Note: The Hittie Mulitplier HMC 578 in combination with a WR-42 waveguide filter, and a MCLI PS8-54 combination successfully supplies the upper LO for the TYPE I and II downconverters.
Wilkinson Two Way Divider Circuit.
Below, Single two-way power divider assembly in a test fixture, 24.8 - 34.3 Ghz,
Shown with 3dB attenuators, short alumina lines, and profiled top cover
Cavity dimensions, cover off: 0.67 wide x 1.02 long x 0.24 deep (inches)
Photo by Bill Saxton
Below, Single two-way power divider in a test fixture, 24.8 - 34.3 Ghz,
Divider only, with profiled top cover. Without attenuators or short alumina microstrip lines
Cavity dimensions, cover off: 0.48 wide x 1.02 long x 0.24 deep (inches)
ShortWilkin_time_cover_K.bmp: Wilkinson Divider CST transient analysis with cover and K connectors.(Note: convergence problems, TS replaces Nichrome Ohmic sheet with PEC).
POWER DIVIDER TEST RESULTS:
Wilen_only_MMlaunch.pdf: Small section with Matt Morgan K-connector launchers: S21 and S11 measure.
Below, Quartz filter test fixture shown with profiled top cover, 24.8 - 34.3 Ghz,
Cavity dimensions, cover off: 0.44 wide x 1.54 long x 0.24 deep (inches)
Photo by Bill Saxton :
Below, Close-up of quartz filter test fixture, 24.8 - 34.3 Ghz, Photo by Bill Saxton:
Cavity dimensions, cover off: 0.44 wide x 1.54 long x 0.24 deep (inches)
Photo by Bill Saxton :
QUARTZ FILTER MODELING RESULTS:
The discrepancy between the CST model and MWO model is primarily the effect of the tapered quartz transition between the 50 Ohm line (25 mils) and the 9.4 mil alumina line. The MWO office model suggests a better S11 and S21 performance than measured. This model did not include the tapered transition with only the 50 Ohm line as input (Model M1). Once the taper is included(Model M2), the match tends to agree with the CST models and the measured results. Several various CST models were executed in an attempt to simulate the measured results. The quartz filter without the K connector beads in both a closed and open cavity, Models C1 and C2, respectively. The quartz filter in an open cavity with K connector beads, Model C3. The quartz filter without the alumina transitions, but with the quartz transitions with K connectors in a closed and open cavity, Model C4 and C5, respectively. And finally a CST model without the quartz and alumina transitions and without the vias in to compare the MWO models and CST models(C6,C7,C8).
CONCLUSION
The quartz CPW transitions did not provide adequate impedance match. The model C6-8 of the scribed and broken substrate indicate an improvement, although the wider bandwidth differs from the MWO model, M3.
CST models predict better than -10 dB return loss with no CPW transition. Either a spark plug K connector transition (Matt Morgan's design) or a straight thru connection should work, although the straight thru is much more dependent upon machining. Both types were modeled giving similar results. Addition of the Corning glass 7070 bead with an air line coax between the bead and microstrip line degrade the return loss by 10 dB.
403arch_nobead.bmp: CST (TS) model of spark plug transition without glass bead.
403arch_bead.bmp: CST (TS) model of spark plug transition with glass bead.
403arch_bead_CPW_10Al.bmp: CST (TS) model of spark plug transition with glass bead, CPW transition, and 10 mil Alumina microstrip line.
More evidence to support poor match from transitions. Measured data from one section with narrow bandwidth design: Note the mismatch at 32 GHz and dip in S21 response.