Skip to content

Switching chatter after an inductive opening ​

Open in simulator →

A breaker opens on a 10 A inductor current with nowhere for the current to go, and the trapezoidal rule answers with a voltage that flips sign every step and never decays.

A 100 V DC source drives 10 A through a breaker, a 10 Ω resistor and a 10 mH inductor. The breaker opens at t=5 ms. Physically the current would stop and the inductor voltage would spike once; at a 50 µs step the simulated inductor voltage vL instead alternates between about +4 kV and −4 kV on every step for the rest of the run.

Before the opening the circuit is an RL step response: the current climbs with a 1 ms time constant and has reached 9.93 A by 5 ms. The breaker then opens with no arc model and an off resistance of 1 GΩ, which leaves the inductor current no path.

The trapezoidal companion model of an inductor behaves, over one step, like a resistance of 2L/Δt in parallel with a current source carrying the history. When the breaker cuts the current, the integrator cannot remove the inductor's stored current without overshooting it. The solution settles into a mode that alternates sign every step: the current flips between about ±4 μA and the inductor voltage between ±(2L/Δt)I, which at Δt=50 μs is 400 Ω×9.93 A≈3970 V. The trapezoidal rule adds no damping, so the alternation keeps its amplitude for the remaining 100 steps of the run. Plot vL with markers to see it: two samples per cycle, at the Nyquist rate of the step, with no physical frequency behind it.

The engine has no automatic critical-damping adjustment, so what the plot shows is the raw integrator. Chapter 6 of the EMT course explains the mechanism and the standard remedies.

Things to try ​

  • Give the current a path. Add a 400 Ω resistor from the inductor's top node to ground. That is 2L/Δt, and the alternation is gone one step after the opening: the inductor current decays through the resistor instead. A snubber across the breaker or a freewheeling diode does the same job, which is why well-designed converters do not chatter.
  • Shrink the step. At Δt=10 μs the alternation is still there, but its period halves with the step and its amplitude grows five-fold, to about ±19 900 V, because 2L/Δt grows. A smaller step does not cure chatter; changing with the step is how you recognise it as numerical.
  • Plot iL. The physical answer, a current that stops, is buried in the microamp-level alternation.

The run is 10 ms at a 50 µs step.

In the course ​