Example 5: Doherty Power Amplifier

The second example that will be considered is a Doherty power amplifier found in the example library of Keysight’s Advanced Design System (ADS). The amplifier is built with two Freescale MRF8S21100H transistors for a centre frequency of $2.14\mathrm{GHz}$ (Figure E5.1). The main transistor is biased in class-AB with a quiescent current of $0.7\mathrm{A}$. The auxiliary transistor is biased deep in class-C with a quiescent current of $1.1\mathrm{mA}$.

Figure E5.1The Doherty power amplifier.

The amplifier is excited by bandpass multisines that have $41$ spectral lines in a band of $10\mathrm{MHz}$ around $2.14\mathrm{GHz}$. The RMS of the input multisines is $22\mathrm{dBm}$. The steady-state response of the circuit to the different-phase multisine excitations is obtained by HB simulation. The spectrum of the output wave $B_t$ around the centre frequency is shown in Figure E5.2.

Figure E5.2The output spectrum $B_t$ of the Doherty power amplifier. In a band-pass system only odd non-linear distortion (red stars) falls inside the band. The RMS of the distortion is also shown (magenta dashes).

This circuit can be considered to be strongly non-linear, especially because of the auxiliary amplifier. To obtain the BLAs of the two transistors in the circuit, tickler multisines are added to the circuit at the output of the total amplifier. The added multisines are current sources which insert an RMS current of $1\mathrm{\mu A}$ on frequency bins in between the frequencies of the main multisine. 60 different-phase multisines were used to estimate the MIMO BLA and derive the distortion present in the circuit. The total simulation time for this Doherty Power amplifier was 2 hours (approximately 2 minutes for one different-phase multisine). Increasing the number of different-phase multisines decreases the variability of the distortion estimate in Figure E5.2 but inevitably increases the computational complexity and cost.

The main distortion contributor is found to be the main transistor (Figure E5.3). This can be expected, as the auxiliary amplifier only kicks in for limited amounts of time in this Doherty configuration. A similar Doherty amplifier was analysed in [31]with a Volterra-based DCA under two-tone excitation. It was concluded there that the auxiliary amplifier only contributes significantly to the distortion for very high amplitudes in the two-tone. With modulated signals, like the multisines used in the BLA-based DCA, the peaks only occur from time to time, so the average contribution of the auxiliary amplifier to the total distortion is low.

Figure E5.3The distortion contributions show that the main amplifier is the dominant source of non-linear distortion in the Doherty amplifier.

The information given by the BLA-based DCA is limited for the Doherty power amplifier, because only the signals around the centre frequency are used here. Designers are also interested in how the low-frequency signals in the bias and supply networks are up-converted in-band through the second-order non-linearities [2]. The current implementation of the DCA doesn’t split the up-converted low-frequency signals from the high-frequency odd-order non-linear distortion appearing in-band. A more advanced BLA-based DCA can be implemented using the higher-order BLA [32]. With the higher-order BLA, one could obtain a similar result to [31], but for modulated signals, instead of two-tones.