Abstract-This paper presents a high-efficiency, linear power amplifier (PA) for 28GHz mobile communications in 40nm CMOS technology. The design and layout are optimized for high linearity while maintaining high gain and output power. A capacitance neutralized differential pair with source degeneration inductor for linearity enhancement is discussed. The inductive degeneration technique greatly increases the optimal load impedance, which enables a low loss parallel power combining. The complete PA achieves a measured saturated output power of 18.1dBm with 41.5% power-added efficiency (PAE). With 6 Gb/s QAM-64 signals, the proposed PA achieves an average output power of 8.4dBm and 8.8% PAE, with -25 dBc EVM. All measurements are performed with a fixed bias condition.Index Terms-CMOS, power amplifier, inductive degeneration, linearity, EVM, 5G.
I. INTRODUCTIONSeveral frequency bands in millimeter-wave (mmWave) range have been licenced for future fifth generation (5G) wireless communication systems, enabling Gb/s data rate with low latency. To meet the demand of large data traffic within limited spectrum resources, high order modulation scheme like 64 QAM will be adopted, which presents large peak-to-average power ratio (PAPR).To amplify these modulated signals with high fidelity, PA designs for 5G applications have to meet the stringent AM-AM and AM-PM linearity requirements. In RF and microwave range, multiple correction techniques have been investigated to improve linearity of the front-ends. However, it is challenging to deploy these techniques into 5G applications, such as phased arrays and MIMO systems, where the baseband bandwidth is relatively large and the function units might be distributed into the systems. Digital pre-distortion (DPD) for each PA can be prohibitively complex for large-scale arrangement. In any case, the nonlinearity property of the PA itself needs to be dealt with in the first place.Recently, lots of efforts have been put to improve the PA performance for 5G applications. In the work of [1], high efficiency and linearity are achieved by using subthreshold biasing and inductive degeneration. And as a variable, the size of the transistor is optimized for -25dBc EVM. In [2], the linearity is greatly improved by using second harmonic control and deep class-AB biasing. However, the overall performance of the PA, in terms of