Fiber-Optical Analogue Systems

• Fiber-optical analogue is a family of fiber-based systems that use inherent optical dynamics for analog transport, processing, and modeling complex wave phenomena.
• These systems enable experimental emulation of physical paradigms, such as analogue gravity via Kerr nonlinearity and RF signal equalization in coherent interconnects.
• They offer practical benefits like reduced electronic conversion, energy efficiency, and unique platforms for studying nonlinear wave dynamics and mesoscopic analogies.

Searching arXiv for relevant papers on fiber-optical analogue systems and analog fiber-optic signal processing. Fiber-optical analogue denotes a family of fiber-based systems in which the optical fiber is used as an analogue medium rather than only as a passive transmission line. In one sense, the term refers to non-digitized transport and processing: RF waveforms, detector pulses, or coherent-receiver equalization are handled in the analog domain and mapped directly onto optical carriers or analog integrated circuits. In another sense, it refers to physical analogues in which fiber propagation reproduces the mathematics of other wave systems, including curved spacetime, chaotic cavities, and Anderson-localizing media. Across these uses, the common element is that propagation, scattering, Kerr nonlinearity, and photodetection are treated as the operative dynamics of the system itself (Moreno-Ruiz et al., 2021, Nambath et al., 2019, Redding et al., 2023, Karadimitrakis et al., 2016).

1. Scope and usage of the term

The expression spans at least three technically distinct usages. The first is analogue transport, where the RF waveform itself is sent over fiber, as in radio-frequency-over-fiber and analogue radio-over-fiber systems. The second is analogue processing, where functions usually assigned to ADCs, DSP, or matrix accelerators are instead executed by analog circuitry or passive fiber dynamics. The third is analogue modelling, where a fiber or a nonlinear dielectric reproduces an effective equation or metric associated with another physical problem, such as a Schwarzschild-like horizon or a mesoscopic scattering cavity (Mena et al., 2013, Li et al., 2021, Nambath et al., 2019, Moreno-Ruiz et al., 2021).

A common misconception is to treat fiber-optical analogue as synonymous with analogue gravity. The literature is broader. It includes coherent-receiver equalization implemented as a 4×4, 2-tap FIR filter in analog hardware, disordered ytterbium-doped fibers used as a fiber-optical analogue of a transverse Anderson-localizing medium, and passive standard single-mode fiber used as a random convolution engine through Rayleigh backscattering (Nambath et al., 2019, Bassett et al., 2021, Redding et al., 2023).

Another misconception is that analogue operation is simply the absence of digitization. In the cited work, analogue behavior is usually architectural: the signal chain is reorganized so that the physically evolving optical or electrical waveform performs the required function. In coherent interconnects, equalization occurs immediately after optical-to-electrical conversion and before any ADC stage. In Kerr-based analogue gravity, an intense pump pulse produces a refractive-index landscape that is interpreted as an effective spacetime for a probe. In radio astronomy RFoF, the analog RF signal is carried optically to avoid long coaxial runs and their associated loss, weight, and interference pickup (Nambath et al., 2019, Moreno-Ruiz et al., 2021, Mena et al., 2013).

2. Effective metrics and gravitational analogues in fiber

In the fibre-optical analogue of gravity, an intense pump pulse changes the refractive index through the optical Kerr effect, and that perturbation acts as a curved background for a weaker probe field. The refractive index is written as

$n(\tau)=n_0+\delta n(\tau),$

with comoving coordinates

$\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$

In this frame the effective $(1+1)$D optical metric is

$$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$

and the analogue horizon occurs when

$u\,n(\tau)=c.$

This is the point at which the probe is trapped relative to the moving pump (Moreno-Ruiz et al., 2021).

The same work relates the fiber geometry to accelerated-mirror constructions and introduces explicit Schwarzschild and Schwarzschild-Planck analogues. In the regularized case, the optical refractive-index profile is engineered so that the analogue metric reproduces a Schwarzschild-Planck line element, with the regularization length $\ell$ mapped to a material time scale $\tau_\ell$. For silica fibre, the paper states

$\tau_\ell \sim 1\ \text{ps}.$

This is significant because the regularization is not merely formal: in the fibre-optical system it is interpreted as the response time of the fibre molecules to the optical field. The regularized geometry preserves the horizon while rendering the Ricci and Kretschmann scalars finite at the horizon for $\tau_\ell>0$ (Moreno-Ruiz et al., 2021).

A broader treatment starts from Maxwell’s equations in a step-index fiber, derives guided modes, and then incorporates cubic nonlinearity to obtain a general wave-envelope propagation equation and, under further simplifying assumptions, the NLSE. In that setting, the fundamental $HE_{11}$ or $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$0 mode is well approximated by a Gaussian transverse profile, and a soliton-induced refractive-index perturbation acts as a moving effective spacetime for a probe. The horizon condition is written as

$\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$1

and the probe equation can be matched to a Klein–Gordon equation on a curved metric. The same source distinguishes Hawking-type mode conversion from resonant radiation: both arise from the same dispersive soliton background, but the former is the analogue-gravity process associated with positive- and negative-norm mode mixing, whereas the latter is a classical phase-matched emission process (Kranas et al., 17 Dec 2025).

The gravitational analogy is extended further to quasinormal modes. Under suitable approximations, perturbations of a soliton reduce to a Schrödinger-type equation with a $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$2 potential, producing a discrete complex spectrum analogous to black-hole ringdown. This suggests that fiber systems can emulate both horizon physics and barrier-induced damped oscillations within a single nonlinear-optical platform (Kranas et al., 17 Dec 2025).

3. Analogue transport over fiber

In large-array radio astronomy, a prototype 425–850 MHz RF-over-fiber link was developed for CHIME to transport the analog RF signal from the antenna/feed to the processing room over optical fiber instead of coaxial cable. The system uses intensity modulation with direct detection (IMDD), a directly modulated Fabry-Perot (FP) laser, 1310 nm operation, and single-mode fiber. Measured link parameters include Gain: 4 dB, Input compression point (CP): > 10 dBm, Output third-order intercept (OIP3): 29 dBm, Output noise floor: -143 dBm/Hz, Noise figure: 27 dB, and SFDR: 115 dB·Hz^2/3. In the CHIME two-element interferometer, the estimated system temperatures of coax and fiber versions were consistent to within about 5 K across the band, supporting the claim that the RFoF link was not the system bottleneck (Mena et al., 2013).

For cryogenic detector instrumentation, an analog optical readout system was designed to replace coaxial transmission through cryostat feedthroughs. The architecture converts detector pulses into optical signals with light intensity linearly proportional to the electrical amplitude, transmits them through single-mode optical fiber, and converts them back to electrical signals while preserving waveform shape. At -100 degree Celsius, the system achieves a -3dB bandwidth of larger than 150MHz, a dynamic range of up to 500mV, and 70mW per channel cryogenic-region power consumption. The same work investigates low-temperature coarse wavelength division multiplexing (CWDM) and demonstrates four-channel multiplexing at $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$3 through a single fiber, although insertion loss increases and the best-matched CWDM channel becomes blue-shifted by about 40–80 nm relative to room-temperature matching (Zhou et al., 11 Feb 2026).

In radio access networks, analogue radio over fiber (A-RoF) is framed as an alternative to digitised radio over fiber (D-RoF) for massive-MIMO and mmWave fronthaul. The cited analysis states explicitly that in A-RoF, “no analogue-to-digital conversion (ADC) and digital-to-analogue conversion (DAC) is required.” For a 256-QAM, 28 GHz, $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$4, $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$5 configuration, the aggregate transport bandwidth is reported as about 48 GHz for A-RoF and about 1954 GHz for D-RoF, which the paper interprets as roughly 40× less optical bandwidth for A-RoF. The same comparison, however, also reports A-RoF EVM ≈ 8% and D-RoF EVM ≈ 3.5%, with only D-RoF meeting the cited 3GPP EVM target of 3.5% in that case. This establishes a central trade-off: bandwidth and power savings do not eliminate the analogue-fidelity problem (Li et al., 2021).

A later network-scale demonstration moves analogue transport from architectural argument to production deployment. It shows coexistence of digital coherent optical (DCO) channels and analog radio-over-fiber (ARoF) channels for 60 GHz mmWave and 210 GHz sub-THz transport over a 77 km live metro network with 6 ROADMs, a 400 GHz OSaaS window from 192.8 to 193.2 THz, and a downstream 25 km PON. The experiment reports that all four DCOs were successfully received, the ARoF channels did not degrade the DCO QoT, and the mmWave and sub-THz services did not interfere with each other. Reported ARoF data rates include 2.34 Gb/s and 5.8 Gb/s for the mmWave channels, with HD-FEC limit $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$6 and SD-FEC limit $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$7 used as performance references (Dass et al., 23 May 2025).

4. All-analog processing and fiber-based computation

One of the clearest examples of fibre-optical analogue processing is the all-analog adaptive equalizer for coherent data-center interconnects. In the proposed receiver chain, fiber channel → coherent optical front-end → analog equalizer → CPRC/CDR, the output of the coherent front-end—four electrical baseband signals corresponding to the I/Q components of the X and Y polarizations—is fed directly into an analog equalizer before digitization. The prototype is a fractionally spaced two-tap continuous-time CMA equalizer, implemented as a 4×4, 2-tap FIR filter in 130 nm SiGe BiCMOS technology. Its adaptive laws are continuous-time integrals based on the constant modulus algorithm (CMA), so blind coefficient adaptation is performed entirely in analog hardware. Experimentally, the system was validated at 40 Gb/s over 10 km standard single-mode fiber channel; after the analog equalizer and behavioral CPRC, the reported EVM values were 28% back-to-back, 32% at 5 km, and 33% at 10 km, with estimated pre-FEC BERs $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$8, $\zeta=\frac{z}{u}, \qquad \tau=t-\frac{z}{u}.$9, and $(1+1)$0, respectively (Nambath et al., 2019).

A different analogue-processing paradigm uses the fiber itself as a computation engine. In fiber optic computing using distributed feedback, an input vector

$(1+1)$1

is temporally encoded as an optical pulse train and injected into standard single-mode fiber (SMF). Rayleigh backscattering from many weak scatterers acts as a distributed set of partial reflectors, producing delayed, randomly weighted copies of the input. The backscattered field is described by

$(1+1)$2

or equivalently

$(1+1)$3

where $(1+1)$4 is a Toeplitz-like transfer matrix. Photodetection introduces a nonlinearity through

$(1+1)$5

Because the fiber length can exceed the encoded pulse-train length, one launch can yield multiple distinct random convolutions in one pass, which the paper identifies as grouped convolutions (Redding et al., 2023).

The reported demonstrations position passive fiber dynamics as a front-end for machine-learning workflows rather than as a complete optical computer. On the SONAR benchmark, linear SVM achieved 75% accuracy, while the fiber-based nonlinear projection increased SVM accuracy to 90.4%. For MNIST digits and Fashion-MNIST, with training restricted to the final ridge-regression readout, the reported accuracies were 96.7% and 85.3%. The power analysis further states that at $(1+1)$6, the RBS system uses about 30× less power than a GPU, while the experimental setup used 500 m fiber, 200 MHz encoding, and about 10 µs per MVM for $(1+1)$7 (Redding et al., 2023).

These two lines of work exemplify different meanings of analogue processing. In the equalizer, analog circuits replace a high-speed ADC+DSP stage. In the distributed-feedback computer, passive fiber scattering performs the transform and a digital back-end only reads out and trains the final layer. This suggests that fibre-optical analogue processing is best understood as a redistribution of functionality across optics, analog electronics, and only selected digital stages.

5. Disorder, localization, and mesoscopic analogies

A disordered ytterbium-doped silica glass fiber has been developed as a fiber-optical analogue of a transverse Anderson-localizing medium. The structure is approximately invariant along the propagation direction but highly disordered in the transverse plane, so light is inhibited from diffusing freely across the cross-section. In the reported implementation, the final fiber has outer diameter $(1+1)$8, core diameter $(1+1)$9, and transverse index contrast $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$0 at $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$1. Localization is observed over lengths up to $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$2 using 633 nm and 532 nm visible lasers. Pump propagation at 975 nm is modeled with effective parameters adapted to the disordered geometry, yielding $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$3, $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$4, $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$5, $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$6, and $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$7 (Bassett et al., 2021).

The significance of these measurements is that the fiber is not treated as a conventional step-index guide with a single well-defined core mode. The measured effective pump radius is much larger than the nominal localized-core radius, which the paper interprets as evidence that cladding modes participate strongly. In this context, the analogue is not metaphorical but structural: the fiber reproduces the transport properties of a random transverse potential while retaining the longitudinal usability of a fiber device (Bassett et al., 2021).

A second, more abstract analogy treats a multimode or space-division-multiplexed optical fiber as the optical counterpart of a chaotic cavity from mesoscopic physics. The channel is represented by a $$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$8 scattering matrix, with negligible backscattering so that the transmission block carries the essential dynamics. Random distributed crosstalk is modeled through a random internal Hamiltonian, and mode-dependent loss (MDL) is represented by a diagonal loss matrix. The resulting effective MIMO channel obeys

$$\mathrm{d}s^2={u^2 n(\tau)^2}\,\mathrm{d}\zeta^2-c^2\left(\mathrm{d}\tau+\mathrm{d}\zeta\right)^2,$$9

where $u\,n(\tau)=c.$0 is deterministic, $u\,n(\tau)=c.$1 is a random Gaussian matrix, and $u\,n(\tau)=c.$2 encodes MDL. Using replica theory and random matrix theory, the paper concludes that in the large-$u\,n(\tau)=c.$3 limit the mutual information becomes approximately Gaussian, with mean and variance obtained from saddle-point equations (Karadimitrakis et al., 2016).

These two cases illustrate different uses of analogy in fiber science. The Anderson-localizing fiber is an experimental medium whose internal disorder directly realizes a localization problem. The chaotic-cavity model is a channel-theoretic analogy used to compute capacity statistics. Both, however, rely on the same conceptual move: fiber propagation is treated as a representative instance of a broader wave-scattering class (Bassett et al., 2021, Karadimitrakis et al., 2016).

6. Distributed nonlinear-wave analogues and self-similar propagation

In optical fiber amplifiers and fiber lasers with distributed parameters, the generalized nonlinear Schrödinger equation with varying coefficients and gain supports exact periodic and solitary self-similar waves. The governing equation is

$u\,n(\tau)=c.$4

with $u\,n(\tau)=c.$5 the distributed group-velocity dispersion coefficient, $u\,n(\tau)=c.$6 the Kerr nonlinearity coefficient, and $u\,n(\tau)=c.$7 the gain/loss coefficient. Exact self-similar propagation requires a compatibility relation linking these distributed parameters: $u\,n(\tau)=c.$8 The amplitude scaling function is written as

$u\,n(\tau)=c.$9

with

$\ell$0

and therefore

$\ell$1

This makes the existence of the self-similar waves a parameter-management condition rather than a generic feature of any inhomogeneous fiber (Kruglov et al., 2022).

The derived families include periodic cn, sn, and rational-elliptic waves, as well as bright and dark solitary waves. For example, the bright solitary solution is

$\ell$2

and the corresponding full field is

$\ell$3

Dark and kink-like solitary waves are obtained in the $\ell$4 regime (Kruglov et al., 2022).

The dynamics become especially transparent in a periodically distributed amplification system with

$\ell$5

for which

$\ell$6

The paper interprets this as periodic dispersion producing oscillatory pulse steering, described as “snakelike” motion, while constant gain amplifies the peak intensity without changing the profile. Numerical split-step Fourier simulations with

$\ell$7

show that the periodic wave, bright soliton, and dark rational-soliton remain stable under 10% white noise (Kruglov et al., 2022).

In this usage, the analogue is not directed toward gravity or mesoscopic transport, but toward robust nonlinear-wave behavior in managed media. The fiber serves as a controllable analogue platform in which distributed dispersion, nonlinearity, and gain reproduce exact self-similar structures and perturbative robustness. Taken together with analogue gravity, RF-over-fiber transport, all-analog equalization, and disorder-based platforms, this establishes fiber-optical analogue as a heterogeneous but coherent research category: the fiber is treated as an active analog medium whose intrinsic dynamics encode, transport, process, or emulate the phenomenon of interest (Kruglov et al., 2022).