Today’s designers of cutting-edge phones face the same challenge that their predecessors wrestled with: extending the time between battery charging. To succeed now, these designers must build efficient front-end systems offering the high levels of linearity demanded by digital networks operating with complex coding schemes. There are many options for fulfilling these requirements, and they all have their pros and cons, argues Chris Novak, General Manager of 3G/4G Solutions at RFMD.

We are in the midst of a connectivity and mobility revolution. This is driving a dramatic change in communication technologies throughout the world, and the engineers that are responsible for networks and the development of mobile devices are facing tremendous technical challenges.

These challenges are spurred by soaring sales of hand-held smart devices, which are ramping up the number of high-speed data connections. According to Cisco, wireless data usage is going to increase at a compound annual growth rate of almost 80 percent through 2016, as more consumer use handheld devices for social networking and entertainment. This explosion of wireless traffic will be driven by the uptake of sleek-form-factor devices that combine seamless connections with an increase in time between charges.

To keep pace with this rocketing demand for data, smart devices and networks are migrating to broader frequency ranges, wider signal bandwidths and higher data rates. This is fuelling innovation in the RF front-end industry, with engineers developing new designs for making best use of the limited energy stored in the battery.

RF system challenges

Efforts to increase data rates and improve the network’s spectral efficiency have included the introduction of new standards employing advanced coding techniques, such as orthogonal frequency-division multiplexing (OFDM). One of the downsides of these complex coding techniques is that the modulations can lead to a significantly increased variation in the amplitude of the modulated signal – this is needed to encode more information for each transmitted symbol (see Figure 1). In addition, improvements in spectral efficiency diminish the efficiency of the RF transmission system. This reduces in-use time for the mobile device, and also heats it up.

 



Figure1. An example showing modulation constellations with increasing amplitude variation.

Increasing spectral efficiency is only going be part of the solution, due to the relentless pace of growth of wireless data. To help to meet this demand, additional frequency spectrum is being allocated to networks. However, these allocations to digital networks are haphazard. They are regulated by independent governments, and the decision makers are not concerned with designing a cost-effective global radio system. What is happening is that the frequency bands being given over to digital networks differ greatly from region to region. This uncoordinated approach means that smart devices have to cope with an increasing number of frequency bands of varying bandwidths – there are now more than 36 cellular transmit bands of various bandwidth, ranging from 695 MHz to 3800 MHz. Compounding this issue are the incremental, yet challenging, RF specifications for ensuring that digital networks can operate without interference with other wireless systems, such as public safety bands, global positioning systems, and wireless LAN (WiFi).

One of the demands that the latest digital network modulation schemes place on smart devices is the requirement for linear amplification. To ensure a high-quality wireless network, the RF transmit system must accurately and proportionally amplify the input signal – significant distortions in either phase or amplitude cannot be tolerated. Excessive distortion has two unwanted consequences: Data errors in the transmitted signal; and a spilling of the transmission energy into other bands, which results from modulation products. In the latter case, this leads to interference with other data devices.

In general, improvements to the spectral efficiency of the modulation scheme increase the crest factor or peak-to-average ratio (PAR) of the signal (see Figure 2). PA efficiency peaks when this amplifier is at or near its compression point (see Figure 3 for the compression curve and efficiency of a typical power amplifier). However, when PAR increases, the PA spends more time operating further from compression. In other words, efficiency is compromised to maintain sufficient linearity and ultimately preserve peak amplitudes.



Figure 2. Increase in peak-to-average ratio (PAR) for common digital network modulation schemes.



Figure3. The compression curve and power-added efficiency (PAE) for a typical power amplifier showing the effect of greater PAR.

Transmission: The big battery drainer

When smart devices are transmitting, especially at higher power levels, the power amplifier draws a great deal of energy from the battery. The time that the smart device can go between charging, which is also known as the battery life, is governed by the efficiency of the amplifier portion of the RF front-end. To extend battery life, engineers are developing amplifier front-end designs that excel in efficiency and linearity, so that the PA can operate closer to compression and thus hit higher efficiency.

There are several approaches for meeting these goals. One is pre-distortion – making the system compensate for PA non-linearities by applying inverse non-linearities to the amplifier’s input signal, so that the PA operates closer to compression. Alternatively, one can use ‘feed-forward’, a technique based on correction of non-linearity by subtracting an estimate of the non-linearity at the PA output. Yet another option is ‘feedback (Cartesian or Polar)’: Detecting the output signal, comparing it to the desired input signal and correcting the input. And last but by no means least, there is ‘envelope elimination and restoration’ (EER). This involves amplification of a constant envelope signal featuring phase information, onto which superimposed amplitude information is produced by modulating the amplifier’s power supply.

At RF Micro Devices, which is headquartered in Greensboro, NC, we believe that envelope tracking (ET) – a variation of EER with additional back-off to guarantee linearity – is rapidly rising to the top as the most effective power optimization technology for higher power RF transmission. One of its greatest attributes is that it works well across a wide frequency range, a wide bandwidth and across multiple modulations. This makes it well suited to satisfying the needs of the global smart device market. However, it’s worth noting that in some applications the best performance might result from a combination of technologies previously discussed (see Figure 4 for a graphic representation of relative efficiency improvements for several methods).



Figure 4. A graphic representation of PAE improvements for several enhancement methods, neglecting conversion efficiency associated with implementation.

Throttling back

In modern digital networks, power control schemes reduce the total transmit power during the times when full power is not necessary. The benefits are not limited to saving energy, as this approach also minimizes unnecessary interference, leading to a higher system capacity. The key to working well in this type of network is to design the RF front-end for efficient transmission over a large dynamic range of output powers.

There are many ways to enhance RF efficiency over a large dynamic range. Two of the most obvious are to turn off the final PA stages when the power is low enough to be delivered by just the driver stages, and to trim the bias current at lower powers to cut current consumption. Alternatively, engineers can adopt a Doherty configuration, tuning the main amplifier to favourable (high) impedance at low power and turning on an auxiliary amplifier at higher powers to maintain correct impedance. Another option is to employ load manipulation techniques, such as ‘chain matching’ or ‘load switching’, with favourable impedances maintained over a wide power range through the switching in and out of matching elements. And there is also average power tracking (APT) or multi-state power management. With this approach, the PA is kept close to saturation by turning down the amplifier’s power supply voltage when the average power falls.

In practice there are trade-offs in complexity, cost and benefit for all these techniques that can deliver efficiency enhancement over wide ranges of modulations, frequency bands, bandwidths, and output power requirements. However, we have seen a rise in the popularity of systems taking advantage of PA power management, such as DC-DC converters. Go down this route and a single converter technology can be cost-effectively applied across all the different PA components and configurations needed to cover the required transmit bands and modulations (see Figure 5). Many alternatives are inferior, because they solve similar problems with techniques that potentially add to the cost and complexity of each component.



Figure 5. A representation of an RF front end showing PA power management (DC-DC converter) to multiple PA components.

We can illustrate this point with a block diagram of an ET design that we have put together. (See Figure 5). Further improvements in high-power efficiency may result from adding pre-distortion – this could increase high-power efficiency by allowing the PA to operate closer to compression (see Figure 3 for an example of how this can increase system efficiency).

However, effective, high-bandwidth envelope tracking demands a DC-DC converter that is capable of delivering the rapid, accurate changes in voltage required to follow the envelope of high PAR modulations. Unfortunately, gains in PA efficiency resulting from this approach are offset by declines in the conversion efficiency of the ET DC-DC converter and the circuitry needed to rapidly follow the modulation envelope (operating at a wider bandwidth reduces the converter efficiency). The good news, however, is that the system ‘up-lift’ is favourable. In other words, the PA efficiency improvement, minus the ‘cost’ of the ET implementation circuitry, is beneficial rather than detrimental.

As transmit power is lowered, a point can be reached where the PA benefit is effectively cancelled by the conversion efficiency of the ET circuitry. When this happens, the efficiency enhancement method must be changed for optimum performance. Several options are available: Switching the system to an APT mode with greater DC-DC conversion efficiencies; adjusting PA bias, or using similar low-power efficiency enhancement techniques; or adopting optimum combinations of these techniques, which come together to ensure high-efficiency transmit operation over the full range of conditions.

Bringing it all together

We believe that these new generations of RF platforms will require closer co-ordination of transceiver, power management and power-amplifier blocks. These latest designs must also be capable of handling complex communication, as well as timing signals to ensure proper operation with the transceiver – there needs

to be a seamless, well executed transition between the various efficiency enhancement techniques providing transmit operation over a wide range of modulations, frequencies and power levels. Desirable qualities from the power management include rapid, accurate changes in voltage capable of following a high PAR envelope if ET is used, plus the prevention of excessive switching noise, preservation of spectral purity and maintenance of sufficiently high power conversion efficiencies to reap the rewards of this approach. Additionally, the PA must be designed with the optimum load line for operation across the dynamic range. This implies proper device scaling, and careful trade-offs, such as judicious selection of the supply capacitance so that the smart device fulfils the complex requirements of the wireless network. 

It is clear that for the foreseeable future, consumer demand for mobility and high data rate connectivity in smart devices will continue to drive complexity and challenges in the RF front-end market. In response, engineers will continue to improve and optimise the energy usage, thermal performance and battery life of modern data devices through innovation in the individual components, as well as integration of these components into a cohesive RF platform.