Advanced ceramic materials and a high-resolution circuit fabrication process can combine to provide excellent electrical performance in small sizes. As demonstrated in Part 1 of this white paper, the combination has been applied at Dielectric Laboratories, Inc. (DLI) to the production of XTREME-Q? high quality-factor (Q) resonators at microwave and millimeter-wave frequencies. This concluding part of the article series will show that the high-Q nature of these materials can also be applied to microwave and millimeter-wave filters at frequencies exceeding 67 GHz. In recognition of the unconventional nature of these new filters, DLI refers to the departure with conventional frequency-management design and manufacturing approaches as a " disruptive technology."
These capabilities in ceramic resonators and filters are not the results of new-found knowledge but come from years of designing and manufacturing extensive lines of high-Q Single-Layer Capacitors (SLCs) and Multi-Layer Capacitors (MLCs). Manufactured to precise tolerances, these extremely stable capacitors maintain their values over time, temperature, and vibration in the most demanding commercial, industrial, and military applications. The capacitor expertise is a building block for the higher-level design and manufacturing services offered as part of DLI's disruptive technology.
The disruptive technology essentially uses capacitors, inductors, and other high-frequency structures and components to form high-performance components on ceramic substrates. DLI offers a variety of different proprietary ceramic formulations in support of the technology, including CG, CF, and FS materials. These low-loss materials feature excellent stability over temperature (see Fig. 1 in Part 1 of this White Paper), as judged by their temperature coefficients of frequency (a measure of the variations in frequency over a certain temperature range). For temperatures from ?20 to +120°C, the temperature coefficients of temperature for the CG, CF, and FS ceramic materials range from 2.3 to 8.8 PPM/°C. In terms of resonators and filters, these numbers translate into frequency shifts of less than one tenth of one percent at extremely wide temperature extremes.
Compare this to the temperature stability of alumina (Al2O3) substrates commonly used in microwave circuits, at about 120 PPM/°C. These ceramic materials offer greater miniaturization capabilities compared to alumina and other substrates as a result of the higher dielectric constants. And they are stable over time, having been proven for decades in the company's lines of SLCs and MLCs.
Understanding Filters
The design and production expertise detailed in Part 1 of this White Paper (Resonators) also serves the synthesis and fabrication of highperformance, extremely temperature stable microwave filters from 500 MHz to beyond 67 GHz. The disruptive ceramic technology, with the aid of precisely controlled photolithography, has been applied to a wide range of lowloss, high-rejection filters, including microstrip, cavity, ring, and symmetrical dual-mode resonator types. DLI has already produced lowpass, highpass, and bandpass filter designs, including Bessel, Chebyshev, edge-coupled, hairpin, interdigitated, and custom variations. With passbands as wide as 30 percent, typical performance levels include insertion loss of less than 2 dB, return loss of 15 dB and minimum rejection of 45 dB, with tight amplitude ripple and group-delay variations.
The performance of individual filters depends upon a number of factors, including the number of filter poles, the architecture, the resonator Q, and the percentage bandwidth. In order to appreciate the performance of the filters fabricated with DLI's disruptive ceramic technology, it might help to review the basics of how a filter's performance is evaluated. Although the essential function of an RF/microwave filter is simple—to remove unwanted signal energy across a specific band of frequencies while leaving remaining portions of the spectrum unaffected—the design and implementation of high-frequency filters with acceptable performance traits is often a challenge even for experienced engineers.
Filters come in essentially four types: bandpass, band-reject, lowpass, and highpass filters. A bandpass filter channels signals with minimal attenuation through a range of frequencies known as the passband, and rejects signals at frequencies above and below the passband. A band-reject filter (also known as a notch filter) is essentially the opposite of a bandpass filter. It rejects signals across one band (known as the stop band) and allows signals to pass with minimal attenuation at frequencies above and below the stop band. A lowpass filter channels signals with minimal attenuation below a specified cutoff frequency, while rejecting signals above that cutoff frequency. The cutoff frequency is commonly a point at which signal attenuation reaches 3 dB. A high pass filter is essentially the opposite of a lowpass filter, rejecting signals below the cutoff frequency and passing signals with minimal attenuation above the cutoff frequency.
Filters are judged in terms of a number of performance parameters, including insertion loss, return loss (or VSWR), rejection, ripple, amplitude-versusfrequency response (or selectivity), group delay (how long a signal takes to propagate through a filter), phase response, and even quality factor (or Q). In a bandpass filter, insertion loss is the amount of signal attenuation above a 0-dB level that would be represented by an ideal transmission line in place of the filter. Insertion loss occurs due to a filter's dissipative elements (the resistors, inductors, capacitors, and transmission lines). Rejection is the amount of signal attenuation at specified points above and below the passband or center frequency, including the insertion loss.
Every filter has characteristic impedance, measured in ohms, with 50 ohms being typical. While the characteristic impedance may be consistent across the passband, it tends to vary once stop-bands are approached. At a system level, matching the terminal impedances at the I/O of a filter is critical to achieving good filter performance.
Bandpass filters are defined in terms of their center frequency (CF) and the width of their passband. But defining the CF of a bandpass filter is not obvious, since it can be done arithmetically or geometrically. Generally the geometric definition is employed in the filter design process, and the arithmetic definition is used to specify a filter. The arithmetic center frequency is simply the sum of the lower and upper bandedges divided by two. For example, for a bandpass filter with 3-dB frequencies of 900 and 1000 MHz, the arithmetic center frequency is found by (900 + 1000)/2 = 950 MHz.
Like a resonator, a bandpass filter exhibits a Q. In a filter, the Q is the ratio of the midband frequency to the bandwidth. A narrowband filter, for example, with 3-dB band edges of 950 and 1000 MHz (center frequency of 975 MHz) has a Q of 975/50 = 19.5. A bandpass filter with wider 3-dB bandwidth of 500 to 1000 MHz would have a much lower Q of 750/500 = 1.5. Essentially, filter Q is related to the bandwidth, with narrower filter bandwidths resulting in higher filter Q values.
In fabricating filters, high-Q circuit elements (such as transmission lines, capacitors and inductors) are desirable for high-performance filter responses. Low-Q circuit elements tend to yield higher passband insertion loss and lower stopband attenuation. And lower-Q elements lead to a rounding of a filter's response, with poorly defined filter skirts. In general, circuit elements should have a Q of 100 or better commensurate with filter Q. Filter insertion loss is proportional to the ratio of filter Q to the resonator unloaded Qu. Where Loss = Q/Qu.
With additional sections, the rejection of a given filter design can be increased, but with additional complexity and insertion loss (due to additional resonant elements). Historically this resulted in difficult and time-consuming tuning when a fabricated filter misses its design targets (such as CF or BW). Filters fabricated with DLI's disruptive ceramic technology provide excellent performance with no tuning and can be produced in large quantities with the repeatability of a photolithographic process.
Fabricating Filters
The high-performance ceramic materials can do for filters what they do for resonators: bring the advantages of stability with temperature and time, small size, and high performance to designs from 500 MHz. The response of DLI's miniature filters remains stable across wide temperature ranges, as shown for the 3.6-GHz filter in Fig. 9. The measured S21 and S11 responses of this filter remain virtually unaffected at temperatures from ?20 to +120°C with the filter exhibiting no changes in amplitude or frequency.
Table 3 offers a sampling of typical filter specifications for different frequencies and bandwidths. The 2.14-GHz filter is an interdigitated type. The 3.5-, 4.2-, and 6.5-GHz filters are symmetrical dual-mode resonator filters (SDMRFs). The 5.6-GHz filter is an edge-coupled design, while the 9.7- and 37-GHz filters are end-coupled designs. In addition to these types of filters, DLI offers a wide range of filter types. These miniature ceramic filters are about 1/15th the size of printed-wire-board (PWB) filters with greatly improved repeatability and temperature stability.
The first filter in Table 3 is a 2.14-GHz seven-pole Chebyshev bandpass filter with 1.8-dB insertion loss (Fig. 10). The filter, which does not require tuning in production, is ideal for surface-mount applications. It exhibits outstanding temperature stability of ±15 PPM/°C fabricated on a ceramic substrate with dielectric constant of 23. The filter measures 0.4 × 0.75 × 0.035 in.
The 5.6-GHz edge-coupled filter is a five-pole Chebyshev bandpass design with 3-dB bandwidth of 410 MHz (7% BW). It has insertion loss of 2.2 dB with almost 40-dB rejection of unwanted signals at 5.2 and 6.2 GHz (Fig. 11). The ultimate rejection is more than 75 dB. As with the 2.14-GHz bandpass filter, the 5.6-GHz design is fabricated on a ceramic substrate with dielectric constant of 23. It achieves temperature stability of ±15 PPM/°C and measures a mere 0.9 × 0.25 in.
Moving up in frequency, the 9.7-GHz bandpass filter in Table 3 is a seven-pole, end-coupled design that was fabricated to match a footprint so it is larger than it needs to be. It has a 4% (400-MHz) bandwidth. The insertion loss is less than 2.7 dB at the center frequency, with as much as 70-dB outof-band rejection (Fig. 12). This type of end-coupled resonator topology is suitable for low percentage bandwidths of 2% to 5% and high out-of-band rejection. The filter measures 1.1 × 0.1 × 0.03 in. For comparison, the same response filter, if fabricated on standard printed-circuit-board materials, would exceed 3 inches in length.
The highest-frequency filter in Table 3 is a 37-GHz bandpass filter. The three-pole Chebyshev design features thin-film gold metalization on fused silica, with circuitry precisely patterned using state-oftheart photolithography techniques. It measures 0.32 × 0.10 × 0.01 in., features a 600-MHz bandwidth (1.6%) centered around 37 GHz with less than 2.2 dB insertion loss at the center frequency (Fig. 13).
These measurement results are an example of test capability that extends to 67 GHz using vector network analyzers with both test fixtures and coplanar probes.
DLI has also designed and fabricated high-performance dual-mode resonator filters using highly integrated surface-mount technology (SMT) as well as wire-bond approaches. Available with gold and nickel/gold metallizations, the filters share the temperature stability of DLI's other ceramic filters with low insertion loss (about 0.5 dB) and excellent rejection ( exceeding 65 dB). As an example, the low insertion loss and high rejection can be seen in the measured responses of a 2-GHz dual-mode resonator bandpass filter (Fig. 14).
This design has less than 0.5 dB insertion loss across a 500-MHzwide passband, with filter skirts that cleanly drop to more than 60 dB rejection.
Additional wideband dual-mode resonator filters include compact four-pole designs with bandwidths as wide as 20 percent. For example, a bandpass filter with 7-GHz center frequency that exhibits less than 0.7 dB insertion loss and rejection that drops below 40 dB. It measures 0.3 × 0.15 × 0.03 in. for ease of placement on the tightest circuitboard designs. And a 10-GHz, four-pole, dual-mode resonator bandpass filter (Fig. 15) that maintains the excellent insertion-loss performance of less than 0.7 dB across the passband, with outofband rejection exceeding 30 dB. The 10-GHz bandpass filter measures 0.15 × 0.08 × 0.03 in.
For outstanding selectivity, DLI has produced several symmetric dual-mode resonator filters with as much as 40-dB rejection measured only 1 percent bandwidth from the center frequency. For example, a 4.2-GHz symmetric dual-mode resonator bandpass filter exhibits insertion loss of only 2.2 dB in the passband (Fig. 16) while achieving rapid rejection just 40 MHz from the center frequency and more than 70-dB rejection at the extreme band edges. Similarly, a 6.5-GHz symmetric dual-mode resonator bandpass filter (Fig. 17) matches the performance of the 4.2-GHz filter with similarly high rejection. Both filters feature integrated traps to suppress harmonics, and are designed on temperature-stable ceramic material measuring 0.7 0.35 0.03 in. DLI has also designed patentpending ceramic cavity filters. One example is is shown with a center frequency of 10.5 GHz (Fig. 18). Suitable for local-oscillator filtering, image filtering, and selecting desired harmonics in multiplier chains, the compact filter exhibits a passband of 270 MHz with less than 1.3 dB insertion loss. It measures 0.17 × 0.2 × 0.03 in.
In cases where higher-order bandpass filter responses are required, ceramic resonators can be cascaded in series to achieve desired results. A single ceramic cavity resonator, which generates one transmission zero, is the basic building block for a cascaded filter. A higher-order bandpass filter is designed by cascading several of these building blocks to generate the required transmission zeros (Fig. 19). The synthesis and design of these cascaded filters is based on models that cascade the designed cavity resonator in the vicinity of the center frequency. In designs already fabricated relative bandwidths as wide as 3% have been achieved.
Using a novel resonator-toresonator attachment approach, DLI has developed high-performance filters with their patent-pending resonator technology that can potentially obsolete waveguide cavity filters. These miniature, resonatorbased filters offer electrical performance comparable to waveguide-filters, at a fraction of the size, cost, and general burdens associated with waveguide devices. Figure 20 shows an example of these miniature " waveguidereplacement" filters.
In fabricating these miniature ceramic filters, lower-frequency designs rely on folded resonators while higher-frequency designs generally use end-coupled configurations. By controlling the amount of coupling, the filter's response can be tailored to a particular requirement. The filter's impedance is managed by careful placement of the input/output structures. Through the use of classical filter models and computeraidedengineering (CAE) tools, practical filter solutions can be developed to meet the most demanding set of requirements. Final implementations are verified by means of electromagnetic (EM) simulation tools and proprietary software.
As with the custom resonators, designing a miniature custom ceramic filter requires establishing a set of requirements, including the type of filter (BP, HP, and LP), the center frequency, the 3-dB bandwidth, insertion loss at the center frequency, and the return loss at the center frequency. Other considerations include the amount of rejection required and at what low-side and high-side points, the power-handling capability, the required operating-temperature range, what type of mounting (surface-mount or microstrip) is required, and the size limitations for the filter.
In addition to filters, the ceramic-technology can also be applied to the fabrication of miniature diplexers and duplexers. These filter-based components are used to combine and/or separate frequencies, typically in communications receivers and transceivers. A diplexer is a three-port device that acts as two different filters, covering several different frequency bands or communications channels, such as cellular and PCS frequencies. A duplexer is a threeport device with a common input port that feeds two frequency bands, such as a transmit band and a receive band. As with the filters, miniature ceramic diplexers/ duplexers can be fabricated from 500 MHz with typically less than 2-dB insertion loss and minimum return loss of 15 dB. These ceramic diplexers and duplexer can be fabricated with gold or nickel/gold metallization and achieve isolation between bands of better than 50 dB. As an example, DLI fabricated a duplexer for UMTS applications. The thin-film ceramic duplexer (Fig. 21) eliminated two separate filters and an isolator in 1/15th the size of a PWB implementation, with greatly improved temperature stability and unit-to-unit repeatability.
This high-performance ceramic technology is also well suited for bias filter networks and gain equalizers. DLI has developed a series of bias filter networks designed to filter RF signals from bias and control lines from 10 MHz to 40 GHz. Ideal for biasing GaAs FET devices and monolithic microwave integrated circuits (MMICs), a single bias filter network can simplify assembly by replacing numerous separate discrete components. Similarly, the gain equalizers have a positive "gain slope," which can be used in commercial and military applications to flatten the gain response of an active module from DC to more than 40 GHz.
In summary, for resonator and filter-based solutions that must be small in size, stable with time and temperature, and with the performance associated with high-Q circuits, these patent-pending XTREME-Q ceramic components show how a disruptive technology can improve on the status quo.
