FMCW LiDAR: Principles, Techniques & Applications

• FMCW LiDAR is a coherent optical ranging technique that uses frequency-chirped lasers and coherent detection for precise distance and velocity measurements.
• It achieves sub-centimeter to micrometer resolutions with robust signal processing, leveraging integrated photonic systems and advanced calibration techniques.
• FMCW LiDAR enables multi-dimensional imaging and long-range sensing for applications in autonomous vehicles, industrial metrology, and high-resolution 4D mapping.

Frequency-Modulated Continuous-Wave (FMCW) Lidar is a coherent optical ranging technique that achieves distance and velocity measurement by mapping time-of-flight and Doppler shift directly onto laser frequency. Unlike pulsed time-of-flight (ToF) direct-detection LiDAR, FMCW leverages a frequency-chirped, highly coherent source and coherent detection, enabling sub-centimeter to micrometer-scale range resolution, direct velocity measurement, ambient light immunity, and superior sensitivity suitable for long-range, high-precision, and multi-dimensional applications. FMCW architectures are found in state-of-the-art photonic integration, ultra-high resolution 4D imaging, and multi-channel massively parallel systems, forming a foundation for next-generation perception in autonomous vehicles, industrial metrology, and beyond (Riemensberger et al., 2019, Chen et al., 2024, Lukashchuk et al., 2023).

1. Physical Principles and Signal Model

In FMCW LiDAR, a single-frequency laser's optical frequency $\nu(t)$ is swept linearly (chirped) across a bandwidth $B$ over a period $T$: $$\nu(t) = \nu_0 + k t \qquad \text{with} \quad k = \frac{B}{T}.$$ A reference portion of the optical field acts as a local oscillator (LO), and the majority is transmitted to the target. Reflections are delayed by $\Delta t = 2d/c$, where $d$ is the round-trip distance.

By mixing the LO and return signals on a photodetector, a beat note at frequency $$\Delta f = \nu(t) - \nu(t-\Delta t) \approx k \Delta t$$ is generated. Given $\Delta t = 2d/c$, the target distance is directly mapped from

$d = \frac{c}{2k} \Delta f.$

For radially moving targets, the Doppler effect introduces a shift $f_D = 2 v \nu_0 / c$, enabling velocity extraction via two sweeps of opposite slope: $$f_u = \Delta f + f_D;\qquad f_d = |-\Delta f + f_D|;\qquad v = \frac{\lambda}{2} f_D.$$ Fundamental range resolution is limited by the total chirp bandwidth, $\Delta d = c/2B$; increasing $B$ improves resolution at the cost of greater requirements on sweep linearity, bandwidth, and calibration (Riemensberger et al., 2019, Mateo et al., 2015, Okano et al., 2020, Cai et al., 2024).

2. Source Engineering and Chirp Linearity

FMCW LiDAR performance depends critically on the coherence, linewidth, and sweep linearity of the optical source. Photonic integration advances have produced RSOA-based external-cavity lasers and Vernier-filter schemes with monolithically integrated piezoelectric micro-actuators, achieving mode-hop-free, high-speed tuning (>1.5 GHz bandwidth, >100 kHz sweep rates) and sub-kHz intrinsic linewidth (Lihachev et al., 2023, Lukashchuk et al., 2023). Advanced predistortion and feedback corrections (including on-the-fly deconvolution and phase-locked loops) drive RMS chirp nonlinearity into <0.1% for high-fidelity ranging (Chanelière, 23 Dec 2025, Lihachev et al., 2023).

Soliton microcombs and electro-optic combs enable massive parallelization: frequency-chirped pump lasers transfer coherent sweeps to entire comb spectra, producing tens to hundreds of mutually coherent, wavelength-multiplexed channels. Pump laser frequency chirps within the soliton existence range ($\sim$[1.2,2.9] GHz detuning) can be tracked with $\eta_\mu \approx 1$ chirp transfer efficiency across the C-band, enabling true parallel FMCW operation (Riemensberger et al., 2019, Cai et al., 2024, Chen et al., 2024).

3. Parallelization, Scalability, and Multi-Dimensional Imaging

Microcomb-based and electro-optic-comb-based architectures route each comb tooth through wavelength-dispersive optics (gratings, OPAs), forming massively parallel, mutually coherent FMCW channels. Demonstrated implementations achieve:

• 30 simultaneously operating channels (amplifier-limited) with >90 possible (comb-limited), each independent, enabling true parallel 4D ranging (distance, velocity, $\theta_{x,y}$).
• Frame/pixel rates up to 3 Mpx/s (30 channels × 100 kHz scan), with projections to >100 Mpx/s for full-comb (179 lines, 1 MHz scan) operation (Riemensberger et al., 2019, Chen et al., 2024).
• Adaptive, reconfigurable channel spacing (20–44 GHz) and zoom-in imaging (ROI densification) for dynamic spatial resolution—ROI resolutions of $0.012^\circ$ (up to $15\times$ conventional) (Chen et al., 2024).
• Dense 2D/3D mapping via hardware-multiplexed (fiber bundle, photonic lantern) arrays or angularly dispersed beams; solid-state beam-steering using OPAs or coupling of spectral channels to nanophotonic grating arrays (Riemensberger et al., 2019, Okano et al., 2020).

Multi-dimensional sensing is directly achieved: simultaneous extraction of $x, y, z$ (from comb index and grating angle) and $v$ (from Doppler), and, when fused with auxiliary sensors (e.g., color cameras), enables color-enhanced depth imaging (Chen et al., 2024).

4. Signal Processing, Compressive Sensing, and Precision

Coherent detection in FMCW LiDAR employs heterodyne gain, yielding shot-noise-limited sensitivity (even under ultralow return power) and direct background suppression. Balanced detection, background subtraction, and single-pixel compressive sensing architectures enhance SNR further and enable compressive depth-mapping with drastically reduced channel counts; solving only two linear equations suffices to recover full-depth scenes (Lum et al., 2018).

Depth and velocity are recovered per channel with fast FFT-based demodulation; localization precision down to sub-millimeter (axial $\sim$40–60 µm, lateral $\sim$200 µm) is demonstrated with robust centroid localization and time-frequency analysis (Qian et al., 2020). Compressive architectures reduce the number of independent spatial measurements for a $n$-pixel map to $O(m) \ll n$, retain high SNR, and handle multi-return (multi-depth) targets via sparsity-promoting optimization (Lum et al., 2018).

Absolute accuracy in FMCW LiDAR is fundamentally tied to the traceable calibration of the chirp rate $k$. Molecular spectral references (HCN, CO) yield chirp calibration uncertainties at the 7–21 ppm level, supporting sub-millimeter absolute error at >100 m ranges—sufficient for industrial metrology applications (Mateo et al., 2015, Mateo et al., 2015).

5. Advanced Functionalities: Doppler, Odometry, and Quantum Limit Extensions

Doppler extraction is inherent: simultaneous measurement of up- and down-chirp beat frequencies separates range and velocity. Each channel provides independent radial velocity, facilitating per-return 6-DOF vehicle odometry when combined with IMUs; continuous-time Kalman filtering on direct Doppler measurements offers real-time pose estimation with sub-percent error at <10 ms/frame computation (Yoon et al., 2023).

Quantum extensions propose exploiting entangled, frequency-chirped photon states (NOON states) in Mach–Zehnder configurations, enabling sub-shot-noise scaling. For $n$-photon entangled FMCW, the range accuracy scales as $1/\sqrt{n}$ and velocity as $1/n$—outperforming classical coherent detection at equal photon number, and transducing all measurements to electronic (microwave) frequencies, eliminating the need for ultrafast photon timing or optical Doppler measurement (Huang et al., 2023). This approach circumvents inefficiencies plaguing pulsed quantum LiDAR, since only the generation of entangled photons (via SPDC) incurs nonlinear optical loss.

6. Photonic-Electronic Integration and System-Level Considerations

Wafer-scale co-integration of high-coherence tunable lasers (Vernier-filter, RSOA, PZT-driven), high-voltage waveform drivers (e.g., SiGe BiCMOS ASICs), erbium-doped waveguide amplifiers (EDWA), and on-chip coherent receivers, enables truly "plug-and-play" FMCW LiDAR engines, manufacturable at scale and compatible with both focal-plane array and OPA front-ends (Lukashchuk et al., 2023, Lihachev et al., 2023, Chen et al., 2024). Precise chirp linearity ($<$0.1% nonlinearity), mode-hop-free GHz sweeps, and sub-kHz linewidths are achieved in fully integrated PICs consuming tens of mW, supporting >10 cm range resolution at modulation rates exceeding 100 kHz.

Integrated systems leveraging photonic combs, dynamic ELF frequency control, and on-the-fly predistortion algorithms (Chanelière, 23 Dec 2025), together with calibration against spectral references, maintain accuracy, robustness, and flexibility—enabling solid-state, mass-producible systems for automotive, robotics, and emerging 4D and color-augmented machine vision (Riemensberger et al., 2019, Chen et al., 2024, Cai et al., 2024, Lukashchuk et al., 2023).

7. Limitations, Trade-offs, and Outlook

FMCW LiDAR’s performance is limited by practical factors: achieving wide, high-linear chirps challenges modulator electronics and tuning mechanisms; power division in multi-channel and photonic-lantern arrays sets SNR-per-channel; data acquisition bandwidth dictates maximum range and voxel rate; and signal processing must address speckle, multi-path, and calibration drift (Riemensberger et al., 2019, Qian et al., 2020, Mateo et al., 2015).

The trade-off between scan rate, range/resolution, and angular FOV is fundamental—wider bandwidth and more pixels require greater optical power, faster detectors, and more sophisticated photonic integration. Upcoming advances include fully monolithic PIC-based LiDAR combining microcombs, OPA beam-steering, and integrated coherent receivers, pushing towards 100 Mpx/s 4D imaging, integration with on-chip photonic quantum sources, and large-scale automotive deployment (Chen et al., 2024, Lukashchuk et al., 2023, Riemensberger et al., 2019, Huang et al., 2023).