Modelling of nonuniform magnetic field effects on partially magnetised Y-junction circulator

The success of modelling partially magnetised linear 3-port junction circulator, which is non-uniformly biased, depends chiefly on the precise representation of the strength of the intrinsic magnetic field in the axial direction across the ferrite material placed at the centre of the junction. The inadequateness of using conventional Polder tensor, which is established with the assumption that the ferrite material is saturated by a constant uniform magnetic field, in modelling partially and non-uniformly magnetised 3-port ferrite junction circulators is presented. It is proposed that the combination of static and microwave finite element (FE) solver overcomes the limit of Polder tensor in modelling partially and non-uniformly magnetised ferrite junction circulator. The proposed combination method requires the generation of the steady magnetic field in a magnetostatic solver, where the magnetic state of the ferrite material is determined, and the information of the static magnetic field of ferrite junction is transferred to the microwave solver for use in the microwave permeability tensor of the partially magnetised ferrite junction. Using the coupling method, the reflection coefficient responses of the 3-port ferrite junction circulator are determined in the low field loss region which coincides with the partially magnetised region and accompanies a few undesirable problems in communication systems, such as the reduction of bandwidth, and intermodulation product interference. The determined reflection coefficient is used to obtain the split frequencies which are important in designing and analysing the performance of a 3-port ferrite circulator in partially magnetised region. The main modelling factors considered are the effects from the number of mesh, the type of the magnetic field, the biasing configurations and the value of the dielectric constant. In each case, various split frequencies are obtained and, by considering all factors step by step, optimal modelling conditions are decided. The modelling results are compared with the experiment and show good agreement