Interpreting results#
A retrieval returns a complex field, a predicted trace and a convergence history. Interpreting it means inspecting the measured and predicted traces and their residual together with the temporal and spectral profiles, settling the trivial ambiguities, and checking that summary numbers such as the FWHM describe the pulse you think they do. This chapter also explains how croak’s models were validated against simulated instruments whose pulse is known.
Which page do I need?#
I want to… |
Read |
|---|---|
plot a retrieval, compute durations, phases and peak power |
|
resolve the SHG direction-of-time ambiguity |
|
check whether retrieved spectral wings are real |
|
know how well each model recovers a known pulse |
Key concepts#
The trace error \(R\) measures agreement with the model, not with the true pulse; on simulated data the known-truth errors measure the latter. A FWHM is one summary of a pulse: it says nothing about satellites or phase. The edge-energy fraction (spectral energy near the edges of the retrieval band) flags spectral wings that are artefacts of the retrieval rather than of the pulse. Durations are only comparable between fields defined at the same plane.
Main entry points#
process_result(): durations, phases, Taylor coefficients and residuals in oneProcessedResult.plot_retrieval(): the twelve-panel summary figure.resolve_time_direction(): settle the SHG time ambiguity.edge_energy_fraction(),post_filter(): spectral-edge diagnostics and windowing.truth_errors(): errors against a known pulse.
The first retrieval introduces the summary figure, and the simulation examples compare retrievals with a known field. Continue with propagation and dispersion tuning or uncertainty.