Fig 2—k/A transmission-line phase shifter: A) circuit; B) LO Drive current; C) RF output voltage.

Comparison of Three Phase-Shift Networks

For the next analyses we will compare three different 90° phase-shift networks, a X/4 transmission line, a pi network, and a tee network. At their design frequency, these networks behave identically; if you put them in a black box you couldn't tell them apart—if you can only test them at the design frequency. Each of the networks provides exactly 90° of phase shift at the design frequency, when driving 50 Q.

Fig 2A shows the circuit for the ^./4-transmission-line phase shifter. Fig 2B shows the LO drive current and Fig 2C shows the RF port voltage. As we would expect, the results are exactly like those of Fig 1 except for the phase shift, which is exactly 90°. Great performance; it's totally insensitive to LO drive level, but it is difficult to adjust the phase shift. (Later, I will show that you can adjust for RF

phase errors in the audio portion of an SSB rig.) Even without the ability to trim the phase shift, this is a good circuit when you need only modest unwanted-sideband rejection, and you are working at a frequency where the line length is practical.

Fig 3 is just like Fig 2, except that it uses a pi-network phase shifter. Note that the drive-current waveform is narrower and more peaked than in the previous case. The RF output waveform reflects the narrower conduction time. The RF output phase-shift error is +8.7°, and the phase change with drive level is large enough to be visible in the waveforms. The RF output level is -6.5% compared to that with resistive LO drive. The output capacitor of the pi network increases the drive voltage rise time during that portion of the cycle when the diodes are not conducting.

This makes the pi-network phase shifter the most sensitive, of the three, to LO level.

Fig 4 is just like Figs 2 and 3, except that it uses a tee-network phase shifter. Here the LO drive-current waveform is wider and less peaked than in the other two cases. The RF waveform is missing the nonconducting portions seen in the previous two cases; it is almost a square wave. The phase-shift error is -4.2°, and the RF output level is +3.5%. It's most interesting that the output phase shift is almost as independent of LO drive level as it was with the A/4 line. Note that the tee network must be physically close to the DBM to minimize shunt capacitance [from the tee-output-to-mixer-input node to ground—Ed}.

Fig 5 compares the three networks for their change in amplitude and phase with an LO drive level change of ±10%. It is clear that the transmission-line network is the best, the tee network is next and the pi network the poorest.

Hybrid Power Splitters

To make a real SSB rig, we must feed one of the phase shifters described via a hybrid power splitter, to minimize interaction between the two DBMs. Fig 6A shows a suitable hybrid using just one core. This hybrid has three ports referenced to ground and one that floats. Ports B and C have equal impedances, while port A has half the impedance of ports B and C, and port D has twice the impedance of ports B and C. This hybrid is commercially available with a second core containing an autotransformer to step up the impedance of port A, as shown in Fig 6B. This is a broadband hybrid,

0 0

Post a comment