Study on microstrip filter design with emphasis on surface roughness, anisotropy and simulation

I try to present reasons for deviations between micro strip filter simulations and measurements in this thesis. A short introduction to the printed circuit board manufacturing process is given to explain problems associated with it. I present a comparison of simulation results between some published micro strip models for a single micro strip line and parallel coupled lines. I identified attenuation factors of the micro strip, and I simulated the effect of the surface roughness of the conductors on the insertion loss of the micro strip filters. The latest model, Huray's snowball model, tries to model the surface roughness as stacked snowball piles. However, this model is complicated and requires knowledge of the complex relative permittivity over the whole frequency range of interest. I also discussed losses due to the surface finish. I reviewed different methods for measuring the relative permittivity of the micro strip substrate, and implemented and tested the cavity resonator method. Anisotropy of the permittivity exists in many of the commercially available microwave printed circuit board laminates. I presented design of three different microwave micro strip filter realizations (edge coupled, hairpin, and stub) on different frequency bands. I have simulated the relative permittivity which would match the measured pass band centre frequency. This value differs for each filter, and therefore I also simulated anisotropy of the substrate. I found that anisotropy shifts the pass band to lower frequencies and matches out of band response better compared to increasing the relative permittivity alone. I simulated the effect of the nickel layer on top of copper conductors and found that this shifts the centre frequency nearly as much as anisotropy of the permittivity. I iterated values of the surface roughness model for one of the edge-coupled filter designs to match the measured losses. However, the same values, when I used them for other filters, did not explain the measured higher losses whereas simulation with a nickel layer seemed to increase losses.