Spurious Emission (SEM)

Spurious emissions are defined by bodies like the ITU-R and FCC as emissions on frequencies outside the necessary bandwidth and the out-of-band domain. They include harmonic emissions, parasitic emissions, and intermodulation products, but exclude spectral regrowth in adjacent channels. Compliance is mandatory for market access and requires testing across a wide frequency range, often up to the 10th harmonic of the carrier.

Unlike intermodulation distortion, spurious emissions often originate from non-ideal hardware behaviors:

• Harmonic Emissions: Integer multiples of the carrier frequency generated by nonlinear active devices.
• Parasitic Oscillations: Unintended resonances in bias networks or PCB traces.
• Phase Noise: Broadband noise skirts from the local oscillator extending far beyond the modulated signal.
• Digital Noise Coupling: Clock harmonics and digital switching noise coupling into the RF chain.

Regulatory bodies specify spurious emission limits in two ways:

• Absolute Limits: Fixed power levels (e.g., -36 dBm for frequencies >1 GHz in some bands) that must not be exceeded, regardless of the transmitter's output power.
• Relative Limits: Limits defined relative to the transmitter's carrier power (e.g., -50 dBc). Absolute limits are particularly challenging for high-power base stations, requiring extremely high-Q filtering and careful shielding.

Spurious emission testing is a critical step in device certification. It requires specialized equipment and environments:

• EMI Receivers and Spectrum Analyzers: With quasi-peak and average detectors per CISPR standards.
• Anechoic Chambers: To isolate the device under test from external ambient signals that could mask true spurious levels.
• Substitution Method: Replacing the DUT with a signal generator to calibrate path loss and accurately measure absolute power levels at the antenna port.

Suppressing spurious emissions requires a holistic design approach beyond digital predistortion:

• Low-Pass and Band-Pass Filtering: High-order cavity or ceramic filters placed after the power amplifier to attenuate harmonics.
• Electromagnetic Shielding: Board-level and enclosure-level shielding to prevent parasitic coupling and radiation.
• Spread-Spectrum Clocking: Dithering digital clock frequencies to spread noise energy and reduce peak spurious levels.
• Linear Power Supply Design: Minimizing ripple and switching noise that can modulate the RF carrier.

It is critical to distinguish spurious emissions from out-of-band (OOB) emissions. OOB emissions are the immediate spectral regrowth caused by the modulation process and nonlinearity, falling close to the assigned channel. Spurious emissions occur far from the carrier, are often caused by completely different physical mechanisms (like harmonics or parasitic oscillations), and are subject to separate, often stricter, regulatory limits that cannot be addressed by digital predistortion alone.

A spectral mask is a regulatory power-versus-frequency envelope that defines the absolute maximum allowable emissions. Unlike ACLR, which is a relative ratio, the mask sets an absolute power limit (dBm/Hz) for spurious domains.

• 3GPP TS 38.104 defines specific masks for 5G NR base stations
• SEM measurements must account for the transmitter's operating band and carrier aggregation configurations
• Violations typically occur due to harmonic distortion or parasitic oscillations in the PA

IMD generates spurious products at sum and difference frequencies when multiple signals pass through a nonlinear PA. Third-order products (IMD3) fall close to the carrier and are the primary cause of spectral regrowth.

• Occurs due to cubic nonlinearity in the amplifier transfer function
• IP3 (Third-Order Intercept Point) is the theoretical metric used to predict IMD levels
• DPD must model and cancel these odd-order products to suppress SEM

Nonlinear amplitude and phase conversion are the root physical mechanisms behind spectral regrowth. AM-AM distortion causes gain compression at high envelope peaks, while AM-PM distortion introduces phase shifts that vary with input power.

• Memoryless AM-AM/AM-PM curves are insufficient for wideband signals
• Memory effects cause frequency-dependent asymmetry in the SEM profile
• Neural network DPD excels at learning the complex mapping between AM-AM/AM-PM trajectories and the required correction

CFR reduces the Peak-to-Average Power Ratio (PAPR) of the transmitted waveform before the PA to prevent hard clipping. By smoothing signal peaks, CFR directly limits the generation of spurious emissions.

• Peak windowing applies a smooth time-domain filter to peaks exceeding a threshold
• Noise shaping pushes clipping distortion into frequencies far from the carrier
• CFR and DPD work synergistically: CFR reduces the peak excursions, and DPD cleans up the residual nonlinearity

Memory effects cause the PA's output to depend on past input states, not just the instantaneous envelope. This creates frequency-dependent nonlinearity that manifests as asymmetric spectral regrowth.

• Thermal memory arises from slow substrate heating and cooling
• Electrical memory stems from bias network impedance and envelope frequency coupling
• Trapping effects in GaN HEMTs introduce long time-constant dynamics
• Generalized Memory Polynomial (GMP) models capture these effects for DPD correction

SFDR quantifies the ratio between the fundamental signal and the highest spurious component within a specified bandwidth. It is the definitive metric for a receiver's ability to detect weak signals in the presence of transmitter-generated spurious emissions.

• Expressed in dBc (relative to carrier) or dBFS (relative to full scale)
• A high SFDR transmitter prevents desensitization of co-located receivers
• DPD directly improves SFDR by suppressing distortion products below the noise floor