Questions, comments, complaints, bugs about this page? We would like to hear from you.
A Packaged MEMS-Based 5-bit X-Band High-Pass/Low-Pass Phase Shifter
Matthew A. Morton, Member, IEEE, and John Papapolymerou, Senior Member, IEEE
Sep 29, 2008:
Abstract- This paper demonstrates the first packaged microelectromechanical systems-based phase shifter using high-pass/low-pass circuit topology. The device was fabricated with a process resembling available low-cost commercial Si and SiGe processes, and demonstrates an average loss of 4.5 dB with an rms phase error better than 10° for 8–12 GHz. Thermal compression bonding is used to package the phase shifter with hermeticity confirmed by Military Standard 883G, Method 1014.12. The total shifter area is 9.2 mm2. Performance is compared to and found competitive with phase shifters using more expensive GaAs and BST technologies, which are less easily used in system-on-chip applications.
Index Terms—Microelectromechanical systems (MEMS), MEMS phase shifters, monolithic microwave integrated circuit (MMIC) phase shifters, phased arrays, phase shifters.
I. INTRODUCTION
The high–low-pass phase-shifter topology is ideal for the broad bandwidth and small size required for monolithic transmit/receive (T/R) modules [1]–[3]. Numerous X-band phase shifters have been demonstrated on SiGe with extremely low phase error and large signal performance [4]–[6]. However, each of these suffered from the same setback—insertion loss. While amplifiers can be used in the signal path without significant increases in die space or power requirements, the overall impact on the T/R noise figure hinders system performance.
The largest component of this loss comes from the switches that have an insertion loss on the order of 1.4 dB each. With ten switches used in the signal path of a 5-bit phase shifter, this loss can add quickly. The ability to utilize p-i-n diode switches instead of MOS-based technology would allow for a potential 0.5-dB decrease in loss per switch (5 dB total). However, this increases the cost of fabrication, and for many technologies is not an available option.
Microelectromechanical systems (MEMS) switches can provide the low insertion loss desired for this application. In addition to extremely low loss, MEMS have an inherently higher IIP3 and P1dB [7]–[11]. This enables higher signal power levels to be handled and maintained, reducing the noise and signal power burden on the rest of the T/R system.
There have been a number of MEMS-based phase shifters with exceptional size and loss characteristics presented in the literature, many with loss performance that exceeds the work presented here. However, each of these have significant challenges. The Rockwell Scientific MEMS-based phase shifters by Tan et al. [14], [15] provide high-performing 4-bit shifters with loss below 2 dB at X-band, yet are built on expensive GaAs substrates with metal thicknesses as large as 5 µm. The very low-loss (< 2.5 dB) 6-bit shifters by Nordquist et al. [16] are large due to the use of transmission lines as delay elements, but here too the alumina substrate and thick 6.5-µm gold lines are unrealistic and expensive for a phased-array application. The shifters utilizing BST technology from Chen et al. [17] also provide desired performance in a compact footprint, but the requisite sapphire substrate is hardly a cost-effective approach to the many thousands or tens of thousands of phase shifters needed for a phased-array system.
Each of these phase shifters represent the state of the art in MEMS-based phase-shifting technology, and each of them may have similar or better performance than the phase shifters presented in this study. However, none of these methods are compatible with a system-on-chip (SoC) approach, with the goal of reducing the majority of the RF front end of phased arrays to a single chip. Each of these methods require separate components, some quite costly, for each individual phase shifter. In addition, low loss associated with these solutions is largely attributed to the use of thick metal layers for transmission lines, which are not available on commercial processes used for SoC integration. Realistic losses of transmission lines on integrated platforms are on the order of 6 dB/ΰg [18].With this constraint, the 6.5-cm delay lines in the Nordquist et al. topology [16] make it far from low loss. Separate substrates with different materials or thick metals may be acceptable for systems with very few phase shifters, but for large numbers, the cost associated with any of these methods is prohibitive. The question here is not “how can we make the highest performance or smallest sized phase shifter?,” but “how can we do so cost effectively?”
This paper seeks to answer that question with a MEMS-based phase shifter that is fabricated in such a way as to realistically represent what fabrication processes are available on low-cost commercial technology such as silicon and SiGe. No metal thicknesses greater than 1 µm are used (not 5 > µm), the SiO dielectric thickness is only 10 µm (not 200- µm GaAs), and no materials such as sapphire, BST, GaAs, or alumina are used. The work presented here has been done in such a way as to enable full SoC integration with existing commercial processes, and provide a phase-shifting solution that may be slightly more lossy, or slightly larger than the above solutions, but provides a practical solution that can be physically realized in a large phased-array system.
A wideband micromachined method was used to package the phase shifter, requiring both input and output RF interconnects as well as ten individual dc interconnects. This method provides the lowest loss and bandwidth of the alternative popular approaches [19]–[22], while remaining a suitable bonding temperature for MEMS devices and leaving the primary sample wafer free of bulk micromachining.