1
\$\begingroup\$

For example, one wire of a twisted pair is longer than the other wire (imbalance starts from the middle of the twisted pair, everything is balanced before the middle point). What problems will it cause?

\$\endgroup\$
4
  • \$\begingroup\$ Do you mean that you have a regular twisted pair but it is terminated with unequal-length wires? Or do you mean something like a twisted pair with one straight wire and the other wrapped on it in a spiral? I'm pretty sure that in the latter case your assembly will act like a regular twisted pair, except that it may show more radiated emissions and susceptability, and its velocity factor may be slower. It makes me wish for enough spare time to run off and do the experiment. \$\endgroup\$
    – TimWescott
    Jun 17 '19 at 20:17
  • \$\begingroup\$ Mathematically / theoretically speaking, you'd want your differential pair to be perfectly matched in length and impedance across the entire link to achieve the maximum benefits. Of course, we can't actually do this in real life. The maximum deviation from that depends on your link; while PCIe may require +/- 0.1mm in terms of matching, something like RS-232 or something slow may let you get away with inches of mismatch. \$\endgroup\$ Jun 17 '19 at 21:28
  • \$\begingroup\$ Depends on what is driving the transmission line, and if it's being driven single ended or differentially. \$\endgroup\$
    – Voltage Spike
    Jun 18 '19 at 20:37
  • \$\begingroup\$ @laptop2d can you explain both? Thanks \$\endgroup\$ Jun 18 '19 at 20:53
1
\$\begingroup\$

The inverted signal will be phase shifted compared to the non-inverted. If the imbalance is a significant portion of the wavelength, there will be noise issues. Noise is emitted because the EM fields of the signals don't cancel each other like they do on a matched pair.

\$\endgroup\$
1
\$\begingroup\$

For a diff pair, having a mismatch not only reduces the data eye pattern margin, but it also causes common-mode current for the uncancelled portion of the signal. This will show up as noise on the ground reference alongside the pair and can be a big EMI headache.

This is a huge problem for example with HDMI on the driver side: unequal trace length adding differential skew to the pairs, resulting in common-mode noise at harmonics of the data rate. One remedy is to apply common-mode diff filters to knock down this non-cancelled noise component.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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