People like wireless data communications: if you walk into a typical high street coffee house, you're likely to see customers updating their Facebook page over a Wi-Fi network or receiving 'tweets' from favourite celebrities on their smartphones.
Most Wi-Fi and cellular communications networks operate at frequencies between 800MHz and 3GHz. Such frequencies are attractive for wireless communications as they allow moderate bandwidths, have good propagation characteristics and there is a wide range of available component parts. Whilst the use of these frequency ranges is unlikely to lessen any time soon, there is increasing interest in wireless communications at much higher frequencies. The key reason for the interest in higher operating frequencies is the increased data rate available. Put simply, the higher the carrier frequency, the higher the practical data rate and the link between carrier frequency and data rate is well illustrated in figure 1. Unfortunately, there are inevitable downsides to increasing the carrier frequency: the propagation environment becomes more problematic; the capability of available technologies and component parts reduces; and the cost of components, sub systems and test equipment soars. Propagation The transmission path in current cellular and Wi-Fi systems is seldom a direct 'line of sight' between the base station and the mobile. At these frequencies, signals bounce off objects pretty well and can pass through the walls of buildings with acceptable attenuation. These facts are a large part of the attraction in using such frequencies for mass market mobile communications. Under free space conditions, the propagation path loss increases as the square of the operating frequency for given antenna gains; for example, there is a 20dB increase in loss for every decade increase in frequency. Whilst this indicates that operating at a lower frequency should be advantageous in reducing path losses, practical antenna implementations reduce the benefits. Broadly speaking, in order to have significant, gain an antenna must be electrically large. Put another way, its physical size must be significant compared to the wavelength. While the wavelength at 1GHz is 30cm, at 100MHz it is 3m high gain antennas can become physically unwieldy at low frequencies. At millimetre wave frequencies (greater than 30GHz) high gain antennas are physically small, which in part offsets the greater path loss at these higher frequencies. They are also highly directional, which results in line of sight propagation. For communication links without a line of sight connection, propagation takes place via multiple paths, resulting in signal fading due to constructive/destructive interference and this can limit data rates. It also implies the typical mean path loss will increase at a rate of about 30 to 40dB for each factor of 10 increase in distance, compared to 20dB for a line of sight link. This limits practical cell sizes for mobile communications to between a few hundred meters and a few kilometres, depending on the local environment. The high directivity antennas used at mm wave frequencies help mitigate multipath effects such as fading and allow very high data rate operation. In planning mm-wave links, an additional margin needs to be considered for losses due to rain (rain fade margin). These losses increase with frequency and vary with location, time of year and type of rain. Statistical data is used to determine the necessary rain fade margin based on the required link availability. In comparison, rain losses are generally insignificant at cellular frequencies. Additional signal attenuation is observed at mm-wave frequencies due to atmospheric absorption by water vapour and oxygen. Oxygen, in particular, has an absorption peak at around 60GHz, bringing an additional attenuation of about 15dB/km, which limits practical ranges. Viewed positively, this enables frequency reuse over relatively short distances; even stealth operation. Windows of spectrum exist between absorption peaks, where signals suffer less attenuation. The bands at 77GHz and 94GHz lie in such windows, which makes their use potentially attractive and has seen their adoption for high resolution radar applications. Applications and frequency bands The largest current application for millimetre wave communications is point to point links for mobile communications back haul and microwave point to point links in the 6 to 40GHz range are a well established technology. Looking to the immediate future, there is a lot of interest in frequencies between 40.5 and 43.5GHz, the so called the 42GHz band. This is viewed as a likely extension to the current range of fixed link frequency allocations. Although the availability of components to support 42GHz communications is still relatively limited, this is starting to change with recent product announcements from a number of suppliers. E-band spectrum – at 71 to 76GHz and 81 to 86GHz – is also receiving a lot of interest. The use of E-band offers worldwide availability of a large amount of spectrum under a 'light license' basis, a scheme operating in the US, the UK and many other countries that allows licences to be obtained quickly and cheaply whilst retaining the benefits of interference protection. Despite the attractions of E-band, it is not used in anywhere near the volumes of links at, say, 32GHz or 38GHz. The catalyst for high volume deployment will be significant reductions in equipment cost, but this requires wider availability of component parts with adequate performance at acceptable costs. There is also a significant allocation of mm-wave spectrum at around 60GHz, which is receiving interest from some quarters. The most extensive and flexible allocation is in the US, where the 57 to 64GHz band is available for unlicensed use. Two applications are normally cited for spectrum around 60GHz: medium range point to point outdoor links; and very high data rate Wireless LANs or Wireless personal area networks (PANs). However, the higher atmospheric attenuation caused by oxygen absorption makes the 60GHz bands look less attractive for medium range point to point links than E-band. Nevertheless, spectrum around 60GHz is an attractive option for very high data rate WLAN/WPAN applications, where short range links enable very high data rates. And the oxygen absorption problem could actually be beneficial; allowing the band to be reused in relatively close proximity. Technology The development of mm-wave components is far from trivial. The number of suitable commercially available processes is limited and, by the time operation at E-band is considered, the available transistor gain is modest. Achieving adequate output power and linearity is also problematic as larger transistors have even less gain and combining multiple smaller transistors incurs additional combining network losses. Short geometry cmos and SiGe processes have transistors with a high enough ft to provide gain at frequencies to E-band and numerous circuits operating at high mm-wave frequencies have been demonstrated using these technologies. However, all current commercially available mm-wave links for wireless back haul incorporate GaAs based front end monolithic microwave ics. The reason behind this is that acceptable noise figure (NF) and adequate linearity are essential requirements and GaAs technology offers superior performance in these respects. GaN technology shows a lot of promise for the future, in particular for the realisation of mm-wave power amplifiers. However, the commercially available GaN foundry processes are only suitable for operation to around 20GHz so, for the immediate future, GaAs technology is the best candidate. Of today's commercially available GaAs processes, the best choice for output power and linearity at E-band is the 0.1µm gate length pseudomorphic high electron mobility transistor, or pHEMT. Obviously process costs also need to be considered. GaAs processes at 0.1µm geometries tend to use e-beam written gates, which can push up costs. Some 0.13µm and 0.15µm processes have optically defined gates, which should result in lower production costs and these can be used to implement cost competitive components for some functional blocks. Figure 2 shows a Plextek designed single sideband up converter operating at E-band, which is implemented on a low cost pHEMT process with optically defined 0.13µm gates. The reliance on e-beam written gates is likely to change over time and it is expected that GaAs processes with optically defined gates at lengths of 0.1µm and smaller will become commercially available in due course. It is likely that cmos or SiGe will be selected for WLAN/WPAN applications in the 60GHz band. For these applications, the potential product volumes would be extremely high, while the required performance (NF, linearity and transmit power) would be less stringent than for point to point links and the cost targets very low. These factors lead to the conclusion that this is an application likely to be dominated by highly integrated silicon transceivers, with the possibility of a small role for low cost transmit amplifiers realised in GaAs. Although the desire clearly exists for the very high data rates that increasing operating frequencies can allow, the move up in frequency is slowed by the cost of equipment and the performance of available component parts. Processes exist that allow the development of component parts that can be produced at a low cost in high volume at frequencies up to E-band. The real barrier is the absence of a commercial leap of faith to believe that market volumes will justify the development costs. Liam Devlin is director of Plextek's RF Integration Group. Marcus Walden is senior technical consultant with the company's Radio Group.