Power, Linearity, and Distortion Compression and Intercept Points Informational

How does gain compression affect the error vector magnitude of a digitally modulated signal?

Q: Can DPD improve EVM?

Yes. Digital Predistortion (DPD) applies an inverse distortion to the signal before the PA. The PA nonlinearity and the DPD cancel, producing a linear output. DPD can improve EVM by 10-20 dB (allowing the PA to operate 5-10 dB closer to P1dB while meeting the EVM spec). This directly improves efficiency: without DPD at 10 dB back-off: PAE ≈ 5-10%. With DPD at 3 dB back-off: PAE ≈ 25-40%. DPD is standard in all cellular base stations and is increasingly used in handsets and WiFi APs.

Q: Does the PA class affect EVM?

Yes. Class A: most linear (best EVM at a given back-off) but least efficient. Class AB: good compromise (used in most PAs without DPD). Its EVM performance is slightly worse than Class A but much better efficiency. Class F, J, inverse F: high-efficiency classes that achieve good EVM with DPD correction. Doherty: maintains efficiency at back-off (where the PA operates most of the time); EVM is managed through DPD. All high-efficiency PA architectures rely on DPD for acceptable EVM.

Q: How do I measure compression-induced EVM?

Method: generate a known modulated signal with very low EVM (< 0.5%) using an AWG or VSG. Apply it to the amplifier at increasing power levels. Measure the output EVM using a VSA. Plot EVM vs output power. The EVM increases as the output power approaches P1dB. The power level where EVM reaches the specification limit is the maximum allowable output power for that modulation format.

Gain compression in an RF amplifier directly degrades the EVM of a digitally modulated signal by distorting the amplitude and phase of the transmitted symbols: (1) Amplitude distortion (AM/AM): as the amplifier compresses, the gain decreases for large signal peaks. The outer constellation points (which have the largest amplitude) are compressed more than the inner points. For QPSK: all symbols have the same amplitude, and compression shifts all symbols equally (relatively benign). For 16-QAM: the outer symbols are compressed more than the inner symbols, distorting the amplitude ratio (AM/AM). For 64-QAM and 256-QAM: even small compression (0.5-1 dB) significantly distorts the multi-level amplitude, causing EVM > 5%. (2) Phase distortion (AM/PM): compression is accompanied by a phase shift (AM-to-PM conversion). The phase of the signal changes depending on the instantaneous amplitude. This rotates the outer constellation points relative to the inner points, creating additional EVM. (3) Rule of thumb: for a given compression level, the EVM is approximately: at P1dB (1 dB compression): EVM ≈ 8-12% (-18 to -21 dB). At P1dB - 3 dB (3 dB below P1dB, ~0.3 dB compression): EVM ≈ 3-5%. At P1dB - 6 dB (~0.1 dB compression): EVM ≈ 1-2%. At P1dB - 10 dB (deep back-off): EVM ≈ 0.3-0.5%. (4) Required back-off: for QPSK (EVM < 17.5%): operate at or slightly below P1dB (0-3 dB back-off). For 16-QAM (EVM < 12.5%): operate 3-6 dB below P1dB. For 64-QAM (EVM < 8%): operate 6-10 dB below P1dB. For 256-QAM (EVM < 3.5%): operate 10-15 dB below P1dB. The higher the modulation order: the more back-off is required, reducing the PA efficiency.

Category: Power, Linearity, and Distortion

Updated: April 2026

Product Tie-In: Amplifiers, Mixers, Attenuators

Compression and EVM

The relationship between amplifier compression and EVM is one of the most important trade-offs in modern wireless transmitter design.

Peak-to-Average Power Ratio

Modern modulation schemes (OFDM, SC-FDMA) have high PAPR (peak-to-average power ratio): LTE: PAPR ≈ 8-12 dB. 5G NR (OFDM with 256-QAM): PAPR ≈ 10-13 dB. WiFi 6/6E (OFDM): PAPR ≈ 10-12 dB. The PA must handle the peaks without compression while delivering the average power. Required P1dB: P1dB_required ≈ P_average + PAPR + margin. For P_average = 23 dBm (200 mW, typical UE transmit power) and PAPR = 12 dB: P1dB ≈ 23 + 12 + 2 = 37 dBm (5 W). The PA is producing only 200 mW average but must be sized for 5 W peak capability (efficiency at average power: typically 5-15% without DPD).

Compression vs EVM

At P1dB: EVM ≈ 8-12%
At P1dB-6dB: EVM ≈ 1-2%
64-QAM: need 6-10 dB back-off
256-QAM: need 10-15 dB back-off
P1dB_req ≈ P_avg + PAPR + margin

Common Questions

Frequently Asked Questions

Can DPD improve EVM?

Yes. Digital Predistortion (DPD) applies an inverse distortion to the signal before the PA. The PA nonlinearity and the DPD cancel, producing a linear output. DPD can improve EVM by 10-20 dB (allowing the PA to operate 5-10 dB closer to P1dB while meeting the EVM spec). This directly improves efficiency: without DPD at 10 dB back-off: PAE ≈ 5-10%. With DPD at 3 dB back-off: PAE ≈ 25-40%. DPD is standard in all cellular base stations and is increasingly used in handsets and WiFi APs.

Does the PA class affect EVM?

Yes. Class A: most linear (best EVM at a given back-off) but least efficient. Class AB: good compromise (used in most PAs without DPD). Its EVM performance is slightly worse than Class A but much better efficiency. Class F, J, inverse F: high-efficiency classes that achieve good EVM with DPD correction. Doherty: maintains efficiency at back-off (where the PA operates most of the time); EVM is managed through DPD. All high-efficiency PA architectures rely on DPD for acceptable EVM.

How do I measure compression-induced EVM?

Method: generate a known modulated signal with very low EVM (< 0.5%) using an AWG or VSG. Apply it to the amplifier at increasing power levels. Measure the output EVM using a VSA. Plot EVM vs output power. The EVM increases as the output power approaches P1dB. The power level where EVM reaches the specification limit is the maximum allowable output power for that modulation format.

Keep Reading