A Case Study in Successful First-Pass Filter Design

by Bill Weedon, Applied Radar, Inc. & Chris DeMartino, Modelithics

Filters are critical components used in nearly every radar and communications system to pass signals within a certain frequency range while rejecting others that fall outside the desired band. When a filter is needed, one option for system manufacturers is to simply purchase one from a vendor. A range of filter vendors are currently on the market that could potentially provide a solution for a given requirement. Alternatively, a system manufacturer could choose to design its own filter rather than purchase one from an outside vendor.

Buying or designing a filter is a decision that must be determined by various factors like cost, lead times, design tools, etc. One company that recently had to make such a decision is Applied Radar, Inc. Located in North Kingstown, R.I., Applied Radar manufactures defense-based radar, communications, and electronic-warfare (EW) systems as well as products for commercial applications.

This article presents a case study of a bandpass filter that Applied Radar designed with first-pass success. Both Ansys Nuhertz Filter Design software and the Cadence AWR Design Environment were used for the design process. In addition, Modelithics passive-component models were utilized to simulate all of the lumped components in the design.

Choosing to Design Rather Than Buy 

Applied Radar builds many frequency converters, many of which must provide a final intermediate-frequency (IF) range that is specifically determined by the customer. In addition, many newer systems require wideband instantaneous-bandwidth performance that often ranges from 500 MHz to 1 GHz. Of course, the proper filters are needed in these systems to achieve the desired performance.

A recent project required a bandpass filter with a passband frequency range of 950 to 1,450 MHz. This filter was needed for use in the IF stage of a frequency converter. For this requirement, rather than purchase a filter from a vendor, the company decided to design one itself that would be realized by mounting surface-mount inductors and capacitors on a printed-circuit board (PCB). The chosen substrate for this filter was 20-mil-thick Rogers RO4003C.

Using the Right Design Tools

Without question, designing a filter that achieves the desired specifications with only one design iteration requires the use of the proper simulation tools. As mentioned, for this filter, Applied Radar used a combination of three design tools: Cadence® AWR Design Environment® software, Nuhertz FilterSolutions, and Modelithics Microwave Global Models™. Applied Radar made the decision to utilize these design tools after failing to achieve success with a “cookbook-based” design approach that the company attempted earlier.¹

The design process began in FilterSolutions, which is a tool used to design various types of filters, such as lumped-element and distributed implementations. Figure 1 shows the FilterSolutions user interface in which Applied Radar specified a seventh-order, lumped-element Chebyshev filter. The design is also specified to have equal inductors. Figure 2 shows the filter schematic with ideal component values.

Next, Modelithics Microwave Global Models are specified for use (Figure 3). Microwave Global Models represent entire families of inductor and capacitor components. These models scale with part-value, substrate, and solder pad dimensions, among other designer-friendly features. For this filter design, the AVX 08051A series of surface-mount capacitors is selected for all capacitors, while the Johanson L-07 wirewound series is selected for the inductors. FilterSolutions automatically sets the part values of the Modelithics Global models. Once the user is finished selecting the synthesis options and component families, the design can be pushed from FilterSolutions to AWR Design Environment software for further optimization. 

 In addition, before the design is pushed from FilterSolutions to AWR Design Environment software, the dimensions of the microstrip interconnects can be specified. Note that the interconnects are selected for optimization, meaning that their dimensions can later be optimized in AWR Design Environment software. Thus, both the part values of the component models and the interconnect dimensions are configured to be optimized.

The next step is to export the filter to the AWR Design Environment software. Figure 4 shows the schematic of the filter in AWR® Microwave Office® software after exporting from FilterSolutions. Notice how the microstrip interconnect lines together with the microstrip T-junction elements have been automatically placed in the filter schematic.

All that is needed at this point is to perform an optimization, in which both the part values of the components and the microstrip dimensions are automatically adjusted to achieve optimal performance. As mentioned earlier, FilterSolutions automatically handles the configuration, simplifying the process for the designer. Figure 5 shows the results of the AWR Microwave Office circuit simulation after optimization.

 While not performed in the original design process for this particular filter example, designers may also want to use the AWR AXIEM® 3D planar electromagnetic (EM) simulator to perform an EM/circuit co-simulation. In an EM/circuit co-simulation, the microstrip interconnects and vias would be analyzed with AWR AXIEM software, while the components would still be analyzed within the Microwave Office circuit simulator. For comparison purposes, we can show the EM/circuit co-simulation results here. Figure 6 shows the EM structure of the filter that will be analyzed with AXIEM. Note that for both the circuit simulation already presented and the EM/circuit co-simulation, the Sim_mode parameter of all models is set to 0. This setting enables the models to account for all real-world parasitic, substrate, and solder-pad effects.

 Figure 7 shows both the circuit simulation and EM/circuit co-simulation results of the optimized filter. The results reveal that the EM/circuit co-simulation results do not stray too far from the results of the circuit simulation. Therefore, EM/circuit co-simulation may not be necessary for this design. Moreover, Table 1 shows the final part values of the filter.

 Comparing Real and Simulated Performance

Following the design process, the filter was built and tested with the same components from the final simulation (Figure 8).

 Figure 9 shows the measured data of the filter along with the circuit simulation and EM/circuit co-simulation results. One can see that the measured data agrees well with the results from both simulations. It is believed that the slight difference between measured and simulated S₁₁ is either due to component tolerances or the fact that the SMA connectors used to measure the filter were not incorporated into the simulation.

 In addition, Figure 10 shows wideband measured data of the filter along with the final simulated results. These results illustrate how the simulations not only predicted the passband performance, but they also predicted the out-of-band performance well.

 In closing, the design flow presented here made it possible to achieve first-pass success with a solution that converged in minutes rather than hours or days. The combination of AWR Design Environment software, Nuhertz FilterSolutions, and Modelithics Microwave Global Models for lumped passive components allowed for a very direct workflow that produced excellent results. Those with similar needs may want to consider the tools and semi-automated design flow illustrated with this example. 

Note that a complete AWR simulation file for this filter can be downloaded from the Modelithics website.

References

1. G. Matthaei, L. Young, E.M.T. Jones, “Microwave Filters, Impedance-Matching Networks, and Coupling Structures,” Artech-House, Norwood, MA, 1980.

(676)