Rotary-saturation based method, such as spin-locked rotary excitation (SLOE) ^[1], detects the shift of magnetization in the transverse plane, and has demonstrated superior sensitivity in detecting oscillatory neuronal currents ^[1-3] and magnetic nanoparticles ^[4]. A key assumption in these methods is that the oscillating field should be precisely triggered or in phase with the spin-lock periods in order to create robust signal change in the time series. This requirement is often not satisfied in practice, especially in vivo where the signal of interest may process an arbitrary phase that renders the detection sensitivity. In this work, we analyzed the time series of rotary-saturation from an oscillatory view, based on which a new frequency domain detection method is proposed and verified for improved sensitivity at low field.In the view of SLOE, a doubly rotating frame was introduced ^[1,2] with the direction of lock field (B_SL) as the longitudinal direction as illustrated in Fig. 1. During the spin lock period, if the frequency of oscillatory fields (B_osc) matches that of the spin-lock field and its phase is triggered (the initial phase offset of B_osc needs to be known a priori), B_osc will align with x’’ (Fig. 1a) and flip the magnetization M (shown in green) a small angle of from the z’’ axis. Then an increase of magnetization along y’’ axis is used as the contrast (M_contrast). However, if the initial phase of B_osc is not known so that is not triggered, then the effective B_osc that flips the magnetization will be a cosinoidal projection (Fig. 1b), resulting a changing flip angle for each TR, which will be reflected as oscillating signal in the time series. In this case, M_contrast could be considerably decreased leading to deteriorated detection sensitivity. Instead, given its oscillating nature, a spectral peak will be present and hence we may detect the oscillatory field by detecting the spectral peak in the frequency domain. The SLOE method was implemented on a 3.0 T whole body scanner (GE Discovery 750). Phantom study was used to verify the detection of the non-triggered oscillatory field, where an oscillating current loop to create subtle oscillatory fields. The single coil copper wiring was wound around a plastic tube filled with NiCl₂ solution. A function generator was connected to the coil to generate oscillatory field in the direction of static field. A block-designed experiment was used with current on and off. Identical single slice acquisition parameters were used: TE = 24 ms, slice thickness = 6 mm. Oscillatory current of 100 Hz was used with corresponding the BSL of 2.35 μT. The magnetic field at the center of the loop was calculated due to Biot-Savart law. To mimic non-triggered phase, a TR of 1002 ms that was not dividable by the period of oscillatory currents was used, so that phases of oscillatory fields will accumulate over consecutive TRs. Conventional magnitude derivation based SLOE detection ^[1] and the newly proposed spectral detection, which performs a paired t-test on the power of different frequency components, were performed and compared. A ROI containing 2 x 2 voxels at the center of the loop was use to depict the time series signal. With an oscillatory field of 1 nT ( Fig. 2a), the time series signal showed significantly higher level of oscillation in the ON block as compared to OFF block; whereas with an oscillatory field of 0.25 nT (Fig. 2c), the level of oscillation of the time series signal dropped significantly due to the phase inconsistency, and traditional derivation test may fail to for such detection. On the other hand, the spectral component (Fig. 2b and Fig. 2d) showed obvious peaks at 0.2 Hz with both levels of oscillatory field used allowing for robust detection. The increased sensitivity of spectral detection at lower field can be seen in Fig. 3, where the activation map was clearly more complete with the spectral method at lower field.Up to date, SLOE has been reported to be the most sensitive oscillatory magnetic field detection method that features a sub-nanotesla detection sensitivity. However, the original rotary saturation methods that are based on signal level change may fail if the phase of the oscillatory field is not precisely known, which is usually the case in practice. In this work, we further proposed and verified a spectral detection method that may overcome this limitation by detecting the spectral peak related to the phase offset. Phantom experiment demonstrated significantly improved detection sensitivity at lower field.