Design of Linearization of Ultra-Linear Power Amplifier

　　With the rapid development of mobile business, especially the development of CDMA and third-generation mobile communication technology, the system has higher and higher requirements for the linearity of power amplifiers. In mobile communication systems, in order to ensure a certain range of signal coverage, we usually use power amplifiers to amplify the signal, and then transmit it through the RF front end and antenna system. In CDMA or WCDMA and TDSCDMA base stations, if a general high power amplifier (usually working in Class AB) is used, the spectrum regeneration effect will be produced due to the influence of nonlinearity. In order to better solve the spectrum regeneration and EVM (error vector magnitude) problems of the signal, linearization technology must be used for the power amplifier. In addition, the power amplifier accounts for about 50% of the cost of the base station amplifier. How to effectively and cost-effectively solve the linearization problem of the power amplifier is very important.

　　1. The proposal of ultra-linear power amplifier solution

　　The traditional method to solve the linearity of power amplifiers is to use power back-off to ensure the intermodulation components of the power amplifier, that is, to ensure that the power amplifier works in the linear range, so as not to affect the signal coverage and communication. Figure 1 shows the curves of the third-order intercept point, 1dB compression point and third-order intermodulation with input power.

　　Figure 1. Output power at decibel compression point

　　As can be seen from the figure, the traditional solution is to reduce the input power. If the input power is reduced by 1dB, the intermodulation component of the system will be improved by 2dB, and so on. In other words, in order to ensure linearity, for CDMA or WCDMA power amplifiers, we can only use a 100W amplifier tube to output 5W power. However, since the tube is designed for 100W, its static operating point is still very high, and the static current is still very large. Therefore, the overall current of the power amplifier will be very large. The large current means that the efficiency of the power amplifier is very low, and a large part of the heat can only be released to the tube and the circuit board. This heat is not only a waste of energy, but more importantly, it will reduce the service life of the chip. In terms of benefits, the price of an amplifier tube that can provide such a high power is very expensive.

　　Based on the above considerations, the intermodulation that can be obtained by simple power backoff is limited. As the power increases further, relying on power backoff still cannot solve the problem. Therefore, a feedforward predistortion design is proposed here to solve the linearity, efficiency and cost problems at the same time.

　　2. Feedforward predistortion amplifier design

　　At present, the more mature and popular super-linear solutions include feedforward technology, pre-distortion technology (including analog pre-distortion and baseband pre-distortion), feedback technology and other methods. Considering that the use of feedforward technology alone has high requirements for the error amplifier and cannot reduce costs and improve efficiency too much, the use of pre-distortion technology alone can improve linearity and efficiency but cannot meet the requirements of super-linearity. Combining the shortcomings of the two technologies, a feedforward combined with pre-distortion technology is proposed here. The detailed principle block diagram is shown in Figure 2.

　　Figure 2 Block diagram of feedforward predistortion solution

　　As shown in Figure 2, the input signal first passes through the directional coupler and passes through the delay line to prepare and cancel the output signal, thereby detecting the cancellation situation, and the other path is sent to the pre-distortion unit (PD) to generate a distorted signal, thereby improving the linearity of the main power amplifier. At the same time, a part of the output coupling of the main power amplifier is cancelled with the delayed main signal, the main signal is removed, and only the error signal is retained. Through the power divider, on the one hand, it is used as a detection of the cancellation effect and thus as a reference for closed-loop control; on the other hand, it is sent to the error power amplifier to amplify and couple with the main power amplifier to cancel the intermodulation signal, thereby further improving the intermodulation. If the improvement effect here is still not ideal and does not meet the super-linear requirement of 70dBc, the feedforward loop can be increased by one level to form a 3- or 4-level loop, thereby improving the improvement effect.

　　The above is only an open-loop solution. Considering that factors such as input power and temperature can affect the cancellation effect, a closed-loop control link must be designed here so that the attenuator and phase shifter in the system can automatically track changes and automatically adapt to adjustments according to changes in environmental parameters, thereby ensuring the overall linearity requirements.

　　The realization of the closed loop is first based on the sampling of several reference points in the entire loop to guide the changes of various constants, including input power, output error power, ambient temperature, main signal and error signal cancellation, and other factors that determine the changes of various parameters. At the same time, the system requires that the response speed of the adaptive algorithm must be within 20ns to ensure that once the parameters change, the overall intermodulation can track the changes in time and avoid the phenomenon of short-term intermodulation degradation.

　　The following will introduce the system implementation methods and core technical issues in each unit.

　　2.1 Pre-distortion generator (PD)

　　The pre-distortion part adopts an analog pre-distortion solution. This solution has been verified by preliminary tests and can improve the intermodulation of 600KHz dual-tone signals by more than 15dB, and the ACPR of 1.28MHz modulation signals by more than 10dB. The overall block diagram of the pre-distortion generation unit is shown in Figure 3.

　　Figure 3 Predistortion generation block diagram

　　The input signal is divided into two paths through a 3dB bridge. The 0° end is used as the main signal and is sent to the synthetic 3dB bridge through a delay line; the -90° end is used as the error signal generation end and then passes through a 3dB bridge. The 0° end here generates a distorted signal. By adjusting the bias of the amplifier tube FP2189, its intermodulation component is very large. The phase is adjusted through a phase shifter to prepare for the main signal to cancel. The -90° end first adjusts the amplitude through an attenuator and then passes through the FP2189 with very good bias adjustment to generate a very good intermodulation signal. In this way, when passing through the -90° port of the synthetic bridge, the main signal and the error signal differ by -180°, so that the part that generates the intermodulation signal is removed from the main signal and only the error signal is retained. By adjusting the attenuator and phase shifter to make its phase differ from the phase of the main signal passing through the delay line by -90°, the phase difference between the main signal and the distorted signal can be achieved by -180° with the help of the -90° end of another 3dB synthetic bridge, which means that the distorted signal is inverted, thereby canceling the intermodulation components of the main signal during the amplification process of the main power amplifier.

　　The generation of the above predistortion signal is the first technical difficulty of this project. However, after two months of experiments, the feasibility of this method has been demonstrated. As mentioned above, the intermodulation of the 600KHz two-tone signal at 2.14GHz can be improved by more than 15dB, and the ACPR of the 1.28MHz CDMA modulation signal can be improved by more than 10dB. 　　2.2 Feedforward Unit

　　The above pre-distortion scheme has been verified by experiments and can improve intermodulation by 15dB and ACPR by 10dB, so that a smaller tube can be used to push out a larger power. However, this still cannot meet the requirements of super linearity (i.e. -70dBc), so the feedforward method is introduced again to further improve the linearity. If the feedforward method is used alone, the power requirement for the error power amplifier is high, because the intermodulation product of the main power amplifier is high, so a large power must be pushed out at the error power amplifier to offset it. Therefore, the price of the error power amplifier is increased and the efficiency is reduced. However, if a pre-distortion unit is added before the main power amplifier, the intermodulation product can be greatly reduced, the requirements for the error power amplifier can be reduced, and the efficiency can be improved.

　　The principle block diagram of the feedforward part is shown in Figure 4

　　Figure 4 Feedforward principle diagram

　　The input signal is amplified by the main power amplifier, and due to the nonlinearity of the main power amplifier, intermodulation components will be generated. The amplified main signal is subtracted from the input signal through a coupler, so that the amplified main signal only has a distorted signal. The distorted signal is amplified by an error amplifier to make its amplitude the same as the intermodulation product amplitude of the main signal, and then adjusted by a phase shifter and attenuator to make it exactly -180° out of phase with the main signal, thereby canceling out the intermodulation products in the main signal and further improving the linearity of the power amplifier.

　　In actual design, if the primary loop cancellation effect is not ideal and fails to meet the ultra-linear requirement of -70dBc, you can consider adding more loops to further cancel the distorted signal.

　　2.3 Closed-loop adaptive unit

　　Both the pre-distortion unit and the feed-forward part can greatly improve the overall linearity of the power amplifier, but through experimental verification, their cancellation effect will be affected by the amplitude and phase of the signal. If the phase difference of the two canceled signals exceeds 2°, and the amplitude difference of the two signals exceeds 5dB, the improvement effect will be very poor. However, since the power amplifier itself will be affected by many factors such as the improvement of ambient temperature and the change of input signal strength, it is necessary to require our power amplifier to automatically adapt to various environmental applications. Therefore, in order to meet the above requirements, the entire system must have an adaptive unit to automatically adjust various parameters according to environmental changes to ensure that the power amplifier works in super linearity.

　　The closed-loop adaptive unit will be the difficulty of the entire project. On the one hand, there are many parameters that need to be controlled, and the available input signals are very few. At the same time, the overall mathematical model is difficult to establish, and the relationship between input and output cannot be described by a mathematical model. On the other hand, the time delay from the input signal change to the output signal is within about 20ns, which requires the entire algorithm to complete the calculation from input to output within 20ns, that is, the real-time performance of the algorithm is required to be very high.

　　The starting point of the algorithm will be to obtain a large number of control voltages of phase shifters and attenuators obtained in different environments through experiments based on the table lookup method, and use this as a sample to design a neural network machine. The neural network is trained through a large number of samples to enable it to quickly determine the control voltages of each attenuator and phase shifter based on environmental variables and input power and other factors. Ultimately, the neural network machine will be implemented in FPGA.

　　To obtain the test data, a program that can operate the signal source and spectrum analyzer will be designed on a PC, the power amplifier will be placed in a high and low temperature box for high and low temperature tests, the input power will be adjusted at the same time, and a search algorithm will be designed to automatically complete the test process through the host computer, obtain a large number of test samples, and train the neural network on MATLAB. And finally implemented in FPGA.

　　The above is just an assumption. The closed-loop adaptive algorithm will be the difficulty of the entire ultra-linear power amplifier. It will take a lot of time to collect information, try various solutions, and finally propose and design a most suitable implementation solution. Therefore, a lot of time and energy may be spent here.

　　3. Ultra-linear power amplifier product implementation solution

　　Figure 5 Pre-distortion + main power amplifier implementation scheme

　　Figure 6 Predistortion + feedforward + main power amplifier solution

　　4. Key technical issues of ultra-linear power amplifier

　　Bandwidth issue.

　　As the frequency increases, both the phase and delay will change, which requires that the frequency response characteristics of each module be consistent across the entire frequency band.

　　Offset problem.

　　The fundamental starting point of pre-distortion and feedforward is the problem of signal cancellation, that is, to make the two signals 180° out of phase. If the error of phase difference cancellation exceeds 2°, the cancellation effect will be greatly reduced. In other words, whether the phase can be matched is the key to determine the effect.

　　Attenuators and Phase Shifters

　　Attenuators and phase shifters are the main adjustment units for predistortion and feedforward. If the attenuator has additional phase shift or the phase shifter brings additional attenuation, the entire system will be uncontrollable. The time delay of the other two is also the key to determining the delay line. After experimental demonstration, although the attenuator and phase shifter built with a bridge can meet the requirements, the additional phase shift and attenuation are too large, and the consistency is poor, which will be unfavorable for production. Therefore, some integrated attenuators and phase shifters are further selected for testing.

　　Adaptive Algorithm

　　As mentioned in the closed-loop control algorithm, the response time and response speed of the algorithm are very critical technical indicators. At the same time, because there are too few input parameters and the entire mathematical model is difficult to establish, the research and development of the adaptive algorithm will be the bottleneck of the entire project, and a lot of time and energy will be spent here.

　　5. Advantages of ultra-linear power amplifier

　　As a very popular technology nowadays, ultra-linear power amplifier has many advantages over the traditional method of designing power amplifiers. It not only can make the linearity of the power amplifier very good, but also can greatly reduce costs, improve efficiency, increase benefits, etc.

　　Cost advantage

　　Although the ultra-linear power amplifier is much more complicated in structure than the traditional power amplifier, and has added additional chips such as attenuators and phase shifters, which may increase some expenses, it can greatly improve the linearity of the power amplifier, that is to say, instead of using a 100w amplifier tube to produce 20w power, now only a 45w tube may be needed to produce 20w. In this way, the expenses saved on the amplifier tube will far exceed the expenses of those attenuators and phase shifters.

　　Advantages in efficiency

　　Traditional power amplifiers all use the power fallback method, which will make the amplifier tube work at a very high static working point but the output power is not large, and a lot of energy will be wasted on static current and heat. The use of super linearization technology can further improve the working state of the power amplifier tube by improving the nonlinearity of the power amplifier, reduce the waste of static current, and greatly improve efficiency. At the same time, it is also equivalent to improving the overall stability.

　　Advantages in high power amplifiers

　　Take a 200W power amplifier for example. For a base station amplifier with four carrier frequencies, four 200W power amplifiers are needed. Because the power is too high, the linearity of the power amplifier cannot be guaranteed. Only one amplifier can be used for each carrier frequency. This requires a series of expenses such as the combiner and power supply system, which makes the whole machine very large. If an ultra-linear power amplifier is used instead, it can be said that one power amplifier can do the job, which greatly reduces the cost and size. This will be very competitive in the market.