Microwave Office Stripline Model vs EM

[CharlieTheTuttle]

Has anyone tried to do an EM model of a stripline and compared it with the model
S1LIN. Test Conditions :

Substrate : 1mm thick FR-4, Er=4.7.

Stripline in center, thus 0.5mm FR-4 each side. Line is 80mm long, 0.3mm wide.
Test Freq is 450Mhz.
RL = Return Loss with other side of line in
50R.

My results are :

RL of S1LIN Model : -21.2 dB
RL of EM Model : -13 dB
Linear Sim with Aplac : -21 dB.

Why am I suspicious? Because I've done
the exact same thing but with a Microstrip
line and the results (linear vs EM) are
very similar. Has anyone else done some
tests like this with Microwave Office
or have some light to shed on the subject.

Thanks
Charlie The Tuttle

 

[flugbottom]

Could you specify a strip thickness?

 

[goldensol]

On 2001-10-11 08:56, CharlieTheTuttle wrote:
Has anyone tried to do an EM model of a stripline and compared it with the model
S1LIN. Test Conditions :

Substrate : 1mm thick FR-4, Er=4.7.

Stripline in center, thus 0.5mm FR-4 each side. Line is 80mm long, 0.3mm wide.
Test Freq is 450Mhz.
RL = Return Loss with other side of line in
50R.

My results are :

RL of S1LIN Model : -21.2 dB
RL of EM Model : -13 dB
Linear Sim with Aplac : -21 dB.

Why am I suspicious? Because I've done
the exact same thing but with a Microstrip
line and the results (linear vs EM) are
very similar. Has anyone else done some
tests like this with Microwave Office
or have some light to shed on the subject.

Thanks
Charlie The Tuttle

You can carefully read it's help.Maybe you using it in wrong way

 

[CharlieTheTuttle]

I have spent many hours on this problem, and consulted the manuals, without finding a
solution. It's interesting to note that in the supplied examples, there are stuff all stripline examples.

There is one example that compares EM striplines at 15Ghz, with various cell widths. They certainly do vary a lot!
However, only microstrip applications
have examples of "volataire (the linear
simulator) vs EM model".

One could say that after you reach some
20dB of return loss, any small errors
in the models will cause large errors in
a RL graph of dB. However, we are talking
about 13dB vs 21dB! And why does the
microstripline volataire model match the
EM model so well, even down to 35dB of RL?

Maybe there is something wrong with my
interpretation of how you should do things,
but why are there so few examples of
stripline applications with Microwave
office (in fact, no where in the manual
does it mention stripline, only micro
stripline).

I think I might have to simulate the
line in Eagleware to see what model is
wrong.

Oh, the line thickness is set at 0.035mm
which corresponds to 1/2 oz copper.

CharlieTheTuttle.

 

[bug]

I ran your case in a field solver, and the line impedance appears to be 54.6 ohms. It translates to about -26dB return loss.

 

[flugbottom]

>Oh, the line thickness is set at 0.035mm

All troubles start here. EMSight doesn't model accurately thick conductors. It uses approximate surface impedance model that cannot account for different currents flowing over the upper and lower surfaces of a thick conductor. However, S1LIN implements approach that accounts for this phenomenon. You can model thick conductors in EMSight adding thin dielectric layer and placing two strips on its top and bottom; add also several shorting vias along conductors.
Thickness of strips metal doesn't matter much.

For your case this approach gives RL=-18 dB, that is closer to -21 than -13 :smile:

Of all known MOM 2.5 EM solvers only Ensemble 8 claims to be able to model thick metall adequately...

 

[CharlieTheTuttle]

On 2001-10-13 08:52, flugbottom wrote:
>Oh, the line thickness is set at 0.035mm

All troubles start here. EMSight doesn't model accurately thick conductors. It uses approximate surface impedance model that cannot account for different currents flowing over the upper and lower surfaces of a thick conductor. However, S1LIN implements approach that accounts for this phenomenon. You can model thick conductors in EMSight adding thin dielectric layer and placing two strips on its top and bottom; add also several shorting vias along conductors.
Thickness of strips metal doesn't matter much.

For your case this approach gives RL=-18 dB, that is closer to -21 than -13 :smile:

Of all known MOM 2.5 EM solvers only Ensemble 8 claims to be able to model thick metall adequately...

Thank you very much for confirming that I'm not insane! However, I don't understand why the "metal thickness model
issue" would effect stripline EM models
and not microstripline EM models. In both
of my test examples, the thickness was
the same, but the error in the microstrip
EM model was very small.

CharlieTheTuttle

 

[Troubleshooter]

hi,
it's not exactly matching to the question posted here, but maybe it's a little help.
under ads the results from the build in "line calc" in the scematics simulations
don't fit to the momentum simulations in the layout. so you have to optimize the
width of a microstrip line to get a 50 ohm line.

w*w.rogers-corp.com/mwu/mwiform.htm

here you find a small free tool which results in better matching to momentum.
maybe it's better for striplines, too.
regards

<font size=-1>[ This Message was edited by: Troubleshooter on 2001-11-05 20:39 ]</font>