ConvertsDevelopers of GSM handsets are looking for more sophisticated RF architectures than the tried and trusted superhet receiver…but that’s not so easy. Roy Rubenstein explains why
Whenever advances in silicon processes are mentioned, it is usually in the context of shrinking geometries and ever greater on-chip integration. Less often mentioned is the attendant increase in the underlying transistors’ operating frequencies. For wireless communications, this is helping to shift the boundary between RF and baseband towards the antenna, simplifying the RF chain.
That, at least, is the theory.
Developers of GSM RF and baseband ICs, although aware of this trend, are having to contend with developments, such as multi-standard, which are making the RF design anything but simple.
“The design of GSM handsets is not easy but it is very tractable and amenable to the components available,” said Dr Walter Tuttlebee, radio comms business development manager at Siemens’ Roke Manor Research. A commonly used receiver architecture for GSM is the dual IF (intermediate frequency) superheterodyne receiver (see diagram). Yet while being a tried and trusted design (the superhet dates back to 1918), engineers are looking to use more sophisticated RF architectures for the handset.
The motivation for this is one of cost. “The more conventional the architecture, the more stages there are and the greater the cost,” said John McNicol, GECPlessey Semiconductors’ marketing manager for communications.
While RF functions such as the amplifiers and mixers are being swept up within an RF IC, additional off-chip components such as the SAW filters are required.
According to McNicol, SAW filters are not only expensive but also adversely affect a design’s size and power consumption. This is an important issue, especially for appliances such as the Nokia 9000, where the radio communications is but one of the product’s features. Direct Conversion
One receiver architecture which does away with the IF stages and associated filters is direct conversion. Here, a local oscillator with a frequency equal to that of the carrier is applied to the mixers, resulting in an output comprising the in-phase (I) and quadrature (Q) baseband data.
Since only one mixing stage is used, its performance (a function of such factors as the IC component tolerances) is key. For dual IF, while more mixing stages are used, there is also more opportunity for processing to compensate for the signal impairments introduced. For direct conversion, these include intermodulation products, phase error and a DC offset in the I and Q channels, which reduce the dynamic range. Moreover, the burden of the filtering, to extract GSM’s 200kHz-wide channels, now falls on the baseband processor. Consequently, while direct conversion is a highly attractive prospect, likely to reduce the handset cost by between 10 and 20 per cent, it is still to be proven commercially viable for GSM.
According to Roke Manor, advances in the power consumption and dynamic range of A/D converters are allowing for ‘creeping’ direct conversion with the boundary between the RF and baseband moving up the signal chain. Roger Hopper, a cellular terminal designer at Roke Manor, claims that to implement direct conversion digitally would require at least a 14-bit A/D converter – to achieve the necessary dynamic range – capturing the signal channels at 900MHz .
The attraction of mixing and filtering digitally is that the signal impairments of direct conversion can be solved using DSP techniques. Such a development is unlikely to be used commercially before the year 2000, claims Hopper, as only then will such an A/D converter, and hence the resulting handset, be competitive in terms of power consumption. Dual-band and dual-mode
Another developing complicating RF receiver architectures is the advent of GSMdual-bands phones, handsets that operate at several frequencies (for GSM, 900 and 1,800MHz or 900 and 1,900MHz for use in the US) to enable the handset to be used with several service providers.
The RF design of the first dual-band mobile phones, while more complex, is being tackled relatively straightforwardly as manufacturers are keen to get products to market. The choice of the local oscillators and the IFs require careful consideration to avoid any mutually inflicted interference.
Manufacturers typically use two RF front ends comprising separate low noise amplifiers and mixers, with the two signal chains down-converted to a common IF stage. Future designs are expected to better share the RF components.
If dual band phones are exercising RF designers, one development set to kick-start architectural innovation is multi-mode phones, handsets that support more than one cellular standard.
Ericsson has already developed a dual-mode handset that supports GSM and DECT. Other envisaged combinations include GSM and the IS-95 US code division multiple access (CDMA) standard, and phones supporting terrestrial and satellite-based standards.
Unlike a dual-band phone which share the one standard, dual-mode handsets implement distinct radio protocols. “All the things that people are trying to get rid of, such as filters, are [for dual-mode] doubled up,” said Hopper.
Moreover, each standard has its own reference frequency, used to define the system’s parameters. For GSM it is 13MHz while for IS-95 it is 19MHz. This complicates the RF chain’s design due to the scope for conflicting harmonics.
At present there is uncertainty as to which style of dual-mode handset will gain market acceptance. Until then, semiconductor manufacturers will remain reluctant to develop custom RF designs.
The superhet is likely to be with us for awhile yet.   Dual IF receiver at a glance For the GSMhandset receiver path, 200kHz-wide channels containing the encoded voice are down-converted from 900MHz (or 1,800 and 1,900MHz) to baseband. Extracting such a narrowband signal, in the likely presence of a stronger signal in an adjacent channel, is commonly performed by down-converting to baseb and in two stages, filtering at each. Such an architecture is referred to as a dual IF (intermediate frequency) or dual conversion superheterodyne receiver.
It is worth pointing out that certain chipset manufacturers have adopted designs that use a single IF stage. These include Philips Semiconductors and Lucent Technologies. Single IF can be seen as a compromise between dual IF and a direct conversion approach.
For a dual IF design, the received signal from the antenna is filtered to reject signal components outside the GSMband. It is passed through the low noise amplifier stage before being image filtered and mixed to the first IF.
The mixing to the IF is performed with a frequency-agile synthesiser which changes its frequency to track the appropriate channel containing the signal of interest. The choice of the IF varies between handset designs – examples include 71MHz and 246MHz. The IF is chosen at a part of the spectrum deemed ‘quiet’, where few unwanted spectral components reside.
The image filtering is needed prior to mixing to suppress components at the image frequency which otherwise would coincide and interfere with the desired signal.
For example, a signal at 946MHz to be shifted to an IFof 246MHz, is mixed with a local oscillator frequency of 700MHz. Without image filtering, any signal at 454MHz would be superimposed on the desired one.
The SAW filter filters the 200kHz channel of interest. Since the mixer is lossy, the signal is amplified and mixed again, this time with a static oscillator, to the second IF where further filtering occurs.
The signal is finally taken to baseband where processing uncovers the encoded speech.

Leave a Reply

Your email address will not be published. Required fields are marked *