Network wiring notes - 8P8C / RJ45
For other network related things, see:
|
What you were probably looking for: Ethernet
(...for the pedants: mainly for 10Base-T, 100BaseTX, and 1000Base-T which are but three of a good deal more methods of the wider concept of Ethernet, but the others are much less relevant to consumer networking)
With pin positions are counted from left to right with the contacts pointing up, clip on the back, and the and pointing up (cable coming out the bottom):
Used by Used by Color (568B) Pin Color(568A) 10/100Mbit 1Gbit/10gBit Orange-white 1 Green-white ✓ ✓ Orange 2 Green ✓ ✓ Green-white 3 Orange-white ✓ ✓ Blue 4 Blue ✓ Blue-white 5 Blue-white ✓ Green 6 Orange ✓ ✓ Brown-white 7 Brown-white ✓ Brown 8 Brown ✓
You care mainly about 568B, because yout typically want to make a straight cable, a.k.a. patch cable, which is 568B-to-568B.
The other type is a crossover cable (568A-to-568B). There were useful in the 10/100 era when switches were pricy, and you wanted to connect a NIC to another NIC directly. Any gigabit-capable NIC can do this crossover internally, making life simpler and these cables mostly unnecessary.
Also, yes, 568A-to-568A is functionally identical to 568B-to-568B but really only a way to confuse yourself.
When wiring plugs
- Cut the outer insulation (that holds the cables together)
- Untangle the wire ends, so that they won't pull back when crimped
- ideally, for ≥1gBit it is suggested that you keep the untwisted length as small as possible, to minimize near-end crosstalk[1]
- order them according to the colors you want
- cut them to equal length (if more than slightly different)
- ideally, the outer insulation should make it into the plug far enough so that the crimping digs into it - this provides a little more protection against pull the cable. which means the wires should stick out ~15mm from the sheath. And you want to do that at this stage, because ordering them will change length a little.
- insert into plug
- probably check the order again, they may have jumped around, and checking is faster than redoing
- check that all the wires go to the end (seeing shiny copper from the tip is a good sign)
- use a crimping tool, which shoves the sharp part of the pins into the wires
- (optional: put some solid glue in the socket, to spread force when accidentally pulling the cable)
Notes:
- technically, there are different plugs for braided and solid-copper wiring, that clamp/pierce the wire in different ways (see e.g. stranded solid contacts)
- most of the time you'll work with solid
- (in home use you can often get away with mixing this, in professional use too much of your network may depend on doing one thing properly)
- other things that may be using 8P8C includes
- ISDN
- RS-232 (varied pinouts)
- FireWire
- DMX
- ...and many others. It's a convenient plug.
More notes on ethernet wiring
On standards
The wiring used on 10Mbit, 100Mbit (specifically 10-BASE-T and 100-BASE-TX) ethernet over 8P8C (informally RJ45) plugs is defined by TIA/EIA-568-B, which define two plug wiring alternatives, 568A and 568B.
Notice the lack of dashes; 568-B is the standard that 568A and 568B are part of, 568-A a completely different standard (yes, that's stupidly confusing).
On coexistence
- 10Mbit and 100Mbit networking use only pair 2 and 3 (orange and green) in a 568-style cable
- Countries using analog phones with 4P or 6P plugs (in 8P sockets) can have analog phones and FE coexist, because they use only pair 1 (blue, in the center). Large setups can do this to make their wiring simpler
- It's also sometimes used for the (somewhat riskier) DIY variant of power over ethernet
- on-the-cheap two-pair cables will work in most places, but only because the NICs fall back to 100MBit
- 1GBit and 10gBit ethernet uses all four pairs in a 568-style cable, so can't do the above trick,
- You want cables rated Cat5e or better for 1gBit, Cat6 or better for 10gBit
- so you can put in the wiring for whatever highest speed you expect to use in the next X years, and can do the actual speed upgrade later by replacing only the switch
On loopbacks
Loopbacks connect a port to itself.
It's useful to check
- whether a patchbay was wired properly
- whether a wallplug or the cable to the nearest switch is broken
- whether a (long) cable is broken
- whether a switch/router port is broken
...all mostly by whether the link light lights up.
However, it became a much less useful when gigabit became standard.
Gigabit NICs have crosstalk detection (detects how much signal interferes onto other wires), and will often decide that a loopback cable/plug is actually an extreme amount of crosstalk - any may not show link at all.
Which defeats the point entirely, and unless you know you can and have disabled crosstalk detection on the NIC and switch, you won't know what lack of link-light means.
There were always nicer tools for checking cables and sockets, and these days that's what you want to have instead.
If you still want to make one:
Connect: (If you're wiring a plug as a loopback, make sure you're not confused about which end is pin is pin 1)
Pin 1 to 3 Pin 2 to 6 (Pin 4 to 7 for a gBit loopback) (Pin 5 to 8 for a gBit loopback)
You could take an existing cord (e.g. one that has one the retainer clip broken), cut it in half, and connect:
Orange-white to Green-white Orange to Green (Blue to Brown-white for a gBit loopback) (Brown to Blue-white for a gBit loopback)
(This is the same for both 568A and 568B)
Other 8P8C wiring
Phone
FireWire
ISDN
DMX
RS-232
(varied pinouts)
More technical
On crossover cables
On cable standards (Cat5, etc.)
According to the specs:
- Cat5
- rarely seen - it has fallen out of favour and is barely sold
- ...so informally refers to Cat5a
- Cat5a
- Currently still quite common
- designed to support 100 MHz signals
- e.g. (originally) 100mBit at <100m
- but e.g. decent cable is fine to do 1gBit at <100m
- Cat6
- designed to support 250MHz signals
- 100mBit, 1gBit at <100m
- e.g. <55m for 10gBit (less if many are bundled)
- Cat6a
- designed to support 500MHz signals
- e.g. 10GBit at <100m
- Cat7
- designed to support 600MHz signals
- stricter about crosstalk (pairs individually insulated)
- Cat7a
- designed to support 1000MHz signals
- Cat8 (verify)
Notes:
- Any cabling that is not shielded will crosstalk
- with itself (hence pair shielding, see below)
- with others, meaning permissible distances are lower when many are bundled (relevant to company wiring)
- length of a single cable seems to be restricted more by propagation speed than e.g. category specs
- hence commonly seeing ~100m regardless of exact cable/speed. It is a generic figure to keep in mind.
- cable quality is usually more about cheap cables that don't actually conform
- So a decent brand may matter more than category specs
- category specs do become more relevant at higher speeds
- bringing in some tradeoff between speed and range
- shielded cable may in some environments make a larger difference than category
- e.g. when you're putting it close to (significant) power lines
- Often called copper, but need not be pure.
- eg. CCA for Copper Clad Aluminium, CCS for Copper Clad Steel
- which needs to be a little thicker to meet the resistance-per-length requirement
- and may indicate lower quality in general - you may want to avoid it (verify)
In practice
Companies may want Cat6 (or perhaps Cat6a) for future compatibility, to be able to use 1gBit and 10gBit now (also somewhat assuming that the next big speed upgrade is still over copper TP).
Cat7 can go beyond 10gBit at short-ish distances, but chances are you won't be needing that any time soon. Only data centers might care.
Home LAN can often still get away with Cat5a for gBit, but if you're pulling new wires, you may as well go for Cat6 (little price difference now, good for the next switch/speed upgrade or two).
On cable shielding
Cables with shield
tl;dr:
- at home you don't need shielded cable (unless maybe you're doing 10gBit)
- it seems that using shielded cable, even if that shield is not grounded, will still noticeably reduce RFI(verify)
- in large installations you need your network people to think like electricians, and preferably be certified
Home use is a fairly friendly environment, so UTP (Unshielded Twisted Pair) is typical, and fine for 1gBit in most cases.
Some variants like F-UTP (foiled-whole) is not much more expensive, so is also seen.
With larger installations, longer cables, interference becomes more relevant.
At higher speeds, crosstalk between pairs becomes relevant.
Having powerful electronics and sensitive electronics near each other may also cause you some headaches.
So shielded versions become interesting.
The more-letter variants (look like U/STP or U-STP) are more specific (Earlier three-letter acronyms aren't entirely consistent, in part because the naming first got confusing, got used inconsistently, and got changed. Basically, anything with three letters (except UTP) is potentially not what you think).
The parts:
- overall shielding type
- U - unshielded
- S - braided shield
- F - overall foil
- SF - braid and foil
- - or /
- pair shielding type (to reduce crosstalk, common on long cables, 10gBit cables(verify))
- U - unshielded
- F - foil around each pair
- TP (twisted pair)
Notes:
- prefabbed patch cables can make assumptions (e.g. that certain shielding is not relevant when short and probably not bundled)
- assumptions you cannot an make when wiring a building.
- When planning for higher speeds (e.g. 10gBit), it's sometimes handier to go slightly overkill
- not because it's typically necessary, but because you may avoid some specific problems.
- The common types may be U-UTP, F-UTP, U-FTP, S-FTP, SF-UTP (verify)
- To make things even more interesting, there are 'industry standard' names that looks imilar but are different
- some are pretty clear, e.g. UTP is U/UTP
- others are fuzzier, e.g.
- STP can refer to U/FTP, F/UTP, S/UTP, SF/UTP
- SSTP can refer to S/FTP, or to SF/FTP
- SFTP can refer to SF/UTP, S/FTP, SF/FTP
- see e.g. https://en.wikipedia.org/wiki/Twisted_pair#Cable_shielding
Naming pendantics and telephony
When we say RJ45, we often mean something like "Ethernet wiring on an 8P8C plug."
RJ is the group of plugs that can be described by their positions and connectors, such as 8P8C in Ethernet, while RJ45 actually refers to a specific telephone wiring on the 8P-style plug (probably the most common one among several), while 8P8C refers to that plug itself and no specific wiring. Regardless, most people call the plug RJ45, regardless of wiring.
See also Common plugs and connectors#Modular_connector_.28.3FP.3FC.29.3B_Registered_Jack
Plugs may have fewer actually present conductors than they have positions, so 8P2C, 8P4C, 8P6C, 8P8C, 6P2C, 6P4C, 6P6C, 4P2C, 4P4C all exist.
When there are less connectors than positions, they are in the middle positions; RJ-style wiring is from the middle out.
For most of us, the is interesting only in that you can plug a phone with 6P plugs into a 8P (ethernet-plus-phone) socket and have the phone work - the clip aligns the plug in the middle.
If more more than the middle two wires are used in telephone wiring, they carry either power, or a second (RJ14) or even third (RJ25) telephone line on the same wire, but consumers rarely see this type of phone wiring.
There are exceptions to the 'always start in the middle', but they tend to be intentionally working around RJ-style wiring.
The most common use of 6P outside of the US is probably phone wiring according to RJ11, which often use just a single pair in the middle. In the US, 8P connectors with the RJ45 phone wiring is common.