Setting RF Amplifier Backoff for a Modulated Signal’s Peaks

An RF amplifier that handles a continuous tone at a particular power may distort a modulated waveform with the same average power. Envelope peaks can enter the nonlinear region even while the average appears modest. Select operating backoff from the waveform and required modulation quality rather than from average power alone.

Characterize the waveform

The Analog Devices discussion of transmitter lineup evaluation connects high peak-to-average ratio with power amplifier distortion and the use of backoff or linearization. The appropriate operating point depends on the actual signal and system requirement.

Record modulation, occupied bandwidth, allocation, pulse shaping, crest factor reduction, and observation duration. A nominal modulation label does not uniquely define the distribution of envelope peaks.

Relate peak and average power

A hypothetical waveform with a 10 dB peak-to-average power ratio reaches peaks ten times its average power over the stated observation. If average output is 20 dBm, those observed peaks correspond to 30 dBm before compression changes the waveform.

This calculation does not mean the amplifier needs an arbitrary fixed margin below its published P1dB. The acceptable level depends on AM-to-AM and AM-to-PM behavior, spectral regrowth, EVM, and the permitted probability of high peaks.

Sweep the operating point

Apply the representative modulated waveform and sweep average drive while measuring output power, EVM, adjacent channel behavior, and DC consumption. Keep the signal generator sufficiently linear so it does not dominate the observed distortion.

Watch for a loss of crest factor at the output as peaks compress. A modest increase in average output accompanied by degraded modulation quality can identify the region where additional drive is no longer useful.

Include thermal and linearization states

Repeat after thermal stabilization and across the required frequencies and supply conditions. If digital predistortion or another correction is used, document its training and enabled state. Its benefit should be verified with the actual waveform and operating range.

Check startup and burst behavior separately if the application does not operate continuously. Bias settling and memory effects can make a steady-state result incomplete.

Record the waveform file and generator scaling with the result. A regenerated random payload can have different rare peaks, making a small change in apparent clipping difficult to interpret.

Save waveform statistics, power sweep, distortion limits, efficiency, and temperature. A satisfactory operating point meets the required modulation and spectral performance with a documented power tradeoff. A CW compression number is useful characterization, but it does not directly specify the average output power available for every high crest factor signal.