Period mapping
Pulsed-source runs can interleave acquisition conditions frame by frame — laser on/off at half the muon pulse rate in photo-μSR, RF on/off alternating every few seconds, stepped delays in pump–probe sequences. Each condition accumulates into its own period histogram set within one file, retaining its own beam exposure bookkeeping. Two-period files load as a single dataset with the Red/Green selector (Red, Green, G−R, G+R) in the grouping dialog; this page covers runs with more than two periods.
The Map periods… button (shown beside the Red/Green selector when the reference run has three or more periods) opens a matrix: one row per period, with its good-frame count, and a three-way choice
Red |
summed into the red set |
Green |
summed into the green set |
Ignore |
left out entirely |
On Apply, a combined dataset is created whose red and green sets are the
count-level sums of the chosen periods (independent supplied variances added,
raw Poisson counts added when no variance is supplied, good frames added per set), and the full Red/Green machinery applies to it:
plot either set, or the G−R difference that isolates the effect of the
stimulus. The mapping is recorded with the grouping for provenance.
Asymmetry also checks that selected list inputs are original acquired periods,
because flattening a nested set would discard its own clocks or coefficients.
A selected unit singleton is resolved to its original clock and file facts,
preserving displayed scalar overrides. Pass an existing mapped dataset directly
to remap its originals. Direct
remapping of unit-weight arithmetic remains available; a selected nonunit
coefficient requires original periods or Ignore, rather than silently
replacing that coefficient with one. Ignored coefficients remain recorded for
later remapping. Ignored unsupported nesting retains compact source identities,
not another raw replay tree; selecting it later gives original-period guidance.
Explicit mapping replaces inherited implicit counterpart intent. These checks
affect mapping/preparation requests, not loading or inspection of source rows.
Mapping acquired periods… prepares the mapped counts and replay snapshot
on a worker. Cancel leaves the original rows available and does not add an
incomplete mapped result.
Defaults follow WiMDA’s period-mapping form: period 1 to red, period 2 to green, the rest ignored; with no green periods the result behaves as a single-set run and the G±R modes do not apply.
When to use this. Multi-condition sequences where several periods share a condition — e.g. a laser-delay scan where periods 1 and 3 are light-off and 2 and 4 are light-on at different delays: map {1, 3} → red and {2} → green to compare one delay against the off reference, then remap for the next delay without reloading anything. Each period retains its own measured good-frame count. G+R forms one asymmetry from the pooled prepared counts; G−R subtracts independently reduced set asymmetries without an exposure multiplier or a factor of one half. Summing periods with different conditions into one set is statistically fine but physically meaningless — the sums are labelled by your mapping, not by what the hardware did, so record the period legend (the acquisition log’s period table) with the analysis.
The rows identify stored data periods, numbered from one, rather than every
slot in the acquisition’s DAE sequence. Declared NeXus period_type and
period_labels remain inspectable in file metadata. When the declared data
slots correspond to the stored arrays, each selected period records its DAE
slot separately; absent or inconsistent declarations record that limitation.
The dialog does not add dwell rows or fix them to Ignore. Zero counts do
not establish a dwell interval, and ROOT detector banks are not inferred to be
acquisition periods.
Original acquisitions, including ignored ones, retain their counts, supplied variances, timing, good-bin limits, exposure, and file coefficients. Mapping leaves the individually loaded observations available for another selection or mapping. A red-only mapping reduces and can be selected as period 1; a green selection or G±R operation requires an actual green set. G±R establishes common observed native-bin support across both sets before binning. Incompatible widths or empty support affect the requested combination; the original observations remain individually inspectable.
For a requested set, Asymmetry resolves current grouping, exclusions, and timing and correction policies against each original acquisition. It corrects each required detector with that acquisition’s applicable coefficient and supported exposure, aligns whole bins using its own resolved t0, and pools counts and retained variances on common observed support. Group-background preparation then acts once on the requested set, followed by count binning and the ratio. There is no effective deadtime-coefficient average or fractional resampling. Fractional timing residuals remain recorded with the acquisition/channel mapping.
One explicitly chosen α/β applies to the comparison; individual-period calibration requires an explicit source selection. Missing required File coefficients leave the complete requested result uncorrected with a warning, including required references. Ignored inputs and unused raw bins do not become correction requirements. Scoped preparation retains raw context outside consumed bins; its excluded variance contribution does not certify those bins as observations. Signed contributors require supplied variance in consumed bins. Held raw-index arrays remain available for inspection; the quick look and its counts-first replay use native bins, file support, selected α, β one, and corrections off. Point errors retain first-order propagation at fixed α/β, with existing shared background terms. This does not provide full cross-period/reference covariance. Asymmetry also checks that detector lists correspond and that signed input has supplied variance before pooling multiple contributors. Single acquired observations remain loadable without a new statistical-eligibility inference. An unsupported sum raises an operation error; the original observations remain available.
Reference-background selection
A Background run chosen from a loaded period retains that acquired period when loaded from its source file or reopened in a project. Period exposure is selected with its histograms, and need not equal another period’s exposure. Distinct acquired periods in one file may serve as independent inputs; selecting the identical observations as an independent reference is refused. See Background correction for the selection and uncertainty contract.
Compared with other programs
WiMDA 1.326 pools mapped raw counts before correcting the resulting set.
Asymmetry corrects acquired constituents before pooling, following the order
of Mantid’s documented per-workspace MuonPreProcess. musrfit 1.12.1 also
corrects selected histogram contributions before grouping, and its grouping
and ADDRUN support whole-bin timing shifts. Asymmetry also checks compatible
native widths and common observed support for the requested combination;
a failed combination leaves individual observations available. These are selection
and operation comparisons, not equivalence of exposure or error models.
WiMDA’s G+R forms one pooled-count ratio, as does Asymmetry. In Mantid,
MuonPairingAsymmetry pools SummedPeriods before taking the ratio.
With equal-total 20% and 80% inputs, G+R is 50% and G−R is 60%. Unequal totals
do not generally give the arithmetic mean. One group-background operation acts
on pooled G+R, including Manual levels in the requested observation’s count
units. Reference subtraction uses summed evidenced sample exposure for Automatic
scaling or the explicit Manual coefficient once, retaining both-side variance.
G−R keeps separate set preparation and independent marginal variance addition. See Asymmetry in the μSR software landscape.
References
S. J. Blundell, R. De Renzi, T. Lancaster, and F. L. Pratt, Muon Spectroscopy: An Introduction (Oxford University Press, Oxford, 2022). — photoexcitation and RF alternating-frame acquisition.
Pratt, Physica B 289–290, 710 (2000).
Mantid, MuonPreProcess v1 and LoadAndApplyMuonDetectorGrouping v1 (6.16 development documentation).
musrfit, User manual (1.12.1).