I've been trying to find information for the above question and I can't seem to find anything. I have 1000 BASE-T lines coming from an Ethernet cable to a PCB mounted connector. I'm pretty much confined to the area shown below:

enter image description here

The lines are not matched inter-pair (some are mismatched by 100 mils), however, they will be matched intra-pair. How important is it for the differential pairs to be length matched (inter-pair wise)? I've gotten conflicting responses from my peers that they do not need to be matched inter-pair wise and I can't seem to find any info online in regards to 1000-T.

Much help appreciated.


1 Answer 1


There is no need to have the pair lengths balanced. The standard for 1000base-T (IEEE Std 802.3-2008) states: Link delay skew

The difference in propagation delay, or skew, between all duplex channel pair combinations of a link segment, under all conditions, shall not exceed 50 ns

So, 100 mil difference are no issue at all. In fact, even normal RJ-45 cables are not balanced on purpose: Each of the pairs usually has a different number of turn on the twisted pairs to reduce the amount of cross-talk. Over the length of dozens of meters, this can accumulate to some difference in the length of the pairs that is larger than the difference in trace length on your PCB.

  • \$\begingroup\$ If you need a refresher on nanoseconds, see the wonderful illustration from Grace Hopper. \$\endgroup\$
    – David
    Jun 10, 2019 at 18:02
  • 4
    \$\begingroup\$ @David In twisted pair (100 ohms) the signals are slower, so a nanosecond is about 20 cm only. Nice fact, speed of light is one of the few cases where imperial units are simpler than SI: It's about a foot per nanosecond. \$\endgroup\$
    – asdfex
    Jun 10, 2019 at 18:09

Your Answer

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

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