Many of us got our hands on our first mobile phone in the late 1990s. Back then we could make calls, send text messages, and play some very basic games.
In comparison, today’s smartphones are incredibly powerful devices. In addition to calling and texting, they can take pictures and video, send e-mails, open documents and surf the web – and the price for doing all of this compares very favourably with the tariffs of yesteryear.
The foundation underpinning this evolution in the bang-per-buck provided by the mobile is an ever-advancing wireless infrastructure. It supports, in a cost-effective manner, higher transmission rates and increases in data traffic.
To continue to accommodate the growth in wireless traffic throughout the 2020s and beyond, there will be further changes to networks. They will have a denser spatial distribution, with greater deployment of smaller base stations, such as picocells and femtocells; and they will increase bandwidth by exploiting spatial diversity – with multiple-in, multiple-out transmissions. In addition, the transmitter amplifier will be integrated into the antenna to increase compactness and improve energy efficiency, via a reduction in cable losses. And to top it all, hardware components will support multi-band, multi-mode operation (MIMO) to deliver frequency and service agility.
However, just increasing data rates is not good enough. A major issue is base station inefficiency, which is a major contributor to the higher levels of energy consumption in wireless communication. This stands in the way of demands for a ‘green IT’.
Today’s mobile base station architectures for 3G and the first 4G are moving in the right direction, as there are requirements for a higher efficiency for the RF PA, as well as the needs for an output power of more than 100 W and a speed in excess of 2 GHz. The good news is that GaN is now fulfilling these requirements in real systems, thanks to progress in semiconductor technology and PA design.
Figure 1.Digital transmitter concept, including modulator, (digital) PA stage and filter.
However, it is not clear how sales of GaN PAs will fare when delivering the move toward distributed power in MIMO systems for 4G and 5G communication, and the associated increase in picocell and femtocell deployment. Critics argue that GaN devices are pricey, particularly when formed on a SiC substrate, and savings can be made with SiGe, siliconBiCMOS and even GaAs technologies. These are cheaper, and they offer a high level of integration.
However, silicon-based millimetre-wave PAs suffer from a low power density, an inferior efficiency and larger parasitics. When judged in these terms, GaN has the upper hand. What’s more, it has reached a level of maturity where it is possible to reliably fabricate larger designs with a higher level of integration. For these reasons, our team at FBH Berlin is investigating potential applications of GaN in future wireless networks, especially for smaller cells.
Base station woes
High levels of operating loss in base station PAs are compounded by the need to satisfy increasing bandwidth requirements in modern communication standards, such as 4G and 5G. This means that the amplifiers are operated well below their full-scale input power – typical values for power back-off are 6 dB or 12 dB. Consequently, for common analogue PAs optimised for maximum input power, the energy efficiency drops below 10 percent, with efficiency values for the whole base station being even smaller.
Due to the high total power consumption of base stations, interest is growing in novel system concepts for nextgeneration communication infrastructure that combine improved efficiency at power back-off with very high modulation bandwidths.
Of these designs, the most disruptive is that of our digital transmitter (see Figure 1). We have been developing GaN-based digital PAs, a technology that is well very known within the audio industry, for almost a decade. Now this digital GaN PA is piquing interest in the mobile communications sector.
With our approach, a digital baseband signal feeds the modulator, which generates the input voltage bit sequence to the digital PA. The digital input signal is amplified by a highly efficient PA switching stage, and the analogue filter at the output reconstructs the wanted signal.
Figure 2. Circuit schematic of a four-stage digital PA building block in push-pull configuration (T1/2) and realised H-bridge digital PA module (top right, next page; area: 68 x 42 x 15 mm3 ). Two preamplifiers in source configuration (1) amplify the very small input voltage swing of 0.5 Vpp only, which feeds the input of the differential PA (2). This part generates the different voltage swings for the final-stage transistors (3).
A great strength of this technology is that a very high power-added efficiency can be realised independently of the power back-off level. Losses are low, thanks to GaN HEMTs only being in either the onstate or the off-state. There is also a high degree of flexibility, because a broadband approach eliminates the requirement for narrowband impedance matching techniques, which are found in the likes of class AB and class E amplifiers.
GaN has several great attributes for a digital PA. It can form very fast, broadband transistors that are needed to cater for applied advanced modulation schemes, which may include frequency components down to kHz and up to several GHz, depending on the signal frequency encoded. The devices can also combine high speed with a low output capacitance for a given current.
We have made major contributions towards establishing the digital approach in mobile communications. Our successes include, in collaboration with NEC Corp. in Japan, the world’s first digital transmitter for a 450 MHz signal frequency. One of our key findings is that the finalstage configuration with the highest potential is the voltage-switching, push-pull approach. Unfortunately, complementary GaN HEMTs for the efficient driving of the final-stage are not available. So, working within this restriction, we have built a module with our 0.25 μm GaN-HEMT process (see Figure 2 (a) for a typical circuit diagram, and 2 (b) for the corresponding module).
This module features a pair of preamplifiers in source configuration that amplify the very small input voltage swing – it is just a TTL-level input voltage swing of 0.5 Vpp. The amplified signal is fed into the input of the differential PA, which generates swings for final-stage transistors. Combining this PA with an off-chip, 800 MHz band-pass filter has enabled us to set a new benchmark for the microwave amplification with a digital approach: a maximum power-added efficiency of 60 percent. This PA also produces a very high power gain of 40 dB.
One of the great strengths of the digital power amplifier is that it provides a versatile building block with multi-band capability. Currently, the only suitable candidates for the core amplifier of such a digital PA are those of the class-D/S type. These essentially act as power switches.
We have used our latest generation of GaN MMICs as building blocks for the construction of various power amplifier modules with increased digital content. They include the first tri-band amplifier of class-S type for the 0.8/1.8/2.6 GHz band, and an H-bridge and a single-chip classD/S PA for the 800 MHz band. Our portfolio of digital amplifiers has a maximum output power of 14 W and final-stage drain efficiency of up to 90 percent. Although full-scale output powers in the 10 W-range are insufficient for today’s macrocell base stations, they are more than adequate for smaller cells, such as picocells and femtocells.
Moreover, by applying a digital Doherty operation to the H-bridge PA enables final-stage drain efficiencies of 75 percent at power back-off levels of 6 dB, and 40 percent at 12 dB. These results demonstrate the benefits of the digital PA for realising a range of circuits with the same IC, formed with changes that just involve the periphery. The high flexibility and compactness of the circuits, combined with the performance of our digital power amplifiers, demonstrates that our technology is a very promising candidate for inclusion in picocells and femtocells for next-generation mobile communication infrastructure.