Error Vector Magnitude (EVM)

EVM is the ratio of the error vector power to the reference signal power, expressed as a percentage or in decibels. The error vector is the magnitude of the difference between the measured complex signal and the ideal reference constellation point at the exact symbol timing instant.

• RMS EVM: Averaged over all symbols in a frame.
• Peak EVM: The maximum error vector magnitude observed.
• Formula: EVM_RMS = sqrt(P_error / P_reference) × 100%

EVM is a composite metric that aggregates multiple physical-layer impairments into a single figure of merit. Key contributors include:

• Power Amplifier Nonlinearity: AM-AM and AM-PM distortion causing constellation warping.
• IQ Impairments: Gain imbalance, quadrature skew, and DC offset in the modulator.
• Phase Noise: Random phase fluctuations from the local oscillator spreading symbol points.
• Carrier Leakage: Unwanted feedthrough of the local oscillator appearing as a DC offset in the constellation.

EVM serves as the primary cost function for optimizing digital predistortion (DPD) coefficients. A well-trained DPD system minimizes EVM by pre-distorting the signal to cancel the PA's nonlinear characteristics.

• Direct Learning Architecture (DLA): Minimizes the error between the desired signal and the PA output.
• Indirect Learning Architecture (ILA): Identifies a post-inverse model of the PA and copies it to the predistorter.
• EVM Floor: The residual EVM after linearization, limited by memory effects and observation path noise.

Wireless standards define strict EVM limits to ensure reliable demodulation and high data throughput. Exceeding these limits results in compliance failure.

• 5G NR (FR1): 3.5% for 256-QAM, 1.5% for 1024-QAM.
• 5G NR (FR2/mmWave): 5.0% for 64-QAM.
• Wi-Fi 6 (802.11ax): -35 dB (1.8%) for 1024-QAM.
• LTE: 3.5% for 64-QAM.

EVM and Adjacent Channel Leakage Ratio (ACLR) are often inversely related in DPD optimization. Aggressive linearization to suppress spectral regrowth (ACLR) can sometimes increase in-band distortion (EVM) due to peak-to-average power ratio expansion.

• EVM: Measures in-band signal quality.
• ACLR: Measures out-of-band spectral containment.
• Joint Optimization: Modern DPD algorithms use multi-objective cost functions balancing both metrics simultaneously.

EVM is measured using a vector signal analyzer (VSA) that demodulates the received signal and compares it to an ideal reference generated from the detected symbols.

• Constellation Diagram: Visual scatter plot showing symbol point spreading.
• EVM vs. Subcarrier: Plots EVM per OFDM subcarrier to identify frequency-selective impairments.
• EVM vs. Symbol: Time-domain analysis revealing transient or memory effects.

A critical regulatory metric quantifying the ratio of transmitted power within an assigned channel to the power leaking into adjacent frequency channels. While EVM measures in-band signal quality, ACLR measures out-of-band spectral containment. Digital predistortion must simultaneously optimize both metrics, as aggressive linearization for EVM improvement can paradoxically worsen ACLR if spectral regrowth compensation is not properly balanced. Regulatory bodies like 3GPP specify strict ACLR limits (typically -45 dBc for 5G NR base stations) that transmitters must meet alongside EVM requirements.

The ratio of instantaneous peak power to average power in a transmitted signal, expressed in dB. High PAPR signals like OFDM force power amplifiers to operate with significant back-off from their compression point, reducing efficiency. This back-off directly impacts EVM because operating closer to saturation increases nonlinear distortion. Modern 5G and Wi-Fi signals exhibit PAPR values of 8-13 dB, requiring crest factor reduction techniques to balance efficiency against EVM degradation.

The unwanted appearance of signal energy in adjacent frequency channels caused by intermodulation products generated when a nonlinear power amplifier processes a modulated signal. Spectral regrowth is the physical mechanism behind ACLR degradation and is directly correlated with EVM deterioration—both originate from the same nonlinear transfer function. Wideband signals experience more severe regrowth because higher-order intermodulation products spread across broader frequency ranges, demanding DPD systems with sufficient linearization bandwidth to capture and cancel these products.

A signal conditioning technique applied before the power amplifier to reduce the peak-to-average power ratio of the transmitted waveform. CFR algorithms clip or shape signal peaks while managing the resulting in-band distortion to stay within EVM budgets. The technique creates a deliberate trade-off: aggressive CFR improves PA efficiency by reducing back-off requirements but introduces in-band distortion that degrades EVM. Modern systems cascade CFR with digital predistortion, where CFR handles peak reduction and DPD corrects the remaining nonlinearity.