Insulator (12 glass discs with horns)
Model against paper: −10 kA mid-span 2.00/1.99 µs (paper 2.02/2.01); −8.5 kA 2.36 µs (2.27–2.33); u = 0.75: T21 1.31, T20 2.75 µs (1.33, 2.72); −5 kA no flashover; T19 and T22 do not flash at −10 kA. After flashover the path is 180 Ω (tower surge impedance).
Gapped arrester (EGLA)
Power-frequency voltage
Relief wave
Simplifications: one conductor rather than the eight-conductor frequency-dependent line, no coupling to shield wires or other phases, no corona, no tower or footing dynamics before flashover, a fixed-threshold gap, and no arrester lead inductance. The timings are close to the paper's EMTP-RV results at 8.5–12.5 kA; above about 15 kA the model flashes slightly earlier, and late flashovers at T19/T22 can appear that the paper's model does not show. Treat it as a teaching model.