Lumped HPF
A lumped 2 GHz elliptic highpass filter is designed on a 20 mil thickness RO4003C substrate. The design goals are the following:
| Parameter | Value | Band |
| Insertion loss | < 1 dB | [2, 4] GHz |
| Stopband rejection | > 30 dB | < 1.5 GHz |
| Return loss | < -15 dB | [0, 2] GHz |
An order 2 elliptic filter (2 resonators) satisfies the requirements above. The filter designer tool from this website is used for the first steps. Then Sonnet Lite, Qucs-S, and Kicad are used.
Ideal 2 GHz elliptic HPF Circuit
Once having obtained the filter values from the filter designer tool, the filter is built with ideal components and simulated using Qucs-S.
The next step is to synthesize the inductors with PCB traces. A RO4003C substrate will be used. As a first approach the MLIN component with the RO4003C substrate properties is used for estimating the length of the traces. A width of 200 μm to maximize the length of the trace. Once the estimation of the inductor length are obtained, the inductors are laid out into Kicad with a meander shape in order to have a compact design.
Once the meander are simulated in Sonnet Lite, it is convenient to compare their performance with the ideal inductors. For doing this, the S-parameter data obtained from Sonnet Lite are exported into a Touchstone file and then compared those of an ideal inductor in Qucs-S.
Both inductors have an inductance sufficiently close to the ideal one. The Q factor of both inductors at the attenuation poles frequency are good enough to make decent resonators.
As the meander inductors are no longer ideal inductances, it is need to readjust the filter capacitors so as the elliptic shape is preserved. Also it is needed to replace the ideal capacitors by commercial SMD parts. In this sense, Murata SMD ceramic capacitors will be used. Murata Simsurfing is a nice tool and allows the download of S2P files that can be used in Qucs-S.
Sonnet Lite does not allow to include ports inside the simulation box, just only on the box edges. Consequently, having SMD capacitors in the design imply that the filter components must be simulated as independent unit. Unfortunately, this approach misses EM couplings between the inductors and other traces, but the result should be close enough for a 2 GHz HPF. Proceeding this way, the feed lines, and GND via holes are simulated separatedly and then after exporting the S-parameters, simulated all together inside Qucs-S.
Layout
