How to find characteristic impedance of microstrip with an unknown substrate?

Hi all,

I probably have a simple question.

I have to design a circuit with a multilayer substrate which i cannot know his real permitivity nor other parameters but dimensions, I can just use it to my simulations and momentum.

But i need to make a 50 Ohms / 90o microstrip line. After doing the simulations and obtaining the S parameters, how could I obtain Z0 just with S parameters?

I put a 50 Ohm load at the line and made the Z0 = ZL (1-S(1,1))/(1+(S(1,1)) relation with a known line and substrate (W and L for 50 Ohm and 90o), but I can't obtain feasible results (Z0 = 0.2 < 80o)

Thanks ! :)

Hi,

If you have to design it then:
* either you need to know it
* or you need to specify it

--> Discuss this with your PCB manufacturer

Klaus

I may explained wrong.

I get the substrate specifications, but is multilayer, so I have differents permitivities for each layer but its behaviour is different all together. I did the momentum simulation of a kind of 50 Ohm / 90o degrees line (based on a medium of all layers permitivities, but it isnt the real permitivity) and obtained the S Parameters.

Now i want to know if its really 50 Ohms / 90o degrees, with the objetive of knowing how far of it is it and how to correct it later.

In conclusion, know the Zc and electric length of a microstrip line based on just the S parameters.

Thanks anyway!

Convert S={S11, S12 ; S21, S22} to F={A, B ; C, D}.

You can get Z0=sqrt(B/C).

2*pi*Length/Lambda=acos(A)=acos(D)

Nice! It worked like a charm. Now i think i could work with any substrate if i cant know its real permitivity.

Thanks! :)

What i do...i take my simulation of the thru transmission line, plot S11 on a smith chart and sweep the frequency from say 1 to 10 GHz. It will draw a small circle on the smith chart, and the CENTER of that circle is the characteristic impedance of the line.

So this is showing a 40 ohm characteristic impedance line (assuming your simulation ports were 50 ohms).

Is that what you were asking?

Correction.

It should be :
2*pi*Length/Lambda=imag(acosh(A))=imag(acosh(D))