The tester says the cable is perfect. The laptop agrees. The uplink disagrees — because the fault is not in whether the wires connect, but in which wires are twisted together.
Of all cable faults, the split pair is the one that makes experienced people feel cheated, because every cheap test says the cable is good. The wires are continuous. The colours match at both ends. And yet, as the video demonstrates, the moment the cable is asked to join two switches — especially at gigabit, over a real distance — it falls apart. The explanation is one sentence: Ethernet signals travel on twisted pairs, and a split pair quietly untwists the signal while leaving every connection in place. This guide explains how it happens, why the usual tests miss it, and how to prove it for free.
Inside a network cable, wires travel as four twisted pairs. The twist is the technology: the two wires of a pair carry opposite copies of the same signal, and interference hits both equally, so the receiver cancels the noise out. A split pair happens when the plug is wired so that a signal and its return path come from different twisted pairs. Every pin still connects to the correct pin at the other end — continuity is perfect — but the two wires carrying one signal are no longer twisted together, so nothing cancels the noise they pick up along the run.
The natural instinct when crimping is to arrange colours in tidy groups: orange pair, green pair, blue pair, brown pair, side by side. It looks professional and, as the video's story goes, it is how many self-taught installers have always done it — matching the same tidy order at both ends so "the colours meet". The catch is that Ethernet's standard pinout is not tidy: in T568B, the green pair is split around the blue pair (pins 3 and 6, with blue on 4 and 5). The standard looks wrong precisely because it is protecting the pairs. Grouping the colours neatly breaks exactly one thing — the pairing of pins 3 and 6 — which happens to be a working pair at 100 Mbps and part of the four-pair conversation at gigabit.
A basic cable tester sends a pulse down each pin and checks it arrives at the same-numbered pin. Split pairs pass 8 out of 8 — the tester has no idea which wires were twisted together inside the jacket. And a PC link at 100 Mbps over a desk-length cable survives because the run is short, the noise picked up is small, and Fast Ethernet only needs the two pairs involved. Stretch that same cable to an inter-floor uplink at gigabit — four pairs, both directions, real distance, electrical noise from the building — and the uncaptured noise drowns the signal. Same cable, same tester result, different physics.
Wire-map testers that specifically detect split pairs do exist and are worth owning if you crimp often — but the plug-reading test above is free and settles most cases.
There is no setting, driver or prayer that fixes split pairs; the cable must be re-terminated. Cut the plugs off (the fault lives in the last centimetre, not the cable itself) and crimp fresh plugs to T568B on both ends, keeping each pair twisted until it enters the plug and stripping as little jacket as possible. Retest: tester, link speed, then a real file copy across the uplink. Teach the finding to whoever crimped it — the goal is not blame, it is that the next hundred cables the building gets are wired to the standard the first time.
See the live troubleshooting in the video — the same cable tested with a laptop, then between two switches, and the wiring fault found on camera.
▶️ Watch Full VideoA basic tester checks continuity only — that pin 1 reaches pin 1, pin 2 reaches pin 2, and so on. In a split-pair cable every pin does reach its partner, so it scores 8 out of 8. The tester cannot see that pins 3 and 6 are not twisted together inside the jacket. Detecting that needs either a wire-map tester with split-pair detection or simply reading the plug colours against T568B.
At 100 Mbps over a short desk cable, the noise a split pair picks up is small enough for the signal to survive — Fast Ethernet also uses only two of the pairs. Stretch the same cable to a long uplink and ask for gigabit, where all four pairs run in both directions at once, and the uncancelled noise swamps it. Short and slow forgives; long and fast does not.
No. The fault is in how the wires were landed in the plugs themselves, so both plugs must come off and be re-crimped following T568B (or T568A) exactly, at both ends. The cable in between is fine. Keep the pairs twisted right up to the plug when you re-crimp — excessive untwisting at the ends recreates a milder version of the same problem.