Question About Microstrip T Power Divider

[chenchangfu]

I want to design a microstrip antenna array.So I need a power devider.I referred the book Microwave Engineering,and designed a microstrip T power devider,you know,just the 50ohm to 100ohm.But I use the HFSS to simulation and found that S21 and S31 are always in the near of -5dB.I do not know why.Can some people tell me why?By the way the center frequency is 12.5 GHz.

 

[vfone]

Here are more ideas how to do it:

 

[chenchangfu]

Thank you,but can you tell me detailly how to choose the W1 and W2 in the picture?

 

[tony_lth]

W1=W2 is the 50 ohms .
You can simulate it.

 

[chenchangfu]

OK,I will have a try!

---------- Post added at 06:02 ---------- Previous post was at 05:21 ----------

I can not understand if W1=W2 is the 50ohms how the devider can realize equal power distribution?Do not need impedance match?

 

[sv1437]

W2 = 50 ohms and W1 = 100 ohms (twice the impedance of the feedline) for a T-junction power divider.

-sv

 

[volker_muehlhaus]

I can not understand if W1=W2 is the 50ohms how the devider can realize equal power distribution?Do not need impedance match?

Yes, you also need impedance matching using 1/4 lambda line.

You can choose if you want to
-connect the two outputs first (50||50=25) and transform that 25 Ohm to 50 Ohm with a line of 50/SQRT(2) or
-transform each output individually from 50 Ohm to 100 Ohm with lines of 50*SQRT(2) and then have 100||100 = 50 Ohm at the T-junction for the parallel connection.

In post #1 you asked why this is not matched from the reverse side, i.e. looking from the output. You can easily understand that if you look at the resulting parallel connections and impedance transformations.

 

[Azulykit]

You might find this site useful:
**broken link removed**

I would also recommend running down a copy of Pozar's, Microwave Engineering down as he discusses dividers in some detail.

You need to take care of small details if you are attempting to work in X band. Start by making sure your transmission lines are working and that you have the ports modeled correctly. Then start working on the t-junction. There are numerous references on the subject available. HFSS is an excellent tool to explore this question. I would build a parametric model to make tuning less of a chore.

One additional comment, HFSS is an excellent analysis tool. You need to figure out want you want to synthesize.

 

[chenchangfu]

Thank you for your reply.Now I could not exactly calculate the widths of the transmission lines with 50ohms and 100ohms.The tool APPCAD is used to calculate them.But I compared the result with another tool and I found that they are different.I don not know how to choose.Then I build mode in the HFSS,use the lumped port as excitation ports and the solution model is Driven Terminal.Can you tell me what other details should I take care of?By the way,I want to know how to determine the length of the transmission line with 100onms?A quarter wavelength or no restrictions?

 

[sv1437]

Hi chenchangfu,

Firstly, why did you use driven terminal simulation in HFSS rather than "Driven Modal" and What are the properties of the substrate ( i.e., what are the dielectric constant and height of the substrate) ?

-sv

 

[chenchangfu]

I use the driven terminal as solution type because I read some examples and all of them chose the type.At first I also did not know the reason,then I searched related articles including the help of HFSS.They gave the answer that Driven terminal is more suitable for the question about power devider when we want to consider the coupling between port2 and port3.
The properties of the substrate is just FR4.The dielectric constant is 4.4 and height of the substrate is 0.8mm.

 

[sv1437]

FR4 substrate for frequencies more than 4GHz is not a right choice...and your center frequency is 12.5 GHz. I would rather prefer Rogers substrate for such high frequencies.

 

[chenchangfu]

I know FR4 is not suitable for high frequency,but now I just have FR4 at my hand.So I wanted to simulate it and have a try then I found that the the insertion loss was very high.I will take your advice and use the Rogers 5880 as the substrate.Thanks again.