Challenges in Chipset Design for EDGE Terminals
Challenges in developing handsets and other terminal devices for EDGE
By Dan Fague and Doug Grant, Analog Devices, Inc.
With second-generation (digital) cellular standards fully deployed worldwide, it is time to look forward to the future of wide-area wireless communications in practical terms. Several versions of third-generation "3G" standards have been developed in an effort to provide higher data rates for advanced mobile services, and some commercial deployments have begun (primarily Japan), it will be quite a while before true 3G is available worldwide. There are several versions of 3G air interfaces developed and maintained by different organizations: the 3GPP organization is responsible for the standard generally referred to as W-CDMA in many countries, this system will be used in conjunction with GSM networks to provide "UMTS" the 3GPP2 organization is driving the cdma2000 family of standards the "1XRTT" version uses a single radio carrier to provide higher data rates, while the "3XRTT" version uses three carriers to provide even higher data rates; a Chinese-government-backed group is backing the third version, called TD-SCDMA.
The W-CDMA standard is expected to be the most widely deployed of the 3G systems, with cellular operators in Japan, Europe, and North American committed to using it. However the economic realities of the past few years have forced operators to take a hard look at the costs associated with deploying a new system in newly allocated spectrum, since it means new cell sites, towers, and equipment. Most have delayed or even cancelled their plans for 3G deployment until economic conditions improve. Some observers point out that operators using the cdma2000 1XRTT system already provide 3G-like services in Korea, Japan and some parts of the U.S. However, switching to the cdma2000 standard is not an option for operators with major investments in GSM protocol-related infrastructure.
Figure 1. Block diagram of typical Cellular Terminal Chipset
EDGE to the Rescue
Fortunately, in the GSM family of cellular standards, an intermediate step exists between today's GSM system and true 3G. It is called "EDGE", for Enhanced Data rates for GSM Evolution, and offers 3 to 5 times higher data rates than standard GSM. It uses the same frequency bands, the same channel width and spacing, and the same protocol as GSM/GPRS. In addition, since EDGE has been in the 3GPP roadmap, most currently available GSM/GPRS infrastructure equipment (base stations) has been designed with EDGE capability. An operator can upgrade the network from GSM/GPRS to EDGE almost instantly. All that is needed for widespread deployment of EDGE (and the delivery of services that can take advantage of the higher data rates) is handsets. And at present, there are very few available, owing to the lack of complete chipsets available until recently.
Figure 1 shows the block diagram of a typical high-end cellular terminal. There are serious challenges in developing handsets and other terminal devices for EDGE, compared to GSM/GPRS terminals. The biggest challenge is in the radio section, which requires higher performance in the receiver chain compared to conventional GSM/GPRS receivers to accommodate the EDGE signal, and a different architecture than usually used in GSM/GPRS radios to generate the EDGE waveform as well as a more linear power amplifier than required for GSM/GPRS. The baseband section includes additional challenges, such as better A/D converters and getting a fast enough processor core to handle the more complex equalizer and demodulation tasks. Without belittling the challenges in the baseband design, the architecture is not much different than standard GSM/GPRS. However, the radio section requires significant redesign.
How EDGE Differs from GSM/GPRS
To achieve the higher data rates that EDGE promises, the modulation spectral efficiency and complexity is increased. The rather simple GMSK modulation used in GSM/GPRS is replaced with a more complicated 3pi/8-shifted 8 PSK. The pre-modulation filter that shapes the spectrum is also more complex.
Figure 2. Comparison of GMSK signal used in GSM/GPRS with 8PSK signal used in EDGE
The 3pi/8-shifted 8-PSK modulation of EDGE presents a complicated picture to the data receiver (see Figure 2). The number of phase states is increased (from 4 to 8), and the pre-modulation filter is not a simple Gaussian filter. Instead, the filter is constructed from a series of segments that are assembled in the time domain to create a response that is compatible with the EDGE equalizer. This makes for an efficient equalizer implementation, but the computations required are still formidable.
In most implementations of the modulator in a GSM/GPRS transmitter, a form of an offset PLL or "translation loop" is used. These transmitters rely on the fact that the modulation is phase only, with no modulation information in the amplitude (envelope) of the signal. The transmitter VCO is locked in some way to the local oscillator of the modulator, typically with feedback from the transmitter VCO's output driving a phase detector and being compared to the modulator's output frequency or the local oscillator output.
While this method can work for the phase portion of the EDGE signal, it is not enough to create the signal necessary for EDGE. The EDGE modulator must create the phase portion and also create the envelope fluctuations that occur due to the modulating signal. Several methods of creating the EDGE signal are possible, though most architectures are moving towards a polar modulation scheme. In a polar modulator, the phase and amplitude of the modulated signal are generated and applied to the carrier separately.
An advantage of the polar modulator is that existing GSM hardware can be used for the phase modulation portion of the EDGE modulator. An amplitude path of some sort must be added to complete the platform. The amplitude can be added by modulating the power amplifier directly, or by modulating the output of the phase path prior to the power amplifier input. In either case, the timing of the amplitude and phase blocks is critical to achieving good error vector magnitude (EVM) performance.