Signal-to-Noise Ratio Estimation

Relies on a known pilot sequence or preamble multiplexed into the transmitted signal. The receiver correlates the received samples against the stored replica to isolate the noise component.

• Maximum Likelihood (ML) estimators achieve the Cramér-Rao Lower Bound (CRLB) asymptotically.
• Advantage: High accuracy and low computational variance.
• Trade-off: Consumes valuable spectral efficiency by dedicating time-frequency resources to pilot symbols rather than payload data.
• Commonly used in standardized systems like LTE/5G NR where reference signals (DMRS) are predefined.

Also known as blind estimation, this approach computes SNR directly from the received signal's statistical moments without any prior knowledge of the transmitted symbols.

• M2M4 Estimator: Uses the second and fourth-order moments of the received signal envelope to separate signal power from noise variance.
• Decision-Directed (DD): Demodulates symbols first, treats them as correct, and subtracts the reconstructed signal to estimate noise.
• Advantage: Maximizes throughput by eliminating pilot overhead.
• Disadvantage: Suffers from error propagation at low SNR where symbol decisions become unreliable.

A specialized NDA technique that operates on M-PSK modulated signals by partitioning each symbol period into two halves and computing the sum and difference statistics.

• Exploits the constant envelope property of PSK constellations.
• The variance of the sum channel contains signal-plus-noise power, while the difference channel contains only noise power.
• Provides unbiased estimates even in the presence of carrier phase offset.
• Particularly effective in burst-mode satellite communications where rapid acquisition is critical.

Operates on the sample covariance matrix of the received signal vector, typically in multi-antenna or oversampled systems. SNR is derived from the ratio of eigenvalues.

• Minimum Description Length (MDL) or Akaike Information Criterion (AIC) can simultaneously estimate SNR and the number of signals present.
• The largest eigenvalue corresponds to the signal subspace, while the noise floor is estimated from the remaining eigenvalues.
• Robust in frequency-selective fading channels where time-domain methods fail.
• Computationally intensive due to matrix decomposition but highly accurate for MIMO systems.

Leverages the frequency-domain structure of orthogonal frequency-division multiplexing. Estimators exploit null subcarriers, pilot patterns, or the redundancy of the cyclic prefix.

• Null Subcarrier Method: Measures the power in unused guard band subcarriers where only noise is present.
• Cyclic Prefix Correlation: Correlates the CP with its repeated segment at the end of the symbol to estimate signal power.
• Pilot-Aided Frequency-Domain: Interpolates channel estimates across pilot-bearing subcarriers and computes the residual error variance.
• Must account for inter-carrier interference (ICI) caused by Doppler spread in mobile channels.

Integrates SNR estimation into the turbo decoding loop of modern error-correcting codes. The estimator refines its estimate with each iteration using increasingly reliable soft-output log-likelihood ratios (LLRs).

• Expectation-Maximization (EM) Algorithm: Treats transmitted symbols as hidden variables and iteratively maximizes the likelihood function.
• Achieves near-DA performance without pilot overhead after sufficient iterations.
• Critical for near-Shannon-limit systems like LDPC-coded 5G and DVB-S2X.
• The variance of the estimate decreases monotonically with each turbo iteration.