Signal Processing

Coherent Integration

Q: What is the difference between coherent and non-coherent integration?

Coherent integration sums complex (I/Q) pulse returns preserving both amplitude and phase, giving an SNR gain of N. Non-coherent integration sums magnitude or magnitude-squared values, discarding phase information, and provides a gain of approximately N^0.5 to N^0.8 depending on the initial single-pulse SNR. Coherent integration requires a stable local oscillator and knowledge of target Doppler, while non-coherent integration works without phase reference. For 100 pulses, coherent integration gains 20 dB versus roughly 10 to 16 dB for non-coherent methods.

Q: What limits the coherent processing interval in practice?

The CPI is limited by three factors: target coherence time, oscillator stability, and dwell time allocation. A target accelerating at a m/s squared decorrelates after approximately sqrt(lambda / a) seconds, where lambda is wavelength. At X-band (10 GHz) with 10 m/s squared acceleration, the coherence limit is about 55 ms. Oscillator phase noise must remain below roughly 1 radian RMS across the CPI. In multi-function radars, timeline scheduling constrains dwell time to 1 to 50 ms per beam position, further limiting integration length.

/koh-heer-ent in-teh-gray-shun/

A radar signal processing technique that sums N received pulses while preserving their amplitude and phase information, producing a signal-to-noise ratio improvement of exactly N (linear) or 10 log₁₀(N) dB. The pulses must be collected within a coherent processing interval (CPI) during which the transmitter, receiver, and target maintain phase stability. Coherent integration is the foundation of pulse-Doppler processing, synthetic aperture radar (SAR) image formation, and moving target indication (MTI), and is typically implemented via FFT-based Doppler filter banks operating on complex I/Q samples.

Category: Signal Processing

SNR Gain: 10 log₁₀(N) dB

Typical CPI: 1 to 100 ms

Understanding Coherent Integration

Coherent integration exploits the deterministic phase progression of a target's echo across successive pulses. When a target moves at constant radial velocity v_r, each returned pulse accumulates an additional phase shift of 2πf_dT_r, where f_d = 2v_r/λ is the Doppler frequency and T_r is the pulse repetition interval. By applying a matched Doppler filter (or equivalently computing the DFT across slow-time samples), the processor aligns all N pulse returns in phase before summation, causing the signal to add coherently (proportional to N) while noise, being random, adds in power (proportional to N). The resulting SNR improvement is N, the maximum achievable for any linear integration scheme.

In modern pulse-Doppler radars, coherent integration is performed using an N-point FFT across the slow-time dimension for each range bin. The FFT simultaneously creates N Doppler filter outputs, each representing a different radial velocity hypothesis. A 64-pulse CPI, for example, produces 64 Doppler bins and 18.1 dB of processing gain. Extending the CPI to 256 pulses increases the gain to 24.1 dB but requires four times the dwell time and tighter oscillator stability. Designers balance integration time against scan revisit requirements, target maneuver limits, and phase noise budgets.

Integration Gain and CPI Limits

Coherent Integration SNR Gain:
SNR_out = N × SNR₁

Processing Gain (dB):
G_I = 10 log₁₀(N)

Maximum Coherent CPI (acceleration-limited):
T_coh,max ≈ (λ / a)^½

Where N = number of integrated pulses, SNR₁ = single-pulse SNR, λ = wavelength (m), a = target radial acceleration (m/s²). At X-band (3 cm) with a = 10 m/s², T_coh,max ≈ 55 ms.

Coherent vs. Non-Coherent Integration

Parameter	Coherent Integration	Non-Coherent Integration	Binary Integration	Design Impact
Input	Complex I/Q samples	Magnitude or \|x\|²	Binary (0/1) detections	Hardware complexity
SNR gain (N pulses)	N (= 10 log N dB)	N^0.5 to N^0.8	Variable (M-of-N)	Detection range
Phase coherence required	Yes, across full CPI	No	No	Oscillator cost
Doppler resolution	Δf = 1/T_CPI	None	None	Velocity measurement
Example: 64 pulses	18.1 dB gain	9 to 14.4 dB gain	~12 dB (M=5)	Sensitivity budget

Common Questions

Frequently Asked Questions

How much SNR improvement does coherent integration provide?

Coherent integration provides an SNR gain equal to the number of integrated pulses N, expressed as 10 log₁₀(N) dB. Integrating 64 pulses yields 18.1 dB of processing gain, while 256 pulses provides 24.1 dB. This is the theoretical maximum when all pulses maintain perfect phase coherence. In practice, oscillator phase noise, target acceleration, and platform motion reduce the effective gain by 0.5 to 2 dB for typical airborne pulse-Doppler systems integrating 32 to 256 pulses.

What is the difference between coherent and non-coherent integration?

Coherent integration sums complex I/Q pulse returns preserving both amplitude and phase, giving an SNR gain of N. Non-coherent integration sums magnitude or magnitude-squared values, discarding phase information, and provides a reduced gain of approximately N^0.5 to N^0.8 depending on single-pulse SNR. For 100 pulses, coherent integration gains 20 dB versus roughly 10 to 16 dB for non-coherent methods. Coherent integration requires a stable oscillator and Doppler knowledge, while non-coherent works without a phase reference.

What limits the coherent processing interval in practice?

Three factors limit the CPI: target coherence time, oscillator stability, and dwell time allocation. A target accelerating at a m/s² decorrelates after approximately (λ/a)^0.5 seconds. At X-band with 10 m/s² acceleration, this is about 55 ms. Oscillator phase noise must remain below roughly 1 radian RMS across the CPI. In multi-function radars, timeline scheduling constrains dwell time to 1 to 50 ms per beam position, further limiting the number of integrable pulses.

Related Terms

← Coherent Gain Coherent Jammer →