Complex Baseband Representation

The complex baseband signal decomposes into two orthogonal real-valued components: I (In-Phase) and Q (Quadrature). The I component modulates the cosine carrier, while the Q component modulates the sine carrier. This decomposition allows a single complex number I + jQ to represent both the instantaneous amplitude (envelope) and instantaneous phase of the modulated signal. For behavioral modeling, this means the PA's nonlinear dynamics can be captured entirely in the baseband domain without simulating the carrier frequency.

Any real bandpass signal x(t) = A(t)cos(2πf_c t + φ(t)) has an equivalent complex baseband representation x̃(t) = A(t)e^(jφ(t)). This transformation shifts the spectrum from around the carrier frequency f_c down to zero frequency. The critical property is that all information is preserved—the original bandpass signal can be perfectly reconstructed from its baseband equivalent. For PA modeling, this reduces the required simulation sampling rate from the Nyquist rate of the RF signal to the Nyquist rate of the modulation bandwidth, typically a 10-100x reduction.

From the complex baseband signal x̃(t) = I(t) + jQ(t), the instantaneous envelope is computed as |x̃(t)| = √(I² + Q²) and the instantaneous phase as ∠x̃(t) = arctan(Q/I). These parameters directly feed behavioral models:

• AM-AM distortion: Output envelope vs. input envelope
• AM-PM distortion: Output phase shift vs. input envelope
• Memory effects: Envelope-dependent time constants This explicit separation of amplitude and phase information is essential for understanding and correcting PA nonlinearities.

The PA's nonlinear behavior can be expressed as a complex gain function G(|x̃|) that depends only on the instantaneous input envelope magnitude. The output baseband signal is ỹ(t) = G(|x̃(t)|) · x̃(t). For memoryless nonlinearities, G is a simple complex-valued function of the envelope. For systems with memory, G becomes a Volterra kernel or memory polynomial operator. This formulation is the mathematical foundation for digital predistortion, where the predistorter implements the inverse complex gain characteristic.

Simulating a 2.4 GHz carrier with a 20 MHz modulation bandwidth would require sampling at >4.8 GHz to satisfy Nyquist. The complex baseband representation reduces this to ~40-60 MHz sampling. Key benefits for PA behavioral modeling:

• Computational efficiency: 100x fewer samples to process
• Memory reduction: Proportionally smaller data matrices for model extraction
• Real-time feasibility: Enables hardware-in-the-loop testing at baseband rates
• Algorithm development: Faster iteration on coefficient estimation and model validation

The complex baseband representation directly maps to the physical IQ modulator in transmitter hardware. The digital I and Q baseband signals are converted to analog, mixed with a local oscillator (cosine for I, 90°-shifted sine for Q), and summed to produce the RF output. This means:

• IQ imbalance errors appear as complex baseband distortion
• DC offset manifests as carrier leakage
• Quadrature skew creates cross-talk between I and Q Behavioral models operating on complex baseband signals can naturally incorporate these hardware impairments alongside PA nonlinearity.

A low-frequency signal representation that contains all the information of the original bandpass signal, used to simplify simulation and analysis of RF systems. The baseband equivalent captures the complex envelope—both amplitude and phase modulation—while omitting the high-frequency carrier. This representation is mathematically exact: the original bandpass signal can be reconstructed without loss by multiplying the baseband equivalent by the carrier frequency. In power amplifier modeling, working at baseband reduces the required sampling rate from gigahertz to megahertz, dramatically lowering computational complexity while preserving all nonlinear distortion characteristics.

The complex baseband signal is decomposed into two real-valued components: the in-phase (I) and quadrature (Q) channels. The I component modulates the cosine carrier, while the Q component modulates the sine carrier, which is 90° out of phase. This orthogonal decomposition allows any modulation scheme—QPSK, 16-QAM, OFDM—to be represented as a single complex number I + jQ at each symbol instant. For behavioral modeling, the I and Q paths are processed jointly to capture cross-channel distortion effects introduced by power amplifier nonlinearities.

The complex envelope A(t) = I(t) + jQ(t) is the time-varying complex amplitude that modulates the carrier. Its magnitude |A(t)| represents the instantaneous amplitude, while its phase ∠A(t) represents the instantaneous phase deviation. This representation is fundamental to understanding AM-AM and AM-PM distortion: the power amplifier's nonlinear transfer function acts directly on the complex envelope magnitude, compressing the amplitude and rotating the phase. Behavioral models such as memory polynomials operate entirely on the complex envelope, predicting the distorted output envelope from the input envelope.

Quadrature modulation is the physical process of impressing the complex baseband signal onto an RF carrier. The I component multiplies cos(2πf_c t) and the Q component multiplies sin(2πf_c t); summing these produces the bandpass signal. This architecture is universal in modern transmitters. However, imperfections in the analog quadrature modulator—gain imbalance between I and Q paths or phase error deviating from exact 90°—create IQ imbalance, which appears in the complex baseband representation as an undesired conjugate term. Behavioral models must account for this impairment alongside PA nonlinearity.

The analytic signal is a complex-valued time-domain signal whose imaginary part is the Hilbert transform of the real part, containing only positive-frequency components. The complex baseband representation is derived by frequency-shifting the analytic signal down by the carrier frequency. This mathematical construction ensures that the baseband signal is band-limited and free of negative-frequency artifacts. In digital predistortion systems, the analytic signal framework guarantees that predistorter processing does not introduce aliasing or spectral folding when the corrected signal is upconverted back to RF.

When modeling power amplifiers at complex baseband, the Nyquist sampling criterion applies to the modulation bandwidth, not the carrier frequency. For a 100 MHz 5G NR signal at 3.5 GHz, baseband sampling requires only ~200-500 MHz (depending on oversampling for nonlinearity capture) versus ~7 GHz for bandpass sampling. However, to accurately model spectral regrowth into adjacent channels, the baseband sampling rate must be 3-5x the signal bandwidth to capture third- and fifth-order intermodulation products that spill into neighboring frequency allocations.