Designing RF Filters

When to Use

• Designing bandpass filters (coupled-line, hairpin, open-loop, combline, interdigital, SIW)
• Designing lowpass filters (stepped-impedance)
• Designing bandstop or notch filters (spurline, stub-based)
• Extracting coupling matrices from measured S-parameter data (measuredFilter)
• Selecting a filter topology for a given bandwidth, selectivity, or size requirement

When NOT to Use

• Designing transmission lines for impedance control — use matlab-design-pcb-txline
• Designing couplers or splitters — use matlab-design-pcb-coupler
• Designing passive components (inductors, capacitors, baluns) — use matlab-design-pcb-passive
• Setting up substrate or conductor materials — use matlab-manage-pcb-material
• Optimizing filter dimensions after design — use matlab-optimize-pcb-design

Typical Workflow

1. Before: matlab-manage-pcb-material — set up substrate and conductor
2. This skill: Design the filter (catalog object or custom geometry)
3. Check mesh/memory: memoryEstimate(obj, fc, 'RetainMesh', true) — inspect auto-mesh density before committing to a full solve
4. After: matlab-analyze-em — validate S-parameters → matlab-optimize-pcb-design — tune dimensions → matlab-write-pcb-layout — export Gerber

Quick Reference — Filter Selection

Bandpass Filters

Coupled-Line Filter

f = filterCoupledLine;
f = design(filterCoupledLine, 3e9);    % Design at 3 GHz
show(f);
sp = sparameters(f, linspace(1e9, 5e9, 101), 'SweepOption', 'interp');
rfplot(sp);

Key properties: FilterOrder, CoupledLineLength, CoupledLineWidth, CoupledLineSpacing, PortLineLength, PortLineWidth.

Hairpin Filter

Folded coupled-line resonators for compact size:

f = design(filterHairpin, 3e9);
show(f);
memoryEstimate(f, 3e9, 'RetainMesh', true);  % Check mesh density before solving
sp = sparameters(f, linspace(1e9, 5e9, 101), 'SweepOption', 'interp');
rfplot(sp);

Key properties: FilterOrder, CoupledLineLength, CoupledLineWidth, CoupledLineSpacing, PortLineLength, PortLineWidth, Spacing, ResonatorOffset, FeedOffset.

Chebyshev Response

Pass FilterType and RippleFactor to design() for equiripple passband response:

f = design(filterHairpin, 1.8e9, FBW=10, FilterType='Chebyshev', RippleFactor=0.5);

FilterType options: 'Butterworth' (default), 'Chebyshev'. FBW sets fractional bandwidth (%). RippleFactor sets passband ripple in dB (default 0.5) — only applies to Chebyshev.

Fifth-Order Hairpin

f = filterHairpin;
f.FilterOrder = 5;
f = design(f, 2.4e9);
show(f);

Open-Loop Filter

Quasi-elliptic response with cross-coupling:

f = filterOpenLoop;
f.NumPoles = 6;
f.FeedOffset = 0.5e-3;
show(f);
sp = sparameters(f, linspace(1e9, 5e9, 101), 'SweepOption', 'interp');
rfplot(sp);

Key Properties: NumPoles (4/6/8), ResonatorLength, ResonatorWidth, SplitGap, GapHorizontal, GapVertical, FeedOffset, CoupledResonatorGap, QuadrupletGap, QuadrupletOffset.

Combline Filter

Short-circuited resonators, excellent for narrow-band. Note: filterCombline does not have a design function — set properties manually:

f = filterCombline;
f.FilterOrder = 3;
f.Height = 1.6e-3;
show(f);

Key Properties: FilterOrder, ResonatorLength (scalar or vector), ResonatorWidth, ResonatorSpacing (scalar or vector), ResonatorOffset, FeedOffset, Capacitor (loading capacitance — distinctive to combline).

Interdigital Filter

Alternating short-circuited resonators, wideband. Note: filterInterdigital does not have a design function — set properties manually:

f = filterInterdigital;
f.FilterOrder = 4;
f.Height = 1.6e-3;
show(f);
sp = sparameters(f, linspace(3e9, 7e9, 101), 'SweepOption', 'interp');
rfplot(sp);

Key Properties: FilterOrder, ResonatorLength (scalar or vector), ResonatorWidth (scalar or vector), ResonatorSpacing (scalar or vector), ResonatorOffset, ViaDiameter (scalar or vector — distinctive to interdigital), FeedOffset, IsShielded, Connector.

Lowpass Filters

Stepped-Impedance Lowpass

f = filterStepImpedanceLowPass;
f = design(filterStepImpedanceLowPass, 2.5e9);
show(f);
sp = sparameters(f, linspace(0.1e9, 5e9, 101), 'SweepOption', 'interp');
rfplot(sp);

Key properties: FilterOrder, HighZLineWidth, LowZLineWidth, HighZLineLength, LowZLineLength.

Bandstop / Notch Filters

Spurline Filter

Compact notch using coupled-line section on one side:

f = filterSpurline;
show(f);
sp = sparameters(f, linspace(1e9, 6e9, 51), 'SweepOption', 'interp');
rfplot(sp);

Double spurline for deeper rejection:

f = filterSpurline;
f.LineType = 'Double';
show(f);

Key Properties: LineType ('Single'/'Double'), CoupledLineLength, CoupledLineWidth, CoupledLineSpacing, LineGap (gap between coupled line and output line — distinctive to spurline), IsShielded, Connector.

Stub Filters (Open/Short)

The filterStub object supports open-circuit stubs (bandstop) and short-circuit stubs (highpass):

f = filterStub;
f.StubLength = [6e-3 6e-3 6e-3];
f.StubWidth = [0.5e-3 0.5e-3 0.5e-3];
f.StubOffsetX = [-6e-3 0 6e-3];
f.StubShort = [0 0 0];         % 0=open (bandstop), 1=short (highpass)
f.StubDirection = [0 0 0];     % 0=below, 1=above trace
f.SeriesLineWidth = 1.8e-3;
f.SeriesLineLength = 12e-3;
show(f);

Key Properties: StubLength (vector), StubWidth (vector), StubFeedOffsetX (vector), StubShort (0=open, 1=short; vector), StubDirection (0=down, 1=up; vector), SeriesLineLength, SeriesLineWidth, IsShielded, Connector.

Radial Stub

rs = stubRadialShunt;
rs = design(stubRadialShunt, 5e9);
show(rs);

SIW Bandpass Filter

SIWFilter uses NumResonators (not FilterOrder), and Substrate is read-only (set via internal resonator objects):

f = SIWFilter;
f.NumResonators = 4;
show(f);
sp = sparameters(f, linspace(8e9, 14e9, 51), 'SweepOption', 'interp');
rfplot(sp);

Measured Filter Extraction

measuredFilter extracts a coupled-resonator circuit model (coupling matrix, external Q, unloaded Q) from measured or simulated 2-port S-parameter data. This enables filter tuning, diagnosis, and comparison against ideal synthesis targets.

Key Properties:

• Sparameters — 2-port S-parameters data (sparameters object)
• FilterOrder — Number of resonators in the model
• CenterFrequency — Passband center frequency (Hz)
• BandWidth — 3-dB bandwidth (Hz)
• CouplingMatrix — Extracted N+2 coupling matrix (populated after extraction)
• QualityFactor — Unloaded quality factor (populated after qualityfactor())

Key Methods:

• residue(mf) — Extract lowpass admittance residues and poles from S-parameter data
• transversalMat(mf) — Calculate transversal coupling matrix from residues
• canonicalCouplingMat(mf) — Rotate transversal matrix to canonical (folded) form
• optimize(mf) — Isospectral optimization of the coupling matrix
• sparameters(mf, freq) — Synthesize S-parameters from the extracted circuit model
• qualityfactor(mf) — Calculate unloaded quality factor

Complete Workflow

% Step 1: Load measured S-parameters and inspect visually
S_meas = sparameters('measured_filter.s2p');
rfplot(S_meas);
% Identify center frequency and bandwidth from the plot

% Step 2: Create measuredFilter with matching initial guess
mf = measuredFilter(Sparameters=S_meas, ...
    FilterOrder=8, ...
    CenterFrequency=2114.6e6, ...
    BandWidth=9.6e6);

% Step 3: Extract residues and poles
residue(mf);

% Step 4: Build transversal matrix, then rotate to canonical form
transversalMat(mf);
M = canonicalCouplingMat(mf);
disp(mf.CouplingMatrix);

% Step 5: Optimize coupling matrix (isospectral flow)
mf = optimize(mf);

% Step 6: Compare extracted model vs measured data
freq = linspace(2.08e9, 2.15e9, 501);
S_model = sparameters(mf, freq);
rfplot(S_meas); hold on;
rfplot(S_model, '--');
legend('Measured', 'Extracted Model');

% Step 7: Quality factor
Q = qualityfactor(mf);
disp(mf.QualityFactor);

Custom Filter Assembly

For topologies not in the catalog, build with pcbComponent:

% Example: custom 2-pole open-loop filter from shape primitives
sub = dielectric("FR4");
sub.Thickness = 1.6e-3;
cond = metal("Copper");

% Build resonator shapes using traceLine, traceRectangular, Boolean ops
% (see matlab-assemble-pcb-layout skill)

pcb = pcbComponent;
pcb.Layers = {filterShape, sub, groundPlane};
pcb.BoardShape = groundPlane;
pcb.BoardThickness = sub.Thickness;
pcb.Conductor = cond;
pcb.FeedDiameter = feedWidth/2;
pcb.FeedLocations = [x1 y1 1 3; x2 y2 1 3];

Multi-Layer Filters

All filter catalog objects support multi-layer dielectrics:

f = design(filterCoupledLine, 3e9);
sub = dielectric("FR4", "Teflon");
sub.Thickness = [0.8e-3 0.4e-3];   % Set Thickness BEFORE assigning to filter
f.Substrate = sub;
f.Height = 1.2e-3;
show(f);

Filter Selection Guide

Design Adjustments

Pitfalls

1. Use interpolating sweep for S-parameters: Always use sparameters(obj, freq, 'SweepOption', 'interp') for MoM solves. Direct sweeps solve at every frequency point individually and are significantly slower.

1. Check mesh density before solving: Catalog filters generate dense auto-meshes that dominate runtime. Always run memoryEstimate(obj, fc, 'RetainMesh', true) before sparameters(). If memory is excessive, coarsen: mesh(obj, 'MaxEdgeLength', lambda/6). See matlab-analyze-em for full mesh inspection workflow.

1. FilterOrder vs NumPoles: Some objects use FilterOrder, others use NumPoles. Check the specific object's properties.

1. PortLineLength affects response: The port feed line length contributes phase and can shift the filter's apparent center frequency. Ensure sufficient length for proper excitation.

1. design() sets all dimensions: After design(obj, fc), all geometric parameters are overwritten. Customize properties after calling design, not before.

1. Coupled-line filter bandwidth: Controlled by CouplingSpacing — smaller gaps = tighter coupling = wider bandwidth, but fabrication-limited.

1. StubShort convention: In filterStub, StubShort=0 means open-ended stub (creates bandstop); StubShort=1 means short-circuited (creates highpass). This is counterintuitive.

1. No design() for combline/interdigital/stub: filterCombline, filterInterdigital, and filterStub do not support design(). Set dimensions manually based on resonator theory.

1. SIWFilter Substrate is read-only: You cannot directly set SIWFilter.Substrate. The substrate is controlled through the internal resonator and transmission line element objects.

1. measuredFilter initial guess matters. The CenterFrequency and BandWidth must closely match the actual passband of the measured data. Poor initial values cause residue() to extract incorrect poles. Inspect the S-parameter data visually first.

1. ResonatorSpacing vector length = FilterOrder - 1: For combline, interdigital, and hairpin filters, ResonatorSpacing specifies gaps between resonators. An N-th order filter has N resonators but only N-1 gaps. Do NOT pass N elements — pass N-1.

1. GroundPlaneLength is read-only on most filters: GroundPlaneLength and GroundPlaneWidth are auto-computed from resonator dimensions on filterCombline, filterHairpin, and filterInterdigital. Do NOT attempt to set them.

Related Skills

• matlab-manage-pcb-material — Substrate selection for filter performance
• matlab-analyze-em — S-parameter extraction and field visualization
• matlab-optimize-pcb-design — Optimizing filter dimensions
• matlab-assemble-pcb-layout — Custom filter topologies via pcbComponent
• matlab-design-pcb-passive — Resonators, baluns, split-ring structures for filters

----

Copyright 2026 The MathWorks, Inc.