Copper Link Won't Certify: What the Tester Is Actually Telling You

A certification tester does not just say PASS or FAIL. Every parameter it measures points at a different fault, and the failing parameter is the diagnosis — if you read it instead of just re-testing and hoping. A wiremap failure is a termination mistake. A NEXT failure at high frequency is usually untwist at the jack. A length failure is usually a settings error, not a long cable. The tech who treats the tester as a yes/no box re-terminates the wrong end twice and then pulls a new cable run; the tech who reads the parameter fixes the right thing once.

First: are you certifying, or just checking?

Three different tools get called "testers," and they answer different questions. A verifier checks continuity and wiremap — is it wired right, is anything open or shorted. A qualifier runs a signal test and estimates whether the link will carry a given speed, but it is not a standards test and its result will not settle a warranty claim. A certifier measures the full parameter set against a published test limit — TIA-568 or ISO 11801 — and issues a pass/fail with a margin at every frequency. Only a certifier's result settles a handover or warranty dispute. Before you argue about a FAIL, confirm which tool produced it and which test limit was selected: a Cat6A permanent-link limit and a Cat6A channel limit are different tests, and testing a permanent link against a channel limit (or the reverse) produces a meaningless verdict.

The parameters, and what each failure means

Wiremap is each conductor landing on the right pin — no opens, shorts, reversals, crossed or split pairs. A fail is a termination mistake, not a cable fault. Fix it at the jack or patch panel and check the punch-down order against T568A/B.

Length is physical length calculated from signal delay and the NVP setting, with a 90 m permanent-link limit and 100 m for a channel. A fail usually means the wrong NVP for the cable type, or a coiled service loop being counted as length. Confirm the NVP against the cable datasheet before believing it.

Insertion loss is signal lost along the run, growing with length, frequency, and temperature. A fail means an over-long run, a poor connection, or a hot pathway. Check length first, then re-seat connections and look for kinks and tight bends.

NEXT and PSNEXT are near-end crosstalk — signal leaking between pairs at the same end. A fail usually means too much untwist at the termination (over about 13 mm on Cat6A) or a low-grade jack. Re-terminate keeping pairs twisted to the contact, and replace the jack if it repeats.

ACR-F and PSACR-F are far-end crosstalk, formerly ELFEXT. This one is less termination-sensitive than NEXT, so if NEXT passes and ACR-F fails, suspect the cable itself rather than the ends.

Return loss is signal reflected back by impedance mismatches. A fail means mixed-category components, a crushed or kinked cable, or a bad crimp. Every component in the link must be rated for the category being certified, patch cords included.

Propagation delay and delay skew are how long each pair takes and the spread between the fastest and slowest pair. Gigabit and faster Ethernet send data down all four pairs at once, so a large skew means one pair arrives late. A fail with a clean wiremap usually means split pairs or two different cable types joined in one run — look for a splice or a mid-run repair.

Wiremap first — always

Wiremap is the first parameter for a reason: every other measurement is meaningless on a miswired link. The usual suspects are a split pair (two conductors from different pairs landed together — the classic cause of a link that "works" at 100 Mb/s and fails at gigabit), a reversed pair, a crossed pair, or an open/short from a conductor that missed the contact. All of them are termination faults, fixed at the jack or the patch panel, not in the ceiling. Re-terminate cleanly — pairs twisted right up to the IDC contact, conductors fully seated, no nicks from the punch tool — and re-test before you touch anything else.

NEXT: the untwist tax

NEXT is the most common certification failure on an otherwise good run, and it is almost always paid at the termination. Untwisting a pair to punch it down destroys the geometry that cancels crosstalk, and the damage grows with frequency — so a link can pass NEXT at 100 MHz and fail it at 350 or 500 MHz, which is exactly the Cat6A failure signature. Keep the pairs twisted until the last possible millimetre, about 13 mm of untwist maximum on Cat6A, and use a jack rated for the same category as the cable. If NEXT fails on one pair only, re-terminate that pair's end first; if it fails on all four pairs at both ends, the jacks themselves are suspect — a Cat5e jack on a Cat6A run will fail NEXT no matter how neatly it is terminated. And remember the tester measures both ends: a NEXT fail reported at the far end is fixed at the far end.

Return loss: the mismatch detector

Return loss measures signal reflected back toward the tester by impedance changes along the link. A clean, uniform run reflects very little; a kink, a crush, a bad crimp, or a component of the wrong category creates a reflection. Return loss failing across all pairs points at a damaged section of cable or a mismatched component in the path; failing on one pair points at that pair's termination. The most common real-world cause in a data center is mixed components — a Cat6A cable terminated into a Cat6 jack, or a patch cord of a lower category in a channel test. If the run passes return loss but fails NEXT, the problem is the termination geometry; if it fails return loss, something in the path does not match.

Length and insertion loss: check the settings before the cable

A length failure is the parameter most often misread, because the tester does not measure length with a tape — it calculates it from signal delay and the NVP (nominal velocity of propagation) you selected. Pick the wrong NVP for the cable and an 85 m run reports as 95 m and fails. Before you believe a length fail, confirm the NVP matches the cable manufacturer's datasheet for that exact cable type, and confirm you selected the right test — permanent link (90 m limit) versus channel (100 m, including patch cords). A genuine over-length run usually shows up in insertion loss too: loss grows with length and frequency, so a run that is long and lossy at high frequencies is genuinely long. A run that fails length but passes insertion loss with margin is a settings error until proven otherwise. Also check for a coiled service loop: a 10 m coil in the ceiling is 10 m of length the tester counts.

The pass/fail call: read the margin, not just the verdict

A certifier gives three verdicts, not two: PASS, PASS* (pass with a margin inside the tester's accuracy — a marginal pass), and FAIL. A PASS* is a pass, but it is a pass with no headroom: the link meets the limit today and may not after the next re-patch or the next summer's heat, because insertion loss rises with temperature. Record the worst-case margin — the smallest margin across all parameters and frequencies — not just the word PASS. A link that passes with 6 dB of worst margin is a different asset from one that passes with 0.4 dB, and the next tech on the ticket deserves to know which one they are inheriting. When a link fails, record the failing parameter, the pair, the frequency it failed at, and the margin — "NEXT fail pair 3-6 at 350 MHz, -2.1 dB" — because that line alone tells the next person where to look.

A certification check before every handover

1. Right test, right limit. Permanent link (90 m) or channel (100 m)? TIA Cat6A or ISO Class EA? Confirm on screen before the first test.
2. Right NVP. Match it to the cable datasheet for that exact cable — a wrong NVP makes a good run fail length.
3. Calibrated tester, good reference cords. Reference cords wear out; a failing cord makes good links fail. Replace them on schedule, not when they start lying.
4. Wiremap first. Fix every wiremap fault at the termination and re-test before reading any other parameter.
5. One fix, one re-test. Re-terminate one end, re-test, and read the new worst margin — changing both ends at once hides which end was at fault.
6. Record the margin. Worst-case margin, failing parameter and frequency if any, and the test limit — the verdict alone is not a record.

Recording the certification in the ticket

A bare FAIL tells the next tech nothing — the parameter and the margin tell them exactly where to look. The minimum useful line records the link, the test limit, the worst margin, and what changed:

Rack B12 port 14 ↔ Patch panel A-07, Cat6A,
permanent link, 62 m.
NEXT FAIL pair 3-6 at 350 MHz, margin -2.1 dB.
Re-terminated far-end jack (untwist was ~25 mm).
Re-test PASS, worst margin 4.8 dB (NEXT, 500 MHz).

The shape to copy: link, test limit, length, failing parameter plus pair plus frequency plus margin, the fix, re-test verdict with worst margin. Six fields, and the next person on the ticket can see whether the link is solid or marginal without re-testing it — which is the whole point of certifying it in the first place.

A question for you

What's the most common reason a copper link fails certification in your experience — wiremap, NEXT, or something else?