I am aware that punch-down junction boxes and other solutions are the recommended methods for continuing an UTP cable that is too short, but you are not in control of replacing (for example, solid-built in the wall).

Having a bit of electrical background in my past, I am really perplexed by the fact that carefully soldering UTP cable is seen as bad for high speeds (usually cited as an 'impedance bump'), but junction boxes which add potential resistance and risk of oxidation at the edges, are seen as good.

What are the exact details of why soldering is considered bad for this type of operation?

Later edit: I did a small, fast and not-the-most-scientific experiment (video of testing it) and cut a patch cord in two and soldered it back. I used then www.speedtest.net to measure the speed and it was the same as before (300 Mbps as in my subscription's specifications).

  • 1
    Nobody disputed that soldered cables may work. This site is about professional networking and stuff like this isn't. Neither is measuring cable reliability by running an Internet speed test. – Zac67 Oct 10 '17 at 20:57

I am aware that punch-down junction boxes and other solutions are the recommended methods for continuing an UTP cable that is too short...

You cannot splice or otherwise extend a UTP cable that is too short. If a UTP cable is too short, you must run a new cable of the proper length, not to exceed the maximum length allowed (properly installed cabling will have a pull-string, or you could use the old cable for a pull-string). Splices, taps, couplers, etc. are expressly forbidden by the ANSI/TIA/EIA 568, Commercial Building Telecommunications Cabling Standard. Splicing the cable adds impedance and decreases the return loss.

UTP cabling is limited to 100 meters. That length assumes up to 90 meters of solid-core (better performance, but fragile) horizontal cable, and no more than 10 meters of stranded (poor performance, but less fragile) patch cord combined between both ends.

The frequencies used by modern networking require adherence to tight specifications.Each cable run must be tested to meet or exceed all the tests in the test suite for the cable category. Simple connectivity from end-to-end (like for telephone cable) is not sufficient. The high, modern networking frequencies require tight specifications to be met. The primary tests are:

  • Wire Map - Checks for proper pin to pin termination, and for each of the 8 conductors the wire map checks for: Continuity to the far end, Shorts between any two or more conductors, Reversed Pairs, Split Pairs, Transposed Pairs, Any other miswiring.

  • Length - The physical length of the cable is the actual length derived by measurement of the cable(s) between the two end points. The electrical length is the length derived from the propagation delay of the signal and depends on the construction of the cable. The maximum physical length of the horizontal cable (permanent link) one end of the cable to the other is 90 meters. The maximum length of the channel model is 100 meters.

  • Insertion Loss - Insertion loss is the loss derived from inserting a device into a transmission line. The insertion loss for both the permanent link and the channel models are the total insertion losses of all the components.

  • Near End Cross Talk (NEXT) - Pair to pair NEXT loss is the measurement of signal coupling from one pair to another. The result is based on the worst pair to pair measurement.

  • Power Sum Near End Cross Talk (PSNEXT) - Power sum NEXT takes into account the statistical crosstalk between all pairs while energized. This is a calculated amount derived by adding up the crosstalk results between all pair combinations.

  • Equal Level Far End Cross Talk (ELFEXT) - FEXT is the unwanted coupling of a signal induced by a transmitter at the near end, measured on the disturbed pair at the far end. ELFEXT is the same measurement of FEXT, less the effect of attenuation.

  • Power Sum Equal Level Far End Crosstalk (PSELFEXT) - As in Power Sum NEXT, these are computed values based on the sum of all the possible pair combinations under the respective tests.

  • Return Loss - Return loss is the value of energy reflected by impedance variations when devices are inserted into the cabling system.

  • Propagation Delay - Is the time it takes the signal to travel from one end of the cable/system to the other. The maximum channel propagation delay is 555ns (nanoseconds) and for the link it is 498 ns, both measured at 10Mhz.

  • Delay Skew - Delay skew is the signalling delay difference in time (nanoseconds) between the fastest pair and the slowest pair. The maximum channel delay skew is 50 ns, and in the permanent link it is 44 ns.

Any tests that are out of specification will fail the test, and the condition must be corrected and the test suite performed again until the cable passes or is replaced.

You can also permanently damage a cable by exceeding the pulling tension or minimum bend radius when installing the cable run.

  • Splices in structured cabling are technically prohibited, but I see it done a lot -- couplers, and patch panels inside crawl spaces. I suspect the answer he/she is looking for would best come from an EE -- it's an electrical ("RF") signalling problem. (I can explain it in terms of fluid dynamics, but not RF) – Ricky Beam Oct 8 '17 at 3:39
  • 1
    Yes, but I have seen a lot of flaky cabling that must be replaced because it is spliced. It's one of those things that may or may not work, and it may simply cost a lot more time, trouble, and money than it does to do it right in the first place. Cabling that seems to mostly work fails the test suite, then we find the splice, replace the cable, and everything starts working correctly. – Ron Maupin Oct 8 '17 at 3:48

'Impedance bump' is the exact keyword: the solder point changes the impedance of the cable. When this happens, energy is reflected back to where it came from and - obviously - this is bad, attenuating the signal and directly adding to crosstalk.


When soldering two UTP cables together, the new cable will technically work as long as there is a good connection between the wires. But the twists that you see on each pair is intended to be there. Which are engineered. By cutting the ends off and soldering them together, you lose some of the twists and therefore some of the performance. So the network performance will almost definitely suffer, and you may get a lot of packet loss.

  • 2
    You could solder preserving the twisting (provided it's the same direction) - that's not the only point. – Zac67 Oct 10 '17 at 6:26
  • Indeed, if you carefully solder (with a thermostat-enabled soldering station) each of the 8 conductors, carefully cut at offsets (not all of the same length), you can preserve almost all of the twisting. I intend to make an experiment soon by speed testing a patch cable as it is and then speed testing it after I cut it in half and solder the two halves. – Andrei Rînea Oct 10 '17 at 11:13
  • @AndreiRînea, what do you mean by speed testing it? You need to use a cable tester. They are very expensive, but you can usually rent one by the hour. It will run through the full test suite for the cable category. – Ron Maupin Oct 10 '17 at 12:47
  • By "speed testing" I meant running a long, large data transfer using that cable and measuring the actual speed of transfer. – Andrei Rînea Oct 10 '17 at 14:43
  • 1
    @AndreiRînea, that really doesn't measure what you seem to think it does. It's not like the bandwidth slows down or speeds up as the cable gets worse or better. Really, your business is depending on good cabling, and it should be willing to make a capital expenditure that should last 20 or so years. For example, if you had invested in properly installed and tested Category-5e cabling in 1999, it could run the modern 1 Gbps ethernet speeds today, but not the new 10 Gbps. A current installation of Category-6a should last well into the future. – Ron Maupin Oct 10 '17 at 20:56

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.