Defining waveport for wilkinson power divider

[omena]

Hello!
I'm trying to design a Wilkinson power divider for corporate antenna array feed network using microstrip technology in HFSS for 2.4GHz. I'm using FR4 epoxy (epsilon=4.4, tan delta = 0.002). I have modeled the structure but have a problem with defining the port. The results from my simulations are inconsistent and the isolation (S23) is only -5dB. For the isolation resistor, I've tried both Lumped element BC and impedance BC. Can someone help me on how to model the ports. The help manual proposes port width of 10W which is impractical in my case since the trace width is 2.8mm and substrate height is 1.60mm.

 

[eraste]

Hi,
What simulator do you use ?
If it is ADS or Ansoft designer , no problem the port are easy to define.
If you use HFSS, a lumped port is advised.

Bye.

 

[omena]

I am using HFSS version 11. I have tried lumped port but the resonant freq for port 1 is highly shifted upwards from that of port 2&3: The isolation between ports 2&3 still didn't improve. I have tried doing the same simulation in CST MWS using discreet ports but still the same problem. I really can't understand the source since the model seems to be correct

 

[eraste]

Hi,

One comment : the loss of your FR4 substrate seems to be to low (0.02 for FR4 is more common).
Second comment : at 2.4 GHz and if your impedance is 50 ohms the width of the line must be 3 mm.

And finally : Could you send me your HFSS project ?

By for now...

 

[omena]

The first comment is right, It was a typo error the loss tangent is 0.02. The width of the 50ohm trace is 3mm and that of the 70.71 is 1.5mm, according to my calculations. I'll send you the project when I get to the office. Thank you in advance for the assistance.

 

[omena]

@eraste: I tried to upload the file but there is an error message of an invalid file. So I don't know what to do next

- - - Updated - - -

ok now I got it. It's the file format which was wrong. Now here it is in zip firmat

 

[eraste]

Hello,

I had a look to your design.
Thirst of all, it seems to me that your "V" pattern is not adequate.You must have coupling near the junction of the "V". I suggest you a pattern like this :

https://fr.wikipedia.org/wiki/Diviseur_de_Wilkinson

Plus, you have to optimize the Lambda/4 branch and their width because it appears that the frequency you obtain is above 2.4 GHz so branch are too small.
The values you calculate are nominals values which are a good start point but don't account for the geometry you use.

About the lumped port, it looks strange to put it in the FR4. In fact your lumped port have to represent the in/out energy (field) and suppose that field varies a little bit in the surface (rectangular shape) defining the lumped port.

Bye for now.