Asymmetry Logo

Contents

  • Getting started
  • Analysis workflows
  • Reference
    • Find a feature
    • Scripting
    • Data and projects
    • Reduction and calibration
      • Grouping and calibration walkthrough
      • Detector Grouping and Layout
      • Data processing
      • Asymmetry units across the API
      • Alpha calibration
      • Beta calibration
      • Background correction
      • The background ladder: which stage removes what
      • Histogram binning
      • Time-zero search
      • Detector exclusion
      • Period mapping
      • Run arithmetic: co-add and co-subtract
      • Exclusions: a glossary of six mechanisms
    • Setting up and running fits
    • Fit wizards
    • Fit-function library
    • Frequency-domain analysis
    • Trending and downstream models
    • Specialised modes
    • Simulation
    • The graphical interface
  • Background and design
  • API reference
  • Contributing
⚠️ Alpha software. Descriptions of µSR data, fitting models, and workflows here may be incomplete or in error — verify the physics against trusted sources.
Asymmetry
  • Reference
  • Time-zero search
  • View page source

Time-zero search

Time zero establishes the origin of the analysis time axis. File timing is an instrument calibration; a count feature provides a candidate anchor for review. Neither agreement with a feature nor a quiet diagnostic certifies physical time zero. Timing offsets change phase and early-time interpretation.

Selected timing

From file retains each selected channel’s usable file calibration. Integer-only calibration is sufficient, including a legitimate zero. NeXus integer/exact conflicts retain usable integer anchors and discard only the conflicting exact declaration. Format-specific unspecified markers remain separate from explicit zero; see Loading Data.

Legacy LEM ROOT v4/v5 uses numeric file resolution and t0 rather than display strings or histogram axes. The supported v4 four-channel NPP/PPC banks share header slots 0–3 while retaining separate count arrays. Zero and fractional anchors remain usable; negative or nonfinite entries supply no file timing. Valid file timing remains usable with unknown source metadata, without an automatic search to compensate for a missed header.

When a required channel has no usable file timing, From file searches its raw counts under an established source strategy. Other selected channels retain their usable file calibration. The fallback records estimated provenance; it never relabels the estimate as acquired calibration. Manual applies a signed whole-bin offset to the resolved base. An offset alone cannot supply a missing base. Switching periods reapplies a Manual offset to that period’s own resolved base; an estimated base from another period is not transferred.

Auto-detect requests replacement timing for the actual selected acquired channels, after exclusions, for each pair and period. Every required channel must supply its own usable estimate. A failed unused channel does not block a valid pair. A failed selected channel never receives another detector’s anchor or a median shift. The diagnostic run-wide median is not a detector calibration.

Asymmetry also checks that automatic timing has a known source strategy and that every required channel resolves. These checks protect the requested alignment, rather than file admission. An unresolved request shows Preview unavailable:, keeps Apply disabled, and preserves the acquired histograms and editable settings. Repair it by selecting usable channels, returning to valid file timing, or supplying supported explicit calibration or an API strategy. Valid file timing remains usable when an advisory search fails.

Source and feature conventions

Known source metadata selects the existing continuous or pulsed strategy. NeXus definition=pulsedTD supplies an explicit pulsed-source fact. Missing or ambiguous source facts remain unknown. A filename does not select a strategy. The API find_t0_for_run(..., pulsed=False) explicitly selects the continuous strategy; pulsed=True selects pulsed search and takes precedence over metadata. resolve_selected_timing accepts the same explicit argument when resolving a selected request. T0Policy.strategy records a last result for display; it is not an authored override of unknown source metadata.

  • Continuous search takes the maximum-count bin of each acquired histogram, choosing the earliest bin on ties. This is a prompt-peak feature, provided that the maximum represents that feature.

  • Pulsed search walks back from the maximum to the adjacent leading half-height crossing, interpolates linearly, and rounds to the nearest bin. An early flash can therefore be separate from the chosen leading edge. A histogram starting above half-height has no supported leading crossing.

These feature definitions do not identify the physical pulse centre for every source response. Background, flash, overlap, and detector response can affect what feature is found. A physical timing calibration requires independent instrument evidence; changing the estimator requires a separate benchmark.

Search uses raw, unbinned acquired counts. Display binning, Fourier crops, and good-window choices do not change the feature estimate. The grouping window reuses a background raw-channel scan across selections of the same acquisition; acceptance follows the current pair and period. Cached-only periods without raw histograms are not searched.

Diagnostics and controls

Auto-detect timing for a selected pair whose own prompt estimates agree with file anchors; an unused empty channel does not block the request.

Auto-detect uses the selected pair’s own estimates. Agreement in this synthetic example is a feature check, not physical calibration evidence.

The line shows File: bin N or File: none, followed by Detected: … while the worker is pending, Detected: unavailable on failure, or Detected: bin M (strategy, spread S). The detected bin is the maximum of selected own estimates. The signed Δ diagnostic is the median selected detected_i − file_i shift; the spread is the range of those shifts. They are different summaries when detector delays differ. A failed selected estimate is named in the unavailable diagnostic.

A ⚠ button exposes advisory messages about file disagreement, detector-shift spread, and a good window opening on the detected feature. Valid file timing is not rejected by these messages. The display threshold is the larger of a family floor (two bins continuous, three bins pulsed) and the median feature width. Prompt FWHM, pulsed 10–90 % rise width, spread, and thresholds are descriptive diagnostics, not confidence intervals, localisation limits, or calibration tests.

Find t0 fills the Manual common offset from the cached selected estimate. It is available after the raw scan completes; Apply commits the request. Use Auto-detect to apply each channel’s own estimated anchor rather than a shared manual offset.

Alignment and windows

Alignment transports whole acquired bins. Exact file origins retain their fractional residuals under integer policy shifts; the search does not estimate new fractional calibration. The selected maximum anchor wins, with acquired ordinal breaking ties. A pair or projection change resolves the exact origin from its own selected reference detector, even when the integer anchor is unchanged. Recorded alignment includes the actual selected channel origins, effective timing sources, requested search strategy, and residual spread. Acquired t0_source remains a separate file fact, including when estimated and file anchors are numerically equal.

File-good intervals remain attached to acquired bins. Manual good-window endpoints follow effective timing as signed offsets; omitted endpoints retain mapped file bounds. Only common observed support is reduced. See Detector Grouping and Layout, Project files, and Count reduction and transformed signals.

For Background run…, sample and reference retain independent timing requests. Changing sample timing cannot overwrite a reference’s File, Manual, or Auto-detect anchors. The same whole-bin mapping and exact-origin residual records apply to F/B and grouped spectral/MaxEnt inputs. Native bin widths and required observed support must be compatible at subtraction; raw loading stays available. See Background correction for reference replay and compatibility details.

Compared with other programs

WiMDA’s Search for T0 finds a grouped maximum, including for pulsed data. musrfit 1.12.1 musrt0 -g writes detector prompt maxima. Unique synthetic maxima at bins 20 and 31 agree exactly with Asymmetry’s continuous feature. Asymmetry’s leading half-height feature for pulsed data is a different choice; a maximum comparison does not validate physical pulse-centre accuracy.

Mantid 6.16.1 MuonPreProcess applies a supplied timing table and shifts spectrum coordinates. That operation tests timing application, not automatic feature detection. Whole-bin Asymmetry alignment and Mantid’s retained shifted coordinates agree only when selected observations and origin conventions match. The per-selected-success safeguard above is an Asymmetry request rule; no claim is made that all three programs enforce it. See Selected automatic timing for conditions and limits.

Mapped acquired constituents

Each selected acquired constituent resolves its own timing before count pooling. The common grid uses the maximum selected effective integer anchor, breaking ties by acquired period and channel order. Each file-good interval stays attached to its own raw bins. Fractional residuals are recorded without resampling or padding; incompatible widths or empty common support make the requested combination unavailable while the original observations remain inspectable. Automatic searches use original acquired counts; a Red/Green held-set search cannot supply an anchor for several distinct acquisitions. See Period mapping.

References

  • S. J. Blundell, R. De Renzi, T. Lancaster, and F. L. Pratt, Muon Spectroscopy: An Introduction (Oxford University Press, Oxford, 2022), time-zero calibration.

    1. Suter and B. M. Wojek, Phys. Procedia 30, 69 (2012).

  • musrfit user manual.

  • Mantid MuonPreProcess.

  • WiMDA.

Previous Next

© Copyright 2026, Asymmetry Contributors.

Built with Sphinx using a theme provided by Read the Docs.