Checking Simulation Waveform Resolution Before Measuring Peaks or Slopes

A smooth simulation plot can conceal how many data points support a peak, slope, or short pulse. The solver’s time steps, saved waveform compression, display interpolation, and exported file are related but different stages. Check the data path before treating a measured feature as a precise circuit prediction.

Identify the stored data behavior

In Analog Devices’ LTspice support discussion of exported waveform resolution, an ADI employee explains waveform compression controls, including disabling compression. That guidance concerns saved data. It does not mean that changing a display or export option alone improves every aspect of numerical accuracy.

Record simulator version, analysis settings, maximum time step where specified, compression settings, and the export method. Identify whether downstream analysis receives the original adaptive time points or resampled data.

Test the measurement’s sensitivity

1. Choose the actual metric, such as peak voltage, pulse width, or settling time.
2. Inspect the saved points around the feature of interest.
3. Repeat with a suitably refined time-step limit.
4. Compare with reduced or disabled waveform compression where appropriate.
5. Check that the export and analysis preserve time and units correctly.

For example, a derivative calculated from a sparsely saved edge can differ substantially from the slope suggested by a plotted line. Adding interpolated points later does not recover information that was never retained. Evaluate the metric using data with adequate resolution.

Separate numerical stability from model validity

If successive refinements produce materially different results, investigate before selecting the most favorable run. The circuit may contain a very fast event, a model discontinuity, or a numerical issue. Smaller time steps can clarify a calculation but cannot supply a missing physical parasitic or an undocumented device behavior.

Keep the same stimulus and model files during this sensitivity check so only the intended numerical settings change. Record run duration and file size if they affect the practical choice of settings, but do not use those concerns to hide a decision-critical uncertainty.

Archive the metric calculation and representative raw data with the simulation. A screenshot alone is insufficient when a result depends on a few fast points or a small residual error.

The release should state that the reported metric is stable enough under appropriate numerical refinement for its purpose. Combine that evidence with the model’s limitations and hardware validation where needed. This makes the result more credible than simply displaying extra decimal places in a measurement box.

Check exported decimal precision as well, since rounding can obscure a small difference even when the internal waveform is sufficiently resolved.