MEMS switch technology has many potential benefits over conventional electronic devices for switching microwave and millimeter-wave signals. MEMS switches possess very low insertion loss, miniscule power consumption, and ultra-high linearity. These characteristics make MEMS technology an ideal candidate for incorporation into passive circuits, such as phase shifters or tunable filters, for implementation in the passive front-end of phased antenna arrays at Xband and above.
Emphasis in RF MEMS research has shifted towards achieving a better understanding of MEMS switch reliability and improving the packaging of MEMS switches. Efforts in the former have centered on methodologies to quantify the amount of charging present in capacitive MEMS switches, and relating that charging back to the electromechanics of the switch. In this manner, improvements in materials and mechanical designs can be quantified relative to increasing the operational lifetime of MEMS capacitive switches.
Prior to the start of life testing, a thorough characterization of the electromechanical, RF, and dielectric charging performance was completed.
Electromechanical Performance — The bistable switching characteristics of the MEMS devices were tested by probing the switches with a swept voltage through a capacitance meter. An HP 33220A arbitrary waveform generator, driving a Tabor 9100 high-voltage amplifier, provided voltage drive to a Boonton 7200 capacitance meter, with its analog output captured and digitized using a Tektronix 2440 oscilloscope. These measurements were taken at multiple locations across the wafer.
RF Properties — From the switch operating curves and appropriate test structures, the RF capacitive characteristics of the MEMS switch were extracted. The switch itself averages 44 fF of shunt off-capacitance, of which approximately 30 fF is plate capacitance and 14 fF is fringing capacitance. This switch capacitance is partially compensated with inductive feedlines in and out of the switch, yielding an effective shunt capacitance of ~ 15-20 fF in off-state. The RF properties of these switches at microwave and millimeter-wave frequencies were measured on a HP 8510 vector network analyzer. The RF insertion loss of the MEMS switch is typical of most modern MEMS switches, with less than 0.1 dB insertion loss through 40 GHz.
Dielectric Charging Characterization — The transient current measurements were made over a voltage range of 30-60 volts to quantify the change density as a function of both time and voltage. These techniques were applied to the sputtered silicon dioxide. This characterization was only completed for negative polarity, as this material commonly exhibits significantly higher charging for a positive polarity drive voltage. The test devices used for this characterization were MIM capacitors fabricated on the same wafers as the MEMS switches.
Dielectric Charging Reduction
The simple and most effective methods for reducing dielectric charging within MEMS switches are to reduce the operating voltage, reduce the dielectric area, and/or reduce the operating duty cycle. In order to achieve high cycle lifetime, all three of these techniques were utilized to minimize the amount of dielectric charging present in the switch.
Reduced Operating Voltage — Higher operating voltages exponentially decrease the operating lifetime of the switch. The lower limit of operating voltage is determined by the minimum restoring force necessary to ensure that surface forces do not induce stiction of the switch membrane. Further, the minimum control voltage is also determined by the ability to achieve a repeatable tensile stress of the switch membrane during fabrication.
Reduced Dielectric Area — Minimizing the dielectric contact area of MEMS switches is becoming more common as the difficulties of overcoming dielectric charging become evident. By changing from a continuous sheet of dielectric beneath the switch to an array of posts, the active area for charging can be reduced.
Reduced Duty Cycle — In order to obtain high cycle counts in switch lifetime measurements, it is necessary to operate the switch as quickly as possible. With typical switching times of ~5-8 microseconds, operating at cycling frequencies above 50 kHz yields short on or off times, with most of the period spent in transit between the two states. These switches were operated at 60 kHz, with a cycle period of 16.7 µS. In this case, the effective duty cycle for on time was on the order of 10%.
A switch as described above was operated at -30 volts bias using a trapezoidal waveform at a repetition frequency of 60 kHz. The effective duty cycle of the switch was 10%. At this rate, the switch accumulated cycles at the rate of 216 million cycles/hr. The switch was run for a total of 476 hours without failure, accumulating a total of 102.8 billion cycles. During this process, snapshots of the slow-speed (1 kHz) and high speed (60 kHz) operating curves were recorded twice a day.
The high-speed curves yielded information regarding operating at high cycling frequencies, while the slow-speed curves yielded diagnostic information regarding the pull-in and release voltages over the duration of the test. From the high-speed operating curves, data points were accumulated at the minimum and maximum (fully up and fully down) states to determine if there was any drift in the RF performance of the device over the duration of the test. Figure 2 displays the minimum and maximum RF detector signal for the duration of the test. The off state (contact up) was very stable over time. Though there was some drift in the on state (contact down) detector signal over time, it is hard to quantify how much is generator or detector drift over three weeks of testing.
While achieving 100 billion switching cycles without failure is a noteworthy achievement, it is important to understand how those results were achieved. These results pull together several years of development in quantifying dielectric charging, improving testing techniques, and designing for reduced dielectric charging. The results show that a switch capable of order-of-magnitude impedance change can operate for extended periods without charging-induced failure. It also demonstrates that the underlying mechanics of MEMS devices are robust and can cycle for extended periods.
This article was written by C.L. Goldsmith and D.I. Forehand of MEMtronics Corporation, Plano, TX, and Z. Peng and J.C.M. Hwang of Lehigh University, Bethlehem, PA. For more information, visit http://info.hotims.com/22922-545.