shack-1166
Polyphase networks for phased SSB equipment
Many homebuilders have gone to phased principle in their SSB rigs and CW too. I have not done so yet but I think I'm on the way...
The most critical component of such SSB radio is the audio phase shifter. So called Polyphase network has become "the industry standard".
I did not want to just copy other peoples drawings but wanted to understand a little bit better, how good they can be
and how sensitive they are. I found in the web at least three good sources of information. Certainly there are many more. 
PA2PIM
Niess
JA1KO
I played with my modeling tool and finally was able to get out performance curves over frequency band 100Hz-10 kHz like:
Voltage loss
Phase shift / channel
Phase difference between the channels, deviation from ideal 90 deg
Sideband suppression
I used examples from literature as starting point and continued trimming them according to my preferences. 
Trial and error was my method in finding the best component values.
I have chosen four examples here to show what I did. Schematics are shown as in transmitter.
1. SSB network where all capacitors are equal
2. SSB network, where all resistors are equal
3. SSB network, where voltage loss is zero, standard resitor values
4. SSB network, where voltage loss is zero, standard capacitor values added on 18.3.2009
4. CW network, where voltage loss is zero
Lessons I learned were:
Voltage loss depends on direction of signal in the network. There can be 10dB difference, in zero loss model more than 35dB.
Equal capacitor models shall have low frequency components on output side of the network
Equal resistor models shall have high frequency components on output side
Zero voltage loss models shall have high frequency components on output side
Equal capacitor/resistor models (6 stages) have 9.5dB voltage loss on their center frequency, 1.4dB less on both ends.
Input impedance of the following amplifier shall be very high, in my case > 5Mohms, which is not a problem however.
Absolute component values are not very critical but in each vertical column they shall be almost equal.
absolute values can deviate a few %,
but inside columns 0.1% cause about 5 dB worsening of sideband suppression and  1% almost 20dB.
For SSB suitable frequency range is 300-3000Hz so that suppression maximums are on those band limits.
It is reasonable to make audio filters which limit enough frequencies outside the band. 
Pekka, OH1TV 1.3.2009
Back to projects page