Surges & arresters · Insulation Lab

Why an insulator flashes, or does not

Each phase hangs on a string of 12 glass discs with arcing horns. In a lightning surge the discs do not matter much: the flashover runs through the air between the horns. It is not instant. Above a critical field a leader grows from the horns across the gap, and it needs time to bridge it. A short, high pulse can pass without a flashover, while a lower but longer one may not.

String
12 × U120B
glass discs, arcing horns
Effective gap g
1.44 m
paper calibration
Leader starts at
E₀·g = 913 kV
E₀ = 634 kV/m
CFO (specification)
884 kV
used by the Compare Lab
t = 0.00 µs

Across the horn gap

leader length to scale

Voltage and the strength of the remaining gap

kV
voltage across the insulatorE₀·(g − l): leader grows only above this

Leader length

cm, of 144 cm
Volt–time curves

Higher voltage, faster flashover; shorter tail, higher strength

Each curve is the same insulator under one wave shape. A point is the peak voltage and the time at which the leader bridges the gap. The left end of each curve is its critical peak: below it, the wave ends before the leader can finish.

Critical peak by wave shape

WaveLeader modelCIGRE short-tail

CIGRE short-tail column: CFONS = (0.977 + 2.82/τ)·CFO with CFO = 912.5 kV and τ the tail time constant, the formula the Compare Lab and the grounding modules use. Both agree that short tails need a higher peak. The simple formula climbs much faster for tails below about 6 µs, where it is outside its intended range.

What to take away

Six lessons

Below 913 kV nothing grows

The leader advances only while the voltage exceeds E₀ times the remaining gap. At the start that is 634 kV/m × 1.44 m = 913 kV. Any voltage below it can last as long as it likes. This is why the paper's calibration says the string "withstands 912.47 kV and flashes at 912.56 kV" on a long wave.

A peak is not a flashover

Once started, the leader shortens the gap, the field at its tip rises, and it accelerates. But it needs a microsecond or more. In the paper's −69 kA backflashover case, phase S saw 1.04 MV and the leader grew only 6.2 cm before the voltage fell. A fixed-CFO check would have called it a flashover.

Short tails are easier to withstand

In a backflashover the footing and the adjacent towers send relief waves within 0.4–4 µs, so the insulator sees a short-tailed pulse. That is why designs use a "non-standard" CFONS above the standard CFO.

The time to flashover is the arrester's window

An arrester only helps if its gap sparks over and its relief wave arrives before the leader finishes. For a 10 kA direct strike the far insulator takes about 0.7–0.9 µs after the surge arrives. That gives the 60–90 m reach of a single arrester in the Relief Wave Lab.

884 kV or 912.5 kV?

884 kV is the specification CFO: a 50% flashover voltage for the standard wave, a statistical figure. 912.5 kV is the deterministic calibration of the leader model in EMTP. They differ by 3%. The screening pages use 884 kV with the CIGRE formula; the travelling-wave labs use the leader model.

The arrester gap is the weaker link, on purpose

The arrester's series gap (650 ± 50 mm) sparks over at about 570 kV, well below the horn gap. On a protected phase the gap always goes first and the insulator is spared. No volt–time curve is published for the gap, so the paper uses a fixed 570 kV and tests 450 and 500 kV.

Model behind the page

Leader progression

dl/dt = k·u·[u/(g − l) − E₀] while u/(g − l) > E₀; flashover when l = g. g = 1.4395 m, E₀ = 634 kV/m (paper calibration, 4.8% above the 605 kV/m negative-polarity guideline), k = 1.8 m²/(kV²·s) (set in the Relief Wave Lab to reproduce the paper's flashover times). Leader drawn as growing equally from both horns.

Wave shapes

Standard: double exponential 1.2/50 µs. Short tail: a smooth (1 − cos) rise to the peak at the chosen front, then an exponential decay to half at the chosen tail time. Chopped: 1.2/50 µs falling to zero in 0.1 µs at the chop time. The three event waves are the phase-S insulator voltages from the Backflashover Lab and the Relief Wave Lab, computed with flashover suppressed and scaled by the slider.

Leader progression is a deterministic model of the 50% behaviour. Real flashover voltages scatter by a few percent (standard deviation about 3% for the CFO).