Maximize Range in Mobile Handsets with CMOS-on-Sapphire RF Switches
You canít build a multi-band RF front-end switch with PIN diodes. But think twice before going to GaAs.
Mobile handsets need to include many high-selectivity filters in order to accommodate the growing number of frequency bands that must be handled by today's portable multimedia devices. Each handset generation has supported more frequencies, so the number of filters has grown, forcing designers to add switching elements in order to optimize the link budget.
Figure 1: Typical multi-band handset system layout
including a SP9T switch
As with all other components in a mobile handset, the cost, performance, and form factor of these switches has been under great scrutiny because they operate in the highest-volume consumer market in the world. In addition, the switch manufacturer needs to have a technology roadmap that will allow it to accommodate new bands and functionalities for next-generation phones and beyond. Rising to this challenge, CMOS-on-sapphire RF switches, already available in single-pole, nine-throw (SP9T) configurations, offer the benefits of high linearity and ease of design in the face of shrinking footprints.
Adding Bandwidths, Changing Needs
For numerous regulatory, political, and technical reasons, the frequency bands used for cellular communications varies across the Americas, Europe, and Asia. As a result, products designed into the handset are required to offer multi-band support. For instance, it is not uncommon for a GSM phone to support up to four bands (850/900/1800/1900MHz). UMTS, generally regarded as a successor to GSM, will also require multi-band support to enable global roaming and data transmission. When this multi-band requirement is combined with the need for high integration, low power, small size, and maximized link budget (high linearity), then the challenge facing systems designers becomes even more intriguing.
As cellular telephone service proliferates around the world (there are nearly 3 billion subscribers worldwide, with service reaching 80% of the world's population), frequency bands continue to be added. Due to the stringent performance requirements dictated by the range needed for effective network operation, high selectivity filters are used liberally in handset front-ends. It works this way, as the frequency band count increases, handset designers incorporate switching elements to optimize the link budget.
Back when handsets were single band, designers used PIN diodes for switching due to their high performance and low cost. However, once handsets combined GSM/EDGE and WCDMA, PIN diodes no longer met size and performance requirements. Requiring long quarter-wave transmission lines and large forward bias currents to operate, PIN diodes ceased to be effective with the introduction of quad-band GSM.
Ongoing Performance Challenges
In order to fill the technology gap created by the four-band requirement, handset designers began using IC-based solutions manufactured using UltraCMOS or GaAs. These technologies readily solved the numerous implementation problems with PIN diodes for multi-band operation, and, thus, essentially replaced PIN diodes for the switching function.
In the early days of 3G wireless, designers added WCDMA to GSM/EDGE handsets. This introduced a challenging third-order intercept (IP3) requirement of +68 dBm on the front-end switch in order to ensure network robustness. This means that the switch cannot generate a distortion signal larger than three trillionths of the WCDMA transmit signal, or the equivalent of one silicon atom in a half-mile string of atoms. UltraCMOS meets this difficult linearity requirement, and it has emerged as the leading solution for RF front-end switching applications due to the benefits of a low-loss, low-capacitance sapphire substrate combined with the fundamentally exceptional linearity of an intrinsic MOS device.
Recently, the need for more switching has been growing at an impressive speed. For instance, in 2007, the majority of 3G handsets shipped with quad-band GSM/EDGE and single-band WCDMA, which requires a SP7T switch. By 2008, the integration of three WCDMA bands was required, increasing the switch complexity to a SP9T (Figure 1). This trend of increasing switch complexity is expected to continue. For instance, at the time of this writing, there are now nine WCDMA bands defined worldwide, and that number is expected to climb.
By incorporating an UltraCMOS SP9T switch, designers can expect performance specifications which often exceed those of GaAs-based SP9Ts, while meeting the demand for smaller footprint. Unlike GaAs switches, UltraCMOS devices do not require external bypassing or filtering components, and are available as a flip chip IC, which requires less attention to control line routing and simplifies the printed circuit board (PCB) layout.
Figure 2: SP9T switches in UltraCMOS and GaAs differ
significantly in die size.
Selecting a System Design
An RF front-end switch must support WCDMA and GSM/EDGE signals simultaneously, so it must be the most linear element in the handset in order to ensure the expected range. In fact, it is required to be the most linear solid-state device of any high-volume application in the world. Handset systems designers are effectively faced with two choices in switches: UltraCMOS or GaAs. In addition to RF performance characteristics that will maximize range and optimize the link budget, some additional considerations include: voltage handling, die size, system size, ease of design, ESD tolerance.
In UltraCMOS, the transistors on sapphire are dielectrically isolated from one another, so they can be placed in series (or stacked) to tolerate the very high voltages levels that can be present in an antenna switch. Although CMOS is low-voltage technology, peak-to-peak voltage handling of 50 V is readily tolerated. Additionally, on-chip bias generation provides for optimized performance and eliminates the need for external DC blocking capacitors.
Figure 2 compares a WEDGE SP9T implementation in UltraCMOS and GaAs. Note the significant difference in die size: the UltraCMOS die measures 1.43 mm2, which is approximately half the die size of the GaAs SP9T of 2.85 mm2. With the fine design rules of aluminum metallization and the flexibility to place FETs in any orientation, UltraCMOS allows RF switches to be implemented more compactly than GaAs. The complementary devices, analog control capability, and MOS capacitors allow for control of both series and shunt RF FETs with a simple four-wire interface and low current drain. Figure 2 also shows the implementation areas for the 4:16 decoder in the two designs.
Figure 3: Moving to standard flip-chip packaging can reduce the
module area by 43%, even while adding more functionality to the
switch (SP7T vs SP9T).
Switches that use standard flip-chip chip-scale packaging technology can significantly reduce the footprint of handset front-end modules (FEMs). These FEMs are usually fabricated in costly low temperature cofired ceramic (LTCC), so reductions in module area directly impact the total solution cost. Figure 3 shows how migration from wirebonding to flip-chip reduced the LTCC substrate area consumed by 43% while simultaneously increasing the supported frequency band count from five to seven.
Ease of Design
With flip-chip technology, the switch can be placed as a surface-mount device (SMD). Since all of the other components in the FEM are typically SMDs, the need for any wirebonding equipment is eliminated. Shipped in tape-and-reel form, these flip-chip switches are inserted in the module using standard placement machines, resulting in the lowest manufacturing cost.
Additionally, at less than 250 μm mounted height, flip-chips reduce the overall module thickness, which is critical to meeting the height demands for handset slimming.
Including shunting FETs on all ports in the UltraCMOS die also has tremendous impact on ease of module design and performance. Because all signal paths are shunted to ground by the shunt FET, noise generated throughout the rest of the system has no impact on the active path. By controlling the impedances of the isolated ports at the plane of the switch ports, the module designer does not have to optimize the entire signal chain simultaneously, but can instead focus only on the active path. Shunt FETs also provide very high isolation performance which relaxes the spurious requirements on transceivers, and ensure protection of receive SAW filters from high RF power and ESD events.
Table 1: Performance Comparison between UltraCMOS and GaAs SP9Ts
Because it is a CMOS technology, UltraCMOS offers an advantage of integrated ESD protection devices. As a result, devices manufactured using this process have Class 2 (2000 V) HBM tolerance on control pins and 1500 V HBM tolerance on RF pins, so UltraCMOS switches are highly robust against ESD damage. This reduces module fallout during manufacturing, and also dramatically reduces the required ESD circuitry at the antenna to meet stringent IEC61000-4-2 requirement, recovering circuit board area and eliminating some insertion loss.
Table 1 summarizes the SP9T performance of the two technologies. Insertion loss performance is approximately equal, while linearity performance of UltraCMOS far exceeds that of GaAs. The wide VDD range of UltraCMOS aligns with existing handset supply voltages, and bias generators are in development for direct operation from battery voltage or 1.8 V supplies
UltraCMOS RF switches provide tremendous value to handset front-ends, particularly in multi-band applications where very high IP3 is required. As 3G handsets continue to increase in complexity, the switches will simultaneously need to scale to higher throw counts. We are likely to also see the integration of multiple switches on a single die. These developments will fully exploit the advantages of UltraCMOS, and are increasingly difficult for GaAs switches to achieve. In this market, an achievable, long-term improvement roadmap will likely prove to be a critical factor in a technology’s success. The good news is that UltraCMOS is positioned to handle the future needs of handset designers.
This article first appeared in the October/November issue of Portable Design. Reprinted with permission.
San Diego, CA