Improved Trmable Ferrimagnetic Filter Using a Magic-T

It is well known by workers in the field that a disadvantage of tunable ferrimagnetic bandpass filters is the necessity for large magnetic fields in order to achieve resonance at the higher microwave frequencies (above 20 GHz). Large magnetic fields require heavy, bulky magnets and high tuning powers. The hgh dc field requirement arises principally' as a result of the "stacked" orthogonal waveguide configuration, which had previously been necessary to achieve high isolations outside the passband. A geometry of the conventional type is shown in Fig. 1, where it is seen that the external dc tuning field must bridge the heights of two stacked waveguides. In an effort to ameliorate the high field problem, Carter' has proposed the configuration of Fig. 2, which achieves isolation by means of a cutoff section of waveguide between input and output lines. Ths scheme, however, has several disadvantages. It requires a minimum of two ferrimagnetic resonators, tends to introduce magnetostatic modes as a result of asymmetries, and imposes a tradeoff between maximum values of sphere-to-sphere coupling and off-band isolation. Additional dii€iculties are caused by the presence of the slot, which, as Carter has pointed out, decreases coupling from the external circuit to the spheres and reduces the unloaded Q factors of the resonators. The proposed geometry, shown in Fig. 3, alleviates these disadvantages while still retaining the advantages of a nonstacked configuration. The operation is explained as follows. As is well known, if the collinear arms of a magic-T are short-circuited at equal distances from the center of the junction, and an input signal is applied to the sum arm, practically no signal appears in the difference arm. This isolation is dependent solely.on mechanical symmetry and, typically, may be greater than 30 dB. Now if a resonant, overcoupled ferrite sphere is placed directly in front of one of the collinear arm short circuits, the phase of reflection will be changed to that of an open circuit. If there were no losses, and the junction were perfect, complete transmission would result from sum to difference ports. For the lossy case, the actual transmission may be derived as follows. Referring to Fig. 4, let