Rotating-frame projection ========================= .. image:: /_generated/screenshots/rotating_frame_projection.png :alt: The plot in the rotating frame: the Lab | Rotating switch on Rotating, the frame bar filled by Auto-detect, and stacked P′_x, P′_y and P_z subplots of a synthetic RF nutation run :width: 100% *A synthetic two-period RF run on an idealised eight-detector octant polarimeter,* *shown in the rotating frame after* **Auto-detect…** *(see* `Worked example`_\ *).* *The spin leaves* :math:`+z` *and nutates about* :math:`B_1 \parallel x'`\ *:* :math:`P'_y` *carries the nutation,* :math:`P'_x` *stays at zero, and* :math:`P_z` *swings through zero and back. The frame bar above the title holds the* *setup and run fields; the badge above the top subplot names the frame.* In a radio-frequency (RF) μSR experiment a coil adds a field :math:`B_1` oscillating at the generator frequency :math:`\nu_\mathrm{RF}` to a static field :math:`B_0 \parallel z`. Near resonance, :math:`\nu_\mathrm{RF} \approx (\gamma_\mu/2\pi)B_0`, the muon spins turn away from :math:`z` — they *nutate* — and the physics of the experiment (nutation, spin locking, echoes) lives in a frame that turns about :math:`z` with the drive. In the laboratory the same motion is a fast precession at :math:`\nu_\mathrm{RF}` whose envelope carries the nutation, so the lab-frame :math:`P_x` and :math:`P_y` are hard to read and harder to fit. The rotating frame is the standard tool of magnetic resonance for exactly this reason [1]_ [2]_ [3]_. When a grouping measures both transverse projections — the :math:`P_x` and :math:`P_y` of EMU's vector-polarisation mode (:doc:`vector_polarization`) — Asymmetry reaches that frame *exactly*. Each time bin's transverse pair is rotated about :math:`z` by the angle the frame has turned through, giving :math:`P'_x` and :math:`P'_y`, while :math:`P_z`, which a rotation about :math:`z` leaves alone, is shown unchanged. No filter is involved: every bin is transformed on its own, its error bar is propagated through the rotation, and retained counting noise is propagated with the pair. Display bunching is applied *after* the rotation, which is valid for that reason, and the rotated curves can be fitted like any other projection (`Fitting P′_x and P′_y`_). This is a different tool from the :doc:`rotating-reference-frame display ` of a single forward–backward asymmetry. With only one projection measured, the frame can be reached only by demodulating and low-pass filtering, which correlates neighbouring points and is meant for looking rather than fitting. That filtered display stays available for single-pair data under **Options → Advanced → Rotating reference frame**; it is not offered when the exact projection is. Switching to the rotating frame ------------------------------- Whenever the displayed projections include :math:`P_x` and :math:`P_y`, a **Lab | Rotating** switch sits to the left of the projection chips. Clicking **Rotating** relabels the transverse chips ``P′_x`` and ``P′_y`` and shows the frame bar in the slot above the plot title; **Lab** returns to the laboratory projections. The chips keep their tint, selection and memory in either frame. When the toolbar is narrow the switch shortens to **Lab | Rot**, and when the chips fold into a single button the switch moves into its menu as a **Frame** section above the projections, with the button reading, for example, ``Rot · 2 of 3 ▾``. The switch is saved with the project's plot state. The first time, nothing is drawn: the plot reads "Enter ν_RF to show the rotating frame." and the bar's only editable field is **ν_RF**, outlined and marked ``required``, under the hint "Enter ν_RF (the generator frequency) and the B₁ axis. Auto-detect… proposes the rest and changes nothing until you apply." Typing :math:`\nu_\mathrm{RF}` creates a frame on every displayed run, with every other field at its default, and the rotated curves appear. :math:`\nu_\mathrm{RF}` is the one value you must supply, because it is the generator's frequency and is not recorded in the data files. Nothing else is estimated until you ask. The frame bar ------------- The bar groups the frame's values by what they belong to. **SETUP** fields describe the spectrometer and drive, and are shared by every displayed run: an edit is written to each of them. Where the displayed runs disagree, the field is empty and reads ``mixed``. ``ν_RF`` The generator frequency, typed in MHz or, with the **MHz | G** toggle, as the field in gauss that resonates at it (:math:`\nu = (\gamma_\mu/2\pi)B`). The frame turns at this frequency. ``B₁ ∥`` **x** | **y** The lab axis of the linear RF field, which you know from the coil. Its sign carries no information for a linear field, and the choice between :math:`x` and :math:`y` is a 90° phase; what it fixes is which rotated component is parallel to :math:`B_1`. The y labels say so: with :math:`B_1 \parallel x`, :math:`P'_x` is labelled :math:`\parallel B_1` and :math:`P'_y` :math:`\perp B_1`. ``φ_RF`` (°) The phase of the RF drive at :math:`t_0`. At a fixed frequency and trigger it is a constant of the setup, so one value serves a whole series of runs. Sense (``↻ −1`` | ``↺ +1``) The rotation sense of :math:`P_x + iP_y` ("Rotation sense of P_x + iP_y"), toggled by clicking. It is :math:`-1` when :math:`(P_x, P_y, P_z)` are right-handed about :math:`B_0`, the usual case, and :math:`+1` when the labels are left-handed about it — for example with :math:`B_0` reversed. **RUN** fields (the header reads ``RUN`` and the run number) belong to the run selected in the Data Browser: ``b_x``, ``b_y`` (%) The transverse baselines left after α: the constant parts of :math:`P_x` and :math:`P_y`, removed before the rotation. Each period of a two-period run keeps its own pair (`Periods`_). ``a_y/a_x`` The transverse gain :math:`g`, the amplitude of :math:`P_y` over that of :math:`P_x`. It rescales :math:`P_y` so that the two transverse projections measure the polarisation on one scale. Each field shows where its value came from. A default is grey italic, an estimate written by **Auto-detect…** looks plain, and a value you typed carries ✎ ("Typed by you — Auto-detect never proposes over it"). The frame is saved on each run in the project, values and provenance together. The badge above the top subplot names the frame the curves were drawn in, for example ``Rotating frame · ν = 1.355 MHz · B₁ ∥ x′ · φ_RF = 39.9° · s = −1``, so an exported figure describes itself; a value the displayed runs disagree on reads, for example, ``φ_RF mixed``. When the plot panel is narrow the bar wraps onto two rows, and below that the run fields fold into a ``Run`` *n* ``▾`` button that opens them in a popover, with **Auto-detect…** shortened to **Detect…**. Auto-detect ----------- **Auto-detect…** estimates everything except :math:`\nu_\mathrm{RF}` and the :math:`B_1` axis: "Estimate the sense, φ_RF, baselines and gain from the displayed runs over the visible window; nothing changes until you apply." It is enabled once every displayed run has a frame and the runs share :math:`\nu_\mathrm{RF}` and the :math:`B_1` axis. It works from the time window currently visible on the plot, which must span at least two turns of :math:`\nu_\mathrm{RF}`; every sum it forms is weighted by inverse variance, so the noisy late bins of a long window do not swamp the signal and there is no need to zoom in first. It reads every period of each run on its own, never the combination the plot is showing. - **Baselines**, for each period of each run, are fitted beside the signal: a least-squares fit of :math:`b + c(t)\cos 2\pi\nu t + d(t)\sin 2\pi\nu t`, in which :math:`c(t)` and :math:`d(t)` are smooth cubic splines that absorb any slowly varying amplitude and phase. A plain mean over the window is pulled by the strong early turns, and an average over whole turns leaks the :math:`\nu \pm \nu_1` sidebands of a nutation; the fitted baseline suffers from neither. - **Gain**, per run, is the ratio of the RMS transverse swings. In the lab the transverse polarisation turns in a circle, so the two projections should swing equally. - **Sense**, from the runs together, is the side, :math:`+\nu` or :math:`-\nu`, of the complex spectrum of :math:`P_x + iP_y/g` that holds more power. The ratio of the two (as amplitudes) is the estimate's *contrast*. - **φ_RF**, from the runs together and for that sense, turns the transverse polarisation onto the direction a nutation from :math:`+z` takes, :math:`\hat z \times \hat B_1` (`Conventions`_). The axis is the principal direction of the transverse polarisation in the frame; its sign comes from early times, when a spin leaving :math:`+z` first moves that way. :math:`\nu_\mathrm{RF}` is never estimated. Under RF the transverse spectrum is not a single line at :math:`\nu_\mathrm{RF}` but a sideband pair at :math:`\nu \pm \nu_1`, which can lie far from :math:`\nu` when :math:`B_1` is large, so a peak search would find a sideband rather than the drive. .. image:: /_generated/screenshots/rotating_frame_review.png :alt: The Auto-detect rotating frame review with the sense and phase row, a gain row and a baseline row for each period, all pre-ticked :width: 75% *The* **Auto-detect rotating frame** *review for the run above: the setup row,* *then the run's gain and the baselines of its red and green periods, each* *estimate beside the current value and its contrast.* The estimate opens in the **Auto-detect rotating frame** review rather than being applied. Its header names the :math:`\nu_\mathrm{RF}` and :math:`B_1` axis it worked at, and each row sets the current value beside the estimate (``0.0° → 40.1°``): - **Sense, φ_RF** — one row for the runs together, since :math:`\varphi_\mathrm{RF}` is estimated for that sense, written to every run. - ``Run`` *n* ``· gain`` — the run's gain. - ``Run`` *n* — the run's baselines; a two-period run has one row per period, ``Run`` *n* ``· red`` and ``Run`` *n* ``· green``. Rows still on defaults or earlier estimates start ticked; a row holding a value you typed never does, and is marked ``typed``. An estimate whose contrast is below 3 cannot be applied ("contrast … < 3 — cannot apply"): the data carry too little transverse signal to say which way it turns. Nothing changes until **Apply** *n* **changes**, which writes the ticked values as estimates; the title strip then reads, for example, "✓ Applied 4 estimates · contrast 65". **Cancel** leaves every frame as it was. If the estimate cannot be made — a window shorter than two turns, for example — the status bar reads "Auto-detect failed — see the log" and the log names the reason. Periods ------- On a two-period run (red is period 1, green period 2) each period keeps its own baselines, with their provenance; the gain, sense and :math:`\varphi_\mathrm{RF}` are shared. The distinction matters on RF data, where the red and green transverse baselines can differ by about 0.1 % — for example when the RF is on in one period and off in the other. A single pair of baselines would leave the difference behind as a spurious oscillation at :math:`\nu_\mathrm{RF}` in the rotated curves. While the Grouping window's **RG Mode** shows **Red** or **Green**, the bar shows and edits that period's baselines. **G minus R** removes :math:`b_G-b_R`. Pooled **G plus R** removes one common baseline pair. Asymmetry also checks that Red and Green transverse baselines agree exactly, including zero, before rotating pooled G+R. Unequal baselines show "Pooled G+R rotation unavailable: Red and Green transverse baselines must agree. Use Lab, Red, Green, or G minus R." The derived baseline fields are empty and read-only. Lab G+R, rotating Red, Green, and G−R remain available. musrfit's scalar RRF and Mantid's two-component ``RRFMuon`` are different observation workflows. These comparisons do not establish a native policy for pooling unequal period baselines. The equality check is an Asymmetry safeguard. Read a difference with care. A period with the RF off carries no transverse signal, so when the RF is on in red and off in green, **G minus R** holds minus red's transverse polarisation: the nutation appears turned over, into :math:`-y'` for :math:`B_1 \parallel x'`. That is correct arithmetic, not a wrong :math:`\varphi_\mathrm{RF}` — which is why **Auto-detect…** estimates from each period separately and never from the combination on display. Conventions ----------- The rotation removes the baselines, puts the labelled projections on physical axes, and then undoes the precession at :math:`\nu_\mathrm{RF}`: .. math:: z_\mathrm{phys}(t) = \bigl(P_x - b_x\bigr) + i\,(-s)\,\frac{P_y - b_y}{g}, \qquad z'(t) = z_\mathrm{phys}(t)\,e^{+i(2\pi\nu_\mathrm{RF} t + \varphi_\mathrm{RF})}, with :math:`P'_x = \mathrm{Re}\,z'` and :math:`P'_y = \mathrm{Im}\,z'`. Here :math:`s` is the sense, :math:`g` the gain, and :math:`b_x`, :math:`b_y` the baselines of the displayed period (or their combination). The muon's gyromagnetic ratio is positive, so its spin precesses in the negative sense about a field along :math:`+z`, and :math:`e^{+i\theta}` undoes that precession. With :math:`\varphi_\mathrm{RF}` the drive's phase at :math:`t_0`, the co-rotating half of the linear RF field is static in the frame and lies along the chosen axis. A spin leaving :math:`+z` then nutates about :math:`B_1` towards :math:`\hat z \times \hat B_1`: - with :math:`B_1 \parallel x'`, into :math:`+y'` — :math:`P'_x` is the component along :math:`B_1` (the spin-locked one) and :math:`P'_y` the one across it; - with :math:`B_1 \parallel y'`, into :math:`-x'`. On resonance the nutation runs at :math:`\nu_1 = (\gamma_\mu/2\pi)B_1`, with :math:`B_1` the amplitude of the co-rotating component. The frame turns at the generator's :math:`\nu_\mathrm{RF}`, so that :math:`B_1` is static in it, not at the muon's Larmor frequency. During free evolution — with the RF off, or between pulses — the spins turn at :math:`\gamma_\mu B_\mathrm{local}`, so a local field away from resonance appears in the frame as a slow drift at the offset :math:`\gamma_\mu B_\mathrm{local}/2\pi - \nu_\mathrm{RF}`. That drift is information about the field at the muon, not an error in the frame. .. dropdown:: Error propagation With :math:`\theta = 2\pi\nu_\mathrm{RF}t + \varphi_\mathrm{RF}`, :math:`x = P_x - b_x` and :math:`y = (P_y - b_y)/g`, .. math:: P'_x = x\cos\theta + s\,y\sin\theta, \qquad P'_y = x\sin\theta - s\,y\cos\theta, The transverse counting covariance is propagated as :math:`C'=MCM^T`, where :math:`M` is the matrix of the two equations above. With :math:`c=\cos\theta`, :math:`q=\sin\theta`, and :math:`k=\operatorname{Cov}(P_x,P_y)/g`, the marginal variances are .. math:: \sigma'^2_x &= c^2\sigma_x^2 + q^2(\sigma_y/g)^2 + 2scqk,\\ \sigma'^2_y &= q^2\sigma_x^2 + c^2(\sigma_y/g)^2 - 2scqk. The rotated cross-component covariance is :math:`cq[\sigma_x^2-(\sigma_y/g)^2]+s(q^2-c^2)k`. Overlapping projections can have positive, negative, or cancelling counting covariance even though each individual F/B pair is disjoint. Equal marginal errors alone do not imply independent rotated components. .. _vector-rotation-counting-errors: Counting observations and conditional errors ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For fixed :math:`\alpha` and :math:`\beta`, each projection uses :math:`A=(F-\alpha B)/(\beta F+\alpha B)`. Writing :math:`D=\beta F+\alpha B`, its fractional count derivatives are .. math:: \frac{\partial A}{\partial F}=\frac{\alpha B(1+\beta)}{D^2},\qquad \frac{\partial A}{\partial B}=-\frac{\alpha F(1+\beta)}{D^2}. Asymmetry applies the percentage conversion once and contracts these derivatives against the actual retained acquisition and reference-bin variances. A detector used at a different aligned raw bin is a different observation. Signed constituent coefficients and reference scales enter before covariance propagation. G−R combines the reduced period responses, including any reused reference; G+R forms the ratio after corrected-count pooling. Existing projection calibration rules are retained; the general derivative does not enable a new beta setting. Range-background and tail-background estimates retain their count influence, including overlap with the displayed bins. Within a projection cohort, tail-fit uncertainty uses acquisition counting variance (or supplied variance), rather than a separate fitted Poisson expectation for each overlapping group. Fitted background levels and point values are unchanged. Display packing averages the rotated responses and retains this shared estimation noise; it is not always an independent-bin quadrature sum. This is first-order uncertainty propagation for correlated inputs, following `JCGM 100:2008, section 5.2 `__. Errors remain conditional on alpha, beta, baseline, gain, and the selected correction values. They do not include calibration or geometry uncertainty, certify a full cross-bin likelihood, or turn several correlated projections into independent fit data. Undefined and defined zero-variance ratios retain their Gaussian-fit omission classification, including after shared-reference cancellation. Large histogram preparation and rotation run in the cancellable **Preparing vector projections…** worker; **Cancel** leaves the source observation unchanged. Asymmetry also checks that retained transverse projections belong to the same preparation cohort; reduce them together rather than mixing independently prepared snapshots. Caller-supplied curves without retained counting observations require the explicit API argument ``assume_independent=True``. Compared with other programs ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In Mantid, `RRFMuon v1 `__ rotates two supplied transverse spectra. Signs, phase, and units must be matched before comparing that coordinate operation: set its real input to :math:`x`, imaginary input to :math:`sy`, and negate its imaginary output to obtain :math:`P'_y`. The comparison uses the v1 documentation associated with the 6.16 development snapshot; no native Mantid RRFMuon execution or shared-detector uncertainty agreement is established here. musrfit 1.12.1's scalar rotating-frame workflow uses different observations and packing. Those operations do not supply a verified drop-in covariance contract for overlapping EMU ratios. Asymmetry's additional counting-noise accounting follows the retained observations and the derivatives above. Independent analytic and seeded-count checks establish the propagation; acquired-file execution establishes operation support, not physical vector geometry or numerical agreement with another program. Fitting P′_x and P′_y --------------------- A rotated subplot is a fit target like any projection: click it, and the Single tab fits exactly the rotated curve drawn — in the run's current frame and period combination — cropped to the fit range. The fit is saved among the run's saved fits under its own projection, beside the run's lab-frame fits (:ref:`saved-single-fits`), and its curve is drawn only on the rotated subplot. Each such fit records the frame its data were rotated in. If you later change a value that alters the rotated curve — :math:`\nu_\mathrm{RF}`, the :math:`B_1` axis, :math:`\varphi_\mathrm{RF}`, the sense, the gain, the baselines the displayed period combination uses, or the combination itself — the fit no longer describes the data on show. Its entry in the **Saved fits** menu then carries ⚠, and while the form still describes it the line under its name reads "Fitted in a different rotating frame — re-run the fit to update it." Fitting again refreshes it. Provenance alone never makes a fit stale: re-typing a value, or an estimate becoming a typed value, changes nothing that was fitted. The Batch tab fits a rotated projection across runs in the same way. A draft fits whatever the plot's fit target shows, and follows it when you click another subplot. Each member is rotated in its own run's frame, and the series records the projection and each member's frame. Runs with no frame are listed in the members, unticked and disabled, with the tooltip "No rotating frame — set ν_RF in the rotating view". A series whose members' frames have since changed shows ⚠ on its pill in the Parameters panel, with "Fitted in a different rotating frame — re-run the fit to update it." on the tooltip, and ⚠ in the Batch tab's series menu. Rotated series are not yet offered to :doc:`joint fits `, where they are listed as ``Rotating-frame series``. No dedicated nutation model exists yet; the usual oscillation components apply. Note what a fit of a damped oscillation at one frequency to :math:`P'_y` measures: an *effective* nutation frequency, set by :math:`\nu_1` together with anything that spreads or shifts it — the coil's ring-up at the start of the pulse, the spread of :math:`B_1` over the sample, and any detuning from resonance, which raises the nutation frequency to :math:`\sqrt{\nu_1^2 + \Delta\nu^2}`. Treat it as :math:`\gamma_\mu B_1/2\pi` only when those effects are known to be small. Worked example -------------- The main rotation and Auto-detect review figures use a synthetic run with invented numbers, made for the documentation: an idealised eight-detector octant polarimeter with directions :math:`(\pm1,\pm1,\pm1)/\sqrt{3}`, in :math:`B_0 = 100` G along :math:`z`, so that :math:`\nu_0 = (\gamma_\mu/2\pi)B_0 \approx 1.355` MHz. - **Red** (period 1) has the RF on at 1.355 MHz with :math:`B_1 \parallel x` and :math:`\nu_1 = 0.2` MHz (a co-rotating :math:`B_1` of about 15 G), damped at 0.2 μs⁻¹ as an inhomogeneous :math:`B_1` would damp it. - **Green** (period 2) has the RF off: the spin stays on :math:`z` and relaxes slowly. - In the lab, the drive's phase at :math:`t_0` is 40°, the labels have sense :math:`-1`, :math:`P_y`'s amplitude is 0.92 of :math:`P_x`'s, and the grouped projections' baselines differ: :math:`(b_x, b_y) = (0.25, -0.10)/\sqrt{3}` % in red and :math:`(0.10, -0.15)/\sqrt{3}` % in green. The geometric factor comes from averaging the unit octant directions within each half-plane. 1. Select the run with the ``P_x``, ``P_y`` and ``P_z`` chips on, and click **Rotating**. The plot asks for :math:`\nu_\mathrm{RF}`. 2. Type ``1.355`` into **ν_RF** and leave **B₁ ∥** on **x**. The rotated curves appear in the default frame, with :math:`\varphi_\mathrm{RF} = 0` — still turning, because the frame's phase and the baselines are wrong. 3. Click **Auto-detect…**. The review proposes sense :math:`-1`, phase near 40°, gain near 0.92, and a baseline pair for each period. The estimates are pre-ticked for application. 4. Click **Apply 4 changes**. :math:`P'_y` now shows the nutation into :math:`+y'`, :math:`P'_x` stays at zero within its errors, and :math:`P_z` swings through zero and back at the same frequency, both damped, as in the figure at the top of the page (bunched ×4 after the rotation). The estimates depend on the selected time range and these invented counting statistics; they do not establish acquired detector geometry. Changing **RG Mode** to **G minus R** would then show the nutation turned over into :math:`-y'`, as described under `Periods`_, without any change to the frame. Assumptions and limitations --------------------------- - **Both transverse projections are needed.** The switch appears only for groupings that declare :math:`P_x` and :math:`P_y`. A single forward–backward pair cannot be rotated exactly; use the filtered :doc:`rotating_frame` display for viewing such data. - **The frame assumes** :math:`B_0 \parallel z` **and orthogonal transverse projections.** The rotation is about the :math:`P_z` axis. Detector pairs that only approximate orthogonal axes (see the limitations in :doc:`vector_polarization`) leave a corresponding error in the rotated components; the gain corrects a difference of scale, not of direction. - **Rotating-wave approximation.** Only the co-rotating half of the linear RF field is static in the frame. The counter-rotating half turns at :math:`2\nu_\mathrm{RF}` in it, and its small effect on the spin dynamics (the Bloch–Siegert shift [4]_) is not part of the picture of a static :math:`B_1`. It is negligible when :math:`B_1 \ll B_0`; at :math:`B_1/B_0` of a few tenths it leaves a small ripple at :math:`2\nu_\mathrm{RF}` on the rotated components and shifts the resonance slightly. - **Auto-detect assumes a nutation from** :math:`+z`. Its :math:`\varphi_\mathrm{RF}` puts the transverse polarisation on :math:`\hat z \times \hat B_1`, which is right for a spin starting along :math:`z` and driven about :math:`B_1`. For other preparations, type :math:`\varphi_\mathrm{RF}` from a reference run measured that way. - **One phase for the whole run.** :math:`\varphi_\mathrm{RF}` is the drive's phase at :math:`t_0`, which holds for every pulse only if the RF source runs continuously and is gated. A source that restarts its phase at each gate drives later pulses about other axes of the frame — a second pulse that turns the spin about :math:`y'` instead of :math:`x'`, for example — and the frame cannot show both at once. - **One frame per loaded run.** A two-period run (red and green) keeps a baseline pair for each period. A file of three or more periods loads as one run per period, and each gets its own one-period frame. - **The rotated components share noise within a bin** (see the error propagation above). Single-component fits use the corrected marginal errors. Shared background estimates can also correlate neighbouring bins. These errors do not establish an independent joint likelihood; rotated series are not offered to joint fits. - **No dedicated models yet.** Nutation, spin-locking and echo models are not in the fit-function library; the effective-frequency caveat under `Fitting P′_x and P′_y`_ applies to the general-purpose ones. References ---------- .. [1] S. J. Blundell, R. De Renzi, T. Lancaster, and F. L. Pratt, *Muon Spectroscopy: An Introduction* (Oxford University Press, Oxford, 2022) — polarisation projections and the rotating reference frame in μSR. .. [2] M. H. Levitt, *Spin Dynamics: Basics of Nuclear Magnetic Resonance*, 2nd ed. (Wiley, Chichester, 2008) — the rotating frame, resonance offset and nutation. .. [3] I. I. Rabi, N. F. Ramsey, and J. Schwinger, Rev. Mod. Phys. **26**, 167 (1954). .. [4] F. Bloch and A. Siegert, Phys. Rev. **57**, 522 (1940). - `WiMDA overview `__ (1.326). - musrfit, `Asymmetry RRF fit `__ (1.12.1). - Mantid, `RRFMuon v1 `__. Related topics -------------- * :doc:`vector_polarization` * :doc:`rotating_frame` * :doc:`data_reduction/period_mapping` * :doc:`fitting` .. figure:: /_generated/screenshots/pooled_frame_safeguard.png :alt: Synthetic pooled G+R with unequal transverse baselines and an explicit unavailable message. :width: 100% Pooled G+R remains available in **Lab** when unequal transverse baselines prevent rotation. The baseline fields show the reason in their tooltips.