How to Design microstrip array antenna at 110GHz with Taylor distribution?

[uri1636]

1. For a PhD study, I am designing a serial array of microstrip antenna at a frequency of 110 GHz with Taylor distribution to reduce the side lobes to -30 dB.
2. In the first stage, I designed a set of 16 lines, with 20 elements in Taylor distribution, each with an input port and the simulation software (CST) summarizing the effects to one result, so I got the gain of 20.3dB.
3. In the second stage, I designed a Wilkinson 1:16 power divider to the frequency of 110 GHz to divide the main input into 16 branches to connect to the 16-line serial array and got an appropriate band width around 110 GHz.
4. In the third stage, I combined the two systems, and although each one works independently, the desired 16 dB gain around is not obtained after the reductions from the power divider, However, an excellent bandwidth of around 110 GHz is obtained.
5. How can I solve this problem?

 

[Georgy.Moshkin]

What is S21 value for power divider at 110GHz?
Are you sure antenna sub-array and dividers are matched regardless of line lengths at all ports?

 

[uri1636]

The S21 is -25dB.
The ouputs of the wilikinson 1:16 divider are 50 Ohm lines and the inputs of the serial microstrip antenna lines are 50 Ohm so they matched.
When I design 120 GHz serial microstrip array antenna with equal distribution at the same technique, I get working antenna without any problem(only the gain was reduced by 5dB).
Can you help me???

 

[Georgy.Moshkin]

I think by -25dB you meant S11. What are losses in your Wilkinson power divider (S21, S31)?
Also you may plot electrical field distribution for your array, for example in metal plane or 0.1mm above it. It may help to see the problem (lost metal connection, air instead of dielectric, out-of phase in some sub-arrays which cancel radiation instead of combining,etc.)

 

[uri1636]

The losses are 25 dB as my previous design at 120GHz that worked(equal distribution) without problem.
Do you want to see my design?

 

[Georgy.Moshkin]

You may post a picture here. What do you mean by losses are 25dB? It's a 99% loss.
First you need to check how good your Wilkinson divider is, without using an antenna.

 

[uri1636]

https://imgur.com/LmPOEAO
Please see the above picture

 

[Georgy.Moshkin]

s21=-20dB is too low. S21 is usually a little less than -3dB for 2 port power divider.

 

[Georgy.Moshkin]

It turns out that s21=-20 db is for 16-way power divider.
Single power divider S21=-5dB, signal goes through 4 cascaded dividers resulting in -5dB-5dB-5dB-5dB=-20dB at output.
Solution is to optimize single power divider for better s21, or use another power dividing network configuration with less losses.

 

[volker@muehlhaus]

I guess this is an on-chip design, where lines are quite lossy. Did the simulation include proper losses of metal and substrate, including lossy silicon substrate?

"excellent bandwidth of around 110 GHz is obtained" sounds like your power is dissipated rather than radiated.

 

[FvM]

It's relative easy to calculate the losses of an equal 1:16 power divider. Ideal s21 is -12 dB, so -20 dB means 8 dB loss, -25 dB even 13 db loss.

But how is the feed network of your "taylor distributed" array designed and how do you calculate losses? I would expect an (ideally lossless) power divider with unequal sx1 pattern. If you have equal sx1 followed by attenuators, the overall gain will be of course bad.

 

[Georgy.Moshkin]

uri1636 sent me images of design, not sure if i can post it here.
Design is similar to https://www.instockwireless.com/16way-power-splitter-combiner-divider-app.htm
Currently it is equal power division. Single power divider S21 is -5dB, so output is -5dB x 4 = -20dB
For better results single power divider building block must be optimized for better S21.