RE: [SI-LIST] : Dual Stripline impedance

Michael A. Baxter ([email protected])
Wed, 15 Apr 1998 13:14:34 -0400

We have used both edge-coupled and broadside-coupled differential pair
constructions successfully in past and current designs. We have not found
registration tolerance to be a problem to date but this is always a
concern. We typically try to use wider lines of 8 mils or wider which
gives you more metal to withstand off-set registration. Our calculations
show that the lines can be off set by a couple of mils without significant
changes in differential impedance. Wider lines are used to minimize loss
effects in very high-speed designs; we also typically use 1oz. copper
wherever possible (more cross-section to reduce resistive losses). This
becomes very important with today's very fast edge rates and signaling
at 1Gbps and beyond. We typically use broadside-coupled pairs where
routing channels are scarce.

By the way, we use internally developed Method of Moments field solvers
to do our computations. Any number of field solver tools should be able
to handle this geometry.


- Michael Baxter

p.s. A simple formula for predicting differential impedance for coupled
pairs is: Z,diff = 2*Z0[(1-kb)/(1+kb)] where Z0 is the single-ended
impedance and kb is the backward crosstalk coefficient. As the coupling
goes to zero, Z,diff = 2*Z0 as it should for uncoupled pairs.

At 11:33 AM 4/15/98 -0400, you wrote:
>This type of differential impedance configuration is called Broadside
>Coupled Stripline. This model has some benefits and liabilities. The
>pair is created by having identical routing paths for the two traces and
>placing them on adjacent layers. It is mechanically similar to the dual
>stripline model for characteristic impedance and makes constructive use
>of the interplay between overlapping circuits. A Broadside Coupled
>Stripline is theoretically predictable but has poor controllability in a
>fabrication environment. The dielectric thickness variation between
>Plane1 & L1 and Plane2 & L2 causes signals to have non-identical
>reference plane locations.
>The dielectric between the two signals becomes critical. Natural
>variation of this dielectric causes a large amount of variation in the
>differential impedance along the entire trace and in localized sections
>of the trace. This variation is caused by variations in the laminate
>composition, changes in trace geometry and pressure distribution during
>lamination. Layer to layer shift (variation in registration) causes
>variation in the differential impedance because of a change in the
>overlap between the circuits. In addition, these will most likely be a
>different etched trace width distribution from one layer to the other.
>This will result in an effective overlap equaling the width of the
>smaller trace.
>When you think about using a Broadside Coupled Stripline configuration
>the variables noted above must be considered and taken into account when
>considering impedance tolerancing.
>Regarding differential impedance calculators which can calculate
>Broadside Coupled Striplines, I would suggest Polar Ltd. As a possible
>source. Polar is a manufacturer of TDR's and has published some
>differential impedance calculators in the past. Polar's address is
>Polar Instruments Ltd, Garenne Park, St. Sampson, Guernsey,Channel
>Islands GY2 4AF, UK.
> Tel: +44 (0)1481 53081 Fax: +44 (0)1481 52476
>Email: [email protected]
>Steve Silbert
>Viasystems. Inc.
>[email protected]
> -----Original Message-----
> From: [email protected] [SMTP:[email protected]]
> Sent: Wednesday, April 15, 1998 10:45 AM
> To: [email protected]
> Subject: [SI-LIST] : Dual Stripline impedance
> 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
> of it, it does sound like a good idea (not mine). Now, the hard
> is getting the right geometry... Any help from various programs
> 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

| Michael Baxter e-mail: [email protected] |
| NESA, Inc. |
| 636 Great Road Tel +1.978.897-8787 |
| Stow, MA 01775 USA Fax +1.978.897-5359 |