quarter wavelength transformer

hi, everyone.

I am doing a printed circuit design in 60GHz range.

The basic structure is 2-dielectric substrate tapes with 3 copper layers on top/in between/ on bottom. The total thickness is 203um+406um=609um, and with DK=3.55.


Structure:

copper
-----------------
Rogers 4003
-----------------
copper
-----------------
Rogers 4003
-----------------
copper (GND)



The broad-side coupled strip line is going to be used as transmission line, printed on the top and middle. The characteristic impedance is obtained by IE3D modua -> Process -> Find transmission line parameters.

The quarter-wavelength transformer is designed with 100Ohm-70Ohm-50Ohm cascaded transmission line.

However, it seems the simulation is not pretty stable, since the simulation results changed with altering the 50Ohm/100Ohm feeding line length.

I am not quite sure in this range, the higher-order/surface wave mode will affect the simulation and the results.

Has anyone designed this type of structure at such high frequency before? Any experiences or advised will be appreciated. Thanks in advance.

The attachments are the transmission line for 50/70/100Ohm and the quarter-wavelength transformer IE3D geo files.

hi jayleung:
I think there have many problens in your model:
1: port define is error,
2: Modeling of stripline which have finite thickness is error;
3: substrate layer define is error.

to solve these problem , you can refer to manual of Ie3d,usual in the directory C:\Program Files\Zeland\ie3d\manual

Hi, antboy. Can you give more details about what kind of the errors you mentioned for the port
difinition, and substrate layers.

In my structure, the differential excitation (balanced ports) is needed to feed the energe into the top and middle copper layers. And actually, my IE3D version is V11. And it can run the simulation with this geo files and without showing any error messege. The concerns are whether the higher-order modes or surface modes will affect this 3 conductor simulation results.

Hi jayleung,

go thr Chapter-9 of IE3D Manual

Hi, Balaguru thavamani. I am not quite sure which version of IE3D you are using. My IE3D version is 11, and the chapter 9 is "Lumped element extraction and signal integrity". It does include the "Finding transmission line characteristic" and I have already followed the introduction to get the characteristic impedance of the differential broad-side coupled stripline with RF ground.

The problem is that when I use the derived stripline with 50Ohm/70Ohm/100Ohm to build a quarter-wavelength transformer, the results is not stable.

When the feeding 50Ohm/100Ohm line is pretty short, the performance of the 1/4 transformer is OK, but when I made the feeding linea longer, the magnitude of the S-parameters are different.

But according to the microwave network theory, if the characteristic impedance of the feeding transmission line is uniform, it is equivalent to add exp(j*bete*l) to the S-parameters, so the magnitude of the S-parameters should be the same.

So I am wondering whether the "find transmission parameters" function in IE3D has some limitation, or the differential setting of the "advanced extent" ports has some limitation", or physically in such high frequency, the higer-order modes are excited and the microwave network theory cannot be applied.

Thanks.

Hi:

It seems to me the simulator is working fine. Also, the higher order mode is not there for a single uniform TLN. However, the dimensions of your TLN segments are too big. There are some big steps there. Those big steps can't be neglected. You can't not get expected results by neglecting those big steps. Please try to optimize the dimensions to compensate the discontinuities effects from the big steps. Regards.

Thanks, Jian. So do you mean that if I use multi-section impedance transformer or taper transformer, the design will be improved?

And just one more question, why the magnitudes of the S-parameter of different feeding line length have different value? Is it due to the mutual coupling of the feeding line to the transformer or the de-embed technique I used in the differential ports?
Will using a longer feeding line benefit to obtain a stable results?

Thanks.

Hi, Jayleung:

I double checked the structure. I think it is due to multiple modes involved. The 2 metal strip layers are located at 406 um and 609 um. However, you have also defined an infinite ground at z = 0. There are 3 conductors involved. There are at least 2 propagating modes. The transformer should not work as you expected.

best regards,

Thanks, Jian. I reviewed the structure and find that once I reduced the distance between the top and the middle metal strip layers, the transformer worked better.

So I think even though I restricted the ports in this two layers as differential ports, but due to the distances of these two ports to the RF ground are different, so they are not strictly differential, and 2 propagation modes cannot construct the common mode and differential mode as other symmetry structure. And also due to the large distance compared with the operation wavelength, the leakage/radiation of the structure is significant and cannot be neglected.

Hi, Jayleung:

Yes. There are multiple modes involved due to the RF ground which is about 400 um below the traces. It makes the transformer working different from what it is supposed to be without the ground. I think you got it correctly now. Thanks!