2P Innovation · free, no sign-up

Distributed-element geometry synthesiser

Choose a topology and generate CAD you can use: a KiCad footprint, a DXF, dimensions and a layout to scale. Every design is checked against six fabricators' published limits before you order the board.

Edge-coupled bandpass filter

Parallel-coupled half-wave resonators. The classic microstrip bandpass.

validated Response validated against published data

Compactness 38 x 8 mm, N=5 at 4.7 GHz. The long one.

Bandwidth Happy from about 5 to 20%. Widest of the three.

Manufacturability Single layer, no vias, nothing to fit.

Power handling Set by the coupling gaps, 0.13 mm at 10%.

Model confidence Validated against Pozar 8.7, in the regression suite.

Notes

Spurious at 2f0. Plan a lowpass after it if the second harmonic matters.
Below roughly 5% FBW the first gap drops under 0.1 mm. Check the fab panel before ordering, not after.
At 1 GHz this stops being a sensible shape. Go hairpin or interdigital.
The one to reach for when you want the answer to be right rather than small.

Synthesise Edge-coupled →

Hairpin bandpass filter

Edge-coupled resonators folded into hairpins.

unconfirmed Closed form not yet confirmed against full-wave EM

Compactness 21 x 9 mm, same spec as the edge-coupled. Half the length.

Bandwidth About 5 to 15%. Folding costs you the top end.

Manufacturability Single layer, and the tapped feed dodges the tightest gap.

Power handling Same gap limit as edge-coupled, tighter bends.

Model confidence Closed form only. Not confirmed against full-wave EM.

Notes

Folding couples the two arms of each resonator and the equivalent edge-coupled model does not see it. Expect f0 to land low.
Simulate before committing. This is the only topology here whose response has not been checked against EM.
The tapped feed puts the tightest feature inside the array rather than at the port, which is usually what you want.
Buys length, spends width. Not a smaller part, a differently shaped one.

Synthesise Hairpin →

Interdigital bandpass filter

Quarter-wave bars grounded at alternating ends.

indicative Response is indicative, geometry is the deliverable

Compactness 9 x 10 mm, N=5. Smallest filter here by area.

Bandwidth About 5 to 20%.

Manufacturability 7 vias and a real ground plane. Not just a copper pattern.

Power handling Wide bars, no fine edge gaps.

Model confidence Idealised inverter model. Band edges are indicative.

Notes

First spurious near 3f0 rather than 2f0. The clean window above the passband is much wider than either half-wave filter gives you.
Via inductance is not in the model, and the bar-to-ground transition is a fab detail this tool cannot check.
Of the three filters, the one most worth an EM run before you commit copper.
Bars side by side suit a narrow board, which is often the deciding factor.

Synthesise Interdigital →

Wilkinson power divider

Two-way split with matched ports and isolation between the outputs, set by a single bridging resistor.

validated Well-behaved closed form

Compactness 11 x 11 mm at 4.7 GHz.

Bandwidth 20 to 30% single section, then it falls away.

Manufacturability Copper is easy, but a resistor has to be fitted.

Power handling The resistor takes the imbalance. Size it for the fault case, not the matched one.

Model confidence Exact closed form.

Notes

Isolation is set by the resistor's placement and its parasitic inductance, not by its value. At these frequencies the mounting matters more than the part.
Unequal splits work as readily as equal ones.
Past about 30% go multi-section. A single section will not hold both match and isolation.
The only part here that is not purely copper.

Synthesise Wilkinson →

Branch-line coupler

Quadrature hybrid: an equal split with 90° between the outputs. Four quarter-wave arms in a ring.

validated Well-behaved closed form

Compactness 11 x 11 mm at 4.7 GHz. Smaller than it looks.

Bandwidth 10 to 20%. Narrowest thing on the site.

Manufacturability Pure copper, wide traces, no vias, nothing to solder.

Power handling 35 ohm arms, no fine gaps anywhere.

Model confidence Exact closed form, and symmetric enough to check by eye.

Notes

Series arms run at Z0 over root two, about 35 ohm, so they come out wide. The area goes up fast as frequency comes down.
Four junctions, all treated as ideal. Compensating them is an EM job.
90 degrees between the outputs is the whole point: balanced amplifiers, I/Q mixers.
Split ratio and phase both drift outside the band. It degrades gracefully, but it does degrade.

Synthesise Branch-line →

Things to note before using this tool

Fabrication. Geometry is checked against six shops' published limits. If it cannot be built, you are told, and by whom it can be.
Loss. There is no dissipation model. Every response is lossless and peaks at 0.000 dB. Insertion loss is not something this tool knows.
Models. Every closed form is cited with the range it is valid over. Outside that range the engine warns rather than extrapolating quietly.
Substrates. Permittivities are datasheet design values, not IPC process values. On a 4.75 GHz hairpin that is about 95 MHz of centre frequency.
Confidence. Each topology states how far its response has actually been verified. They are not all the same.
Scope. Closed-form synthesis, not simulation. Verify in EM before you fabricate.

Coming soon, maybe

Coupled-line directional coupler. Built, but its directivity is an unverified prediction, and directivity is the only number that decides whether the part is usable.
Radial / butterfly stub. Not built. The curved-geometry primitive is done; the synthesis is not.

2P's own rule for the engine is that if a capability is not built and functional, it is not claimed. The website inherits that rule.

Want a design verified full-wave?

Closed-form synthesis is a starting point, not a guarantee. Send us a design and we will run it through a full-wave EM solve and come back with the S-parameters and an engineer's read on what they mean.

Ask for an EM verification →

These get built in Australia.

2P Innovation designs, fabricates and assembles RF hardware locally, to Class 3, fully documented.

About 2P Innovation →