Filter design using CAD. I. Linear circuit simulation

Due to chronic spectral congestion in the lower reaches of the RF spectrum and the envisaged demand for future high bit rate services, a number of wireless communication systems have been developed or proposed for the 5-20 GHz frequency range. For transceivers operating below 3 GHz, surface mount lumped element components or coaxial resonator structures are typically employed to realise the RF front end filtering requirements, due to size and cost considerations. However, for systems operating at these new, higher frequencies, distributed filter structures implemented in stripline geometries become attractive and this is likely to drive a resurgence of interest in distributed filter technology. Design methodologies for distributed element bandpass filters are well established and widely available. A linear circuit simulator CAD package is typically used to verify the design and characterise second order effects such as impedance discontinuities and field fringing. In addition, such packages usually include performance optimisation tools, allowing the designer to perform a rough hand calculation and then use the CAD package to centre the design. However, attempting to optimise a distributed element filter using an ideal filter for the performance goal can sometimes produce undesirable implementation parameters. The distributed filter is composed of inherently periodic structures that possess line loss and dispersion that fundamentally limits the "fit" of the characteristic to that of the ideal. This paper discusses optimisation strategies based on a knowledge of the specific limitations of the distributed filter, using two case studies for illustration. It argues that, in many cases, the optimiser is only required to absorb second order effects and merely centre the design frequency, rather than be used to perform "blind" optimisation.