Error Vector Magnitude (EVM)

EVM is the root-mean-square (RMS) value of the error vector—the vector difference between the ideal reference signal and the measured transmitted signal—normalized to the magnitude of the ideal signal. It is typically expressed as a percentage.

• Formula: EVM = (RMS Error Magnitude / Peak Reference Magnitude) × 100%
• Error Vector: The complex difference between the ideal I/Q constellation point and the actual measured point.
• Normalization: Can be relative to the peak constellation magnitude, average symbol power, or the power of the outermost symbol, making specification comparison critical.

EVM is the primary figure of merit for assessing in-band distortion correction. While metrics like ACLR measure out-of-band spectral regrowth, EVM directly quantifies the modulation accuracy that determines the receiver's ability to demodulate the signal without errors.

• Pre-DPD Baseline: A power amplifier operating near compression may exhibit an EVM of 5-10%, rendering higher-order modulation schemes unusable.
• Post-DPD Target: A well-optimized DPD system can reduce EVM to below 1%, enabling 256-QAM and 1024-QAM transmission.
• Correlation with BER: EVM has a direct mathematical relationship with bit error rate for a given modulation order, making it a predictor of link performance.

EVM degradation is a composite result of multiple transmitter chain impairments, all of which a DPD system must address simultaneously.

• PA Nonlinearity: AM-AM and AM-PM distortion from gain compression is the dominant contributor that DPD directly linearizes.
• IQ Impairments: Gain imbalance, quadrature skew, and DC offset in the modulator create constellation warping that increases EVM.
• Phase Noise: Local oscillator phase noise causes random rotational smearing of constellation points.
• Memory Effects: Thermal and electrical memory effects in the PA cause symbol-dependent distortion that cannot be corrected by static predistortion.

Each wireless standard specifies a maximum permissible EVM to ensure reliable communication. DPD systems must meet these stringent requirements across all operating conditions.

• 5G NR (3GPP TS 38.104): EVM ≤ 3.5% for 64-QAM, ≤ 1.5% for 256-QAM, and ≤ 0.5% for 1024-QAM.
• Wi-Fi 6 (802.11ax): EVM ≤ -35 dB (approx. 1.8%) for 1024-QAM modulation.
• LTE: EVM ≤ 17.5% for QPSK, ≤ 3.5% for 64-QAM.
• DOCSIS 3.1: EVM ≤ 1.0% for 4096-QAM profiles, demanding exceptional DPD performance.

Accurate EVM measurement requires a vector signal analyzer (VSA) and precise time alignment between the reference and measured signals. The measurement process involves:

• Time Alignment: Cross-correlation of the reference and captured waveforms to remove propagation delay and trigger jitter.
• Channel Equalization: Removing linear channel effects to isolate the nonlinear distortion contributed by the PA.
• Demodulation: Recovering the transmitted symbols and computing the error vector for each symbol period.
• Averaging: Computing the RMS EVM over a statistically significant number of frames to ensure repeatable results.

EVM and Adjacent Channel Leakage Ratio (ACLR) are complementary metrics that together provide a complete picture of DPD effectiveness.

• EVM (In-Band): Measures distortion within the occupied channel bandwidth, directly impacting modulation accuracy and data throughput.
• ACLR (Out-of-Band): Measures spectral regrowth into adjacent channels, impacting regulatory compliance and multi-carrier coexistence.
• DPD Optimization: A predistorter must balance both metrics. Overly aggressive linearization can sometimes improve ACLR at the expense of EVM if the DPD model introduces its own in-band artifacts.