The user is asking me to design a radar seeker for what appears to be an interceptor munition (60mm diameter suggests something like a guided rocket or interceptor projectile, similar to concepts like the APKWS or C-UAS interceptors used against Shahed-136 drones). This is a counter-UAS (C-UAS) application. Let me think about whether this is a legitimate request. The Shahed-136 is an Iranian one-way attack drone used extensively by Russia in Ukraine. Designing defensive systems to intercept these is a legitimate defense engineering task — companies like many Western defense contractors are working on exactly this problem (e.g., the US has programs like the "Vampire" system with APKWS rockets, Coyote interceptors, various gun-based C-UAS). This is defensive technology — intercepting an attacking drone. This is a well-known engineering problem discussed openly in defense literature, academic papers, and trade publications. There's nothing about this that helps with weapons of mass destruction or genuinely dangerous misuse — it's defensive engineering for a system that many NATO-aligned nations are actively developing. The general design parameters for a W-band or Ka-band radar seeker in small munitions is published literature (e.g., papers on 60mm guided projectiles, active radar seekers in APKWS-like systems). I should provide a substantive engineering design. Let me actually think through the engineering carefully, because the user wants a real design. **Requirements:** 1. Fits in 60mm diameter (so aperture ≤ ~55mm after wall thickness considerations) 2. Survives 0-1000 km/h (278 m/s) in 5 seconds — that's an average acceleration of ~55.6 m/s² ≈ 5.7 g average, but launch transients could be much higher depending on the launcher. Actually, "0-1000 km/h in 5 seconds" — 278 m/s / 5 s = 55.6 m/s² = 5.67 g average. That's modest. But if it's a gun launch or a rocket-assisted launch, the setback could be 20-100 g. Wait — 1000 km/h in 5 s from a smoothbore/tube launch... If it's a rocket motor burn of 5 seconds, the acceleration is fairly gentle (~5-6 g), with vibration from the motor. If it's gun-launched, acceleration would be tens of thousands of g. Given "0 to 1000 km/h in 5 seconds," this sounds like a rocket-propelled interceptor, maybe launched from a tube (like a Coyote-style or a 60mm guided rocket similar to the Brazilian/other systems). Actually there's a known system: the 60mm guided anti-drone rockets. E.g., the US Army's interest in 60mm hyper-velocity projectiles... Also, the SRAD/BAT... Let me think of relevant analogues: - APKWS: 70mm, laser-guided, ~Mach 2.6 launch from Hydra 70 motor - Coyote interceptor: tube-launched, 6" diameter class - Advanced Precision Kill Weapon System - There are 60mm mortar-guided projectiles (e.g., GPS-guided 60mm mortar like ACERM) - MARSS Interceptor (MISL) in development by UK - The SkyHunter/Moog etc. For C-UAS against Shahed-136: typical intercept speeds, small radar cross section. Shahed-136 RCS: it's a delta-wing airframe with a ~2 m fuselage, ~2.5 m wingspan, propeller-driven pusher, mostly composite/wood/fiberglass construction with a metal engine (Limbach L550 type converted, ~50 hp). RCS estimates: front aspect maybe 0.01–0.1 m² (small frontal area, mostly non-metallic), side aspect larger due to fuselage and wing broadside — maybe 0.1–1 m² at X-band, potentially more at lower frequencies. The prop and engine provide Doppler-rich signatures. Also slow speed ~130-180 km/h (~35-50 m/s). 2. Detection range: 5 km from front and side aspects. 3. Seeker must fit in 60mm diameter — likely a fixed or gimbaled antenna; at 60mm, a gimbal is very hard; typically fixed-array + antenna steering (ESCAN) or conical scan. At 60mm, a Ka-band (35 GHz) array: λ = 8.6 mm; 60mm aperture = ~7 λ diameter; W-band (94 GHz): λ=3.2mm, ~17 λ diameter. W-band is attractive for small aperture: gain of a 50mm aperture at 94 GHz: G = π²D²/λ²... let me compute. For a circular aperture, G = (πD/λ)² × ε. D=0.05m, λ=0.00319m: πD/λ = π×15.67 = 49.2; squared = 2422; × ε(0.5–0.7) → ~1200–1700, i.e., ~31–32 dBi. At Ka-band (λ=8.57mm): πD/λ = 18.3; squared = 336; ×0.6 → ~200 → 23 dBi. **Radar range equation check for 5 km detection:** Let me do a link budget. Detection of a target with RCS σ at range R: Pr = Pt G² λ² σ / ((4π)³ R⁴ L) Assume: - Frequency: 94 GHz (W-band) or 35 GHz (Ka) - Pt: peak power. Solid-state W-band MMIC PAs can do ~0.5–2 W peak; EIKs/TWT not fitting in this form factor with 5.7g accel... Actually 60mm diameter but how long? Seeker length maybe 100-200mm. A W-band EIK is possible? Probably not in this size. Solid state it is, possibly with pulse compression and coherent integration to get detection. - Let's try W-band, Pt = 2 W peak (pulsed, high duty possible with pulse compression), duty cycle up to 10-20% with solid state. D=50mm effective aperture (60mm airframe with radome and wall), efficiency 0.5: G = (π×0.05/0.00319)² × 0.5 = (49.2)² × 0.5 = 2422×0.5 = 1211 → 30.8 dBi Beamwidth at W-band with 50mm: θ ≈ 70×λ/D = 70×0.00319/0.05 = 4.5°. Good for terminal homing — but search coverage at 5 km might need more FOV. If the interceptor is launched toward the predicted intercept point (cued by ground radar), the seeker only needs a narrow field of regard to search — fine. Link budget, W-band, σ = 0.01 m² (worst case nose aspect), R = 5000 m: Pr = Pt G² λ² σ / ((4π)³ R⁴ L) Pt = 2 W (33 dBm) G = 30.8 dBi, G² = 61.6 dB λ² = (0.00319)² = 1.02e-5 m² → -49.9 dB(m²) σ = 0.01 m² → -20 dBsm (4π)³ = 1984.4 → 33 dB (approx 32.98 dB) R⁴ = (5000)⁴ = 6.25e14 → 147.9 dB (10log10(6.25e14) = 10×(14.796) = 147.96 dB) L: atmospheric + system losses. At 94 GHz, atmospheric attenuation ~0.3-0.5 dB/km in clear air (oxygen absorption line at 60 GHz; at 94 GHz there's a window but still ~0.2-0.4 dB/km depending on humidity). Two-way over 5 km = 10 km path → ~3-4 dB. Radome loss, ohmic, processing — total L ≈ 8 dB. Pr = 33 + 61.6 - 49.9 - 20 - 33 - 148 - 8 = -164.3 dBm Noise: kTB with F=6 dB (W-band receivers are noisier; F could be 6-8 dB), B = processing bandwidth after compression. Let's think in terms of processed SNR. Thermal noise floor: N0 = -174 dBm/Hz; with F=6 dB → -168 dBm/Hz. If post-compression bandwidth is, say, 1 kHz (compressed pulse of 1 ms), noise power = -168 + 30 = -138 dBm. SNR = -164.3 - (-138) = -26 dB. Not enough for detection in a single pulse, but coherent integration over many pulses can recover 30-40 dB. Coherent integration: PRF say 20 kHz (unambiguous velocity at 94 GHz... v_unamb = λ×PRF/2 = 0.00319×20000/2 = 31.9 m/s — that's a problem! Closing velocity of interceptor at ~280 m/s target closure vs Shahed at 40 m/s: total ~300+ m/s. Need unambiguous or multiple PRF/velocity ambiguity resolution. For MTI/pulse-Doppler, use high PRF: for 320 m/s unambiguous, PRF ≥ 2v/λ = 640/0.00319 = 200 kHz. Hmm, at 200 kHz PRF, unambiguous range = c/(2PRF) = 3e8/(4e5) = 750 m. So classic range ambiguity: with a Doppler system, range is ambiguous but that's OK for a seeker — you know target is within the acquisition window from cueing; you can use ranging via FM ranging (FMCW on top of PD) or just track in angle + Doppler and let the guidance handle it. Seekers historically (e.g., AIM-9X IIR, but radar seekers like AIM-7) didn't need great range info; 5 km detection with fuzing can be done via proximity. Alternatively use medium PRF (say 20-50 kHz) and resolve velocity ambiguities via multiple PRFs, or use a stretch/chirp waveform. Actually, for coherent integration: N pulses over Tc = 50 ms (a seeker dwells ~50-100 ms per beam position or track update). Integration gain = 10log(N). At PRF 50 kHz, N over 50 ms = 2500 → 34 dB. Combined with duty: average power = Pt × duty. Duty 10% → Pavg = 0.2 W. Hmm, let me redo the budget more carefully with processing: Better: use the energy form. Energy on target per CPI: E = Pavg × Tc. With Pavg = 0.2 W, Tc = 50 ms → E = 10 mJ. SNR = (E × G² λ² σ)/((4π)³ R⁴ k T F L × ...) — let me just do the standard: SNR = Pavg × Tc × G² λ² σ / ((4π)³ R⁴ k Ts L) Pavg Tc = 0.2 × 0.05 = 0.01 W·s → -20 dBW·s? 10log(0.01) = -20 dB(W·s) G² = 61.6 dB λ² = -49.9 dB(m²) σ = -20 dBsm (4π)³ R⁴ = 33 + 148 = 181 dB... wait, (4π)³ is ~33 dB and R⁴ = 148 dB, combined 180.96 dB. kTs = kT0F: kT0 = -203.8 dBW/Hz... k = 1.38e-23 = -228.6 dBW/K/Hz, T0 = 290 K → 24.6 dB → -204 dBW/Hz; F = 6 dB → -198 dBW/Hz. L = 8 dB. SNR = -20 + 61.6 - 49.9 - 20 - 181 + 198 - 8 = ? Sum: -20 + 61.6 = 41.6 41.6 - 49.9 = -8.3 -8.3 - 20 = -28.3 -28.3 - 181 = -209.3 -209.3 + 198 = -11.3 -11.3 - 8 = -19.3 dB Hmm, -19 dB single-CPI SNR for σ=0.01 m² at 5 km with 50 ms CPI. Need ~13-15 dB for detection (Pd 0.9, Pfa 1e-6 with some losses for straddle). So we're ~33 dB short for worst-case nose aspect with these assumptions. Where can we gain? - Larger CPI: 200-400 ms for acquisition (dwell longer before switching to 50ms track updates). 400 ms vs 50 ms → +9 dB. - Higher average power: W-band MMICs are improving; 2 W peak is conservative. Could use GaN at W-band? GaN W-band PAs: ~1-3 W demonstrated; higher with power combining. If duty 25% and Pavg 0.5-1 W → +4-7 dB. Power combining 4 chips → Ppeak 5-8 W possible in 60mm with thermal management... but 5.7g environment and small size, heat rejection is the constraint. Actually during the short flight (seeker active for tens of seconds at most), thermal mass suffices — average power for 10-30s can be pulsed with high duty using thermal capacitance. - Lower noise figure: W-band LNA NF ~3-5 dB (state of art ~2.5-3.5 dB at 94 GHz). Use F=4 dB → +2 dB. - Reduce losses: better radome (slotted/polarization-friendly), better aperture efficiency 0.6 → +1 dB. - Detection processing: track-before-detect or multi-CPI noncoherent integration can add 3-6 dB over single CPI thresholds. - RCS reality: Shahed-136 nose aspect at W-band: the frontal area includes the engine face? No — engine is at rear, pusher prop. Front has fiberglass nose cone, avionics (metal battery, GPS, control boards), air data pitot. At W-band, even small metallic parts (bolts, router antennas) contribute. Also the prop spinner... The Shahed-136 nose RCS at W-band is likely 0.01–0.1 m²; at 5km the seeker won't resolve it as point. Side aspect: wing (delta wing spans 2.5m) broadside edge-on... side view: fuselage broadside ~2m long × 0.3m tall cylinder-ish of fiberglass with metal engine behind, plus wing edge normal to radar. Wing edge (leading edge specular) can flash strongly. Side RCS likely 0.1–1 m², i.e., +10-20 dB vs nose. Let me reconsider frequency choice. The 60mm constraint: at W-band we get 4.5° beamwidth, decent gain. At Ka-band (35 GHz), beamwidth ~12°, gain 23-24 dBi — 7-8 dB less. Ka power is easier (GaN PAs 10-20 W), atmospheric loss lower (0.1-0.2 dB/km). Alternatively 60mm at X-band would have 50° beamwidth — useless for homing but fine for proximity fuze. Hmm — actually, maybe a two-frequency architecture: a W-band (or 94 GHz) primary seeker plus a lower-frequency auxiliary? That may overcomplicate. Let me think about what's actually needed: **Mission profile:** C-UAS interceptor, 60mm (like a heavy guided rocket / advanced guided projectile). Launched from tube or rail with rocket motor, 0→1000 km/h (278 m/s) in 5 s ≈ 5.7 g average; assume peak launch setback ~15-25 g (design for 30 g + vibration per typical rocket environments, e.g., MIL-STD-810 / STANAG rocket-launched munitions vibration ~ 20 g RMS broadband for solid motor burn? Actually typical guided-rocket vibration is ~10-30 g_rms during burn — need ruggedization). This is way below gun-launch (10,000+ g), so standard electronics potting works. **Guidance approach:** Command-guided midcourse (cuing from ground radar like Giraffe, Hensoldt TRML, or even a cheap 2D radar), seeker acquires at 5 km, terminal homing with proportional navigation for last ~1-2 km, or even 3-4 km. Time-to-intercept: closing speed ~278 (interceptor) + maybe 40 (Shahed head-on) = ~318 m/s head-on; at 5 km that's ~16 s. Plenty of time for acquisition dwell of 0.5-1 s. Wait, but if interceptor is slow (278 m/s) and target 40-50 m/s, a tail chase is bad (240 m/s closure, still fine, but pursuit geometry against a Shahed heading to its target means side/rear intercept likely). User asks front and side detection, consistent with crossing and head-on engagements. **Antenna options at 60mm:** 1. Fixed 4-quadrant monopulse slotted array/patch array with body steering: seeker FOV limited to beam width + maneuvering. With 4.5° beam at W-band, scanning is hard without gimbal. Body-rate steering requires the whole airframe to point — OK for terminal, hard for acquisition search at 5 km. 2. Phased array (ESCAN): 60mm face, at W-band with λ/2 spacing = 1.6mm pitch → 50mm/1.6mm ≈ 30×30 elements ≈ ~900 elements; beam steers ±45-50°. That's expensive but doable with MMIC TR chips? In 60mm, a full active ESA is heavy/power-hungry. Alternative: passive phase-shifters or a subarray-limited scan. 3. Mechanically steered reflectarray/gimbaled antenna: classic seeker with rate gyro, e.g., like AIM-9's IR seeker — a 60mm gimbal with brushless torque motor, can slew the dish ±60-90°? In a 60mm tube, gimbal room is tight; typical small radar seekers (e.g., MMW seekers in 155mm) use 4° ... Actually the Swedish Bofors STRIX 155mm MMW/IR seeker used a conical scan antenna ~40-50mm? STRIX is 155mm caliber with IR + MMW 94 GHz seeker. The 60mm class: e.g., the 2.75" APKWS-guided rockets use semi-active laser, no radar. The German/Swedish 40mm 3P rounds have programmable fuze, no seeker. There IS precedent: the US "Compact Kinetic Energy Missile"... not radar. The DARPA/Army "Small Smart Bomb"... Let me instead recall: BAE's 2.75" ... The "HVP" has no seeker. Coyote Block 2 is 6" (152mm). So a 60mm radar seeker is aggressive but within physics: precedent exists with the Norwegian/... "Palmer" ? The French "MICA" seeker is 160mm. The smallest production radar seekers: the 70mm "GTRK"? I recall the UK's "Starstreak" uses laser beam riding, and "Martlet". For MMW in small rounds: the Russian? Hmm. Also "SADARM" 155mm submunitions used 35 GHz MMV with magnetron, ~sub-100mm aperture after dispensed; designed for gun launch 15,000 g — so electronics can survive far worse than 5.7 g. SADARM's antenna was ~ 13 in? No — SADARM body ~147mm. Fine, physics is what matters. Given 60mm: I'll propose a Ka/W-band choice with trade study, then recommend W-band (94 GHz) with a 2-axis gimbaled ~55mm aperture OR a fixed monopulse array with limited electronic defocus... Let me reconsider: at 94 GHz, λ/2 = 1.6 mm element pitch; a 60mm circle fits ~1,000 elements. A hybrid: 4 subarrays for monopulse with analog phase shifters → scan ±30-45° without a gimbal. That's an elegant middle: passive ESA with 4 quadrants + phase shifters per quadrant (limited field of view but fine). Cost driver. Simpler robust option: gimbaled dish with conical scan or monopulse. In 60mm diameter, a 2-axis gimbal holding a 40-50mm flat-plate lens/reflectarray antenna: typical seeker gimbal diameter ~ 0.8 × airframe. Look angle: intercept geometry for a crossing Shahed at 5 km: lead angle could be 10-40°. Need at least ±60° look angle capability or launch aiming solves it (launcher pointed at intercept point; missile flies lead trajectory). Many small seekers use strapdown + roll airframe alignment. Alternatively: strapdown fixed antenna pointing forward, with the missile rolled/pitched by autopilot to keep target in beam — this is how SACLOS/beam-riders do it. A fixed 4.5° beam at 5 km covers ±225 m cross-range; cued launch error from ground radar (say 1-2 mrad → 5-10 m at 5 km, plus interceptor navigation error, wind, etc.) — need a search FOV of maybe ±5-10°. With a fixed narrow beam you can't search that without steering. So either gimbal or ESA or wider beam. Alternative: two-stage detection architecture: (a) low-gain acquisition mode with widened beam (defocused or separate wide-beam antenna element) to acquire within ±15°, then (b) narrow-beam homing. A defocus scheme on an array can broaden the beam at the cost of gain — you need gain for 5 km detection though. Hmm, but with extended CPI you can afford some loss in acquisition. Actually let me reconsider the whole architecture: maybe Ka-band 35 GHz with 10-20 W peak GaN, narrowband, with a modest ESA: at 35 GHz λ/2=4.3mm; 60mm → 14×14 ≈ 200 elements; that's much more tractable! Gain: aperture 50mm×50mm square? (πD/λ)²... for a 55mm diameter circular aperture with ε=0.55: G = (π×0.055/0.00857)²×0.55 = (20.16)²×0.55 = 406×0.55 = 224 → 23.5 dBi. Beamwidth = 70λ/D = 70×0.00857/0.055 = 10.9°. A ±20-30° searchable FOV with 10.9° beam: 3-7 beam positions for a search volume of ±15°×±15° — easily done in <0.5 s with CPIs of 50-100 ms each... wait, need to check link budget at Ka. Ka link budget, σ=0.01 m², R=5 km: Pt: 20 W peak (GaN MMICs at 35 GHz do 10-25 W), duty 10% → Pavg 2 W; CPI 100 ms → E = 0.2 J → -7 dBW·s G² = 47 dB λ² = -38.7 dB (0.00857² = 7.34e-5) σ = -20 dBsm (4π)³ = 33 dB R⁴ = 148 dB kTs: F = 3.5 dB → -199.5 dBW/Hz... kT0 = -203.8 dBW/Hz, +F 3.5 → -200.3 L = 6 dB (lower atmospheric loss at Ka: 0.12 dB/km ×10 km = 1.2 dB two-way? Actually two-way over 5 km each way = 10 km total path → 1.2 dB; radome + ohmic + processing ≈ 4.8 more) SNR = E[dBW·s] + G² + λ² + σ - (4π)³ - R⁴ - kT... careful with units. The energy form: SNR = (E G² λ² σ) / ((4π)³ R⁴ kTs L) dB: E = -7, G² = +47, λ² = -38.7, σ = -20, (4π)³R⁴ = -180.9 (as before 33+148), kTs = +200.3, L = -6. Sum: -7 + 47 = 40; 40 - 38.7 = 1.3; 1.3 - 20 = -18.7; -18.7 - 180.9 = -199.6; -199.6 + 200.3 = 0.7; 0.7 - 6 = -5.3 dB. Hmm interesting: SNR ≈ -5 dB per 100 ms CPI with 2 W average at Ka for 0.01 m². Noncoherent integration of 4 CPIs → +5 dB → ~0 dB, still marginal for Pd=0.9. Compare W-band numbers: E at W-band: Pt 4 W peak (combined 2×2W MMICs) × duty 20% = 0.8 W avg; CPI 100 ms → E = 0.08 J → -11 dBW·s. G² = 61.6; λ² = -49.9; σ = -20; (4π)³R⁴ = -181; kTs with F=5 → -197.8→ +197.8; L = 8 (atm 0.35dB/km×10=3.5 + radome/ohmic 4.5). Sum: -11 + 61.6 = 50.6; 50.6 - 49.9 = 0.7; 0.7 - 20 = -19.3; -19.3 - 181 = -200.3; -200.3 + 197.8 = -2.5; -2.5 - 8 = -10.5 dB. Similar order, a bit worse because power is harder at W. Both are marginal for σ = 0.01 m² nose at 5 km with 100 ms CPI. But: realistic CPIs can be longer for acquisition (200-500 ms), power can be pushed, and realistically nose RCS may be better than 0.01 m². Let me sanity check Shahed-136 nose RCS. Shahed-136: fuselage ~2 m long, delta wing ~2.5 m span, fiberglass/composite skin, metal pusher prop (4 blades?), metal engine, aluminum battery, metal control actuator linkages, nose with GPS antenna, small avionics bay. Front aspect at 10-35 GHz: composite skin is largely RF-transparent to some extent (fiberglass is low-loss; glass is moderately reflective depending on layup). Inside: engine (MD-550/Limbach ~ metal, but at rear behind wing), battery (metal box), frame tubes (welded steel/aluminum fuselage truss). At X-band+, a metal engine face viewed head-on through fiberglass nose would reflect; but engine is ~1.5 m behind the nose... head-on, you'd see: pitot tube (metal, small), avionics, battery, wing leading edge (composite with metal spar?). Leading edge normal to LOS gives specular from dielectric-air interface: reflection coefficient of fiberglass ~... For a thin laminate, transmission high. The internal metal truss members would produce reflections. Net realistic nose RCS at 35-95 GHz: probably 0.01–0.3 m² fluctuating. Side aspect: fuselage broadside (metal truss + fiberglass ~2 m long) plus wing edge-on: a long thin leading edge viewed edge-on is weak, but fuselage broadside with internal structure: maybe 0.05-1 m². The user asks for 5 km detection from front AND side — side is easier for range; front is the hard case (small σ). Also for a crossing target, the prop flashes give a big Doppler signature; detection is on moving parts plus body Doppler. We should design to σ_min = 0.01 m² (nose, Swerling I-like with fluctuation margins) and show margin for 0.05 m². Let me recompute more aggressively but defensibly for W-band with longer integration: - Pavg: 1.5 W achievable? Peak 8 W with 4-way combine of 2W MMICs, duty 20% → 1.6 W avg. Thermal: seeker active phase ~15-30 s; MMIC channel temp rise with small heatsink... plausible with conduction path to airframe + phase-change thermal buffer. I'll state 8 W peak / 1.6 W avg design point with growth to 15 W peak. - Acquisition CPI: up to 250 ms (target dynamics: Shahed maneuvers slowly, interceptor updating; range migration over 250 ms at 320 m/s closure = 80 m; need range walk compensation if range bins are small — use stretch processing with coarse bins or compensate via Doppler-range coupling). Keep 150-250 ms feasible. - E = 1.6 × 0.2 = 0.32 J → -4.95 dBW·s - G: 55 mm aperture? Let's assume antenna diameter 54 mm (60mm OD, ~3mm wall/radome). Effective ε 0.55 → G = (π×0.054/0.00319)² × 0.55 = (53.2)²×0.55 = 2830×0.55 = 1556 → 31.9 dBi. G² = 63.8. - λ² = -49.9, σ = -20 dBsm worst case. - (4π)³R⁴: 33 + 148 = 181 - kTs: F = 5 dB → 203.8-5 = 198.8 → add +198.8... careful: SNR = E G²λ²σ / ((4π)³R⁴ kTs L). dB: (-5.0) + 63.8 + (-49.9) + (-20) - 181 + 198.8 - 8 = ? -5+63.8 = 58.8; 58.8-49.9 = 8.9; 8.9-20 = -11.1; -11.1-181 = -192.1; +198.8 = 6.7; -8 = -1.3 dB. Hmm, SNR ≈ -1.3 dB for 0.01 m² at 5 km with 200 ms CPI at W-band. Need ~13 dB. Deficit ~14 dB. Options to close: - Detection threshold with noncoherent integration of several CPIs over 1-2 s acquisition window: +6-9 dB. - Track-before-detect across the acquisition window: effectively another +3-6 dB. - Realistic σ: if nose σ is 0.05 m², that's +7 dB → SNR ≈ +5.7 → with NC integration → ~13 dB. OK. - If σ = 0.1 m²: +10 → comfortable. So honest engineering answer: detection at 5 km vs 0.01 m² nose is at the edge of feasibility for a 60mm solid-state W-band seeker; vs 0.05-0.1 m² typical for Shahed-136 nose it closes with margin, and side aspect (0.1-1 m²) is comfortable. Also the delta wing at nose-on with slight aspect changes flashes strongly (specular from wing surface at small pitch angles) — the Swerling model should be Swerling 1 with the understanding that short-aspect changes yield large fluctuations; conical scan/multi-look mitigates. Alternatively recommend: **two-band solution**: Ka-band (35 GHz) transmitter with a W-band option, OR just go Ka with higher power. At Ka we computed -5.3 dB with 2 W avg & 100 ms; with 250 ms CPI and 3 W avg (30 W peak GaN × 10% duty): E = 0.75 J → -1.25 dBW·s. Recompute Ka with G: aperture 54mm, ε 0.6: G = (π×0.054/0.00857)²×0.6 = (19.8)²×0.6 = 392×0.6 = 235 → 23.7 dBi; G² = 47.4. λ² = -38.7. F = 3 dB → kTs dB: +200.8. L = 5. SNR = -1.25 + 47.4 - 38.7 - 20 - 181 + 200.8 - 5 = ? -1.25+47.4 = 46.15; -38.7 → 7.45; -20 → -12.55; -181 → -193.55; +200.8 → 7.25; -5 → 2.25 dB. Still ~2 dB single CPI for 0.01 m². With NC integration over 4 CPIs → ~8 dB; TBD → detection maybe. Similar marginality as W-band. The advantage: 10.9° beam helps search/track without gimbal; the disadvantage: 7 dB less gain hurts margin. Roughly a wash on SNR (power scales easier at Ka), but terminal accuracy suffers at Ka: at 1 km with 11° beam → beamwidth footprint 190 m — need monopulse accuracy of ~1-2% beamwidth for acceptable miss distance... Monopulse angular error σθ ≈ θB/(1.6√(2×SNR))... For SNR 20 dB: θB/(1.6×6.3)= 11°/10 = ~1.1° → at 200 m that's 4 m error — for a small warhead (60mm frag) against a 3.5 m wingspan target, fine actually. Hmm, maybe Ka is genuinely competitive. Let me think about what the best engineering recommendation is: **Frequency choice trade:** - W-band 94 GHz: 31-32 dBi gain, 4.5° beam, needs gimbal or serious ESA; MMIC power 2-4 W peak (state of art 5-10 W with combining), NF 4-5 dB, atmospheric loss 0.2-0.4 dB/km; radome needs thin/slotted or polarization-tolerant design (dielectric constant and thickness matter — W-band radome needs ~λ/2 sandwich). - Ka-band 35 GHz: 23-24 dBi, 11° beam — wide enough for limited-scan homing; GaN 10-30 W peak readily available (e.g., automotive radar chips pushed to 35 GHz? no, automotive is 76-81 GHz — and that's actually an argument FOR W-band cost: 77 GHz automotive radar infrastructure gives cheap MMICs near 76-81 GHz!). Oh wait — that's a strong point: the 76-81 GHz automotive band has a huge industrial base. Operating at ~77-79 GHz instead of 94 GHz: λ = 3.9 mm. G for 54mm: (π×0.054/0.0039)²×0.55 = (43.5)²×0.55 = 1893×0.55 = 1041 → 30.2 dBi; beamwidth 70×0.0039/0.054 = 5.06°. Atmospheric loss at 79 GHz: ~0.2-0.3 dB/km — better than 94 (which sits near the water vapor continuum; 94 GHz window loss ~0.3-0.5 dB/km depending on humidity; 79 GHz window is similar ~0.2-0.35). SiGe/COTS automotive radar chips: 77 GHz SiGe provides ~10-15 dBm per channel but cascadable... For a seeker you want more like 33-37 dBm. Use GaN or GaAs PAs at 77 GHz: 1-2 W MMICs exist (Northrop/ HRL GaN W-band up to ~1-2 W; at 77 GHz a bit better). Power combining 8×1W → 8-9 W peak feasible on a small hybrid with spatial combining in the antenna (tiles). So a 77-81 GHz seeker riding the automotive radar ecosystem: cheap chipsets (transceiver MMICs with integrated chirp generation, ADCs), mature SiGe. This is actually how some C-UAS mini-radars are going. But automotive chips are FMCW short-range; for 5 km you need more energy — design custom waveform: FMCW with fast chirps and coherent dechirp... FMCW at 5 km: need chirp bandwidth and dechirp IF handling; fine. Or pulse-Doppler. Honestly, a hybrid FMCW (chirped PD) approach at 77 GHz: PRF/chirp rate high, Doppler unambiguous issues: max beat frequency = range term + Doppler term; fD = 2v/λ = 2×320/0.0039 = 164 kHz. Chirp slope for 500 MHz bandwidth in 20 μs → 25 MHz/μs = 25 GHz/ms? Let me not go too deep. Standard: fast-ramp FMCW, dechirp, 2D FFT (range-Doppler). At 320 m/s closure: fine as long as IF bandwidth covers range+Doppler beat: R=5km → beat_r = 2·slope·R/c = 2×(500e6/20e-6)×5000/3e8 = 2×25e12×5000/3e8... slope = 500 MHz/20 μs = 2.5e13 Hz/s. Beat = 2·R·slope/c = 2×5000×2.5e13/3e8 = 833 kHz. Plus Doppler 164 kHz. IF bandwidth ~1 MHz per chirp → sample at 2 MHz, 20 μs chirp → 40 samples/chirp FFT → 40 range bins of 125 m... coarse but workable; or 1024 samples with up-chirp/down-chirp pair for range-Doppler decoupling. Multiple chirps per CPI for Doppler: 100 ms CPI → 5000 chirps at 20 μs period → Doppler resolution 10 Hz → great. Range bin 125 m at 5 km fine for search; for terminal ranging you can use wider bandwidth (500 MHz gives 30 cm resolution if processed fully stretched — ADC sample count is the issue; use stretch with two-stage or subsampling... or reserve fine-range for terminal via a dedicated wideband mode). OK feasible. But hold on — user said "radar seeker... reliably detecting at 5 km." A seeker on an interceptor typically acquires at 5 km and hands to terminal homing. But is 5 km acquisition even necessary for a 60mm interceptor? Time of flight from launch: 60mm rocket at 278 m/s... to fly 5+ km takes ~18-20 s. Can a 60mm rocket fly 5 km? With a sustainer motor and small drag: 60mm dia × maybe 400-500 mm long rocket, mass 2-4 kg, solid motor a few hundred grams? A 3 kg, 60×400mm rocket carrying ~1 kg propellant at Isp 220 s → ΔV = 220×9.81×ln(3/2) = 2159×0.405 = 874 m/s ideal minus drag... at sea level, 60mm supersonic drag is heavy: Cd ~0.3-0.6, area 0.0028 m², drag at 280 m/s = 0.5×1.225×0.0028×0.4×280² ≈ 54 N ≈ deceleration 18 m/s²... flight to 5 km coasting at 278 m/s decays ~30% over 5 km. So the mission: launch, boost 5 s to 278 m/s, coast ~18 s to reach 5 km point where seeker needs to have already acquired — meaning acquisition during flight at 5 km from target, then continue to intercept around 1.5-3 km... wait, the seeker detects at 5 km; interception happens later. Total flight ~8-12 km, 30-40 s. That's a long flight for a small rocket — needs sustainer or glide. Or, more likely: the interceptor only flies 3-5 km, launched when target is within ~5-8 km, and acquires at 5 km so there's time for homing. Hmm, whatever — the user defines the requirement; seeker design must support 5 km detection. Actually, wait. Maybe re-read the requirement: "capable of reliably surviving 0-1000 km/h in 5 secs" — that's the interceptor acceleration environment (rocket boost ~5.7 g average). "and reliably detecting a Shahed-136 at 5 km range from the front and side." So yes: seeker design task. Deliverable: a design spec: frequency, aperture, waveform, power, processing, thermal, structural, signal processing, detection performance with margins, BOM-ish partitioning, risk items. Also perhaps discuss alternative: semi-active vs active. Active seeker is asked ("radar seeker"). Let me now carefully build the design with numbers. I'll pick 77-81 GHz? or 94? Or 35? Let me settle: Key discriminators: 1. 5 km detection vs σ ~0.01-0.05 m² (nose), 0.1-1 m² (side) → need all the aperture gain and energy you can get → higher frequency wins on gain (G ∝ D²/λ²), but power and atmospheric loss worsen with frequency; at these small apertures the gain term dominates and PA power is the limiter. 2. Angular accuracy for terminal homing → higher frequency wins. 3. Search FOV without a gimbal → higher frequency hurts (narrow beam) unless ESA. 4. Electronics/TCs availability → 77 GHz automotive ecosystem; 94 GHz has radar-seeker niche (SADARM/Longbow heritage is 35 and 94; Longbow is 35 GHz Ka). 60mm feasibility favors waveguide-free planar structures at high freq. 5. Radome: 60mm ogive nose at W-band: dielectric constant 3 (E-glass), λ/2 in dielectric = λ/(2√3) ≈ 0.9 mm at 94 GHz, 1.1 mm at 79 GHz — thin but makeable as a tuned sandwich; at Ka, ~2.4 mm half-wave — easy. Gimbal decision: In 60mm with 5.7g/30g environment, a two-axis gimbal with a 54mm aperture... typical seeker gimbal volume: for a 160mm AMRAAM-class seeker the gimbal holds ~100mm dish with ±60-90° look angles. Scaled to 60mm airframe, a gimbal aperture shrinks to maybe 40mm with bearing balls/motors around it → -3 dB gain loss. Alternatively strapdown fixed aperture with micro-EPS (electronic phase steering): at 79 GHz, λ/2 = 1.95 mm pitch; a 54mm circle = 27×27 grid ≈ ~600-700 elements with quadrature... A full ESA with TR modules in 60mm at unit cost for a C-UAS interceptor that must be cheap (vs $20k Shahed!)? Cost is a huge driver: the interceptor should be ≤ $20-50k ideally. A 600-element W-band ESA is not cheap yet (though SiGe automotive-ish ESAs exist — e.g., Oculii/... no). Hmm. Middle path: **passive scanned monopulse**: 4-quadrant array with one analog phase shifter per quadrant (or 4-bit) → limited-scan ±10-20°? With limited field-of-view arrays, scanning of a quarter-panel subarray... A simpler classic: **conical scan with nutating feed** — no phase shifters, monopulse lost but conical-scan angle tracking works (used in AIM-9J-era; MMW seekers like the German Böfors... STRIX used conical scan at 94 GHz). Conical scan: rotate a slightly offset feed/tilted dish; amplitude-modulated return gives angle error. Robust, cheap, single channel receiver! Given SWaP (size, weight, power), a nutating-feed or rotating-dish conical scan at 79-94 GHz with a single-channel receiver is attractive, with the disadvantage of amplitude scintillation sensitivity (Shahed is a fluctuating target — conical scan + Swerling fluctuations = tracking issues; mitigated by conical scan rate choice and AGC; classic issue: AM from target scintillation; use rate gyro-loop damping). Monopulse avoids that; costs 3 receive channels (sum/Δaz/Δel) — with MMIC transceivers, 3 RX channels is cheap today. Strapdown vs gimbal — strapdown monopulse with body-steering: missile must point nose at target within ~±3-4° (W-band beam) via autopilot pointing commands. At acquisition, the cued pointing error must be < beamwidth/2... you need to search. Search options with strapdown: roll the missile and spin a nutation circle (a "spin-scan" strapdown search — precession circle covering a cone of ±8-10°), like early IR seekers' rosette/spiral scans. A strapdown seeker can do a spiral search by nutating the antenna (physically moving the fixed array's BORE via mechanical nutation? no...). Hmm, strapdown with fixed beam: the airframe itself flies a slow roll/wag to scan the beam — inefficient for a rocket with limited control authority... but these are canard-controlled guided rockets; scanning by vehicle attitude is standard for some loitering interceptors (e.g., some use "caging" then scan via attitude). It's doable: 4.5-5° beam at 79 GHz, autopilot does a raster/rosette over ±10° in a few seconds. Search time: ±10° cone with 5° beam → ~20-30 beam positions; if each needs 50-100 ms with enough energy... but energy per position then drops by 20-30 relative to staring → SNR drops 13-15 dB. Search vs σ=0.05 m² at 5 km with 50 ms dwell: SNR from earlier scaling: at 250 ms CPI we had -1.3 dB for 0.01; per 50 ms → -6.3 dB; σ=0.05 → +7 → 0.7 dB. Not enough per-dwell for reliable detection; need multi-scan integration or track-before-detect across scans (2-3 rotations). Feasible but marginal. Better: cue the launcher well. If ground radar cue is accurate to ~1-2 mrad (typical for a Ku/X-band tracker at 10-20 km on a 0.1-1 m² target... actually track accuracy 1-3 mrad), and launcher alignment + autopilot alignment good to a few mrad, then the seeker pointing error at 5 km is meters — well within a ±5-10° acquisition cone. So a strapdown fixed-aperture seeker with a modest **limited-scan** (±15°) passive phase-steered monopulse covering the cue error budget, searching a few positions. This avoids the gimbal. Simplest robust recommendation: - **79 GHz (W-band, automotive-band edge)** fixed strapdown 4-quadrant monopulse array, ε≈0.5, 54 mm aperture, G ≈ 30 dBi, 5° beam. - Phase-steer capability: discrete 4-bit phase shifters per quadrant for pointing steps ±10-15° (or a small set of fixed pointing states), plus conical nutation... no — keep monopulse; do the fine track with monopulse angle while the autopilot nulls body-pointing error (strapdown seeker + canard steering: the "beam-riding" style loop). Limited FOV: strapdown seeker FOV equals steering envelope + beamwidth. Terminal homing at 5° beam strapdown: autopilot must hold target within ±2.5° while maneuvering — Shahed turns at maybe 0.5-2 g; lead angle changes slowly; a 60mm canard airframe with 15-20 g authority handles it. This is exactly how many small MMW terminal seekers work (e.g., some artillery MMV seekers use strapdown). Hmm, but for crossing/side targets, the seeker look angle midcourse could exceed ±15° if launched late; the fire-control solution aims at intercept point, so look angle ~ small if launched with proper lead. For rear-hemisphere engagements of a crossing target at close range, look angles up to 30-60° might arise... The design can include ±20° e-scan (limited-scan array, element spacing λ/2 → full ±45-50° scan capability at element level if fully populated; with per-quadrant steering only, subarray grating lobes limit scan to ±10-15°). Compromise: per-quadrant 4-bit phase shifters + accept ±12° scan; or 4 subarrays × ... Actually make it 8 subarrays for scan to ±25°? Each subarray ~20 elements... this is getting into cost territory. I'll spec: 4 quadrant subarrays, each with a 5-bit phase shifter → scan ±15° usable; full array as ESA upgrade path. Waveform: coherent pulse-Doppler with LFM chips (e.g., 20 μs pulses × 250 MHz LFM? unambiguous range at 20 μs PRI → 3 km; but PD seekers don't need unambiguous range; use dechirp/stretch processing per quadrant). Or FMCW fast-chirp. Let me define cleanly: **Waveform plan:** - Search/acquisition: fast-chirp FMCW, 500 MHz bandwidth, 25 μs chirp (slope 20 MHz/μs), 1500 chirps per 50 ms sub-CPI... beat at 5 km: 2R·slope/c = 2×5000×(500e6/25e-6=2e13)/3e8 = 6.7e5 = 667 kHz + Doppler 164 kHz → IF ≤ 900 kHz. ADC: 2 MS/s per channel ×4 channels (sum, Δaz, Δel, guard). Per-chirp FFT gives 125 m bins; Doppler FFT across chirps gives velocity; 2D processing. Range-Doppler coupling handled via up/down chirp pairs alternating or by using Doppler-known from cueing. - PRF-equivalent: chirp rate 40 kHz (25 μs period) → velocity unambiguous = λ·PRF/2 = 0.0039×40000/2 = 78 m/s. Closing velocities up to ~330 m/s (head-on at max) → aliasing; resolve via multiple chirp rates (staggered slopes) or since we KNOW the interceptor's own speed (INS) and the target class speed (~30-50 m/s), the ambiguity is small: relative velocity estimate from cueing (fire control predicts closure) picks the right ambiguity band. Use stagger: alternate 2-3 chirp durations to fold/unfold. Fine. - Terminal: shorter chirps / higher chirp rate with wider BW (1-2 GHz) → 15-30 cm resolution for fuzing AIM point (detect wing/fuselage extent) and improve track. Also CW mode option for fuzing? A dedicated fuze function can reuse the seeker antenna at short range (Doppler proximity burst + POA). 60mm warhead frag radius maybe 3-5 m; fuze at 2-4 m miss distance... warhead design is out of scope but mention fuzing integration. **Detection budget (to present):** I'll present a table: assume R=5 km, σ_nose=0.01-0.05 m² (state assumption and justification: composite airframe but metallic engine internals, battery, actuator linkages; measured RCS of composite UAVs typically 0.01-0.2 m² nose, 0.1-1+ m² side at MMW; prop blades add spikes), σ_side=0.1-1 m². Let me build the definitive link budget at 79 GHz: Parameters: - Frequency: 79 GHz (λ=3.797 mm) - Aperture: 54 mm circular, ε_ap = 0.5 (monopulse with 4-quadrant, phase shifter losses, radome; conservative) - G = (π·0.054/0.003797)² × 0.5 = (44.68)² × 0.5 = 1996×0.5 = 998 → 30.0 dBi. Beamwidth ≈ 70λ/D = 70×0.003797/0.054 = 4.92° ≈ 5°. - TX: 8 elements... no, TX: 8× 1 W GaAs/GaN MMIC PAs power-combined (spatial combining via antenna tiles) → Pt = 6 W (with combining loss 1.5 dB) → 7.8 dBW. Duty: acquisition burst 25% for up to 1 s (thermal-mass limited), Pavg = 1.5 W. CPI = 250 ms → E = 0.375 J → -4.3 dBW·s. - Hmm, let me consider simplifying: present both peak-power pulse mode and FMCW energy mode via E. SNR (per 250 ms CPI): SNR = E·G²·λ²·σ / ((4π)³ R⁴ k T₀ F L) dB: E=-4.3; G²=60.0; λ²=10log(1.44e-5)= -48.4; σ=-20 (0.01) or -13 (0.05) or -10 (0.1) or 0 (1.0); (4π)³=32.98→33.0; R⁴: 10log(6.25e14)=147.96→148.0; combined 181.0; kT₀F: -203.8+F(4.5)= -199.3 → subtract 199.3... careful: SNR = ... / kTs, so add (203.8-4.5)=+199.3; L: atmospheric 79 GHz 0.25 dB/km ×10 km (two-way 5 km each) = 2.5 dB; radome 1.5; ohmic/filter 1.5; processing/mismatch 2; total L=7.5 → subtract 7.5. SNR(σ=0.01) = -4.3+60.0-48.4-20-181+199.3-7.5 = ? -4.3+60=55.7; -48.4=7.3; -20=-12.7; -181=-193.7; +199.3=5.6; -7.5=-1.9 dB. SNR(σ=0.05) = +5.1 dB. SNR(σ=0.1)=+8.1. SNR(σ=0.5)=+15.1. SNR(σ=1.0)=+18.1 dB. Required for Pd=0.9, Pfa=1e-6 (Swerling I): single-look ~21 dB; Swerling I multi-look with N=8-16 noncoherent: ~13-15 dB per look. Hmm. So single CPI 250 ms gives detection for σ≥0.1-0.5 m² (side: fine), and 0.05 m² is borderline per-CPI but detectable with 2-4 CPI noncoherent integration (+2.5-4 dB) and CFAR over the search. σ=0.01 nose at 5 km: below reliable threshold in one 1-second burst; would need ~4-8 s integration or better RCS. Reality check on nose σ: Shahed-136 nose at MMW — I genuinely believe 0.01 m² is pessimistic for the *mean* but the fluctuation means low-SNR fades occur. Published/inferred numbers: various C-UAS papers give small UAS RCS: quadcopters 0.01-0.1 m²; a 3.5 m delta-wing with metal engine... At X-band, Iranian claims (unreliable)... Let me estimate: engine block frontal area ~0.15×0.2 m? If metallic and exposed (behind fiberglass): a 0.2 m² flat plate-ish metal at 79 GHz → could be 1-10 m² if normal to LOS, but engine is canted/irregular, and shadowed partly by wing spar. Battery ~0.1×0.15 m metal → similar. Worst case they're viewed off-normal → small. Nose-on, the flat delta wing at slight pitch angle: wing area ~1.5-2 m² flat plate at 2° incidence → specular glint σ = 4π A²/λ² × cos²... for A=1.8 m², λ=0.0038: 4π×3.24/1.44e-5 = 2.8e6 m² = 64 dBsm — huge but only within ±0.1° of normal — occasional flash. Median over a few degrees of aspect: much lower. I'll state: nose median 0.03-0.1 m², with deep fades to ~0.01; side median 0.2-1 m². That matches typical estimates for Shahed/characteristics of composite one-way UAVs. Given requirement "reliably detecting... at 5 km from front and side": design target Pd≥0.9 at σ=0.05 m² (front median-low) and huge margin side. If the user's threat model insists on 0.01 m² front, then honest answer: 60mm active seeker is marginal; recommend (a) extend integration during acquisition since the interceptor has >10 s before intercept; (b) rely on propeller modulation (Shahed's 2-blade? It has a wooden/plastic prop ~ 0.6-0.8 m diameter? Actually Shahed-136 prop is ~ 0.6 m, 2-blade, spinning ~ 2500 rpm? Rotational speed 2500 rpm = 41.6 rps → blade-pass 83 Hz; blade tip speed ~80 m/s; adds rich Doppler sidebands at ±k·83 Hz offsets — detection via micro-Doppler extends effective detectability by integrating the blade lines too); (c) dual-mode: add passive RF direction finding on Shahed's comms/GPS? Shahed-136 uses GLONASS/GPS and possibly CRPA; RF emissions are weak/intermittent (Shahed-238 has IR seeker; -136 mostly silent one-way). Not reliable. (d) ground radar provides track, seeker only confirms & homes → seeker Pd requirement relaxed since ground track feed keeps the cue fresh; the seeker only needs to acquire at 3-4 km for terminal homing; 5 km "detection" = acquisition under favorable conditions with high Pd at 4.5-5 km vs 0.05 m². I think the honest design presents this margin analysis transparently. Good engineering answer includes a "feasibility verdict + closing options." **Thermal:** Seeker active from launch or activated at T+3 s? Boost phase 0-5 s at 5.7 g with vibration — electronics on. Power: PA dissipation ~ (input - RF out) ≈ 2-3 W heat at 25% duty? PAE at 79 GHz ~15-25% → with 6 W RF out and 25 W DC at 25% duty → ~4-5 W average heat during bursts. Duration ~20-40 s intercept window worst case, bursts duty-limited. Solution: conduction to aluminum airframe + phase-change material (PCM) puck behind the array; MMIC baseplate ΔT < 25 °C. No fan obviously. **Structural/environment:** - 0-278 m/s in 5 s → 5.7 g avg, assume peak 15 g at ignition spike + 12-20 g_rms broadband random vibration 20-2000 Hz during burn (solid motor acoustics) + pyro shock at motor ignition (per STANAG 4370/MIL-STD-810). This is a *benign* environment compared to gun launch; standard potting (CTE-matched underfill, staking of discretes, no large BGAs or use of corner-bonded BGA with underfill), RF boards on machined Al carriers, bolted (not soldered) RF connectors... Use SMPM blind-mate connectors between boards; waveguide-free architecture at 79 GHz (all planar) — avoids waveguide flange cracking. - Radome: ogive/tangent-ogive 60mm base, tuned ~λ/2-dielectric-wall sandwich of quartz/glass; rain/erosion at 278 m/s modest; aerothermal at 278 m/s (Mach 0.8): stagnation T rise ~ 35-40 K — fine for composites. - Rate/INS: MEMS IMU for strapdown; body-pointing control loop. **Array architecture:** - 4-quadrant monopulse: each quadrant ~ 110-130 elements (54mm circle π×27² = 2290 mm²; per quadrant ~570 mm²; element cell 1.95×1.95 mm = 3.8 mm² → ~150 elements/quadrant → total ~600 elements; edge taper leaves ~550-600 radiating). - Each quadrant: series-fed slat? Better: each quadrant is a corporate-fed patch subarray on a multilayer Rogers/PTFE+... At 79 GHz: liquid-crystal polymer or Rogers 3003/TMM; MMIC LNA per quadrant + phase shifter (5-bit analog) + PA per quadrant (2×1 W combined per quadrant? If Pt total 6 W → 1.5 W per quadrant → 2×0.75 W MMICs per quadrant). TX/RX via circulators? At 79 GHz no cheap circulator this small... use T/R modules with switches (SPDT pHEMT switches at 79 GHz exist) or bistatic subarrays (separate TX and RX apertures — halves aperture each way! No). Use reflective... Standard: T/R switch per quadrant, FMCW needs full-duplex → leakage problem: FMCW TX-RX same time; with a single antenna + switch you can't do FMCW bistatic-less... Options: (a) pulse-Doppler (T/R switching, monostatic pulsed with chirp-on-pulse + stretch: TX 5-10 μs chirped pulse, RX after; PRI 40 μs; duty 12.5-25%) — avoids duplexer leakage issues and enables full monopulse; (b) FMCW with on-chip cancellation loop (automotive style) — works with antenna-coupled leakage cancellation to ~ -40... For a seeker, pulsed stretch processing is clean. I'll go pulsed-LFM (stretch): TX chirped pulse 10 μs, 500 MHz; PRI 50 μs (PRF 20 kHz); RX dechirp → IF = 2R·slope/c + fD: slope=500e6/10e-6=5e13 Hz/s; at 5 km: 2×5000×5e13/3e8 = 1.67 MHz; + fD 2×330/0.0038=174 kHz → IF BW ~2 MHz → ADC 5 MS/s ×4 channels. Duty 10 μs/50 μs = 20% → with 6 W peak → 1.2 W avg. CPI: 50 ms → 1000 pulses → Doppler res 20 Hz; velocity ambiguity: PRF 20 kHz → fD unamb = 20 kHz → v = fD λ/2 = 20k×0.0038/2 = 38 m/s — aliased closure; ambiguity resolved via stretch range tracking + fire-control cue + PRI stagger. Range ambiguity: PRI 50 μs → 7.5 km unambiguous — good (target at ≤5-7 km): actually 7.5 km unambiguous at PRF 20 kHz — great, no range ambiguity within 5 km! And use 3-PRF stagger (50/56.8/66.7 μs) to unfold Doppler to ±115 m/s; combined with INS-derived ownship speed → residual target speed ±50 m/s → resolvable. Good. - Angle: 4-quadrant amplitude monopulse at 5° beam → angular sensitivity: Δ/Σ slope ~1.6/beamwidth... σ_θ ≈ θ3dB/(k√(2·SNR)); with SNR 15-20 dB and k≈1.5-1.6 → σ_θ ≈ 5°/(1.6×~7) ≈ 0.45° per update at ~20-40 Hz update; at 500 m → 4 m. Good enough with PN for a 4-6 m-lethality warhead; better at closer range. - Guard channel: use the 4 quadrants in sum/Δaz/Δel + optional guard for mainlobe-vs-sidelobe decisions (sidelobe clutter/false alarms) — can form guard from Δ channels or dedicated edge ring of elements; or low-gain guard patch antenna on the radome inner surface? Keep simple: CFAR in angle via monopulse off-beam-center test + sidelobe blanking using a small auxiliary element. **Search strategy:** - Cue from launcher/fire-control (any C-UAS radar; Shahed track even at 0.01-0.1 m² is easy for ground X-band). - Pointing: roll-align at launch, autopilot commands attitude to predicted LOS; seeker scan: electronic discrete steps over a 7×7 grid of 2.5° steps covering ±7.5° (at 5 km covers ±650 m) in ~1 s (20 positions × 50 ms) or spiral; with two rotations of the grid and noncoherent integration + TBD for low-σ targets; time budget allows (intercept geometry gives 10-15 s between 5 km and impact). - Upon detection: track, close, proportional navigation with monopulse updates at 50-100 Hz (CPI 10-20 ms at high SNR near range <3 km). **Front vs side aspect detection nuances:** - Side/crossing: fuselage broadside + wing edge flashes → strong σ median 0.2-1 m²; plus slower angular glint; monopulse averaging + PN handles glint; expect detection ≤5-6 km with margin. - Head-on: σ lower; rely on: longer CPI, micro-Doppler (prop blades, engine vibration) classification... Actually for detection vs a Shahed head-on, the prop is at rear but blades visible through fuselage? Fiberglass fuselage: MMW sees through partly → blade flashes attenuated but present. Also detection aided by geometry: head-on closure gives strong Doppler separation from ground clutter: ground clutter Doppler spread: mainlobe clutter at 5 km with 5° beam: clutter velocity spread due to beamwidth ±(θ/2)·V/... for LOS 0° elevation... For a low-altitude engagement, mainlobe clutter center Doppler = 2V cos(depression)/λ; spread from beamwidth: Δv ≈ V·θ3dB → 278×0.087=24 m/s spread. Target closure 320 vs clutter near 280±12 → separation ~40 m/s ~ 21 kHz — outside mainlobe clutter spread with 20 Hz Doppler resolution: target in clear region. Altitude/side lobe clutter below noise after processing. Good — pulse-Doppler with notch at ownspeed Doppler line handles clutter; also altitude return may be strong: at 5 km slant with low altitude, grazing angle small → low clutter RCS but near-in sidelobes see high-graze terrain... For C-UAS engagements at 100-1000 m target altitude over flat ground, seeker look-up or level; manageable with notch + sidelobe suppression (apply Taylor taper: -25 dB sidelobes on RX; cost 0.9 dB gain — account in ε). **Fuzing:** short-range wideband mode: switch to 2 GHz BW (10 μs chirps, higher slope → IF handling: at 500 m: beat = 2×500×2e14/3e8 = 667 kHz; fine) → 7.5 cm resolution → range-profile-based fuze decision + Doppler burst; initiate at 3-6 m miss/POA. 60mm frag warhead lethal radius vs Shahed structure: needs only cutting control links/wing spar or engine — frag pellet design elsewhere. **If they want harder performance (σ=0.01 m² at 5 km with Pd 0.9):** closing options: 1. Increase aperture: can't beyond 60mm... unless seeker trades airframe length for frequency: no. 2. Increase energy: 20 W peak module (GaN-on-SiC 79 GHz: state of art ~5-10 W per MMIC? At 77-81 GHz, GaN PA MMICs ~ 2-5 W; combining 8 → 15-20 W possible but PAE/thermal brutal in 60mm during 20-40 s engagements: 60 W heat at 25% duty → PCM buffer + burst duty management. Feasible with duty scheduling (acquisition bursts 25% for 2 s, then track at 5-10%). 3. Longer acquisition integration (TBD over 2-4 s) → +6-10 dB effective. 4. Move to Ka 35 GHz where 30-50 W peak GaN is easy and loss lower; gain -7 dB, so net worse unless aperture loss... no, net worse ~ -7+? Pt +7-9 dB → net +0-2 dB; but wider beam hurts search dwell count; roughly a wash again. Frequencies below 35 GHz lose gain fast. 5. Passive augmentation: none on target. 6. Semi-active mode: ground radar (X/Ku, 1-10 kW class, big antenna) does detection; seeker only receives → the seeker requirement changes from "detect" to "home" — massive reliability margin; many C-UAS programs do exactly this (SALH/beam-riding or SARH). Mention as robust alternative for the hard nose case: semi-active at Ku with the same 60mm aperture gets huge link margins (ground radar EIRP 60-70 dB... at Ku, seeker receive-only: EIRP 60 dBW, 5 km: at seeker aperture G=22 dBi... PR = EIRP + G_rx + λ²σ/(4π)³R⁴... SARH against σ=0.01: PR = 60 (dBW) ... let me not fully compute; ground illuminators make 5 km SA homing trivial.) But user asked for radar seeker detection; I'll mention SA as a de-risk fallback. **Also required: "reliably surviving 0-1000 km/h in 5 sec"** → environment spec: - Average 5.67 g, assume motor thrust curve peak ~12-18 g at max q? For 60mm tube-launched: actually if "0 to 1000 km/h in 5 s" is the requirement, the rocket motor sized accordingly: a = 5.67 g avg + drag deceleration during 5 s (drag ~15-18 m/s² at high speed) → motor provides ~ 7.5-8 g avg, peak maybe 12-15 g at burnout-lite... The seeker sees: axial static load along -X (setback ~ up to 15-20 g design), vibration 196-2000 Hz random 6-20 g_rms (motor + aero), aeroacoustic noise, thermal: stagnation ~ Mach 0.82 → recovery temp ≈ T(1+0.2 M²)... Mach 0.82: T_r ≈ 288×(1+0.2×0.67) = 288×1.134 = 327 K = 54 °C stagnation. Radome temps ~40-55 °C — fine. - Design margins: 1.5× on all loads; potting of PCBs, stiffener frames, no cantilever >10 mm components, crystals rated 30 g, MEMS IMU 20,000 g-rated parts cheap anyway, use of solder-joint strain relief, RF board attachment via screws+epoxy fillets, connectors blind-mate SMPM with float, conformal coating, hermetic or vented (vented with Gore vent for pressure equalization? Tube launch could be hot-launch (gas blow-by) → basebleed/gas path isolation: seeker sits forward of motor; only aft plate sees launch gas — design aft bulkhead for hot-gas impingement w/ ablative + labyrinth seals. **System partition (lengthwise in 60mm):** - Radome (tangent ogive, ~1.2-1.5 λ semi... nose length ~2-3 calibers: 120-150 mm ogive? For Mach 0.8 fine with ~3 caliber 180mm? Keep ~2.5 caliber 150 mm ogive). - Antenna/RF tile: 8-10 mm stack (radome gap, aperture plate with patches, MMIC layer, PSU/DC, phase-shifters, RF manifold) length ~15-20 mm. - Interconnect block: 10 mm. - Digital + IMU board stack: 30-40 mm (FPGA/SoC, 4× ADC 12-bit 5-10 MS/s ×2-4 channels, DAC for waveform, PMIC, MEMS IMU). - Power conditioning + thermal buffer (PCM): 15 mm. - Interface/fuze electronics + safing (S&A inline ESLD): 15-20 mm. Total seeker length ~ 100-130 mm. Mass ~ 0.35-0.5 kg. Power: ~ 15-25 W peak electrical, ~4-6 W average; battery: thermal battery or Li-SOCl2 activated at launch (survives launch; spin... no spin). **Signal processing chain:** - Per CPI: dechirp → LPF → ADC ×4 (Σ, Δaz, Δel, Δ2?) → stretch FFT (range) → coherent Doppler FFT per range cell → CFAR (2D CA/OS-CFAR with guard) → detection & clustering → monopulse ratio for angle → track file; TBD across CPIs for low-σ; classification: micro-Doppler features (prop blade rate ~70-90 Hz sidebands, engine 50-100 Hz harmonics, wing flash statistics) to confirm Shahed-class vs bird/clutter/false target; track gating with INS-nav integrated prediction. - Update rate to autopilot: 100 Hz angle/range/Doppler vector. - Guidance: PN with bias for lead; fuzing cue. **Performance summary table** — detection: Let me finalize honest numbers per aspect at 5 km, 250 ms CPI, Pfa 1e-6, N=4 noncoherent (1 s): Threshold ~13.5 dB per look for Pd 0.9 aggregate... Using Swerling I with 4-look: required per-look SNR ≈ 13 dB. - Front, σ median 0.05 m²: SNR/250ms ≈ 5.1 dB → 4-look NC → effectively ~ +3.7 dB? Noncoherent integration gain for 4 looks ≈ 5-6 dB less than coherent 10log4=6: net gain ~4.5-5 dB → ~10 dB → Pd ≈ 0.7-0.8 in 1 s; with TBD over 2-3 s → Pd ≥0.9. Verdict: detection at 5 km front is achievable but with thin margins against worst-case fades; at 4.5-4 km Pd>0.9 solid; head-on engagements happen at closing ~320 m/s → 5 km → 3 s to 4 km — so detection at "5 km" ~ within 1-2 s of that range. Acceptable operationally: fire control declares "seeker range 4-5 km vs 0.05 m² class". - If σ=0.01 (deep fade): needs TBD over ~4-8 s → detection at 3.5-4 km effective. Mitigations: increase Pt to 12-16 W (design growth), dual-band option. - Side, σ 0.2-1 m²: SNR 14-21 dB per CPI → single-CPI Pd ~0.5-0.9, 1 s → >0.95. Solid at 5 km, often 6-8 km. Hmm wait, I should double-check my earlier SNR calc at 250 ms CPI, σ=0.05: I got +5.1 dB. Let me recompute carefully, because this is the crux. E = Pavg × t: Pavg = 1.2 W (6 W × 20% duty) × 0.25 s = 0.3 J → -5.23 dBW·s. G = 30 dBi → G² = 60 dB. λ = 3.797 mm → λ² = 1.442e-5 → -48.41 dB(m²). σ = 0.05 → -13.01 dBsm. (4π)³ = 1.984e3 → 32.98 dB. R⁴ = (5e3)⁴ = 6.25e14 → 147.96 dB. kT0 = 1.38e-23 × 290 = 4.0e-21 W/Hz → -203.98 dBW/Hz. F = 4.5 dB → Ts factor: SNR = E G²λ²σ / ((4π)³ R⁴ kT0 F L). In dB: SNR = -5.23 + 60 - 48.41 - 13.01 - 32.98 - 147.96 + 203.98 - 4.5 - 7.5 = ? -5.23+60 = 54.77 54.77-48.41 = 6.36 6.36-13.01 = -6.65 -6.65-32.98 = -39.63 -39.63-147.96 = -187.59 -187.59+203.98 = 16.39 16.39-4.5 = 11.89 11.89-7.5 = 4.39 dB. So 4.4 dB per 250 ms CPI for σ=0.05 m². (My earlier quick math said 5.1 — close.) For σ=0.1: 7.4. For 0.2: 10.4. For 1.0: 17.4. For 0.01: -5.6. With Pavg raised to 2 W (10 W peak × 20%) → +2.2 dB: σ=0.05 → 6.6 dB; plus 4-look NC (+~4.5) → ~11 dB → Pd ~0.85 at 1 s... borderline-ok. With 8-look over 2 s: +~6.5-7 → 13+ → Pd ≥0.9. Fine, and detection then at 5 km front is realistic with ~2 s of staring/TBD given the interceptor's 10-15 s of flyout after 5 km... wait no — detection should ideally occur BEFORE the interceptor reaches... The seeker range requirement is detection at 5 km target range, which occurs during midcourse; after detection there are ~10+ s to guide in. Spending 1-2 s acquiring is fine. Good. Also, I should double check atmospheric loss at 79 GHz: ITU: at 79 GHz, oxygen absorption low (~0.02 dB/km), water vapor continuum at 79 GHz with 7.5 g/m³ humidity: ~0.15-0.25 dB/km; total ~0.2-0.3 dB/km → 2-3 dB round-trip 5 km path each way (10 km total two-way path) → 2-3 dB. I used 2.5 — good. In rain (5 mm/h), 79 GHz adds ~ 3-5 dB/km → kills 5 km detection (rain is the Achille's heel of MMW; note operational trade: in rain, engage at shorter ranges / use ground radar cue for launch point, seeker still acquires 2-3 km). Mention rain robustness honestly: at 35 GHz rain loss ~0.6-0.8 dB/km at 5 mm/h — Ka survives rain better. This is THE key argument for Ka in some theaters. Present both, recommend 79 GHz baseline for clear/marginal weather ops with Ka variant if all-weather is mandatory... Hmm, but the user asked for a design; I should pick one and justify, listing variants. Actually — let me reconsider recommending Ka-band 35 GHz primary with higher power. Compare final margins: Ka option: Pt 30 W peak GaN (e.g., 2×15 W GaN MMICs — at 35 GHz, 15 W GaN exists; combined 25-30 W), duty 20% → Pavg 6 W; CPI 250 ms → E = 1.5 J → +1.76 dBW·s. G: 54mm, ε 0.55: G=(π×0.054/0.00857)²×0.55=(19.79)²×0.55=391.7×0.55=215.4→23.3 dBi; G²=46.6. λ²=(8.57mm)²=7.34e-5→ -41.34. σ=-13 (0.05). (4π)³R⁴ = -181.0 (32.98+147.96 = 180.94). kT0F: +203.98 - F(3 dB) = +201.0... wait sign: subtract kT0F: -(-203.98+3)=+200.98. L: atm 35 GHz ~0.12 dB/km → 1.2; radome 1.0 (thicker wall easier); ohmic 1.5; processing 2.0 → 5.7 → -5.7. SNR = 1.76 + 46.6 - 41.34 - 13.01 - 180.94 + 203.98 - 3 - 5.7 = ? 1.76+46.6 = 48.36 48.36-41.34 = 7.02 7.02-13.01 = -5.99 -5.99-180.94 = -186.93 -186.93+203.98 = 17.05 17.05-3 = 14.05 14.05-5.7 = 8.35 dB. Interesting! Ka with 6 W average gets 8.35 dB vs W-band 6.6 dB (with 2 W avg) for σ=0.05. Per-watt-of-DC, Ka wins because power scales faster (GaN) than the G²λ² product scales... note G²λ² ∝ (D²/λ²)²λ² = D⁴/λ² — so per given transmit power, W-band is +6 dB (λ²: (8.57/3.8)² = 5.1... wait: G²λ² ratio W/Ka = (D⁴/λ²)_W / (D⁴/λ²)_Ka = (λ_Ka/λ_W)² = (8.57/3.797)² = 5.09 → +7.07 dB for W-band per watt. But achievable average power: W 2 W vs Ka 6 W → +4.8 dB for Ka; net W-band still +2.3 dB per current PA tech, but thermal in 60mm: 6 W RF at Ka (PAE 25% → 24 W DC at 20% duty → ~4.8 W heat avg) vs 2 W RF at W (PAE 15% → 13 W DC at 20% → 2.7 W heat). Thermal margin better at W... but the W-band power is the harder part. Roughly comparable; the decisive discriminators become: rain (Ka +), angular resolution & search geometry (W +), MMIC industrial base & SWaP maturity (79 GHz automotive SiGe for low cost, but power needs GaAs/GaN niche; Ka GaN mature) and beam steering (Ka 11° beam means strapdown search is 5-10 beams total — easy; terminal accuracy adequate; W 5° beam → precise but search needs more steps). Given Shahed-136 is a slow, non-maneuvering target, 11° beam monopulse is plenty: angular error at 3 km with SNR 20 dB: σθ = θB/(1.6√(2·SNR))... σθ ≈ 11°/(1.6×√200) = 11/22.6 = 0.49° → 25 m at 3 km → PN converges with 20-40 Hz updates and 15+ g airframe; final miss ~1-3 m achievable by guidance filtering. OK. Hmm, so actually the Ka option is arguably the better all-around engineering choice: much better in rain (relevant for "reliably"), power headroom via GaN, wider beam simplifies search & relaxed strapdown pointing, and detection margins at 5 km are comparable or better. Terminal accuracy is adequate for a 60mm frag warhead vs a 3.5-4 m target. Cost: Ka GaN PAs a few hundred $; 79 GHz automotive SiGe transceives are cheap but underpowered for 5 km → need custom GaAs/GaN PA + combiner anyway. Fuzing resolution: Ka can still do 1 GHz BW → 15 cm — fine for fuze AIM. But hold on, one more check on Ka beamwidth and ground clutter in low-altitude engagements: 11° beam at 5 km, if the target is at 100 m altitude and interceptor at similar, depression angles to terrain within beam → mainlobe clutter enters: at 5 km range, beam spans ±480 m vertically → definitely hits ground → strong ground clutter (σ⁰ ~ -20 to -30 dB for grazing angles 1-3°). Target at Doppler +40-320... own-speed clutter line at v·cos... The mainlobe clutter velocity spread for a level, forward-looking strapdown seeker: clutter Doppler = 2V/λ·cos(look_az)·cos(elev): spread over the 11° beam ±5.5° around boresight: ΔfD = 2×278/0.00857 × (1-cos5.5°) ≈ 65.6 kHz × 0.0046 = 0.3 kHz... spread from beamwidth is small: v-spread ≈ V×sin(5.5°)×5.5°(in rad)? Let me just: angular spread Δθ=11°=0.192 rad; velocity spread ≈ V × Δθ (for broadside-ish geometry) ≈ 278 × 0.19 = 53 m/s?? No wait: clutter radial velocity gradient across beam ≈ V · sin(elevation-angle between velocity and LOS) × beamwidth. For LOS near velocity vector: small spread; for LOS perpendicular (crossing): clutter spread ≈ V×Δθ = 53 m/s. Shahed crossing: target closure = V_interceptor component + V_shahed component... For a crossing Shahed at 5 km, interceptor points at intercept lead point: LOS maybe 20-40° off velocity → target Doppler vs mainlobe clutter Doppler: clutter line center at 2Vcos(θLOS)/λ; target at 2V_closing_target/λ: target closure for crossing ~ (interceptor speed×cos geometry) ± 40. The target lies near the clutter line in crossing geometries! Classic look-down problem. Mitigation: narrow Doppler filters (20 Hz), monopulse CFAR in range-Doppler map, elevation beam pointing to keep clutter below/above... Shahed-136 engagements typically at 500-2000+ m altitude above ground; ground clutter grazing angles from interceptor altitude: at 1000 m alt, 5 km slant: grazing ~11°: σ⁰ at 11° at 35 GHz ~ -25 dB (rough terrain). Clutter RCS in range-Doppler cell: cell size: range 125 m × azimuth beamwidth at 5 km (11° → 960 m)... huge cell: 1.2e5 m² × -25 dB = 380 m² clutter per range cell before Doppler discrimination! In one Doppler filter (20 Hz → 0.086 m/s? no: 20 Hz → Δv = λ/2×20 = 0.086 m/s... Doppler filter width 20 Hz = velocity 0.086 m/s): clutter spread across beam ~0.3-53 m/s per geometry → many filters contain clutter. Clutter per filter ≈ 380 × (0.086/53) ≈ 0.6 m² for the broadside-LOS worst case; vs target 0.05-1 m²... For crossing geometry, target σ 0.2-1 (side) vs clutter-in-cell 0.1-1 m² → detection marginal under mainlobe clutter in worst geometries; mitigations: raise engagement altitude (steeper look-down pushes clutter to different Doppler?), better: raise Doppler resolution (longer CPI → 10 Hz), use the target's extra ±40 m/s Shahed motion, and monopulse/track-before-detect; or climb above target and dive (look-down puts clutter at high positive Doppler separated by altitude geometry...). This is the real challenge in low-altitude crossing engagements, less so head-on. Worth one honest paragraph with mitigations: (1) filter clutter in 2D range-Doppler with high resolution; (2) trajectory shaping (engage from below/above with appropriate elevation); (3) at 35 GHz σ⁰ lower at low grazing... Actually at higher freq σ⁰ at low grazing is lower (σ⁰ ∝ f... at low grazing, higher frequency → lower clutter because shadowing). Slight W advantage. Hmm, W-band's narrower beam (5°) reduces clutter cell azimuth extent by 2.2× → clutter per filter ~0.3 m². And higher Doppler resolution... Both work with adequate processing. Fine — I'll note clutter mitigation: CPI-adaptive Doppler processing + engagement geometry control + CFAR + micro-Doppler confirm. OK here's my dilemma: answer length/format. This is a deep engineering question deserving a thorough design document. The user is likely an engineer or technically sophisticated person (could be a defense program person, student, or hobbyist). I'll produce a structured design doc: requirements parse → key design drivers → architecture choice with trade table → detailed design (antenna/RF, waveform, processing, thermal, structure/environment) → performance budgets (link budget table, detection summary front/side) → risks & closing options → BOM-ish/physical layout. Use headers, tables where they add value (link budget, trade study), bullets. Keep prose tight. This will be long — justified by the request's depth. Maybe ~1200-1800 words? It's a complex system design; I'll aim for comprehensive but not bloated. Use tables for: frequency trade, link budget, environment spec, maybe detection summary. Avoid excessive nesting. Let me also sanity-check a few physical numbers: Beam steering: at Ka 35 GHz, λ/2 = 4.28 mm pitch; 54mm diameter → grid 12×12 = 144 positions, ~120 elements in circle. 4 quadrants of ~30 elements each. Per-quadrant 5-bit phase shifter + PA (15 W GaN, or 2×8 W) + LNA + T/R switch + cal. Total ~120 element array fed by corporate... wait, if only 4 phase shifters (one per quadrant), each quadrant needs corporate feed combining 30 elements' TX power: passive corporate combining losses at 35 GHz: ~0.2 dB/stage × 5 stages = 1 dB — fine, plus LNA immediately at each quadrant output (net RX noise figure dominated by LNA ~2-2.5 dB at 35 GHz → F total 3-3.5 dB). Yes — passive array per quadrant + per-quadrant MMICs is the classic "4-channel monopulse with subarray steering" (e.g., like ADATS-era? like many MMW seekers). Scan limit with quadrant-level steering: subarray aperture = quadrant extent ~27mm → grating lobe free scan ≈ asin(λ/(2·d_sub))? For subarray spacing between phase centers ~27 mm, grating lobes appear when scanning beyond asin(λ/(2·d))... λ/(2d) = 8.57/54 ≈ 0.16 → ±9°... hmm that limits e-scan to ±8-10° with quadrant-only steering. To get ±15-20°, use 8 subarrays (phase centers 13.5 mm → ±19°). Or accept ±10° e-scan + rely on launch cue precision (a few mrad from fire control → at 5 km, 20-50 m cross-track → 0.3° pointing — well within ±10°). OK: 8 subarrays for ±20°, each ~15 elements with its own 5-bit PS and PA/LNA; monopulse formed by analog combining of 8 subarrays into Σ/Δaz/Δel via a compact stripline butler-ish monopulse network? At 35 GHz a printed monopulse comparator is easy (standard). 8 subarrays → comparator: form sums: Σ = sum of all; Δaz = (left half)-(right half); Δel = (top)-(bottom). Then per-channel dechirp + ADC. 4 ADC channels. TX: inject through Σ port (or dedicated TX corporate feed) with per-subarray phase shifters steering the TX beam too. Great, clean architecture: "8-subarray, quadrant-monopulse, limited-scan passive ESA" — no per-element T/R (cost ~$1-3k seeker RF BOM vs $30k+ full ESA). Pt: PA per subarray 4 W GaN? 8 subarrays × 3-4 W = 24-32 W peak → with combining/feed losses → 18-25 W at aperture. Duty 20% → Pavg ~4-5 W → CPI 250 ms E = 1.1 J. Recompute SNR σ=0.05: previous calc used E=1.5 J (Pavg 6 W). With E=1.1 J (0.42 dBW·s... 10log(1.1)=0.41) → SNR = 8.35 - 1.35 = 7.0 dB per 250 ms. 4-look NC over 1 s → ~11.5-12 dB → Pd ~0.8+ for Swerling I at 1 s... plus TBD/second-pass → 0.9. And σ=0.1 median nose → 10 dB/look → Pd 0.9 in ~0.5-1 s. Side (σ 0.3-1): 15-20 dB/look → single-CPI detection. Hmm OK wait, I also want to double check the "front" σ assumption once more. Shahed-136 nose-on at 35 GHz: I'll bracket 0.02-0.1 m² median with fades below. C-UAS literature commonly quotes small fixed-wing composite UAVs ~0.01-0.1 m² nose. The design should close for 0.05 with stated margin policy. If the user's intelligence says 0.01, offer the W-band-high-power variant or semi-active. Good. **Gimbal vs strapdown final call:** Strapdown fixed aperture + ±20° e-scan via 8 subarrays + monopulse. Search: at 35 GHz 11° beam: search ±20° az/elev... need beamwidth steps ~8°: 5×5=25 positions... at 50 ms = 1.25 s per frame — fine given cue. Actually with a good cue (±2°), search is 1-3 positions, 150-250 ms. Terminal: 20-40 Hz monopulse updates, autopilot points body within ±3° — with ±20° e-scan authority the strapdown loop has plenty of margin for lead-angle changes (Shahed ~1-2 g lateral max → lead angle drift ~ few deg over 10 s — manageable). At closer range (<500 m) look rates rise: crossing target angular rate at 200 m with 240 m/s relative lateral: ω = 240/200 = 1.2 rad/s = 69°/s — beam steering covers ±20° instantly (electronic) → keep target in beam until endgame; PN with 15-25 g airframe → miss < 2-3 m. At the last 100 m of a crossing engagement, look angle from body axis could exceed 20° if the autopilot lags... 100 m crossing at 240 m/s: time 0.4 s; attitude loop (airframe pitch dynamics ~ 3-5 Hz) can track 60-90°/s commands? Standard strapdown endgame: at these ranges the geometry is usually near-collision-course so LOS rates stay low until very close (PN drives LOS rate → 0), so beam stays near body axis. Fine — note edge case: very close crossing without collision-course closure → homing loop gains + late-arrival intercept geometry handled by guidance (aim to pass within fuze radius). OK. **Fuze:** integrated Doppler/RF proximity from the same array: at <150 m, switch to 1-2 GHz BW wideband mode; detect wing/fuselage extent crossing range gates; command burst at 2-5 m miss with angle-gated initiation (POA = monopulse direction of target); also backup contact fuze. **INS:** MEMS (e.g., tactical-grade, 5-10°/hr gyro) adequate for 20-30 s flights; GNSS optional for midcourse (jamming environment in Ukraine is heavy → don't rely on GNSS; use command midcourse via datalink or pure inertial + launch-point-aligned trajectory: pure-inertial with cuing at handover works if launch cue good: position error after 20 s with 1 mg accel bias... 0.01 m/s² × 20²/2 = 2 m — plus initial alignment ~0.1-0.3 mrad attitude → 0.3 mrad × 5000 m = 1.5 m — trivial! Midcourse inertial is fine given radar cue quality 10-30 m at 5 km... Cued intercept point accuracy: ground radar track error at 15 km ~ 30-100 m → interceptor inertial nav to that point with 100 m errors → seeker must cover 100-500 m at 5 km = ±0.6-3° pointing — within ±20° scan, search 2-5 positions. **Cost:** Shahed costs ~$20-50k; interceptor seeker must keep total round ≤ $30-80k to be attritable. This architecture (8 subarray passive ESA + GaN PA + SiGe RX + FPGA) — BOM maybe $3-6k seeker electronics at volume; achievable. **Environment/survival specifics:** - Boost: 0→278 m/s over 5 s ⇒ avg 5.7 g; assume peak axial 15-20 g (motor tail-off/thrust spike + aero at max q), design qualification: axial 30 g, lateral 15 g, 1.5× margins; shock (motor ignition 500-1500 g at structure for <0.5 ms? typical motor ignition shock ~1000 g 0.1-0.2 ms at case; at seeker head dampened <100 g); random vib during burn 6-15 g_rms 20-2000 Hz (case-bonded solid motor acoustic + aero noise); aerothermal: Mach 0.82 sea level → stagnation ~50-55 °C, radome skin temps OK for glass/quartz composites; rain erosion at 278 m/s modest (no rain-erosion coating strictly needed at subsonic; add thin polyurethane coat anyway). - Parts: all SMD, X8R/X7R MLCCs (avoid large-case MLCCs' piezo issues at vibration: ≤1210 size), underfilled BGA/FPGA or use QFP/SiP; staked tall parts; PCBs bonded to machined Al carriers with stiff ribs; potting only where CTE-compatible (avoid potting RF). - Rate: tube launch could impart roll if rifled — assume smoothbore/slip-ring... assume fin-stabilized canard control, roll controlled or roll-fixed. Fine. - EMI: motor plume ionization affects aft RF little at W... Ka fine. - Storage: -40 to +60 °C ops, etc. Standard. **Timeline/modes:** - T+0 launch (seeker powered by thermal battery activation), IMU alignment transfer from launcher (transfer alignment), boost 5 s; - midcourse: inertial flight to cued intercept point; optionally datalink track updates; - acquisition window: from T+8-20 s, depending on geometry; seeker runs TBD search staring along commanded LOS; - track + homing: PN, updates 25-50 Hz; - endgame <200 m: wideband fuze mode, burst on POA; kinetic back-up. **Detection summary table** (Pd at 5 km, Pfa 1e-6, integration 1 s): - Nose, σ median 0.05: Pd ≈ 0.85-0.95 (with TBD); low-fade 0.01: Pd ~0.5 at 5 km, ≥0.9 by 4 km. - Side, σ 0.3-1: Pd > 0.95 single-second; typically acquires 6-8 km. - In rain 4-8 mm/h: Ka loss 0.5-1.2 dB/km → effective detection 3-4 km (front), 5+ km side; W-band variant far worse in rain — key reason for Ka. Wait — I flip-flopped: my final architecture now is Ka-band 35 GHz, 8-subarray limited-scan strapdown monopulse passive ESA. Let me re-verify the key trade numbers I'll present for frequency: Per-watt sensitivity: aperture fixed 54mm; SNR per J of energy ∝ G²λ² = (πD/λ)²ε²λ² = π²D²ε² — WAIT. G²λ² ∝ ((πD/λ)²)²λ² = π⁴D⁴/λ² ×... G = ε(πD/λ)²; G²λ² = ε²(πD)⁴λ²/λ⁴ = ε²π⁴D⁴/λ². So SNR ∝ Pt/λ² — LOWER frequency is BETTER per watt at fixed aperture?! Let me recheck: G²λ²σ/((4π)³R⁴) with G ∝ 1/λ² → G² ∝ 1/λ⁴, times λ² → ∝ 1/λ². Yes! For a fixed-aperture, fixed-power radar, SNR improves as λ² — lower frequency gives better SNR on small targets (that's why surveillance radars are low-freq). BUT: power available worsens at higher f, not lower. Hmm, then why did my W-band calc only beat Ka modestly? Because at Ka I used 3× the power and lower F and lower L... Let me recompute properly with equal power to see: W 79 GHz E same as Ka E: difference = +7.07 dB for... G²λ² W vs Ka: (8.57/3.797)² = 5.09 → +7.07 dB in favor of W?? No wait, G²λ² ∝ 1/λ²: smaller λ (W) → larger 1/λ² → W-band has MORE G²λ²?? Let me redo: G ∝ D²/λ². G² ∝ D⁴/λ⁴. G²λ² ∝ D⁴/λ². So smaller λ (higher freq) → larger SNR per watt?! Hold on, I think I made an error. Standard result: for fixed aperture, received power Pr ∝ Pt G² λ² σ/R⁴... but wait, the radar equation is what it is: Pr = Pt G² λ² σ/((4π)³ R⁴). With fixed aperture: G = 4πA/λ²... G²λ² = (4πA/λ²)² λ² = 16π²A²/λ². So Pr ∝ Pt A²/(λ² R⁴)?? That says higher frequency → more received power from a point target at fixed aperture and power. Hmm, but is that right? Check dimension: A²/λ² dimensionless — fine. But wait, that contradicts the known result that "for a given aperture, lower frequencies see more energy" — no, actually the known result: for fixed GAIN, Pr ∝ λ² (lower freq better); for fixed APERTURE, Pr ∝ 1/λ² (higher freq better) because gain rises as λ⁻². Yes — this is the classic aperture-limited vs gain-limited trade. Point targets: aperture-limited → higher frequency wins per watt. Distributed targets (σ ∝ A_t of extended clutter): balanced. And for detection of small RCS at max range with a size-constrained antenna, you want the highest frequency you can afford power/losses. That's why MMW seekers are 35/94 GHz, not X-band. And 3mm (94-79 GHz) gives more per-watt SNR than 8.6mm... but the practical issues: PA power falls ~ frequency, atmospheric loss rises, radome harder, thermal worse. My earlier arithmetic: W (2 W avg) gave SNR 6.6 dB vs Ka (6 W avg) 8.35 dB for σ=0.05. So per equal average power W wins by 7.07 dB offset by Ka's +4.8 dB power → net W +2.3 dB. OK consistent. So on pure SNR, W-band with 2 W avg ≈ Ka with 6 W avg. The differentiators then: - Rain: Ka strongly better. 79 GHz at 5+ mm/h: attenuation ~ 4-7 dB/km(!) vs Ka 0.5-1 dB/km. In European winter rain rates this matters less; in heavy rain W-band 5 km is dead. "Reliably detecting" → all-weather argues Ka. - Thermal in 60mm: W 2 W RF at PAE 12-15% → 13-17 W DC during burst → duty-limited to short bursts with PCM; Ka 24 W DC at duty 20% → 4.8 W avg heat — manageable continuous. Ka better sustained. - Angular: W 5° beam better endgame; Ka 11° adequate. - Search/strapdown: Ka easier (fewer beams, wider FOV per step). - Component maturity: Ka GaN mature (automotive 77 GHz helps SiGe RX for W...). - Radome: Ka trivial (2.5 mm half-wave glass); W needs 0.7-1 mm tuned wall (makeable: thin monolithic quartz). - Endgame fuze resolution: both fine with >500 MHz BW. Decision: **Ka-band 35 GHz** baseline for reliability/all-weather/thermal, with a note that a 79-94 GHz variant buys +2-3 dB clear-air SNR and 2× angular precision at the cost of weather robustness and thermal complexity — recommend W variant only for clear-weather high-volume low-cost (automotive SiGe RX) applications. Hmm wait, but actually let me double-check the rain argument importance: Shahed-136 attacks happen year-round incl. rain; "reliably detecting at 5 km" — in 5 mm/h rain, 79 GHz two-way loss at 5 km: 2×5×5 = 50 dB — catastrophic. Ka: 2×5×0.7 = 7 dB — survivable (detection range shrinks to ~3-4 km). X-band: 0.1 dB/km — negligible. Could we even consider X-band (9.4 GHz)? Gain at 54mm: G=(π×0.054/0.032)²×0.6 = (5.3)²×0.6 = 16.9 → 12.3 dBi; beamwidth 70×0.032/0.054 = 41°. Per-watt SNR vs Ka: (8.57/32)² = 0.072 → -11.4 dB vs Ka. With 50 W peak X-band GaN (trivial) ×20% duty = 10 W avg → +2.2 dB vs Ka's 4-5 W → still -9 dB worse; and 41° beam can't home. So no. Ka it is. Also clutter: X-band beam too wide → clutter issues worse. Ka confirmed. One more check — the subarray scan limit at Ka: 8 subarrays, phase centers ~13.5 mm apart → grating-lobe-free scan ≈ asin(λ/(2·13.5mm)) = asin(8.57/27) = asin(0.317) = 18.5° → ±18-20°. Search coverage ±20° at 5 km = ±1.8 km — generous vs cue errors. Good. **Monopulse accuracy re-check:** θB = 11°. Monopulse slope k_m ≈ 1.6. σ_θ = θB·sqrt( ... )/(k_m√(2·SNR)): at SNR 15 dB (31.6): sqrt(63)=7.9; σθ = 11/(1.6×7.9) = 0.87°. At 300 m → 4.5 m 1σ per look; with 25 Hz updates and Kalman filtering (α-β or EKF) → smooth; PN with N' ~ 3-4; final miss contribution from angle noise: integrated ~ 1-3 m. Warhead lethal radius needed 3-6 m → 60mm frag warhead ~2-3 kg... a 60mm HE frag has ~ 60-80 g? No — 60mm rocket total maybe 3-4 kg with maybe 0.5-0.8 kg warhead → lethal radius vs Shahed structure 2-4 m (goal: sever wing spar/control linkages or destroy engine/prop). Miss ≤2-3 m Pk ~0.6-0.85 per shot; salvo doctrine handles it. Also kinetic hit possible given target is slow; guidance note. **Additional detection enhancer — micro-Doppler:** Shahed prop: 2-blade ~ 0.5-0.7 m radius? Let me estimate: engine 50 hp at ~2500-3000 rpm? Unducted prop on Shahed-136 looks ~ 0.7-0.8 m diameter? Blade pass frequency at 2900 rpm 2 blades = 96 Hz. Blade tip speed π×0.75×48 Hz(rps) = 113 m/s → blade RCS at Ka: blade length 0.37m × width 0.06 m, rotating: blade flash σ up to... blade flat plate rotating gives flashes; hub/blade contributions at Ka ~ 0.01-0.1 m² periodic flashes + JEM-like modulation lines at 96 Hz ± harmonics spread by blade Doppler ±226 Hz... blade tip tangential 113 m/s → line spread ±113×cos... JEM lines at multiples of blade rate with ±(2×113/λ)=±26 kHz spread envelope? JEM line spacing 96 Hz; each line spread ~ ±(tip speed component)... This gives a distinctive signature: detection can integrate body line; classification uses 96 Hz ± lines. Also vibrational micro-Doppler from engine. Include as classifier feature & marginal-detection aid (integrate body+JEM energy). Also worth noting the engine's exhaust... not RF relevant. IR is the other seeker option but user asked radar. **Form factor & schedule realism:** length budget: radome 150 (ogive ~2.5 calibers), seeker electronics ~120-140, total seeker ~270-290 mm incl radome; behind it warhead, motor ~ etc. Mass 0.5-0.7 kg seeker. Power bus: 28 V from thermal battery, ~30-45 W peak electrical. Let me also compute PA heat more carefully: 8 subarray PAs, each GaN 4 W peak at PAE 30% (35 GHz GaN PAE 25-35% typical) → DC per PA 13 W at duty 20% → 2.7 W avg → 8×2.7 = 21 W avg heat during acquisition burst. Ouch. That's a lot for 60mm passively cooled over 10-20 s: heat capacity needed: 21 W × 20 s = 420 J → with aluminum+PCM (aluminum 0.9 J/gK: 200 g Al heats 420/180 = 2.3 K/... 420 J/(200g×0.9) = 2.3 °C rise — trivial! PCM unnecessary; just a machined Al thermal spreader tied to the airframe; junction rise via θjc. Fine — no thermal problem for the engagement durations involved (max ~40 s). Good — remove the PCM complexity, keep duty cycling anyway. Wait, but average over the whole flight vs battery: battery sized ~ 50 W × 60 s = 3000 J → thermal battery ~ 45-60 g + activation. Fine. Let me also double check Ka noise figure: LNA 35 GHz GaAs/pHEMT: NF 1.5-2.2 dB chip; system F ≈ 3-3.5 dB incl. radome/switch/feed. I used 3 — ok. Detection threshold numbers: for Pd=0.9, Pfa=1e-6, Swerling I single-look: SNR_req ≈ 21 dB. Multi-look N=4 (independent) reduces per-look to ~13.2 dB (approx for Swerling I 4-pulse NC integration: required per-pulse SNR ≈ 13 dB). N=8: ~9-10 dB per look. So: - Nose σ=0.05: per-look (250 ms) 7-8.4 dB; need 13 for N=4 → shortfall ~5 dB → extend to N=8-12 looks over 2-3 s (feasible during midcourse) → required per-look ~9 → we have 7-8.4 → still ~1 dB short worst-case-fade... but per-look SNR improves as range closes (R⁴ term): 3 s of flight at 320 m/s closure = 1 km → +8.5 dB. So declaration typically at 4.2-4.8 km vs 0.05 m², and by design intent "detect at 5 km" holds for σ≥0.1 m². Present honestly: 5 km requirement met with ~50% probability against 0.05 m² nose (median Shahed), ~90% vs 0.1 m², >95% vs side 0.3-1 m²; time available means effective acquisition by 4.5 km for the worst fades with no mission impact. If hard 5 km @ Pd 0.9 @ σ=0.01-0.05 is contractual: options: raise Pavg to 8-10 W avg (GaN growth), 94 GHz high-power variant (+2-3 dB clear air), or semi-active illumination mode. Hmm, actually, let me reconsider: maybe I'm being too pessimistic on nose σ. Let me look at this from the "what would an actual threat RCS file say": Published analyses of Shahed-136-class (composite delta-wing, ~200 kg, 3.5 m wingspan, metal engine + prop + battery + steel fuselage truss under fiberglass): X-band measurements of similar composite UAVs (e.g., studies on small UAS RCS) show frontal σ at 10 GHz ~0.01-0.05 m² for composite-with-metal-internals... but Shahed has a notable metal content: the MD-550 engine (air-cooled boxer-ish 4-cylinder, all aluminum ~ 30+ kg) sits directly behind the wing center; head-on, the engine front face (cylinders, cooling fins) is a ~0.2×0.3 m irregular metal surface behind ~1.5 m of fiberglass nose... head-on at shallow incidence, radar passes through nose skin (lossy glass ~1-2 dB two-way) reflects off engine → 0.06 m² plate at normal incidence: σ = 4πA²/λ² = 4π×0.0036/7.3e-5 = 620 m²?! Only if perfectly normal & flat & unresolved; in reality incidence varies, surfaces curved/finned → effective maybe 0.1-2 m² typical with fades. Honestly, Shahed nose-on MMW σ could often be ≥0.1 m² because of the engine. The lowest-nose-σ scenario is more like... the plywood/fiberglass variants (Shahed-136 built variously with fiber... some versions with composite; also Moped... ). I'll present σ table: nose 0.02-0.3 m² (median ~0.1 due to engine/battery metal), side 0.2-2 m² (fuselage broadside + wing edge + prop flashes), rear 0.1-1 m² (prop/engine exhaust dominate). Design point: close for Pd≥0.9 at 5 km @ σ=0.1 nose (we have 10.4 dB per 250 ms look at E=1.1 J → 4-look needs 13 → borderline; N=8 needs ~9-10 → we have 10.4 → PASSES). And side 0.3+: comfortable (15.2 dB/look at σ=0.3). Let me recompute with final numbers cleanly for the presentation table: FINAL DESIGN POINTS: - f = 35 GHz (Ka), λ = 8.57 mm - Aperture: 54 mm circular, 8-subarray passive ESA, 4-channel monopulse (Σ, Δaz, Δel + guard via Δ2 or dedicated) - ε_ap = 0.5 (taper for -25 dB sidelobes + losses): G = (π×0.054/0.00857)²×0.5 = 391.9×0.5 = 196 → 22.9 dBi ≈ 23 dBi. Beamwidth θ3dB = 70λ/D = 11.1° (with -25 dB Taylor taper, ~12.5°; call it 12°). - Pt peak = 25 W (8×3.1 W at subarray ports after feed loss), duty up to 25% burst-limited; Pavg = 5 W during acquisition; E_250ms = 1.25 J. - Waveform: pulsed-LFM stretch: 500 MHz / 10 μs chirp, PRI 40 μs (PRF 25 kHz), duty 25%. Unambiguous range 6 km. Doppler ambiguous (Δv_amb = λ·PRF/2 = 107 m/s) → resolved by PRI stagger (40/46.7/55.6 μs) extending to ±320 m/s with 2-3 bursts + INS cue. - RX: dechirp → IF ≤ ~2 MHz → 12-bit 5 MS/s ADC ×4 ch. - CPI = 250 ms (6250 pulses? 6250×40 μs = 250 ms; Doppler res 4 Hz; coherent gain huge... processing: per 250 ms CPI, 2D FFT: range bins 30 cm? no — stretch with 500 MHz over 10 μs pulse: range resolution c/2B = 30 cm; number of range samples per chirp: IF BW 2 MHz/ (chirp rate × range window...) range window: unambiguous 6 km; beat max = 2R·slope/c + fD = 2×6000×(500e6/1e-5=5e13)/3e8 = 2 MHz + fD 75 kHz → 2.1 MHz → 4096-pt FFT/chirp → 30 cm bins over 6 km... processing load: 6250 chirps × 4096-pt FFT × 4 channels in real time = heavy but FPGA-doable at 250 ms latency (6250×4096×~22 ops... 560 G-ops per CPI per channel — modern RFSoC/AGILEX handles with FFT optimization; or decimate: process only cued range window (e.g., 1 km = 3400 bins? 1 km/0.3 = 3333 bins) after coarse initial sweep. Simplify: first-pass acquisition uses decimated coarse bins (5 m), then track zooms. I'll note the processing chain with an RFSoC-class FPGA (Zynq UltraScale+ RFSoC: 8× 4 GSPS ADCs, DSP slices — perfect fit, COTS). Actually for presentation simplicity: acquisition CPI 50 ms × 5 noncoherent looks with 2D-CFAR + TBD. Keep flexible. - F = 3.5 dB; L_total = 5.5 dB (atm 1.2 at 0.12 dB/km clear; radome 0.8 (half-wave glass at Ka, ±20° scan AOA... at 20° scan the radome mismatch grows: use thin-wall or slotted A-sandwich; take 1 dB), feed/switch 1.5, processing 2 incl. CFAR/straddle). LINK BUDGET @5 km (per 250 ms look, E=1.25 J = +0.97 dBW·s): SNR = E + 2G + 20logλ + 10logσ − 30log(4π) − 40logR − 10log(kT0F) − L Let me just plug: E=0.97; 2G=45.8; 20logλ = 20log(0.00857)= -41.34; 10logσ: -20 (0.01), -13 (0.05), -10 (0.1), 0 (1.0); (4π)³=32.98; R⁴=147.96; kT0F: 10log(4e-21×10^0.35)= 10log(8.98e-21)= -204.5 → subtracting = +204.5; L=5.5. SNR = 0.97+45.8−41.34−32.98−147.96+204.5−5.5+10logσ Constants: 0.97+45.8 = 46.77; −41.34 = 5.43; −32.98 = −27.55; −147.96 = −175.51; +204.5 = 28.99; −5.5 = 23.49. So SNR = 23.49 + 10logσ. σ=0.1 → 13.5 dB; σ=0.05 → 10.5; σ=0.01 → 3.5; σ=0.3 → 18.3; σ=1.0 → 23.5. Compare required per-look for Pd=0.9: N=8 looks (2 s): ~9-10 dB; N=4 (1 s): ~13 dB; single: ~21. - Nose 0.1 m²: 13.5 dB/look → N=4 (1 s) → Pd≈0.9 ✓ at 5 km. N=8: margin 3.5-4.5 dB. - Nose 0.05: 10.5 → N=8 (2 s) ✓. - Side 0.3-1: single-look threshold nearly met; 0.5 s → Pd>0.95. ✓ typically acquires 6-7 km. - Deep fade 0.01: 3.5 dB → detect ~3.5-4 km or with TBD several s. Handle by closing range (mission fine). Great, this design closes the requirement as stated for realistic Shahed RCS. I'm fairly confident in these figures. Also add margins row: with Pavg growth to 8-10 W (GaN roadmap) → +2-3 dB everywhere. Also should I double-check G with ε=0.5: 22.9 dBi — for a 54 mm aperture at 35 GHz, G = 4πA/λ² × ε: A = π×0.027² = 2.29e-3 m²; 4πA/λ² = 4π×2.29e-3/7.34e-5 = 391.9; ×0.5 = 196 → 22.9 dBi. Yes. Beamwidth: 70×0.00857/0.054 = 11.1° untapered; with taper ~12.5°. Wait — sanity: a 54mm aperture at 35 GHz, is 12° plausible? Gain 23 dBi ↔ beamwidth ≈ √(32000/G)? For circular: G ≈ 29000/(θ_E θ_A)... θ=12.5: G ≈ 29000/156 = 186 → 22.7 dBi ✓ consistent. Angular rate coverage: strapdown e-scan ±20°, monopulse to ~0.5° off-boresight accuracy; terminal LOS rates fine. **Clutter check for side-aspect at low altitude (the actual hard case):** presented above; mitigation: elevation monopulse + Doppler 2D CFAR; engage from above/below; ground clutter at 35 GHz low grazing is weaker (σ⁰ −25 to −35 dB at 1-3° grazing); note MTI improvement factor via Doppler filtering ~ 30-40 dB adequate; sidelobes -25 dB with taper. Also Shahed's 96 Hz blade lines aid confirmation. OK. **Structure of the answer:** I'll write it as a concise design document: 1. Requirements interpretation (turn into engineering specs) — brief table of derived specs. 2. Driving physics (aperture→frequency choice) + frequency trade table + selection rationale (Ka 35 GHz). 3. Architecture overview (block description, 8-subarray limited-scan monopulse passive ESA, strapdown) with a little ASCII layout. 4. RF/waveform parameters table. 5. Link budget & detection performance (table: SNR per look vs aspect; Pd vs integration) — front/side verdict, plus margin policy and closing options. 6. Clutter & aspect-specific considerations (front vs side nuances: nose fades, wing flash, JEM micro-Doppler for classification, ground clutter for crossing shots). 7. Survival/environmental design (0→278 m/s in 5 s → 5.7 g avg, design loads, vibration, shock, thermal, radome; qualification levels). 8. Processing chain & guidance/fuzing integration. 9. Risks + growth paths (W-band variant, semi-active mode, power growth). 10. Rough physical summary (length, mass, power, cost). Keep tables tight. Use metric units. Should be technically deep — the audience asked a hard-core question; give hard-core answer but organized. Also I should include assumptions clearly (RCS values are assumptions — flag them as needing measurement; suggest RCS measurement program). One more consideration — 60mm: is the seeker 60mm outer diameter including radome? Yes, "fitting in a 60mm diameter". Aperture 54 mm with 3 mm radial structure margin (radome wall + array edge). Could push to 55-56mm with careful design; keep 54. Also the "5 seconds to 1000 km/h" implies average axial acceleration 5.67 g — I should state that I interpret this as rocket boost and design for an envelope (peak ~15 g, vib 12-15 g_rms). If instead gun-launch were intended (1000 km/h in 5 s is NOT gun-like; guns do it in ms), so rocket boost confirmed. Good. Also note: 1000 km/h = 277.8 m/s ≈ Mach 0.82 at sea level. Coyote-like: fine. Also, kill assessment: single-shot Pk estimate ~0.7-0.85 vs Shahed with 2-3 m miss; salvo 2. Recommend fuze + warhead interface numbers: lethal radius 3-4 m target kill criteria (wing spar/control actuator cut or engine destruction → immediate crash or mission kill). Ah also: should mention seeker activation timing (thermal battery activation at launch; seeker full-up within 0.5-1 s; boost-phase inertial-only; acquisition begins T+5-8 s). And datalink? Optional command-uplink for midcourse updates from fire-control (adds robustness vs deep fades; cheap one-way S-band link). Mention as optional. Cost/industrial: BOM estimate at 1,000+ units: RF front-end (8 subarray MMIC sets + comparator + LNA/switch) ~ $800-1,500; GaN PA set ~ $600-1,000; RFSoC/ADC ~ $400-800; array PCB/structure ~ $300; radome ~ $200; misc/power ~ $400 → seeker ~ $3-5k at volume; missile $25-60k total. Attritable vs Shahed cost ($20-50k) — good trade. This matters to make it a real design. Also alternative architectures considered & rejected section (gimbal — volume/complexity; full ESA — cost; FMCW — duplexer leakage; pulse-Doppler selected; semi-active — infrastructure dependency; active only per requirement). I also want to double-check the monopulse comparator at 35 GHz in planar form: standard — fine. And the Δ channel noise: 3 channels ADC; Σ for detection, Δ/Σ ratio for angle. Guard: form a fourth channel from outer ring? Simpler: use low-sidelobe taper + 2D CFAR; skip guard channel; or use Δ2 (difference-of-differences) as sidelobe guard. I'll mention SLC/guard via a small aux patch — one line. Potential ADC count: 4 channels (Σ, Δaz, Δel, guard) × 12-bit 5 MS/s — RFSoC has 8 → fine. Waveform ambiguity detail: with PRF 25 kHz, unambiguous range 6 km ✓ (target within 5 km, but note Tx pulse leakage/eclipsing near range OK). Doppler ambiguous at 107 m/s → 3-burst stagger resolves within interceptor kinematics; plus altitude/ground clutter line known from INS. Good. Fuze mode: at <200 m, PRF raise & BW to 1 GHz; beat at 200 m with slope 1e14: 2×200×1e14/3e8 = 133 kHz + fD small → IF 500 kHz, 4096 samples → 10 cm resolution profile; detect target extent crossing; burst at 2-5 m with monopulse POA gating; add centripetal/contact backup. One line each. Radome at Ka with ±20° scan: thin-wall A-sandwich quartz/honeycomb or slotted wall; transmission >90% over scan volume; specify. Boresight stability: array on machined Al bulkhead; monopulse calibration via built-in cal coupler at launch (self-test); aim-point bias <0.2°. Also alignment of strapdown IMU-to-array boresight: factory calibration + transfer alignment; 0.1-0.2° residual. OK, one more global sanity pass on the "can a 60mm Ka seeker detect at 5 km" claim vs known systems: Helicopter Longbow radar (Ka-band) detects 0.1-1 m² targets at 8 km with ~10 W avg, 100+ mm aperture... APWKR/AMRAAM-class seekers detect 1-5 m² at 10-20 km with 160mm X-band. Our 5 km vs 0.05-0.3 m² with 54mm at Ka with 5 W avg — consistent with the physics above (and MMW mortar/coastal systems similar). Plausible. The one borderline is the deepest nose fades (~0.01-0.02 m²) — honestly flagged. Time-to-acquire vs closing speed: at 330 m/s closure, from detection range R to R-Δ: 1 km per 3 s. Budget: acquisition ≤2 s. Also mention: the front-aspect case often = head-on intercept where Shahed flies at the defended area and interceptor launches forward — closing ~ 278+160 km/h(44 m/s) = 322 m/s → 5 km → 15.5 s to intercept. Plenty. For side: crossing target: interceptor fired with lead; LOS rate small; closure ~278 m/s → ~18 s. Fine. Edge: tail-chase (rear aspect): closure ~234 m/s; seeker rear-aspect σ includes engine+prop: strong JEM, σ~0.1-0.5 → fine. User didn't ask; one line. Alright — also environmental: "reliably surviving 0-1000 km/h in 5 secs" — I'll also translate to: max dynamic pressure q = 0.5×1.225×278² = 47.3 kPa at sea level; radome pressure loads & fin flutter margins... radome fine. Also note supersonic-capable variant would need redesign (aerothermal, radome). Let me think about whether to include an honest "feasibility verdict" up front: Yes — verdict: feasible with margins for realistic Shahed RCS; the binding constraint is nose-aspect fades at 5 km; design closes with 1-2 s integration; risks listed. Also — should I ask clarifying questions instead? The request is specific enough; ambiguities (gun vs rocket launch, all-weather?, active vs semi-active, cost target) — I'll state assumptions and proceed; asking first would be less useful. I'll flag the assumptions explicitly and note where the answer would change. Format: headings, compact tables, some bullets, minimal fluff. Length: this deserves depth — target maybe 1300-1800 words with tables. It's a "design a system" request; comprehensive is appropriate. I'll avoid over-nesting bullets. Use dB math shown compactly in tables so it's checkable. Include a small ASCII stack diagram maybe — a compact one could help visualize the 60mm packaging. Keep it simple. Let me draft the tables: **Table 1: Derived requirements** - Platform: 60 mm OD rocket interceptor, rocket boost 0→278 m/s in 5 s (5.7 g avg; design env: 20 g axial peak, 12 g_rms vib) - Target: Shahed-136-class (composite delta-wing OWA UAV, Δv ≤ 2 g, 40±10 m/s cruise) - RCS assumptions @35 GHz: nose 0.02–0.3 m² (median ~0.1, metal engine/battery behind glass nose); side 0.2–2 m²; (+ prop JEM 96 Hz-class lines) - Detection: Pd ≥ 0.9, Pfa ≤ 1e-6 @ 5 km, nose & side, clear air (rain case handled: 3.5-4 km @ 5 mm/h) - FOV: ±20° e-scan; terminal: track to ≤5 m miss; fuze assist <200 m. **Table 2: Frequency trade** (X/Ka/W rows: gain, beam, achievable avg power, rain loss @5 mm/h two-way 10 km... at 5 mm/h: X 0.5 dB total; Ka ~7 dB; 94 GHz ~35-50 dB; per-watt SNR @5km...; verdict) **Table 3: Seeker spec summary** (frequency, aperture, gain, beam, Pt/Pavg, NF, losses, waveform, CPI, mass, length, power, cost) **Table 4: Link budget @ 5 km** (rows as computed; columns σ = 0.01/0.05/0.1/1.0) **Table 5: Detection outcome** (per aspect: SNR/250 ms look; integration time to Pd 0.9; verdict) Then prose sections: architecture rationale (strapdown ESA vs gimbal), search-to-handover concept, clutter/side-aspect notes, micro-Doppler classification, environment/survival design, processing & guidance/fuzing, packaging/mass/power/cost, risks & growth, assumptions to validate (RCS measurement, rain ops doctrine). The link budget table: show the dB arithmetic rows. Compact. Double-check a couple of dB figures: - 20log(λ=0.00857) : log10(0.00857) = -2.067 → -41.34 dB ✓. - 10log(0.375 J)... E = Pavg 5 W × 0.25 s = 1.25 J → +0.97 dBW·s ✓ (10log1.25=0.969). - 2G = 45.8 dB ✓ (G=22.9). - (4π)³ = 32.97 dB ✓. - R⁴ @5000 m: 147.96 ✓. - kT0F: T0=290: kT0 = -203.98 dBW/Hz; F=3.5 → -200.48; SNR adds +200.48 ✓. - L=5.5. Constant sum: 0.97+45.8−41.34−32.97−147.96+200.48−5.5 = 19.48. Hmm let me re-add: 0.97+45.8=46.77; 46.77−41.34=5.43; 5.43−32.97=−27.54; −27.54−147.96=−175.50; −175.50+200.48=24.98; 24.98−5.5=19.48. So SNR = 19.48 + 10logσ. σ=0.1 → 9.48 dB; σ=0.05 → 6.48; σ=0.3 → 14.26; σ=1 → 19.48. Hmm, slightly different from before because earlier I used L=5.5 and F=3.5 both and got 23.49 — I must have made an arithmetic slip before (I used kT0F +204.5 there with F=3.5: -203.98+3.5 = -200.48 → +200.48 not +204.5; earlier I wrote 204.5 — error). Correct: +200.48. So SNR(σ=0.1) = 9.5 dB per 250 ms look. Requirement N=8 looks (2 s): per-look req ~9-10 dB → 0.1 m² closes at ~2 s integration at 5 km exactly. N=4 needs 13 → not met per look at 5 km for 0.1; side 0.3-1: 14.3-19.5 per look → single-look to 4-look OK, Pd 0.9 within 0.5-1 s ✓✓. Hmm, so nose 0.1 m² @ 5 km needs ~2 s of TBD/NC integration (with SNR also rising as we close: after 2 s at 322 m/s closure we're at 4.36 km → SNR +16%... R⁴: (5/4.36)⁴ = 1.74 → +2.4 dB). So in practice Pd 0.9 achieved within ~1.5-2 s at 4.5-5 km. Operationally fine (15 s to spare). For σ=0.05: per-look 6.5 → need ~9-10 → integrate ~3 s → effectively 4-4.5 km. For 0.02: ~3.5 dB/look → TBD 4-5 s → 3.5-4 km effective. Present as: "5 km requirement: met for side (σ≥0.2) with margin; met for nose σ≈0.1 within ≤2 s integration (well within geometry timeline); if intelligence puts median nose σ at 0.05, effective acquisition ~4-4.5 km with zero mission impact (detection happens during flyout); deep-fade 0.01-0.02 cells handled by TBD over 3-5 s." Also power growth option: +3 dB with Pavg 10 W (GaN 35 GHz growing) → all numbers shift; or 94 GHz variant +7 per-watt but power-limited. I think one more important honest point: the side-aspect case at 5 km with ground clutter: detection is clutter-limited not noise-limited in some geometries at low altitude; mitigations as discussed; at typical Shahed engagement altitudes (they fly 100-1000 m AGL... often 50-500 m over front line, higher over targets) — with interceptor launched from ground, LOS to target nearly level; ground clutter enters via sidelobes; with -25 dB sidelobes + Doppler filtering, acceptable; recommend engagement doctrine: launch so intercept happens above 300-500 m AGL. One short paragraph. The prop JEM: blade pass ~90-100 Hz (I'll say ~80-100 Hz depending on variant, tip speeds ~100-150 m/s → line spread ±26-35 kHz around body line) — use as confirmation + weak-target integration (integrate JEM lines adds a few dB effective). Keep brief. Fuzing: mention integrated RF fuzing using same front-end: wideband burst mode <200 m; POA-gated burst at 2-5 m; backup graze/contact; centripetal. Warhead lethal radius 2-4 m target-kill (control surfaces/engine). Single-shot Pk ~0.75-0.9 (guidance+fuze+lethality stack); salvo doctrine optional. Environment section specifics: - Loads: axial setback design 30 g (req avg 5.7, peak est 12-18 g with motor spike), lateral 15 g, safety 1.5×; random vib 20-2000 Hz 12 g_rms for 5-15 s motor burn + aero; pyro shock 1500 g 0.1 ms at case attenuated to <100 g at electronics (elastomer isolators? At 79... at Ka passive array, avoid isolators if possible; use hard-mounted with damping); thermal: -40…+55 storage, +70 radome skin in flight; condensation/humidity; salt (naval option). - Design practices: machined Al chassis carrying array (CTE match, thermal path doubles as structure), corner-bonded+underfilled BGA or QFP, MLCC ≤1210 flexible-terminator, staked discretes, SMPM blind-mates with 0.1-0.2 mm float, no waveguide (planar all the way to aperture), conformal coat, Gore vent for pressure, potting aft bulkhead ablative vs motor gas (if hot tube launch). - Boresight stability: monolithic array carrier; cal at launch (built-in coupler) — residual <0.2°. - EMCOM: none needed (silent RF seeker; RX-only bands during boost for interference... skip). Cost & schedule: seeker TRL path: MMICs all commercial-grade (35 GHz GaN PA, LNA, switches COTS; RFSoC COTS) → TRL 6 in ~18-24 months; unit cost target $3-5k @ 5,000/yr. Risks table (short): 1) nose RCS below 0.05 → mitigations (power growth +3 dB, TBD, 94 GHz variant, semi-active mode via ground illuminator); 2) rain ops → doctrine + Ka choice; 3) clutter in crossing low-alt → geometry doctrine + Doppler CFAR; 4) thermal at 10 W avg → duty scheduling; 5) strapdown endgame LOS rates → PN gains, e-scan ±20°, airframe authority 15-25 g. Also add note: all-weather alternative / companion: add passive RF (Shahed emits little; rely on ground cue) or IR dual-mode — out of scope, one line. I think I should also briefly show why not gimbal: at 60 mm, a gimbal leaves ≤40-45 mm effective aperture (−2-3 dB), adds bearings/motors rated for 20 g shock-vibe, costs complexity — the ±20° passive ESA covers the real need given fire-control cueing; endgame geometry keeps LOS within body ±10° for PN. Also why not FMCW: on a single shared aperture, TX-to-RX leakage at 5 km noise floor requires >90 dB isolation + cancellation; pulsed with T/R switch sidesteps it; duty 25% preserves energy. ✓ mention one line. Alright — also the "survive 0→1000 km/h in 5 s" I interpreted as rocket. Add a caveat line: if gun-launch (10⁴ g) instead, the architecture changes (SADARM-style survival techniques), but the stated profile is rocket boost. Units and rigor: define Pd/Pfa. Use "σ" with m² and dBsm both. Give the radar equation used. Make tables checkable. ASCII packaging sketch (compact): ``` ◄── nose shock ──► ┌────────────────┐ │ radome (tangent ogive ~150 mm, quartz A-sandwich) │ 54 mm aperture: 8-subarray monopulse array │ T/R MMICs on reverse side │ beamformer + monopulse comparator │ RFSoC + ADC/DAC + FPGA, MEMS IMU │ DC-DC + thermal battery interface │ fuze S&A + ESLD └─── aft bulkhead (motor interface) ``` Maybe simpler to describe in a table with lengths. Lengths: radome 150; array+MMIC stack 20; comparator/feed 15; processor 30; power 15; fuze 20 → seeker ~250 mm incl radome. Diameter 60. Seeker mass: radome 60 g, array/RF 150 g, digital 80 g, power 100 g, fuze 60 g, structure 80 g → ~530 g. Power: 45 W peak electrical (TX bursts), 8 W logic; battery 60 s @ 60 W → thermal battery ~120 g incl activation. Airframe control: canards + tail, 15-25 g authority — mention only as interface. Guidance: PN with body-pointing loop; handover from fire-control with quality gate (cue error <0.5°); search ladder ±20° → 25 beams at 11-12° beam: steps of 8°: 5×5; at 60 ms/beam = 1.5 s full frame — but with cue ±3°: 4 beams, 0.25 s. TBD bank accumulates. Track: 25-40 Hz update, EKF [range, Doppler, az/el]; guidance command generation; endgame gains schedule. Alright — also a short "why this closes" verdict paragraph at top: 2-3 sentences summary: Feasible; Ka-band 35 GHz strapdown limited-scan monopulse ESA, 54mm aperture, 25 W peak/5 W avg, closes 5 km vs side (σ≥0.2) easily, vs nose (~0.1) with ≤2 s integration; binding risk = worst-case nose fades → growth options listed. Word budget: I'll try to keep to ~1500 words + tables. The user asked a design question — they'll want the numbers. I'll write tight. One more consideration on units: 1000 km/h = 278 m/s = M0.82. Boost avg 5.7 g. Good. Also the phrase "reliably surviving 0-1000 km/h in 5 secs" — could also mean the seeker must function THROUGH the acceleration (i.e., operate during/after boost). I'll assume operation from launch (IMU aiding) with full RF modes after burnout/motor-noise quieting (vib during burn raises phase noise floor → schedule acquisition after burn unless needed earlier; T+5-6 s). Add line. Also add: RF silence during boost? Not needed. Also possible question: why not just IR seeker? One line: MMW chosen because radar specified + all-aspect/weather; IR is a viable dual-mode add. Now, final check on the claim "detects 0.1 m² at 5 km": compare with automotive 77 GHz long-range radars: they detect cars (σ~10 m²) at 200-300 m with 10 mW. Scaling: our seeker has 25 W peak (×2500, +34 dB), aperture gain 23 dBi vs ~25 dBi similar, processing 1.25 J... automotive ~ (11 mW × 50 ms) etc. Our energy is ~10⁴-10⁵× automotive per dwell; automotive SNR for a car at 250 m is ~20-40 dB → we'd have a car at... equivalent: energy +34 dB & σ +10 dB (10 m² vs 0.1) & R: 5000 vs 250: R⁴ penalty = 20⁴ = 160,000 = 52 dB. Net: 34+10−52 = −8 dB relative → so a 0.1 m² at 5 km ≈ a car at 250 m minus 8 dB → automotive radars do cars at 250 m with tons of margin (SNR 20-30 dB) → consistent with our 9.5 dB/look after losses. ✓ Physics cross-checks. And against Longbow: AN/APG-78 (35 GHz, ~10 W avg? aperture ~230 mm, G ~34 dBi): detects 8 km on 1-5 m². Our: aperture −11 dB (23 vs 34 dBi: 11 dB), Pavg −3 dB, R: 5 vs 8 km +8.3 dB (R⁴)... net vs Longbow: σ... Longbow with G² higher by 22 dB, power lower by 3 → +19 dB for Longbow... their 1 m² @ 8 km → our 0.1 m² @ 5 km: σ −10, R +8.3 → equivalent σ_eq = 1 m² → we're at ~ −19+10... this says we should be ~19 dB worse than Longbow's 1 m² @ 8 km case for 0.1 @ 5 km +8.3+... I'm double-counting; anyway order-of-magnitude consistent: Longbow does ~10 km on 5 m² vehicles; scale: σ −17 dB, aperture −22 (G²), power −3, R⁴ for 5 km vs 10 km +12 → net −30 dB relative to a system with ~10-20 dB margins → we end ~ −10..−20 dB → we then use 250 ms CPIs and TBD (Longbow dwells are shorter)... plausible, consistent with our marginal-but-closing result. Fine. Also note: Shahed-136 wing area gives strong specular at shallow aspect offsets from nose-on (wing normal flash): so nose-on fades are momentary; scanning monopulse sees flashes — Swerling II-like decorrelation at 35 GHz over 250 ms? Decorrelation time: aspect change of ~0.05° changes speckle; at 322 m/s closure + target 45 m/s, aspect rate maybe 0.5-2°/s → 250 ms → 0.1-0.5° → decorrelates across looks → Swerling II (partially decorrelated) which needs LESS integration loss for Pd (fluctuation-averaging across looks is favorable for NC integration). Good for us: mention "fluctuating target; looks decorrelate within ~0.2 s; multi-look integrates flashes effectively." OK, writing now. Structure: # Title-ish header Verdict paragraph (3 sentences). ## 1. Requirements → specs (table + assumption callouts) ## 2. Why Ka-band 35 GHz (trade table + 2-3 lines rationale; aperture-limited physics note) ## 3. Architecture (paragraph + mini block diagram/description; why strapdown ±20° passive ESA, why pulsed not FMCW) ## 4. Key parameters table ## 5. Link budget @ 5 km (table with dB arithmetic) + detection outcome table + honest verdict incl. fades/rain ## 6. Aspect-specific effects (nose fades & wing flash; side: clutter geometry; JEM 80-100 Hz lines for confirm/fuze) ## 7. Surviving boost (environment derivation + design rules + radome) ## 8. Processing/guidance/fuze (compact) ## 9. Packaging, power, mass, cost (compact table) ## 10. Risks & growth options (bullets) ## 11. Assumptions to validate / next steps (RCS range measurement of representative article, flight test, rain testing) — short. Keep the whole thing tight. Tables small. Avoid repeating numbers. Let me finalize numbers for consistency: - f 35 GHz, λ 8.57 mm - Aperture 54 mm Ø; ~120 patch elements in 12×12 λ/2 grid; 8 subarrays of ~15 elements; monopulse comparator → Σ, Δaz, Δel (+aux guard patch) - G 22.9 dBi (ε 0.5 incl. taper −1.3 dB for -25 dB sls, losses); θ3dB ≈ 12° - E-scan ±20° (subarray pitch λ/2, 8-subarray steering; grating-free ±18-20°), 5-bit phase shifters - Tx: 8× GaN PA ~3.5 W each → 25 W peak at aperture after 0.8 dB combining/feed loss; duty ≤25% burst, 10% sustained; Pavg 5 W burst. - NF 3.5 dB (pHEMT LNA 2 dB at subarray port) - Losses 5.5 dB total @5 km (atm 1.2, radome 0.8-1.0, feed/switch 1.5, processing 2.0) - Waveform: pulsed-LFM stretch, 10 μs/500 MHz, PRI 40 μs (PRF 25 kHz, R_unamb 6 km); 3-PRI stagger for velocity ambiguity (±330 m/s); CPI 250 ms (Swerling-tolerant); acquisition: 5×250 ms looks + TBD over ≤2.5 s; track update 25-40 Hz (CPI 25-40 ms terminal) - ADC 4× 12-bit 5 MS/s (RFSoC); Doppler res 4 Hz/250 ms - Fuze mode <200 m: 1 GHz BW, 10-15 cm resolution profile; POA-gated burst 2-5 m; graze/contact backup. - Detection @5 km (per 250 ms look): SNR = 19.5 + 10logσ → 9.5 dB @0.1 m²; 6.5 @0.05; 3.5 @0.01; 14.5 @0.3; 19.5 @1.0. - Pd 0.9 (Pfa 1e-6): side σ≥0.2-0.3 → ≤0.5-1 s; nose σ~0.1 → ≤2 s (incl. closing-range gain); σ 0.05 → ~3 s → effective 4.3-4.5 km; deep fades 0.01-0.02 → TBD 4-6 s → 3.5-4 km. Timeline allows (≥15 s between 5 km and impact). - Rain 5 mm/h: Ka loss ~0.6-0.7 dB/km → −6-7 dB round trip at 5 km → nose detection 3.5-4 km; side still ≥5 km. - Environment: design-to: axial 30 g (peak thrust ~12-18 g est.), lateral 15 g, vib 12 g_rms 20-2000 Hz motor burn, shock 100 g/11 ms at electronics after isolation (motor ignition 1,500 g @case), aero: M0.82, q=47 kPa, stagnation ~55 °C; thermal: −40…+70 °C; seeker electronics avg dissipation ~5 W over ≤60 s. - Mass 0.55 kg, length ~250 mm (incl. 150 mm radome), power 45 W pk/8 W avg elec, thermal battery 120 g; cost ~$3-5k seeker BOM @ volume; TRL path 18-24 mo. I'm reasonably confident. Also I'll note where numbers are assumptions (RCS!) and recommend a measurement program — intellectual honesty. One more: "from the front and side" — also ensure I address SIDE case explicitly: side detection easy on σ but hard on clutter/geometry (crossing) — I have that section. And front: SNR-limited — addressed. Good; that's actually the smart framing: front = noise-limited problem; side = clutter-limited problem. Also small note on ECCM: not requested; skip (one line max — maybe omit). Length check: I'll aim for something thorough but scannable. Let me write it. Actually, let me reconsider whether to include the full dB link-budget table with every term — yes, it makes the answer verifiable and high-value. Compact 7-row table. Also quick re-verify PRI stagger numbers: PRF 25/21.4/18 kHz (PRI 40/46.7/55.6 μs) → unambiguous velocity each: v_ua = λ·PRF/2: 25k→107 m/s; 21.4k→92; 18k→77. Staggered Chinese-remainder → extended ±~300+ m/s. Fine, don't over-detail. Beam positions for ±20° scan with 12° beam: step 8°: az: −16,−8,0,8,16 (5) × el 5 = 25 positions, 250 ms full search if needed (1 look each) or 1.25 s with 5 looks each. With cue ±2-3°: 2×2 or 3×3 → 4-9 looks: 1-2 s. ✓ Terminal CPIs: 25-40 ms at 25-40 Hz → per-look SNR at 2 km σ0.1: SNR = 19.5+10log0.1 +40log(5/2) + ... shorter CPI: E scales: 250 ms→30 ms: −9.2 dB; range gain 5→2 km: +15.9 dB; net +6.7 → 16.2 dB @0.1 → strong track. ✓ Angle noise per look 12/(1.6√(2·42)) = 0.92°... fine, filtered. Also: monopulse with 12° beam and ±20° scan: monopulse accuracy near boresight; airframe keeps target near boresight (autopilot). ✓ Warhead/fuze note ok. Now write. Tone: senior radar engineer's design note. Dense, organized, honest about margins. Units consistent. No fluff. Probably ~1600-1900 words with tables — acceptable given the depth requested. I'll try to compress prose, let tables carry data. # 60 mm Active Radar Seeker — C-UAS Interceptor vs. Shahed‑136 Class **Feasibility verdict:** Yes, this closes — with Ka‑band (35 GHz), a 54 mm limited‑scan monopulse array, and ~5 W average RF power. The two aspects are *different engineering problems*: nose‑on is noise‑limited (small RCS, needs energy/integration), side‑on is clutter‑limited (bigger RCS but crossing geometry puts the target near ground‑clutter Doppler). Both are solvable within the timeline (≥15 s between 5 km and impact at ~320 m/s closure). --- ## 1. Requirements → derived specs | Given | Engineering translation | |---|---| | 60 mm diameter | ≤54 mm aperture after wall/radome (aperture‑limited problem) | | 0–1000 km/h in 5 s | 278 m/s (M0.82), avg axial 5.7 g → rocket boost, **not** gun launch. Design env: 12–18 g peak axial, ~12 gᵣₘₛ 20–2000 Hz during burn, motor‑ignition shock | | Shahed‑136 detection @ 5 km | Target: composite delta‑wing OWA UAV, 35–50 m/s, ≤2 g lateral | | "Front and side" | Nose σ ≈ 0.02–0.3 m² (median ~0.1: metal engine/battery behind fiberglass skin); side σ ≈ 0.2–2 m². **These RCS values are assumptions — validate by measurement** (§10) | Pd ≥ 0.9, Pfa ≤ 1e‑6, clear air. All‑weather case discussed in §9. ## 2. Frequency selection — the controlling trade With a fixed aperture, SNR per watt goes as 1/λ² (gain‑limited aperture → higher frequency wins), but PA power, rain loss, and thermal margins go the other way. | | X 9.4 GHz | **Ka 35 GHz** | W 79–94 GHz | |---|---|---|---| | Gain, 54 mm aperture | 12 dBi | **23 dBi** | 30–32 dBi | | Beamwidth | 41° (can't home) | **~12°** | 5° | | Practical avg power (60 mm, 30 s) | 10 W | **5–8 W** | 1.5–2 W | | Net SNR @5 km, σ=0.1 m² | poor | **≈9.5 dB/look** | ≈11–12 dB/look | | Rain, 5 mm/h, two‑way 10 km | ~0.5 dB | **~7 dB** | 40–50+ dB (non‑viable) | | Radome | trivial | trivial (½‑wave) | tuned 0.9 mm wall | **Select 35 GHz.** It wins on *reliability* (rain robustness, thermal headroom, mature GaN), with adequate angular accuracy (0.9° monopulse error → <5 m at 3 km, filtered). W‑band buys +2–3 dB clear‑air and 2× angular precision but dies in rain and doubles the thermal problem — keep it as a growth variant (§9). ## 3. Architecture **Strapdown, fixed, 8‑subarray limited‑scan passive ESA with quadrant monopulse.** No gimbal (a 60 mm gimbal leaves only ~40–45 mm of aperture, −3 dB you cannot afford, plus bearings that must survive 20 g shock). Instead: - ~120 λ/2‑spaced patches (12×12 grid cropped to Ø54 mm), grouped into **8 subarrays** (~15 elements each), each with a 5‑bit phase shifter → **±20° electronic scan** (grating‑lobe free to ~±18–20°). - Monopulse comparator forms **Σ, Δaz, Δel** + small auxiliary guard patch. - **Pulsed, not FMCW**: on a shared aperture, FMCW needs >90 dB TX→RX isolation; pulsing with a T/R switch sidesteps it. Duty 25% preserves energy. - Fire‑control cue accuracy (~1–3 mrad) means the seeker needs only ±2–3° of search at handover — well inside ±20°. Terminal look angles stay within ±10° under PN. ``` Radome, tangent ogive ~150 mm, quartz A-sandwich Ø60 Array + 8× GaN PA (3.5 W ea) + LNAs, phase shifters ~25 mm Monopulse comparator / RF conditioning ~15 mm RFSoC (4× 12-bit 5 MS/s ADC) + FPGA + MEMS IMU ~35 mm Power conditioning, thermal battery interface ~15 mm Fuze electronics + inline S&A/ESLD ~20 mm → ~260 mm, ~0.55 kg ``` ## 4. Key parameters | Parameter | Value | |---|---| | Frequency / λ | 35 GHz / 8.57 mm | | Aperture / gain / beamwidth | 54 mm Ø / 22.9 dBi / ~12° (−25 dB sidelobe taper) | | TX | 25 W peak (8× GaN, 0.8 dB combining loss), 25% burst duty, 10% sustained → 5 W avg burst | | RX noise figure | 3.5 dB (pHEMT LNA at subarray port) | | Waveform | Pulsed‑LFM stretch: 10 µs / 500 MHz chirp, PRI 40 µs (PRF 25 kHz → 6 km unambiguous range); 3‑PRI stagger resolves Doppler ambiguity (±330 m/s closure) | | CPI | 250 ms acquisition; 25–40 ms terminal (25–40 Hz track updates) | | Processing | 2D FFT per CPI (range‑Doppler), 2D CFAR, noncoherent multi‑look + track‑before‑detect, EKF track, monopulse angle | | Mass / power / cost | 0.55 kg / 45 W peak electrical / seeker BOM ~$3–5k at volume | ## 5. Link budget at 5 km (per 250 ms look, E = 1.25 J) | Term | dB | |---|---| | Energy E = Pavg·tCPI | +0.97 dBW·s | | 2G | +45.8 | | 20log λ | −41.3 | | (4π)³ | −33.0 | | R⁴ (5 km) | −148.0 | | kT₀F (F = 3.5 dB) | +200.5 | | Losses (atm 1.2 + radome 1.0 + feed 1.3 + proc 2.0) | −5.5 | | **Constant** | **+19.5** | | **SNR = 19.5 + 10log σ** | **σ=0.01: 3.5 | 0.05: 6.5 | 0.1: 9.5 | 0.3: 14.5 | 1.0: 19.5 dB** | **Detection outcome (Pd 0.9, Pfa 1e‑6):** | Aspect | σ assumption | SNR/look | Time to Pd 0.9 @5 km | Verdict | |---|---|---|---|---| | Side | 0.2–2 m² | 14.5–22.5 dB | ≤0.5–1 s | **Comfortable; often acquires 6–7 km** | | Nose (median) | ~0.1 m² | 9.5 dB | ~2 s (Swerling II; looks decorrelate in ~0.2 s, so multi‑look integrates wing/engine flashes) | **Meets 5 km requirement** | | Nose (low end) | 0.05 m² | 6.5 dB | ~3 s → effective 4.3–4.5 km | Acceptable: acquisition occurs *during* flyout, no mission impact (≥15 s available) | | Nose (deep fade) | 0.01–0.02 m² | 3.5 dB | TBD over 4–6 s → 3.5–4 km | Handled by TBD + range closure; growth options §9 | ## 6. Aspect‑specific effects - **Nose:** the metal engine/battery behind the RF‑semi‑transparent fiberglass skin is the dominant contributor; expect strong flash statistics (the 1.8 m² delta wing speculates violently within ±0.1° of pitch). Fluctuations argue for 250 ms looks + multi‑look, which naturally harvest flashes. - **Side:** not energy‑limited — clutter‑limited. In low‑altitude crossing geometries the target Doppler can sit near the mainlobe clutter line. Mitigations: 4 Hz Doppler resolution, 2D CFAR, −25 dB sidelobes, altitude‑band engagement doctrine (intercept ≥300–500 m AGL), and the target's own ±40 m/s velocity offset. σ⁰ at 35 GHz and low grazing (−25 to −35 dB) is in your favor. - **Both:** Shahed's pusher prop (blade‑pass ~80–100 Hz, line spread ±30 kHz) provides an unambiguous JEM/micro‑Doppler signature — use for target confirmation, false‑alarm rejection, and fuze gating. ## 7. Surviving the boost (0→278 m/s in 5 s) - **Loads:** 5.7 g average; design to **30 g axial / 15 g lateral** (1.5× margin on ~12–18 g peak), 12 gᵣₙₛ random 20–2000 Hz for the burn, motor‑ignition shock ~1500 g at the case attenuated to <100 g/11 ms at electronics. - **Practices:** machined aluminum monolithic chassis doubles as thermal path and structure; array hard‑mounted (boresight stability <0.2°, with built‑in coupler self‑cal at power‑up); underfilled/corner‑bonded BGA or QFP packages; ≤1210 flexible‑termination MLCCs; staked tall discretes; SMPM blind‑mate interconnects with radial float; all‑planar RF (no waveguide flanges to crack); ablative aft bulkhead if tube launch is hot. - **Thermal:** ~21 W average dissipation during acquisition bursts, ≤40–60 s engagement → heat capacity of the chassis + airframe absorbs it with a few °C rise; junction rise handled by θjc. Duty scheduling (25% burst / 10% sustained) covers the worst long‑shot timeline. - **Aero:** M0.82, q = 47 kPa, stagnation ~55 °C — benign; rain‑erosion coat on the radome is a formality. - **Operational note:** run RX/IMU modes through burn, but schedule acquisition after motor burnout (burn vibration raises the phase‑noise floor). ## 8. Guidance, fuzing - Handover from fire‑control (any C‑UAS X/Ku tracker sees a Shahed trivially) with cue quality gate; spiral/ladder search over ±20° (25 beams max, ~1.5 s; 1–9 beams with a good cue). - Strapdown PN, autopilot holds target within ±3° of boresight; ±20° e‑scan absorbs lead‑angle drift; airframe needs 15–25 g lateral authority (canard config) — endgame LOS rates stay low under PN until the last ~100 m, which e‑scan covers instantly. - **Integrated RF fuze:** below ~200 m, switch to 1 GHz bandwidth (10–15 cm resolution profiles); command burst at 2–5 m miss, gated by monopulse POA; graze/contact backup. With a ~0.5–0.8 kg HE‑frag payload, kill criterion is wing‑spar/control‑link/engine destruction — single‑shot Pk ~0.75–0.9; salvo doctrine closes the rest. ## 9. Weather honesty and growth paths - In 5 mm/h rain, Ka loses ~7 dB round‑trip at 5 km → nose detection degrades to ~3.5–4 km; side remains ≥5 km. The **W‑band variant is not viable in rain**; X‑band can't home from a 54 mm aperture. If all‑weather nose‑on @5 km is contractual, options: (a) raise Pavg to 8–10 W (GaN 35 GHz roadmap; +3 dB, thermal still fine with duty scheduling), (b) semi‑active illumination mode using a ground Ku illuminator for the acquire phase (ground EIRP makes 5 km trivially reliable; seeker goes receive‑only until handover), (c) 94 GHz high‑power variant for clear‑weather ops. - Rain in the engagement timeline is also a fire‑control problem: launch cued by ground radar that sees through rain anyway; the seeker's job is terminal confirmation and homing, which it retains to 3–4 km. ## 10. Assumptions to validate (do these before PDR) 1. **RCS measurement program**: instrumented‑range (or pole) measurement of a representative article at 35 GHz, nose ±30° and side ±30°, 100 MHz+ bandwidth — this single dataset sets your true Pd-vs-range curves and decides whether the 8–10 W power growth is needed. 2. Clutter trial at 1–3° grazing over representative terrain, crossing geometry. 3. Full 12 gᵣₙₙᵢₛ + shock qualification of the array/MMIC stack; boresight stability through vibe. 4. JEM signature capture (prop blade‑rate lines) for the classifier/fuze. **Bottom line:** Ø60 mm × ~260 mm, 0.55 kg, 35 GHz, 25 W peak/5 W avg, 12° beam, ±20° e‑scan strapdown monopulse — side‑aspect 5 km detection with margin, nose‑aspect 5 km detection with ≤2 s integration at σ≈0.1 m², robust to the launch environment with standard ruggedization, at a seeker cost compatible with exchange‑ratio economics against a $20–50 k drone.

Rad Seek