Energy efficient 5G transmitters – thanks to Doherty designs...
01 June 2020
High efficiency is a key factor in the development of profitable 5G base station transmitters. The use of Doherty amplifiers is a viable and popular solution for achieving the corresponding output power.
This tutorial was originally featured in the June 2020 issue of EPDT magazine [read the digital issue]. Sign up to receive your own copy each month.
Here, Dr Patrick Agyapong, Product Manager for Signal Generators & Wireless Communications and Gareth Lloyd, Application Developer & senior expert in RF, µ-Wave/mm-Wave frontends and 5G/radar industries at electronics & wireless T&M experts, Rohde & Schwarz tell us how R&S is providing developers with a new, flexible test solution that allows them to take full advantage of all the strengths of this design.
Power amplifiers in base stations constitute 30%-60% of the costs and 20%-60% of the energy consumption of the entire transmitter infrastructure, making their optimisation a major concern for operators. For RF power semiconductor and infrastructure vendors alike, a key focus is therefore on increasing the efficiency of base stations and radio relay transmitters.
The most effective ways to reduce amplifier power dissipation have been known for a long time, but only in the last few years has it been possible to implement them in practice. That’s because energy efficient operation near the saturation limit comes at the cost of nonlinearities that have to be compensated by upstream or downstream measures, such as digital predistortion, that are only possible with advanced circuit technology.
On the infrastructure side, conventional Doherty power amplifier designs have established themselves in the market, whereas envelope tracking dominates on the device side. In envelope tracking, the supply voltage of the power transistors is dynamically adapted to the signal envelope. The Doherty method splits the input signal into two parallel amplifier paths. The main amplifier handles the base load and is permanently operated at energy efficient full modulation. If the input signal exceeds a certain level, the second amplifier switches on and handles only the load peaks. This task sharing is particularly promising for digital signals with their high crest factor. Many years of R&D effort were necessary to make this theoretically attractive concept possible in practice. And the workload will increase even further when it comes to using the design for high microwave frequencies and wideband applications, such as 5G or satcom.
Every implementation of a Doherty amplifier is always an approximation, since perfect Doherty operation is a theoretical concept that cannot be achieved in practice. But even with rough approximations, performance is usually better than the baseline class AB amplifier at low operating frequencies and small bandwidths. However, the technique becomes less tolerant to rough approximations and hence less efficient as operating frequency, bandwidth and output power increase. New and repeatable design processes are required to harness the full potential of the technique. The R&S SMW200A vector signal generator, together with the R&S FSW signal and spectrum analyser, effectively supports the development of such designs.
In the Doherty design, effects on both the input side and the output sides play an important role in the overall concept. The overall efficiency of the design is determined on the input side – and thus ultimately how high or low the energy costs actually are. The output side, with the combiner, determines the maximum potential.
The signal to be amplified must be split in order to drive the two amplifier paths. There are different approaches to implementing this split: classic implementations perform this split in the analogue domain; dual-input solutions perform the split in the digital domain. Various studies suggest several advantages of the dual-input split process compared to a classic Doherty implementation. Among other things, up to 60% more RF output power, 20% less energy consumption and 50% wider bandwidth can be realised. In between, there are dispersive splitters that replace the fixed analogue splitter. These allow variable tuning of the signal distribution depending on operating modes and operating points, such as frequency and level.
But how does the developer find the optimal tuning? The challenges remain the same regardless of whether the amplifier is being developed for 5G and satcom transmitters or for other applications that require demanding high performance and reproducibility. Developers want to maximise the performance of their amplifiers under the intended operating conditions as efficiently as possible. Unfortunately, these are mutually contradictory goals: an improvement in performance results in a loss of efficiency, and vice versa. It is necessary to find an operating point and a parameter set that minimise these costs.
In order to do this, the sensitivity of the design to frequency, phase and level variations in the amplifiers paths must be known. The usual development process depends heavily on reference designs and manual fine-tuning of prototypes, making it difficult to explore beyond a few local performance and efficiency optima in the multidimensional parameter space. Due to a lack of insight into the sensitivities of the chosen design, the designer typically specifies conservative metrics to accommodate part-to-part variations in a production environment. The result is usually suboptimal, because the true potential of the design was not fully explored and specified. With the new solution from Rohde & Schwarz, developers understand the optimum interaction between the two Doherty paths and can choose their design accordingly.
Application example: Doherty amplifier development with the R&S SMW200A & the R&S FSW
The Doherty design setup consists of the R&S SMW200A vector signal generator and the R&S FSW signal & spectrum analyser, and includes two input ports, matching networks for inputs and outputs, active components, bias networks and a Doherty combiner (see Figure 1). By designing the Doherty prototype as an instrument with two inputs, a more accurate picture of performance limits, trade-offs and expected reproducibility in a production environment is obtained.
The test setup requires two signal paths, whose signals can be varied relative to each other. In addition to using precise, stable and repeatable amplitudes and phase shifts, it is recommended to be able to transmit at least one of the signal paths nonlinearly. All this can be achieved with the R&S SMW200A. The measurement algorithm can be either faster or more thorough, depending on whether it is programmed to find the optimum values for the desired parameters or to characterise a wide range of parameters. In a simple experiment, for example, the optimum measured values, as well as their relative amplitude and the phase compensation values, can be determined. An example of a more complicated application would be a detailed search as a basis for performing a sensitivity analysis or to determine a solution.
The possibilities of such an approach are demonstrated by a test setup in which a 3.5 GHz GaN Doherty amplifier is to be developed for a 5G base station. The prototype design allows direct access to the inputs of the main and external amplifiers. The R&S SMW200A two-path vector signal generator produces two input signals for controlling the GaN amplifier. The dependent quantities are measured by connecting the individual RF output to the R&S FSW signal and spectrum analyser. An R&S HMP power supply, which measures the DC current consumption, provides the DC power supply for the instruments. The amplifier is stimulated using differential linear and nonlinear signals, with the first signal passing through the input power, amplitude and phase. The nonlinear tests use a variable, magnitude-dependent function at two frequencies. After measuring the output power, the ratio of peak power to average power, adjacent channel leakage power (ACLR) and power consumption, MATLAB analyses the results.
During linear measurement analysis, the efficiency at a given power level and the saturated power versus amplitude and phase differences are displayed (see Figure 2), with Figure 2c presenting the worst efficiency and output power. The operating mode of the classic Doherty design is a quasi-constant amplitude/phase split. Efficiency and saturated power for these amplitude/phase values are obtained by extracting the worst performance at the test frequencies.
Selecting the nominal amplitude/phase split makes it possible to add an artificially generated interference signal which simulates the maximum natural variation of production series components. The bulk effect of these part-to-part variations can be observed from a look-up table, shown in Figure 3. Figure 3a shows the drain efficiency and saturated output power at two frequencies. Figure 3b shows the estimated parameter scatter of saturated output power and drain efficiency, compared to the nominal values for the same two frequencies. The cumulative parameter scatter, which summarises the results from the two frequencies, can be found in Figure 3c. Paradoxically, the largest value for the part-to-part variation in this case occurs in the target variable, efficiency.
An alternative approach to the design of the input splitter can reduce this deviation. A dispersive input distributor is used for this purpose. Thanks to amplitude and phase differences at the two design frequencies, the contour diagrams shown in Figure 3a can be “superimposed” on each other. Using the same data for part-to-part variation with this dispersive splitter design leads to a better result (see Figure 4), with a higher efficiency and a lower standard deviation.
Using a digital Doherty approach, the two paths can always be operated at the optimum amplitude and phase point. By splitting the signal for the two paths in the digital range, further optimisation with different pre-distortion of the two Doherty channels is possible. Figure 5, which illustrates the distributions of average power versus efficiency and peak envelope power (PEP) versus average power, shows an example of such optimisation potential.
The saturated output power is 1.7 dB higher than that of the conventional Doherty amplifier, resulting in a 48% increase in performance. This indicates that 1.2 dB of the improvement – or 32% more power – is due to better adjustment of the amplitudes and phases of the signal paths. The increase in average power is also associated with an increase in efficiency by 5 points (from 44% to 49%), or alternatively, instruments that are 48% smaller can be used and still achieve the original target output power.
The R&S SMW200A and R&S FSW application example illustrates how new T&M approaches lead to new insights. They are the basis for designing an ideal input-split model that meets all the requirements and operating conditions.
The Doherty amplifier is a commonly used approach to optimise the efficiency of output amplifiers in 5G infrastructures and other systems such as satcom. With the R&S SMW200A and the R&S FSW, Rohde & Schwarz offers a powerful solution for optimising working conditions in terms of frequency and level. Based on the new insights, designers can choose the optimum implementation for their application, such as a classic analogue Doherty splitter or a fully digital implementation. The optimal solution is probably somewhere between the two. Replacing an analogue splitter with a dispersive one makes it possible to get much closer to optimum performance without increasing the complexity of the output. This is an essential prerequisite for large-scale use, such as in 5G base stations.
Contact Details and Archive...