How does the Doherty amplifier maintain efficiency at back-off?

At low power levels (below the transition point), only the carrier amplifier is active. It sees a load impedance of 2 times R_opt (transformed through the quarter-wave inverter), which means it reaches voltage saturation at half its maximum current. At this point it operates at the efficiency of a saturated Class AB amplifier (roughly 50 to 60 percent). As input power increases past the transition point, the peaking amplifier turns on and injects current into the common load. Through the impedance inverter, this current dynamically reduces the load impedance seen by the carrier from 2R_opt toward R_opt. The carrier amplifier stays in saturation (maintaining peak efficiency) while the peaking amplifier handles the additional power demand. The result is an efficiency curve with two peaks instead of the single-peak curve of a conventional PA.

What is the difference between symmetric and asymmetric Doherty?

In a symmetric (classical) Doherty, both amplifiers are the same size. The efficiency peak at back-off occurs at 6 dB below peak power, matching WCDMA signals but too shallow for LTE and 5G signals with 9 to 12 dB PAPR. An asymmetric Doherty uses a larger peaking amplifier (typically 2 to 3 times the carrier size), extending the high-efficiency region to 9 to 10 dB back-off. Three-way Doherty designs add a second peaking amplifier to extend the range to 12 dB or more, matching 5G NR PAPR requirements.

Why has Doherty become dominant in 5G base stations?

5G NR signals have 10 to 12 dB PAPR, meaning a conventional Class AB PA operates at 5 to 8 percent average efficiency. An asymmetric Doherty with DPD achieves 40 to 50 percent average efficiency with the same signal, reducing the DC power consumption of a macro base station from 1500 W to 400 W per sector. With electricity costs and carbon targets driving operator OPEX, the Doherty architecture saves 10,000 to 15,000 dollars per site per year. Nearly every major base station PA (from NXP, Wolfspeed, Ampleon, and MACOM) is now designed for Doherty operation.

Active Components

Doherty Power Amplifier

William Doherty invented this architecture in 1936 to solve AM radio's efficiency problem, and 90 years later it dominates the multi-billion-dollar 5G base station PA market for the exact same reason. The core insight: use two amplifiers, one always on (the carrier) and one that switches on only for peaks (the peaker), connected through a quarter-wave impedance inverter that makes the peaker's current dynamically change the load impedance the carrier sees. At average power levels, the carrier operates into a high impedance and stays saturated; at peak power, both amplifiers contribute and the load drops. The result is near-peak efficiency at both average and peak power, not just at one.

Category: Active Components

Efficiency at 10 dB BO: 40 to 50% (with DPD)

Market Share: >90% of macro base station PAs

Two Amplifiers, One Quarter-Wave Trick

The Doherty output combiner is a quarter-wave transmission line that performs impedance inversion: when the peaking amplifier is off (high impedance), the carrier sees 2×R_opt and reaches voltage saturation at half the peak current. When the peaker turns on and injects current into the common node, the quarter-wave line transforms this into a load reduction for the carrier, pulling its impedance from 2R_opt down toward R_opt. The carrier stays saturated throughout, and the peaker provides the additional power for the peaks.

Doherty Architecture Variants

Variant	Back-Off Efficiency Peak	Peaker Size	Complexity	Best Signal Type
Symmetric (classical)	6 dB	Equal to carrier	Low	WCDMA (7 dB PAPR)
Asymmetric (1:2)	9 to 10 dB	2× carrier	Moderate	LTE, 5G NR (10 to 11 dB)
Three-way	12 dB	2 peakers	High	Multi-carrier 5G (12+ dB)
Digital Doherty	10 to 12 dB	Varies	Highest	Massive MIMO sub-arrays
Inverted Doherty	6 dB	Equal	Low	Broadband (no λ/4 at output)

Efficiency Comparison: Doherty vs. Conventional

Average efficiency with a 10 dB PAPR signal:

Class AB (conventional):
η_avg ≈ η_peak × 10^−PAPR/20 = 60% × 10^−0.5 = 60% × 0.316 ≈ 19%

Symmetric Doherty (6 dB BO peak):
η_avg ≈ 35 to 40%

Asymmetric Doherty (10 dB BO peak):
η_avg ≈ 45 to 50%

For a 200 W average-power macro cell PA: conventional Class AB dissipates 850 W as heat; asymmetric Doherty dissipates 200 W. That is $12,000/year in electricity savings per sector at $0.12/kWh.

Common Questions

Frequently Asked Questions

How does Doherty maintain efficiency at back-off?

At low power, only the carrier amplifier runs, seeing 2×R_opt through the quarter-wave inverter and reaching saturation at half current. As input rises, the peaker turns on, injecting current that dynamically lowers the carrier's load impedance. The carrier stays saturated (peak efficiency) while the peaker handles the peaks.

Symmetric vs. asymmetric Doherty?

Symmetric (equal-size amplifiers) peaks efficiency at 6 dB back-off, fine for WCDMA. Asymmetric (peaker 2 to 3× carrier) extends the peak to 9 to 10 dB, matching LTE/5G PAPR. Three-way designs reach 12+ dB for multi-carrier signals.

Why is Doherty dominant in 5G base stations?

5G NR at 10 to 12 dB PAPR drops conventional PA efficiency to 5 to 8%. Asymmetric Doherty with DPD achieves 40 to 50%, saving $10,000 to $15,000/site/year in electricity. Over 90% of macro base station PAs are now Doherty designs.

Related Terms

← Digital Predistortion Doppler Shift →

PA Architecture

Doherty Combiner Design Worksheet

Excel-based tool for computing quarter-wave line impedances, offset lines, and peaker bias points for symmetric, asymmetric, and three-way Doherty architectures.

Download Worksheet