To alleviate issues in switched filter banks typically used in MILCOM radios, a miniaturized tunable UHF filter has been developed using MEMS switches and lumped element components tuning over the 225-512MHz range.
Switched filter banks typically used in military communications (MILCOM) radios have several inherent disadvantages, such as size, weight and high-power dissipation. To alleviate these issues, a miniaturized, high-power tunable UHF filter has been developed using MEMS switches and lumped element components tuning over the 225-512MHz range. The filter is configured in seven discrete steps using multiple single-pole/four-throw (SP4T) switches to adjust the resonant frequency by changing inductor values.
The novel design approach can handle 60W and achieves low insertion loss below 1.5dB, up to 3dB less than traditional switched filter banks, which use solid-state-based switches. More importantly, the single-resonator architecture creates an ultra-small form-factor of 3.4-by-1.6-by-0.7in, which represents a greater than 90 percent reduction in volume when compared to solid-state-based solutions.
The miniaturized UHF tunable filter bank uses MEMS switches developed by Menlo Micro, which have inherent advantages over solid-state technologies. These switches feature low losses, large power input handling, excellent RF linearity, and reliable operation over 3 billion switching operations. One critical feature to note in comparing MEMS-based designs with solid state technologies is the very low on resistance (Ron), which results in the lowest possible insertion loss and an ultra-low off capacitance (Coff), resulting in greatly reduced signal leakage. These lower parasitics offer improved “on” and “off-state” operation, enabling the creation of a high-Q tunable resonator with little degradation in RF performance when compared to a fixed resonator. This enhanced behavior also offers key advantages with increasing RF input power and minimal performance variation over the -40°C to 85°C operational temperature range.
The filter’s unique design topology uses four SP4T MEMS switches with discrete inductors and capacitors, configured as a four-pole bandpass filter. The design can easily be scaled to increase performance by adding MEMS switches and components for additional filter poles and zeros, without drastically changing the losses or the size of the filter. Although this filter design operates in the UHF band, it’s not limited to these frequencies as the MEMS switches can operate from DC to 18 GHz, enabling greater broadband operation. Let’s take a closer look at the design and performance of the UHF discrete tunable filter compared to traditional switched filter banks, as well as the contribution of the MEMS switches to the test results.
Tunable Filters versus Switched Filter Banks
Switched filter banks are key building blocks for frequency selective RF and microwave portions of many systems in both receive and transmit paths. The topology of an “N channel” switched filter bank is simply “N” number of filters placed between a pair of single-pole/N-throws (SPNT) switches. This approach is very effective when the number of channels (N) is small, but as N increases, design complexity and size rapidly become unmanageable. The degraded insertion loss performance due to cascading SPNT switches causes further power dissipation, which can create significant thermal challenges for the entire system.
For larger numbers of channels, a tunable filter can provide a higher-performance, more compact solution than a switched filter bank. Traditional tunable filters based on varactor diodes had significant limitations in terms of linearity and power handling. The use of MEMS switches in the frequency selective elements gives substantial advantages.
Figure 1 Switched filter bank (a) versus tunable filter (b).
The key challenges in designing a tunable filter include maintaining high Q factors, maintaining performance with high power handling and temperature variation, and ensuring filter performance is maintained over the full tuning range.
Designing an RF-MEMS-based Tunable Filter
The tunable filter shown in Figure 2 uses four MEMS-based MM5130 RF SP4T switches from Menlo Micro to realize a four-section filter tunable from 225MHz to 512MHz in discrete steps. The switched filter bank (SFB) measures 2.5 x 1.25 x 0.25 inches without connectors.
The filter is configured to tune over the frequency range in seven discrete steps. With 4-bit binary control from the SP4T switches, a total of 16 steps can be implemented if required.
Figure 2 Example tunable filter covering 225MHz to 512MHz.
Comparison to Typical Solid-State SFB Designs
Table 1 compares the specification of the tunable filter with a typical high-power switched filter bank, and Figure 3 depicts the tunable filter compared to a typical switched filter bank. Key performance benefits include low insertion loss, small size, less weight and low power consumption.
Table 1 RF-MEMS based Tunable Filter specifications
Figure 3 Tunable filter compared with a typical switched filter bank.
Theory of Operation
The key aspect in designing a multi-section tunable filter is to use a structure that maintains the desired filter response over the full tuning range while varying as few component values as possible. This 4-pole lumped element bandpass filter design varies the inductor value associated with each pole to tune the filter’s center frequency.
This topology, shown in the Figure 4 filter schematic, inherently provides good harmonic rejection performance due to the design’s large number of shunt capacitors.
Figure 4 Filter schematic.
The capacitors between each inductor (i.e., the “PI” section) have a constant ratio of reactance between each other as frequency varies. This design allows filter response to be maintained over a wide frequency range. Since only the inductors are varied and the capacitors remain constant, the bandwidth percentage remains constant with frequency. As a result, the filter bandwidth increases with rising frequency, which is generally desirable as the insertion loss and harmonic rejection remain constant. The variable inductors are realized by switching fixed inductors in and out of the circuit using MEMS switches, as shown in Figure 5.
Figure 5 Switched inductors.
Switches S1 through S4 are the arms of an SP4T MEMS switch. Each arm may be controlled independently, providing a total of 16 possible switching states. Inductors L1 through L4 may be switched in parallel with the fixed inductor Lf, effectively lowering the overall inductor value. At high frequencies where minimum inductance is required, most of the switches will be closed, which effectively reduces the switch’s ON resistance as multiple paths are closed in parallel.
The MEMS switch is an ideal element for this type of application as it offers a combination of very low ON resistance (<0.5Ω) and very small OFF capacitance (0.03pF), which means parasitic resonances appear well out of the band of operation. When compared to other solid-state switch technologies, these are very low parasitics.
The MEMS switches are activated by electrostatic force and thus require a high-voltage source for switching operation. The switch’s gate is set for a bias of 0VDC, which places the metal cantilever beam in a non-deflected (off) state. Thus, the path between RF input and output is isolated with an air gap, similar to a traditional mechanical relay. When the gate is set to its required actuation voltage of +88V, the electrostatic force existing between the gate and cantilever beam is strong enough to cause downward deflection, forming a connection with the contact and closing the switch (on state). Given the low supply current necessary for the electrostatic operation, the design uses a single charge pump circuit to generate the +88V required by each of the sixteen (4 x SP4T) MEMS switches.
The filter circuit is very tolerant of different layouts due to the large number of shunt capacitors that can be incorporated into PCB traces to enhance design flexibility. It’s important to minimize stray capacitance in the interconnection between the inductors and the switch, which may form a parasitic resonance when combined with the lumped inductor. For this reason, the switched inductors are grouped closely around each individual switch, keeping trace lengths and widths as small as practical.
Small Signal Performance
Measured insertion loss and return loss met the target requirements, as shown in Figures 6 and 7. The filter response has good high side selectivity as expected, and the insertion loss is constant over the tuning range. The MEMS switch works very well for this application, and most of the insertion loss stems from the finite Q factor of the inductors, capacitors and PCB material. Of particular note is the consistency of insertion loss as the frequency is tuned across the band of interest.
Figure 6 Insertion Loss performance for 7 frequencies of the tunable filter.
Figure 7 Insertion Loss and Return Loss performance for channel 3: 250-340MHz of the tunable filter.
Power Handling and Thermal Performance
The MEMS switch selected for this design exhibits very low insertion loss of 0.15dB at 4GHz as well as third-order intercept greater than 85dBm, while capable of handling 25W of RF input power in a 50Ω system. In addition, the switch is designed to handle up to 160 V without damage, enabling it to be used safely in the resonant element of the filter. It was important to use a relatively low internal impedance level in this filter design. Figure 8 shows the maximum voltage gain relative to 50Ω seen by the switches. Within the filter passband, this voltage gain reaches a maximum of 3. This ensures that at higher power operation, the peak voltages are relatively low. However, when approaching continuous wave (CW) input power levels of 60W, the switches must handle >100V. This can be problematic for most solid-state technologies that must resort to transistor-stacking, which further degrades the switch parasitics.
Figure 8 Peak voltage gain seen by switch element relative to 50 ohm input.
The MEMS switch’s low loss and high voltage handling provide the filter with significant gains in power density and thermal performance. Compared to a switched filter bank experiencing as much as 3 dB additional insertion loss, the input power requirements for a given output level are significantly reduced, which provides additional power dissipation benefits.
Compared with the use of varactors, switches have much better linearity and power handling for tunable applications. The only limitation is the number of discrete tuning states available. Solid-state switches do not have the same combination of almost zero power consumption and high IP3 exhibited by the MEMS switch. This makes the MEMS switch an ideal candidate for discrete tunable filters.
The filter was tested up to 35 Watts CW (test set limitation) as referenced in the thermal plot in Figure 9. As expected, the majority the filter’s power dissipation appears in the inductors.
Figure 9 Filter thermal performance.
The elimination of solid-state devices from the filter design helps to significantly improve insertion loss stability over temperature. Figure 10 shows the small-signal testing over temperature.
Figure 10 Filter performance over temperature.
Summary and Future Directions
By closely integrating the switch and inductor combination to minimize parasitic capacitance, the frequency range can be extended significantly up to 6 GHz with existing switch topologies. This approach will outperform existing filter technologies, especially for emerging sub-6 GHz 5G NR RF infrastructure.
Given the switch’s very low ON resistance, the frequency of operation easily can be extended down to the HF/ VHF band without using large-value inductors as the frequency reduction can be achieved mostly by increasing the shunt capacitor values. As a result, air-cored (non-ferrite) inductors can be used at much lower frequencies, minimizing size and improving stability and linearity. This property also facilitates the realization of more sophisticated filters with greater selectivity to maximize rejection close to the passband.
Using additional switch arms allows more discrete steps to be implemented, further approximating a continuously tuned filter. This approach also allows the filter bandwidth to be reduced as the switching circuits are not a significant contributor to the overall insertion loss. A practical limit is in the region of 6-bit control, enabling a total of 64 discrete frequencies. In addition, using more switches to select multiple tunable resonators allows even more frequency bands without making significant sacrifices in overall insertion loss or size.
The enhanced performance of the miniaturized tunable filter stems from the switch’s unique, breakthrough MEMS-based core technology. The very low levels of parasitic off capacitance and very low on resistance result in a high-Q resonator that can replace multistage solid-state components and their associated biasing circuitry and thus reduce the filter’s size and weight. Since the switches are inherently low loss, this filter design can handle much higher power levels with lower heat dissipation than a traditional switched filter bank. These advantages reduce total filter cost while meeting the stringent requirements for emerging multiband radio systems.