Inconsistent results between Terminal Wave Port and Terminal Lumped Port when simulating a flat FCC cable ... can someone help me ?

[rdpdo2002]

Hello,

I would like to simulate the system {connector1 + 24 inches cable + connector2}, something like this :

Obviously, I cannot simulate this because of the computing cost....


But under HFSS, I can use deembedinbg using wave ports (but not using lumped ports) so my idea is to simulate :
- the S-parameters of the connector1 inputs / output
- a little length of cable using wave ports
- dembed to get the 24 inches lenght of cable
- the S-parameters of the connector2 inputs / output (which are the same as connector1)
- Add all the circuits in series

I am just wondering if I can add the S-parameters circuits in series and if it is correct to add them to get the total system s-parameters ?

Thanks you for your advice

 

[D.A.(Tony)Stewart]

You can simulate anything if you have tested it for the Zo and crosstalk or have all the L/C parameters.

Is it 200 ohm or shielded 100 ohm cable or ?

 

[rdpdo2002]

Hello,

It's a shield 100ohms differential cable
I can get the Z by TDR.

Thanks

 

[rdpdo2002]

Hello,

I am trying to simulate a 100 ohm differential impedance FCC flat cable (Molex) but I can't find consistent results when using the Lumped or Wave terminal ports.

My aim is to simulate 24 inches long cable. But as it is very processor demanding, I want to simulate a portion of the cable (here 5 cm) and virtualy increase the lenght of the cable by de-embeding it.

Problem is I cannot use lumped ports because using lumped port do not give the possibility to use de-embeding for extend lenght of cable. So I need to to use wave ports.

The structure of the cable is following and I am trying to simulate common signal to get 50 ohm impedance :

​

When using terminal Lumped port, I got 50 ohm result for a single signal with two GND around it (GND/ SIG / GND ) which is consistent with the 100 ohm diff cable :

​

Unfortunatly, when using terminal wave port, I got 38 ohms instead of 50 :

I have tried to modify a little the port geometry so that there is 3 digit scale for the E-field on the wave port, and a better e-filed vector distribution but I still get 38 ohm instead of 50 ohm...

I think the geometry for my port is OK because if I increase it like this, I got oscillation on the TDR response, but still 38 ohms:

Do someone have any idea to get better result using wave port ?

Thanks you very much !

 

[rdpdo2002]

Hello,

I still having this problem and I cannot resolve it... I asked on official ANSYS forum but nobody can epxlain me what's happen...

 

[D.A.(Tony)Stewart]

I cannot see a clear 2D dwg with crosssection with all dimensions signals, Dk for your single & dual signal coplanar stripline.

 

[rdpdo2002]

Hello @D.A.(Tony)Stewart ,


I am now trying simulating this portion of the cable :

I made some tests under Polar. Here are the data I used under polar (note the 100mm H1 substrate height to simulate no ground plane):

With these data, I got 115 ohm diff impedance :

Under HFSS, material are :

Under HFSS with terminal Wave port, I got about 70 ohm impedance :

And using lumped terminal port, I got about 130 ohm impedance :

I'm stuck on these problems I can't understand why I don't have consistent values with HFSS between modal & lumped ports and why I cannot get the right values polar give me ...

I tried to ask on ansys forum but get no answer.

I am trying to get an acces to ansys learning hub to ask for help but it's difficul to get a free acces...

--- Updated Nov 1, 2024 ---

I also tried this lumped port configuration:

And still get about 130 ohm:

--- Updated Nov 1, 2024 ---

Using polar, I get these different values for the impedance :

It seems that:
- Lumped port configuration give me the Even Mode impedance
- Wave port configuration give me the Common mode impedance

Do you think it is chance ??

My HFSS project is here : https://drive.google.com/file/d/1jEWDVxm6hTSQ0ILNnNiVQMt1N1CC-aNG/view?usp=sharing

 

[volker@muehlhaus]

I'm stuck on these problems I can't understand why I don't have consistent values with HFSS between modal & lumped ports and why I cannot get the right values polar give me ...

You compare apples to oranges ...

Polar gives you the differential impedance between the two center lines, with the outer lines as ground.
In the HFSS wave port solution, it seems that you excite a single wire with neighbors left and right as ground (somewhat like coplanar)
In HFSS lumped port solution, you excite the outermost two wires in differential mode.

 

[rdpdo2002]

You compare apples to oranges ...

Polar gives you the differential impedance between the two center lines, with the outer lines as ground.
In the HFSS wave port solution, it seems that you excite a single wire with neighbors left and right as ground (somewhat like coplanar)
In HFSS lumped port solution, you excite the outermost two wires in differential mode.

Thanks you so much ! So it seems I do not use port correctly...

I would do the same as Polar, that is to say use the out wires as reference (GND) for each of the internal signal wire

Can you please explain me what is wrong in the way I use the port... ? Here is how I do :

Using wave port, I use the internal two wires as terminal and out wires as reference for these terminals :

And after that, I construct my differential pair using the two internal wire (+ sig / - sig):

Using lumped port, I also use external wires as reference. Here is example for + signal of the differential pair (I do the same way for the neg signal, with reference to the other wire at left) :

And internal wire as terminal :

And after that, I construct my differential pair using the two internal wire (+ sig / - sig), same way as modal port.

 

[rdpdo2002]

Hello,

I think the Aluminium is the problem...

I have tried a single ended simulation, using Aluminium and FR4 :
- Using FR4, I got coherent values with Polar
- Using Aluminium, I got non coherent value !

I think it's because the aluminium is treated like a PEC by HFSS and it "block" the electric field of the lumped port wich normally tell how is the field for boundary condition ? Any idea to resolve this ??

Thanks so much again

--- Updated Nov 2, 2024 ---

By modifying the "Bulk conductivity" of Aluminim to 0, I got now good result :

But can I use this tips ? I don't know if using a 0 bulk conductivity for aluminium instead of the real value will induce bad results for S-parameters ???

I dont know if HFSS or Polar give the right value when using Aluminium ....

 

[volker@muehlhaus]

And after that, I construct my differential pair using the two internal wire (+ sig / - sig), same way as modal port.

Ok, that was not visible from your field plot.

But yes, I now see that your Polar model has a bottom ground, and your HFSS model (and cable) does not have that ground. So that is a big difference indeed.

In this case, HFSS is the more accurate model. If conductivity = zero in Polar gives the same result, that is fine, but in general I would trust HFSS then because it models your physical structure (no ground plane).

 

[rdpdo2002]

In this case, HFSS is the more accurate model. If conductivity = zero in Polar gives the same result, that is fine, but in general I would trust HFSS then because it models your physical structure (no ground plane).

Thanks for your advice.

Now that I figured out how to properly define my lumped ports and that aluminum can be a problem, I will try to use my nanoVNA on the real cable to see if I get better result for caracteristic impedance measurement under HFSS by using bulk conductivity=0 or not for aluminium...

 

[volker@muehlhaus]

to see if I get better result for caracteristic impedance measurement under HFSS by using bulk conductivity=0 or not for aluminium...

This makes no sense - if you don't have a metal plane there, then don't model it at all.

 

[rdpdo2002]

This makes no sense - if you don't have a metal plane there, then don't model it at all.

My cable have a aluminium shield.

It's not connected to GND and so it's not a GND plane for the diff. signals, so no need to simulate it ??

 

[volker@muehlhaus]

Your posts are a bit confusing - in #10 you show an arrow to a dielectric (=insulator) and write that this is aluminium (=metal), so it is difficult to understand what you really build.

My cable have a aluminium shield.

If your cables have a shield, then inlcude it in the model (with the real conductivity) and leave it floating, not connected to the ports. It will have an effect on line impedance, even if it is floating.

 

[rdpdo2002]

Your posts are a bit confusing - in #10 you show an arrow to a dielectric (=insulator) and write that this is aluminium (=metal), so it is difficult to understand what you really build.

Yes, I was making some test to compare results between Polar software and HFSS but polar software cannot take into account all the layers of the real cable so I needed to remove some layers for making the tests.

If your cables have a shield, then inlcude it in the model (with the real conductivity) and leave it floating, not connected to the ports. It will have an effect on line impedance, even if it is floating.

Thanks you this is what I was planning to do

 

[D.A.(Tony)Stewart]

Pardon my intrusion, but I cannot verify Molex's marketing specs either.

this is what I think.

This is an edge-coupled, embedded, differential waveguide geometry.
For Zdm to be 100 Ohms +/-10, and trace pitch = 0.500 mm with space, s =0.23 to trace, and height, h1 =0.10 mm above ground plane(=ref), s/h1 = 2.3 then I estimate Zo must be 100/1.6 = 62.5.

Yet with 0.100 mm PET Er= 3.4 , between any conductor and gnd plane, the I estimated Zo= 33 ohms.
So I look forward to your results.
Tony EE since 1975 , practising retirement

It's not connected to GND and so it's not a GND plane for the diff. signals, so no need to simulate it ??

View attachment 195094

With 0.1mm thick PET (Kapton), between Aluminum foil shield and Cu traces, you have a waveguide and thus a controlled Zo It will be about 30 pF / sq.cm which like a high pass filter will shield the RF signals and provide low impedance Zo where the prop. delay time is greater or equal to the risetime. For short lengths or slower rsietime signals, Zo does not apply with T-line effects yet the shield now acts like an antenna to AM/FM signals and couples into Zo but not Zdm "iff" it is perfectly balanced.

Anecdotal: Alum. foil is lossy and will combust in a microwave oven with wrinkles but when flat is an excellent reflector.
I can estimate your DM impedance accurately from Zo from the gap ratios for DM/CM, so make sure Zo is accurate.

On one page they call it a 90 ohm USB 3 cable and another page advertise it as a 100 Ohm HDMI cable. The Engineering doc indicates 100 Ohms +/-10 for Zdm which with this geometry tells me Zo must be 100/1.6=62.5 yet I compute 33 Ohms.

The other thing of concern is if the shield is not terminated, there will be CM EMI coupling to the CM signals and any mismatch in the traces will translate into loss of signal integrity. Anecdotal: I was delivering a custom designed Gantry for EM WPT measurements to U of T researchers and the Laptop USB would not communicate with an Arduino on a 1m cable if the 19V DC laptop charger was connected, yet when charger was disconnected, error-free and same when connected to a PE grounded PC tower on USB and this wasn't even USB3 signal rate.

 

[rdpdo2002]

Pardon my intrusion, but I cannot verify Molex's marketing specs either.

this is what I think.

This is an edge-coupled, embedded, differential waveguide geometry.
For Zdm to be 100 Ohms +/-10, and trace pitch = 0.500 mm with space, s =0.23 to trace, and height, h1 =0.10 mm above ground plane(=ref), s/h1 = 2.3 then I estimate Zo must be 100/1.6 = 62.5.

Yet with 0.100 mm PET Er= 3.4 , between any conductor and gnd plane, the I estimated Zo= 33 ohms.
So I look forward to your results.
Tony EE since 1975 , practising retirement


With 0.1mm thick PET (Kapton), between Aluminum foil shield and Cu traces, you have a waveguide and thus a controlled Zo It will be about 30 pF / sq.cm which like a high pass filter will shield the RF signals and provide low impedance Zo where the prop. delay time is greater or equal to the risetime. For short lengths or slower rsietime signals, Zo does not apply with T-line effects yet the shield now acts like an antenna to AM/FM signals and couples into Zo but not Zdm "iff" it is perfectly balanced.

Anecdotal: Alum. foil is lossy and will combust in a microwave oven with wrinkles but when flat is an excellent reflector.
I can estimate your DM impedance accurately from Zo from the gap ratios for DM/CM, so make sure Zo is accurate.

On one page they call it a 90 ohm USB 3 cable and another page advertise it as a 100 Ohm HDMI cable. The Engineering doc indicates 100 Ohms +/-10 for Zdm which with this geometry tells me Zo must be 100/1.6=62.5 yet I compute 33 Ohms.

The other thing of concern is if the shield is not terminated, there will be CM EMI coupling to the CM signals and any mismatch in the traces will translate into loss of signal integrity. Anecdotal: I was delivering a custom designed Gantry for EM WPT measurements to U of T researchers and the Laptop USB would not communicate with an Arduino on a 1m cable if the 19V DC laptop charger was connected, yet when charger was disconnected, error-free and same when connected to a PE grounded PC tower on USB and this wasn't even USB3 signal rate.
View attachment 195104

Thanks you very much, i will keep updated with my result... I hope I will be able to do some measurement using my little nanoVNA...

I think I will try firstly to get a single ended with 50ohm terminated single port S parameter and try to compare my result with HFSS.
Secondly, I will try to measure differential response single port with 2x50 ohms ended as my nanoVNA seems to have two ports. I will also need to deembed the SMA connector/cable before the molex.

For my project, I want to transmit VDSL data using 24inches of this molex cable, SPI speed at about 15Mhz with rise/fall time not lower than 15ns so I think it should be ok... But I would like to test it under HFSS/Circuit...

It's a personnal project so I dont care too much about EMI, only about SI...

 

[D.A.(Tony)Stewart]

edge-coupled, embedded, differential waveguide geometry.

Perhaps I made an incorrect assumption analyzing the Molex FPC flatcable, but I don't see how Zdm is 100+/-10 ohms

Critical ratios:
Trace width/height ratio = 2.7:1
Coplanar trace space/height = 1.6
This results in Zo much less than 45 Ohms, so how can Zdm reach 90 Ohms over a dielectric comm mode coupling to a ground plane?

Molex show the ground plane on top and the critical dimensions are for the connector fit and not to control Zo but the critical dimension between stiffener and alum foil inside the connector. I show this on my top left and rotated below bottom right in plan view or below this in in the tail cross-section. Again I ignored this short wavelength sensitive zone ~ above say 300 MHz.in this thread.

 

[rdpdo2002]

I now get consistent values when using modal terminal ports or lumped terminal ports ! the way I was initialy configuring lumped port for differential pairs was not well done... This is now one problem solved !

From my previous tests under HFSS, to get coherent values between Polar (polar do not take into account the bulk conductivity, only the delectric constant) and HFSS for a single ended caracteristic impedance, I needed to set bulk conductivity of aluminium to 0. So I have made some tests using both bulk conductivity to 0 or not for aluminium (I also made a test using the option "solve inside" when bulk conductivity is not 0) :

When bulk conductivity of aluminium is not 0, I got a Zdiff between Z0 and 2xZ0, like @D.A.(Tony)Stewart said. But I do not get a Z0 value much less 45 ohms... I got 50 ohms. And I got Zdiff = 70 ohms. Ratio Zdiff/Z0 = 1,4 (when bulk conductivity is not 0 and "Solve Inside" option is enable, I got same values)

When bulk conductivity for aluminium is 0, I got Z0=88 ohms, Zdiff= 115 ohms and Zdiff/Z0= 1,3.

So it seems to get Zdiff around 100 ohms I need to set bulk conductivity to 0...


All I have to do now is test the actual cable to see what I get... but I don't know if with my nanoVNA I can measure Z0 and also Zdiff... I don't now if I can sent a differential signal on the two ports... But I hope it will give me some guidance ...