Hello Folks,
we are in a heated discussion in our group, about simulation and the fabric properties of the material.
A coworker is insisting that the Hyperlynx simulation is bogus because they do not take into account the woven glass fabric style of the material. The tool does not take account of the cloth type nor the direction, thus it can not calculate the right value of the dielectric value according to the direction a signal is traveling on. The resulting change in phase for a diff pair traveling the board on different direction should be different, out of this consideration.
On my side I am of the opinion that the effects are here, but insignificant, as they are are of lower impact. I am unable to quantify this, and show that they are of lower impact, since I have no other simulator as Hyperlynx.
I assume that all simulation have some relation to measurement too :-)
Can somebody help me in this.
Thanks
Matija
Matija,
The fiber-weave effect depends on multiple factors, and it is a statistical issue. Therefore, I think it unwise to make a categorical statement about a modeling methodology or frequency or routing method related to fiber weave. There is a lot of good information published to describe the problem, how to simulate it (usually the worst case), and how to route traces to mitigate the problem. In just a quick search I found this paper from DesignCon last year.
https://www.signalintegrityjournal.com/ext/resources/article-images-2017/4103/4103_SIJ_GlassWeave.pdf
The abstract states that the problem can be (worst case) as much as 7ps/in, but with some good choice of dielectric material, and some typical manufacturing variation, the skew is less than 1ps/in.
Now the question is about the significance of 1ps/inch in your particular design. You can easily add a few picoseconds worth of T-line in a LineSim schematic to observe the impact of the skew on your circuit. Then you can decide if it is important to model the fiber weave in any particular design.
Another question is, what post-layout simulation tool _can_ simulate the fiber weave geometry? If you have to extract the model of a differential pair to another tool for fiber-weave modeling, why not do it in an easy-to-use tool, LineSim?
Regards,
Weston