Open app

Split pair, continuity fine, NEXT gone

common

“My little tester says it's wired perfect but the certifier fails it hard on crosstalk.”

Likely causes

  • Voice-style (USOC) punch: right pins, wrong pairs

    ~40%

    Wiremap, length and continuity all pass because USOC is straight through pin to pin. The pairing is the giveaway: USOC nests its pairs 4-5, 3-6, 2-7 and 1-8, so 3,6 and 4,5 test clean while 1,2 and 7,8 are each made of conductors from two different pairs. Look for white-brown/brown on pins 1 and 8 and white-green/green on 2 and 7. Every failing combination includes 12 or 78.

  • Punched straight down the colour order instead of the jack sequence

    ~25%

    Blue lands on 1,2. And blue belongs on 4,5 in every T568 scheme, so that alone calls it. This habit splits 3,6 and 4,5 while leaving 1,2 and 7,8 intact, so 45 shows up in the failing combinations. Same signature on every jack punched that shift; it's a habit, not a slip.

  • Single conductors drawn from two different pairs at one end

    ~15%

    Pin numbers are correct end to end so continuity and wiremap look clean, but one conductor of 3,6 came off green and the other off orange. The certifier names the split. This is the one that survives a quick visual, because the pins are right.

  • Coupler, line doubler or splitter in the path wired USOC

    ~12%

    Fault vanishes when the device is removed, and measured length includes a short stub the drawing does not show.

  • Two-pair jack or voice splitter left in line

    ~8%

    Opens on 4,5 and 7,8 alongside the crosstalk failure.

What to bring

  • Certifier
  • Continuity tester (to show the customer why it lied)
  • Punch tool with correct blade
  • Jack instruction sheet (phone)
  • Inspection light

Safety

Split pairs live on reused and voice-era runs, which is exactly where an unrecorded live PoE port turns up on the far end. Unplug the switch port before you punch: breaking a loaded pair arcs and pits the contacts. Type 4 is four-pair by definition and Type 3 may be two-pair or four-pair, so treat all four pairs as live until the port is out (IEEE 802.3 cl. 145).

Steps

  1. Strip every device out of the path and retest

    Couplers, surface boxes, splitters, line doublers, old adapters. Free, fast, and it is the answer often enough to try first.

    If that doesn’t do it

    Path is clean end to end: the split is in a termination.

  2. Read which pair combinations fail, not just the verdict

    A certifier names a split pair; a continuity tester cannot see one because every pin is electrically correct. Then read the combinations: 12 and 78 in the failures points at a USOC punch, 45 in them points at a straight colour-order punch. 12-36 fails under both habits, so on its own it tells you nothing.

    If that doesn’t do it

    Combinations are ambiguous: go read colours against pins at both ends.

  3. Check colours against pins at both ends

    What holds in both T568A and T568B: 4,5 is always the blue pair, 7,8 always brown, and 1,2 and 3,6 must each be the two conductors of one pair. Orange on 3,6 in A, green on 3,6 in B. Blue or brown appearing on 1,2 or 3,6, or one green and one orange conductor sharing 3,6, is your split however right the pin-to-pin map looks.

    If that doesn’t do it

    Both ends correct: you have an unrecorded block or punch field mid-run.

  4. Re-punch the offending end to the jack sequence

    Unplug the switch port first: reused and voice-era runs are where live PoE hides. Punch off the jack body, keep pairs twisted to the contact, retest before you close the plate.

    If that doesn’t do it

    Still split: the other end is wrong too. Re-punch it as well.

  5. Hunt the mid-run block

    Old 110 or 66 field, a consolidation point, a furniture feed. Anywhere the run was broken, someone may have landed it voice-style.

    If that doesn’t do it

    Nothing found: place the fault with whatever distance diagnostics your unit has (TDR length, return loss anomaly, crosstalk trace if fitted) and walk to that distance.

  6. Retest and keep the result

    Print or save the passing result against the same link ID. This is the fault that gets argued about later.

    If that doesn’t do it

    NEXT improves but still fails: you had a split plus a dress problem. Go to NEXT failing at the panel end.

References

  • ANSI/TIA-568.2-D cl. 6 (field test requirements)
  • ANSI/TIA-568.1-D cl. 6 (horizontal cabling)
  • IEEE 802.3 cl. 40 (1000BASE-T, all four pairs in use)

More Copper & Certification guides

All 54 guides