With A

Extended Bandwidth and Increased Efficiency Quasi-Doherty Power Amplifier Design Revised Approach For Load Modulation Abstract This paper presents an approach to power added efficiency (PAE) increase for Quasi-Doherty power amplifier (Q-DPA) design. For this aim, active feedback is utilized instead of a passive quarter wavelength transmission line ( 4  TL) usage, which is conventionally used in the DPA schematic. PAE increase can be done by applying an accurate load modulation to the main amplifier (PA main ), especially for technologies in which output impedance of the main power amplifier (Z out,main ) considerably varies in both low and high power regions. Because such precise modulation is still based on a modified TL, this approach suffers from the inherent narrowband behavior of that TL. As a consequence, expecting a wideband DPA may not be satisfied in all cases. To deal with this issue, active feedback is used to play a role in reaching PA main , which is not saturated before, to its maximum efficiency at the highest level of received input power (P in ) in the high power region. Following Z out,main trajectories in power and frequency sweeps simultaneously just by a passive TL are not needed anymore. Still, for the sake of preventing total PAE degradation due to the consummated power by the feedback path’s power amplifier (PA feedback ) should be limited, analytical confinement is provided in this work. A comparison is made between GaAs pHEMT 0.25um MMIC technology-based conventional DPA and the proposed revised approach based-DPA to verify the mentioned approach. The proposed PA shows maximum output power of 33.4 dBm, maximum PAE of 41.6, fractional bandwidth of 11%. The Q-DPA works with a maximum power gain of 24.16.


Introduction
In the era of traveling to untouchable space spots, the nadir of the oceans, wirelessly connected systems are strongly needed. The world of wireless networks is interminably insatiable for developing the current high speed and high-efficiency subsystems for the sake of realizing a modern, power-efficient globe. This future world's perspective entices researchers to introduce modified structures and theories for the communication subsystems. By studying the field of designing PAs, in retrospect, it can be understood several different PA design methods are introduced to increase PAs' PAE, bandwidth, to name but a handful. Such methods comprise envelope tracking [1][2][3][4][5], switching mode class PAs [6][7][8][9][10], harmonically tuned class PAs [10][11][12][13][14][15], Cherix outphasing [16][17][18][19][20], and DPA [21][22][23][24][25]. DPA is one such PA that is still under more scrutiny to subdue the related restrictions of exerted technology and try to reach the best possible performance of that PA [26][27][28][29][30]. This PA comprises two power amplification paths. The first one is called main; its related PA works as class AB-PA and contributes to amplification in the whole input power range. The second path is named auxiliary and is biased as a class C PA. The latter path is only active when the input level power increases. Hereafter we call this power region the high-power region. To prevent efficiency degradation due to saturation of the main power amplifier in the high-power region, load modulation is done at the output of PAmain using a 4  TL. Many developing strategies have been applied to the conventional DPA [25]. The common point of almost all of them is that these designs are based on a 4  TL-based load modulation in which Zout,main trajectory change in power, and frequency sweep are not precisely modeled.
According to [23], revising load modulation can be done based on linear modeling of the variations in Zout,main's trajectory in power sweep at modulation equation. This load modulation modification led to the PAE increase in comparison by the condition in which a conventional modulation was applied to DPA. Since that revised modulation-based DPA does not provide a broadband performance, in order to design a wideband DPA, more consideration should be attended to. In this work, to present a high PAE and wideband DPA in GaAs technology in which Zout,main shows two disparate trajectories in power and frequency sweeps, the strategy of load modulation is modified.
Fulfilling an ideal wideband load modulation that precisely follows Zout,main variation in power sweep leads to significant complexity in design. Therefore, in this work, active wideband feedback is applied to PAmain to reach this PA to its high efficiency in high power region. PAmain design is done by keeping it away from saturation before presenting its highest efficiency. According to the non-saturated status of the PAmain, point-to-point narrowband following of Zout,main variation in high power region is not needed anymore. Expunging the narrowband modulation can pave the way to design a wideband DPA.
Furthermore, bias and size selection of PAfeedback should be made to reach the required Pout feedback while its consumed power is not high. Such a design of feedback path contributes to keeping the PAE as high as possible. Analytical studies of PAfeedback design for PAE's sake are done in the next section, design theory.

Active feedback utilization
According to the conventional DPA concept, in the high-power region, when PAaux contributes to power amplification, 4  TL-based load modulation averts DPA's efficiency roll-off [25].
However, this approach has its own disadvantages: bandwidth degradation due to narrowband behavior of 4  TL, not expecting Zout,main variation with power in which such changes are needed to be applied in a revised load modulation [23], to name but a handful. Load modulation begins Still, it highly depends on 4  TL usage to fulfill the task of modulation. By considering 4  TL inherent narrowband feature and load modulation lockage in multi-stage DPA[Grebini], some modifications for the load modulation approach in some cases are needed. Class AB PAs are designed for providing the maximum efficiency in the highest input power level, and after that, applying higher Pin leads to their considerable efficiency roll-off. In this work, we designed PAmain so that it does not reach its maximum efficiency at the end of the low-power region. Active feedback starts injecting extra power to PAmain's second stage as power increases-whole PA works in the high-power region. By doing so, PAmain provides its maximum drain efficiency when the input power is high.

Efficiency concerns in the presence of active feedback
From another scope, to have a higher efficiency of PAmain and active feedback than conventional AB-class PA in TL-based DPA, more considerations should analytically discuss that hereunder we delved into them.
First, the total efficiency of PAmain and the active feedback should be more than PAmain in conventional DPA. To realize this efficiency priority, the consumed power of PAfeedback should follow a restriction by which presenting higher efficiency of the feedback-based PA is guaranteed.
The following equations mathematically show how this condition can be satisfied.
The restriction of Eq.3 indicates the condition for bias and size selection of PAfeedback, a 6×150 transistor with -1V and 5V bias voltages for its gate and drain.

Feedback junction to the main path
One of the typical conjunction of the feedback path to the main loop is the resistive one. Due to power dissipation in this method, capacitive or inductive ones are preferred. In the case of the capacitive junction, due to the degradation effect of large capacitors in the circuit's performance and realization in MMIC, inductive junction is selected.
Among several configurations of the inductive junction, quasi T-junction is chosen. The insertion of this junction should be so that no significant bandwidth degradation has occurred in OMN of the main path and whole DPA. Since OMNs designing both paths are correlated together, bandwidth degradation prevention due to inductive junction insertion should also be done. To deal with this issue, both networks are designed with a bandwidth margin to ensure that even with inductive insertion, the proposed DPA will provide the desired bandwidth. Figure 1. exhibits the bandwidth of inductive junction, OMNs of PAmain, and PAaux. This figure proves that the DPA is able to provide the aimed bandwidth.

Class of PAfeedback selection
To satisfy Eq.2, designing PAactive in class C seems to be a proper strategy to take. Still, by considering class C PA's inherent narrowband performance, some more considerations for choosing the class in which PAfeedback operates should be scrutinized. For the sake of presenting a wideband performance, PAfeedback can be biased as an AB-class PA, and for controlling its consumed power, the drain voltage of PAfeedback should be as low as possible.
In the aspect of stability, two main issues should be addressed here: preventing oscillation after applying active feedback in the circuit. The other is the transistor's stability. For the sake of coping with undesired oscillation due to feedback presence, prevention of the ring oscillator presence, feedback path mandatorily should enjoy the odd number of PA stages so that total there are even PAs in the feedback loop, PAmain is also included. PAauxiliary. These appropriate low-Q networks present the desired bandwidth (see Figure 1.).

Input power divider
According to the strategy of this work to prevent PAmain saturation during high-power region, input power should be controlled so that the main application path delivers a small portion of power.
According to what proposed in [31], the characteristic impedance of main branch is matched to main path's optimum impedance at low power level to satisfy this condition. The auxiliary branch's impedance characteristic is matched to the optimum impedance of the auxiliary amplification path. By doing so in the high power region, most of the input power is assigned to the auxiliary power amplification path.

Phase compensation
In the conventional DPA, a 4  TL at the input of the auxiliary path of amplification is placed to compensate for the 90° difference in the main path phase. This difference is due to the 4  TL contributing to the conventional load modulation. In this work, by modifying the strategy of modulation, the second 4  TL utilization is not needed anymore. Still, due to the difference in these two paths of amplification, both paths' consequent phase differences should be compensated.
According to [21], the difference in the phase of main and auxiliary paths is compensated in designing the matching networks of main path.

Bandwidth
One of the primary concern in this design is providing wideband performance. To satisfy the desired 11% fractional bandwidth with a revised approach for load modulation, all sections of DPA, such as input power divider and matching networks, should provide the wideband performance. Lumped model of the Wilkinson power divider presents the aimed 11% bandwidth.
Besides, the appropriately low-Q LC networks can serve as wideband matching networks. Figure   1. shows the bandwidth of each section and the whole DPA. It can be easily understood that the desired bandwidth is reached.

Power combination concerns
The low impedance of Zout,main, which is due to its saturated status, in the high-power region can destructively affect the power combination of DPA's output port. According to DPA's concept [25], to address this issue, the 4  TL-based load modulation plays a role. In our case, in which saturation of PAmain is prevented, the output impedance of this PA doesn't engender any concern.
It should be noticed the only thing worthy of attention is the output matching networks of both paths should be designed so that the output port of our proposed Q-DPA be matched to standard 50 Ω. Figure 2. shows the output impedance of main and auxiliary paths of amplification. The resultant impedance is matched to the mentioned standard one. Figure x1. It also indicates the bandwidth providing by these networks and confirmed their wideband performance.  Therefore, the design of harmonic rejection networks is neglected.    For a better understand, the performances of both designs in power sweep are shown in Figure 9.
and Figure 10. These figures clearly confirm that using the revised approach helps in presenting a high PAE power amplifier. To indicate more about the performance of the Figure 11. and Figure 12 show output powers of PAmain, PAaux, the proposed Q-DPA and its related power gain.

Conclusion
This work presented the design of a wideband Q-DPA. Considering Zout,main, the main variation in power sweep, which makes realizing a proper load modulation hard, is done in the load modulation's revised strategy. That modification is grounded on active feedback. The feedback made PAmain to be reached its maximum PAE in the highest level of power. Therefore there is no concern about making a delicately ideal load modulation by which increased PAE can be obtained.
To draw an accurate conclusion, a comparison between the proposed DPA and that of the conventional DPA in which a 4  TL-based load modulation is done is made in this work. Both In this work, this task is done analytically. Besides, PAfeedback's bias and size selection are also made to provide desired Pout,feedback, and satisfy the limit of consummated power.