Microwave Frequency Fiber Interferometer

• MFFI is an interferometric sensing technique that modulates a microwave signal onto an optical carrier in telecom fibers to measure integrated phase changes.
• It detects minute fiber delay variations linked to strain, temperature, and pressure, enabling seismic, tidal, and ocean wave monitoring.
• The method leverages standard telecom hardware for low-cost, scalable deployment while offering integrated measurements with limited spatial resolution.

Searching arXiv for the cited MFFI papers to ground the article in the relevant literature. Microwave Frequency Fiber Interferometry (MFFI) is an interferometric sensing technique in which a stable microwave carrier is modulated onto light in standard telecom fiber and the phase of that microwave modulation is used to detect tiny changes in the fiber’s propagation delay. Those delay changes reflect mechanical strain, pressure, temperature, and, in deployed terrestrial or subsea cables, geophysical and oceanographic signals including earthquakes, tides, storms, and ocean waves. Across the literature, MFFI is characterized as an “integrated” fiber-optic sensor when it yields a single time series for an entire link, and as a per-span sensor when timing structure or loop-back architecture allows separate link segments to be resolved (Karydis et al., 20 Jun 2026).

1. Definition and conceptual scope

Microwave Frequency Fiber Interferometry operates by imposing a microwave or RF signal, typically around 10 GHz, on an optical carrier, transmitting it through fiber, and comparing the received microwave phase with a local reference. In this usage, the fiber acts as one arm of an RF interferometer, and the measured observable is the phase change induced by delay fluctuations in the optical path. In the 2021 terrestrial demonstration, this was described explicitly as a “lumped” fiber interferometer based on transfer of a stable microwave frequency over a telecom fiber and back, with the mixer output proportional to the instantaneous phase difference between the two microwave tones (Bogris et al., 2021).

The distinction between MFFI and other fiber-sensing modalities is central. Distributed Acoustic Sensing recovers a spatially distributed strain field with meter-scale resolution using coherent optical backscatter, but it is difficult to deploy over thousands of kilometers in traffic-carrying submarine cables and requires high-complexity interrogators. Optical interferometric sensing directly measures optical phase at hundreds of THz and requires very low-linewidth lasers and high-performance electronics. Polarization-based sensing tracks state-of-polarization changes and is simpler but less sensitive than phase-based methods. MFFI occupies an intermediate position: it retains the phase sensitivity of interferometric methods while moving the phase measurement into the microwave domain, where off-the-shelf oscillators have sub-Hz linewidths and very high spectral purity (Karydis et al., 20 Jun 2026).

Within the literature, MFFI is also framed as complementary to DAS rather than a replacement. Theoretical and experimental comparison work emphasizes that MFFI yields a single time series representing integrated deformation of the entire fiber, whereas DAS yields many local strain-rate channels. The same comparison shows that integrated DAS strain-rate can be quantitatively compared with the MFFI response, and that MFFI sensitivity depends strongly on points of curvature and at cable endpoints (Bowden et al., 2022).

2. Physical principle and observables

The core relation used throughout the MFFI literature is the mapping from group delay to microwave phase. If the propagation delay is $\tau(t)$ and the microwave frequency is $f_\text{RF}$ or $f_\text{mw}$, the received phase is written as

$\phi(t)=2\pi f_\text{mw}\tau(t)$

or, equivalently,

$\phi = 2\pi f_\text{RF}\frac{n_g L}{c}.$

A perturbation of fiber length and refractive index therefore produces a phase change. One formulation given in the terrestrial work is

$$\Delta \phi = 2\pi f_\text{RF}\frac{n_g \Delta L + L \Delta n_g}{c},$$

while the subsea per-span formulation emphasizes

$$\Delta \phi(t)=2\pi f_\text{mw}\,\Delta\tau(t)=2\pi f_\text{mw}\frac{n_g}{c}\Delta L(t).$$

In both forms, the measured phase is a proxy for optical-path variation and therefore for integrated strain, temperature, pressure, and related perturbations (Bogris et al., 2021).

For seismic and geophysical interpretation, the phase is commonly transformed into delay, then length change, then strain. The subsea per-span work gives the conceptual sequence

$$\phi_k(t)=\arctan\!\bigg(\frac{Q_k}{I_k}\bigg),\quad \Delta \tau_k(t)=\frac{\Delta \phi_k(t)}{2\pi f_\text{mw}},\quad \Delta L_k(t)=\frac{c}{n_g}\Delta \tau_k(t),\quad \epsilon_k(t)=\frac{\Delta L_k(t)}{L_{0,k}}.$$

In the 2021 terrestrial work, the phase was unwrapped and then differentiated twice to form an “acceleration” proxy for comparison with accelerometers. In later calibration work, strain acceleration rather than strain rate was selected as the primary observable because it had higher temporal smoothness and better consistency within distance clusters (Karydis et al., 20 Jun 2026).

A further theoretical refinement links MFFI to distributed measurements. The integrated-system analysis writes the measured phase rate as proportional to the line integral of axial strain rate along the cable,

$$\partial_t\phi(t)\approx \frac{\omega}{c_0}\,\partial_t\int_0^L \tilde{n}(s)\,\epsilon(s,t)\,ds.$$

That same work derives an equivalent form showing sensitivity concentrated at endpoints and where the tangent vector changes along the cable, thereby connecting the integrated response to cable geometry and curvature. This suggests that the effective response of an MFFI is not spatially uniform even when the readout is a single scalar time series (Bowden et al., 2022).

3. Instrument architectures and implementations

Across reported implementations, MFFI hardware is intentionally close to standard microwave-over-fiber and telecom practice. A typical architecture comprises a CW telecom laser, a LiNbO$_3$ Mach–Zehnder modulator driven by a phase-locked microwave source, a fiber link with loopback or reflection, a fast photodiode, a mixer or I/Q demodulator, and low-speed digitization. The 52 km Attika experiment used a CW DFB laser at 1550 nm, a 10 GHz PLL microwave generator, an EDFA to boost launch power to about 9 dBm, a 10 GHz photodiode, a microwave mixer, a 10-bit ADC at 100 Hz, and a looped-back commercial WDM fiber path of approximately $f_\text{RF}$0 km (Bogris et al., 2021).

Submarine deployments retained the same basic architecture while adapting it to shore-station operation. The Cephalonia–Ithaca experiment used a 10 GHz RF oscillator, an optical transmitter, a photodetector, and a 16-bit ADC sampling at 800 samples/s, with a passive loop-back at the far end so that only one active shore station was required. The total one-way fiber length was 15.6 km, including terrestrial and submarine sections, and the effective MFFI path was approximately 31.2 km out and back. The paper reports phase resolution of about 0.02 mrad, approximately 30 times better than the earlier terrestrial demonstration (Bogris et al., 7 Apr 2025).

The 2026 per-span North Atlantic implementation introduced a pulsed round-trip RF interferometer on an operational repeatered cable. A tunable laser at about 1565 nm was intensity-modulated by a Mach–Zehnder modulator driven by a 9.8 GHz PLL, gated by an acousto-optic modulator into approximately 0.5 ms pulses, amplified, and launched into the IRIS cable. The return path used the existing High-Loss Loop Back configuration so that, from Galway, the cable behaved as 17 loops with roughly 100 km spacing and a total round-trip time of about 17.5 ms. Returned pulses were detected by a 10 GHz photodetector, coherently down-converted in an analog I/Q mixer using the same 9.8 GHz oscillator, digitized at 1 Msps, and decimated in FPGA logic to about 16 ksps per channel (Karydis et al., 20 Jun 2026).

This per-span architecture is distinct from a conventional Mach–Zehnder or Michelson optical interferometer. In the authors’ description, the “interferometer” is formed by phase comparison between the transmitted microwave reference and the received microwave modulation after propagation around the loop. That design also avoids the coherence demands of optical-frequency interferometry, because the phase comparison is in the microwave domain rather than at the optical carrier frequency (Karydis et al., 20 Jun 2026).

4. Integrated links, per-span sensing, and calibration

The earliest reported MFFI deployments yielded one integrated measurement per link. In the Athens comparison study, the MFFI followed the same path as a DAS system and then was looped back to the starting location, producing a single integrated strain-rate record over approximately 24 km of deployed telecom fiber. The paper showed that summing the DAS channels produced a composite signal that agreed with the MFFI output to within pre-event noise in both low-frequency and higher-frequency bands, thereby validating the interpretation of MFFI as an integrated line measurement (Bowden et al., 2022).

Per-span sensing modifies that integrated paradigm by exploiting timing structure in repeatered submarine infrastructure. In the 1,770 km Ireland–Iceland experiment, the AOM generated pulses separated by roughly 1 ms, while each loop return incurred about 1 ms of additional round-trip delay. The FPGA therefore separated the pulse train into 17 channels, each associated with a different span between repeaters. The authors note that later pulses include contributions from multiple spans, but in the implemented HLLB configuration each channel is associated with a distinct approximately 100 km span, following the known routing and timing of the cable (Karydis et al., 20 Jun 2026).

Quantitative seismic calibration was addressed in the 2025 Cephalonia–Ithaca study. That work treated MFFI as a long, integrated strain meter and developed a data-driven calibration framework for magnitude and rough distance estimation. Continuous hourly recordings were screened against the NOA earthquake catalogue; only earthquakes with ML $f_\text{RF}$1 were retained, producing a set of 75 events. The working observable was strain acceleration filtered with a zero-phase Butterworth bandpass filter from 1 to 7 Hz, and one-minute event windows were extracted for analysis (Deligiannidis et al., 28 Jul 2025).

Distance estimation did not use travel times, because P- and S-wave arrivals were not consistently visible in the MFFI data. Instead, events were grouped into four distance clusters—0–10 km, 10–20 km, 20–40 km, and $f_\text{RF}$2 km—from the geometric center of the fiber, and normalized cross-correlation with representative templates showed that intra-cluster correlations were significantly higher than inter-cluster correlations. Magnitude estimation used the empirical relation

$f_\text{RF}$3

with fitted coefficients

$f_\text{RF}$4

Tabled error metrics gave NMSE and standard deviation by distance cluster, and the study states that errors are mostly below 0.25 magnitude units in the main distance range (Deligiannidis et al., 28 Jul 2025).

5. Demonstrated geophysical and oceanographic observations

MFFI has been demonstrated on both terrestrial and submarine telecom infrastructure as a sensor for earthquakes, tides, storms, ocean waves, and microseisms. The 2021 Attika experiment showed efficient detection of seismic waves from distant epicenters greater than 400 km on commercially deployed fiber. Two events were highlighted: the 11 July 2021 Thebes earthquake with $f_\text{RF}$5 at approximately 47 km from the fiber, and the 12 October 2021 Crete earthquake with magnitude 6.3 at approximately 410 km. In both cases, the fiber-derived acceleration proxy showed high correlation with a nearby accelerometric station, with S-waves clearly visible and, for the more distant event, a pronounced coda (Bogris et al., 2021).

Submarine operation extended MFFI to marine geophysics. On the Cephalonia–Ithaca cable, the 2025 conference paper reports operation for two months and detection of more than 110 earthquakes with magnitudes between 1.5 and 3.0, including representative events of magnitude 2.5, 1.6, and 2.1. For those examples, P and S arrivals were clearly visible in the MFFI data and aligned with the FSK.HP seismic station, while comparison with DAS showed synchronized coda waves and similar spectrogram structure. The same experiment identified a strong 12-hour peak attributed to tidal waves affecting the submarine part of the cable, a strong 24-hour peak attributed mainly to temperature changes affecting the terrestrial sections, and a 5 dB higher strain-rate PSD on a stormy day than on a calm day in the 0.04–0.1 Hz band associated with sea-wave frequencies (Bogris et al., 7 Apr 2025).

The most extensive subsea demonstration to date is the 1,770 km IRIS deployment. Operated continuously since November 2025, it resolved semi-diurnal tides across all 17 spans, with the strongest amplitudes in spans 3, 4, and 8 near steep bathymetric gradients in the Rockall Basin. It also recorded enhanced secondary microseism energy in the 0.1–0.3 Hz band during North Atlantic storm periods, with correspondence to WAVEWATCH III hindcast significant wave height maps. For teleseismic events, the paper presents records from the 8 December 2025 Hokkaido earthquake of magnitude 7.6 and the 1 April 2026 Indonesia earthquake of magnitude 7.4. Channel 17, near Iceland, shows surface waves, indications of S waves and SSS, and visible surface-wave dispersion in spectrograms; comparison with the BORG seismic station shows strong agreement for the surface-wave phase (Karydis et al., 20 Jun 2026).

Taken together, these results establish several regimes of demonstrated MFFI sensing.

Regime	Demonstrated observations	Representative paper
Terrestrial lumped sensing	Local and distant earthquakes on commercial fiber	(Bogris et al., 2021)
Short submarine link	Micro-earthquakes, tides, sea-wave band changes	(Bogris et al., 7 Apr 2025)
Long subsea per-span sensing	Tides, storms, microseisms, teleseismic earthquakes on 17 spans	(Karydis et al., 20 Jun 2026)

6. Advantages, limitations, and research directions

The literature consistently presents MFFI as low-cost and scalable because it uses standard tunable telecom lasers rather than ultra-narrow-linewidth metrology lasers, commercial 9.8–10 GHz PLL oscillators, Mach–Zehnder modulators, off-the-shelf EDFAs, photodiodes, analog I/Q mixers, low-speed ADCs, and lightweight embedded control. Relative to OFDR-based or optical-frequency interferometric methods, the complexity is shifted from optical to microwave/RF hardware, simplifying deployment and real-time processing. The method is also compatible with WDM operation and, in the subsea deployments, either used a dedicated dark fiber or a C-band high-loss loop-back wavelength that did not interfere with live traffic (Karydis et al., 20 Jun 2026).

At the same time, MFFI has intrinsic limitations. In its integrated form it provides no meter-scale spatial resolution, and even per-span architectures resolve only coarse cable segments rather than continuous distributed strain. Sensitivity depends on link geometry, particularly curvature and endpoint motion, and the integrated response can suppress small events on long straight sections. The 2025 calibration study further shows that distance estimation presently relies on waveform clustering rather than continuous kinematic inversion, because P- and S-wave arrivals are not consistently visible. The empirical magnitude model is not yet a full physics-based transfer function from cable strain to earthquake source parameters (Bowden et al., 2022).

Instrumental and environmental limitations also recur across the literature. The 2021 Attika study identified ADC quantization noise, optical noise from EDFAs, thermal fluctuations, anthropogenic vibrations, and possible polarization-induced fading as important practical constraints. The 2026 IRIS study mentions inline EDFA gain and noise, chromatic-dispersion-induced fading over many loops, and long-term sensitivity drift from environmental noise and aging. In the IRIS configuration, tuning the microwave from exactly 10 GHz to 9.8 GHz was used specifically to mitigate chromatic-dispersion-induced fading over 17 loops (Bogris et al., 2021).

Future directions stated in the papers include extending MFFI toward distributed sensing and comparison with DAS, building compact “black-box” units for wider deployment, improving sensitivity through microwave frequency, span length, and signal processing, and integrating MFFI with DAS, polarization sensing, or optical interferometry for richer physical interpretation. The calibrated seismic study explicitly states that the long-term goal is early warning and that the work “lays the foundation for implementing MFFI in real-time EEW architectures, in regions with sparse seismic instrumentation.” A plausible implication is that MFFI will remain most valuable where long-haul terrestrial and submarine telecom infrastructure provides coverage beyond the practical range of DAS and beyond the density of conventional seismometer networks (Deligiannidis et al., 28 Jul 2025).