1. Introduction and Learning Objectives
"How far can this radar see?" is the question every radar engineer — and every customer — asks first. The honest answer is: it depends on the target, the environment, and the required probability of detection. Radar performance quantifies the capability of a radar system to fulfil its mission: detecting targets at range, measuring them accurately, tracking them reliably, and resisting interference.
These notes build the performance framework of radar engineering from the ground up: the radar range equation, receiver noise, detection statistics, system losses, search and track budgets, propagation effects, and electronic-counter-countermeasures (ECCM).
Learning Objectives
- Derive the radar range equation from first principles and manipulate it for prediction, detection-range, and required-power problems.
- Compute receiver noise power, noise figure, and system noise temperature.
- Explain the trade-off between probability of detection Pd and probability of false alarm Pfa.
- Apply the Albersheim and Marcum/Swerling detection curves and understand non-coherent integration of pulses.
- Account for the major system losses (beam-shape, scan, atmospheric, collapse, CFAR, processing).
- Evaluate search-radar performance (cumulative detection, scan rate) and tracking accuracy (monopulse, track-while-scan).
- Analyse the effects of the atmosphere, multipath, and interference on radar performance.
- Summarise ECCM techniques and their performance costs/benefits.
Why it matters: A radar that is "powerful" on paper can still fail in the field if the engineer ignored a 3 dB atmospheric loss at the horizon or a beam-shape loss during a fast scan. Performance analysis is where electromagnetics, signal processing, and systems engineering meet.
2. The Radar Range Equation
2.1 Derivation from First Principles
Consider a radar transmitting peak power Pt through an antenna of gain Gt. The power density at range R is
S = PtGt / (4πR²)
A target of radar cross-section σ (m²) intercepts and re-radiates a fraction of this energy. The target's echo power density back at the radar is
Secho = (PtGtσ) / (4πR²)² · (1/(4π))
The receiving antenna, with effective aperture Ae = Grλ²/(4π), collects this energy. Assembling everything (and noting Gt = Gr = G for a monostatic radar with one antenna):
Radar Range Equation (monostatic):
Pr = Pt G² λ² σ / [(4π)³ R⁴ L]
where L lumps all system losses (Section 6). The R⁴ dependence is fundamental: doubling the detection range requires 16× (12 dB) more received power.
2.2 Forms of the Range Equation
| Form | Equation | Use |
|---|
| Received power | Pr = PtG²λ²σ/[(4π)³R⁴L] | Link-budget analysis |
| Maximum range (pulse radar) | Rmax = [PavG²λ²σtsn/((4π)³kTs(S/N)reqL)]1/4 | System design & prediction (energy form) |
| Energy form | Pr/N = Pavtsn · G²λ²σ / [(4π)³R⁴kTsL] | Shows detection needs average energy × integration |
with Pav = Pt·duty cycle, ts = time-on-target per scan, n = number of hits integrated, k = Boltzmann's constant, Ts = system noise temperature.
Worked Example 2.1 — A radar transmits Pt = 100 kW, G = 35 dB, λ = 0.1 m. Target σ = 5 m² at R = 80 km, total loss L = 6 dB. Find Pr.
Linear values: G = 103.5 = 3162; L = 3.98.
Pr = (10⁵)(3162)²(0.1)²(5) / [(4π)³(8×10⁴)⁴(3.98)]
Numerator: 10⁵ × 10⁷·⁹⁹ × 0.01 × 5 ≈ 10⁵ × 10⁷·⁹⁹ × 0.05
(4π)³ ≈ 1984; (8×10⁴)⁴ = 4.096×10¹⁹.
Pr ≈ (10⁵ × 10⁷·⁹⁹ × 0.05)/(1984 × 4.096×10¹⁹ × 3.98) ≈ 1.0×10−12 W = 1 pW — typical! Radar echoes are astonishingly weak.
Worked Example 2.2 — Same radar: if the receiver needs (S/N)req = 13 dB and the noise level is −104 dBm, what is the single-pulse maximum range for this target?
Required Pr,min = −104 dBm + 13 dB = −91 dBm = 7.94×10−13 W.
Rmax = [PtG²λ²σ/((4π)³Pr,minL)]1/4
= [(10⁵)(10⁷·⁹⁹)(0.01)(5)/(1984 × 7.94×10−13 × 3.98)]1/4
≈ [7.96×10¹¹]1/4 ≈ 168 km.
2.3 The R⁴ Law and Its Consequences
- Range grows only as the fourth root of any improvement: 4× more power → only 1.41× more range.
- Antenna gain enters as G² (or Ae²) — aperture is the most "efficient" lever.
- Large targets (σ ↑ 10×) extend range by only 101/4 = 1.78×.
R⁴
Range dependence — the radar's tyranny
PtG²λ²σ
Numerator levers (power, aperture, wavelength, target)
(4π)³
Two-way spherical spreading constant
1 pW
Typical order of a received echo
3. Receiver Noise and System Noise Temperature
3.1 Thermal Noise
Every receiver adds noise. The available thermal noise power in bandwidth B is
N = k T0 B (k = 1.38×10−23 J/K, T0 = 290 K)
At T0 = 290 K: N ≈ −114 dBm per MHz of bandwidth — the absolute floor set by physics.
3.2 Noise Figure and Effective Noise Temperature
F = (S/N)in / (S/N)out = 1 + Te/T0
Effective input noise temperature Te = (F − 1)T0. System noise temperature:
Ts = Ta + Te = Ta + (F − 1)T0
where Ta is the antenna noise temperature (sky/ground brightness seen by the antenna: ~30–50 K pointing at cold sky, 290 K pointing at warm ground).
3.3 Cascaded Stages (Friis Formula)
Ftotal = F₁ + (F₂ − 1)/G₁ + (F₃ − 1)/(G₁G₂) + …
Design rule: The first stage dominates. This is why a low-noise amplifier (LNA) with F ≈ 1 dB is placed right at the antenna feed — losses before the LNA (cable, duplexer) add dB-for-dB to the system noise figure.
Worked Example 3.1 — Receiver chain: antenna + cable loss 1.5 dB (F₁ = 1.41, G₁ = 0.71), LNA F₂ = 2 dB (1.58) with G₂ = 25 dB (316), mixer/IF F₃ = 10 dB (10).
Ftotal = 1.41 + (1.58−1)/0.71 + (10−1)/(0.71×316)
= 1.41 + 0.82 + 0.04 = 2.27 → 3.6 dB.
The cable dominates! Moving the LNA ahead of the cable would give ≈ 1.6 dB.
4. Probability of Detection and False Alarm
4.1 The Detection Problem
The radar decides each range-Doppler cell: target present (H₁) or noise only (H₀). Because noise is random, two error types exist:
| Decision / Truth | Noise only (H₀) | Target present (H₁) |
|---|
| Declare "no target" | Correct | Miss — probability 1 − Pd |
| Declare "target" | False alarm — probability Pfa | Correct detection — Pd |
4.2 The Threshold Trade-off
A single threshold T is set on the envelope detector output:
Pfa = exp(−T²/2σ²) (noise-only, Rayleigh envelope, linear detector)
Raising the threshold ↓ Pfa but also ↓ Pd. You cannot have both — this trade-off is the core of radar detection theory.
Worked Example 4.1 — A radar processes N = 10⁶ range-Doppler cells per second. The operator tolerates one false alarm per second on average. Find Pfa.
Pfa = 1/N = 10⁻⁶.
Note: this is per-cell. Practical radars achieve this with CFAR (Section 6) rather than a fixed threshold.
4.3 Swerling Target Models
| Model | RCS behaviour | Typical target |
|---|
| Swerling I | Constant during a scan, fluctuates scan-to-scan (exponential PDF) | Complex aircraft, viewed long enough between scans |
| Swerling II | Fluctuates pulse-to-pulse | Aircraft with rapid aspect changes |
| Swerling III / IV | As I / II but chi-square PDF with 4 degrees of freedom | Targets with one dominant scatterer + smaller ones |
| Marcum (0) | Non-fluctuating | Spheres, corner reflectors, calibration targets |
Fluctuation degrades single-pulse detection but helps with integration diversity — a target that dips on one pulse may peak on the next.
5. Detection Theory and Integration
5.1 Single-Pulse Detection (Non-Fluctuating Target)
Pd = ∫T∞ p(r|H₁) dr, with required single-pulse SNR (Marcum case):
for Pfa = 10⁻⁶, Pd = 0.9 → single-pulse SNR ≈ 13.2 dB (non-fluctuating, no integration)
5.2 Albersheim's Approximation
A closed-form approximation avoiding Marcum Q-function tables (valid for non-coherent integration, n pulses):
SNR₁(dB) = −5 log₁₀n + (6.2 + 4.54/√(n+0.44)) · log₁₀(A + 0.12AB + 1.7B)
A = ln(0.62/Pfa), B = ln[Pd/(1 − Pd)]
Worked Example 5.1 — Required single-pulse SNR for Pd = 0.9, Pfa = 10⁻⁶ with n = 10 non-coherently integrated pulses.
A = ln(0.62/10⁻⁶) = 13.34; B = ln(9) = 2.197.
SNR₁ = −5(1) + (6.2 + 4.54/√10.44) · log₁₀(13.34 + 0.12×13.34×2.197 + 1.7×2.197)
= −5 + 7.61 · log₁₀(19.8) = −5 + 7.61×1.297 ≈ 4.9 dB per pulse.
Integration buys ≈ 8 dB: ten 4.9 dB pulses do the job of one 13.2 dB pulse.
5.3 Coherent vs Non-Coherent Integration
| Aspect | Coherent (predetection) | Non-coherent (post-detection) |
|---|
| What is summed | Complex I/Q samples (phase aligned) | Detected envelopes/magnitudes |
| SNR gain with n pulses | n (full) | ≈ √n–n (loss of ~1–2 dB at small n) |
| Needs | Phase coherence, Doppler alignment across pulses | Only amplitude; trivial to implement |
| Loss vs ideal | None | Integration loss Li(n) |
5.4 Cumulative Detection on Scan
If a target is illuminated on m successive scans with per-scan Pd = p, the probability of detection at least once ("cumulative" Pd) is
Pd,cum = 1 − (1 − p)m
This is why search radars can accept a modest per-scan Pd: the operator cares that the target is seen regularly, not every scan.
6. System Losses
Every real radar underperforms the textbook range equation. Losses L (linear, >1) enter the denominator. Knowing them is the difference between a paper radar and a working one.
| Loss | Physical cause | Typical value |
|---|
| Beam-shape loss | Target not always at beam peak during scan | 1–3 dB |
| Scanning loss | Beam motion during the pulse/echo | 0.5–1.5 dB |
| Atmospheric absorption | O₂/H₂O molecular absorption | 0.01–0.1 dB/km (S-band, weather-dependent) |
| Duplexer/circulator & feed | RF plumbing between antenna and receiver | 1–2 dB |
| Integration loss | Non-coherent combining inefficiency | 1–2 dB |
| CFAR loss | Adaptive threshold raises effective threshold | 1–3 dB |
| Filter mismatch loss | Received pulse not perfectly matched to receiver filter | 0.5–1 dB |
| Collapse loss | Range/Doppler folding compresses the display | 1–2 dB |
| Quantization loss | Finite ADC bits, phase-shifter quantization | 0.5–1 dB |
| Propagation (multipath) loss | Nulls from ground reflection at low elevation | Can exceed 20 dB at a null |
Field lesson: Losses add in dB. A radar with 6 dB of ignored losses has half the detection range of its paper version (range ∝ (1/L)1/4 → 6 dB → ×0.71). Discipline in the loss budget is the mark of a professional radar design.
7. Search-Radar and Tracking Performance
7.1 Search Performance
A search radar must cover a solid angle Ω (steradians) with beamwidth ΩB ≈ θazθel per revisit time ts. The number of beam positions per scan:
nbeam = Ω/ΩB ; frame time tframe = nbeam · tdwell
Hits per dwell: n = tdwell/Tr = tdwell·fr. Design tension: more hits per dwell → better Pd (integration), but fewer beam positions → longer frame time → slower target update. The search radar design equation balances these against Rmax⁴ via the energy form of the range equation.
7.2 Tracking Accuracy
Tracking quality is measured by the rms error σ of range, angle, and Doppler estimates:
σR ≈ c/(2B√(2·SNR)) (range, B = waveform bandwidth)
σθ ≈ θB/(km√(2·SNR)) (angle, km ≈ 1.6–1.8 for monopulse)
σf ≈ 1/(T√(2·SNR)) (Doppler/velocity, T = observation time)
All errors shrink with 1/√SNR — every 6 dB of SNR halves the theoretical error bound.
7.3 Tracking Architectures
| Architecture | Principle | Strength |
|---|
| Single-target tracker (STT) | Dedicated servo locks one beam on one target | Highest accuracy & data rate |
| Track-while-scan (TWS) | Search scan updates many filters (e.g., Kalman/α–β) in parallel | Many tracks, no dedicated dwell |
| Phased-array track | Adaptive beam scheduling: revisit as needed by track quality | Optimal time–energy allocation |
Worked Example 7.1 — Monopulse radar, 3 dB beamwidth θB = 2°, SNR = 20 dB. Find σθ.
σθ ≈ 2°/(1.7√(2×100)) = 2/(1.7×14.14) ≈ 0.083° ≈ 5 mrad.
8. Propagation Effects on Performance
8.1 Standard Atmosphere and Refraction
The standard atmosphere's decreasing refractive index bends rays downward, effectively extending radar horizon beyond the geometric one:
Rhor ≈ 4.12(√ht + √hr) km (h in metres, standard 4/3 Earth-radius model)
8.2 Multipath and Fading
Ground-reflected signals interfere with the direct ray. At low grazing angles the two paths are nearly antiphase → deep nulls (up to 20–30 dB fade), making low-altitude targets intermittently invisible. The interference region extends to elevation ≈ 0.5θB above the horizon — a classic performance cliff for air-surveillance radars.
8.3 Attenuation and Weather
- Gases: O₂ resonance near 60 GHz; H₂O near 22 and 183 GHz. Below 10 GHz, <0.01 dB/km.
- Rain: attenuation ∝ f² (roughly) above ~3 GHz; heavy rain at X-band can cost several dB per km — why weather radars prefer S/C-band for long range and X-band for fine resolution at short range.
- Ducting: strong gradients trap rays, creating anomalous long-range detections and clutter beyond the horizon.
9. Figures of Merit for Radar Performance
| Metric | Definition | Typical design target |
|---|
| Instrumented range | Longest range the receiver is gated to process | Mission set |
| Detection range R50/R90 | Range at which Pd = 50%/90% for the design target | Spec-driven |
| False-alarm time Tfa | 1/(Pfa·Ncells·rate) — mean time between false alarms | > hours |
| Track accuracy | σR, σθ, σv of the maintained track | Mission set |
| Coverage | Volume scanned per unit time at full performance | Ω/tframe |
| Availability / MTBF | Operational uptime, mean time between failures | Military: > 99% |
10. Electronic Counter-Countermeasures (ECCM)
Radar performance in contested environments is preserved by ECCM design:
| Threat | ECCM technique | Performance cost |
|---|
| Noise jamming | Frequency agility, burn-through mode (high ERP), sidelobe blanking/cancellation | Reduced dwell per frequency |
| Deception (range gate steal) | Pulse-to-pulse agility, leading-edge tracking, guard gates | Processing complexity |
| Chaff | MTI/Doppler discrimination (covered in the MTI unit), polarization diversity | Already budgeted |
| Anti-radiation missiles | Low-sidelobe antennas, random emission control (EMCON) | Reduced own coverage |
| Sidelobe jamming | Sidelobe cancellers (SLC), adaptive arrays | Degrees of freedom spent |
The jamming equation replaces the thermal noise floor with the jammer spectral density: (S/J) = PtGtσRj²Bj / (4πPjGjR⁴L) — a self-screening jammer effectively imposes its own "range equation" the radar must beat.
11. Chapter Summary — Key Equations and Facts
R⁴
Range law of the radar equation
kT₀B
Thermal noise floor (−114 dBm/MHz)
13.2 dB
Single-pulse SNR for Pd=0.9, Pfa=10⁻⁶ (non-fluct.)
1/√SNR
Accuracy improvement law
F₁ + (F₂−1)/G₁ + …
Friis cascade formula
1 − (1−p)^m
Cumulative detection over m scans
4.12√h km
Radar horizon (standard atmosphere)
L budget
6 dB of ignored loss = 29% range loss
- The radar range equation is the master tool: predict range, required power, or required SNR — always in the energy (average-power) form for system design.
- Detection is a statistical decision: Pd vs Pfa trade-off governed by threshold setting and target fluctuation model (Marcum/Swerling).
- Integration (coherent or non-coherent) recovers per-pulse SNR at the cost of dwell time; Albersheim gives quick answers.
- Losses are real: budget every dB (beam-shape, scan, atmospheric, CFAR, mismatch).
- Accuracy scales as 1/√SNR; search performance trades coverage against dwell; propagation (multipath nulls, rain, refraction) can dominate the link at low elevation and high frequency.
12. Self-Test Questions
- Derive the radar range equation, stating every assumption (monostatic, matched polarization, point target, free space).
- A radar's peak power is quadrupled. By what factor does maximum range change? What change would double the range?
- Compute the noise power in dBm for a receiver with B = 2 MHz and F = 4 dB at T₀ = 290 K.
- Using Friis' formula, show why an LNA must precede a long cable run, quantifying with your own example.
- For Pd = 0.95, Pfa = 10⁻⁵, use Albersheim to find the per-pulse SNR with n = 5 integrated pulses.
- A target is detected with Pd = 0.6 per scan. What is the probability it is seen at least once in 4 scans? At least once in 10?
- List six system losses and estimate a total loss budget for an S-band air-surveillance radar.
- A monopulse radar has θB = 1.5° and SNR = 16 dB. Compute the theoretical angle accuracy. What SNR would halve it?
- Explain the low-elevation multipath null and why it is worst for tall targets at long range… and why it matters most for detecting low-flying aircraft.
- State the radar horizon formula and compute the horizon for an antenna at 30 m looking at a target at 10 000 m altitude.
- Explain how a self-screening noise jammer modifies the range equation, and define "burn-through range".
- Compare coherent and non-coherent integration gain for n = 20 pulses, including the integration loss.
Show brief answers (1–6)
- Spreading 1/(4πR²) each way × target σ × aperture λ²G/4π; assume isotropic-equivalent gains, matched pol., single point scatterer, no multipath/atmosphere.
- Range ×41/4 = 1.41; doubling range needs 16× power (12 dB) — or 4× aperture (G²) or 16× σ.
- N = kT₀BF = 1.38e−23×290×2e6×2.51 = 2.0e−14 W ≈ −107 dBm.
- First-stage loss adds directly: e.g. 2 dB cable before an LNA costs 2 dB in F; after the LNA it costs 2/316 ≈ 0.006 dB.
- A = ln(0.62/1e−5) = 11.13; B = ln(19) = 2.944; SNR₁ ≈ −5log5 + (6.2+4.54/√5.44)·log₁₀(11.13+0.12·11.13·2.944+1.7·2.944) ≈ −3.5 + 8.15·log₁₀(17.3) ≈ 6.4 dB.
- 1−(0.4)⁴ = 0.974; 1−(0.4)¹⁰ ≈ 0.9999.
Show brief answers (7–12)
- Typical budget: beam-shape 1.6 + scan 0.7 + feed/duplexer 1.5 + atmospheric 0.5 + CFAR 1.5 + mismatch 0.7 + integration 1.0 ≈ 7.5 dB.
- σ_θ = 1.5/(1.7√(2·39.8)) = 1.5/15.2 ≈ 0.099° ≈ 1.73 mrad; +6 dB SNR (4× linear) halves σ.
- Direct + ground-reflected rays interfere; near grazing incidence path difference ≈ 2h_t h_r/R → phase opposition → cancellation up to 20–30 dB; low-flying targets dwell in this region their whole approach.
- R_h ≈ 4.12(√30 + √10000) = 4.12(5.48+100) ≈ 435 km.
- Jammer noise replaces kT₀B with P_jG_j/R²B_j; burn-through = range where S/J = required (S/J)_req, i.e. where radar power wins over jammer ERP.
- Coherent: 13 dB (10log20); non-coherent: ≈ 13 − 1.5 ≈ 11.5 dB effective.
References for Further Reading
- M. I. Skolnik, Introduction to Radar Systems, 3rd ed., McGraw-Hill — range equation, noise, detection chapters.
- M. I. Skolnik (ed.), Radar Handbook, 3rd ed., McGraw-Hill — loss budgets, propagation, ECCM.
- N. Levanon, Radar Principles, Wiley-Interscience — detection theory and waveforms.
- B. R. Mahafza, Radar Systems Analysis and Design Using MATLAB, CRC Press — worked range-equation examples.
- D. K. Barton, Radar Equations for Modern Radar, Artech House — the definitive loss-budget treatment.
- Note: Swerling/Maccum numerical thresholds vary slightly between references; always state Pd, Pfa, fluctuation model, and integration count when quoting a required SNR.