Branchline Coupler????????

[haydar]

I need to design branchline coupler for 63.87 MHz. Lambda/4 becomes 117.4 cm. by using dielectric substrate with epsilonr equals to 4.7, lambda/4 is 54.16 cm.
I fabricated a coupler with this value. however I get isolation equals to 0.3 and other S parameters are not as I expected.
Is there anybody who constructed a coupler before and knows how to design and what to do to obtain best results?
I have read some posts about coupler but my design is at low frequency and I want to learn design hints.
thanks.

PS:can you suggest some companies to buy coupler?

 

[flatulent]

Are you meandering the lines or are they straight? There may be some coupling if they meander.

You can also use coaxial transmission lines. for the 35 ohm sections, put two 70 ohm lines in parallel.


You can also use LC networks instead of the transmission lines and get a much smaller physical size.

 

[BigBoss]

I need to design branchline coupler for 63.87 MHz. Lambda/4 becomes 117.4 cm. by using dielectric substrate with epsilonr equals to 4.7, lambda/4 is 54.16 cm.
I fabricated a coupler with this value. however I get isolation equals to 0.3 and other S parameters are not as I expected.
Is there anybody who constructed a coupler before and knows how to design and what to do to obtain best results?
I have read some posts about coupler but my design is at low frequency and I want to learn design hints.
thanks.

PS:can you suggest some companies to buy coupler?

I recommend you to build this by using double aperture ferrite cores.In that case, transmission lines will be twisted copper lines with proper length and wire diameter. Or if you have money, buy ready to use from Epcos(formerly Siemens+Matsushita).

Look at that link.

**broken link removed**

 

[radiohead]

I would say: go for ferrite coupler because the design will be much more compact. Toko also some couplers. Watch this link

**broken link removed**

 

[haydar]

Flatulent:
yes, I must meander the lines, because 54.16 cm is too long. you are right, when I draw lines as straight, result changes.
can you explain this LC network subject, or give site address that explains?

BigBoss:
I cannot use ferrite core, it must be non-magnetic. I will check the site.

radiohead:
the same, must be non-magnetic, I will check the site. I love Pink Floyd, radiohead cannot be alternative of PF.

 

[VSWR]

Can you post a picture or drawing of your coupler so we can take a look?

 

[flatulent]

I cannot find the sources I remember. The lines are pi LC single section networks. The result is a square of inductors with capacitors to ground at the corners. Try this in your simulator. The reactances of each part is the same as half of the the desired line impedance at your operating frequency. (That is for each isolated pi section. The capacitors at the corner are the sum of the two end capacitors of the individual pi sections connected together at that node.)

 

[haydar]

this is the design. but gives error.

 

[flatulent]

Here is an old Ham Radio Magazine article which contains the LC version.

 

[VSWR]

Can you post a layout drawing with dimensions, instead of a HFSS design file?

this is the design. but gives error.

 

[haydar]

this is jpeg file

 

[VSWR]

Looking on the layout I suspect RF coupling between the meander sections of the transmission lines. Can you separate the transmission lines or are you bound to the outer dimensions of the substrate board?

If the latter, I suggest you use coiled coaxial cables as transmission lines instead or go for discrete LC design. See above postings.

 

[haydar]

ok, let's think I can use coax cables. how can I obtain Zo/sqrt(2)
impedance?

 

[flatulent]

ok, let's think I can use coax cables. how can I obtain Zo/sqrt(2)
impedance?

You can put transmission lines in parallel to get new impedances. Two 70 ohm lines in parallel make 35 ohms. In parallel means that two equal length lines have at each end the outer conductors connected together and the inner conductors connected together.

One advantage of coax is that you can coil them to make the physical size of the circuit smaller.

 

[VSWR]

Here is the result using paralleled 75 ohm RG59 and 53 ohm RG58 cables. Just as an example. You have to replace them by cables with smaller outer diameter in order to fit on your board.

 

[haydar]

Here is the result using paralleled 75 ohm RG59 and 53 ohm RG58 cables. Just as an example. You have to replace them by cables with smaller outer diameter in order to fit on your board.

VSWR
thank you very much for your close interest. this forum is addictive, I can't help reading all time.
anyways, I will ask a question: S13 is very low, ok, but values of S12 and S14 should be equal to 0.5 (linear magnitude) for my application.
In HFSS design, I get S12 and S14 about 0.65 and 0.75 (linear magnitude, not dB). what is the problem?
thanks.

 

[flatulent]

Conservation of energy still holds. The sum of the power going to the three passively terminated ports should equal the input power. Do your calculations show this? One experiment is to do four simulations using each port in turn as the input.

 

[haydar]

flatulent,

you are right, but I built a coupler (it does not work properly), and it gives below results: (ports 1 and 4 are inputs)
S14=0.297 (should be 0)
S12=0.490 (should be 0.5)
S13=0.660 (should be 0.5)
Phase of S12=-108 (should be -90)
Phase of S13=-190 (should be -180)

this does not hold the conservation. can you comment on these results? interestingly, hfss simulation gives the same results.

 

[flatulent]

My guess is that the coupling between sections caused by the meandering is causing the impedance and/or the electrical lengths of the lines to be different than what you want.

Try this simulation. Extract just one of the lines with the present physical shape. Put a passive termination on it and meassure the input impedance and phase shift at your operating frequency. From this you can back out the line impedance and electrical length.

Do this for both impedance lines.

Then do some cut and try with the physical length and trace width to get the measurements to quarter wave and the desired impedance.

Then change your complete circuit to have these trace dimensions and simulate again.

 

[flyhigh]

Hi,

yes, this forum is really adictive. I think one should be amazed how many literature and approaches are developed for such a fundamental (and often regarded as very simple) microwave component

Here are some of mine contributions that uses both lumped and distributed elements to build a coupler. Also some equations are given.

flyhigh