Overcoming the RF challenges of multiband mobile handset design

Demands on components in the RF signal chain of a mobile handset increase with the number of bandwidths and modes, especially for the antenna and its switch. The integration capabilities of RF CMOS devices solve many of these issues while maintaining peak performance.

CMOS has long been regarded as the technology of choice for integration, particularly in high-volume applications where cost is a major driver. During the last few years, advances in CMOS technology have led to its use in analog devices, and then into the intermediate frequency (IF) and radio frequency (RF) domains that were once dominated by BiCMOS and GaAs. As the industry drives forward with multiband applications, these developments have come not a moment too soon.
 

The bottom line is that the RF signal path in a mobile handset has become extremely crowded. The complexity of cellular phones has grown at a rapid pace, moving from dual-band, to tri-band, and now quad-band. In addition, these phones also need to handle a variety of signals for peripheral radios, such as Bluetooth, Wi-Fi, and GPS. This trend is expected to continue as WiMAX and LTE (4G) capabilities are added. In a mobile handset, the antenna switch is the gatekeeper that controls antenna access for all of the radio signals. Currently, designers are specifying new single-pole, 9-throw (SP9T) switches, and we can reasonably expect that SP10T is not far away.

RF challenges

Handsets now being developed incorporate tri-band WCDMA and quad-band EDGE platforms, an architecture that demands at least seven radios in a single handset. Most likely, the market will standardize on at least seven bands, with room for an eighth (for LTE). Regardless, complexity will continue to rise due to the increased popularity of peripheral radios and functions that also need access to the antenna.
 

All of this has greatly complicated the RF front-end by more than tripling the number of high-power signal paths engineered in today’s quad-band EDGE handsets. By its nature, a multiband handset must handle numerous RF paths that all operate on different bandwidths. Yet, they all need to access the same antenna. The most effi cient approach is to path all of the signals through a single RF switch.  For switch manufacturers, this has meant a corresponding evolution from single-pole 4-throw (SP4T), to SP7T, to now SP9T confi gurations in order to handle the increased number of signals. Increased functionality also impacts the antenna, which must effectively radiate from 800 MHz to 2200 MHz from a tiny footprint. Antenna designers are addressing these issues by taking advantage of performance-enhancing matching, switching, and lumped tuning elements. Ultimately, though, it is the RF switch that must be capable of switching up to nine paths or more of high-power RF signals with low insertion loss, high isolation, and exceptional linearity. Next-generation designs require to/from fi lter banks, and, as a result, the burden on the switch element is high. In terms of specifi cations, new applications tend to require very low insertion loss due to the signal going through multiple switch paths; very high linearity due to a WCDMA platform; very high isolation for critical paths; and a small, effective switch solution for independent (yet simultaneous) signal paths with up to 14 control states. In a mobile handset, RF designers are typically responsible for the antenna switch module (ASM), front-end module (FEM), and the transmit module. The ASM typically includes a switch, decoder, power amplifi er (PA) low-pass fi lters, ESD circuitry, and a voltage generator.

 Figure 1. Next-generation mobile handset design must handle a multitude of receive and transmit signal paths.

A multimode, multiband mobile handset generally uses a single PA module to handle all of the quad-band GSM/EDGE signals. In contrast, each WCDMA signal requires its own individual PA. Figure 1 shows a next-generation mobile handset design. The orange area shows the additional PAs and filters that are required to handle the multitude of receive (Rx) and transmit (Tx) signal paths.

 A quad-band GSM handset with one WCDMA band requires at least a single-pole, 6-throw (SP6T) switch to manage all of the signal paths. Designers could use a diplexer and two SP3Ts (which is a popular GaAs configuration), but this results in higher insertion loss than when using a single SP6T switch. In all of these designs, a multiband scenario means significant architectural, performance, and cost challenges. And, any design trade offs in a multiband phone still require the handset to meet or exceed the performance levels of all the standards supported.

Insertion loss

Insertion loss is important in multiband designs because it directly impacts the effective power-added effi ciency (PAE) of the PA. GSM PAs are typically run in saturation at up to 3 W, with an average PAE of 55%. The PA is responsible for half of the total current drain in the handset, so any deterioration of efficiency has a direct impact on battery life. Maintaining the PA’s PAE, then, needs to be a high priority. Early multiband WCDMA/GSM handsets featured separate signal chains for WCDMA and GSM that were then pathed to separate antennas. While this worked for prototypes and fi rst-generation designs, market pressures required a more cost-effective, spacesaving approach. Clearly, the industry needed integrated ASMs that handled seven or more signals.

Initially, SP7T switches were used to support handset architecture with one WCDMA and four GSM bands. Later, designs such as the PE42693, a monolithic SP9T, were developed on UltraCMOS process technology. Figure 2 shows insertion loss over frequency for this switch in terms of multiple transceiver (TRx), Tx, and Rx signals. This level of performance is necessary to ensure effi cient RF front-end designs. The SP9T switch is one of the latest advancements in switch architectures. It can be confi gured to handle multiple bands of WCDMA, GSM, and peripheral radios. The switch in Figure 1, for example, is handling three bands of WCDMA, with paths to duplexers and three PA modules. The switch also handles quad-band GSM/EDGE, which has a single PA module associated with it. In effect, this device has to route fi ve highpower signals through a single switch that is controlled by a simple decoder.

Figure 2. The PE42693 SP9T from Peregrine demonstrates TRx insertion loss of 0.70 dB at 850 MHz WCDMA.

Linearity

Adding more bands to the handset has greatly increased the technical requirements of the switch, and the linearity and harmonic requirements of WCDMA have put a large strain on the performance. For instance, a switch is now generally agreed to need a third-order intercept point (IP3) of better than +65 dBm. In previous GSM-only designs, there was no comparable linearity requirement. The increased front-end complexity of newer designs makes this high level of linearity extremely difficult to achieve for an active device on any manufacturing process. However, by leveraging the linearity
advantages of the UltraCMOS process, the monolithic PE42693 SP9T in Figure 1 is able to maintain the +68 dBm IP3 of its SP7T predecessor with third-order intermodulation distortion (IMD3) performance that surpasses the industry specification of -105 dBm. This level of IMD3 performance
reduces the potential for interference within the mobile handset’s radio.

Isolation

As the number of signal paths in the handset has grown, so has the need for better isolation. The number of input/outputs (I/Os) in an ASM dictates the level of isolation required. For instance, a gallium arsenide (GaAs) SP9T typically requires 18 control lines. In addition to making it challenging to route all of these lines in and out of a singular switch device, it is particularly challenging for the five high-power ports that require good linearity and isolation. The more I/Os there are, the more likely it is for wires to couple during use and bond. An UltraCMOS SP9T, such as the one in Figure 3, requires only four control lines, so this issue is less serious. Isolation is important because the coupling and bonding of signals can be detrimental to a multiband handset’s performance. For instance, the PCS1900 transmit band overlaps with the DCS1800 receive band. Without isolation of 35 dB or better, unwanted in-band signals could pass through the filters and desensitize the receiver, which would result in dropped calls.

Figure 3. The monolithic UltraCMOS PE42693 SP9T antenna switch offers 4-pin logic control with integral decoder/driver that facilitates 1.8 V and 2.75 V control

Small switches
 

Despite the need for more switching capability, the real estate budget for the ASM is shrinking in new mobile handset designs. The need for highly integrated, small antenna switches becomes more pressing. New process technologies for multithrow switches and the use of CMOS is allowing unprecedented integration and shrinking footprints.

For instance, a GaAs SP7T measures 2.4 mm2 whereas a comparable SP7T switch design using 0.5 µm silicon-on-sapphire (SOS) with equal or better small- and largesignal performance measures 1.2 mm2, a space savings of an extraordinary 50%. Currently available GaAs E/D pHemt or J-pHemt SP9T switches measure 2.85 mm2. An SP9T manufactured using an UltraCMOS 0.5 µm process measures 1.87 mm2 and does not require off-chip ESD devices or linearity enhancing matching components. A 0.25 µm UltraCMOS process is in the final stages of deployment, which will further reduce the size of the SP9T another 10%. Until recently, SP9T and SP7T switches were only available as wire-bond devices.

Recent process developments have led switch designers to flip-chip mount the switch to a low temperature co-fired ceramic (LTCC) substrate without underfill, eliminating the area previously required for wire bonding. This further reduces the device footprint. Currently, wafer-level chip-scale packaging (CSP) is in development to produce Ultra- CMOS switches that can be handled like a standard surface-mount package. Besides process technologies, another way to reduce the ASM’s footprint is to increase integration, a common roadmap for all integrated devices. When manufactured in UltraCMOS, switches can eliminate the decoder, blocking capacitors, and the diplexer that are required with other non-CMOS switch technologies. Combined with CSP technology, this process can dramatically reduce the size and thickness of ASMs. In addition, inherent ESD tolerance and a monolithic CMOS interface simplify implementation and use. Finally, the high yield of UltraCMOS processes and scalability to additional switch throws provides a roadmap to higher levels of integration
for future generations of handsets.

Monolithic switches
 

The technical requirements of the multimode, multiband GSM/WCDMA handset have exceeded the capabilities of traditional RFIC technologies such as GaAs (Table 1). Most critically affected by these ultrahigh performance specs are the antenna and the RF switch. As multiband architectures have grown, so has the requirement for the number of PAs and associated filters. But, in reality, the technical demand on the PAs has not changed, only the need for more PAs in a handset design. A critical challenge on the path to multiband handsets is finding an extremely efficient method to route all of the RF signals to the antenna—the monolithic switch. Fortunately, advanced CMOS processes are already in place to satisfy the increased demands on components in the RF signal chain of a multiband, multimode mobile handset. The integration capabilities of RF CMOS devices in the form of UltraCMOS RFICs solve many of these issues while maintaining peak performance.

Publicado por: Jahir Alonzo Linares Mora C.I: 19769430 CRF

Bibliografia: http://rfdesign.com/mag/707RFDF4.pdf