Non-Coherent Detection

The optimal demodulation counterpart requiring carrier phase recovery. A coherent receiver cross-correlates the incoming signal with a locally generated, phase-synchronized replica. While achieving the theoretical minimum Symbol Error Rate (SER) in AWGN, it demands complex Phase-Locked Loops (PLLs) that are vulnerable to cycle slips and phase noise.

• Performance: Achieves a 3 dB advantage over non-coherent FSK.
• Complexity: High; requires accurate channel estimation.
• Failure Mode: Degrades rapidly in high-Doppler environments.

An optimal detection strategy that selects the most probable transmitted symbol sequence rather than deciding symbol-by-symbol. Commonly implemented via the Viterbi algorithm, MLSE is the benchmark against which non-coherent methods are measured. It inherently accounts for Inter-Symbol Interference (ISI) by operating on the entire received sequence.

• Core Mechanism: Maximizes the joint conditional probability density function.
• Trade-off: Exponential computational complexity with channel memory length.

A sub-optimal hypothesis testing framework that replaces unknown deterministic parameters—like carrier phase—with their Maximum Likelihood Estimates (MLEs). GLRT forms the mathematical basis for many non-coherent classifiers. It avoids requiring prior distributions for nuisance parameters.

• Process: Estimates phase, then plugs the estimate into the likelihood function.
• Advantage: Tractable computation compared to optimal Bayesian averaging.
• Application: Standard approach for non-coherent modulation classification.

The process of handling unknown variables—like carrier phase offset or frequency drift—that are not of primary interest but must be accounted for to evaluate the likelihood function. Non-coherent detection is fundamentally a strategy to marginalize or estimate these parameters without explicit recovery.

• Blind Estimation: Operates without training symbols, relying on signal statistics.
• Data-Aided Estimation: Uses known pilots for high accuracy at the cost of spectral efficiency.
• Impact: Mismatched nuisance parameters cause catastrophic classifier failure.

An adaptive thresholding technique critical for robust non-coherent detection in dynamic noise environments. CFAR maintains a fixed probability of false alarm by continuously estimating the background noise power from adjacent range bins or frequency cells and scaling the detection threshold accordingly.

• Cell-Averaging CFAR: Computes the mean noise level from surrounding reference cells.
• Benefit: Prevents receiver saturation from fluctuating interference.
• Relevance: Ensures non-coherent energy detectors remain calibrated.

The fundamental channel model representing thermal noise with a flat power spectral density and Gaussian amplitude distribution. AWGN forms the theoretical baseline for analyzing non-coherent detection performance. The non-coherent combining loss is explicitly quantified relative to coherent detection in this model.

• Mathematical Property: Noise samples are statistically independent.
• Performance Bound: Non-coherent FSK requires roughly 1 dB more Eb/N0 than coherent FSK at 10^-5 BER.