Help with Switched Line Phase Shifters

[Daniel Santos]

Hello to all,

I need help to find a switched line phase shifter high attenuation problem.

I designed a phase shifter with the Switched Line topology. The requirements are:
- Center frequency @ 2.42 GHz;
- 4-bits: 22.5; 45; 90 and 180 degrees;
- low insertion loss and high isolation;
- Built with microstrip line and SPDT switches from HITTIE (model HMC545E);
- Use of the RF Module from COMSOL for simulation;

I have already designed the complete microstrip line (50 Ohm on duroid 5880 Er=2.2) on COMSOL, so I moved on to implementation. 1st stage of Implementation I have made 1 bit with 90 degrees for a test and it works fine (DUT in Vector Network Analyzer from Rohde & Schwarz model ZB4):
- Insertion Loss: 0.8 dB for Ref Line and 1 dB for delay line;
- S11: < -12 dB;
- Difference of phase: 87 degrees between lines;
- Others single bit phase shifters worked fine (22.5, 45 and 180º).

The problem description:
When I build 2 bits together (22.5 and 45 degrees) does not work:
- Insertion Loss in best case is - 4.2 dB and in worst case -8.3 dB;
- S11: in best case is - 12 dB and in worst case is -3 dB;
- Phase Difference between lines is bad for some cases and correct for another’s;

When similar circuits with the same SPDT switches are said to work fine in several papers. The only difference is in microstrip line, the paper use CPW line.
So I need help to find my problem, unfortunately the COMSOL does not have active components libraries like SPDT or pin diodes so I could simulate. I don’t have many background in the RF and Microwave and all I learn until now, was alone reading some books and IEEE articles.

I think my problem is reflections due to some bad adaptation between lines and SPDTs??

Here is the picture of my design, where do you think I did a mistake?

https://obrazki.elektroda.pl/2784486900_1408963979.png

 

[vfone]

Your thought is right. Most probably there is a reflection and mismatch between lines and switches.
Into the simulator, try to separate the stages and check what input/output impedance see each stage separate.
One important thing is that in microstrip design, a smaller strip width leads to higher losses.

 

[biff44]

I am not sure how to say this sounding very technical, but your lines are too fat. They have a poor aspect ratio. microwaves want to go down transmission lines that are long and narrow. When you try to send them down fat and short lines, almost square in shape...you get all sorts of modes of propagation happening, including evanescent modes that have significant reactive effects that CHANGE due to load.

In other words, go to a much thinner substrate material, the widths will shrink up, and it will start to work.

One word of warning, you might need a non-reflective switch for it to work in some of the bits. That is because in the longer bits, if you terminate one arm in SP2T switch ports in isolation, you get full reflection from the switch. If the length of the line is "just right" at some frequency that "turned OFF" arm will resonate, and cause a big insertion loss blip in the arm that is "turned ON"