0
\$\begingroup\$

When the positive and negative tracks of a differential signal are not equal in length, we add some bumpy patterns in the shorter track to make them equal. On the other hand, we usually have a rule for the gap of the two tracks in order to preserve a constant characteristic impedance for the signal (typically 100 ohms for differential signals).

So it seems that the length tuning patterns may change the impedance since they change the gap. How can we minimize the impact of the patterns on the impedance?

Generally, using a few patterns with large amplitude is preferable or more small patterns? For example, which one of these two approaches in this picture is better?

different patterns of differential length matching

\$\endgroup\$

2 Answers 2

1
\$\begingroup\$

When the serpentine pattern features are too close to the trace width, the serpentined segment acts as a wider trace, with effective single-ended impedance much lower than the original trace. Any asymmetrically serpentined differential transmission line will have a serious impedance discontinuity. I would try something like this instead of serpentine:

enter image description here

\$\endgroup\$
1
  • \$\begingroup\$ Thanks @ali-chen, Sometimes in spite of this solution, the serpentine patterns are inevitable. \$\endgroup\$
    – M.H
    Feb 23, 2018 at 8:04
0
\$\begingroup\$

I don't think the impedance variance is going to be enough to worry about. It could be simulated in a 3D EM simulator. I also don't think it matters which pattern you use, it mainly depends on which one fits better into the routing space you have. Maybe of greater concern would be crosstalk that would essentially cause the signal to jump across the zig-zag pattern, again a 3D EM simulation would be helpful.

\$\endgroup\$
1
  • \$\begingroup\$ I've not worked with EM simulators. Isn't there a rule of thumb regarding the question? \$\endgroup\$
    – M.H
    Feb 23, 2018 at 8:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.