---
title: 'Rotating-Frame Acquisition: Methods & Applications'
url: https://www.emergentmind.com/topics/rotating-frame-acquisition
type: topic
---

# Rotating-Frame Acquisition: Methods & Applications

Taken together, the cited literature uses rotating-frame acquisition for measurement configurations in which sensing or readout is carried out in a co-rotating mechanical frame or in an effective rotating spin frame, rather than being treated as a purely static laboratory-frame measurement. This suggests a family of acquisition strategies in which rotation is embedded directly into the forward model, the sensor geometry, the synchronization protocol, or the dynamical observable itself. In the fluid experiment of "Free surface of a liquid in a rotating frame with time-dependent velocity" [1908.04256], the rotating frame is implemented by mounting both the container and the smartphone on a rotating table; in "Photo-acoustic tomography in a rotating measurement setting" [1608.06371], both the optical source and the ultrasonic transducer array are mounted on a rigid rotating frame; and in "Prethermal rotating-frame solid echo in a dipolar nuclear-spin network" [2606.23467], cycle-resolved NMR readout is performed in the rotating frame established by a pulsed spin-lock Floquet drive.

## 1. Rotating-frame acquisition as a measurement architecture

In the mechanically rotating fluid experiment, an LG-G2 smartphone is rigidly fixed to the table beside a narrow prismatic tank, so that the camera and gyroscope measure the evolving free surface and the angular velocity in the same rotating frame. The apparatus is explicitly self-contained: the phone records video of the liquid-air interface through its rear camera, angular velocity $\Omega(t)$ via its built-in three-axis gyroscope logged with the “AndroSensor” app, and a proximity sensor signal used for synchronization [1908.04256].

In the rotating-measurement PAT formulation, the acquisition geometry is defined by a rigid frame rotating about the origin. For each angle $\theta \in [0,2\pi)$, the sensor positions are given by
$$
y^j(\theta)=R(\theta)\,y_0^j,
$$
with the boundary illumination rotated simultaneously according to
$$
f_\theta(x)=f_0\bigl(R(-\theta)x\bigr), \qquad x\in \partial\Omega .
$$
A consequence stated explicitly in the formulation is that one no longer has the option to reconstruct a map of sonic emission corresponding to a given optical illumination, so the optical and acoustic stages must be combined into a single inverse problem [1608.06371].

In the prethermal NMR setting, the rotating frame is not produced by mechanical motion but by a periodic train of spin-locking pulses. After optical hyperpolarization and a conventional $(\pi/2)_x$ tipping pulse, the system is driven by a Floquet sequence of $\vartheta_x$ pulses interleaved with short acquisition windows, and readout windows sample the transverse precession in the rotating frame, reconstructing the cycle-resolved quadratures
$$
S_x(nT_{FC})=\langle I_x\rangle,\qquad S_y(nT_{FC})=\langle I_y\rangle .
$$
Because $T_{FC}$ is much shorter than the typical dipolar timescale $(J^{-1}\simeq 1.5\,\mathrm{ms})$, the pulsed train realizes a Floquet-averaged Hamiltonian and quasi-continuous readout in the rotating frame without repeated re-initialization [2606.23467].

These realizations show that rotating-frame acquisition is not tied to a single discipline. The common feature is that rotation changes what is directly measured and therefore changes the appropriate forward model.

## 2. Continuum-mechanical formulation in a co-rotating frame

For the liquid free-surface problem, the theoretical model is written directly in the rotating frame of angular velocity $\Omega(t)$ about the vertical $z$-axis. In steady rigid-body rotation, the velocity and centripetal acceleration are
$$
\vec u=\Omega\,r\,\hat e_\theta,\qquad \vec a=-\Omega^2 r\,\hat e_r .
$$
The stationary momentum balance in the non-inertial frame is
$$
\nabla p=\rho(\vec g-\vec a)
\quad \Rightarrow \quad
\frac{\partial p}{\partial r}\hat e_r+\frac{\partial p}{\partial z}\hat e_z
=
\rho[-\Omega^2 r\,\hat e_r-g\,\hat e_z] .
$$
Integration yields the pressure field
$$
p(r,z)=-\rho g z+\frac12 \rho \Omega^2 r^2+C,
$$
where $C$ is fixed by the total volume [1908.04256].

At the free surface, $p(r,z')=p_{\mathrm{atm}}$, so the interface becomes parabolic:
$$
z'(r)=H-\frac{\Omega^2}{2g}\left(\frac{L^2}{12}-r^2\right).
$$
Equivalently, relative to the mid-plane,
$$
\Delta h(r,t)\equiv z'(r)-H=+\frac{(\Omega(t))^2r^2}{2g}-\mathrm{const.}
$$
The vertex height at $r=0$ is
$$
z_v=H-\frac{\Omega^2L^2}{24g}.
$$
The assumptions are explicit: inviscid, incompressible fluid in rigid-body rotation, transients neglected; a narrow rectangular prism so end effects are minimal; atmospheric pressure at the free surface; and conservation of volume [1908.04256].

The PAT rotating-frame formulation is likewise encoded at the level of the forward model. The diffusion problem for the fluence $u_\theta(x)$ is
$$
-\nabla\cdot(\nabla u_\theta)+\mu_a(x)\,u_\theta(x)=0,\qquad
u_\theta(x)=f_0\bigl(R(-\theta)x\bigr)\ \text{on }\partial\Omega,
$$
and the initial pressure field is
$$
H_\theta(x)=\mu_a(x)\,u_\theta(x).
$$
The acoustic propagation is then governed by
$$
(\partial_t^2-c^2(x)\Delta_x)\,p(x,t;\theta)=0,
$$
with initial data $p(x,0;\theta)=H_\theta(x)$ and $\partial_t p(x,0;\theta)=0$. Measurements are defined directly on the rotated detector positions through
$$
A[\mu_a](\theta,t;y_0^j)=p\bigl(R(\theta)y_0^j,t;\theta\bigr) .
$$
Here, rotating-frame acquisition is not merely a geometric convenience; it is part of the operator $A$ itself [1608.06371].

In the NMR case, the corresponding rotating-frame structure appears in the effective Floquet Hamiltonian. To leading order in the Magnus expansion, the periodic $\pi_x$-pulse train leaves an effective secular dipolar Hamiltonian aligned along $x$,
$$
H_{\mathrm{eff}}\simeq H_F^{(0)}=
-\frac14\sum_{i<j} d_{ij}\bigl(3I_i^xI_j^x-\mathbf I_i\cdot \mathbf I_j\bigr),
$$
or equivalently
$$
H_{\mathrm{eff}}=-\sum_{i<j} d_{ij}D^{xx}_{ij},
$$
where $D^{xx}_{ij}=(3I_i^xI_j^x-\mathbf I_i\cdot \mathbf I_j)/4$. Under $H_{\mathrm{eff}}$, the total transverse magnetization $I^x$ is approximately conserved, leading to a prethermal plateau [2606.23467].

## 3. Instrumentation, synchronization, and parameter extraction

The fluid experiment provides an explicit example of rotating-frame data acquisition in which synchronization is itself part of the measurement design. The acquisition protocol begins by starting video and sensor logging with a few seconds at zero rotation. A “lens-cover” event, achieved by covering camera and proximity sensor simultaneously, produces a dark frame and a proximity drop and therefore marks a common time fiducial $t=t_{\mathrm{sync}}$ for the video and gyroscope streams. The gyroscope is factory-calibrated, with sampling rate $\approx 50\,\mathrm{Hz}$ and resolution $\sim 0.01\,\mathrm{rad/s}$, and the camera orientation is fixed so that one camera axis is horizontal and aligned with $r$ [1908.04256].

The same experiment also makes the extraction procedure explicit. Individual frames are extracted at selected times $t_i$, with $15$ frames covering the range of $\Omega$. In each frame, Tracker software is used to pick $\sim 8$ points along the liquid-air interface, and these points are fitted to
$$
y=Ax^2+Bx+C,
$$
where $x$ is the horizontal pixel coordinate and $y$ is vertical. The concavity $A$ is proportional to $\Omega^2/2g$, and the vertex height in pixel units is
$$
y_v=-\frac{B^2}{4A}.
$$
A calibration ruler in the field of view provides the pixel-to-centimeter scale in horizontal and vertical directions, and each chosen frame at $t_i$ is associated with $\Omega_i=\Omega(t_i)$ after bias subtraction and conversion to $\mathrm{rad/s}$ [1908.04256].

The uncertainty budget is also resolved operationally. The sources listed are gyroscope resolution $(\pm 0.01\,\mathrm{rad/s})$, frame extraction timing $(\pm 0.02\,\mathrm{s})$, pixel resolution $(\pm \tfrac12\ \mathrm{pixel}\Rightarrow \pm 0.1\,\mathrm{mm})$, and parabolic-fit statistical error in $A$ and $B$. For $A_{\mathrm{exp}}$ versus $\Omega^2$, the slope uncertainty comes from linear regression; for vertex height $y_v$ versus $\Omega^2$, the uncertainty combines fitting and pixel scale [1908.04256].

In the PAT setting, synchronization enters differently: the acquisition is indexed by angle, time, and sensor, and discretization is carried out as $\theta_k=k\Delta\theta$, $t_\ell=\ell\Delta t$, with rotated sensor positions generally off-grid. The implementation therefore uses bilinear interpolation in $2$D or trilinear interpolation in $3$D to evaluate $p(x,t)$ at $y^j(\theta_k)$, and in the adjoint pass the residuals are scattered back to the grid via the same interpolation stencil under the inverse rotation $R(-\theta_k)$. This makes interpolation part of the acquisition operator rather than a post hoc visualization step [1608.06371].

In the NMR experiment, acquisition is cycle resolved by construction. The pulsed spin-lock not only generates the rotating frame but also leaves short acquisition windows interleaved with the drive, allowing continuous interrogation of the same prethermal manifold. A practical consequence stated explicitly is that a single experimental shot can map $S_{\mathrm{echo}}(\tau)$ up to many tens of milliseconds, avoiding the need for repeated tip-and-acquire cycles [2606.23467].

## 4. Inverse problems and reconstruction under rotation

In rotating-measurement PAT, the principal methodological consequence of the acquisition geometry is that the conventional two-step pipeline breaks down. The reconstruction target is the absorption coefficient $\mu_a(x)$, recovered by minimizing
$$
J(\mu_a)=\frac12\sum_{j=1}^{N_s}\int_0^{2\pi}\int_0^T
\bigl|A[\mu_a](\theta,t;y_0^j)-d_{\mathrm{obs}}(\theta,t;y_0^j)\bigr|^2\,dt\,d\theta
+\alpha R(\mu_a),
$$
with either Tikhonov regularization,
$$
R(\mu_a)=\frac12\|\nabla \mu_a\|_{L^2(\Omega)}^2,
$$
or Total Variation,
$$
R(\mu_a)=\|\nabla \mu_a\|_{L^1(\Omega)} .
$$
The formal stationarity condition is
$$
0=A'[\mu_a]^*(A[\mu_a]-d_{\mathrm{obs}})+\alpha R'(\mu_a) .
$$
This is the analytic statement that acquisition and inversion must be handled jointly [1608.06371].

The adjoint is derived by introducing the residual
$$
r(\theta,t;y_0)=A[\mu_a](\theta,t;y_0)-d_{\mathrm{obs}}(\theta,t;y_0),
$$
and the time-reversed adjoint field $q(x,t;\theta)$ satisfying the backward wave equation with terminal data driven by the residual injected at the rotated sensors. The resulting adjoint operator is
$$
(A^*r)(x)=\int_0^{2\pi}\int_0^T u_\theta(x)\,q(x,t;\theta)\,dt\,d\theta,
$$
with the volume-preserving property of the rotation used explicitly through the Jacobian equal to $1$. A steepest-descent update then reads
$$
\mu_a^{n+1}(x)=\mu_a^n(x)-\tau\Bigl[
(A^*(A[\mu_a^n]-d_{\mathrm{obs}}))(x)+\alpha R'(\mu_a^n)(x)
\Bigr] .
$$
The forward operator, the adjoint, and the optimization are therefore all rotation dependent [1608.06371].

The numerical implementation described for a typical $2$D phantom on $\Omega=[-10,10]^2\,\mathrm{mm}^2$ with a $200\times 200$ mesh reports resolution $\sim 0.2\,\mathrm{mm}$, convergence of the residual $\|A\mu-d\|$ to $\approx 10^{-6}$ in $20$–$30$ iterations for TV regularization with $\alpha\approx 10^{-3}$, stable recovery up to $5\%$ additive Gaussian noise, and per-iteration cost consisting of one forward diffusion solve, $N_\theta$ forward acoustic solves, and $N_\theta$ adjoint acoustic solves, totaling $\approx 60\,\mathrm{s}$ per iteration on a single GPU when $N_\theta=60$ [1608.06371].

The fluid experiment uses a much simpler inference problem, but it still exemplifies model-based parameter estimation under rotation. The experimentally extracted parabola concavity and vertex height are each regressed against $\Omega^2$, producing a direct test of the rotating-frame model. The reported plot of $A_{\mathrm{exp}}$ against $\Omega^2$ is linear, with measured slope $20.16(4)\,\mathrm{m\cdot rad^2/s^2}$ and theoretical $2g=19.62\,\mathrm{m/s^2}$, corresponding to agreement within $3\%$. The plot of $\Delta y_v$ versus $\Omega^2$ is also linear, with measured slope $-0.27(1)\,\mathrm{mm\cdot s^2/rad^2}$ and theoretical $-L^2/(24g)=-0.2655\,\mathrm{mm\cdot s^2/rad^2}$; the intercept at $\Omega=0$ gives $H=7.72(3)\,\mathrm{cm}$, in agreement with direct measurement $7.6(2)\,\mathrm{cm}$ [1908.04256].

## 5. Rotating-frame readout in prethermal NMR

The rotating-frame NMR experiment uses a pulsed spin-lock Floquet drive to stabilize a long-lived prethermal manifold of transverse $^{13}\mathrm C$ magnetization in diamond, and then probes that manifold by cycle-resolved inductive readout. If the prethermal $x$-magnetization is tipped into the $y$–$z$ plane with a $(\pi/2)_y$ pulse at $t=0$ and the Floquet drive is continued, the rotating-frame free-induction signal is described by
$$
S_y(t)=\langle I_y(t)\rangle \simeq A\cos(\omega_{\mathrm{pre}}t+\phi)e^{-(t/T_2')^\eta},
$$
with $\eta \simeq \tfrac12$ and fitted decay time
$$
T_2' \simeq 13.4\,\mathrm{ms}.
$$
By contrast,
$$
S_x(t)\simeq \mathrm{const}
$$
up to the prethermal heating time
$$
T_{\mathrm{pre}}\simeq 90\,\mathrm{s}\gg T_2' .
$$
The rotating frame is therefore both the acquisition frame and the dynamical frame in which approximate conservation becomes visible [2606.23467].

Within this same driven manifold, a single echo pulse $(\alpha)_y$ applied after a delay $\tau$ produces a rotating-frame solid echo. After evolution for $\tau$, the pulse $(\alpha)_y$ is applied; after a further evolution of duration $\tau$ under the same pulsed spin-lock, a revival is observed at
$$
S_{\mathrm{echo}}(2\tau)=\langle I_y(2\tau)\rangle>0 .
$$
The minimal two-spin picture attributes this to micromotion-induced transfer between operator subspaces. For a single pair with $d=d_{12}$, one has stroboscopically
$$
I_1^yI_2^y \to \cos(d\tau/2)\,I_1^yI_2^y
+\sin(d\tau/2)\,(I_1^xI_2^z+I_1^zI_2^x) .
$$
The $(\alpha)_y$ pulse rotates these manifolds differently: the $I^y\otimes I^y$ subspace picks up a factor $\cos\alpha$ and can be inverted, whereas the $I^x\otimes I^z+I^z\otimes I^x$ subspace is either left unchanged or mixed into non-refocusable operators. Maximum contrast occurs near $\alpha \simeq \pi/2$, with a small experimental offset $(\approx 5^\circ)$ attributed to pulse-imperfection-induced phase shifts [2606.23467].

For an ensemble of dipolar-coupled spins, the echo amplitude is approximated as
$$
S_{\mathrm{echo}}(\tau,\alpha)\simeq \prod_{i<j}\bigl|\cos(d_{ij}\tau/2)\bigr|^p \sin\alpha,
$$
where $p\le 2$ depends on details of micromotion mixing. The many-body echo envelope is fitted to
$$
S_{\mathrm{echo}}(\tau)\propto e^{-(\tau/T_{\mathrm{echo}})^\eta},
\qquad \eta \simeq \tfrac12,\qquad T_{\mathrm{echo}}\simeq 23.4\,\mathrm{ms}.
$$
Exact simulations of $10^2$ random spin pairs with $|d_{ij}|$ uniformly in $[0,5\,\mathrm{kHz}]$ under the same pulse train reproduce a strong, symmetric revival at $2\tau$ for ideal $\vartheta=\pi$ pulses and a reduced echo amplitude for $\vartheta=\pi\pm 2^\circ$ errors [2606.23467].

The practical acquisition implications are explicit. NV-center optical pumping yields $^{13}\mathrm C$ polarizations $\gtrsim 10^{-3}$, giving cycle-resolved inductive signals $S/N \gtrsim 100$ in a single scan. The condition $T_{FC}\lesssim 0.1\,\mathrm{ms}$ ensures $J T_{FC}\ll 1$, prethermalization is robust, and the prethermal plateau persists for a duration stated to be four orders of magnitude longer than the conventional lab-frame $T_2$ [2606.23467].

## 6. Comparative structure, limitations, and recurring misconceptions

The three cases describe distinct but structurally related forms of rotating-frame acquisition.

| Domain | Rotating-frame mechanism | Acquired quantity |
|---|---|---|
| Fluid free surface | Smartphone fixed to rotating table | Video of interface and gyroscope $\Omega(t)$ |
| Photo-acoustic tomography | Optical source and US array on rigid rotating frame | $d(\theta,t,y_0)=p(R(\theta)y_0,t;\theta)$ |
| Prethermal NMR | Pulsed spin-lock Floquet drive | Cycle-resolved $S_x(nT_{FC})$, $S_y(nT_{FC})$ |

A first recurring misconception is that rotation can always be handled as a coordinate change applied after acquisition. The PAT formulation explicitly rules this out: because the light source rotates at the same time as the ultrasound measurements are acquired, one no longer has the option to reconstruct a map of sonic emission corresponding to a given optical illumination, and the two steps are combined into one [1608.06371].

A second misconception is that rotating-frame readout necessarily implies complete reversal of dephasing. The NMR result is more specific: analytical arguments and toy-model simulations attribute the revival to Floquet micromotion that transfers coherences between operator subspaces, so that only a subset of the many-body dephasing dynamics is inverted by the $y$ pulse [2606.23467].

A third misconception is that co-rotating acquisition automatically yields a fully dynamical fluid model. In the liquid-surface experiment, the theoretical comparison is carried out under the assumptions of inviscid, incompressible fluid in rigid-body rotation with transients neglected. The experiment varies angular speed in small manual steps and compares selected frames with the steady parabolic prediction, so the success of the comparison validates that specific model and data-processing pipeline rather than an unrestricted time-dependent hydrodynamic description [1908.04256].

These limitations are integral to the topic. Rotating-frame acquisition is not a single universal algorithm; it is a measurement strategy whose meaning depends on whether rotation enters through sensor mounting, illumination geometry, effective Hamiltonian engineering, or all three simultaneously.

Source: https://www.emergentmind.com/topics/rotating-frame-acquisition