Alternative A
pins 1/2pins 3/6
The data pairs of 10/100. Power rides as a common-mode voltage on the transformer centre taps, under the signal.
Reference
What the number on the cable actually buys you
A cable category is a bandwidth rating, measured in megahertz and verified over a 100 metre channel. It is not a speed rating. Every confusion in this subject comes from reading it as one.
Cat6 is 250 MHz. 10GBASE-T occupies 500 MHz. The two numbers do not have to meet for a link to come up — the PHY will negotiate, run, and pass traffic — so a 10-gigabit switch plugged into a 90 metre Cat6 drop looks like it works right up until the error counters are read. TIA TSB-155-A puts that link at 37 metres, or up to 55 with a favourable crosstalk environment. The cable is not faulty and the switch is not faulty. The channel was never specified to do it.
This page is the map between the two halves: what each category specifies, what each IEEE PHY demands, and what happens where they do not line up. The reach matrix is computed from the categories and the PHY requirements rather than typed out, and the three documented concessions are held in a separate table so that a concession cannot quietly be presented as a specification.
The categories
Two standards bodies publish these and they do not agree on what exists. ANSI/TIA-568 names categories; ISO/IEC 11801 names channel classes and runs a parallel numbering. Category 7 and 7A are ISO/IEC only — TIA skipped from Cat6A straight to Cat8 — and both need a connector that is not an RJ45 to reach their rated bandwidth.
| Category | Bandwidth | ISO class | Recognised by | Connector | Channel | Status |
|---|---|---|---|---|---|---|
| Cat3 | 16 MHz | Class C | TIA + ISO | RJ45 | 100 m | Withdrawn |
| Cat5 | 100 MHz | Class D | TIA + ISO | RJ45 | 100 m | Withdrawn |
| Cat5e | 100 MHz | Class D | TIA + ISO | RJ45 | 100 m | Current |
| Cat6 | 250 MHz | Class E | TIA + ISO | RJ45 | 100 m | Current |
| Cat6A | 500 MHz | Class EA | TIA + ISO | RJ45 | 100 m | Current |
| Cat7 | 600 MHz | Class F | ISO | GG45 / TERA | 100 m | ISO/IEC only |
| Cat7A | 1000 MHz | Class FA | ISO | GG45 / TERA | 100 m | ISO/IEC only |
| Cat8.1 | 2000 MHz | Class I | TIA + ISO | RJ45 | 30 m | Current |
| Cat8.2 | 2000 MHz | Class II | ISO | GG45 / TERA | 30 m | ISO/IEC only |
- Cat3 — Withdrawn for new installation, still found feeding telephone outlets and older building systems. Two pairs of it will carry 10BASE-T and nothing faster.
- Cat5 — Superseded by Cat5e in TIA-568-B and withdrawn. Same 100 MHz, but without the return loss and ELFEXT limits that four-pair bidirectional transmission turned out to need.
- Cat5e — The floor for anything modern, and still the correct answer for a gigabit drop. The "e" bought return loss, ELFEXT and PSELFEXT limits, not more bandwidth.
- Cat6 — Two and a half times the bandwidth of Cat5e and the most commonly mis-sold category on this list — it is a 10-gigabit cable only over a short, quiet channel.
- Cat6A — The first category specified with alien crosstalk limits, which is exactly what 10GBASE-T needs and what Cat6 lacks. This is the real 10-gigabit cable.
- Cat7 — ISO/IEC Class F. ANSI/TIA-568 has never recognised it. Its full performance needs a non-RJ45 connector, so "Cat7 patch leads" with RJ45 plugs on them are a Cat6A channel with an expensive cable in the middle.
- Cat7A — ISO/IEC Class FA. Same connector problem as Cat7, and no IEEE PHY was ever specified against it — 802.3bq went to Category 8 instead.
- Cat8.1 — Built on Cat6A geometry and backwards compatible with RJ45, which is why it is the one of the two that shipped. The 30 m channel is the trade for 2 GHz.
- Cat8.2 — Built on Cat7A geometry, so it inherits the non-RJ45 connector and, with it, the near-total absence from the market outside a few European installations.
The PHYs, and what each one asks for
The other half of the equation. Note where the pair count changes:10BASE-T and 100BASE-TX use two pairs, everything from gigabit up uses four, bidirectionally, with echo cancellation separating the two directions on the same conductors. That is the moment a split pair stops being survivable, and the moment the spare pairs stopped being spare.
| PHY | IEEE | Year | Rate | Pairs | Symbol rate | Line coding | Min. category |
|---|---|---|---|---|---|---|---|
| 10BASE-T | 802.3i | 1990 | 10 Mbit/s | 2 | 20 MBd | Manchester | Cat3 |
| 100BASE-TX | 802.3u | 1995 | 100 Mbit/s | 2 | 125 MBd | 4B5B / MLT-3 | Cat5 |
| 1000BASE-T | 802.3ab | 1999 | 1 Gbit/s | 4 | 125 MBd | PAM-5, all four pairs bidirectional | Cat5e |
| 2.5GBASE-T | 802.3bz | 2016 | 2.5 Gbit/s | 4 | 200 MBd | PAM-16 / DSQ128 | Cat5e |
| 5GBASE-T | 802.3bz | 2016 | 5 Gbit/s | 4 | 400 MBd | PAM-16 / DSQ128 | Cat6 |
| 10GBASE-T | 802.3an | 2006 | 10 Gbit/s | 4 | 800 MBd | PAM-16 / DSQ128, LDPC | Cat6A |
| 25GBASE-T | 802.3bq | 2016 | 25 Gbit/s | 4 | 2000 MBd | PAM-16 / DSQ128, LDPC | Cat8.1 |
| 40GBASE-T | 802.3bq | 2016 | 40 Gbit/s | 4 | 3200 MBd | PAM-16 / DSQ128, LDPC | Cat8.1 |
- 10BASE-T — Two pairs, one each way. The only PHY here that leaves half the cable spare — which is why Alternative B power exists at all.
- 100BASE-TX — Still two pairs. A 100 Mbit/s link will come up over a badly terminated Cat5e drop that gigabit refuses, which is the classic misleading symptom.
- 1000BASE-T — The break point. Four pairs, each carrying 250 Mbit/s in both directions at once, separated by echo cancellation — so every pair now has to be a real pair.
- 2.5GBASE-T — 10GBASE-T run at a quarter rate, specifically so it would fit down the Cat5e already in the walls. The reason a Wi-Fi 6 access point does not need a recable.
- 5GBASE-T — The same PHY at half rate. Specified against Cat6 for the full 100 m; Cat5e is a conditional case rather than a promise.
- 10GBASE-T — Occupies 500 MHz, which is precisely the Cat6A number. This is the PHY the alien crosstalk limits were invented for.
- 25GBASE-T — Data-centre top-of-rack. Thirty metres is not an oversight — it is the distance at which 2 GHz on copper stops being worth the transceiver.
- 40GBASE-T — Same channel, same 30 m, twice the symbol rate. Widely specified, rarely deployed — the market went to 25G and to fibre.
How far each one actually runs
Row is the cable, column is the PHY. Every cell is computed from the category's rank, the PHY's declared minimum and the concessions table below — none of it is typed in. Conditional means a named document permits it subject to measurement, not that it will work.
| cable ↓ PHY → | 10BASE-T | 100BASE-TX | 1000BASE-T | 2.5GBASE-T | 5GBASE-T | 10GBASE-T | 25GBASE-T | 40GBASE-T |
|---|---|---|---|---|---|---|---|---|
| Cat3 | 100 mFull channel | —Not specified | —Not specified | —Not specified | —Not specified | —Not specified | —Not specified | —Not specified |
| Cat5 | 100 mFull channel | 100 mFull channel | 100 mConditional | —Not specified | —Not specified | —Not specified | —Not specified | —Not specified |
| Cat5e | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mConditional | —Not specified | —Not specified | —Not specified |
| Cat6 | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 37–55 mShortened | —Not specified | —Not specified |
| Cat6A | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | —Not specified | —Not specified |
| Cat7 | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | —Not specified | —Not specified |
| Cat7A | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | 100 mFull channel | —Not specified | —Not specified |
| Cat8.1 | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel |
| Cat8.2 | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel | 30 mFull channel |
The three concessions, and what they are worth
- 1000BASE-T over Cat5 — 100 m. IEEE 802.3ab names Category 5. In practice Cat5e exists because gigabit exposed return loss and ELFEXT as the limiting parameters, and Cat5 does not specify either. Source: IEEE 802.3ab (1999); ANSI/TIA-568-B.2, which is guidance on assessing installed cabling rather than a specification you can order to.
- 5GBASE-T over Cat5e — 100 m. Supported where the alien crosstalk environment allows it. TIA TSB-5021 is the assessment procedure; an untested Cat5e channel is not a 5-gigabit channel until it has been measured. Source: TIA TSB-5021, which is guidance on assessing installed cabling rather than a specification you can order to.
- 10GBASE-T over Cat6 — 37–55 m. Up to 37 m unconditionally; 37 m to 55 m depending on the alien crosstalk environment; beyond 55 m only with mitigation. This is the single most commonly misquoted figure on the page. Source: TIA TSB-155-A.
Everything else in the matrix is the plain rule: at or above the PHY's minimum category, the full channel; below it, not specified. "Not specified" is not the same as "will not link" — it means nobody has undertaken that it will, and no error counter is obliged to tell you which of the two you have.
Reading the screening letters
ISO/IEC 11801 designates a construction as XX/YTP:the code before the slash is the overall screen, the code after it is the per-pair screen. U is none, F is foil, S is braid, SF is braid over foil. Once you know that, every label on every drum decodes without having to be memorised.
"STP" is not one of them. It is a colloquialism that has been used for F/UTP, S/FTP and, historically, the 150 Ω IBM Type 1 cable that is not this system at all. A supplier who writes STP on a quotation has told you the cable has a screen somewhere and nothing else — which of the eight constructions below it is remains an open question.
The two screens do different jobs and it is worth being deliberate about which one you are buying. An overall screen keeps outside interference out. A per-pair screen attacks crosstalk between the pairs inside the jacket — the thing that limits 10GBASE-T — which is why per-pair foil turns up in Cat6A and Cat8 and not below.
| Designation | Overall | Per pair | Also sold as | What it is for |
|---|---|---|---|---|
| U/UTP | No screen | None | UTP | The default, and the right answer for most drops. Rejects interference by twist geometry alone, which works better than its reputation suggests. |
| F/UTP | Foil | None | FTP, ScTP, "STP" | One foil around all four pairs. The usual choice where a cable has to share a containment with mains, and the usual Cat6A construction. |
| S/UTP | Braid | None | — | A braid with no foil. Rare in Cat-rated cable; the braid alone is poor at the frequencies that matter above 100 MHz. |
| SF/UTP | Braid over foil | None | S-FTP, PiMF (wrongly) | Braid over foil, pairs unscreened. Common in industrial and broadcast cable where the jacket takes physical abuse and the screen has to survive it. |
| U/FTP | No screen | Foil | PiMF | Each pair in its own foil, no overall screen. Attacks crosstalk between pairs rather than interference from outside — which is why it turns up in Cat6A and Cat8. |
| F/FTP | Foil | Foil | — | Foil on each pair and one over the lot. Full screening without the bulk and cost of a braid. |
| S/FTP | Braid | Foil | S-STP, "double shielded" | Foil per pair, braid overall. The usual Cat7 and Cat8.2 construction, and about as good as balanced copper gets. |
| SF/FTP | Braid over foil | Foil | — | Everything, everywhere. Specified where the electromagnetic environment is genuinely hostile and nobody wants to argue about it afterwards. |
Rows in grey are legal under the grammar and rare in the catalogue. The four in black cover essentially everything you will be quoted.
T568A, T568B, and why it does not matter
Two termination standards, one difference: the orange and green pairs swap places. The pair membership is identical — pins (1,2), (3,6), (4,5) and (7,8) in both — which is the whole reason a cable terminated the same way at both ends works regardless of which one you picked.
- Pick one and use it everywhere. T568B is the overwhelmingly common commercial termination; T568A is specified in some government and residential work. Neither performs better. A site that uses both will eventually produce a cable with one of each on it.
- An A-at-one-end, B-at-the-other cable is a crossover. It transposes the orange and green pairs, which for 10/100 is exactly the old crossover cable. Since 1000BASE-T and Auto-MDI-X this is no longer useful and no longer harmful — it is just a cable that is hard to account for later.
- Untwist is the parameter nobody measures. Cat5e tolerates about 13 mm of untwisted conductor at the termination; Cat6 and above want less. It costs nothing at the time and it is invisible on a continuity test — it shows up as NEXT failures on a certifier and as a link that negotiates down under load.
- Split pairs pass a continuity test. Wire pin 3 to pin 3 using a conductor from the blue pair and every pin-out light on a cheap tester goes green. The link will run 10/100 and fail gigabit, because the balance the twist provides was never there.
Power over Ethernet
Power rides as a common-mode voltage on the transformer centre taps — both conductors of a pair sit at the same DC potential, so the differential signal on top of it is undisturbed. That is why PoE and data share conductors without a filter, and why the pairs have to be whole for it to work.
| Type | Marketed as | IEEE | Year | At the PSE | At the PD | Pairs | Classes |
|---|---|---|---|---|---|---|---|
| Type 1 | PoE | 802.3af | 2003 | 15.4 W | 12.95 W | 2 | 0–3 |
| Type 2 | PoE+ | 802.3at | 2009 | 30 W | 25.5 W | 2 | 4 |
| Type 3 | PoE++ | 802.3bt | 2018 | 60 W | 51 W | 4 | 5–6 |
| Type 4 | PoE++ | 802.3bt | 2018 | 90 W | 71.3 W | 4 | 7–8 |
The gap between the two power columns is cable loss, and it is the budget you are actually designing to. A Type 4 injector delivers 90 W and the device is guaranteed only 71.3 W of it at 100 m.
Alternative B
pins 4/5pins 7/8
The pairs 10/100 leaves idle. Gigabit and faster use all four regardless, so on those links both alternatives are phantom-powered data pairs.
- The screen is not the return. PoE current goes out on one pair and back on another. A screened cable's screen carries no PoE current in normal operation, and a cable with a damaged pair does not fail safe — it fails hot.
- Copper-clad aluminium is the one to refuse. CCA fails the category specifications outright, and its DC resistance is roughly 1.5 times that of copper. Under a Type 3 or Type 4 load that is heat in a cable that was sold as a bargain. It is also brittle enough to break inside an IDC.
- Type 3 and Type 4 need all four pairs. A four-pair-power device on a two-pair link negotiates down to Type 2 at best, and on a two-pair-only patch lead it may simply not power up.
What reaches the far end
The gap between the two power columns above is cable loss, and it is the budget you are really designing to. This works it out for a specific run — and shows the one structural reason 802.3bt went to four pairs.
Loop resistance——
Lost in the cable——
At the device——
On two pairs instead——
This calculator needs JavaScript. The table above gives the standard's own PSE and PD figures.
The standard's own gap is very nearly this arithmetic. Run a 90 metre channel of 23 AWG at 52 V through the calculator and Type 4 delivers about 72 W against the 71.3 W the standard guarantees, and Type 3 about 52 W against 51 W. The distance between the PSE and PD columns is not a margin somebody chose — it is the resistive loss of a full-length channel. The build asserts it stays that way.
Resistive loss only, at a PSE voltage you set — the standards specify a minimum output voltage rather than a fixed one, and a higher voltage carries the same power as less current and so less loss. Connector and patch-panel resistance are not counted, and neither is the temperature coefficient of copper, which makes a hot bundle worse than this says.
The things that reach site
The 100 metres is 90 plus 10. The channel is a 90 m permanent link of solid-core horizontal cable plus up to 10 m of stranded patch lead at the two ends combined. Stranded conductor is more flexible and measurably more lossy — the 10 m allowance is what pays for that. A 100 m run made of patch leads is not a 100 m channel.
Solid core does not belong in a patch lead. It work-hardens and fractures inside the plug after a few dozen flexes, and the fracture is intermittent rather than open, which is the worst way for a cable to fail. On a touring rack, use stranded and accept the shorter reach.
etherCON is a shell, not a standard. Neutrik's etherCON is a rugged latching housing around an ordinary RJ45; the cable inside it is whatever category was pulled through. It buys mechanical retention and strain relief on a connector that was designed for an office wall, which on a stage is worth a great deal — but it does not upgrade a Cat5e drum to Cat6A.
Sources
- Category specifications and the 30 m Category 8 channel — ANSI/TIA-568.2-D, and ISO/IEC 11801-1 for the Class C to Class II designations and the screening grammar. Both are paywalled; the figures here are cross-checked against Fluke Networks' Category 8 fact sheet.
- 10GBASE-T over Category 6 at 37–55 m — TIA TSB-155-A, as summarised by Fluke Networks on 10GBASE-T field testing. The bulletin itself is paywalled.
- 2.5GBASE-T and 5GBASE-T cabling requirements — IEEE 802.3bz-2016, Clauses 125 and 126.
- PoE type and class powers — Ethernet Alliance, Overview of 802.3bt, cross-checked against Fluke Networks on Type 3 and Type 4.
- Category 7 / 7A and the TIA position — ISO/IEC 11801-1 recognises Class F and FA; ANSI/TIA-568 does not, and no IEEE BASE-T PHY is specified against either.
Assembled 7 September 2026 with AI assistance. The reach matrix is computed from the category and PHY tables above and guarded at build time against its own rules, but nothing here has been verified with a certifier on installed cable. Where a figure comes from a Telecommunications Systems Bulletin rather than a standard it is labelled conditional on this page, and that distinction is the one worth carrying into a specification.
Companion page: what is actually inside a four-channel CAT5 stagebox — the same eight conductors carrying four balanced audio channels, and the nine incompatible ways manufacturers map them.