Optical Vector Analysers (OVA): Fundamentals
- Optical Vector Analysers (OVA) are coherent optical metrology instruments that recover both amplitude and phase to characterize optical devices.
- They employ modulation-based and swept-interferometric techniques to provide high-resolution measurements of loss, dispersion, and coupling parameters.
- OVAs are crucial in integrated photonics, fiber optics, and resonator metrology, offering actionable insights for device calibration and performance optimization.
Optical Vector Analysers (OVA), often also described as Optical Vector Network Analyzers (OVNA) or, in some implementations, Vector Spectrum Analyzers (VSA), are coherent optical metrology instruments that recover the complex response of an optical device or system rather than only its scalar transmission. In the passive-device setting, the measured quantity is the complex transfer function
from which one obtains amplitude or loss, phase, and derived quantities such as linewidth, coupling regime, group index, group delay, and dispersion. In space-division-multiplexing metrology, the same concept is extended to the full linear complex transfer-function matrix of a device or fiber, enabling extraction of insertion loss (IL), cross-talk (XT), and mode-dependent loss (MDL) (Feng et al., 2024, Luo et al., 2023, Kalla et al., 2024, Shi et al., 2024).
1. Concept, nomenclature, and scope
The contemporary OVA literature is not limited to a single instrument topology. Some systems are explicitly called OVAs, some OVNAs, and some VSAs, but in passive-device mode they share the same essential objective: phase-sensitive recovery of optical response versus optical frequency. The fiber-based dual-mode VSA reported for integrated photonics is described as a generalized OVA in passive mode because it measures loss, phase response, and dispersion over 1260 to 1640 nm with 471 kHz resolution and 56 dB dynamic range (Luo et al., 2023). The visible-light VSA for integrated photonics plays the same role in the 766 to 795 nm band, measuring transmission, phase, resonance frequencies, linewidths, and dispersion with 415 kHz frequency resolution and 8.1 MHz absolute frequency accuracy (Shi et al., 2024). In thin-film lithium niobate, an in-situ integrated OVA uses on-chip single-sideband modulation to probe on-chip devices directly, and in long-fiber metrology an optimized OVNA measures the full linear complex transfer function matrix of SDM systems in a single wavelength sweep (Feng et al., 2024, Kalla et al., 2024).
The term is not interchangeable with every optical instrument carrying the words “vector” or “network analyzer.” The “nano-optical vector network analyzer” for infrared optical antennas is a spatially scanning, monochromatic, interferometric near-field instrument that reconstructs local , , and ; it is conceptually adjacent but not a conventional swept-frequency OVA for broadband device transfer functions (Olmon et al., 2010). Likewise, the photonic vector network analyzer based on wideband direct photonic digitizing measures RF and microwave -parameters through photonic undersampling; it is a photonic-assisted RF VNA rather than an optical-domain OVA for optical DUTs (Jin et al., 2020).
2. Measurement principles and signal models
The defining OVA operation is coherent retrieval of both amplitude and phase. In an integrated single-sideband OVA, a continuous-wave optical carrier at is phase-modulated by a swept RF tone at , filtered to retain the carrier and one sideband, and then sent through the DUT. After photodetection, the RF current at is proportional to the product of the carrier-frequency and sideband-frequency optical responses. The paper gives
and, after calibration,
This is the core optical-to-RF vector mapping used in that in-situ architecture (Feng et al., 2024).
A second major class is swept-interferometric OVA. In the fiber-based VSA/OVA, a chirped laser is split into a measurement arm and a delayed reference arm. The delay converts optical chirp into a low-RF beat,
0
with 1 the chirp rate and 2 the arm delay difference. The beat envelope yields amplitude, while the phase is extracted numerically using a Hilbert transform and then mapped from time to optical frequency using a calibrated frequency ruler (Luo et al., 2023). The visible-band analyzer retains the same swept-frequency logic but adds a calibrated fiber cavity for relative frequency reconstruction and saturated-absorption references in rubidium and potassium vapor for absolute frequency anchoring (Shi et al., 2024).
Balanced heterodyne OVA provides the same complex quantity in explicit quadratures. In the free-running Mach–Zehnder implementation, the signal arm is shifted by 40 MHz, recombined with a strong local oscillator, and digitally demodulated to obtain
3
Its sensitivity analysis is expressed through
4
which identifies the coherent-state SQL condition for vector measurement: unit SNR corresponds to approximately one detected photon in the integration bin, up to efficiency (Dasigi et al., 30 Sep 2025).
In matrix OVNA for SDM fibers, the same coherent principle is generalized from a scalar 5 to a full transfer matrix. The explicit demonstration focuses on IL, defined conceptually as the average of the squared singular values of the complex transfer matrix. For kilometer-scale interferometers, the paper identifies a specific polarization-dynamics limit through the polarization rotation rate
6
where 7 is the differential group delay of the reference fiber (Kalla et al., 2024).
3. Principal instrument architectures
A recurring architectural divide is between modulation-based OVA and swept-interferometric OVA. The modulation-based extreme is the asymmetric-signal-generator/asymmetric-signal-receiver architecture. It uses an AOM-shifted carrier plus carrier-suppressed optical double sidebands and recovers the DUT response from beat notes deliberately separated from nonlinear modulation products. In experiment, the system used a 41-tone optical frequency comb with 25 GHz spacing, an 80 MHz AOM shift, and demonstrated 334 Hz resolution, >90 dB dynamic range, and 1.025 THz measurement range (Qing et al., 2019).
Swept-laser interferometric systems emphasize wide optical span with relatively low RF bandwidth. The fiber-based VSA/OVA uses one or more mode-hop-free chirped external-cavity diode lasers, a phase-stable fiber cavity as a frequency ruler with FSR near 55.58 MHz, and low-RF interferometric detection, avoiding high-speed modulators, high-speed photodetectors, and active feedback control (Luo et al., 2023). The visible-light analyzer follows the same principle but obtains the swept visible probe by frequency doubling a telecom-band chirped source in a chirped periodically poled lithium niobate waveguide, then references the result to alkali hyperfine lines (Shi et al., 2024).
Integrated OVA miniaturizes the vector-analysis core onto the photonic chip. The lithium-niobate in-situ system combines a broadband electro-optic phase modulator with a tunable flat-top RAMZI filter to generate a high-fidelity single-sideband probe. A swept RF tone from 10 to 50 GHz drives the modulator, and thermal tuning shifts adjacent channels for stitched wideband measurement (Feng et al., 2024).
Sensitivity-oriented OVA forms a distinct architectural line. The quantum-limited implementation uses a free-running, balanced heterodyne Mach–Zehnder interferometer with a strongly asymmetric 99:1 power split between local oscillator and signal arms, 80 MHz anti-alias filtering, 200 MS/s digitization, and post-processed piecewise phase detrending to suppress residual drift without active phase locking (Dasigi et al., 30 Sep 2025).
Long-fiber OVNA introduces yet another hardware requirement: active polarization management in the reference arm. In the SDM-fiber system, the reference arm contains a long optical delay approximately matched to the DUT path length, but the decisive modification is an automatic polarization controller using optical feedback and a LiNbO8-based polarization transformer to stabilize the reference SOP across the sweep (Kalla et al., 2024).
4. Reported performance dimensions
The literature does not report a single dominant figure of merit; instead, OVA systems optimize different ceilings: spectral span, frequency resolution, dynamic range, sensitivity, absolute frequency accuracy, or robustness to long-delay polarization effects.
| System | Architecture/domain | Reported figures |
|---|---|---|
| (Qing et al., 2019) | Asymmetric comb-assisted OVA | 334 Hz resolution, >90 dB dynamic range, 1.025 THz range |
| (Luo et al., 2023) | Fiber-based swept-laser VSA/OVA | 55.1 THz (1260 to 1640 nm), 471 kHz, 56 dB |
| (Shi et al., 2024) | Visible-light VSA/OVA | 766 to 795 nm, 14.3 THz, 415 kHz, 8.1 MHz accuracy |
| (Feng et al., 2024) | In-situ LN OVA | 50 kHz resolution, 16.2 THz demonstrated bandwidth |
| (Dasigi et al., 30 Sep 2025) | Quantum-limited balanced-heterodyne OVA | 20 THz, about 1 fW at 10 kHz IBW, unit SNR at roughly 0.8 photons |
| (Kalla et al., 2024) | Long-fiber OVNA for SDM MCF | 1530–1570 nm, 100 nm/s; IL deviation up to 4 dB without APC, below 0.5 dB with APC |
Resolution ranges from 334 Hz in the asymmetric comb-assisted system, through 50 kHz in the integrated LN OVA, to sub-megahertz values in wideband swept-source analyzers. These numbers are not interchangeable: the 334 Hz result is tied to a 300 Hz linewidth laser and comb-assisted modulation architecture, whereas the 415 kHz and 471 kHz figures are set by the dynamic linewidth of chirped swept sources (Qing et al., 2019, Shi et al., 2024, Luo et al., 2023).
Bandwidth also depends strongly on architecture. The broadest span reported here is 55.1 THz over 1260 to 1640 nm in a fiber-based swept-laser VSA/OVA, while the visible-band system reaches 14.3 THz over 766 to 795 nm, the integrated LN system demonstrates 16.2 THz, and the quantum-limited heterodyne OVA covers 20 THz from 1480 nm to 1640 nm (Luo et al., 2023, Shi et al., 2024, Feng et al., 2024, Dasigi et al., 30 Sep 2025).
Dynamic-range and sensitivity figures reflect different operating goals. The asymmetric OVA reports >90 dB dynamic range and analyzes a receiver-limited envelope approaching 118 dB in the best channels; the fiber-based VSA/OVA reports 56 dB dynamic range; the quantum-limited OVA instead emphasizes fW-level sensitivity, 9, and operation close to the SQL (Qing et al., 2019, Luo et al., 2023, Dasigi et al., 30 Sep 2025).
5. Derived observables and application domains
A central OVA use case is resonator metrology. In the visible-band analyzer, a Si0N1 microresonator was characterized over 588 resonances across 766–795 nm with 24.25 GHz FSR; the intrinsic linewidth histogram peaked at 90 MHz, corresponding to 2 and 14 dB/m linear loss. The same system measured an example under-coupled resonance with 3 and 4, and fitted integrated dispersion through
5
obtaining 6, 7, and 8 (Shi et al., 2024).
The in-situ LN OVA extends this resonator emphasis to same-chip metrology. It resolves single-ring, double-ring, and triple-ring responses, including phase jumps that identify coupling state, and stitches broadband measurements over 1500 to 1630 nm for a total of 16.2 THz and 84 resonance notches. The same instrument is applied to a dynamically modulated microring resonator realizing a synthetic frequency crystal with FSR 24.8 GHz, allowing direct measurement of collective phase dynamics and density of states of Bloch modes rather than only scalar transmission (Feng et al., 2024).
Broadband waveguide characterization is another mature OVA application. The fiber-based VSA/OVA measured a 1.6394-meter-long Si9N0 spiral waveguide, extracting optical length 3.4214 m, 1 at 192.681 THz, average loss 2, and higher-order dispersion coefficients 3, 4, 5, and 6. The same platform also measured Si7N8 microresonator integrated dispersion up to 9, coherently mapped a 100-GHz-rate soliton microcomb, and was repurposed for FMCW LiDAR (Luo et al., 2023).
OVA capability is not restricted to passive, linear, or high-power regimes. The quantum-limited heterodyne system characterized thin-film lithium niobate microring resonators and extracted an internal quality factor above 5 million, while maintaining SNR > 1 with much less than 1 circulating photon and on-chip probe powers low enough that photorefractive and thermo-optic shifts could safely be ignored (Dasigi et al., 30 Sep 2025).
Matrix OVNA addresses a different application class: long SDM fibers. The optimized system characterized a 10 km 7-core uncoupled multi-core fiber, measuring the full linear complex transfer-function matrix in one sweep and deriving IL, XT, and MDL. The experimental emphasis was IL: without automatic polarization control, wavelength-dependent fading produced distortions up to 4 dB; with control, IL variation was reduced below 0.5 dB, and the standard deviation improved from 1.49 dB to 0.35 dB (Kalla et al., 2024).
6. Technical limitations, failure modes, and recurring misconceptions
The major limitations depend on architecture. In modulation-based OVA, residual sidebands, high-order modulation products, spectral aliasing, and small-signal constraints historically forced tradeoffs among resolution, dynamic range, and bandwidth. The asymmetric ASG/ASR approach was proposed specifically to separate useful beat notes from nonlinear distortion products and thereby remove the dominant modulation-nonlinearity error mechanism of conventional modulation-based OVA (Qing et al., 2019). In integrated photonics, a different problem dominates: ex-situ measurement can obscure intrinsic on-chip response through substantial fiber-chip coupling loss, chip-facet reflections, and Fabry–Perot interference. The in-situ LN architecture addresses that problem by placing the OVA beside the DUT on the same chip (Feng et al., 2024).
Frequency calibration is another persistent constraint. The fiber-based VSA/OVA uses a phase-stable fiber cavity as a relative frequency ruler and a built-in wavelength meter with 200 MHz absolute accuracy; its paper notes that even a 1 K temperature change would cause about 0 MHz cumulative error over the full 55.1 THz if uncorrected (Luo et al., 2023). The visible-band analyzer tightens this ceiling by adding atomic references. It reports an initial mean deviation of 28.0 MHz against K D1 literature values, applies a correction based on the Rb anchor near 384.229 THz, and then obtains 8.1 MHz mean frequency deviation using K D2 lines as an evaluation set (Shi et al., 2024).
High-sensitivity heterodyne OVA is limited less by optical linewidth than by phase drift and integration-bandwidth choices. In the quantum-limited implementation, the interferometer is deliberately free-running rather than actively phase-locked, so phase-sensitive analysis over the full 20 THz range depends on passive arm-length matching, a strong local oscillator, and piecewise phase detrending. The paper states that phase noise is negligible for sweep durations shorter than about 10 ms, corresponding at 200 nm/s to more than 250 GHz phase-stable span (Dasigi et al., 30 Sep 2025).
Long-delay OVNA introduces a failure mode that is specific to interferometric fiber systems: polarization-induced fading. In the 10 km MCF experiment, the long reference-arm delay undergoes wavelength-dependent birefringence, so the reference SOP rotates across the sweep and the polarization overlap at the receiver collapses. The paper states that prior digital polarization equalization becomes impossible in deep-fading scenarios; its remedy is optical-domain reference-arm stabilization with an APC. The tracking-speed requirement is set by
3
and for the demonstrated 1530 nm to 1570 nm sweep at 100 nm/s the required tracking speed is approximately 50 rad/s (Kalla et al., 2024).
A recurring misconception is that scalar transmission alone is sufficient for photonic-device diagnosis. The cited literature repeatedly uses phase to distinguish under-coupled, critically coupled, and over-coupled resonances; to fit internal and external loss rates separately; to recover integrated dispersion; and to observe collective phase dynamics in synthetic frequency lattices. This suggests that, for many integrated and fiber-optic devices, scalar spectral power is not a complete descriptor of device behavior.