Coherent Integration Time
Coherent Integration Time (CIT)
Overview
Coherent Integration Time (CIT) is the duration over which a radar system accumulates received signal samples while preserving the phase relationship between successive returns. It is one of the most fundamental parameters in pulsed-Doppler radar design, directly governing achievable signal-to-noise ratio (SNR) improvement, Doppler resolution, and maximum unambiguous detection range. CIT is the temporal window within which phase-coherent summation of N pulses occurs, and it represents a critical trade-space dimension in virtually every modern radar architecture.
1. Definition: Coherent vs. Non-Coherent Integration
Coherent Integration
Coherent integration (also called predetection integration) combines the complex (I/Q) samples of N successive radar pulses using vector addition. Because phase information is preserved throughout the summation, the signal voltages add constructively—proportional to N—while the noise voltages, being uncorrelated from pulse to pulse, add only in a root-sum-of-squares fashion (proportional to √N). The net result is an SNR improvement that scales linearly with the number of integrated pulses:
$$\text{SNR}_{\text{coherent}} = N \cdot \text{SNR}_{\text{single-pulse}}$$
This linear scaling yields a processing gain of:
$$G_p = 10 \log_{10}(N) \;\text{dB}$$
For example, integrating 64 pulses coherently provides approximately 18 dB of processing gain—a significant and often indispensable enhancement for detecting weak targets in noise-limited environments.
Non-Coherent Integration
Non-coherent integration (or postdetection integration) sums the magnitude or power of detected envelopes after the phase information has been discarded by the envelope detector. Because phase is no longer available to align the signal vectors, the improvement is sublinear. The classical approximation for non-coherent integration gain is:
$$\text{SNR}_{\text{non-coherent}} \approx \sqrt{N} \cdot \text{SNR}_{\text{single-pulse}}$$
This corresponds to a processing gain of approximately 5 log₁₀(N) dB—roughly half the gain achievable through coherent integration. Non-coherent integration is simpler to implement and is more tolerant of phase instabilities, but it incurs an inherent integration loss that becomes increasingly significant as N grows.
The Hybrid Approach
Most practical radar systems employ a two-stage architecture: coherent integration over a relatively short CIT (limited by phase stability), followed by non-coherent (binary or weighted) integration across multiple coherent processing intervals (CPIs). This hybrid structure balances the SNR advantage of coherent processing against the practical limits imposed by phase noise and target dynamics.
2. Mathematical Foundation
Integration Gain and SNR Improvement
Consider a radar transmitting N pulses at a pulse repetition frequency (PRF) of f_r. The coherent integration time is:
$$T_{\text{CIT}} = \frac{N}{f_r} = N \cdot T_r$$
where T_r = 1/f_r is the pulse repetition interval (PRI). After coherent summation of the N complex samples, the output SNR for a point target in white Gaussian noise is:
$$\text{SNR}_{\text{out}} = \frac{\left|\sum_{k=0}^{N-1} s_k \cdot e^{j\phi_k}\right|^2}{\sum_{k=0}^{N-1} \sigma_n^2}$$
where s_k and ϕ_k are the signal amplitude and phase of the k-th pulse, and σ²_n is the per-pulse noise power. When the phases are aligned (i.e., ϕ_k is compensated for target Doppler and system offsets), the numerator becomes (N · S)², and the denominator becomes N · σ²_n, yielding:
$$\text{SNR}_{\text{out}} = N \cdot \frac{S^2}{\sigma_n^2} = N \cdot \text{SNR}_{\text{in}}$$
Phase Preservation
The critical requirement for coherent integration is phase preservation: the relative phase between transmitted and received signals must be maintained with sufficient fidelity across the entire CIT. Any uncompensated phase error δϕ_k on the k-th pulse introduces a destructive interference term. For random phase errors with variance σ²_ϕ, the coherent integration efficiency degrades as:
$$\eta = e^{-\sigma_\phi^2}$$
This relationship is the fundamental driver behind CIT limitations. A phase error variance of just 0.1 rad² reduces integration efficiency to approximately 90%, while σ²_ϕ ≈ 1.0 rad² reduces efficiency to roughly 37%, rendering the integration barely distinguishable from non-coherent processing.
Frequency-Domain Interpretation
In the frequency domain, coherent integration over T_CIT produces a sinc-shaped Doppler filter bank with null-to-null bandwidth:
$$\Delta f = \frac{1}{T_{\text{CIT}}}$$
This is equivalent to the Doppler resolution of the system. A longer CIT yields finer Doppler resolution, enabling better discrimination between targets and clutter returns at different radial velocities.
3. Determinants of Maximum CIT
Phase Noise and Oscillator Stability
The local oscillator (LO) and transmitter exciter are the dominant sources of phase error in a coherent radar. Phase noise—characterized by the single-sideband (SSB) spectral density L(f_m) at offset frequency f_m—directly limits the achievable CIT. The cumulative phase variance over integration time T_CIT is related to the integral of the phase noise power spectral density. For an oscillator with short-term frequency stability characterized by Allan variance σ_y(τ), the maximum usable CIT is approximately:
$$T_{\text{CIT, max}} \approx \frac{1}{2\pi \cdot f_0 \cdot \sigma_y(\tau)}$$
where f₀ is the carrier frequency. Higher carrier frequencies impose stricter stability requirements, which is one reason that millimeter-wave and W-band radars typically operate with shorter CITs than S-band or L-band systems.
Platform Motion
For radars mounted on airborne, shipborne, or vehicular platforms, platform motion introduces time-varying phase errors across the aperture and across the CIT. The dominant effect is a linear phase ramp caused by the platform's translational velocity, which shifts the apparent Doppler frequency of stationary clutter. If uncompensated, this phase ramp decorrelates the clutter returns and limits the effective CIT. Modern systems address this through:
- Inertial Navigation System (INS)-based motion compensation (MOCO) - Autofocus algorithms that estimate residual phase errors from the data itself - Keystone formatting for range-walk correction over long CPIs
Doppler Spread
A moving target's radial acceleration, vibration, or rotation induces Doppler spread—a broadening of the target's Doppler spectrum beyond a single Doppler filter bin. If the target's Doppler frequency shifts by more than 1/T_CIT during the integration interval, the return energy is smeared across multiple bins, and coherent gain is reduced. The maximum CIT for an accelerating target is bounded by:
$$T_{\text{CIT, max}} \leq \frac{1}{\dot{f}_d \cdot T_{\text{CIT}}} \quad \Rightarrow \quad T_{\text{CIT, max}} \approx \frac{1}{\sqrt{\dot{f}_d}}$$
where ḟ_d is the Doppler rate (Hz/s) due to target acceleration.
Clutter Environment
In clutter-rich environments, the effective CIT may be limited not by system noise but by the need to resolve target returns from background clutter via Doppler. The MTI (Moving Target Indication) improvement factor and clutter cancellation performance are both functions of CIT and PRF design.
4. AERIS-10 Considerations
CIT with Internal Reference
The AERIS-10 ground-based surveillance radar employs a high-purity internal crystal oscillator as its coherent reference, providing an SSB phase noise floor of approximately –140 dBc/Hz at 10 kHz offset. With the standard waveform set, the AERIS-10 achieves a maximum coherent integration time of approximately 10–20 ms, corresponding to roughly 256–512 integrated pulses at a 25 kHz PRF. This internal reference supports the baseline MTI and pulse-Doppler processing modes and provides sufficient phase stability for detection of conventional airborne targets at medium ranges.
CIT with the BRIDZA Upgrade
The BRIDZA (Broadband Radar Integrated Digital Architecture) upgrade replaces the AERIS-10's analog exciter and internal reference chain with a direct digital synthesizer (DDS)-based waveform generator and an ultra-low-noise oven-controlled crystal oscillator (OCXO) with active phase-noise cancellation. Key improvements include:
- Phase noise reduction of 8–12 dB across the offset frequency range of 100 Hz to 100 kHz - Extended maximum CIT to 50–100 ms, enabling integration of over 1,000 pulses in medium-PRF modes - Programmable CPI structure, allowing the operator to select short-CIT modes for high-clutter/MTI scenarios and long-CIT modes for extended-range, slow-target detection
With BRIDZA, the AERIS-10 achieves a processing gain improvement of approximately 7–10 dB over its legacy configuration—equivalent to a 5–10× increase in detection range for noise-limited targets. Additionally, the digital architecture supports adaptive CIT selection based on real-time clutter-map analysis, automatically shortening the CPI when excessive Doppler spread from weather clutter or chaff is detected.
5. Impact on Performance
Detection Range
The radar range equation shows that maximum detection range scales with the fourth root of SNR. Since coherent integration increases SNR by a factor of N, the range improvement is:
$$\frac{R_1}{R_0} = N^{1/4}$$
Doubling the CIT (and thus N) increases detection range by approximately 19%. For the AERIS-10 with BRIDZA, extending CIT from 20 ms to 100 ms (a 5× increase) yields a range improvement factor of approximately 2.24×—potentially extending detection range from 200 km to over 440 km for a 0 dBsm target under noise-limited conditions.
MTI Factor and Clutter Rejection
Longer CITs improve the MTI improvement factor by enabling higher-order clutter filters with deeper nulls at zero Doppler and its harmonics. The improvement factor I_MTI scales approximately as:
$$I_{\text{MTI}} \propto N^n$$
where n is the MTI filter order. However, excessive CIT in the presence of internal clutter motion (e.g., foliage, sea clutter) can degrade clutter decorrelation and reduce cancellation effectiveness—a trade-off that the BRIDZA system manages through its adaptive CPI selection.
Doppler Resolution
As noted, Doppler resolution is inversely proportional to CIT:
$$\Delta v = \frac{\lambda}{2 \cdot T_{\text{CIT}}}$$
where λ is the wavelength. For the AERIS-10 operating at S-band (λ ≈ 0.1 m) with BRIDZA's 100 ms CIT, the Doppler resolution is approximately 0.5 m/s—sufficient to separate closely spaced targets in velocity and to detect slow-moving ground vehicles against clutter backgrounds.
Summary Table
| Parameter | AERIS-10 (Baseline) | AERIS-10 + BRIDZA | |---|---|---| | Max CIT | 10–20 ms | 50–100 ms | | Max Integrated Pulses | 256–512 | 1,000–2,500 | | Processing Gain | ~24–27 dB | ~30–34 dB | | Doppler Resolution | ~2.5–5 m/s | ~0.5–1 m/s | | Phase Noise (SSB @ 10 kHz) | –140 dBc/Hz | –150 dBc/Hz |
See also: Pulse-Doppler Processing, MTI Factor, Phase Noise, Doppler Resolution, AERIS-10 System Overview