I/Q Mismatch Estimation

The mathematical foundation of I/Q mismatch estimation models the impaired system as a widely-linear transformation. Unlike standard linear systems, the output depends on both the input signal and its complex conjugate. This formulation captures the generation of the image signal, which is the primary artifact of I/Q imbalance. The estimation problem reduces to identifying the complex coefficients that map the ideal signal and its conjugate to the observed impaired output.

A fundamental batch estimation technique that solves for the I/Q mismatch coefficients by minimizing the squared error between the observed feedback signal and the model prediction. Given a block of transmitted and received samples, the normal equations are formed and solved to yield the optimal complex-valued correction parameters. This method is statistically efficient for frequency-independent imbalance in the presence of additive white Gaussian noise.

A powerful technique that extracts imbalance parameters without a dedicated training sequence. It exploits the statistical property of circularity (or properness), where a perfectly balanced complex baseband signal has zero pseudo-autocorrelation. I/Q imbalance destroys circularity, creating a non-zero relation between the signal and its conjugate. Algorithms like the spectral mean amplitude and phase method estimate the mismatch by measuring the statistical correlation between the received signal and its complex conjugate.

For frequency-dependent I/Q imbalance, estimation is performed in the frequency domain using known pilot symbols. By transmitting pilots on specific subcarriers and observing the resulting image interference on mirror subcarriers, the frequency-selective gain and phase ripple can be characterized. This method constructs a per-subcarrier widely-linear equalizer matrix, enabling precise compensation across the entire signal bandwidth, which is critical for wideband signals like 5G NR.

For time-varying impairments caused by temperature drift or voltage fluctuations, adaptive algorithms such as the Least Mean Squares (LMS) or Recursive Least Squares (RLS) are employed. These methods iteratively update the mismatch coefficient estimates sample-by-sample, minimizing the instantaneous error between the desired and observed signal. The convergence rate and steady-state misadjustment are key trade-offs, making RLS preferable for fast-tracking scenarios despite its higher computational complexity.

In a direct conversion transmitter, the quadrature modulator and the power amplifier form a cascaded impairment chain. Advanced estimation algorithms solve for I/Q mismatch and PA nonlinearity jointly rather than sequentially. This approach models the combined system as a parallel Hammerstein or dual-input nonlinear structure, preventing the error propagation that occurs when one impairment is estimated in the presence of the uncorrected other. This is essential for achieving deep image suppression and spectral regrowth mitigation simultaneously.

A physical impairment in quadrature modulators where the I and Q paths exhibit mismatched gain or non-orthogonal phase. This fundamental distortion creates a mirror image of the transmitted signal, degrading constellation integrity and causing spectral regrowth into adjacent channels. The imbalance is the primary target of estimation algorithms.

The key performance metric quantifying a transmitter's ability to suppress the unwanted image signal generated by I/Q imbalance. Expressed in decibels as the power ratio between the desired signal and its image. High IRR is the objective of compensation. Estimation algorithms directly optimize this metric.

• Uncorrected: 25-40 dB typical
• After compensation: 60-80+ dB achievable

A complex mismatch where gain and phase errors vary across the signal bandwidth. Caused by mismatched anti-aliasing filters, PCB trace length differences, or component parasitics. Requires a complex FIR filter for correction rather than a simple scalar multiplication. Estimation must characterize the mismatch as a function of frequency.

• Causes: Filter mismatch, trace skew
• Correction: Widely-linear filtering

A signal processing technique that extracts I/Q imbalance parameters directly from the statistical properties of the modulated signal without requiring a known pilot or training sequence. Exploits the fact that ideal communication signals are proper (circularly symmetric), while impaired signals exhibit non-circularity. Enables online, in-service calibration.

• Key property: Signal circularity
• Advantage: No bandwidth overhead

A digital linearization technique where baseband I and Q signals are intentionally distorted with an inverse model of the modulator's imbalance before DAC conversion. The pre-distorted signal cancels the analog impairment, resulting in a clean output at the antenna. Requires accurate mismatch coefficients from the estimation stage.

• Input: Estimated mismatch matrix
• Output: Corrected constellation

A comprehensive modulation quality metric measuring the vector difference between the ideal reference constellation point and the actual transmitted signal. Directly degraded by uncorrected I/Q imbalance, which scatters constellation points and increases EVM. Estimation algorithms are validated by their ability to minimize post-compensation EVM.

• Impact: I/Q imbalance increases EVM
• Validation: EVM reduction confirms estimation accuracy