Signal Processing in FMCW Radars

FMCW radars are widely used in short range applications such as autonomous vehicles, heartbeat monitoring, vibration measurement in residential and industrial products, drones ranging and intrusion detection. Building the main concepts from the ground up, I will describe in detail FMCW waveform design, FFT processing and the utility of radar date cube that covers all three dimensions, namely range, Doppler and angle. The talk will conclude with practical tips for designing an FMCW radar system.

What this presentation is about and why it matters

This talk introduces signal processing for frequency‑modulated continuous‑wave (FMCW) radars from first principles. Instead of measuring time‑of‑flight directly, FMCW converts range and motion problems into frequency and phase estimation problems that are easier and more precise to solve with DSP. The presentation covers waveform design (chirps), the mixer/output beat tone, FFT‑based processing to separate multiple targets, and how to build the 3D radar "data cube" (range, Doppler, angle). Practical design tips are included.

Why this matters: FMCW radars are widely used in short‑range applications (automotive/ADAS, drones, proximity sensing, medical vitals, vibration measurement). For DSP engineers working with or transitioning into radar, understanding the FMCW signal chain and the FFT‑centric processing pipeline is essential to make correct choices of bandwidth, sweep time, sampling rate and antenna layout — all of which directly affect resolution, range ambiguity, and target separation.

Who will benefit the most from this presentation

• DSP engineers who understand Fourier analysis and discrete processing but are new to radar concepts.
• RF/embedded engineers implementing FMCW front ends who need to translate system specs into DSP blocks.
• Graduate students or practitioners learning how range, Doppler and angle are estimated using chirp waveforms and FFTs.
• Anyone designing short‑range sensing systems who needs intuition about trade‑offs (bandwidth vs. resolution, sweep time vs. Doppler ambiguity).

What you need to know

To get the most from this talk, be comfortable with:

• Complex baseband / IQ signals. The received RF is typically downconverted to an I (real) and Q (imag) pair so that sinusoids become complex exponentials $e^{j\phi(t)}$ and phase is explicit.
• Chirp waveform (linear FMCW). The basic transmitted chirp can be written as $x(t)=A\,e^{j2\pi\big(ft+\tfrac{\mu}{2}t^2\big)}$, where $\mu$ is the chirp rate (Hz/s). An up‑chirp sweeps frequency up linearly; a down‑chirp sweeps it down.
• Mixing and beat frequency. Multiplying the received delayed echo by the conjugate of the transmit yields a low‑frequency sinusoid (beat note) at $f_b=\mu\tau$ (for ideal linear chirps), where $\tau$ is round‑trip delay. From the beat frequency you can compute range: $D=\dfrac{c\,T_c\,f_b}{2B}$ when using sweep bandwidth $B$ and sweep time $T_c$.
• FFT as a multi‑sinusoid resolver. Multiple targets create a sum of sinusoids at different beat frequencies. The range FFT (fast time) separates targets by beat frequency; peaks map to ranges via the formula above.
• Separating range and Doppler. Motion adds Doppler to the beat. A single chirp conflates range and Doppler; using up/down or multiple chirps gives independent equations so you can solve for both. Across chirps the phase progression is a slow‑time sinusoid whose frequency is proportional to velocity (Doppler). A Doppler FFT (over chirps) separates moving targets.
• Angle estimation with arrays. Multiple receive antennas produce spatial phase differences $\Delta\phi=2\pi d\sin\theta/\lambda$. An FFT (or beamformer) across antennas yields angle‑of‑arrival information; combine this with range and Doppler to populate the radar data cube.
• Practical limits. Be aware of phase wrapping, ambiguity ranges (related to chirp parameters), finite sample length effects (sinc mainlobes), and the need for windowing or zero padding for better peak localization.

Glossary

• FMCW — Frequency‑Modulated Continuous‑Wave: radar that sweeps carrier frequency over time (chirp) and measures beat frequency to estimate range.
• Chirp — A signal whose instantaneous frequency varies linearly with time; key parameters are bandwidth $B$, sweep time $T_c$, and chirp rate $\mu=B/T_c$.
• Beat frequency ($f_b$) — Low‑frequency tone after mixing TX with delayed RX; proportional to target delay and therefore range.
• IQ (Complex baseband) — Representation of a real RF tone as I (in‑phase) and Q (quadrature) components; convenient for DSP and phase measurements.
• Matched filter / processing gain — Correlating received signal with known transmit shape concentrates energy into peaks, improving SNR and detection.
• FFT — Discrete Fourier transform used to separate sinusoids: fast‑time FFT -> range, slow‑time FFT -> Doppler, spatial FFT -> angle.
• Doppler frequency — Frequency shift caused by relative motion; for CW $v=f_D\,\lambda/2$, for FMCW it appears as phase progression across chirps.
• Radar data cube — 3‑D array indexed by range, Doppler (velocity), and angle (direction) containing target energy.
• Range resolution — Minimum separable range ≈ $c/(2B)$; improving with larger sweep bandwidth $B$.

Parting note

This presentation delivers a compact, intuitive bridge from DSP fundamentals to practical FMCW radar processing. It emphasizes the elegant engineering trick of converting time‑delay problems into frequency/phase estimation tasks — a theme that will resonate with anyone experienced in Fourier methods and estimation. Expect clear equations, helpful numerical examples, and a practical perspective on when terms can be ignored and why. If you want a DSP‑oriented, actionable introduction to short‑range radar processing and the radar data cube, this talk is well worth watching.

This overview is AI-generated from the session transcript. Spot an issue? Let us know.