RADAR

PHASE 1: Shout and Listen

Stand in a canyon. Shout. Wait. The echo comes back — and from the delay, you know how far the wall is. Sound travels at 343 m/s. Light travels at 300,000,000 m/s. Same principle, 876,000 times faster. Radar is an echo machine. Transmit a pulse of radio frequency energy. Wait. Listen for the return. The time delay tells you the range. The Range Equation — From Time to Distance The pulse travels to the target and back — a round trip. So the range is half the total travel distance: R = c × t / 2 Where: ├── R = range to target (meters) ├── c = speed of light = 299,792,458 m/s ≈ 3 × 10⁸ m/s └── t = round-trip time (seconds) For a target at 300 km: t = 2R/c = 2 × 300,000 / 3 × 10⁸ t = 600,000 / 300,000,000 t = 0.002 s = 2 milliseconds Shout. Wait 2 milliseconds. Hear the echo. The target is 300 km away.

The Pulse — What You're Actually Shouting You don't transmit continuously — you'd drown out your own echo. Instead, send a short burst of radio energy, then go silent and listen.

Range Resolution — Telling Two Targets Apart Two aircraft flying 100 m apart in range. Can you see them as two separate targets? Only if their echoes don't overlap in time. The minimum separation: ΔR = c × τ / 2 For a 1 μs pulse: ΔR = 3 × 10⁸ × 1 × 10⁻⁶ / 2 ΔR = 150 meters Anything closer than 150 m merges into one blob. Want better resolution? Use a shorter pulse. A 0.1 μs pulse gives 15 m resolution — but puts less energy on target, reducing range. This is the fundamental tradeoff: range vs resolution. Long pulses see far but blur nearby. Short pulses see sharply but go blind at distance. The solution — pulse compression — comes later.

DESIGN SPEC UPDATED: ├── Range equation: R = ct/2 — 300 km target returns echo in 2 ms ├── Pulse width: 1 μs → 150 m range resolution ├── PRF: 500 Hz (limited by maximum unambiguous range) ├── Duty cycle: 0.05% — transmitting only 1 μs out of every 2 ms └── Fundamental tradeoff: long pulse = more energy = more range, but worse resolution

PHASE 2: The Radar Equation

You transmitted a megawatt pulse. It spread out in all directions, hit a fighter jet 300 km away, scattered off its metal skin, and some fraction headed back toward you. By the time it arrives, the signal is one ten-trillionth of a trillionth of what you sent. Can you still detect it? The radar equation answers the most important question in radar: given your transmitter power, antenna, and target — what signal strength comes back? Building the Equation Step by Step Start with what the transmitter sends. A pulse of power P_t radiates outward. An isotropic antenna spreads power equally in all directions across a sphere: Power density at range R = P_t / (4πR²) But we don't use an isotropic antenna. We focus the beam with a dish, concentrating energy by a factor G (antenna gain): Power density at target = P_t × G / (4πR²) The target intercepts some of this energy. Its radar cross section σ (in m²) determines how much: Power intercepted = P_t × G × σ / (4πR²) This intercepted power re-radiates back toward the radar, spreading over another sphere of radius R: Power density back at radar = P_t × G × σ / [(4π)² × R⁴] The radar's antenna captures some of this with its effective area A_e: P_r = P_t × G² × λ² × σ / [(4π)³ × R⁴] Where A_e = G × λ² / (4π) for a matched antenna.

A Real Example — Detecting a Fighter at 300 km Plug in numbers for a military surveillance radar: P_t = 1 MW (10⁶ W) peak power G = 10,000 (40 dB — a large dish) λ = 0.1 m (3 GHz, S-band) σ = 10 m² (typical fighter jet, broadside) R = 300,000 m P_r = (10⁶ × 10⁸ × 0.01 × 10) / [(4π)³ × (3 × 10⁵)⁴] P_r = 10¹³ / [1,984 × 8.1 × 10²¹] P_r = 10¹³ / [1.607 × 10²⁵] P_r = 6.2 × 10⁻¹³ W That's 0.62 picowatts. You transmitted a megawatt and got back less than a trillionth of a watt. The ratio: 1.6 × 10¹⁸ — nearly 10 quintillion to one. Can you detect 0.62 picowatts? Yes — if your receiver noise floor is below it. A good radar receiver has a noise figure of ~3 dB and bandwidth ~1 MHz, giving a noise floor around -111 dBm ≈ 8 × 10⁻¹⁵ W. Your signal is ~80× above the noise. You see it.

DESIGN SPEC UPDATED: ├── Radar equation: P_r = P_t G² λ² σ / (4π)³ R⁴ ├── R⁴ problem: double range → 16× less signal ├── 1 MW transmitted, 0.62 pW returned from 300 km (10 m² target) ├── Dynamic range: 10¹⁸ (182 dB) between transmit and receive └── Detection requires signal above receiver noise floor (~8 × 10⁻¹⁵ W)

PHASE 3: Build the Antenna

Your radio pulse blasts out in all directions. Most of that energy is wasted — illuminating clouds, trees, and empty sky. You need to focus it into a tight beam, like a flashlight instead of a bare bulb. The antenna does this. And the physics that determines how tight your beam can be is the same physics that determines how small you can focus light through a lens. The antenna has two jobs: ├── Transmit: focus energy into a narrow beam toward the target └── Receive: collect the faint echo from that same direction Beamwidth — The Angular Precision A dish antenna's beamwidth is set by the ratio of wavelength to dish diameter: θ ≈ 70 × λ / D (degrees, for a parabolic dish) Where: ├── θ = half-power beamwidth (degrees) ├── λ = wavelength (meters) └── D = dish diameter (meters)

Antenna Gain — Focusing the Energy Gain measures how much the antenna concentrates power relative to an isotropic radiator: G = η × (π × D / λ)² Where η ≈ 0.55-0.65 (aperture efficiency, accounting for imperfect illumination). For our 3 m dish at 3 GHz (λ = 0.1 m): G = 0.6 × (π × 3 / 0.1)² G = 0.6 × (94.2)² G = 0.6 × 8,884 G = 5,330 ≈ 37.3 dB The antenna multiplies your transmitted power by 5,330 in the beam direction. And on receive, it multiplies the collected signal by the same factor. Total round-trip gain: G² = 28.4 million. This is why the antenna is the most critical component in the radar.

DESIGN SPEC UPDATED: ├── Beamwidth: θ = 70λ/D → 3 m dish at 3 GHz gives 2.3° beam ├── Gain: G = η(πD/λ)² → 5,330 (37.3 dB) for 3 m dish at 3 GHz ├── Round-trip gain: G² = 28.4 million — the antenna matters most ├── Bearing accuracy: ~θ/10 with monopulse → ±0.23° at 300 km ≈ ±1.2 km └── Tradeoff: higher frequency = narrower beam, but more rain attenuation

PHASE 4: Find the Speed

You know WHERE the target is. But is it coming toward you or flying away? At 300 km, you might have 20 minutes before it's overhead — or it might already be receding. An ambulance siren rises as it approaches, drops as it passes. The same physics works at the speed of light. The Doppler effect: when a source moves toward you, the waves compress — frequency rises. Moving away, waves stretch — frequency drops. The Doppler Shift in Radar For a target moving with radial velocity v_r (toward or away from the radar): Δf = 2 × v_r × f₀ / c Where: ├── Δf = frequency shift (Hz) ├── v_r = radial velocity (m/s) — component toward/away from radar ├── f₀ = transmit frequency (Hz) └── c = speed of light The factor of 2: the wave is Doppler-shifted once on the way to the target, and again on the way back. For a fighter jet at 300 m/s (Mach 0.88) head-on, at 10 GHz (X-band): Δf = 2 × 300 × 10 × 10⁹ / (3 × 10⁸) Δf = 6 × 10¹² / 3 × 10⁸ Δf = 20,000 Hz = 20 kHz A 20 kHz shift on a 10 GHz carrier. That's a fractional change of 2 parts per million. But 20 kHz is easy to measure — it's in the audio range. Early radar operators literally listened to the Doppler tone through headphones.

MTI — Moving Target Indication The ground doesn't move. Trees barely move. But your jet target is screaming toward you at 250 m/s. The Doppler shift separates them. Pulse-Doppler MTI: transmit multiple pulses, compare echoes from the same range cell across pulses. Stationary objects produce identical echoes — subtract them and they vanish. Moving targets produce phase-shifted echoes — they survive the subtraction.

DESIGN SPEC UPDATED: ├── Doppler: Δf = 2v_r f₀/c → 300 m/s at 10 GHz = 20 kHz shift ├── Velocity resolution: Δv = λ/(2×N×PRI) — more pulses = finer resolution ├── MTI cancels ground clutter: 30-50 dB suppression of stationary returns ├── Pulse-Doppler separates targets by both range AND velocity simultaneously └── Ambiguity: high PRF = unambiguous velocity but ambiguous range (and vice versa)

PHASE 5: See Through Rain

It's a monsoon. Rain at 50 mm/hour — sheets of water filling the sky between you and the target. Every raindrop is a tiny reflector, scattering your signal. And the droplets absorb energy too. At some frequencies, the rain becomes a wall. At others, it's almost transparent. Radio waves interact with rain in two ways: ├── Scattering: rain reflects energy back (rain clutter — false echoes) └── Absorption: rain converts RF energy to heat (signal attenuation) Both effects depend on the ratio of raindrop size to wavelength. Raindrops are 1-5 mm. When the wavelength is much larger than the drop, the wave barely notices it. When the wavelength approaches the drop size, interaction is severe. Attenuation vs Frequency — The Numbers

Why Weather Radars Use S-Band A weather radar wants to measure rainfall. You'd think higher frequency is better — more scattering from drops. But the problem is attenuation shadowing: heavy rain close to the radar absorbs so much signal at high frequencies that you can't see storms BEHIND it. The nearby rain creates a shadow. S-band (3 GHz) penetrates through heavy rain at close range to see the severe thunderstorm 200 km away. The NEXRAD WSR-88D network — 159 radars covering the entire US — all operate at S-band for this reason. X-band (10 GHz) sees rain better at short range — good for ship radar, airport wind shear detection. But it can't see through the storm to what's behind it. Ka-band (35 GHz) can detect individual cloud droplets — but only within ~20 km. It's blind in any real precipitation.

The Radar Cross Section of Rain A volume of rain has a radar cross section per unit volume: η = (π⁵ / λ⁴) × |K|² × Z Where Z is the reflectivity factor (mm⁶/m³) — the sum of the sixth power of all drop diameters in a cubic meter. The λ⁴ in the denominator is why rain returns increase dramatically at higher frequencies. This is the basis of dBZ — the unit weather radars display. A thunderstorm has Z ≈ 50-60 dBZ. Light drizzle: 10-20 dBZ. The relationship between dBZ and rain rate follows Z = 200 × R^1.6 (Marshall-Palmer).

DESIGN SPEC UPDATED: ├── Rain attenuation: S-band 0.2 dB/km, X-band 2 dB/km, Ka-band 12 dB/km in heavy rain ├── Long-range surveillance → S-band (penetrates rain, 300 km range) ├── Fire control → X-band (precision, shorter range, accepts weather loss) ├── Rain clutter RCS ∝ 1/λ⁴ — higher frequency sees more rain backscatter └── Weather radar = S-band to see THROUGH near storms to far storms

PHASE 6: Don't Blind Yourself

Your transmitter fires a pulse: 1,000,000 watts. One microsecond later, you need to detect the echo: 0.000000000001 watts. That's a power ratio of 10¹⁸. If even one billionth of your transmit pulse leaks into the receiver, it's still a million times stronger than the target echo. The receiver doesn't just fail to detect the target — it's destroyed. This is the isolation problem. The same antenna both transmits megawatts and receives femtowatts. Separating these two by 180 dB — the difference between a lightning bolt and a firefly — is one of the hardest engineering challenges in radar. The Dynamic Range Challenge

The T/R Switch — The Gatekeeper The transmit/receive (T/R) switch protects the receiver during transmission and connects it during listening. It must: ├── Pass 1 MW to the antenna during transmit (microsecond switching) ├── Block transmit energy from entering receiver (>80 dB isolation) ├── Connect receiver to antenna during listen (nanosecond recovery) └── Introduce minimal loss on the receive path (100 dB.

The Minimum Range Problem While you're transmitting, you can't receive. The transmit pulse lasts τ seconds, and the T/R switch needs recovery time t_r. Any target closer than the minimum range returns its echo before you're ready to listen: R_min = c × (τ + t_r) / 2 For τ = 1 μs and t_r = 1 μs: R_min = 3 × 10⁸ × 2 × 10⁻⁶ / 2 R_min = 300 m Anything closer than 300 m is invisible. This is the radar's blind zone — the echo arrives while you're still shouting or still recovering from the shout. Military aircraft exploit this by flying at extremely low altitude, making the ground clutter mask their return in the radar's near-field confusion zone.

DESIGN SPEC UPDATED: ├── Dynamic range: 180 dB (1 MW transmit vs 1 pW receive) ├── T/R switch: circulator + PIN diode + gas tube → >100 dB isolation ├── Recovery time: ~100 ns for PIN diode, ~5 μs for gas tube ├── Minimum range: R_min = c(τ + t_r)/2 → 300 m with typical timing └── Any Tx leakage into receiver > -90 dBm saturates or destroys the front end

PHASE 7: Scan Without Moving

Your 3-meter dish rotates once every 10 seconds — a mechanical scan of the sky. But the dish weighs 500 kg and the motor needs 20 kW. The gears wear. The bearings fail. And you can only look in one direction at a time. While you're scanning east, a missile approaches from the west. You won't see it for 5 seconds. In those 5 seconds, the missile covers 5 km. The solution: no moving parts. Replace the dish with thousands of small antenna elements. Steer the beam by controlling the phase of the signal at each element. An electronic beam that can jump from east to west in microseconds. Phased Array — The Principle A phased array is a grid of antenna elements, each with its own phase shifter. By delaying the signal at each element by a precisely controlled amount, you create constructive interference in the desired direction and destructive interference everywhere else.

Modern Phased Arrays — The Numbers The AN/SPY-1 Aegis radar (on US Navy destroyers): ├── Elements: 4,350 per face (4 faces = 360° coverage) ├── Frequency: S-band (~3.3 GHz) ├── Peak power: 6 MW (combined from all elements) ├── Beam steering: ±60° from broadside ├── Beam repositioning time: 100 targets └── Can search, track, and guide missiles simultaneously The AN/APG-81 (F-35 fighter radar): ├── Elements: ~1,200 T/R modules ├── Frequency: X-band (~10 GHz) ├── AESA: each element has its OWN transmitter and receiver ├── Beam agility: >1,000 beam positions per second ├── Functions: air-to-air, air-to-ground, weather, EW — simultaneously └── Mean time between failures: >400 hours (vs ~200 for mechanical)

DESIGN SPEC UPDATED: ├── Beam steering: sin(θ) = λΔφ/(2πd) — electronic, no moving parts ├── Beam repositioning: <50 μs (vs 10 seconds for mechanical rotation) ├── AESA: each element has own T/R module → graceful degradation ├── Aegis SPY-1: 4,350 elements per face, 6 MW, tracks 100+ targets └── AESA enables simultaneous search, track, jam, communicate — one aperture

PHASE 8: Track the Target

Your radar sees a blip. Range 280 km, bearing 045°. Next scan, 10 seconds later: range 276 km, bearing 045.3°. Is that the same aircraft? Or a different one? Or a cloud? Or a bird? Or a random noise spike that happened to look like a target? You have noisy measurements, uncertain identity, and 10 seconds between updates. Welcome to the tracking problem. Detection gives you a snapshot. Tracking gives you a story — position, velocity, acceleration, predicted future position. Tracking is what turns a radar from a flashlight into a weapon system. Track Initiation — Is It Real? A radar has a false alarm rate. With a detection threshold set for P_fa = 10⁻⁶ (one false alarm per million range cells), and a radar that checks 10,000 range cells per pulse at 500 pulses per second: False alarms per second = 10⁻⁶ × 10,000 × 500 = 5 per second You get 5 phantom targets every second from noise alone. How do you tell real targets from phantoms? M-of-N detection: require a target to appear in at least M out of N consecutive scans at approximately the same position. A real target appears consistently. Noise appears randomly. ├── 2-of-3 rule: target confirmed if seen in 2 of 3 consecutive scans │ └── False track probability: P_fa² × geometry ≈ negligible ├── 3-of-5 rule: more conservative, used in long-range surveillance └── Typical time to confirm track: 20-30 seconds (2-3 antenna rotations)

The Kalman Filter — Optimal Estimation Once you have a confirmed track, you need to estimate the target's true position and velocity from noisy measurements. The Kalman filter does this optimally. The state vector at each update: x = [x, y, z, vx, vy, vz] (position and velocity in 3D) Two steps per update: 1. PREDICT: use the physics model to predict where the target should be x_predicted = F × x_previous + noise (F is the state transition matrix — constant velocity model) 2. UPDATE: combine prediction with new measurement x_updated = x_predicted + K × (measurement - H × x_predicted) (K is the Kalman gain — weights prediction vs measurement)

Track Loss and Maneuver Detection The constant-velocity Kalman filter predicts straight-line motion. When the target turns, the prediction diverges from measurement. The innovation (measurement minus prediction) spikes. If innovation exceeds a threshold → maneuver detected → increase process noise → the filter trusts measurements more than predictions until the maneuver stabilizes. If the target disappears for multiple scans: ├── 1 missed scan: coast — extrapolate track, keep searching ├── 2-3 missed scans: tentative loss — expand search gate ├── 5+ missed scans: track drop — target lost, resources reallocated A modern system like Aegis maintains 100+ simultaneous tracks, each with its own Kalman filter, each predicting independently. The computer updates all of them in real time between antenna scans.

DESIGN SPEC UPDATED: ├── False alarms: ~5/second from noise at P_fa = 10⁻⁶ — need M-of-N to confirm ├── Track initiation: 2-of-3 or 3-of-5 consecutive detections at consistent position ├── Kalman filter: predict → measure → update → repeat (optimal state estimation) ├── Position accuracy: 10× improvement over single measurement after 5+ updates └── Track 100+ targets simultaneously, each with independent state estimator

PHASE 9: See Stealth

Your radar was built to detect a fighter jet at 300 km — a 10 m² radar cross section. Then someone reshapes that jet into flat angles and coats it with radar-absorbing material. Its RCS drops from 10 m² to 0.001 m². Same aircraft, same metal, same size — but now it reflects 10,000 times less energy. Your 300 km radar just became a 30 km radar. The R⁴ vs RCS Relationship From the radar equation, detection range goes as the fourth root of RCS: R_max ∝ σ^(1/4)

How Stealth Works — Three Mechanisms 1. Shaping — deflect energy away from the radar ├── Flat panels at precise angles (F-117: all flat facets) ├── Blended curves that scatter in all directions (B-2, F-22) ├── Eliminate corner reflectors (engine inlets, weapon bays internal) ├── Leading edges aligned to reflect only in 2-4 cardinal directions └── Effect: -20 to -30 dB RCS reduction from shaping alone 2. Radar-Absorbing Material (RAM) — convert RF energy to heat ├── Layers of lossy dielectric and magnetic material ├── Impedance matching: minimize reflection at the surface ├── Frequency-dependent — tuned for threat radar bands (X-band typically) ├── Thickness ≈ λ/4 for resonant absorption └── Effect: additional -10 to -20 dB 3. Planform alignment — concentrate all reflections into few narrow spikes ├── Leading edges, trailing edges, panel joints all at same angle ├── Energy reflects in 4-8 narrow lobes instead of all directions ├── If the radar isn't in one of those lobes → very low RCS └── F-22 has ~4 spike directions. Don't park your radar at those angles.

Counter-Stealth — How to Find What Hides Stealth aircraft are optimized against X-band (8-12 GHz) — the frequency of most fighter and SAM radars. But stealth shaping and RAM lose effectiveness at lower frequencies, where the wavelength approaches the aircraft's size. VHF radar (30-300 MHz): wavelength 1-10 m ├── At 150 MHz, λ = 2 m — comparable to aircraft wing edges ├── Stealth shaping doesn't work when λ ≈ feature size (resonance scattering) ├── RAM is impractical at these frequencies (would need meter-thick coatings) ├── Russia's Nebo-M uses VHF — can detect F-22 class targets at ~200 km └── But: VHF beamwidth is huge (30°+) → poor precision, can't guide missiles Bistatic radar: transmitter and receiver in different locations ├── Stealth shaping deflects energy AWAY from the transmitter ├── A receiver elsewhere may catch the deflected energy ├── Forward-scatter: target crosses the line between Tx and Rx │ └── RCS becomes ~target's physical cross-section, regardless of stealth └── Challenge: requires synchronized stations, complex processing Multi-static networks: many receivers, few transmitters ├── Target can't be stealth in all directions simultaneously ├── Some receiver always sees a non-stealth angle └── Modern approach: use cell phone towers as "illuminators of opportunity"

DESIGN SPEC UPDATED: ├── RCS reduction: 10 m² → 0.001 m² → range drops from 300 km to 53 km (R ∝ σ^1/4) ├── Stealth mechanisms: shaping (-20 to -30 dB), RAM (-10 to -20 dB), alignment ├── Stealth optimized against X-band — loses effectiveness at VHF wavelengths ├── Counter-stealth: VHF radar, bistatic, multi-static, passive (cell tower illumination) └── Stealth ≠ invisible — it means visible too late to react effectively

PHASE 10: Friend or Foe