RF·Tools

Microstrip filters

A filter made of copper alone: stepped impedance, radial stubs, quarter-wave stubs, coupled lines, hairpins or a comb. The dimensions are computed for the substrate you name, and the response is computed from those very dimensions, with dispersion and loss. So the plot shows the real insertion loss and the spurious passbands the prototype knows nothing about. Hairpin and interdigital filters are solved with a multiconductor model: there every resonator couples to two neighbours at once, and a chain of pairs cannot describe that.

Type
 

Low-pass needs one frequency only. The refinement fits the dimensions to the response of the layout itself: the textbook calculation misses the cutoff by tens of per cent.

Layout

FR-4, 1.6 mm · size 139×27.2 mm Download SVG Download DXF
C1C2C3C4C5C620 mmrf.akzami.com

The copper is to scale; the circle is a via to ground. Steps and tee junctions are not modelled, so expect to trim the lengths a little in your PCB editor. Ctrl and the wheel, or a pinch, zooms; drag to move; a double click fits it back. The SVG and DXF carry the copper, the vias and the board outline — no dimensions, no scale bar; the sizes are real, in millimetres. The site address sits on a separate layer (MARK in the DXF) so that it cannot end up as copper on import — delete it if you do not want it.

Response of this layout

Download .s2p

A cascade of real line sections with dispersion and copper and dielectric loss — not the ideal prototype.

Dimensions

Element What it is Z, Ω θ, ° W, mm S, mm L, mm
C1 quarter-wave coupled section 93/36 90 1.68 0.13 21.6
C2 quarter-wave coupled section 73/45 90 2.20 0.67 21.2
C3 quarter-wave coupled section 56/36 90 3.36 0.59 20.9
C4 quarter-wave coupled section 56/36 90 3.36 0.59 20.9
C5 quarter-wave coupled section 73/45 90 2.20 0.67 21.2
C6 quarter-wave coupled section 93/36 90 1.68 0.13 21.6
feed system-impedance line 50 — 3.02 — —

What you get

Passband loss3.45 dB
Return loss≥ 9.3 dB
Spurious passband5.79 GHz, -9.4 dB
Board size139×27.2 mm
Substrateεr 4.40, h 1.60 mm

The dimensions were refined against the response of the structure. Before refinement the return loss was 5.7 dB.

a spurious passband at 5.79 GHz (-9.4 dB) — distributed filters always have one; if it matters, follow the filter with a lumped-element low-pass
the gap in section C1 is 0.127 mm, which an ordinary fab will not etch. The end sections of a coupled-line filter are always the tight ones: narrow the band, lower the order or pick a substrate with a higher εr
the even and odd modes differ in speed by 11 %, which turns the null at 2·f0 into a spurious passband. A lower-εr substrate or folded resonators cure it
a return loss of 9.3 dB is the ceiling for this topology at this order: the sections are too long to act as lumped elements. A higher order means shorter sections and a better match