Complex Baseband Equivalent

A real bandpass signal x(t) centered at carrier frequency f_c is represented as:

x(t) = Re{ x̃(t) · e^(j2πf_ct) }

where x̃(t) is the complex baseband equivalent, also called the complex envelope. This decomposition separates the slowly-varying modulation information from the rapidly oscillating carrier, enabling sampling at rates proportional to the signal bandwidth rather than twice the carrier frequency.

The complex baseband signal is expressed in Cartesian form:

x̃(t) = I(t) + jQ(t)

• I(t): In-phase component (real part), modulating a cosine carrier
• Q(t): Quadrature component (imaginary part), modulating a sine carrier

This IQ representation is the native data format for all digital predistortion processing, as it preserves both amplitude and phase information required for nonlinearity compensation.

When a bandpass signal passes through a nonlinear power amplifier, the complex baseband equivalent of the system captures the distortion without simulating the carrier:

• The PA's bandpass nonlinearity is transformed to a baseband nonlinear operator
• Only odd-order nonlinear terms affect the in-band and adjacent-channel response
• Even-order distortion products fall far from the carrier and are typically filtered

This transformation is essential for memory polynomial and Volterra series behavioral modeling.

The key practical benefit of complex baseband representation is dramatically reduced sampling requirements:

• A 2 GHz carrier with 100 MHz bandwidth requires sampling at only ~200-300 MSPS in baseband
• Direct RF sampling would require >4 GSPS per Nyquist
• Enables FPGA-based DPD implementation with commercially available data converters
• Reduces simulation time by orders of magnitude for wideband signal linearization

The complex baseband signal can also be expressed in polar form:

x̃(t) = A(t) · e^(jφ(t))

• A(t) = |x̃(t)|: Instantaneous envelope magnitude — the primary input to look-up table indexing and memoryless predistortion
• φ(t) = ∠x̃(t): Instantaneous phase

This decomposition directly maps to envelope memory polynomial models, where the envelope magnitude drives long-term thermal and bias memory effects.

A power amplifier's nonlinear behavior is often characterized by its complex gain function G(·):

ỹ(t) = G(|x̃(t)|²) · x̃(t)

• The gain depends only on instantaneous input power |x̃(t)|² for memoryless PAs
• For PAs with memory, G(·) becomes a dynamic nonlinear operator
• The digital predistorter implements the inverse: x̃_DPD(t) = G⁻¹(|x̃_original(t)|²) · x̃_original(t)

This formulation is the basis for predistorter synthesis and coefficient extraction algorithms.

The native Cartesian format for the complex baseband equivalent, representing a bandpass signal as I (real) and Q (imaginary) components.

• Eliminates the carrier frequency from analysis
• The complex envelope is defined as: x(t) = I(t) + jQ(t)
• Directly maps to physical modulator inputs on FPGA/DSP hardware
• All memory polynomial basis functions operate on this complex-valued stream

The mathematical operation that shifts a modulated RF signal centered at fc down to zero frequency while preserving its amplitude and phase modulation.

• The physical bandpass signal is: Re{x(t) * e^(j2πfct)}
• The complex envelope x(t) contains all information
• Enables simulation at symbol rate rather than carrier rate
• Reduces computational burden by orders of magnitude for DPD coefficient estimation

The process of extracting the instantaneous magnitude |x(t)| from the complex baseband signal, which serves as the primary input to memoryless nonlinearity models.

• Calculated as: sqrt(I² + Q²)
• Drives look-up table indexing in LUT-based predistorters
• Forms the basis for envelope memory polynomial terms
• Critical for capturing AM-AM and AM-PM distortion characteristics

The power amplifier's nonlinear transfer characteristic expressed as a complex-valued gain G(|x|) that depends only on the instantaneous envelope magnitude.

• Decomposed into AM-AM (magnitude distortion) and AM-PM (phase distortion)
• The baseband output is: y(t) = G(|x(t)|) * x(t)
• Simplifies Volterra series to memory polynomial form
• Enables direct predistorter synthesis by computing the inverse gain function

The minimum sampling frequency required to capture the complex baseband signal without aliasing, determined by the modulation bandwidth rather than the RF carrier.

• Nyquist criterion applies to the complex envelope bandwidth, not the RF bandwidth
• Typically 3-5x the signal bandwidth for DPD to capture third and fifth-order intermodulation products
• Directly impacts FPGA resource utilization and coefficient estimation latency
• Undersampling the baseband equivalent causes spectral regrowth aliasing

The complete transmitter chain model where all components—filters, modulators, and the power amplifier—are represented by their complex baseband equivalents.

• Linear filters become complex-valued lowpass transfer functions
• The PA is modeled as a baseband nonlinearity with memory
• Enables end-to-end simulation without RF carrier generation
• Forms the foundation for indirect learning architecture DPD training loops