Filters play a crucial role in virtually all modern electronic systems, ranging from simple amplifier circuits, where higher-order harmonics need to be eliminated, to complex modulated systems, where stringent adjacent channel specifications must be met. AWR’s new iFilter™ technology in Microwave Office® design suite helps engineers overcome these design challenges by providing a filter synthesis capability tightly integrated in the AWR Design Environment™. While iFilter includes both lumped and distributed filter synthesis, this article will focus on a distributed filter example. The example describes the use of iFilter to synthesize a hairpin bandpass filter with a target center frequency of 5.8GHz very close to required performance. The synthesized structure is then validated using AWR’s AXIEM™ 3D planar electromagnetic (EM) solver, the industry’s most accurate analysis tool. In general, the article will follow this flow: synthesize the filter with iFilter, analyze the synthesized filter with iFilter/AXIEM, optimize the filter with AXIEM, and analyze the optimized filter with AXIEM.

Part of the challenge in distributed filter design is balancing the limitations of the manufacturing process against the tight dimensional tolerances required to meet electrical requirements. A "good" filter design will not only produce the desired measured performance, but will also be well within manufacturing capabilities and, therefore, deliver a realizable and repeatable filter. Since difficult-to-manufacture geometries, such as very wide lines with very narrow gaps or very narrow lines with very wide gaps, are typically related to extreme impedance swings in the resonators, a good filter synthesis tool includes routines to automatically manage the impedances in a design and use all available analytical and empirical techniques to mitigate extreme impedances and produce easy geometries. These techniques include the transformations found in any filter text (such as Norton and Kuroda transforms, and m-derived and constant-k end sections).

For this distributed filter example, the following design goals will be targeted:
Center Frequency (f0): 5.8GHz
3dB Bandwidth: 200MHz
Passband Insertion Loss: <2dB
Passband Return Loss: >15dB
Stopband Insertion Loss: f < 5.4GHz: >40dB
f > 6.2GHz: >40dB
Material: Rogers RO6006 (r = 6.15), 0.010" thick, 1oz Cu
Technology: Simplest manufacturing process (printed structures only, no vias or external components)

This filter will have a fractional bandwidth as follows:

where FBW = Fractional Bandwidth
BW = Bandwidth
F1 = Lower 3dB Frequency
F2 = Upper 3dB Frequency

Common topologies, such as edge-coupled structures, begin to run into problems when fractional bandwidth (FBW) is less than about 15 percent. In general, an edge-coupled filter with this narrow FBW would synthesize with narrow lines and wide gaps—much larger than the line widths—and these possible numeric solutions translate poorly into physical designs. Therefore, alternative topologies such as a shunt stub need to be considered. From Figure 1 (shunt-stub topology), it can be seen that four poles are insufficient to meet the rejection requirement, while five poles provide enough rejection to meet all goals.

Figure 1: Four poles (left) and five poles (right)

But before proceeding with that topology, the choices should first be studied to determine what filter topology fits best within the manufacturing capability. Available options include shunt-stub filters, edge-coupled filters, hairpin filters, interdigital filters, combline filters, and stepped-impedance resonator filters. iFilter provides a very simple comparison feature in which different topologies (all of which meet the filter specifications) can be superimposed on each other and evaluated in terms of footprint and manufacturability. Figure 2 shows this comparison for three different options, shunt stubs with ¼-wave lines and ¼-wave stubs (orange), edge coupled (lighter blue), and hairpin (darker blue).

Figure 2: Comparison of three filter topologies

Based on Figure 2, and the requirement to use the least expensive manufacturing process, the hairpin design will be selected as the solution. This solution has by far the smallest footprint; it is designed to keep line widths constant for easy manufacturing and the entire filter is printed without the need for via holes or external components. (NOTE: the footprint could be reduced by selecting an interdigital or combline configuration, but both these topologies include extra manufacturing steps—interdigital filters require vias, and combline filters require vias and surface-mounted capacitors at the ends of the resonators and the slightly smaller footprint does not offset the increased costs to manufacture.) As part of iFilter's inputs, limits of 0.004" for both minimum line width and minimum line spacing have been specified. This is typical of modern printed circuit board (PCB) processes, so there will be a notification if the design violates those limits (see red-outlined boxes in Figure 1—at this stage of the design, the hairpin topology had not yet been selected and the chosen filter type violated at least one of these requirements.)

With the full set of design requirements in mind, the main iFilter window is set up to produce the desired filter. The electrical requirements are specified, the hairpin topology is selected, and the substrate information has been entered in the "Design Options" window, as shown in Figure 3.

Figure 3: Filter specification in main iFilter Dialog and substrate specification in Technology Dialog

Where applicable, iFilter also provides the ability to adjust secondary parameters in real time and see the results. In the case of hairpin filters, users are able to adjust the filter's nominal impedance (the short connecting lines between the resonators, in the U-turns, will have this impedance and the resonator lines will share this line width). If it is assumed that the area is limited to 0.360" x 0.300", the nominal impedance of the resonators can be adjusted to examine the effects of the line width on both performance and area required, as shown in Figure 4.

Figure 4: Footprint bounding box with Z0 = 45 (left) and Z0 = 55 (right)

Based on the filter's footprint, the Z0 = 55 version for the synthesis will be used. At this point, iFilter's work is almost done and it's time to hand off the rest of the design task to the Microwave Office circuit simulator. The filter's specifications have been set up as optimization goals (visible in all graphics that include the response graph) so these will be added to the project and the design in Microwave Office can be set up with a simple button click. This operation produces the project shown in Figure 5.

Fig 5: Hairpin filter schematic, layout, and electrical response using linear analysis

The Microwave Office filter project was created with all critical parameters set to be tuned and optimized, and iFilter offers an option to automatically run the analysis after creating the project, so the screen shown in Figure 5 is exactly what is seen after sending the filter to Microwave Office.

It can be seen that the filter's passband is very close to the goal, the stopband requirements are almost met, but with little margin on the low side, and the return loss still needs work. So the next step in refining the filter calls for linear optimization followed by EM analysis, and, if desired, further EM optimization. In this distributed hairpin filter example, the linear optimizer was run for 100 iterations with a Simplex optimizer and the resulting filter was analyzed with AWR's AXIEM 3D planar EM tool to produce the response/filter performance shown in Figure 6.

Figure 6: EM analysis of optimized filter

The final step in the design process, which is enabled by AXIEM’s unique simulation speed, is to optimize the filter further via more EM analysis and within the AWR extraction flow. The extraction flow permits the design to be optimized at the schematic level, using the EM engine to perform the analysis for each iteration of the optimizer, and to automatically generate a new EM structure every time the geometry changes based on the optimizer's guidance. With today's powerful multi-core PC systems, and AXIEM’s ability to utilize those resources, EM optimization is a time-effective approach that produces the most accurate designs possible. For this example, the EM optimization approach used AXIEM’s advanced frequency sweep option and resulted in each iteration taking less than 20 seconds on a an engineering desktop PC.

Figure 7: Final filter following EM optimization

Figure 7 shows that nearly all of the goals originally set for this example have been achieved. Passband insertion loss is good. Stopband insertion loss is very good. Plus, there is some margin for manufacturing tolerances. Only return loss is out of spec over a small portion of the band. While there are many options for curing the return loss issue (for example, changing the tap location and/or shortening the first and last lines slightly), that is an exercise readers can pursue on their own. For those who do not have Microwave Office with AXIEM and iFilter, you are invited to download an evaluation copy at and give it a try.


As WiFi and WiMAX wireless technologies move into the higher 5.8 GHz range, filtering is becoming an increasingly important part of any radio network. Designers need to ensure that their designs are as close to required performance specifications as possible by synthesizing and then fine-tuning the design with the most accurate tools available. AWR’s iFilter high performance synthesis module runs seamlessly within Microwave Office, keeping designers' filter designs and their evolution a part of the entire, managed circuit design project. iFilter's intuitive user interface enables users to quickly and easily design filters, connect them directly to circuitry, and make optimization trade-offs that positively impact their designs. iFilter is fast, accurate, and very easy to use. For the example described in this article, the distributed element capability in iFilter was showcased, however, the lumped element capability works exactly the same way. For more information on iFilter visit AWR’s iFilter website at , and be sure to check out real-time, multimedia examples of iFilter in action on AWR.TV.