In a perfect world, an RF design engineer would like to see a switch with a variety of near ideal specifications. These would include low insertion loss, high isolation, extreme linearity, long term reliability, a wide frequency bandwidth, high power handling capability, low actuation voltage, very low cost, small size and a fast switch time. For real world RF design, variances on ideal specifications can be tolerated for many applications. In general, an insertion loss much less than 1 dB, isolation much greater than 20 dB and linearity with IP3 much greater than 60 dBm will meet many current day RF applications. Throw in a frequency bandwidth of between 0.4 and 6 GHz, power handling capability much greater than a watt, an actuation voltage of 2.7 volts and a switch time less than 5 microseconds, and the RF switch could be used for almost any wireless application.

Ideally cost, size and reliability of a switch would also be minimized to the point of being inconsequential. So the designer looks forward to switches that cost much less than a dime, have a size well below 0.5 square millimeters and will last at least three years for typical consumer operating conditions.

Although semiconductor RF switches can be optimized to come close to the ideal for certain specifications, this requires sacrifices in other performance areas. Alternatives to RF-MEMS Switches such as PIN Diodes and Solid State Switches have good high frequency passing capabilities. But PIN Diodes suffer from an increase in resistance (skin effect) above the X-band range (8-10 GHz) as well as higher current consumption due to individual bias requirements for each Diode. While less expensive than PIN Diodes, Solid State FET Switches have unique problems with insertion loss and isolation at frequencies above 6 GHz unless the switch is designed as an absorptive or T-switch.

And their limitations don't stop there. Other problems include their frequency limitations above the 1 GHz range, where their insertion loss and isolation tend to breakdown. Another problem is their non-linear switching behavior and related signal to noise issues.

Overcoming RF-MEMS Actuation Voltages

Actuation voltage has been an issue with RF-MEMS, traditionally, because of the nature of the technology and the types of microstructure used. RF-MEMS have in the past required an actuation voltage of anywhere between 20 volts and 120 volts, which is far too high to be practical in many portable applications without the use of voltage multiplier or charge pump circuits. On the other hand, PIN diodes and FETs have been generally available with switch voltages in the same order as CMOS circuits (anywhere between 1.5 Volts and 5 Volts) – without the need for any extra circuitry.

In order to overcome the actuation voltage issue with RF-MEMS, designers have come up with different mechanical switch schemes that show it is possible to reduce the actuation voltage to levels below 10 Volts. Furthermore since electrostatic driven MEMS devices do not require significant current to actuate them, the higher voltage generation problem is simplified. Requisite voltages can be provided by a very simple and low cost 'charge pump' circuit (the charge pump circuits, which can be integrated into the chip, are used to convert the 3 Volt CMOS supplies to the 20 plus volts needed to actuate the micro electro mechanical relay). Such a circuit is usually very small and low cost since high power and current is not required.

There are other options as a result of new low-cost MEMs packaging techniques. The new packaging enables the MEMS devices to be built on top of the low-cost silicon used to perform control logic and voltage generation functions.

Figure: WiSpry Illustrates One MEMS Fabrication Process that Layers the

Mechanical Switch On Top of the Actual RF Silicon Circuitry

Source: WiSpry, Inc.

RF-MEMS Switching Times Issues

Switching time can be critical for a number of RF applications such as wireless transmit/receive (T/R) switches in GSM cellular phones. For some time this was considered "THE" application for MEMS and so the switching time became the key technical spec that was driving development and research. There are advances on the horizon that could make even this requirement possible. Driving the switch in both directions would certainly improve the speeds but proves difficult in practice. There is also a physical limit to how fast a micro mechanical device can be moved. Its more likely that MEMS will initially find its way into the compelling multi-mode spaces that are rapidly evolving than solve the T/R switching function. Overall, it remains difficult for RF-MEMS to compete with PIN diodes, SOI and GaAs in regards to switching time. These devices have switching times in the 1 nanosecond range, easily 1000 times faster than RF-MEMS switches that now have switching times from 1 to 300 microseconds

RF-MEMS 50 GHz Plus Bandwidths Enable a New World of Applications

The wide bandwidth of RF-MEMS makes them a very strong candidate to replace GaAs MESFETS now used in multi-band and multi-antenna cellular phones. One of the reasons is the bandwidth. One can expect an RF-MEMS switch to have a bandwidth that ranges from DC to beyond 60 GHz. It is not uncommon to see high-end RF-MEMS bandwidths on the order of 100 GHz.

A look at other specifications builds even more of a case for the use of multiband and multimode RF MEMS devices. A typical GaAs MESFET will provide an insertion loss on the order of 0.4 dB and isolation levels in the range of 24 dB. RF-MEMS switches, in general, will have on-resistances 3 to 4 times lower than FETs, translating into losses three or four times lower. With on-resistance levels in the range of 0.5 ohms to 1 ohms, a MEMS switch can offer insertion loss less than 0.1 dB and isolation greater than 40 dB. The advantage of the packaged GaAs SPDT switch is a cost that is less than a quarter . This cost advantage is however eroded when one considers the advantages of a RF-MEMs based system. A RF-MEMS solution will eliminate filters and reduce the component count of the filters used. This adds up to lower power, smaller footprints and associated lower system costs for batteries and packaging.

RF-MEMS Power Handling Capability Climbing Upward

As far as power handling capacity goes, RF-MEMS devices compare favorably to PIN Diodes and FET switches. The typical power handling capabilities of PIN and FET solutions run in the 35 dBm (from 1W to 10 W) range, while a well designed RF-MEMS Switch can pass up to 43 dBm of continuous power and up to 45 dBm in quick pulses.

One of the reasons is package design and the size of a physical MEMS contact head rather than any intrinsic technology limit. Additionally though, the design of the actual MEMS device strongly affects the power handling capability. Careful thermal and electrical design is critical to achieve reliable operation at powers greater than 100 mW. Direct metallic contact switches, while having the greatest bandwidth, concentrate the current through miniscule electrical contacts. For high power operation, these contacts must be closely coupled to a good heat sink. On the other hand, capacitive contact switches spread the current over a larger area and thus can handle more RF power in the closed state for a given thermal design (without the need for a good heat sink). Capacitive MEMS switches are still however limited in the open mode by the forces generated by the RF fields. At sufficient power levels, the capacitive switch will be "self-actuated" by the RF signal. The smaller contact area of metallic contact switches works in their favor leading to much higher self-actuation powers.

The capacitive circuits such as the switch actuators in an electro-static RF-MEMS switch are strictly voltage driven and so their power consumption requirements are very low. Driving high currents into such a switch can result in permanent damage due to the heat and resulting stresses on the RF-MEMS actuator.

Overall, the power handling of MEMS devices can be quite high, but MEMs power capacity depends on the device size and other design material options. The advantages of MEMS are that the dissipation is less due to the low insertion loss and that highly thermally conductive materials are available.

The key obstacle to even higher power handling for MEMs is simply the small size. The heat must be spread quickly away from the tiny device. For solid-state devices the current is spread throughout a larger device, leading to easier heat dissipation problems near the device junctions. However, the larger total power dissipation of solid-state devices requires higher-performance packaging for equivalent power handling. PIN diodes can have very low RF dissipation but they are also conducting DC current for their switching. The heat generated by the DC current lowers their RF power handling capability.

Power Consumption in the Microwatts and NanoJoules

RF-MEMS offer extremely low-power for the static condition (when the switch is closed or open), the dynamic switching condition (the power required to actually switch the relay) and RF signal transmission modes (when an RF signal passes through the actual switch).

For RF-MEMS devices, power consumption during switching is a function of the actuation approach. For electrostatic operation, the drive terminals are a capacitor. Thus there is virtually no static power consumption in the off-state. During a switching operation, charge must be moved onto or off of the capacitor. The dissipation in this operation is set by the resistances of the control lines, which can be very low. If the actuator capacitance is 1 pF and the actuation voltage is 10V, the stored energy in the actuator for the charged state is 50 pico-joules and 25 pico-joules is dissipated per switching event. If the switch is toggled at a 10 KHz rate, the dissipation would be less than a microwatt.

Finally, DC Power is seen across the switch contact points and can run in the 5 to 20Watt range.

Near Perfect Linearity in RF MEMS Leads to Near Zero Harmonic Distortion

Linearity is another key selling point of MEMs devices. Semiconductor devices are voltage controlled non-linearly current sources and hence their resistances and capacitances are easily modulated by the RF voltages. Because the I-V curves of electromechanical switches are much more linear over the switch voltage range than semiconductor devices and because their capacitances are not modulated by the RF voltage, they introduce far lower harmonics than semiconductor based switches do. This results in near perfect linearity – specified on the order of 70 dBm at the third intercept point, resulting in negligible Total Harmonic Distortion (THD) at operating power levels.

Emmanuel Rodriguez C.I. 17208374
Asignatura: CRF
Fuente: http://www.perfectdisplay.com/feature_articles/rfmems/rf_mems_files/rf_mems_page_2.htm