How to run a transient 01 Load a netlist with a `.tran` directive. The RC step preset uses V1 in 0 PULSE(0 5 0 0 0 1 2), R1 in out 1k, C1 out 0 1u and .tran 20u 5m, so the suggested step is 20 µs and the stop time 5 ms. 02 Choose the integrator. Trapezoidal (trap) is the default; backward Euler (be) is more damped; switch adaptive off to run a fixed step for deliberate comparisons. 03 Set UIC and `.ic` when there is no DC solution. A circuit with only capacitive paths has no operating point, so it must start from values such as VC(C1)=1. 04 Name the probes and read the numbers. Node voltages and element currents trace like the operating-point table; the step statistics sit under the plot. 05 Sanity-check against a closed form. For a first-order RC the capacitor voltage is 5·(1 − e^(−t/τ)) with τ = 1 ms; the run should land on it.
Worked readouts
Run
Readout
RC step, .tran 20u 5m, trapezoidal
250 accepted steps, 0 rejected, h = 20 µs , 251 Newton iterations; V(out) = 3.1603 V at t = 1 ms (closed form 3.160603 V ) and 4.966309 V at 5 ms
The same RC run with backward Euler
250 steps, V(out) = 3.142361 V at t = 1 ms — the first-order method lags the trapezoidal curve
Lossless LC tank, .ic VC(C1)=1 I(L1)=0, 1 µs steps over 2 ms
trapezoidal: 2000 steps, energy drift 0.0000 % , V(a) = 0.979220 V at 1 ms; backward Euler: drift −86.4529 % , V(a) = 0.595029 V
Series RLC step (20 Ω, 1 mH, 1 µF), 2 µs steps
peak 1.350369 V at 0.104 ms , overshoot 35.0369 % against a closed-form 35.092 % ; 1000 steps, 0 rejected
Half-wave rectifier with 1 µF smoothing, 5 µs steps
1011 steps, 5 rejected; peak 4.34265 V , settling to a late-run minimum of 1.93868 V
What the step controller actually does Each trial step solves the circuit with Newton's method at the new time point, then estimates the local truncation error (LTE) by comparing the integrator with a lower-order companion formula. The error estimate is compared with a tolerance scaled by trtol (SPICE's default of 7 is used), plus per-quantity relative and absolute floors. If the step is acceptable it is kept and the next step is allowed to grow, up to the step cap you supplied; if it is rejected the step shrinks and the point is recomputed. That is why the RC run above reports exactly 250 steps at 20 µs : the fastest time constant is 1 ms, the cap is 20 µs, and the error is ridiculously small, so the cap is the binding constraint. The rejected points in the rectifier run are the moments when a diode switches and the error estimate spikes.
The integrator choice matters most when the circuit is nearly lossless. The trapezoidal rule is a Cayley transform of the exact exponential, so it preserves the quadratic energy invariant of an ideal LC tank: after 2000 steps the drift is 0.0000 % . Backward Euler is L-stable but dissipative; it removes energy every step by roughly 1/(1 + h²ω₀²), and the same tank loses 86.4529 % of its stored energy over 2 ms. For stiff circuits — widely separated time constants, switching devices — that damping is often what you want, because it suppresses the numerical ringing that trapezoidal can sustain.
Limits and edges
Step statistics are diagnostic, not decorative. A run with many rejected steps is telling you the tolerances are tight relative to the dynamics; a run pinned at the step cap is telling you the cap, not the error, sets the resolution.
The first-order methods have visible phase error. At 250 points across five time constants, backward Euler reads 3.142361 V where the exact answer is 3.160603 V — a 0.6 % lag that shrinks with the step. What it does not do. It does not fit measured waveforms, extract a transfer function, or run a Monte Carlo sweep; it also does not model temperature drift or device noise. Use the AC tab for frequency response, the Engineering Lab for system identification and controller tuning, and the DC sweep for bias curves.
The energy check The LC preset is the cleanest available trap. Trapezoidal conserves the tank energy so honestly (drift 0.0000 % over 2000 steps) that a circuit which is physically lossy can look lossless, while backward Euler's −86.4529 % decay can be mistaken for physical damping rather than numerical dissipation. The rule of thumb: if energy or amplitude changes faster than the component values can explain, switch integrators and rerun before believing it. The half-wave rectifier shows the same issue in a different form — 5 rejected steps near each diode switch, where the waveform has a genuine corner and every integrator has to work.
Where it earns its place Learning the time constant The RC preset connects three quantities that are easy to muddle: τ = RC = 1 ms, the 63.2 % point at 3.1603 V , and the 5τ point where the run reads 4.966309 V . Change C and watch all three move together.
Checking a second-order design The series RLC run measures 35.0369 % overshoot against the closed form 35.092 % , and the peak time 0.104 ms against the predicted 0.10472 ms. That agreement between a numerical integration and a textbook formula is the point of the panel; change R and the damping ratio moves as ζ = (R/2)·√(C/L).
Privacy Integration, plotting and any exported waveforms run locally in the browser; nothing is uploaded.
References
Wikipedia, Numerical methods for ordinary differential equations , en.wikipedia.org (访问日期:2026-10-07)— step control and local error.
Wikipedia, Trapezoidal rule (differential equations) , en.wikipedia.org (访问日期:2026-10-07)— A-stability and energy behaviour.
Wikipedia, Backward Euler method , en.wikipedia.org (访问日期:2026-10-07)— L-stability and numerical damping.
Wikipedia, SPICE , en.wikipedia.org (访问日期:2026-10-07)— .tran, TRTOL and the LTE step controller.
Wikipedia, RC circuit , en.wikipedia.org (访问日期:2026-10-07)— the closed-form step response used as the check.
Calculators in this hub
Hand-picked tools, one click away. The mini versions compute live and carry your values into the full calculator.
Sources & review
Reviewed by CalcX Editorial Team
Updated 2026-10-07