Go to www.awrcorp.com
Back to search page Click to download printable version of this guide.

13.4. iFilter Filter Wizard

The iFilter Filter Wizard is a filter synthesis program. This wizard displays in the NI AWR Design Environment suite if you have the proper license file (FIL-200, FIL-250, FIL-300, or FIL-350) to run the wizard.

13.4.1. Using the iFilter Wizard

The iFilter Wizard uses the same interface for all filter types, the main iFilter dialog box. From this dialog box you can access all options and settings.

13.4.1.1. Starting the iFilter Wizard

You can run the iFilter Wizard to create a new filter or to modify an existing filter.

To create a new filter, open the Wizards node in the Project Browser and double-click iFilter Filter Synthesis.

  • To run Standard iFilter, click Design.

  • To run Advanced iFilter with synthesis capabilities, click Synthesis.

  • To run iMatch for designing impedance matching networks, click Matching.

You can also display this Select dialog box from within the program by clicking the Select Design Mode button.

To modify an existing filter, right-click the filter under the iFilter Filter Synthesis node in the Project Browser and choose Edit.

The main iFilter dialog box displays with properties from any previous design.

13.4.1.2. Running the iFilter Wizard

While the wizard displays, you can change filter and approximation type, edit specification and technology parameters, and configure other options. After every change, the wizard recalculates the element values, redoes the realization (layout or part selection) and calculates and plots the response. You do not need to press a special button after modifications as all the views are kept current.

To enter a different parameter value or specification in the main iFilter dialog box you can use the keyboard, click the up/down arrows next to an option to increase/decrease values, or use the mouse wheel (click in the desired edit box and scroll the mouse to increase/decrease the value). The step size is automatic based on the type and value of the edit box. Press Ctrl while scrolling to increment/decrement with a smaller step size.

13.4.1.3. Closing the Wizard

To close the iFilter wizard:

  • Click Generate Design to create a schematic, graph(s) and other items in the NI AWR Design Environment suite. A filter design item displays under the iFilter Filter Synthesis node in the Project Browser.

  • Click OK to create a filter design item under the iFilter Filter Synthesis node in the Project Browser only. No schematic, graph(s) or other items are created.

  • Click Cancel to close the main iFilter dialog box without saving.

13.4.1.4. Design Properties

In various stages of the wizard, new designs are created, previous designs are recalled or existing designs are modified. To preserve continuity, the wizard continually transfers data between changes. If you start a new design, the latest filter design properties are loaded into the new design rather than prompting you to re-enter all specifications. Also, when you change a filter type, all of the common specifications such as passband corners and technology settings are copied to the new design.

13.4.2. Filter Design Basics

This section provides a brief introduction to filter design with a focus on iFilter terminology. It is not intended to cover every aspect of filters. For more information about designing electrical filters please see a complete reference source.

13.4.2.1. Approximating Function

An ideal filter that passes a desired frequency range with no loss, and stops all undesired frequencies with no leakage is impossible to obtain. It is therefore necessary to approximate a filter response by using some filtering functions (transfer functions) that yield realizable element values, like Chebyshev or Maximally Flat filters. Some of these functions provide sharp stopband attenuation, while others provide flat group delay in the passband. Selecting an approximating function is always a trade-off, while most of the time Chebyshev meets most of the desired characteristics of filtering.

Filters represent varying input impedances with respect to frequency. Depending on the phase and magnitude of the impedance, they either pass or reject frequencies. A series capacitor, for example, has infinite impedance at f=0 Hz which causes all the signals to reflect at zero frequency. A series inductor, on the other hand, has infinite impedance at f=infinity which does not pass any signal.

Transmission Zero (TZ)

A transmission zero is defined as the frequency where no signal transmission occurs. For cascaded element filters (ladder type), TZ's can be created by infinite impedance series elements or zero impedance shunt elements. For example, a series inductor or a shunt capacitor can create a TZ at f = infinity, letting these elements be used for lowpass filters.

Finite Transmission Zero (FTZ)

A finite transmission zero is defined as TZ where frequency is a finite number as in 0 <f< infinity. FTZ's are created by infinite impedance series elements or zero impedance shunt elements at a finite frequency. For lumped element lowpass filters, a series-LC-resonator (SLC) connected as shunt to the circuit creates an FTZ. Likewise, a parallel-LC-resonator (PLC) in the shunt arm also creates an FTZ. You can use both of these elements for lumped-element lowpass or bandpass filters. Not all of the filter types allow FTZ's, however, because the realization technique used may not be suitable.

Monotonic Filters

A filter with no FTZ is sometimes called a "monotonic" filter. For a given filter order, monotonic filters provide less selectivity than filters with FTZ(s). However, their ultimate stopband (regions far away from the passband) contain less spurii.

In iFilter, available approximations for monotonic filters are:

  • Chebyshev

  • Maximally Flat

  • Bessel

  • Linear Phase

  • Gaussian

  • Transitional Gaussian

  • Legendre

Chebyshev filters provide maximum stopband attenuation for a given filter degree. As a consequence, there is a ripple in the passband of the filter. As the ripple gets smaller, the relative stopband attenuation reduces (the filter becomes less selective). Maximally Flat filters do not have a passband ripple but have a good selectivity. The other approximations are mainly used for pulse shaping rather than providing selectivity. Bessel filters, for example, provide an excellent flat delay within the passband. All of these approximations are based on the performance of lowpass prototypes. Once these basic filters have been transformed into the other passbands, such as bandpass, they tend to lose their attractive properties.

Unless the selected filter type does not pose any other limitations, the generic filter degrees are available up to 50 for Maximally Flat and Chebyshev filters. For very high order filters, the element values start losing accuracy due to computation limitations. For other approximations, the maximum filter degree is 10.

Filters with FTZ(s)

In iFilter, the available approximations with FTZ's are:

  • Elliptic

  • Generalized Chebyshev

Elliptic filters have ripples both in passband and stopband. Elliptic filters can provide very sharp selectivity at the expense of wide-spread element values. Because they are not monotonic, they also have a finite attenuation at high frequencies.

Generalized Chebyshev filters are a good balance between monotonic and elliptic filters. Their FTZ's are all concentrated at one single frequency and they tend to result in better element value spread than elliptic filters. There are two available types:

  • Type 1: all TZ's are one single frequency with 1 TZ at infinity

  • Type 3: all TZ's are one single frequency with 3 TZ's at infinity

13.4.2.2. Filter Synthesis

Filter synthesis is the process of constructing an approximating transfer function, then a driving point impedance function, and finally extracting elements from the impedance function. Transfer functions involve poles and zeros in normalized S-domain. For most designers, poles and zeros of the transfer function do not mean much. In reality, cascaded filters with no cross-coupling have explicit pole-zero locations that can be found in textbooks. iFilter, therefore, uses the TZ/FTZ terminology which correspond to the jw-axis, a measurable frequency quantity.

For monotonic and well-behaved FTZ functions, prototype tables and explicit formulas exist. iFilter uses this "LP-Prototype" approach to generate filters. A more generalized filter design includes transmission zero placement and element extraction. This method is called “filter synthesis” and it’s covered in Advanced iFilter product.

13.4.2.3. Design using LP-prototypes

There are two major lowpass prototypes: Ideal LP and Microwave. Ideal LP-prototypes are shown as series inductors and shunt capacitors. Microwave prototypes also include impedance/admittance inverters and they are mainly used for narrowband bandpass filters.

Prototype element values are also called "G-values". G-values are calculated through explicit formulation, or obtained using references such as Zverev’s Handbook of Filter Synthesis. Values are normalized to a source resistance of 1.

G-values are then converted into selected passband of the filter. This process is called "Frequency Transformation". An inductor prototype element is frequency-transformed into:

  • an inductor for lowpass filters

  • a capacitor for highpass filters

  • a series LC-resonator for bandpass filters

  • a parallel LC-resonator for bandstop filters

Likewise, a capacitor prototype element is transformed into:

  • a capacitor for lowpass filters

  • an inductor for highpass filters

  • a parallel LC-resonator for bandpass filters

  • a series LC-resonator for bandstop filters

Inverters are replaced by inductive or capacitive Pi- or Tee-sections for bandpass filters.

The last step of the filter design is Impedance Renormalization. Inductive parts are multiplied, and capacitive parts are divided by the actual source resistance, which is usually 50 ohms.

Although this design method is given for lumped element filters, distributed element filters can be derived in a similar way as described in the following section.

13.4.2.4. Distributed Element Filters

Distributed elements filters are obtained by using cascaded transmission lines, cascaded coupled lines, or multiple coupled lines. When terminated by a load impedance ZL, a transmission line with a characteristic impedance of Zo has the following input impedance:

where θ is the electrical length of the line. For transmission lines built as printed circuits, θ can be expressed as

Stubs

If the termination of the line is a short circuit (ZL=0), then the input impedance is described by

Thus a shorted line shares the same input impedance characteristics as an inductor.

Similarly; an open ended transmission line (ZL=infinity) has an input impedance characteristic as a capacitor,

These two special-terminated transmission lines are also called "stubs".

Periodicity

As found in most textbooks, s=jtanθ is called Richard’s variable. Tanθ is a periodic function and it repeats itself every θ. As a result, distributed element filters obtained by transmission lines periodically repeat their frequency response.

In the following figure, a distributed highpass filter response is given. The filter is designed with 50deg electrical length at 1 GHz. The elements are 90deg (θ/2) at 90/50*1GHz = 1.8GHz. The filter repeats itself at every 180 degrees (θ), (every 3.6GHz). As shown in the figure, the response from 0 to 3.6GHz is identical to the response from 3.6 to 7.2GHz.

In general, distributed lowpass filters and distributed bandstop filters have the same attenuation characteristics. Likewise, distributed highpass and distributed bandpass filters share the same response.

Filter Design

You can obtain distributed element filters using different methods.

The first method is to design lumped element filters using LP-prototypes and replace inductors and capacitors with stub equivalents. The drawback to this method is the size and practicality. A highpass filter calls for a series capacitor, which is equivalent to an open-circuited transmission line in the series arm. This is not possible in microstrip medium, so the technique is limited to lowpass and bandstop filters only.

The second method is to replace lumped elements in the first method with high and low impedance transmission lines. An inductor can be approximated by a high impedance transmission line in a limited-frequency band. Likewise, a capacitor can be approximated by a low impedance transmission line. This method is used in stepped impedance lowpass filters, but the approximation is only as good as the extreme impedances that can be obtained.

The last method is to use coupled transmission lines, either by direct coupled synthesis or by using the coupling coefficient method. There are various types of microwave resonators, such as end coupled, edge coupled, and interdigital. In this method, resonators are either synthesized or selected and coupled using gaps between them. The amount of coupling is dictated by synthesis formulas. Calculating gaps for given coupling is cumbersome, and it involves explicit formulations as well as lengthy EM simulations. However, the end result is rewarding, as this method yields very compact circuits with excellent selectivity.

13.4.3. General Flow of Filter Design

This section describes how to use the iFilter Wizard in the usual filter design process flow.

13.4.3.1. Main iFilter Dialog Box

The main iFilter dialog box is the control center of the iFilter Wizard.

This dialog box is resizable and registers its last state during runtime, so it opens with the size and location at which it previously closed.

The top left-hand side of the dialog box allows you to select the filter type, approximation, filter order and other electrical specifications. The bottom left-hand side of the dialog box allows you to change the physical parameters of the filter, save, and close the wizard. The top right-hand side shows the response of the filter and the associated chart control. The bottom right-hand side provides a view of the equivalent circuit schematic and physical layout. All items in the dialog box, including the layout and the plot, are current with the specifications. You do not need to click a button to re-design or re-analyze.

13.4.3.2. Select Filter Type Dialog Box

To exit the current design and create a new filter in the main iFilter dialog box, click the Filter Type button. The button is at the top left, labeled with the currently selected filter type.

Select Passband, Realization and Main Filter Type in order. Most of the filters have optional types that present subtle variations.

For some filter types, you can design the filter with a series or shunt element at the source side. In this case, a Design Dual Circuit (input: series/shunt) check box is available. If you select this option for a lowpass filter, for example, the design starts with a shunt capacitor rather than a series inductor. A shunt capacitor, however, is only available when source resistance is greater than or equal to load resistance, (for example, RS >= RL). For series inductors, RS <= RL is needed. If the source and load resistances are specified otherwise, the only possible option is selected regardless of the status of this check box.

Click OK to change or confirm the filter type. Properties of the existing filter are copied to the new filter.

NOTE: Some technologies may not be available due to license limitations. Contact NI AWR for license arrangements.

13.4.3.3. Approximation Function Dialog Box

To select the filter approximation function (transfer function), click the second button from the top left in the main iFilter dialog box. The button is labeled with the currently selected approximation function.

Almost all filters can have monotonic approximations, but only some filter types can contain finite transmission zeros. Only available approximations for the selected filter type are listed.

For monotonic filters, there are limits beyond which numeric errors build up or the response cannot be made any more selective. The available filter degree range is based on the following, unless the selected filter topology dictates otherwise.

  • Maximally flat, Chebyshev: 1 to 50

  • Elliptic types: 3 to 15

  • Generalized Chebyshev: 5 to 15 (odd-number). Even orders are not possible with passive elements.

  • For all other types, such as Bessel and Gaussian, the order is limited to a range of 1 to 10. Although these approximations are rarely used, they are included in iFilter for completeness.

In most cases, the available filter order is overridden by the selected filter topology. For distributed filters, first order filter is almost redundant so it is simply skipped. Likewise, a 50-resonator edge coupled filter is not feasible, so the maximum order is set to 15 for that type.

13.4.3.4. Change Passband Ripple Dialog Box

The Chebyshev approximation is used in the majority of filter designs because the Chebyshev approximation supplies reasonable element values and provides significant selectivity among the monotonic filter group. Elliptic and Generalized Chebyshev types provide more selectivity, however not all of the topologies are suitable for finite TZ’s and the response is more sensitive to element variations.

For Chebyshev type, you can specify the passband ripple. Passband ripple is a trade-off between matching and selectivity. If passband ripple is decreased, the passband return loss increases (not desired) and stopband selectivity increases (desired).

You can type the passband ripple in the text box, or click the Ripple[dB] button to display the following dialog box.

Passband ripple [dB], Return loss [dB], and VSWR are all related parameters. You can enter any of them and the wizard calculates and uses the corresponding passband ripple in dB. In this dialog box, the values for Min Zload and Max ZLoad are informational only; they display the impedance termination in order to create the VSWR shown when used in a 50-ohm system.

13.4.3.5. Modifying Specifications

When you select the filter type and approximation type, the relevant parameter editing boxes display in the main iFilter dialog box. You can click an edit box, type, and press Enter to validate the entry, or you can click another parameter instead of pressing Enter.

After you change a parameter value the filter is automatically redesigned. The filter response is plotted and the schematic and layout are updated.

Common specification types are:

  • Degree - This is the prototype degree or number of resonators for a filter.

  • Fp - Passband corner for Lowpass/Highpass filters (S21=Ripple dB for Chebyshev, Elliptic and Generalized Chebyshev, and S21=3.011dB for all other approximations).

  • Fo - Passband center for Bandpass filters, Stopband center for Bandstop filters.

  • BW - Passband width for Bandpass filters, Stopband width for Bandstop filters. This is measured from the ripple or 3-dB corner as previously explained.

  • Stopband IL - Peaks in the stopband for elliptic type approximation. Specifying a high value provides very high attenuation, but the selectivity is not very sharp. A low value may provide sharper attenuation near the passband corner, however it may result in unrealistic element values and the ultimate stopband peaks are very high.

  • Lshunt - Shunt inductor value for lumped, capacitively coupled resonator bandpass filter.

  • Low Zo, High Zo - Lowest and highest allowed impedances for distributed type lowpass filters.

  • Reson Zo, Line Zo - Internal impedance levels for microwave filters.

  • Lshunt - Shunt inductor value for lumped, capacitively coupled resonator bandpass filter.

  • RSource - Source termination (left-hand side of schematic)

  • RLoad - Load termination (right-hand side of schematic)

  • QL, QC, TLatt - Parasitic and loss factors for elements for simple analysis (see “Distributed Model Options Dialog Box” for more information.)

13.4.3.6. Analyzing a Design

Every time you change a filter specification it is concurrently re-designed and analyzed, and its response is plotted. The iFilter Wizard offers three types of analysis:

  • Ideal

  • Lossy

  • Real

You select Analysis mode by clicking the following buttons, located to the left of the plot.

Analysis Mode Lumped Element Filters Distributed Element Filters
IDEAL Elements are analyzed as IND, CAP Elements are analyzed as lossless TLIN and shorted/open circuited stubs.
LOSSY If SRF is disabled, elements are analyzed as INDQ, CAPQ models. If SRF is enabled, elements are analyzed as INDQP, CAPQP models. Not available
REAL Elements are analyzed using the models selected on the Lumped Model Options dialog box, Realization tab (click the Design Options button). Models include real vendor data. Elements are analyzed as lossy TLN.

Analysis mode also determines how the design is generated when you click the Generate Design button.

Analysis Mode Lumped Element Filters Distributed Element Filters
IDEAL Elements are mapped to IND, CAP Elements are mapped to lossless TLIN, and shorted/open circuited stubs.
LOSSY If SRF is disabled, elements are mapped to INDQ, CAPQ models. If SRF is enabled, elements are mapped to INDQP, CAPQP models. Not available
REAL Elements are mapped using the models selected on the Lumped Model Options dialog box, Realization tab (click the Design Options button). If a model does not include SRF, INDQ/CAPQ mapping is used. If model includes SRF, INDQP/CAPQP mapping is used. Real vendor data is also mapped according to this criteria. For multi-coupled line circuits like Interdigital, Hairpin, and MxCLIN/SxCLIN, mapping is used. Otherwise, MLIN/SLIN/MCLIN/SCLIN type cascaded transmission lines and coupled lines are used in mapping.

NOTE: NI AWR does not accept liability for the accuracy of third-party party models, as such data is only available from vendors. However, NI AWR is dedicated to providing the best design software. NI AWR communicates often with component vendors and progressively updates NI AWR simulation models.

13.4.3.7. Plotting Response and Chart Control

On the right-hand side, the filter response is plotted. Analysis is specifically tailored to filter design; therefore only popular measurements are available.

The five buttons below the Plot Settings button set the chart, auto-scale the axes, and add the associated measurements to the plot. The axes are always auto-scaled for simplicity. The iFilter Wizard analyzes circuits in a split second, therefore changing frequency span or type of measurements is not a deterrent to the designer. After analysis, the left Y-axis is scaled to a reasonable range and the right Y-axis is scaled up accordingly. The X-axis range is usually rounded to reasonable values, however, the internal /div scale may not always fit to a reasonable grid count.

The small buttons below the plot setting change the frequency span (X-axis) of the chart. Depending on the passband type (for example, lowpass/bandpass) and bandwidth and corners, span is calculated automatically for narrow/wide/ultrawide buttons. The Passband Analysis Span button scales the X-axis to filter passband only, to see the loss profile of the passband. Clicking the Auto Span button automatically changes the span when Fp, Fo or BW is changed.

The Add Marker button adds markers to the chart. You can also add markers by clicking the Edit Chart Settings button at the top of the main iFilter dialog box and then clicking the Markers button in the Chart Settings dialog box.

13.4.3.8. Chart Settings Dialog Box

To access the Chart Settings dialog box, click on the top button to the left of the filter response.

The first row of buttons select a preset response combination. In filter design, the following response definitions are used more often than standard S-parameters:

  • Insertion Loss (IL) = -dB|S21|. IL > 0 for passive filters.

  • Return Loss (RL) = -dB|S11|. RL > 0 for passive filters.

  • Voltage standing wave ratio (VSWR) = (1+|Rho|) / (1-|Rho|) where Rho is the reflection coefficient.

  • Insertion Phase (PH) = Ang(S21)

  • Phase Variation (PhVar) = variation of phase around a hypothetical linear phase.

The following preset chart types are available for plotting:

  • Insertion Loss and Return Loss

  • Group Delay and Insertion Phase

  • Insertion Loss and Input VSWR

  • Insertion Loss and Group Delay

  • Insertion Loss and Phase Variation

In the Analysis Range section, you enter the minimum and maximum frequency range of the analysis. If the Adjust range automatically when frequency specs are changed check box is selected, iFilter automatically sets the analysis range when you change Fo, BW or Fp specs. If you do not want the analysis range to change, clear this check box.

The IL Axis section provides controls for the Y-axis scaling for plotting the Insertion Loss. In previous iFilter versions, the IL axis scaling was automatic depending on the extent of data within the frequency range. This behavior is still available by selecting the Adjust Automatically check box. If the check box is cleared, you can enter the Y-axis range and this axis scaling remains fixed while the specifications and frequencies change.

In the Markers and Opt Goals section, you can add markers and optimization goals to the filter to visualize the filter response. To add a marker, select Markers and then click the Add button. To edit or delete an existing marker, select the marker and click the Edit or Delete button. To delete all markers, click the Clear button.

To add, edit, delete, or clear optimization goals, select the Opt Goals button and perform the same action.

To speed up the design process (recommended), you can click the Auto button to provide most of the common values for the selected filter type. For example, markers are added to the passband center and corners of a bandpass filter. For the same bandpass filter, optimization goals are added for minimum insertion loss in the passband, and 50dB attenuation is added in the upper and lower stopband.

13.4.3.9. Add/Edit Marker Dialog Box

To add or edit a marker, click the corresponding button in the Chart Settings dialog box.

Select a Data Type and enter a Marker Frequency for the desired marker. You can add markers for any data (measurement) type, but they only display when corresponding data is plotted.

NOTE: In the main iFilter dialog box, you can move markers without opening the Add/Edit Marker dialog box. In the chart area, scroll the mouse-wheel up and down to change the marker frequency. If there is more than one marker, right-click until the desired marker is marked with an "X".

13.4.3.10. Add/Edit Opt Goal Dialog Box

In the Add/Edit Opt Goal dialog box, Fmin and Fmax are the goal range and the Level is the criteria. Greater than and Less than determine which side of the criteria is desired. If you create an Opt Goal for Insertion Loss, and Insertion Loss is a positive number (for example, IL = -S21dB), in the passband, IL should be smaller than a maximum loss level so you should select Less than. For stopband, IL should be greater than a required attenuation so you should selectGreater than.

You also use Optimization Goals in setting up Optimization blocks when you generate the design in the NI AWR Design Environment suite.

13.4.3.11. Viewing the Schematic and Layout

On the bottom right-hand side of the main iFilter dialog box, schematic and layout information display depending on the selected button.

The top two buttons (View Circuit Schematic and View Layout) display the schematic or layout. The schematic is for circuit elements, either in lumped or distributed transmission line form. For lumped element filters, the schematic is the actual filter and there is no associated layout. For distributed element filters, the schematic represents an equivalent circuit from which you can calculate a response. A physical layout displays how the filter will appear, but most of the time you should perform an EM analysis on physical layouts. To increase speed, iFilter only calculates schematic-based responses. Some loss and parasitic information can be included in the analysis. For lumped element filters, the response is exact. For distributed element filters, the schematic-based response is exact for transmission line types and reasonably accurate for coupled line types.

If a warning displays, such as when a parameter value is close to a limit, the corresponding layout element displays in orange. If a limit violation occurs, the element and the View Layout and layout info (View Physical Dimensions) button display in red. Schematics rarely have warnings. In normal conditions, all buttons display in green.

The four small buttons next to the schematic area are for copying information into the Clipboard, toggling the schematic/layout colors between a light or dark background, turning on/off the text display, and toggling the Properties dialog box display. The Properties window shows the details for the selected element in the schematic/layout.

13.4.3.12. Generate Design Dialog Box

To use the extensive analysis capabilities of the NI AWR Design Environment software such as statistics, yield, and optimization, you should generate a design in terms of Microwave Office components: schematic, layout, and graphs. This is exporting an iFilter Wizard item to the Project Browser.

To generate a design in the NI AWR Design Environment suite, click the Generate Design button near the bottom left of the main iFilter dialog box to display the following dialog box.

General Section

Under General, type the Base Name of generated items such as schematics and graphs. You can also use an existing name, although a warning displays to tell you that the exported item already exists. To overwrite the existing item and turn off the warning, select the Overwrite existing items check box.

Schematic Section

In the Schematic section, you set the exported schematic options. You can use variables for parameters. When a parameter is generated as a variable, it is defined as an equation on top of the schematic, and the parameter is referenced to that equation. This is particularly useful when there are common parameters of symmetric elements in the circuit. To generate equations in the schematic and assign parameters to them, select the Use variables for element parameters check box.

You can hide some schematic element properties such as Names, Units, and Labels to simplify the view. If you select the Minor params check box, non-essential parameters in the schematic are hidden. You can also hide the schematic viewing grid by selecting the Snapping gridcheck box.

NOTE: The iFilter Wizard generates a schematic based on the current analysis settings. For more information on analysis settings, see “Analyzing a Design”.

Analysis Section

A generated schematic typically requires an analysis. To analyze the schematic after generating it, select the Analyze design after generation check box.

The NI AWR Design Environment suite can analyze all design items with the same frequency range settings (project defaults), or analyze them individually by setting them at the schematic level. You can use the default settings by clearing the Use range below (not project defaults) check box, or set your own range by selecting this check box and typing the values. When you open this dialog box, the frequency range from the current analysis range displays. To copy this range, click the Set to current range button.

Graphs

the Microwave Office program has an extensive list of measurements and graph types. For designing filters, you only need a limited subset. iFilter allows you to plot various responses and generate them in the Microwave Office program. By selecting chart types (graph types), you can preset graphs and associated measurements.

You can leave the NI AWR Design Environment suite to scale the Y-axes depending on the range of the associated measurements, or by selecting the Use fixed axis settings instead of Auto check box, you can allow iFilter to set the Y-axes.

Tuning and Optimization

iFilter can export tuning variables and optimization goals derived from the current design.

If you select the Mark Tuning Variables check box, iFilter determines the major circuit element parameters and sets them as tuning variables. If you tune the exported circuit in the Microwave Office program (press F9), the tuning variables are set during export.

If you select the Set Optimization Goals check box, iFilter exports the Opt Goals as optimization goals in the Microwave Office program. To add or edit Opt Goals, click the Edit Chart Settings button in the main iFilter dialog box to display the Chart Settings dialog box, and then click the Opt Goals button. iFilter also sets the optimization parameters in the Microwave Office program which are the same as the tuning variables. iFilter defines a rough constraint (20% above and below) for bounding the values.

Microstrip Models

iFilter exports Standard model and X-model microstrip elements. Select the Use X-models check box if you prefer X-models. You can select the following models using this option:

Standard: MTEE$, MSTEPO$, MBEND2$, MLEF, MOPEN, MCROSS$

X-model: MTEEX$, MSTEPX, MBEND90X$, MLEFX, MOPENX, MCROSSX$

After selecting all relevant options, click OK to generate the selected items in the Microwave Office program and close the main iFilter dialog box.

13.4.4. Lumped Model Options Dialog Box

Lumped element filters are initially designed with ideal inductors and capacitors. To build a lumped element filter, you should substitute real life components for ideal elements. In iFilter, you select real life components in the Lumped Model Options dialog box. To open the Lumped Model Options dialog box, click the Design Options button while designing a lumped element filter.

13.4.4.1. Lumped Model Options Realization Tab

You can edit inductor, capacitor, and Real L/C parts options on the Lumped Model Options dialog box Realization tab. To view this tab, click the Design Options button in the main iFilter dialog box and then click the Realization tab.

When iFilter designs a lumped element filter, it holds two sets of circuits:

  • Ideal filter with L,C elements

  • Real filter with selected models and vendors

You can analyze the ideal filter either lossless or with lossy elements and/or parasitics. In the main iFilter dialog box, three buttons control the analysis type.

The Analyze Ideal button is for analyzing the ideal lumped element filter with lossless elements. The output is a textbook type response. The Analyze Lossy button is for analyzing the ideal filter with lossy elements and parasitics. See “Lumped Model Options Parasitics Tab” for details. The Analyze Real button is for analyzing the real filter with selected models.

For inductor models, various preset and fixed ranges are available. An inductor of the selected range can be modeled as:

  • Use AIR COIL (an air-wound coil) - where iFilter calculates the required number of turns based on the maximum coil diameter and wire gauge. iFilter tries to fit the calculated inductors to maximum diameter (bigger coils give bigger Q) and within 20 turns. For calculating coils manually, see “Design Utilities Dialog Box”.

  • INDQ/INDQP - where iFilter treats the element as in the ideal filter case, as lossy and/or parasitics. The INDQ element is a simple lossy inductor. The INDQP element is modeled as an INDQ element with a capacitive effect, and so self-resonant. To set loss and/or parasitics for elements, see “Lumped Model Options Parasitics Tab”.

  • Vendor Part - where iFilter searches its internal vendor database, and chooses the part with the best Q.

An ideal capacitor can be modeled as:

  • CAP/CAPQP - where iFilter treats the element as in the ideal filter case, as lossy and/or parasitics. The CAPQ element is a simple lossy capacitor. The CAPQP element is modeled as a CAPQP element with an inductive effect, so self-resonant. To set loss and/or parasitics for elements, see “Lumped Model Options Parasitics Tab”.

  • Vendor Part - where iFilter searches its internal vendor database, and chooses the part with the best Q.

Every physical lumped element has an associated self-resonant frequency (SRF). An inductor’s reactance increases up to a certain frequency based on L = 2xf and suddenly starts decreasing due to stray capacitances between turns. At f=SRF, the inductor becomes purely resistive, and beyond SRF, it has negative reactance like a capacitor. A capacitor, however, exhibits a positive reactance beyond SRF like an inductor.

NI AWR does not recommend using elements beyond their SRF, however, you should know the following:

  • Capacitors exhibit the highest Q at the lowest frequencies. As frequency gets higher, Q decreases, causing more insertion loss in the passband of filters. In RF and microwave filters, multilayer capacitors are mostly used. Normally, capacitor Q’s are much higher than inductors, so even when they decrease it is not a concern unless the passband is very narrow. SRF, however, can be a major problem. SRF is a result of inductive properties of the multilayer capacitors that are caused by interlinking wires. When SRF is considered, the effective capacitance can drop significantly, causing a bandpass filter to have a narrower passband at a lower center frequency. You should choose filtering capacitors to operate as far away from their SRF as possible. NI AWR recommends using the smallest possible size capacitors, as they tend to have higher SRF. The trade-off is in the Q, as small-sized capacitors may not have high Q.

  • Inductors exhibit an interesting Q-value vs. frequency. They tend to have a Q increasing with frequency up to a point and reduce quite significantly when SRF approaches. Observations show that inductors have their highest Q at about f=SRF/1.5 to 1.7.

In iFilter, you set the SRF criteria on the Lumped Model Options dialog box Realization tab. In Min SRF/Fo ratio to look for within vendor parts, you can enter a value for iFilter to ignore the undesired practical elements. The value is specified as a ratio of Min SRF/Fo. For example, for a filter designed at 200MHz, a value of 3 directs iFilter to only pick elements from its vendor database that have SRF greater than 3*200 = 600 MHz.

This dialog box also includes the following three options:

  • Split shunt capacitors into 2 if not realizable - splits capacitors into a maximum of two pieces if the value cannot be obtained with a single vendor part. For example, 5.9pF is not a standard value. If this box option is selected, iFilter searches all of the combinations that can make 5.9pF and chooses the highest Q combination. A combination of 4.7pF and 1.2pF give 5.9pF, for example.

  • Keep resonance freqs constant for LCs - keeps LC constants the same for each resonator. For LC resonators, iFilter first searches for standard capacitors, as inductors are easier to obtain. If a 2.6pF is needed, and it is only possible to obtain 2.7pF, then iFilter reduces L of the resonator, so that LC multiplication gives the same resonance frequency.

  • Increase inductance lookup margin if value is not found in catalogs – extends lookup margin when searching elements. Catalog inductors values do not cover a wide range as capacitors, so finding a catalog inductor may sometimes be difficult with the default margin (%3). If this option is selected, the software slowly increases this margin until a suitable inductor is found in the inductor catalogs.

For capacitors and inductors modeled as vendor parts, you should click the Inductor Vendors or Capacitor Vendors buttons to set up automatic selection of manufacturer parts as described in the following section.

13.4.4.2. Vendors and Parts Dialog Box

The Inductor Vendors and Parts and Capacitor Vendors and Parts dialog boxes allow you to select vendors and parts from the company inventory. To access these dialog boxes click the Inductor Vendors or Capacitor Vendors buttons on the Realization tab of the Lumped Model Options dialog box.

This dialog box contains many options, yet it is simple to use. To select available vendors and parts:

  1. Between the Available Types and Selected Types list boxes, click the double left arrow button to move all selected types to available types.

  2. In the Vendors section, click the Select None button to de-select all vendor parts.

  3. Select all vendors that are available.

  4. In the Part Sizes section, click the Select None button to de-select all part sizes.

  5. Select all part sizes that are available. Note that the sizes are listed in EIA type which is mostly used in the USA.

  6. Click the Search for Available Types button.

  7. In the Available Types list box, select the parts to use for designs, and then click the single right arrow button to move the type to the Selected Types list box. You can select all types by clicking the double right arrow button. Some models do not provide an acceptable filter response. You should avoid these parts, as they may be selected for use simply based on their high Q.

    NOTE: The iFilter Wizard uses models “as is” (using the manufacturer datasheets). Where individual SPICE models are not provided, iFilter uses simple models based on SRF and Q data. It is not NI AWR’s intention or responsibility to match vendor part datasheets with their actual performance in a circuit.

  8. The bottom right-hand side of this dialog box is for testing values only. When you enter an element value and frequency and click the Find Parts button, iFilter displays all of the found values and lists their associated Q’s and SRF properties.

13.4.4.3. Vendor Part Libraries

Lumped element component manufacturers publish catalog data to support their products. There is no standard format for these data sets. For an inductor, for example, Coilcraft publishes their own models (CCIND in Microwave Office), TDK publishes simple equivalent models; and AVX provides part selection software. Some manufacturers also provide S-parameter data sets. Not all the available data is self consistent and often the information provided illustrates the different modeling approaches employed by manufacturers. Some parts catalogs are more generous in providing measured Q values, the equivalent electrical model (in SPICE format) and/or S-parameter data. To a novice filter designer, this can be more confusing than having no data at all, as often these three pieces of information conflict!

Lack of adequate information can be overcome by traditional design techniques. After the part family type is selected, or selected from the company's warehouse stocks or available lab kits, a designer can go through charts by interpolation and produce his own data. He may also use past experience with the vendor from previously designed filters and other designs.

Searching for vendor parts amongst various data sheets in a reliable fashion is a challenging task for a filter designer to do manually. The goal of a successful search algorithm built into a filter design tool is to scan a single library of pre-processed parts data in a convenient and controlled manner.

iFilter includes built-in vendor part libraries. The libraries are stored and programmed to allow:

  • Nearest model: The available data is interpreted for each vendor and product type separately and they are mapped to the most appropriate component model. For inductor case, it may be a CCIND, an INDQ, and INDQ with self-resonance effects.

  • Fast part-search: iFilter is capable of searching its library of over 7000 parts instantly. Every time you change a filter specification, iFilter automatically searches through more than 25 vendors and chooses the part with the highest unloaded Q at the application frequency. On fast computers, the hourglass cursor that displays during filter design may not be visible.

  • Automated testing: Because of its optimized library, it takes less than one minute to test the full library. This test assures the integrity and accuracy of the NI AWR software.

13.4.4.4. Lumped Model Options Parasitics Tab

Practical inductors and capacitors have two major parasitics: loss and self-resonance. Loss is simulated by using series or parallel resistors that are calculated from the element's unloaded Q data. Self-resonance, however, is a limitation of the element and it is simulated by adding a resonating counterpart at the specified frequency.

You can specify inductor and capacitor parasitics on the Lumped Model Options dialog box Parasitics tab. To view this tab, click the Design Options button in the main iFilter dialog box and then click the Parasitics tab.

Losses

In iFilter, Q can be either simple or advanced depending on the option setting. If the Use fixed QL, QC values (simple) check box is selected, the top QL and QC values in the dialog box are taken as constant throughout the analysis range. This is a simple but effective approximation of loss.

If the Use fixed QL, QC values (simple) check box is cleared, advanced Q settings are assumed.

In the advanced Q settings, Q is a function of frequency and element value. The inductor case is given previously, and capacitor case is intuitive. In the equation:

  • base value - is as specified in the Inductor and Capacitor Losses section QL or QC option.

  • Lref, Cref - are reference values as 1nH and 1pF.

  • Fref - is reference frequency: 100 MHz

  • F - is analysis frequency

  • L, C - are values of the element in the circuit

For example, assume QL=100 for 1nH and QL=150 for 10nH are given at 100 MHz within available inductor stock. By setting QL=100, expV=0.1761 and expF=0, you can simulate Q’s for any inductor of the given stock. Because,

  • QL(f,L) = 100(L=1nH / Lref=1nH)^0.1761 = 100

  • QL(f,L) = 100(L=10nH/ Lref=1nH)^0.1761 = 150

If a frequency dependency exists, you can specify expFL to simulate the effect. ExpL, ExpC, ExpFL, ExpFC are all exponents. To specify

  • 1/x variation, set -1

  • x variation, set 1

  • x^2 variation, set 2

  • no variation if set as 0 (default)

iFilter also makes QL and QC base values available in the main iFilter dialog box. If advanced Q settings are checked, the main iFilter dialog box changes QL,base and QC,base. If the Use fixed QL, QC values (simple) check box is selected, they are used as constant QL,QC values for the simulation.

The For BPF: set Qs to shunt resonators only check box is used to predict losses of narrowband bandpass filters. If a narrowband microwave filter is known to be inductively or capacitively coupled, you can design a lumped narrowband filter of the same degree and passband, and by setting Qs to shunt elements only, you can predict the losses of the microwave filter.

Self-resonance Frequency (SRF)

Every lumped element exhibits a self-resonance frequency where its reactance drops to zero and it becomes purely resistive. For a real life inductor, this is equivalent to a parasitics capacitor connected in series with the inductor element. Likewise, a real life capacitor has an associated inductance (as a result of connecting plates together). Beyond the SRF, an inductor behaves like a capacitor, whereas a capacitor behaves like an inductor.

The SRF affects impedance as well as the unloaded Q of the element. For capacitors, QC decreases as frequency increases and as the SRF is approached. You should therefore use capacitors well below their SRF. For inductors, QL increases with frequency until about SRF/1.5 or 1.7, so it is good practice to select inductors (coils) with the SRF about 1.5-1.7 times Fo.

To simulate the effects of the SRF, enter values in the corresponding boxes for reference inductor (1nH) and capacitor (1pF) values. For a given size of elements (for example, 0402, 0805), the SRF tends to change inversely with the square root of the element. For example, if SRF is entered 10GHz for 1nH, it is taken as 5GHz for 4nH. You can use your component vendor's datasheets to find the SRF values for reference values.

When the Use SRF in analysis (INDQP/CAPQP) check box is selected, the analysis is performed by adding an internal SRF effect to each element. This is in addition to the losses, if selected. When this check box is selected, and the design is generated in the Microwave Office program, the elements are mapped to INDQP and CAPQP, which have SRF effects. If this check box is cleared, analysis and design generation is based on CAP/CAPQ and IND/INDQ models.

13.4.4.5. Lumped Model Options Limits Tab

You can specify inductance and capacitance warning limits on the Lumped Model Options dialog box Limits tab. To view this tab, click the Design Options button in the main iFilter dialog box and then click the Limits tab.

These limits are used to generate automatic warnings in the View Circuit Information window in the main iFilter dialog box. If the limits are reached, iFilter adds a small warning icon to the tree entry.

For lumped element filters, you can set values for:

  • Minimum inductance

  • Maximum inductance

  • Minimum capacitance

  • Maximum capacitance

  • Maximum L or C ratio

Maximum L or C ratio is specified such that either Lmax/Lmin or Cmax/Cmin is to stay below that value. If exceeded, a warning is issued.

13.4.5. Distributed Model Options Dialog Box

The Distributed Model Options dialog box allows you to set up physical model options for microwave filters in general. To open the Distributed Model Options dialog box, click the Design Options button while designing a microwave filter type.

13.4.5.1. Distributed Model Options Realization Tab

You can edit general, coupled line, and layout options on the Distributed Model Options dialog box Limits tab. To view this tab, click the Design Options button in the main iFilter dialog box and then click the Limits tab.

You can scroll the mouse wheel in an edit box to increase/decrease the specified value. For example, doing so for Er steps through popular dielectric values, and doing so for the height parameter of a microstrip steps through 5, 10, 15, 20 ... board thicknesses. For microwave filters in general, the following dialog box displays.

This dialog box includes the following options:

  • Add input and output lines to the layout - adds lines at the filter input and output for the terminations. The extra line widths are calculated for the termination impedances, and a small length is used. You should select this check box, as you can use the extra lines to align the filter to the rest of the layout.

  • Bend/fold long lines when appropriate - turns lines into traces (like MTRACE for MLIN) when the line is excessively long compared to the general layout of the filter. For stepped impedance resonator bandpass filter (SIR), this option is very useful.

  • Split shunt impedances if smaller than Zmin - tells iFilter to split up shunt stub impedances (open or short circuited) into two identical elements and add them to the layout with a CROSS element, rather than a TEE element with one stub. You should do so, as most of the lowpass/bandpass structures have very low impedances. You can specify in the text box the threshold below which the splitting occurs.

  • Alternate input ports to save diagonal space - alternates ports of edge coupled sections of SIR filters so that the layout stays along a horizontal axis. This practice saves diagonal layout space.

  • Auto rotate lines to save space when appropriate - When you specify a rotation angle, this option rotates the filter to the specified degree, so that the layout can be realized at an angle.

  • Draw a reference box for comparison - display a reference box around the layout. In the View Layout mode, when you change a filter parameter, the layout is recalculated and redrawn so you can view the resulting change.

13.4.5.2. Distributed Model Options Technology Tab

You can edit the technology (for example, microstrip or stripline) parameters on the Distributed Model Options dialog box Technology tab. To view this tab, click the Design Options button in the main iFilter dialog box and then click the Technology tab.

You can scroll the mouse wheel in an edit box to increase/decrease the specified value. For example, doing so for the Substrate Er parameter steps through an internal database of popular substrate dielectric constants. For any automatic selection of Er, the corresponding Loss Tangent for that substrate displays. Scrolling the mouse wheel for the Substrate Height(H) steps through industry standard board thicknesses. To access this tab click the Design Options button in the main iFilter dialog box and then click the Technology tab on the Distributed Model Options dialog box.

The Loss Tangent is not used in calculating the dimensions; however it is included to correctly analyze the losses in the Microwave Office program.

13.4.5.3. Distributed Model Options Parasitics Tab

You can set loss parameters for distributed elements on the Distributed Model Options dialog box Parasitics tab. To view this tab, click the Design Options button in the main iFilter dialog box and then click the Parasitics tab.

In this dialog box, you can specify the transmission line attenuation for distributed elements in Real analysis mode only. For RF and microwave filters, it is assumed that the attenuation of microstrip/stripline transmission lines is a linear function of frequency. For example, if a 10mm transmission line has 1dB attenuation at 1GHz, then it has 2dB attenuation at 2GHz. Because the filtering applications are diverse, a general loss factor is not suitable. The dB/cm approach is an old technique, but it works reasonably well. In this case, the attenuation is taking a linear function of frequency. So, 0.001dB at 1MHz increases to 0.010dB at 10MHz.

13.4.5.4. Distributed Model Options Limits Tab

You can edit physical element limits on the Distributed Model Options dialog box Limits tab. To view this tab, click the Design Options button in the main iFilter dialog box and then click the Limits tab.

For printed structures such as microstrip and stripline, the minimum width and minimum spacing (gap) are major manufacturing limits. When these limits are approached or exceeded, iFilter issues a warning by changing the color of the layout icon and elements in the layout, and by displaying a small warning icon next to the layout information entry. Some preset substrate type entries are listed. You can copy these values to Min Width and Min Spacing by selecting the desired entry and then clicking the Load Selected Limits button. Alternatively, you can double-click the desired entry or manually edit the option values.

13.4.6. Lowpass Filters

Lowpass filters are designed to pass signals in a frequency range below a specified frequency, generally called the "cut-off" frequency. For filters that have an equi-ripple passband characteristic (for example, Chebyshev, Elliptic and Generalized Chebyshev), the cut-off frequency corresponds to the passband ripple corner. For other approximation types, it is the 3.011 dB (3-dB) corner frequency. The frequency range below cut-off frequency is called the "passband". The frequency range beyond the cut-off frequency is called the "stopband".

The passband of lowpass filters starts from f=0 Hz (DC). Within the passband the input impedance of the filter is very close to the source impedance, which is usually 50 ohms. In the stopband, impedance of the filter is no longer 50 ohms, and so rejects all the signals. In the stopband, filters are said to attenuate signals, more commonly called "rejection" than attenuation.

There is a limit to the stopband frequency range. For lumped element filters, due to the cavity of the housing, TEM and waveguide modes are excited at higher frequencies, so artificial passbands are observed. For example, a lowpass filter designed to have a 100MHz cut-off may show passbands beyond 1GHz. In these cases, a low order cleanup filter is cascaded to the main filter to suppress the artificial passbands. For distributed element filters, the response repeats itself due to the periodicity of electrical lengths. Therefore, beyond a certain frequency, lowpass filters behave like bandpass filters.

13.4.6.1. Lumped Element Lowpass Filter

Lumped element lowpass filters contains series inductors and shunt capacitors. For filters with finite TZs, SLC resonators replace the shunt capacitors. An elliptic lowpass filter is shown in the following figure.

Properties of lumped element filters are defined in “Lumped Model Options Dialog Box ”.

Typical Specifications
  • Approximation: All available

  • Degree: See standard ranges

13.4.6.2. Lumped Lowpass/Highpass Diplexer

iFilter can design lowpass highpass diplexer using a “single terminated prototype” method. In this method, source impedance is first assumed ZERO ohms for both lowpass and highpass channels, and after combining the two channels in parallel, the source impedance is then set back to 50 ohms.

iFilter designs, analyzes and exports the diplexer channels as in the following figure. The two channels are normally connected at the source end and there is only one source termination. For display purposes, channels are displayed as if they have a source termination each.

When you select the diplexer type the first time, a diplexer setup dialog box displays.

The channel list on the left lists a lowpass and a highpass channel. The cutoff frequency can be set while editing the lowpass or the highpass channel. It is normally set to the same frequency for both channels, however iFilter provides the flexibility of setting them separately, so Fp can be slightly different to optimize the return loss.

This dialog box also allows you to edit bandpass multiplexer channels. The Quick Setup section provides controls for setting multiple channels. You can also specify the number of channels, the lower passband corner of the first channel, and the common bandwidth of all channels.

On the left of the main iFilter dialog box are channel access buttons. The first box shows the selected channel number. The next two buttons toggle between LP and HP channels. The Edit button displays the diplexer setup dialog box where you can edit the frequencies.

13.4.6.3. Stepped Impedance Lowpass Filter

Stepped impedance lowpass filters are very simple to construct. You can use almost any medium that creates a transmission line to make stepped lowpass filters. You can use a coaxial tube with varying inner rod diameters. iFilter primarily focuses on printed technology, therefore microstrip and stripline realizations are included.

A stepped lowpass filter is a series of low and high impedance lines. To make an analogy, a monotonic lumped element filter consists of series inductors and shunt capacitors. If series inductors are replaced by high impedance lines, and capacitors are replaced with low impedance lines, you can obtain a stepped lowpass filter. The line lengths are calculated from inductor and capacitor values and the response is approximated at the corner frequency.

There are a few drawbacks associated with using stepped impedance lines in filter structures:

  • Line lengths must be iteratively adjusted, as the stepped capacitances affect the approximate equivalent circuit parameters. Various optimization routines are available in the literature. A simple approach taken in iFilter is to design the filter by taking the shift in performance into account from the previous iteration. This is a trivial yet effective solution with little need to adjust return loss after design.

  • The input impedance of these filters never reaches a ZERO or INFINITE impedance and they never show full reflection. As a result, these filters do not possess very good stopband attenuation.

  • A drawback (or advantage if used properly), is the recurring passband. It may be shifted by adjusting the impedances to suppress undesired harmonics or spurious regions.

This filter type provides two options: Set Z, varying lengths and Same Length, varying Z’s.

The Set Z, varying lengths option uses lumped element prototypes as a basis, and replaces the prototype elements with transmission lines specified as low and high impedances. Line lengths are calculated for each element, to achieve the best approximation in the passband. For some element values the line lengths may not be realizable for the required impedances. If this occurs, the specified impedances are changed until the approximation yields a positive length for that element. You should specify Low Zo as low as possible to obtain a better stopband rejection.

The Same Length, varying Z’s option uses commensurate line synthesis. The transmission lines all have the same specified electrical length ElecLng (EL) at the passband corner, Fp. EL controls two important aspects of the filter: spurious performance and impedance ratio. As with every distributed filter, there are inevitable spurious passbands. For this type of filter, the spurious passband occurs at Fp * (180-EL)/EL. So, if EL=45 degrees, a spurious passband occurs starting at 3*Fp. When EL is reduced to 30 degrees, the spurious passband corner moves to 5*Fp. So a lower EL moves the undesired spurious passband away from the intended passband. EL also controls the high impedance/low impedance ratio of the filter. The higher the EL, the lower the impedance ratio making the topology more practical to build. Therefore, a trade-off between practicality and spurious response should be decided while adjusting EL.

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges

  • Low Zo 15 to 50 ohms (must be realizable) – for Set-Z option only. Specify as low as possible.

  • High Zo: 50 to 250 ohms (must be realizable) – for Set-Z option only.

  • ElecLng: 10 to 60 degrees – for Same-Length option only.

Tuning and Optimization
  • L_v(n) - are the element line lengths. Usually, the step capacitance between impedance transitions slightly affects the passband corner and the return loss. Some tuning may be required for the lengths.

13.4.6.4. Distributed Stubs Filter

This filter uses lumped element prototypes as a design basis and replaces prototype elements with transmission lines. The distributed stubs lowpass filter consists of transmission lines separated by open-circuited shunt stubs. Transmission line impedances are intended to be user-controlled; so the Line Zo parameter is available for editing. In most cases, however, only the input and output transmission lines can give valid line lengths. For internal lines, the impedances are increased to obtain a valid line length approximation. A higher Line Zo yields better return loss in the passband.

Open-circuited shunt stub impedances are independent of the Line Zo specification. Their impedances and lengths are calculated from the lowpass prototype elements.

Typical Specifications

Approximation: Monotonic

Degree: See standard monotonic ranges

ElecLng: 30 to 150 ohms (must be realizable)

Tuning and Optimization

L_v(n) - are the element line lengths. Usually, the step capacitance between impedance transitions slightly affects the passband corner and the return loss. Some tuning may be required for the lengths.

13.4.6.5. Optimum Distributed Lowpass Filter

Distributed element lowpass filters may contain short-circuited stubs in the series arms and open-circuited (open ended) stubs in the shunt arms. In planar structures like microstrip, short-circuited stubs in the series arm are not realizable. One solution is to convert these stubs into practical open ended shunt stubs using Kuroda transformations, so a 5th order lowpass filter consists of five shunt stubs separated by four transmission lines.

An optimum distributed lowpass filter is much the same in appearance, but with one important difference in the response. The transmission lines in the optimum filter are not obtained by Kuroda transformations, but rather synthesized into the transfer function during the design. Each transmission line therefore adds a transmission zero to the S21 response. Thus, the same 5 stubs + 4 line filter above will have 9th order response for the optimum case, with a much sharper stopband selectivity.

This filter type lets the electrical length (EL) at passband corner be specified. As with the stepped impedance lowpass filter type, the EL controls the spurious passband frequency and impedance ratio of the filter. A trade off might be necessary while setting EL, which is easy to observe in the main IFilter dialog box.

Typical Specifications

Approximation: Chebyshev only with fixed 26dB, 20dB or 16dB return loss.

Degree: 3 to 19

ElecLng: 18 to 85 degrees

Tuning and Optimization

L_v(n) - are the element line lengths. Slight tuning may be required for the lengths for accurate passband corner and return loss.

13.4.7. Highpass Filters

Highpass Filters are designed to pass signals in a frequency range above a specified frequency, generally called the "cut-off" frequency. For filters that have an equi-ripple passband characteristic (for example, Chebyshev, Elliptic and Generalized Chebyshev), the cut-off frequency corresponds to the passband ripple corner. For other approximation types it is the 3.011 dB (3-dB) corner frequency.

There is an upper frequency limit. For lumped element filters, the upper limit is dictated by the performance of the lumped elements, which are limited by loss and self-resonance frequencies. For distributed element filters, the response repeats itself due to the periodicity of electrical lengths. Therefore, beyond a certain frequency, highpass filters behave like bandpass filters and they attenuate beyond a certain frequency. Printed type distributed filters (microstrip and stripline) also have a dielectric loss that limits the maximum frequency for which the performance closes the specification.

13.4.7.1. Lumped Element Highpass Filter

Lumped element highpass filters contains series capacitors and shunt inductors. For filters with finite TZs, SLC resonators replace the shunt inductors. An elliptic highpass filter is shown in the following figure.

The realization of lumped element filters is defined in “Lumped Model Options Dialog Box ”.

Typical Specifications
  • Approximation: All available

  • Degree: See standard ranges

13.4.7.2. Shunt Stub Highpass Filter

A shunt stub highpass filter contains shunt short-circuited stubs and connecting transmission lines. There are three options available for shunt stub highpass filters.

The first option (1/4wave lines + 1/4wave stubs) is a combination of quarterwave length stubs separated by quarterwave length transmission lines. The second option (1/4wave lines + 1/4wave stubs (equal)) is a subtle variation of the first, as the stubs are made equal using exact circuit transformations. The response is the same, and the stub impedances are more realizable in most practical cases.

The last option (1/4wave lines + 1/2wave stubs) is a combination of halfwave length stubs separated by quarterwave length transmission lines. This option has much sharper response than the first two options as it generates a transmission zero below passband corner. However, it also generates a lowpass response around DC.

Shunt stub highpass filters behave exactly like bandpass filters. The upper passband corner of highpass filters does not extend to infinity, but rather a finite frequency. Electrical length (EL) controls the width of the highpass filter.

By specifying EL at Fp, a bandpass filter can be assumed to have a passband of Fp to Fo+2*(Fo-Fp), where the transmission lines are 90 degrees long at Fo. You can deduce the following:

Fo = Fp * 90/EL

For example, a shunt stub highpass filter designed at 4 GHz with 60 degree long lines has a passband of 4 to 8 GHz with Fo=6 GHz.

If the EL is specified high, the resulting passband is narrow and the shunt stubs become very low impedance, necessitating wide strips for planar realization. These filters are therefore more suitable with wide passbands (EL < 70 degrees).

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges (with limitation of 3 to 19 range)

  • ElecLng: 50 to 85 degrees

Tuning and Optimization
  • L_v(n) - are the element line lengths. Slight tuning may be required for the lengths for accurate passband corner and return loss.

13.4.7.3. Optimum Distributed Highpass Filter

An optimum distributed highpass filter is a subtle variation of a shunt stub highpass filter. The optimum type is obtained by exact synthesis of (n-1) transmission lines and a shunt stub, which is modified using Kuroda transformations. The stub lengths are also set to be equal during the transformation stage. The outcome is a filter with perfect return loss and identical shunt elements.

All elements of the filter have specified electrical length (EL) at the passband corner. Like the shunt stub highpass filter, the EL controls the width of the passband and the level of impedances. The optimum distributed highpass filter is suitable for moderate to wide passbands (for example, from EL=40 to 70-75 degrees).

Typical Specifications
  • Approximation: Chebyshev only with fixed 26dB, 20dB or 16dB return loss.

  • Degree: 3 to 14

  • ElecLng: 9 to 85 degrees (extreme specs result in impractical impedances)

Tuning and Optimization
  • L_v(n) - are the element line lengths. Slight tuning may be required for the lengths for accurate passband corner and return loss.

13.4.8. Bandpass Filters

Bandpass filters are the most common RF and microwave filters. They are designed to pass signals within a certain frequency range. The upper and lower limits of the passband are generally called the lower and upper cut-off frequencies. For filters that have equi-ripple passband characteristics (Chebyshev, Elliptic, and Generalized Chebyshev), cut-off frequency corresponds to the passband ripple corner. For other approximation types it is the 3.011 dB (3-dB) corner frequency.

The passband of bandpass filters is also called bandwidth of the filter and often denoted as BW. The middle frequency is termed as center frequency and often denoted as Fo.

An important property of bandpass filters is the insertion loss. The insertion loss is the minimum achievable loss of the filter in the passband, mostly occurring at Fo. Insertion loss plays a great role in receivers as it adds to the system noise figure, reducing the sensitivity. In transmitters, it causes dissipation in the filter and a reduction in power transmitted. For a 10W transmitter, 1-dB insertion loss corresponds to 20% reduction in power (2W). Not only is the delivered power 2W less, but the dissipation of such high power increases the temperature of the filter significantly.

Insertion loss depends on the bandwidth of the filter, unloaded Q of the resonators, and the degree of the filter. If a filter is narrowband, it becomes lossy, so high Q elements are needed to minimize the filter loss. A simple yet accurate expression for bandpass filter insertion loss is given in “Design Utilities Midband IL (Midband Insertion Loss) Tab”.

Most bandpass filters are designed for relatively narrow bandwidths. For lumped element type topologies, filters derived from standard LP prototypes suffer a wide element value range. A 7th degree, 10% standard filter designed at 500MHz yields an inductor range of 0.97nH to 278nH (1:284 impedance ratio) which requires 1-turn and 20-turn coils at the same time. The same filter also has a 1:284 impedance ratio for the capacitor values. Also, filters derived from inverter prototypes yield a more realizable element parameter range. For the same specifications, an inverter-based filter turned into inductively coupled capacitive pi sections uses only 1 value of inductors (126nH) and the capacitance range is limited to 0.15 to 1.45pF (1:9.8 impedance ratio).

For distributed element bandpass filters there are many options based on the available technology. If a microstrip or stripline technology is used, and if area permits, you can design edge coupled or stepped impedance resonator filters.

Hairpin and interdigital filters offer great advantages on microstrip. For machined structures, combline is ideal for machining and tuning.

13.4.8.1. Lumped Element Bandpass Filter

A lumped element bandpass filter is created by applying frequency transformation to the lowpass prototype filter.

For monotonic filters, this corresponds to SLC resonators on the series arm and PLC resonators on the shunt arms. For filters with finite TZs, shunt arms contain LC-quad elements. In iFilter, quad elements are replaced by SLC resonators on the shunt arm.

This type of filter gives accurate response for any bandwidth, however, the element parameter magnitudes can be widely spread. The ratio of maximum to minimum inductance or capacitance can easily reach beyond 100. In that case, you cannot use components from the same part family, or the parasitics effects are difficult to tune out.

Realization of lumped element filters is defined in “Lumped Model Options Dialog Box ”.

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges

  • BW: 0 to 200% (extreme specs result in impractical impedances)

13.4.8.2. Narrowband Lumped Element Filter

A narrowband lumped element filter is derived from inverter prototypes. The main difference between this filter and conventional bandpass filters is the use of series arm components and their parameter range. For conventional filters, the series arm consists of LC-resonators. For narrowband filters, you can approximate the series arms (coupling elements) with single inductors or capacitors. The other difference is the spread in parameter values. For narrowband filters, the spread in parameter values is much less than that found with the conventional type.

There are six options in iFilter for narrowband lumped element bandpass filters. The first three are Inductively coupled options. These contain a series inductor as the coupling elements. Inductively coupled type filters are quasi-lowpass structures, where the stopband attenuation is steeper on the high end of the filter. The other three options use Capacitively coupled shunt sections. These contain a series capacitor as the coupling elements. Capacitively coupled type filters are quasi-highpass structures, where the stopband attenuation is steeper on the low end of the filter. When bandwidth gets wider, the high end attenuation almost disappears for low order filters.

The first option is the generic Inductively Coupled type. It is a popular narrowband filter, however as it contains many inductive components, it tends to be lossy. With proper setting of parasitics and unloaded Qs, this type can simulate filters of quasi-lowpass nature, such as combline filters. For more information, see “Arbitrary Narrowband Filter Simulation Example”.

The second option is the Inductive (identical Shunt C) option where all the shunt capacitors are equated to the same value. This is an attractive feature for applications that require a narrow capacitance range. Tunable filters are prime candidates, where they can be designed with identical tuning diodes using this topology.

The third option is the Inductively coupled Cap-Pi’s where the resonating shunt capacitances of the original inductively coupled type are now replaced by capacitive Pi-sections. You can use this topology to design tubular filters, which are not included in this version of iFilter. This topology has the minimal element value spread of the inductively coupled filters.

The fourth option is the generic Capacitively coupled type. It is another popular narrowband filter, and unlike the inductively coupled type, it does not suffer from insertion loss, as it contains mainly high-Q capacitors.

The fifth option is the Capacitive (identical shunt L) option where all of the shunt inductors are engineered to be the same value. Making all inductors share the same value uses an iterative technique. This process yields inductor values close enough in magnitude that you can use the same coil structure and slightly tune them.

The sixth option is the Alternating Ind/Cap option where the coupling elements alternate between inductor and capacitor. The response is fairly symmetric, not skewing the stopband performance.

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges (with limitation of 2 to 25 range)

  • BW: 0 to 50% (wider bandwidths suffer mismatched return loss)

13.4.8.3. Coupled Resonator Bandpass Filter

The coupled resonator filter is a special form of a Capacitively Coupled Narrowband Filter with equal shunt inductors. It has the advantage of using series input and output capacitors, making it easy to connect to terminations.

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges (with limitation of 1 to 25 range)

  • BW: 0 to 50% (wider bandwidths suffer mismatched return loss)

  • Lshunt: Any suitable range

13.4.8.4. Wideband Lumped Element LP+HP Filter

A wideband lumped element filter is obtained by cascading lowpass and highpass filters. For wide bandwidths, the two filters do not interact, so tuning the passband corners is very easy, as the lowpass side controls the upper frequency corner and the highpass side controls the lower frequency corner.

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges (with limitation of 1 to 25 range)

  • BW: 20 to 200% (narrower bandwidths suffer interactions)

13.4.8.5. Lumped Bandpass Multiplexer

iFilter can design lowpass bandpass diplexers using a “single terminated prototype” method. In this method, source impedance is first assumed ZERO ohms for both multiplexer channels, and after combining all channels in parallel, the source impedance is then set back to 50 ohms. iFilter designs, analyzes, and exports the diplexer channels as in the following figure. The channels are normally connected at the source end and there is only one source termination, however, for display purposes, channels are displayed as if they have a source termination each.

When you select the multiplexer type the first time, a multiplexer setup dialog box displays.

The channel list on the left lists bandpass channels. The lower passband cutoff frequencies can be set while editing each channel. The cutoff frequency is normally set to the same frequency for adjacent channels, however iFilter provides the flexibility of setting them separately when you select the Contiguous Channels check box, so frequency corners can be slightly different to optimize the return loss.

The Quick Setup section provides controls for setting multiple channels. You can also specify the number of channels, the lower passband corner of the first channel, and the common bandwidth of all channels.

On the left of the main iFilter dialog box are channel access buttons. The first box shows the selected channel number. The next two buttons toggle between LP and HP channels. The Edit button displays the diplexer setup dialog box where you can edit the frequencies.

13.4.8.6. Shunt Stub Bandpass Filter

A shunt stub bandpass filter is the same as a shunt stub highpass filter with the specification reformatted. For highpass filters, Fp and EL are defined. For bandpass filters, Fo and BW are defined. The following relations can therefore be established:

Fo = Fp * 90/EL

BW = 2*(Fo-Fp)

The comment for the shunt stub highpass filter is also valid for the shunt stub bandpass filter. The passband of the bandpass filter is controlled by BW rather than EL.

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges (with limitation of 3 to 19 range)

  • BW: 0 to 200% (extreme specs result in impractical impedances)

Tuning and Optimization

L_v(n) - are the element line lengths. Slight tuning may be required for the lengths for accurate passband corner and return loss.

13.4.8.7. Optimum Distributed Bandpass Filter

An optimum distributed bandpass filter is the same as an optimum distributed highpass filter with the specifications reformatted. For highpass filters, Fp and EL are defined. For bandpass filters, Fo and BW are defined. The following relations can therefore be established:

Fo = Fp * 90/EL

BW = 2*(Fo-Fp)

The comment for the optimum distributed highpass filter is also valid for the optimum distributed bandpass filter. The passband of the bandpass filter is controlled by BW rather than EL.

Typical Specifications
  • Approximation: Chebyshev only with fixed 26dB, 20dB or 16dB return loss.

  • Degree: 3 to 14

  • BW: 10 to 180% (extreme specs result in impractical impedances)

Tuning and Optimization
  • L_v(n) - are the element line lengths. Slight tuning may be required for the lengths for accurate passband corner and return loss.

13.4.8.8. Edge Coupled Bandpass Filter (Parallel Coupled Line Filter)

This filter type is also known as Half-Wavelength Parallel Coupled Line Filter, because it uses half-wavelength resonators. The term "edge coupled" is used in iFilter, as the commonest media are microstrip and stripline, where the resonators are coupled through their edges. For Suspended Stripline, or Broadside Coupled Stripline, the term "edge coupled" is not literally correct.

Because of its unique longitudinal shape, it is one of the most popular bandpass filters in moderate bandwidth microwave applications. Another advantage of these filters is that they do not require grounding. On the negative side, two things are worth considering when designing edge coupled filters:

The outermost coupled sections (next to the terminations) require significant coupling, so the spacing is normally tighter than that found with the internal resonator sections. For higher fractional (or percentage) bandwidths, more coupling is required, which necessitates that the resonators be even closer. Due to manufacturing limitations, the minimum gaps between lines must be followed. Naturally, this limits the fractional bandwidths that can be obtained. You can select thicker substrates to obtain wider spacing, but it is at the expense of increasing the cost and size.

The second disadvantage is the spurious frequency response. Theoretically, this topology possesses spurious passbands at odd multiples of the desired center frequency, Fo. If an ideal coupled line filter is designed at 2GHz, it naturally passes 3*Fo, 5*Fo components at 6GHz and 10GHz. For a homogeneous medium like stripline, actual spurious passband content is as calculated in ideal coupled line cases. Microstrip, although it is far more popular than stripline, is not a homogeneous medium, so the even and odd mode mismatch results in the 2*Fo spurious passband emerging. The same 2GHz filter, therefore, if built on a microstrip, has a passband of approximately 4GHz. To counter this, special techniques exist such as wiggly spacings between resonators, and lengthening/shortening resonators beyond the coupled region. In the current version of iFilter only length-adjustment is available.

Edge coupled filters are available in microstrip and stripline.

The microstrip transmission medium exhibits unequal even and odd mode effective dielectric constants. This results in uneven guided wavelengths which in turn cause unwanted spurious passband about 2*Fo. To counter this phenomenon, there are several design techniques such as making the coupled lines "tapered" or "wiggly" or defected ground structures. No such guidance data for wiggly lines has been available for all substrates and thickness. You should design the filter using the technique provided first, then by exporting it into the EM simulator, you should fine-tune it for better spurious performance. It is time consuming to concurrently design and tune a microstrip edge coupled filter structure.

The stripline medium does present the same problem, so it is relatively easier to work with stripline filters. These filters also have a drawback, however. For the same specifications, the first and last resonator sections tend to be more closely spaced for stripline. One solution is to use tapped input/output sections which replace the two coupled line sections with open stubs.

You can specify the source and load impedances as other than 50 ohms, in which case the even and odd mode impedances of the nearest coupled line sections are adjusted accordingly for impedance matching.

There are three options for edge coupled filters: Impedance Controlled, Standard, and Tapped input/output.

For the Impedance Controlled option, the Reson Zo parameter controls the internal impedance level of the coupled resonators. If Reson Zo is specified as close to 50 ohms, the coupled line impedances and line widths are comparable to terminating lines. If a higher impedance is used for Reson Zo, the coupled sections have higher even and odd mode impedances, and the lines are narrower. For tight couplings, a high Reson Zo is recommended, as it results in slightly higher spacings. You should note, however, that the insertion loss increases due to narrower lines. This option works for moderate bandwidths.

The second option is the Standard textbook option where no impedance control is available. This option can achieve slightly higher bandwidths than the Impedance Controlled option.

The third option is the Tapped input/output option. This option is a variation of the Impedance Controlled type. The input/output coupled sections, which have tight spacings, are replaced with an open-circuited shunt stub and a short transmission line. As a result, a more practical circuit is obtained. Another advantage of the Tapped input/output type is the improved spurious suppression. As the tapping sections do not conform to commensurate (identical) lengths of the resonators, the spurious passband may occur at arbitrary frequencies depending on the phases and impedances of lines. The two associated drawbacks of the tapped option are the availability of achievable bandwidths and the imperfect return loss. For wide bandwidths, the tapping section cannot simply provide the replaced coupling, therefore you should slightly tune internal couplings. You should also slightly tune Return loss due to the finite capability of the tapping. The Tapped input/output option is optimized for 20dB return loss (0.0436dB passband ripple), so using other passband ripple values may not result in a good match.

Typical Specifications
  • Approximation: Monotonic (for Impedance Controlled and Standard ) and Chebyshev only (for Tapped input/output)

  • Degree: 3 to 15

  • BW: no limitation, but reasonable performance for up to 40%

  • Resonator Zo: 30 to 120 ohms (must be realizable) – yields thinner lines with increasing Zo

Tuning and Optimization
  • L_v(n) - are the lengths of various line sections. Tuning may be required to set the response to the desired center frequency. For microstrip, you should tune the length parameter which is initially set to 0.000254mm (0.01mil). Negative values of this parameter suppress 2*Fo spurious passband, however the coupled line lengths should be slightly tuned to move the response back to the desired center frequency.

13.4.8.9. Stepped Impedance Resonator (SIR) Bandpass Filter

Stepped impedance resonator bandpass filters look like edge-coupled bandpass filters with extra transmission lines between the coupled sections. They offer the following advantages:

  • They do not suffer 2*Fo spurious response as half-wave edge coupled filters. The spurious response can be shifted by changing the transmission line impedance.

  • Shifting harmonic response at 2*Fo is especially useful in oscillator and amplifier designs.

  • The connecting lines can be made thin enough to bend in any shape, which provides more layout options. The layout can be made very compact.

  • The input and output ports do not have to be placed along the same axis.

The spurious passband of edge coupled filters is not particularly controllable. It appears at odd multiples, for example, 3*Fo and 5*Fo. In addition, for microstrip filters, the even multiples come into effect. The SIR bandpass filter is a slight variation of edge-coupled filters and was created to address this issue. By converting part of the coupled sections into simple transmission lines, the spurious passband can be moved further away from the desired passband. The inserted transmission lines can then be bent for improved realization, resulting in a compact filter. The drawback is that the available fractional bandwidth is narrower than comparable edge-coupled filters. The application is limited to 30% or less bandwidths in practice.

The stepped impedance resonator filter allows editing of the Line Zo (the transmission line pieces). Line Zo controls the location of spurious passband. Specifying a high value for Line Zo causes the spurious passbands to move away from the desired center.

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges (with limitation of 2 to 15 range)

  • BW: no limitation but reasonable performance for up to 25-30%

  • Resonator Zo: 30 to 150 ohms (must be realizable)

Tuning and Optimization
  • L_v(n) - are the lengths of various line sections. Tuning may be required to set the response to the desired center frequency.

13.4.8.10. Interdigital Bandpass Filter

Interdigital bandpass filters consist of tightly spaced vertically oriented resonators, so they offer significant size advantage over most other microwave filters. Originally intended for round rod and rectangular bar technologies, interdigital filters are now used in microstrip more than other filters, due to their simplicity in realization.

Interdigital resonators have fixed 90 degree electrical lengths. They are open-ended on one end and short-circuited on the other. In microstrip or stripline, a short-circuit is provided by vias. In rectangular bars and round rods, the short-circuited end is connected to the housing.

Interdigital filters can be approximated more accurately in homogeneous media such as stripline. In iFilter, there are various options depending on the selected technology.

For microstrip, a non-homogeneous medium, tapped and uniform width line options are available. iFilter uses EM-generated data for calculating spacing between resonators. The resulting structure still needs tuning, mostly tapping lengths and/or the outermost spacings. For other technologies, non-uniform widths are also available.

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges (with limitation of 16)

  • BW: no limitation but reasonable performance for up to 50%

Tuning and Optimization
  • L_v(n) - are the lengths of various line sections. Tuning may be required to set the response to the desired center frequency.

  • LTOT - is the length of the resonators. Increasing/decreasing it moves the center frequency of the filter down/up respectively.

  • LBot - is the tapping length at the input and output measured from the end of the resonator. For the input side, this is the bottom of the resonator. For the output side, it is the top or bottom of the resonator if the number of resonators is an even or an odd number.

  • S(n)_v(n) - are the spacings between resonators. Tuning may be required to set correct passband width.

13.4.8.11. Combline Bandpass Filter

Combline bandpass filters exploit the interdigital filter idea, where all the resonators are coupled together in a small form factor. In addition, combline resonators are adjusted to shorter lengths (down to 30 deg), and the resonance is obtained by adding a tuning capacitor. For planar filters, the capacitor is simply a lumped element. For machined structures like round rods and rectangular bars, capacitance is obtained by tuning rods.

Combline filters possess a topology similar to interdigital filters, however the orientation of combline resonators are all the same (all are short-circuited at the same side). This provides the advantage of tuning the filter at the open ends, which are on the same side of the housing. For planar filters like microstrip and stripline, this is not a big advantage, but for machined structures it is.

Combline filter resonator couplings are inductive, so the response is of quasi-lowpass nature (the selectivity obtained on the upper side of the passband is much sharper than on the lower side).

Typical Specifications
  • Approximation: Monotonic

  • Degree: See standard monotonic ranges (with limitation for 3-16 range)

  • BW: no limitation, but reasonable performance for up to 50%

  • ElecLng: 30 to 85 degrees

Tuning and Optimization
  • L_v(n) - are the lengths of various line sections. Tuning may be required to set the response to the desired center frequency.

  • S(n)_v(n) - are the spacings between resonators. Tuning may be required to set the correct passband width.

13.4.8.12. Hairpin Bandpass Filter

Together with interdigital and combline filters, hairpin filters offer the smallest size for a given number of resonators. Hairpin filters may be thought of as a variation of edge coupled filters. The difference is that the short line connects two coupled sections. Like edge coupled filters, the resonators are 180 degrees long, and as a big advantage, no grounding is needed.

Hairpin resonators are U-shaped and they alternate up and down for orientation. Hairpin filter is based on the coupling between U-resonators. The undesired coupling between arms of U can pose a problem if not properly controlled. In iFilter, the distance between arms is set to 2*W, which limits the coupling to less than 25dB.

Hairpin filters are only built on microstrip and stripline. For other media, supporting U-resonators is not a feasible idea.

Typical Specifications
  • Approximation: Chebyshev only

  • Degree: 3 to 8

  • BW: up to 40% (requires more tuning and tight line spacings as BW% gets higher)

  • Resonator Zo: 40 to 90 ohms (must be realizable)

Tuning and Optimization
  • LTOT - is the length of one arm of the U-shaped hairpin resonators. Increasing/decreasing it moves the center frequency of the filter down/up respectively.

  • LBot - is the tapping length at the input and output measured from the bend of the U towards the open end. Increasing or decreasing LBot causes the passband return loss to go up and down. For speed purposes, the tapping length cannot be calculated very accurately. You should tune this parameter in the Microwave Office program in a few seconds. Other lengths are formulated to keep the center frequency constant as much as possible while tuning LBot.

  • L_v3 (usually) - is the outermost line extension. It represents the amount of line length to add or deduct from the outermost U-section arms. This parameter is usually negative, meaning that the outermost arms should be a little shorter than the inner lines. Decreasing this parameter (making it more negative) usually reduces the upper corner end return loss to even out across the passband. There is no explicit formulation available for this parameter. In fact, examples in the literature assume that it is a positive length. The best approach is to tune it in the Microwave Office program which takes just a few seconds.

  • S1_v1 - is the spacing between arms of U-shaped resonators. It is conventionally taken as twice the resonator width. Tuning this parameter does not affect the response much for most designs. When the dielectric constant is quite high, some adjustment is expected.

  • S2_v1 - is the spacing between the first and second U-resonators (from both ends mostly). Tuning this affects the passband return loss. Passband width is usually not affected by S2_v1.

  • All other spacings - as Sx_v1 are the inner resonator spacings. Increasing them widens the passband and decreasing them narrows it. To achieve a well-matched response for the desired bandwidth, you should tune all inner spacings at the same time. Decreasing one spacing distorts the return loss while widening the passband. Decreasing another spacing widens the passband more, but also decreases the return loss to the desired level.

13.4.9. Bandstop Filters

Bandstop filters are designed to stop signals within a specific frequency range and pass all signals outside of that frequency range. For simplicity, bandpass filter terminology is applied to bandstop filters. The center of the stopband is called Fo and the stopband width is called BW. Here, BW is measured from lower passband cutoff frequency to upper passband cutoff frequency. For filters that have equi-ripple passband characteristics (Chebyshev, Elliptic, and Generalized Chebyshev), this corresponds to the passband ripple corner. For other approximation types it is the 3.011 dB (3-dB) corner frequency.

Bandstop filters are designed to provide infinite attenuation in the stopband center. The depth which the bandstop filter can achieve depends on the quality. Just as bandpass filters have insertion loss due to lossy elements and parasitics, bandstop filters cannot provide infinite or zero impedance, so there is always a leakage towards load side. To achieve very narrowband bandstop filters, you needs high Q elements, just as for very narrowband bandpass filters.

For distributed element filters, the response repeats itself due to the periodicity of electrical lengths, so you can also use a distributed lowpass filter as a bandstop filter.

13.4.9.1. Lumped Element Bandstop Filter

Lumped element bandstop filters contains series PLC and shunt SLC elements. For filters with finite TZs, SLC resonators are replaced by quad LC-resonators, which are simplified to 2 shunt SLC resonators. A Chebyshev bandstop filter is shown in the following figure.

Realization of lumped element filters is defined in “Lumped Model Options Dialog Box ”.

Typical Specifications
  • Approximation: All available

  • Degree: See standard ranges

13.4.9.2. Optimum Distributed Bandstop Filter

Distributed element bandstop filters are identical to optimum distributed lowpass filters with the specification reformatted. A bandstop filter is defined by center frequency Fo and bandwidth BW. An optimum lowpass filter is defined with EL at passband corner Fp. At 90 degrees long, the lowpass filter has no transmission. The S21 response is re-entrant at 90+EL as the corner of spurious passband. The following relations can therefore be established between two filters:

Fo = Fp * 90/EL

BW = 2*(Fo-Fp)

Optimum distributed bandstop filters are not suitable for narrow bandwidths. The shunt open-circuited stub impedances come out very high for bandwidths below 40-45%. Therefore, this filter is more suitable for wide bandwidths.

Typical Specifications
  • Approximation: Chebyshev only with fixed 26dB, 20dB or 16dB return loss.

  • Degree: 3 to 19

  • BW: 10% to 160%

Tuning and Optimization
  • L_v(n) - are the element line lengths. Slight tuning may be required for the lengths for accurate passband corner and return loss.

13.4.10. Auxiliary Dialog Boxes

iFilter includes settings and utility dialog boxes that you can display and use any time during the filter design process.

13.4.10.1. Design Utilities Dialog Box

The Design Utilities dialog box contains basic filter-related conversion and calculation utilities. To display this dialog box, click the Design Utilities button in the main iFilter dialog box.

Design Utilities VSWR (Conversion) Tab

This dialog box tab converts values between well-known filter parameters. To view this tab, click the Design Utilities button in the main iFilter dialog box and then click the VSWR tab. The following parameters are available for converting:

  • Passband ripple

  • Return loss

  • VSWR

  • Ref Coeff (Reflection coefficient)

  • Min ZLoad and Max ZLoad (Minimum and Maximum Load Termination)

Entering any of the first three parameters concurrently updates all other values in the dialog box.

NOTE: The passband ripple is associated with the filter’s insertion loss. It is named so as not to be confused with the dissipated loss. You can use the value for any insertion loss caused by impedance mismatch.

Design Utilities Midband IL (Midband Insertion Loss) Tab

This dialog box tab estimates the midband insertion loss of a bandpass filter. To view this tab, click the Design Utilities button in the main iFilter dialog box and then click the Midband IL tab.

The insertion loss of a bandpass filter can be approximated by the formula:

This formula associates the unloaded Qu of filter elements and the filter Q (Fo/BW) to the loss. Although initially developed for narrowband microwave filters, the concept is applicable to all bandpass filters made up of any technology.

When you edit any of the parameters, the dialog box concurrently calculates and updates the midband insertion loss. If you edit the midband IL, the required unloaded Q is calculated and updated.

Design Utilities Air Coil (Calculation) Tab

This dialog box tab helps you calculate the parameters of an air coil inductor. To view this tab, click the Design Utilities button in the main iFilter dialog box and then click the Air Coil tab.

For lumped element filters, you may need to use wound air coils. On this tab, you can conveniently calculate inductance and Q values. The minimum and maximum coil inductances are based on the gap between the turns. Lmax is given for 0.5wd (tight winding) and Lmin is given for 2wd (loose winding) where the variable wd defines the wire diameter. Although the formula is very accurate, the actual inductor value depends on the orientation and proximity of the coil to the housing.

Qu (unloaded Q) is more empirical, but gives you an idea of the order of magnitude for a realizable Qu.

Design Utilities Capacitance (Gap/Pad) Tab

Two parallel conducting plates separated by a distance d create a static capacitance:

where A is the area of the plates, and Er is the dielectric constant of the gap medium. Both A and d are in meters.

The separation may be air or a dielectric. Lumped element filters are built on printed circuit boards. Pads are needed on PCB where the coils should be connected to capacitors. Pads have a certain area where they create a capacitance to the bottom of the PCB. At high frequencies, capacitors of the filter elements reduce to few pF and below. In these circumstances, capacitance due to pads becomes comparable and may increase or decrease the effective capacitance of the filter elements.

For example, a 4mm2 pad on a 0.005” Duroid has 0.6pF. This should be included in a detailed design as if there is shunt capacitance to ground from every pad of 0.6pF.

13.4.10.2. Environment Options Dialog Box

The Environment Options dialog box displays unit settings and general settings. To view this dialog box, click the Environment Options button in the main iFilter dialog box.

Environment Options Units Tab

The iFilter units settings are initiated from the Microwave Office program, however when iFilter is running, you can set the units locally to the wizard. The settings are saved in each design separately, so if you load a previous design, the displayed units change to those of that design.

13.4.11. Design Examples

To demonstrate iFilter capabilities and simplify the wizard's functionality, the following examples are included.

13.4.11.1. Lumped Element BPF Example

This is a lumped bandpass filter design. Without targeting a specific application, the specification is defined as:

  • 15 dB minimum return loss in a passband of 475 to 525 MHz

  • Maximum insertion loss of 4 dB

  • 50dB attenuation at 600 MHz

  1. Open the Wizards node in the Project Browser and double-click the iFilter Filter Synthesis. The main iFilter dialog box displays with properties from any previous design.

  2. Click the Change Filter Type button at the top left of the dialog box in the Type-Approximation area. The Select Filter Type dialog box displays.

  3. Click the Bandpass button and then click the Lumped button in the row beneath it.

  4. In the Main Filter Type list, select Narrowband Lumped Filter. In the Options list, select Capacitive (identical shunt L), then click OK to close the dialog box.

    You now return to the main iFilter dialog box. The filter type changes but the common properties of the previous design still display.

  5. If the Change Response Approximation Type button (second button from the top) does not display "Chebyshev", click the button to display the Approximation Function dialog box, select Chebyshev and close the dialog box.

  6. In the Ripple text box, type "0.01", which corresponds to 16.4dB return loss.

    For Degree type "5".

    For Fo type "500".

    For BW type "50".

    You may see a plotted response.

  7. Click the Environment Options button in the main iFilter dialog box to display the Environment Options dialog box. Click the Units tab and ensure that Frequency is set to MHz and Dimension is set to mm.

  8. Click the WS (Wide Analysis Span) button in the row of buttons under the plot to set the wideband span to the chart.

  9. To see actual marker values, click the Edit Chart Settings button at the top left of the plot. In the Chart Settings dialog box, click the IL+RL button, then click the Markers button.

  10. Click the Add button to display the Add Marker dialog box. For Fmin type "500", then select Insertion Loss. Repeat this step for an Fmin value of 600 MHz, and then click OK to close the dialog box.

    The markers now display on the insertion loss trace.

    Unfortunately, the filter does not have the required 50dB attenuation at 600 MHz; it has only 45dB of attenuation. By increasing Degree to 6, you can get 61dB attenuation at 600MHz. Increasing the filter order adds extra elements to the circuit, but gives you more attenuation than needed, and also increases the midband insertion loss. By widening the bandwidth of the filter, you can find a suitable attenuation and less insertion loss. Increasing BW to 56 MHz, you now have 55dB attenuation with 5 dB cushion above the specification.

    The filter has shunt resonators coupled with capacitors. The resonators have identical shunt inductors, which is ideal as you only need one inductor value for the filter design. An air wound coil of 1 to 1.5 turns of AWG #32 wire with 2mm coil diameter provides about 2 nH of inductance. The same dialog box reports QL=164 at 500 MHz. A commercially available capacitor has QC=350 at 500 MHz. By setting these values to QL and QC respectively, and clicking the Analyze Lossy button, you can approximate the response with more realistic nonideal conditions. The lossy simulation now shows 3.6dB insertion loss with 55dB attenuation.

  11. Click the Design Options button to set element realization options. In the Lumped Model Options dialog box, select Use AIR COIL and Coil (1-10nH) for the inductor value range of 1-10nH. Set Rmax to "2" mm for coil diameter and Wire to awg32.

  12. Select Vendor Part and then under Capacitors select Vendor 0-10pF and Vendor 10-100pF for capacitor value ranges 0-10pF and 10-100pF. Click the Capacitor Vendors button to display the Capacitor Vendors and Parts dialog box. In this dialog box, select your preferred vendors along with the part sizes. For 500MHz, NI AWR recommends sizes up to and including 0805.

  13. Click the Search for available types button and select 600L and 600S series and move them to the Selected Types list by clicking the right arrow. Click OK to close the dialog box, then close the Lumped Model Options dialog box as well.

    iFilter searches through the selected part types and finds the most suitable parts with highest Q.

  14. Click the View Circuit Information button to see what components are selected for the current schematic. This view details the recommended parts together with the variation of element values. To see the effects of the real-life components, click the Analyze Real button in the Analysis group. The response changes slightly from the lossy model. It may shift along frequency axis and may scale up or down depending on the available Q. If the response becomes significantly different than ideal or lossy models, it is most likely due to the insufficient modeling of the vendor parts selected.

  15. Click the Generate Design button to generate the design in the Microwave Office program. In the NI AWR Design Environment suite you can perform optimization and yield analysis. A PCB layout can be drawn up and exported CAD/CAM data can be generated for the board manufacturer.

13.4.11.2. Microstrip Bandpass Filter Example

This is a tapped input edge coupled filter design. The specification is defined as:

  • 15 dB minimum return loss in a passband of 4500 to 5500 MHz

  • Maximum insertion loss of 2 dB

  • 50dB attenuation at 3600 MHz

  1. Open the Wizards node in the Project Browser and double-click the iFilter Filter Synthesis. The main iFilter dialog box displays with properties from any previous design.

  2. Click the Change Filter Type button at the top left of the dialog box in the Type-Approximation area. The Select Filter Type dialog box displays.

  3. Click the Bandpass button and then click the Microstrip button in the row beneath it.

  4. In the Main Filter Type list, select Edge Coupled Bandpass Filter. In the Options list, select Tapped input/output, then click OK to close the dialog box.

    You now return to the main iFilter dialog box. The filter type changes but the common properties of the previous design still display.

  5. The Tapped input/output type is optimized for 20dB return loss, so click the text box to the right of the Ripple [dB] button and scroll the mouse wheel until the value is 0.0436dB. A 6th order filter with Fo=5000MHz and BW=1000MHz meets the required specs.

    Next is the selection of a suitable substrate. A major selection criterion is typically the cost of the material. Some substrates are cheaper to purchase and process, but these may perform well. Electrically, the quality factor of the substrate that affects the insertion loss is the most important property to be aware of. The quality factor is determined by two losses: Conductor loss and Dielectric loss. Conductor loss is a function of skin depth and line width. As line widths are proportional to the substrate height, selecting thicker substrates yield wider line widths, so low insertion loss. Dielectric loss is substrate-dependent and is determined by the TanD (dielectric loss tangent) parameter. In iFilter, you can specify the TanD parameter, however it is not taken into account for determining the dimensions. When the final simulation is performed using the Microwave Office program to calculate the filter response accurately, TanD is used. Naturally, a low TanD yields better insertion loss in the passband.

  6. In this design example, assume a RO4350B commercial material is available with Er=3.48, 0.030-inch thickness (0.762mm). To enter the parameters, click the Design Options button and click the Technology tab in the Distributed Model Options dialog box. Enter the substrate Height, click in the Substrate Er text box, and scroll the mouse wheel until the 3.48 displays. TanD displays a value of 0.0037. Select a suitable Cond.Thickness (conductor thickness). A ½ oz. copper, corresponds to 17um (0.017mm or 0.67mil).

  7. In the same dialog box, click the Realization tab and select Add input and output lines to the layout to add reference 50 ohms lines to both sides of the filter. Clear the other check boxes and click OK. The filter is ready to analyze and optimize.

  8. Click the Edit Chart Settings button at the top left of the plot. In the Chart Settings dialog box, click the OptGoals button, then click the Auto button to automatically add some OptGoals to the design. Select the OptGoal from list which determines lower stopband with 50dB spec. Because iFilter calculates 50dB points from ideal attenuation functions, you should edit this. You should also edit Return loss to cover 15dB only, rather than the designed 20dB.

  9. Click the Generate Design button and export the filter into the Microwave Office program with response. iFilter generates a design with schematic, graph(s) and optimization goals already set. Due to accurate analysis of microstrip TEEs and steps, the return loss is slightly worse than that shown in iFilter.

  10. In the NI AWR Design Environment suite, choose Simulate > Optimize. In the Optimizer dialog box, check only L_v2, L_v3 and S_v1, the most effective parameters on the return loss. Click the Start button to quickly converge to a solution.

  11. You can now try the effect of TanD on the insertion loss. At 0.0037, IL is 1.01dB. Increasing TanD to 0.01 increases the insertion loss 2.05dB. This demonstrates the importance the dielectric loss tangent is at high frequencies.

13.4.11.3. Arbitrary Narrowband Filter Simulation Example

You can simulate narrowband microwave bandpass filters with lumped elements if you use the proper coupling and enter unloaded Q values correctly.

A combline filter is quasi-lowpass in nature (inductively coupled). Its selectivity in the upper stopband is better than that of the lower stopband. If you select an inverter-based filter type with inductive coupling, you get series inductors and shunt LC-resonators in your filter. Although a combline filter is made in an entirely different medium with metal bars and screws etc., the lumped element filter's stopband response closely resembles the combline filter.

You should set the LC-resonator element Q's slightly higher than what can be obtained with a combline structure to give the passband response of the hypothetical combline filter. To do so, click the Design Options button and click the Realization tab. Select the For BPF: set loss factors to shunt resonators only check box, then enter approximately 1.6 times higher Q (obtained for combline structure) for both QL and QC. The number 1.6 is obtained after some sample designs, but can also be changed to your previous findings.

The response should now provide an idea of the combline filter's response, both in passband and stopband. By placing markers, you may obtain insertion loss values at any frequency.

This method is valid for any microwave filter with known coupling type and unloaded Q-value. Types of microwave filters range from low-Q LTCC's to dielectric resonator with very high element Q's. This method is more advanced, yet equally simple as the industry standard midband loss method, which only addresses the passband center.

Please send email to awr.support@ni.com if you would like to provide feedback on this article. Please make sure to include the article link in the email.

Legal and Trademark Notice