0

I needed to extend a 50m cable by about 1 meter. I know that the best solution would be to make a new run but it's very difficult to swap the cable so I tried to avoid it. As RJ45 couplers would take a few days to be delivered, I bit the bullet and soldered the two cables together. While soldering also I managed to break a small part of the orange wire of one side so to extend that I used some salvaged resistor leg and when I finished the soldering wrapped everything in heatshrink tubing. It was tested afterwards and the average ping (my PC -> router) was either 1ms or <1ms with 0% packet loss.

After 2 weeks and having changed my main router (as per ISP's instructions for something unrelated) comes today. I noticed websites took much longer to load but on speedtest.net the speed was good. So I pinged 192.168.1.1 again and sure enough there was a lot of packet loss and higher delay. I wiggled the junction of the 2 cables a bit and after a couple of minutes the problem was fixed again. Now I'm again having 1ms latency but there are some spikes that go as high as 800ms.

My question: Is it the cable, the new router, something else or is it normal to have spikes after around 1800 ping packets?

5
  • 3
    I don't mean to be rude, but you did everything wrong, and you wonder why things aren't working?
    – Ron Trunk
    Jul 1 at 11:54
  • @RonTrunk I don't mean to be rude, but a "you did everything wrong" reply isn't much helpful either, especially when you are talking to a newbie which you probably know I am based on the question. Jul 1 at 12:45
  • Perhaps you're right. My apologies. I do want to point out that based on your description, you seemed to be aware that you were doing things incorrectly. If you really had no idea, I would have answered differently.
    – Ron Trunk
    Jul 1 at 13:04
  • It's okay! No worries! Perhaps I didn't clarify that I wasn't aware of that. Not that I didn't research the subject before soldering, but a local ISP technician that came here the day before told me I would just be fine with it. At least now I know and I'm going to make the change this week! Thanks so much! Jul 1 at 13:37
  • 1
    You must run a new cable. The standard precludes splicing. Obviously you did not run the category test suite because it will fail on your cable. You should use the pull string (or the old cable if you did not properly install a pull string) to run a new cable and pull string.
    – Ron Maupin
    Jul 1 at 18:36
1

Can a bad ethernet cable cause lag spikes?

Easily.

A bad cable significantly increases the error rate, causing frame drops - an Ethernet frame with a bad FCS is simply ignored/dropped.

Requests or replies inside these frames never reach their destination and need to be retransmitted after a timeout. TCP as transport-layer protocol does that all by itself, protocols using UDP may retry on the application layer (e.g. DNS).

You should check the NIC's and the switch's FCS error counters. Anything in excess of .001% frame error rate is a bad cable to me.

Soldering a network cable is a bad idea. The solder blobs change the impedance of the cable, causing reflections and noise. You should replace the cable. If that isn't possible at all, use a 'proper' junction/splice box ('proper' because the standards don't even allow for this, but done correctly the cable works fine usually).

Of course, the problem could be something else (if you can't tell how should we?), but the cable is the most likely.

1
  • Thanks so much for your time and detailed explanation of everything! A replacement cable is on it's way right now and I will not do it again to avoid that hassle! Jul 1 at 12:45
0

As you said in the first few sentences, the correct solution is a correctly spec'd (length) cable. Anything else is a hack. A coupler would be the common approach -- and it usually will work -- but it's technically a violation of 802.3 specs. (If you're in an environment where compliance is required, such a "patch" would fail -- and potentially cost someone their job.) Soldering(!!!) an extension on is a super bad idea -- that's an end-of-the-world type solution. (comet about to crash into whatever, and you need to get the cure for cancer / proof of alien conspiracy / etc. out of a computer that can't be moved...)

Those solder joints are extremely noisy points in the line, and because a significant amount of twist has been removed, they're sponges for any other electrical noise, including crosstalk. But by far the worse part is the brittle junction it creates. (even more so if the solder job wasn't done well.) It sounds like one or more joints weren't so strong and have broken from the cable being moved around; the only thing maintaining continuity is the heatshrink.

To the point of your question... yes, a bad cable can cause "latency". But not exactly, as it isn't causing the signal to be any slower. It causes errors that cause entire frames to be dropped. Unless something is specifically looking at error counters, those drops will be invisible. Ethernet is not a "reliable link", and has no retransmission capability. When a frame is dropped, it's just gone; nothing even knew it was supposed to be there.

(Of course, your ping times could also be all over the place because the target device isn't answering them. ICMP echo is pretty low priority.)

2
  • I should also add, any seasoned network engineer knows there's no such thing as a temporary fix. Whatever solution that gets done is where it stops; the ticket gets closed, or dropped to zero priority to never be looked at again. (until, and if, it falls apart.) I have decade old tickets like that -- a proper solution will never be done, because I'll never be given the time and resources.
    – Ricky
    Jul 1 at 23:07
  • Ricky thanks a lot for your reply! It made me understand a lot of things I didn't know, especially the part about frame drops. Jul 7 at 9:06

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.