How do I perform swept power measurements on a power amplifier to characterize AM-AM and AM-PM?
AM-AM and AM-PM Measurement
AM-AM and AM-PM characterization is essential for: digital pre-distortion (DPD) design (the DPD linearizer must model and invert both the AM-AM and AM-PM distortion), EVM estimation (AM-PM converts amplitude variations into phase errors, rotating the constellation points and degrading EVM), and system simulation (accurate PA models require both AM-AM and AM-PM data).
Significance
- AM-AM: Causes spectral regrowth (adjacent channel leakage) and in-band distortion. Well understood and easy to compensate with DPD
- AM-PM: Creates asymmetric spectral regrowth and constellation rotation. Harder to compensate because: it is a phase effect (not amplitude), and it varies with temperature and frequency. Typical values: GaAs PA: 5-15°/dB near P1dB. GaN PA: 2-8°/dB. LDMOS PA: 3-10°/dB
AM-PM: Δφ(P_in) = ∠S21(P_in) - ∠S21(P_in_linear) [degrees]
AM-PM coefficient: K_PM = dΔφ/dP_in [°/dB]
EVM from AM-PM: EVM_PM ≈ K_PM × PAPR_signal [degrees → %]
Typical K_PM at P1dB: 2-15 °/dB
Frequently Asked Questions
Why does AM-PM matter?
AM-PM matters because: modern digital modulation schemes (64-QAM, 256-QAM, OFDM) encode information in both amplitude and phase. AM-PM distortion converts amplitude variations (which are inherent in high-PAPR waveforms like OFDM) into phase errors. For example: a 256-QAM signal has constellation points separated by approximately 5 degrees in phase. If the PA introduces 5 degrees of AM-PM distortion: the constellation points rotate differently depending on their amplitude, causing: phase errors (EVM degradation), adjacent channel interference (spectral regrowth), and bit errors. For 5G NR (256-QAM, OFDM): AM-PM must be kept below approximately 1-2 degrees/dB to meet the EVM requirement of less than 3.5%.
How is this data used for DPD?
Digital Pre-Distortion (DPD) uses the AM-AM and AM-PM measurements to create an inverse model of the PA: the DPD applies the inverse AM-AM gain curve (boosting the signal where the PA compresses and reducing it where the PA expands) and the inverse AM-PM phase curve (pre-rotating the phase to cancel the PA's AM-PM). The DPD input signal: x_DPD(t) = F_inverse(x(t)), where F_inverse is the combined inverse of the AM-AM and AM-PM characteristics. With ideal DPD: the PA output is perfectly linear (no distortion, no spectral regrowth). In practice: DPD reduces distortion by 15-30 dB, enabling the PA to operate closer to saturation (higher efficiency) while meeting linearity requirements.
What VNA settings are needed?
VNA power sweep settings for AM-AM/AM-PM: frequency: fixed at the center frequency of interest. Power sweep range: from P_in_min (at least 20 dB below P1dB) to P_in_max (5-10 dB above P1dB or the PA's maximum rated input). Number of points: 100-200 (for smooth curves). IF bandwidth: 1-10 kHz (narrow for accurate phase measurement). Calibration: full 2-port SOLT calibration. Source and receiver attenuators: set to keep the PA output within the VNA's safe input range. Plot: display S21 magnitude (AM-AM) and S21 phase (AM-PM) on separate Y-axes.