QEX

sideration by other do-it-yourselfers.

Description

The basic design is configured for an output IF of 148 MHz. As indicated in the block diagram (see Fig 1), however, an optional filter (FL1) results in image discrimination of at least -20 dB for any IF between 21 and 148 MHz. This filter is a high-pass waveguide type that takes advantage of the excellent natural waveguide cut-off characteristic. One disadvantage is that the waveguide width determines the cutoff frequency, essentially dictating a nonstandard waveguide, in other words, a home-brew assembly. My filter is a 4x11 inch PCB base with a U-shaped 1V4x2'/2x10 inch aluminum cover. Its fabrication is not difficult, but it requires aluminum bending techniques that produce a uniform waveguide width. To accomplish this, I use a simple home-brew bending brake made from hardwood. The filter insertion loss is about 0.25 dB.

Fig 2 is a schematic of the mixer board. The input filter (FL2, a parallel-wire 2400 MHz filter) provides image discrimination of more than 20 dB when using a 148 MHz IF, even without the optional filter. Fig 3 shows the filter details. The wire lengths are A/4, except as adjusted for minor reactance effects from the load or from errors in the wire position. (When it's not exactly parallel to the ground plane.) Peaking for resonance is accomplished by simply bending the wire up or down. Coupling can be optimized by bending the wires sideways while maintaining a parallel wire condition. This changes the wire spacing slightly. Don't let this bending procedure scare you; it's really a very friendly and effective adjustment method.

The filter is, in effect, a resonant matching transformer. In my case, the matching impedances are not particularly accurate. Inserting the filter between a 50 Q reference and a source often results in an insertion gain, rather than a loss.

The SBM rat-race mixer (Fig 4) is made from 0.1-inch-wide PCB strips glued to a substrate. The two barrier diodes interface with the post amplifier through home-brew feedthrough capacitors. Fig 5 shows the capacitor details. Initial tests indicated a SBM balance of about -30 dB. This can be improved to better than -40 dB by simply sliding a small capacitor, referenced to the PCB common, along the X/2 wave section (Fig 4). The capacitor is made from a strip of 0.005xV4-inch brass shim stock bent in an S shape 3/i6x'/i6X3/i6 inches. The dielectric is a piece of Scotch tape on the underside of the end that slides against the XI2 section. This sliding capacitor permits the mixer balance to be nulled to zero. Mismatch at either the signal-input or LO ports will effect the balance.

Lengths of resonant striplines are usually determined by a simple dielectric-constant relationship that assumes pure TEM propagation. Although other conditions have relatively minor effects on this purity, a glued-down stripline is a major consideration. The propagation constant is 0.61, about 30% larger than that for TEM propagation. This number was determined by measuring the propagation along an 8x0.1 inch glued down stripline and recording the XJ2 nodal points along the line. Tests with both 1 and 2.4 GHz sources produced identical results.

The post amplifier uses a common-gate FET amplifier having a relatively broadband input and a fairly high-Q tuned output circuit. This low-frequency input is interfaced with the mixer through 2.4 GHz chokes connected to the mixer-diode feedthrough capacitors. A MAR-6 low-noise MMIC amplifier follows the FET first stage to make a total gain of 28 dB, enough to overcome a poor noise figure in the following IF system. Two inductors, L5 and L6, are tailored to the IF selected to follow the amplifier, 15 to 148 MHz.

Fig 6 shows a schematic of the LO circuit board. It starts with a 44 MHz VXO. I used this simply because of an available junk-box crystal that required some pulling. This VXO is a good performer if we meet two conditions. First, minimize stray capacitance to ground at the crystal/inductor junction. Second, provide a reliable start for this high-Q circuit. (Simply switching on the power usually results in a dormant oscillator.) The condition can be easily corrected by using a low-pass time constant between the supply voltage and the oscillator. The following buffer/tripler 132 MHz output is filtered by FL3, a T-type two-inductor/single-variable-capacitor filter. The following diode (D3) picks off the 395 MHz harmonic. This is filtered by FL4, which is similar to FL3 except for the component values. The output is amplified 18 dB by the MAR-1 that turns on the diode D4. The 1185 MHz third harmonic is filtered by FL5, one of the wire filters shown in Fig 3. The final doubler is on the mixer circuit board shown in Fig 2. The 1185 MHz output overdrives the MAR-l/MAR-4 pair, enhancing the harmonic content through the MAR-4. This produces the 2369 MHz LO output through D5. L24, a resonant Xl\ wire, provides D5 with the necessary high-frequency matching interface. The LO output wire filter L24, L25, appears in Fig 3.

Photo A—FL1 optional filter.

2400 MHz

0 0

Post a comment