I/Q Pre-Distortion

Unlike standard linear filtering, I/Q pre-distortion operates on a widely-linear model. The correction engine processes both the original complex baseband signal and its complex conjugate. This is mathematically necessary because I/Q imbalance creates an image signal that is a conjugate copy of the desired signal. The pre-distorter's 2x2 matrix structure applies a direct filter to the signal and a conjugate filter to cancel the image, a process impossible with a simple scalar multiplier.

Real-world analog impairments are rarely flat across a wide bandwidth. I/Q pre-distortion employs complex-valued FIR filters to correct frequency-dependent mismatch. Key filter design considerations include:

• Anti-aliasing filter mismatch: Compensates for differing passband ripple between I and Q paths
• Trace length skew: Corrects the linear phase distortion from picosecond-level timing differences
• Filter tap count: Balances correction accuracy against computational latency, typically 5-15 taps for moderate bandwidths This transforms a simple scalar correction into a full convolution operation.

A single I/Q pre-distortion block simultaneously corrects three distinct physical impairments without separate processing stages:

• Gain imbalance: Amplitude mismatch between I and Q DACs and reconstruction filters
• Phase imbalance (Quadrature Error): Deviation from the ideal 90-degree LO phase offset
• DC offset: Carrier feedthrough caused by LO self-mixing, corrected by adding a negative DC term This unified approach is more computationally efficient than cascaded single-impairment correctors and prevents error propagation between stages.

I/Q pre-distortion systems often use blind estimation to track time-varying impairments without interrupting live traffic. The algorithm exploits the statistical property of circularity—a properly balanced complex signal has zero pseudo-autocorrelation. By minimizing the correlation between the transmitted signal and its conjugate, the estimator extracts the mismatch coefficient (K). This coefficient directly parameterizes the widely-linear correction matrix, enabling continuous, training-sequence-free adaptation to temperature drift and component aging.

The primary performance metric for I/Q pre-distortion is the Image Rejection Ratio (IRR). The goal is to push the unwanted image sideband far below the noise floor. Typical performance targets include:

• Narrowband correction: Achieves 50-60 dB IRR using a single complex scalar
• Wideband correction: Achieves 40-50 dB IRR across 100+ MHz bandwidths using FIR filtering
• Joint DPD + I/Q correction: Can exceed 65 dB IRR when combined with power amplifier linearization Each 10 dB improvement in IRR directly translates to lower EVM and cleaner spectral emissions.

I/Q pre-distortion is the critical first stage in a complete transmitter linearization chain. The signal flow is:

1. I/Q Pre-distorter: Corrects modulator impairments, presenting a clean signal to the PA
2. Crest Factor Reduction (CFR): Limits peak-to-average power ratio
3. PA Digital Pre-distorter: Linearizes the power amplifier's nonlinear response Without I/Q correction first, the PA DPD would attempt to linearize a signal already corrupted by image interference, leading to model instability and poor ACLR performance. The two systems must be co-designed.