I'm conscious of the fact that this question may be open-ended which is discouraged per the How-to-ask guide, but I'm ignorant enough on the matter that I can't figure out how to ask a more pointed question in the title. I can, however, provide specific details.

I have two different cases relevant to this question. Both involve ruggedized, embedded electronics that utilize FastEthernet networks for communication. In both cases, I already know why they are having problems: it's because the cabling involved does not meet the performance requirements for Category 5e cabling, as demonstrated through the use of a set of Fluke DSX-8000 cable testers. Specifically, both cases shows several frequencies where the NEXT results fall below the threshold for Cat5e. Additionally, the first case shows poor return loss as well.

In one case, the cable in question has worked well (or at least well enough that no one noticed a problem) for a decade or more. But now one of the endpoints on this cable is being replaced by what the vendor told my customer is a "form, fit, and functional replacement" for the previous system, which is no longer available from the manufacturer. However, where the old product used an Intel card (not sure which one, but it uses the igb driver under Linux) to successfully make link on this cable, the new one uses the Ethernet NIC built-in to the Xilinx Zynq System-on-Chip (Cadence GEM driver under Linux), which will repeatedly go up and down.

The the other case, laptops with the Intel LM827579 chip failed to make link on the bad cable, while laptops with Intel I219-LM and some Realtek chip are able to make link.

Note that all these network interfaces will achieve link on a Cat5e-compliant cable, and I have no reason to doubt that they work according to their specifications. Unfortunately both customers have vested interests in not changing the cabling due to the large cost involved with changing anything across the installed base, so they can't let go of the fact we get different results depending on which devices we connect to the network. I've explained that the cables are marginal at best, but they still want a stronger explanation than that.

I have no problem telling them that the cables need to be replaced, but I do feel bad about not being able to communicate why some combinations work while other do not.

  • 2
    There's no good answer. Some engineers design with more tolerance -- others just meet minimum requirements. An analogy: You can drive a car faster than the maximum engine RPM. Some makes will break before others. – Ron Trunk Feb 25 at 20:30
  • 2
    A cable vendor should try to repair a single bad cable or replace it if is is unfixable (remember that there should already be a pull string in place, otherwise you can use the bad cable a s a pull string). The problem is usually a bad termination, and having a competent professional replace both cable terminations can often work wonders. – Ron Maupin Feb 25 at 22:55

Browse other questions tagged or ask your own question.