Multiobjective Bayesian optimization for a 15‐dB back‐off high‐efficiency load modulated balanced amplifier design

In this article, multiobjecive Bayesian optimization (MBO) with a non‐penalization strategy is proposed to design a 15‐dB back‐off high‐efficiency load modulated balanced amplifier (LMBA). Applying the proposed method, the output matching networks of the LMBA are optimized to achieve proper load modulation behaviors that providing good performance. To verify the proposed method, a 2.0‐GHz LMBA is designed and measured. Experimental results show that LMBA achieves an output power of 44.5 dBm with a gain higher than 6.6 dB at saturation. The measured drain efficiency (DE)/power‐added efficiency (PAE) are 67%/52% at saturation, 54%/43% at 9‐dB back‐off, and 52%/46% at 15‐dB back‐off, respectively. Tested with 20‐MHz signal with 10‐dB peak‐to‐average power ratio (PAPR) and a 256‐quadrature amplitude modulation (QAM) scheme, the LMBA achieves adjacent channel leakage ratio (ACLR) levels of −46.7/−47.5 dBc and an error vector magnitude (EVM) of 1.2% after digital predistortion.