Hello Magnus,
The characteristic impedance measured on your differential pair should
be 100 Ohm. Therefore, the odd impedance - frequently defined in the
books as sqrt((L11-L12)/(C11-C12)) for a lossless case - should be
equal to 50 Ohm. In practice, due to losses, skin effect and
proximity effect, your characteristic impedance will be frequency
dependent.
Using the following geometry:
1 oz. Cu gnd plane
--------------------------------------
dielectric layer thickness = th
-----------
|1 oz. Cu |
-----------
dielectric layer thickness = th
-----------
|1 oz. Cu |
-----------
dielectric layer thickness = th
--------------------------------------
1 oz. Cu gnd plane
And the following additional assumptions:
1. Space between ground planes and conductors is filled with FR-4
dielectric for which we assumed has a dielectric constant of 4.6 and a
loss tangent of 0.001.
2. We use the same parameter (th) for all dielectric layers (between
gnd-sig1, sig1-sig2, sig2-gnd) but we could also find a cross-section
with the minimum total thickness of the structure if you have the
flexibility to vary dielectric layers.
3. The 1 oz. Cu used for signal and gnd planes is in fact 0.03556 mm
high with a conductivity of 58.14e6 [S/m]
Our OptEM ID solver provides the following results to obtain a
differential impedance of 100.0 Ohms:
th [mm] width [mm] for DC width [mm] for 1 GHz
0.10 0.0403 0.0274
0.15 0.0689 0.0520
0.20 0.0996 0.0777
0.25 0.1305 0.1043
As a result of skin and proximity effects, the increase of frequency
from DC to 1 GHz results in a reduction of inductance and an increase
of resistance. However, in the above example, changes of inductance
dominate the frequency behavior of the impedance and therefore the
conductor width should be reduced to maintain characteristic impedance
at the level of 100 Ohms when the frequency is increased from DC to 1
GHz. Depending on the signal spectrum, you should probably select
your conductor width between 0.0403 mm and 0.0274 mm if your
dielectric thickness is 0.10 mm.
In general to increase differential impedance you should increase the
separation between signal lines, increase the separation between
signal lines and gnd, and reduce the dimensions of conductors.
We are posting the parametric plot results from simulating the above
structure for different values of conductor width and dielectric
thickness on our web site. It will be found at:
http://www.optem.com under Software, OptEM Interconnect Designer
Sincerely yours
Juliusz Poltz
----------------------------------------------------------------------
----------------- Magnus Homann wrote:
> Hi all,
>
> We're thinking about using a dual stripline for a differential
> pair. The diff. impedance should be 100ohm.
>
> Have anyone used this before? Or seen it in suggested? Come to think
> of it, it does sound like a good idea (not mine). Now, the hard part
> is getting the right geometry... Any help from various programs or
> books? Any references would very helpful, the only one I found is in
> "Transmission Line Design Handbook", and it is not very clear. Can
> you analyze this with a field solver?
>
> The concept goes like this (cut-through view of the board):
>
> Plane
> -----------------------
>
>
> --------
> | L1 |
> --------
>
> --------
> | L2 |
> --------
>
>
> -----------------------
> Plane
>
> The Zdiff should be between L1 and L2.
>
> Homann
"Juliusz Poltz" <jpoltz@optem.com>
*****************************************************
OptEM Engineering Inc. Ph: (403) 289-0499
100 Discovery Place Fax: (403) 282-1238
3553 - 31st Street N.W. Email: info@optem.com
Calgary, AB Canada T2L 2K7
*****************************************************