Quick Answer: A continuity tester costs little and finds about 80 percent of faults if you read four things: wire map, pair count, length and the presence of a split pair. Pass means eight conductors on the right pins, all four pairs used, length under 100 m (328 ft) and no pair split across two cables. Anything else is a re-termination, not a software problem.
Most link faults are mechanical and repeatable: a conductor on the wrong pin, one pair split between pins 3 and 6 and 4 and 5, a jack punched with the wrong colour code at one end only. Each has a distinct signature on a tester, and a link that runs at 100 Mbit/s instead of 1000 Mbit/s is telling you which one it is before any tool is connected.
Reading the wire map
T568B assigns white-orange to pin 1, orange to 2, white-green to 3, blue to 4, white-blue to 5, green to 6, white-brown to 7 and brown to 8. T568A swaps the orange and green pairs. Either is acceptable as long as both ends match; a mixed pair of codes still produces a straight-through link and still passes a continuity test, which is why a map that shows 1-1 through 8-8 is not proof of a good cable.
What matters is that pins 1 and 2 are one twisted pair, 3 and 6 are another, 4 and 5 a third and 7 and 8 the fourth. A split pair puts 3 on the white-green of pair 3 and 6 on the green of a different pair, so the signal travels out on one pair and back on another. The link then radiates and picks up noise, and 1000BASE-T will not train at all.
Fault signatures and what each one means
| Fault | Tester reading | Typical cause | Action |
|---|
| Open pair | No continuity on one pin | Blade missed the conductor | Re-punch and re-test |
| Split pair | Map passes, crosstalk fails | Wrong colour sequence at one end | Re-terminate both ends |
| Short | Two pins read as one | Stray strand in the jack | Trim and re-punch |
| Length fault | 137 m (450 ft) reported | Run too long or a bridge tap | Add a switch or re-route |
| PoE voltage drop | Map clean, 44 V at far end | High resistance joint or thin gauge | Re-terminate, check gauge |
Length is measured by time domain reflectometry, so the tester reports the electrical length of the pair, not the physical route. A cable dressed around three corners and coiled in a ceiling void reads longer than the floor plan suggests, and a reading 20 percent above the plan usually means a service loop was left in place.
Testing in the right order
Before the switch is connected
Test the permanent link with the patch leads removed: jack to jack, no active equipment. A fault that disappears when the patch cable is swapped belongs to the lead, not the building wiring, and swapping leads is the fastest single diagnostic step available.
After connection
Check the port counters for CRC errors and retransmits over five minutes at full load. Zero errors on a 1000BASE-T link means the channel is clean at the frequencies that matter. A handful of errors per million frames is normal; thousands per minute is a marginal pair, usually one with a partially cut conductor.
Where the fault usually sits
Terminations account for the large majority of failures, and the second most common location is the outlet: a faceplate rj45 screwed down onto a cable that was nipped by the box edge will pass a map test and fail under load. Third is the cable itself, usually at a tight bend behind a trunking lid or where a tie was pulled hard enough to deform the pair geometry.
Test at the frequencies the link actually uses. A Cat 5e tester that stops at 100 MHz will pass a cable that was installed as Cat 6 but kinked, because the failure appears between 150 and 250 MHz. If the plant is cat6 cable and the certifier is a 100 MHz unit, the test tells you almost nothing about 10GBASE-T margin.
Practical scenarios
- Situation: A new 24-port patch panel shows two ports linking at 100 Mbit/s into gigabit switches. The map passes, so test crosstalk: both ports have pins 3 and 6 from different pairs, and re-punching to T568B restores 1000 Mbit/s.
- Situation: A camera on PoE++ reboots every 20 minutes. The map is clean and the run is 88 m (289 ft): measure under load and find 46 V at the port dropping to 41 V at the camera, which points to a high-resistance joint at the mid-span coupler.
- Situation: A tester reports 118 m (387 ft) on a run the plans show as 70 m. A 15 m service coil is found above a ceiling tile; removing it brings the reading to 71 m and the link trains at gigabit.
- Situation: An office reports intermittent drops on one desk. A $40 tester passes, and a qualification test at 250 MHz fails return loss by 3 dB, which is a crushed cable under a floor box lid. Replace the last 5 m rather than the whole run.
FAQ
Do I need an expensive certifier?
For a handful of ports, a continuity and length tester plus a known-good patch lead finds most faults. For a contract handover, or for anything carrying 2.5GBASE-T and above, a certifier that reports per-pair margin at the rated frequency is the only defensible evidence.
Why does a cable pass the tester but not link?
Most often a split pair, which a simple map test cannot see. The second cause is length: a link over 100 m may train and then drop. The third is a device expecting four pairs when only two are terminated.
Can a cable test damage equipment?
A passive continuity tester cannot. Some active testers inject a signal that a live switch port will interpret as traffic, so disconnect the run at both ends before testing. Never test a link that is carrying PoE.
