The Rotor's Fingerprint

What motion tells a radar

Hearing speed in the echo's pitch

When a target moves towards a radar, the echo returns at a slightly higher frequency. When it moves away, slightly lower. It is the same effect as a passing siren changing pitch. The shift is twice the target's speed divided by the wavelength. At X-band (about 3 cm), a drone closing at 10 m/s shifts the echo by roughly 670 Hz (our arithmetic). That is tiny next to the 10 GHz carrier, but easy for a coherent radar to measure across a series of pulses.

The ground does not move, so its echo has almost no Doppler shift. Moving target indication (MTI) exploits that, filtering out unmoving returns such as hills and trees.Wikipedia Pulse-Doppler processing goes further and sorts every echo into speed bins, so a small moving target can be pulled out of clutter far stronger than it is. This is how counter-drone radars such as the Ku-band KuRFS on the U.S. Army's M-LIDS vehicle, pictured above, find drones over cluttered ground.Wikimedia Commons

The rotor's fingerprint

A drone is not one moving point. Its body moves at one speed, but its rotor blades spin far faster, with the advancing side of each blade rushing towards the radar and the retreating side moving away. Each blade spreads part of the echo across a wide band of Doppler on both sides of the body's line. That spread is the micro-Doppler signature.

Researchers at the University of St Andrews measured three DJI drones (Phantom 3, Inspire 1 and S900) and four birds of prey with 24 GHz and 94 GHz radars. Their findings set out the fingerprint clearly:Rahman & Robertson

• Drone propellers turn at "of order 100 Hz", and the signature is "the wide signal spread over both sidebands" around the body's Doppler, with blade flashes and rotor modulation lines.
• Birds show "no such wide micro-Doppler spread". Their wings modulate the echo slowly: the measured wingbeat frequencies of all four birds were 4–6 Hz.
• The drone's blade echoes were 20–40 dB below the body's, while the birds' wing modulation was only 0–10 dB below their body echo.

The authors conclude that a high-fidelity drone-detection radar "is possible in the millimetre-wave band or in K-band." The catch sits in the third finding. The feature that best separates a drone from a bird is also its faintest, 100 to 10,000 times weaker than the drone's already bird-sized body echo. A radar has to be close, sensitive and sampling fast to see it. A blade tip on a Phantom-sized drone turning 100 times a second moves at roughly 80 m/s (our arithmetic), so the radar must sample fast enough to capture Doppler shifts many kilohertz wide.

When the drone stops moving

Doppler processing works because targets move and the ground does not. A drone that hovers, as quadcopters do when observing, waiting to drop a munition or sitting in ambush, takes that advantage away. Its body echo sits at zero Doppler, alongside the ground.

Radar engineers call the problem minimum detectable velocity. Any target slower than the radar's MDV "cannot be detected" because its Doppler cannot be separated from the clutter.Wikipedia Recent research on hovering-drone detection describes conventional range-Doppler and MTI methods degrading badly in this case, because the hovering drone's return overlaps the clutter spectrum.Electronics (MDPI)

The rotors still spin, so their micro-Doppler is still there. That faint spread on either side of zero is the only evidence left. Radar makers say their systems use it to hold hovering drones. Weibel, for example, says its XENTA-C radar detects rotor micro-Doppler to separate hovering drones from ground clutter "even at zero speed".Weibel That is a vendor claim; we found no independent test results. It works only if the blade signal, 20–40 dB below an already weak body echo, rises above the noise.

Doppler finds drones because they move. A drone that waits in place leaves only the faint whisper of its blades.

Fingerprints, false friends and fixed wings

Micro-Doppler is powerful, but in the field it has limits:

• Fixed-wing drones have no rotors to fingerprint. A Shahed-type drone, or its Gerbera decoy, is driven by a single pusher propeller at the rear,Wikipedia hidden by the airframe from many angles. Its signature looks more like a small aircraft's, or at worst a bird's.
• Range eats the signal. The fourth-root rule from Part 1 applies doubly: the blade signal is far weaker than the body's, so radars can usually classify a drone only well inside the range at which they first detect it.
• Vendors differ. Radar maker Robin Radar describes drone rotor modulation at roughly 50–100 Hz against bird wingbeats at about 4–10 Hz as the basis for its classifier.Robin Radar How well this works for a given product at a given range is rarely published.
• Bird-like drones exist. Researchers working on drone-versus-bird classification note that the arrival of bird-like drones makes the problem harder still.Mississippi State University