Since the first year of 5G in 2019, 5G construction has been in full swing in the past three years. In the rapid development of 5G, the RF front-end chip will benefit the most. According to Yole's forecast, by 2026, the total global RF front-end market will reach US$21.67 billion, a seven-year growth rate of 74.6% compared with US$12.41 billion in 2019.

The RF front-end chip is the core device of wireless communication. It refers to the functional module after the antenna and before the transceiver. Because it is located at the front end of the communication system, it is called "RF front-end", which generally includes power amplifiers, filters/duplexers. , switches, and low-noise amplifiers.

Figure: The composition and function of the RF front-end

The RF PA (Power Amplifier, power amplifier) ​​is an important device in the RF front-end. Its performance directly determines the signal strength, stability, power consumption and other important factors, and determines the user experience. Its core parameters include gain, bandwidth, efficiency, linearity, maximum output power, etc. Many balanced performance indicators test design capabilities.

Figure: The position of RF PA in chip design

With the development of 5G, the technical performance requirements of RF PAs have been raised again, requiring PAs to have higher operating frequencies, higher power, and larger bandwidths. At the same time, the advent of modularization also requires PA design to meet high integration. Modular requirements.

As a result, PAs with different architecture designs have attracted everyone's attention. In addition to the single-ended PAs that are commonly heard, in recent years, various PA architectures such as Push-pull PA, Balance PA, and Doherty PA have appeared in mobile phone applications. What do these PA architectures refer to? What is its principle? What are the characteristics of its application? This article will discuss the above issues.

PA's Design Philosophy

PA is the abbreviation of Power Amplifier (power amplifier), which is a type of amplifier mainly used for power amplifier output. Unlike other amplifiers, such as low noise amplifiers and driver amplifiers, the main purpose of power amplifiers is to squeeze as much RF signal as possible from DC instead of just focusing on gain.

In order to achieve high enough power output, the output of the PA abandons the conjugate matching for maximum signal "transmission" and adopts load line matching for maximum power "output". In addition, in order to achieve controllable DC consumption, PA design must consider efficiency optimization design.

Load line matching maximizes both the voltage and current swings of the PA, that is, the goal of maximum power output. For the concept of PA load line design, please refer to the article "Load-line and Load-pull of 5G PA" and click the blue word to jump.

Figure: Load Line Theory vs. Voltage/Current Swing

The above are only the design ideas and methods of a single PA unit. In practical applications, the current and voltage swings of a single device may not be enough to meet the overall requirements of the PA; it may also be necessary to combine the architectural design to achieve improved efficiency and improved standing waves. need. At this time, it is necessary to make improvements on the PA architecture side.

Architecture Design of PA

The core goal of PA is "power", and the architecture of different PAs is also based on the concept of "power synthesis". In power combining, power combining can be divided into "simple power combining" and "special power combining" according to different types of power combining.

"Simple power combining" refers to combining multiple low-power devices until the combined power is sufficiently large. Simple power synthesis methods include: current synthesis, voltage synthesis, and power synthesis.

"Special power synthesis" refers to using a relatively special synthesis method to complete some special characteristic designs during synthesis. Common synthetic methods are Push-pull (push-pull), Balance (balance), Doherty (Doherty) and so on.

The design of several PA architectures will be discussed below.

simple power combining

current synthesis

Current synthesis is the simplest power synthesis method. The realization method is to connect multiple smaller devices in parallel to connect the current in parallel.

Figure: Current Synthesis of a Power Amplifier

In terms of implementation, current merging is to connect the bases, collectors, and emitters of multiple transistors respectively in the layout. Because of its simplicity, it is widely used in the design of power amplifiers.

The figure below is a typical GaAs power amplifier power stage layout. The power output stage of the last stage is composed of 4 columns of power arrays (Power Array), each power array contains 5 power cells (Power cells), a total of 20 power cells , together to complete the final power amplifier output.

Figure: Typical Power Amplifier Power Stage Layout

Although current merging is simple to implement, there are a number of points to be aware of when designing:

A single power array is less likely to be too long to ensure currents are superimposed in phase

When combining different arrays, pay attention to the symmetry of the traces to ensure the same phase when the current is combined

The traces used for current merging are relatively wide, and attention should be paid to reasonable design to reduce the parasitic capacitance effect caused by the traces

Current combining is widely used in power amplifiers due to its simple design and convenient implementation.

voltage synthesis

In addition to the current merging method, there is also a simple merging method that combines the voltages. The advantage of voltage combining is that it can improve the optimal output impedance point, and can achieve lower impedance matching loss when doing impedance matching.

The figure below shows a typical implementation circuit of the voltage combining method, as well as an implementation example [1].

Figure: A typical circuit with voltage combining, and a typical implementation example

The voltage combining method is suitable for scenarios where the power supply voltage is much greater than the withstand voltage of the device, such as a high-voltage power supply environment; or when low-voltage devices are used for PA design.

The GaAs HBT breakdown voltage VBCEO used in the current mobile phone PA design is generally between 10 and 25V, and a single device is suitable for a DC Vcc bias voltage within 5V. In 2016, companies in the industry tried to increase the battery voltage from 3.8V to about 11V, and then supply power to the PA design with Cascode voltage combination, in order to achieve the design purpose of improving the optimal output impedance, thereby improving the overall efficiency of the PA. This solution was successfully adopted in some flagship phones around 2016. However, due to the need for an additional boost circuit, which affects the versatility of the solution, this solution has not been popularized in the industry.

power combining

In fact, the combination of current and voltage is for power synthesis. However, there are some power synthesis methods that cannot strictly be said to be current synthesis or voltage synthesis, so they are classified as "power synthesis".

A typical power combining method is to use a power combiner for power combining, and a Wilkinson power combiner (power splitter) is a simple power combiner. The Wilkinson power combiner and its designed power combining PA are shown in the following figure [2]:

Figure: Typical Wilkinson power divider design, and its designed power combining PA

Since the Wilkinson power combiner has the feature that three ports can be matched at the same time, the power synthesis PA designed by the Wilkinson power combiner is simple in design. It only needs to design each PA separately, and then perform power synthesis separately. However, since the Wilkinson power combiner requires two sections of λ/4 transmission lines, and a 100 Ohm resistor needs to be connected in parallel between the two branches, it is not economical to implement Layout, and cannot be effectively used in designs with limited area.

In order to improve the area restriction of the Wilkinson power combiner, the direct two-in-one combination matching (Binary Combine) method is adopted in some designs. Although this method cannot achieve the complete matching of the three ports, due to the Small area, easy to implement, has been widely used in MMIC design. The figure below shows the PA design implemented by the direct 2-in-1 merging method [3].

Figure: Power amplifier realized by two-in-one direct combining

In addition to the research on power combiners, some researches also use Transformer or direct space power combining for power combining [2].

Figure: Schematic diagram of power combining using Wilkinson, transformer, and space combining

special power synthesis

In "simple power synthesis", the idea of ​​power synthesis is to simply combine power to complete the power output of "1+1=2". In addition to the simple synthesis of power, special designs can also be added to the power synthesis to achieve more complex characteristics while completing the power synthesis.

Some common special power synthesis methods are Push-pull (push-pull), Balance (balance), Doherty (Doherty) and so on.

Push-pull PA

Push-pullPA is generally translated as push-pull PA in Chinese. The design of Push-pull PA is to combine two amplifiers with forward and reverse conduction respectively to complete the whole cycle waveform synthesis output. In this way, each individual PA can be designed as a high-efficiency Class B working mode, and the PA as a whole has high efficiency.

Figure: Push-pull PA principle diagram

The Push-Pull PA shown in the figure above is composed of NPN and PNP transistors, which are responsible for the signal conduction of the positive half cycle and the negative half cycle respectively. In actual design, because PNP type bipolar transistors are generally not easy to make high-speed, and in the implementation of integrated circuits, the general Epi (epitaxial layer) contains only one type of transistors, so double NPN type transistors are often used in RF design. circuit.

At this time, the input and output need to use an unbalanced-to-balanced conversion circuit, that is, a balun (Balance to Unbalance), to reverse the two signals. The Push-pull PA designed with balun and dual NPN transistors is shown in the figure below. Different from the push-pull PA designed with NPN+PNP, there is a physical ground, and both power amplifiers flow with the ground as a reference. The push-pull PA designed with a balun and dual NPN transistors flows differentially between the two channels, and the radio frequency is The middle point between the two is the virtual reference ground.

Figure: Push-Pull PA with dual NPN transistors and balun design

It should be noted that the Push-Pull PA can not only use two Class B PAs to improve the PA efficiency, but also combine two Class A PAs to improve power synthesis. When designing with Class A, the efficiency of a single PA does not improve, but the output power synthesis increases. The schematic diagram of power combining using two Class A PAs is shown in the figure below.

Figure: Power synthesis Push-Pull PA with Class A PA design

The balun is an important component of the Push-pull PA. The balun is a circuit that completes the mutual conversion between balanced signals (differential signals) and unbalanced signals (single-ended signals). On the balanced signal side, the signal is transmitted differentially with a phase difference of 180°; on the unbalanced side, the signal is ground-referenced and transmitted single-ended. The two-wire transformer winding method is a common balun implementation method. It adopts the form of double-winding coupling to realize the mutual conversion from unbalanced to balanced signals [4]. In addition, the ratio of the coil can also be changed to realize the transformation of different impedances.

Figure: Balun realized by two-wire transformer winding method

In chip design, the coupling between metal coils is usually realized by means of multi-layer metal edge coupling (Edge Couple) or broad side coupling (Broad Side). The figure below shows the implementation of a transformer balun in an integrated circuit [5].

Figure: Implementation of a transformer balun in an integrated circuit and its equivalent circuit

In order to generate a symmetrical differential signal, the balun generally pays attention to the symmetry of the coil winding in the design; in addition, the two amplification paths of the Push-pull PA also need to be symmetrically designed, which makes the Push-pull PA easier to identify:

There are two symmetrical PA amplification paths

Symmetrical wound baluns before and after the two PA amplification paths

The figure below shows a typical Push-pull PA chip design layout [6].

Figure: Typical Push-pull PA chip design

Balance PA

BalancePA is generally translated as balanced amplifier, which is another special power synthesis method. The Balance PA is the same as the Push-pull PA, and it also uses two PAs to combine power. However, unlike Push-pull PA's 180° power distribution and synthesis, Balance PA uses 90° power distribution and synthesis. The figure below shows the design block diagram of Balance PA.

Figure: Balance PA Design Block Diagram

The biggest feature of BalancePA is that as long as the two PAs are completely symmetrical, the reflected signals of the two PAs will be completely canceled at the input and output ports, so that the input and output have better VSWR in a wide range, which is suitable for S11 /S22 Scenarios with special needs. The cancellation principle of the reflected signal in Balance PA can be represented by the following figure.

Figure: The improvement effect of the balanced amplifier on the input standing wave

Because of the above characteristics, Balance PA has good S11/S22 characteristics. The article [8] shows that the Balance PA architecture can optimize the S22 of the PA with a center frequency of 1.5GHz from -10dB to -35dB. S22 below -15dB is achieved in the GHz range.

However, since Balance PA's cancellation of S11/S22 and power synthesis depend on the precise 90° phase shift of the coupler, and the phase shift accuracy of the coupler is related to frequency, Balance PA can only be controlled effectively within a certain frequency range. According to the article [8], beyond 1.2~1.7GHz, S11/S22 and S21 of Balance PA deteriorated compared with single-ended. At 1.0GHz and 2.0GHz, S22 deteriorated by 5~10dB, and S21 deteriorated by 5%. ~7dB.

Figure: Characteristics comparison between Balance PA and single-ended PA

Because Balance PA shows good S11/S22 near the offset point and has good load insensitivity characteristics, Balance PA is also called Load InsensiTIve PA (Load Insensitivity PA) in some industry companies, or LIPA for short.

The key component of BalancePA design is 90° power distribution and combiner, which can be realized in many ways, which can be designed by phase shifter, directional coupler, or 90° quadrature hybrid network. . The figure below shows the Balance PA design [9] that uses phase shifting to achieve 90° power distribution and synthesis.

Figure: Balance PA design using phase shifting to achieve 90° power distribution and synthesis

In the microwave/millimeter wave frequency band, if lumped inductors and capacitors are used for phase shifting, the device value is small and it is difficult to realize. Generally, in the microwave/millimeter wave frequency band, a 90° coupler is used to realize 90° power distribution and synthesis. The figure below shows the Balance PA design using a directional coupler to achieve 90° power distribution and synthesis [10].

Figure: 90° power distribution with directional coupler

Designed with Synthetic Balance PA

The peculiarities of BalancePA design and characteristics can also be used to identify Balance PA:

There are symmetrical two PA design paths

Has asymmetric power distribution/combining network

Test S11/S22, there is a better S11/S22 near the offset point

Doherty PA

DohertyPA is a framework that has attracted much attention in the use of mobile radio frequency PAs in recent years, and it is very topical. One of the reasons why Doherty PA is eye-catching is its high efficiency, which can well alleviate the power consumption increase caused by 5G power increase.

DohertyPA's optimization of efficiency is through the "dynamic load modulation effect", which relies on "two PAs working in different states cooperate with each other, so that the load of the PA changes, thereby optimizing the fallback characteristics of the PA". Although this is explained in textbooks, the load of one PA can be changed by the working state of another PA. It always sounds like some black technology, which deepens the mystery of Doherty PA.

DohertyPA has been the focus of mobile phone PA design in recent years, but it is not a new architecture developed recently. Doherty PA was invented in 1936 by William H. Doherty, an engineer at Bell Labs, a prestigious laboratory, nearly 90 years ago. The original research goal of William H. Doherty was to develop a high-efficiency transoceanic transmission power amplifier on the order of KW. After the invention of Doherty PA, since the matching application scenario has not yet appeared, in most of the 20th century, only in AM transmission There are some applications in the machine.

Although Doherty PA was not widely used for a long time after its invention, it was all a matter of time. In 1990, with the rapid development of global mobile communications, there was a strong demand for efficient and high-power PAs. At the same time, the development of silicon-based and III-V semiconductor technology and the linearization technology brought by digital signal processing technology have provided a solid application foundation for the application of Doherty PA in mobile base stations. Doherty PA technology has developed rapidly on the base station side, and currently almost dominates the entire macro base station PA market.

The core principle of DohertyPA is the "load modulation" effect, and the basic principle is shown in the figure below. If a load R is excited by two sources, namely a voltage source and a current source, the impedance seen by the voltage source V1 can be expressed by Ohm's law as:

If the output current I2 of the current source changes, the impedance seen by the voltage source V1 will also change. That is, the impedance seen by V1 can be controlled by the size of the current source I2, which is the basic principle of "load modulation".

Figure: Load modulation effect under voltage source and current source excitation

In the actual PA design, the PA is generally equivalent to a current source. Therefore, in the actual design, it is necessary to add an impedance inversion network to convert the current source into a voltage source. In addition, in order to meet the needs of impedance matching and phase alignment, compensation lines will be added to complete the design of the entire Doherty power amplifier.

Figure: The principle and specific implementation of Doherty PA

DohertyPA consists of two PA pathways, the Carrier pathway and the Peak pathway. When working at low power, only the carrier power amplifier is turned on, and its load line is maintained at a high position to maintain high efficiency; at high power, when the peak power amplifier is turned on, the carrier power amplifier load line is modulated to a lower position to generate higher power output . In this way, the efficiency is improved when the power is backed off. (For the relationship between load line and output power, see the article "Load-line and Load-pull of 5G PA", click the blue word to jump).

In fact, when analyzing the work of Doherty PA, it is not as simple as the above description. In the Huizhi Micro PA Q&A group, Doherty PA design experts said: The derivation of Doherty PA needs to be considered in three dimensions of power, voltage and impedance. To derive the Doherty theory, we must first have a concept, which is to decompose the Doherty as a whole. From the perspective of power, there are two nodes, one is the fallback power, and the other is the saturation power. From these two points, analyze the two tubes. The impedance changes, power changes, voltage changes.

The figure below shows a typical Doherty PA efficiency curve. This design is a 60GHz Doherty PA with a saturation power of 20dBm. It can be seen that the efficiency of the Doherty amplifier is 1.45 times higher than that of the Class-B amplifier, and 2.9 times higher than that of the Class-A amplifier [13].

Figure: Doherty's Efficiency Improvement Characteristics at Power Backoff

In mobile applications, Doherty PA is not common. The main reason is that the mobile phone is used in a complex environment and needs to support many frequency bands and modes. However, Doherty PA is load-sensitive, narrow-band, and requires powerful algorithm support, which makes the traditional DohertyPA incompatible in mobile phone applications. The comparison between the application environment of the mobile phone and the base station side is shown in the following table.

Figure: Difference between base station and mobile phone application environment

In recent years, with the continuous increase of PA power consumption in 5G mobile phones, DohertyPA has been reconsidered whether it can be applied to mobile phone applications. Due to the complexity of mobile phone applications, it is necessary to consider meeting the system indicators under high and low temperature and different antenna standing wave ratios. When DohertyPA is designed to meet these indicators, it needs to sacrifice the characteristics of back-off efficiency optimization to compromise compatibility with the above mobile phone applications. characteristic.

The figure below is an example of the Doherty PA design presented in the article [14].

Figure: Doherty PA Design Example

In actual analysis, Doherty PA can be identified according to the characteristics of Doherty PA:

Consists of at least two power amplifiers

The two-way power amplifiers are in asymmetrical state, either by design asymmetry, or by asymmetrical bias state

The power synthesis part is an asymmetric structure

Test S11/S22 to differentiate Doherty PA from Balance PA

The above different power combining architectures have different applications, combining special power

Mixed design of multiple architectures

Different PA architectures are not either one or the other, but can also be combined with each other. For example, in each PA unit of Doherty PA or Push-pull PA, the structure of current combining or voltage combining can be selected for design.

Even different special power synthesis can be combined with each other, for example, Doherty PA's Carrier and PeakPA can choose Push-pull PA. Even two Doherty PAs can be combined to design a Balance PA to reduce the load-sensitive characteristics of the Doherty PA.

The figure below shows the PA designed in [15], which utilizes a Wilkinson power divider, a 90° quadrature coupler, a differential amplifier, and a Doherty PA architecture.

Figure: PA design with multiple architecture combinations


This paper briefly organizes the common PA architectures in engineering, and introduces the design ideas, architecture features, and main implementation methods of PAs with different architectures.

In addition, it should be noted that there is no difference between the advantages and disadvantages of the PA architecture, only whether it is suitable or not. Only by understanding the current needs and constraints, and understanding the characteristics of different PAs, can you choose the right PA architecture.

Welcome to leave a message, do a more in-depth discussion of the PA architecture, and deepen the understanding of different architectures of PA.

In the process of writing this article, I have received the guidance of many industry experts, and I would like to express my gratitude!

Reviewing Editor: Tang Zihong

Leave a Reply

Your email address will not be published.