Ethernet is a network technology developed under the IEEE 802.3 standard that defines how devices share data frames on the same network and how these frames are carried over the physical medium. It was first developed by Robert Metcalfe and his team at the Palo Alto Research Center (PARC) laboratory of Xerox in the 1970s. Today, almost all Local Area Network (LAN) infrastructures run on Ethernet, and applications such as Metro Ethernet have extended it beyond the LAN.
Ethernet should not be thought of as limited to a single speed or cable type. IEEE 802.3 defines copper cable standards such as 10BASE-T (10 Mbps), 100BASE-TX (100 Mbps), 1000BASE-T (1 Gbps) and 10GBASE-T (10 Gbps) together with fibre standards such as 100BASE-FX, 1000BASE-SX and 40GBASE-LR4. In this article, I explain how Fast Ethernet and Gigabit Ethernet operate over copper and fibre, how the CAT cable categories relate to these standards, and the limits to consider when selecting a cable.
1- Fast Ethernet Technology
Fast Ethernet was defined by the IEEE 802.3u standard in 1995, at a time when 10 Mbps Ethernet was starting to become insufficient. It provides a data rate of 100 Mbps and, because it does not change the Ethernet frame structure, it can operate on the same network as 10 Mbps devices. It is implemented by the 100BASE-TX standard over copper cable and by the 100BASE-FX standard over fibre.
⚙️ Operating Principles
Ethernet can operate in 2 different transmission modes. In half-duplex mode, a device either sends or receives data at any given time, and the frames of devices that share the same medium can collide. This mode uses CSMA/CD (Carrier Sense Multiple Access with Collision Detection) to manage collisions: the device listens to the line before it sends, stops the transmission if it detects a collision, and tries again after waiting for a random period.
In full-duplex mode, sending and receiving take place at the same time over separate paths. A device connected directly to a switch has no collisions, so CSMA/CD is disabled and the link runs at full speed in both directions. Today, almost all Fast Ethernet links on switch-based networks operate in full-duplex mode.
Auto-Negotiation determines the speed and the duplex mode that 2 devices use. While the link is being established, each side advertises the speeds and modes it supports, and both sides agree on the highest common value. If one side is set to a fixed value and the other is left on Auto-Negotiation, the result can be a condition called duplex mismatch, which causes heavy packet loss.
CSMA/CD works only in half-duplex mode. Full-duplex links have no collisions, so CSMA/CD is not used.
1.1- Fast Ethernet over CAT Cable
The implementation of Fast Ethernet over twisted-pair copper cable is the 100BASE-TX standard, which was the standard connection type of LAN infrastructures for many years. It uses only 2 of the 4 wire pairs in the cable and requires at least CAT5 cable.
|
Property |
Value |
|
Standard |
100BASE-TX (IEEE 802.3u) |
|
Data Rate |
100 Mbps (in each direction) |
|
Wire Pairs Used |
2 pairs (1 for transmit, 1 for receive) |
|
Encoding |
4B/5B and MLT-3 |
|
Duplex Mode |
Half-duplex and full-duplex |
The table below shows the maximum distance at which 100BASE-TX can operate for each cable category.
|
Cable Category |
Ethernet Standard |
Maximum Distance |
|
CAT5 |
100BASE-TX |
100 m |
|
CAT5e |
100BASE-TX |
100 m |
|
CAT6 |
100BASE-TX |
100 m |
|
CAT6a |
100BASE-TX |
100 m |
|
CAT7 |
100BASE-TX |
100 m |
|
CAT8 |
100BASE-TX |
30 m |
100 m is the channel length. This value consists of at most 90 m of permanent cabling (permanent link) and a total allowance of 10 m for the patch cords at the 2 ends.
1.2- Use of the Wire Pairs in 100BASE-TX
In 100BASE-TX, each wire pair carries data in a single direction. One pair is reserved for transmit (TX) and the other pair for receive (RX), and each pair has a positive (+) and a negative (-) terminal. The signal is carried as the voltage difference between these 2 terminals, so noise that affects both wires at the same time cancels out.
|
Wire Pair |
Pin (+) |
Pin (-) |
Function |
|
Pair 1 |
1 |
2 |
Transmit (TX) |
|
Pair 2 |
3 |
6 |
Receive (RX) |
|
Pair 3 |
4 |
5 |
Not used |
|
Pair 4 |
7 |
8 |
Not used |
The table shows the point of view of the computer's network adapter. The network adapter operates as MDI, so it sends on the 1-2 pair and receives on the 3-6 pair. A switch port operates as MDI-X and uses the same pairs in the opposite direction, which means that it receives on the 1-2 pair and sends on the 3-6 pair. As a result, on a straight-through cable the signal sent by the computer arrives directly at the receive pair of the switch. Connecting 2 similar devices, for example 2 computers or 2 switches, used to require a crossover cable. Today, the great majority of devices support Auto MDI-X, so they swap the pairs automatically and the link is established regardless of the cable type.
1.3- Encoding and Frequency in 100BASE-TX
The 100 Mbps data stream is not sent directly to the cable. It passes through 2 encoding stages. In the first stage, 4B/5B encoding converts every 4 bits of data into a 5-bit symbol. This conversion prevents signals that stay at the same level for a long time and so preserves the clock synchronisation of the receiver. The cost is that the line rate rises from 100 Mbps to 125 Mbaud.
If the 125 Mbaud signal were transmitted with 2 levels, the signal energy would spread to much higher frequencies. In the second stage, MLT-3 (Multi-Level Transmit 3) encoding is applied. MLT-3 moves the signal through 3 levels (-1, 0, +1) in sequence, and even with the fastest-changing bit pattern a complete signal cycle takes 4 symbols. For this reason, the fundamental frequency of the signal falls to 125 / 4 = approximately 31.25 MHz.
100 Mbps data → 4B/5B encoding → 125 Mbaud line rate → MLT-3 encoding → ~31.25 MHz fundamental frequency
31.25 MHz is a characteristic value used to show how the encoding reduces the signal frequency. It is not the only frequency of the signal. The real signal, together with its harmonics, occupies a spectrum that extends to approximately 80 MHz. This spectrum stays within the performance range of CAT5, which is defined up to 100 MHz per pair.
The spectrum applies to each wire pair separately and is not added up across the pairs. The transmit pair and the receive pair use the same frequency band on 2 separate physical paths. The 100 Mbps in each direction is carried only on the single pair that belongs to that direction.
1.4- Fast Ethernet over Fibre
The implementation of Fast Ethernet over fibre is the 100BASE-FX standard. It is preferred for links between buildings, where the 100 m limit of copper cable must be exceeded, and on production sites with heavy electromagnetic interference. Fibre is a dielectric medium, so it is not affected by the interference that sources such as motors and high-voltage lines generate.
IEEE 802.3u defines 100BASE-FX only for multimode fibre. Transmission takes place at a wavelength of 1300 nm and uses 2 fibres, each of which works in a single direction. You can find detailed information about fibre types and OM classes in my Fibre Optic Cable article.
|
Standard |
Fibre Type |
Wavelength |
Distance |
Defined In |
100BASE-FX |
Multimode (50/125 and 62.5/125 µm, OM1-OM5) |
1300 nm |
2 km (full-duplex) |
IEEE 802.3u |
100BASE-LX10 |
Single-mode |
1310 nm |
10 km |
IEEE 802.3ah |
The 2 km value applies to full-duplex operation. In half-duplex mode, CSMA/CD must be able to detect a collision in time, so the distance falls to a few hundred metres. If 100 Mbps is required over single-mode fibre, 100BASE-FX is not used. 100BASE-LX10, which is defined by IEEE 802.3ah, or vendor-defined long-distance optics are used instead.
2- Gigabit Ethernet Technology
Gigabit Ethernet was defined at a time when 100 Mbps was starting to create a bottleneck on backbone links and server ports. It provides a data rate of 1 Gbps in each direction and, because it keeps the Ethernet frame structure, it can operate on the same network as 10 and 100 Mbps devices. It is implemented by the 1000BASE-T standard over copper cable and by the 1000BASE-SX and 1000BASE-LX standards over fibre.
⚙️ Operating Principles
Half-duplex mode and CSMA/CD are defined in the Gigabit Ethernet standards, but they are not used in practice. At Gigabit speed the collision domain becomes so small that the standard had to introduce additional mechanisms such as Carrier Extension for half-duplex. Almost all Gigabit links in real networks operate in full-duplex mode through a switch.
In 1000BASE-T, Auto-Negotiation is used not only to determine the speed and the duplex mode, but also to determine which of the 2 ends of the link is the Master and which is the Slave. The Master side runs on its own clock source, and the Slave side recovers its clock from the incoming data stream. This relationship cannot be established without Auto-Negotiation. For this reason, a port is fixed at 1000 Mbps by leaving Auto-Negotiation enabled and advertising only the 1000BASE-T capability. If Auto-Negotiation is disabled completely, the 1000BASE-T link is not established.
2.1- Gigabit Ethernet over CAT Cable
The implementation of Gigabit Ethernet over twisted-pair copper cable is the 1000BASE-T standard. It was defined by IEEE 802.3ab in 1999, and because it can run on existing copper cabling infrastructures, it allowed Gigabit Ethernet to spread as far as the desktop. Unlike 100BASE-TX, it uses all 4 wire pairs in the cable, and each pair carries data in 2 directions at the same time. Each pair carries 250 Mbps, and the 4 pairs together provide 1 Gbps in each direction.
Sending and receiving on the same pair at the same time is possible because of hybrid circuits and echo cancellation. Each end knows the reflection of its own transmitted signal that returns from the line, so it subtracts this reflection from the received signal and separates the data that comes from the other side. The signal is carried on each pair at 125 Mbaud with 5-level PAM-5 encoding. Half of this symbol rate, 62.5 MHz, is the characteristic value used to explain the encoding. The transmission spectrum of the PAM-5 signal, however, is designed to stay below 80 MHz.
1 Gbps data → distribution across 4 pairs → 250 Mbps per pair → PAM-5, 125 Mbaud → spectrum below 80 MHz
Because 1000BASE-T uses the 4 pairs in 2 directions at the same time, the far-end crosstalk between the pairs (ELFEXT) and the return loss become important. CAT5e is the first category that also defines these values, so it is the lowest cable category required for 1000BASE-T.
|
Property |
Value |
|
Standard |
1000BASE-T (IEEE 802.3ab) |
|
Data Rate |
1 Gbps (in each direction) |
|
Wire Pairs Used |
4 pairs, each bidirectional |
|
Encoding |
PAM-5, 125 Mbaud per pair |
|
Duplex Mode |
Full-duplex |
|
Auto-Negotiation |
Required to determine the Master and the Slave |
The table below shows the maximum distance at which 1000BASE-T can operate for each cable category.
|
Cable Category |
Ethernet Standard |
Maximum Distance |
|
CAT5e |
1000BASE-T |
100 m |
|
CAT6 |
1000BASE-T |
100 m |
|
CAT6a |
1000BASE-T |
100 m |
|
CAT7 |
1000BASE-T |
100 m |
|
CAT8 |
1000BASE-T |
30 m |
2.2- Use of the Wire Pairs in 1000BASE-T
In 1000BASE-T, the pairs are not divided into transmit and receive. In 100BASE-TX, each pair works in a single direction, so the pairs are named TX and RX. In 1000BASE-T, each pair sends and receives at the same time, so IEEE 802.3 gives the pairs a separate name instead of TX or RX.
These names consist of 3 parts, and each part carries a separate piece of information.
|
Part |
Stands For |
Meaning |
BI |
Bi-Directional |
Bidirectional |
D |
Data |
Data |
A, B, C, D |
- |
Letter that distinguishes the 4 pairs |
So BI_DA means "bidirectional data on pair A", and BI_DB means "bidirectional data on pair B".
The table below shows the computer network adapter (MDI) side. On a switch port (MDI-X), the same pins are named with different letters. Pins 1-2, which are BI_DA on the computer, become BI_DB on the switch, and pins 3-6, which are BI_DB on the computer, become BI_DA on the switch. The same swap applies to pins 4-5 and 7-8, which are named BI_DD and BI_DC respectively on the switch. In this way, pair A of one end connects to pair A of the other end. Today, Auto MDI-X makes this match automatically.
|
Wire Pair |
Pin (+) |
Pin (-) |
Function |
|
Pair 1 (BI_DA) |
1 |
2 |
Bidirectional data |
|
Pair 2 (BI_DB) |
3 |
6 |
Bidirectional data |
|
Pair 3 (BI_DC) |
4 |
5 |
Bidirectional data |
|
Pair 4 (BI_DD) |
7 |
8 |
Bidirectional data |
Because all 4 pairs are used, a break or a faulty termination on a single pair prevents the 1000BASE-T link from being established. In this case, the link usually falls back to 100 Mbps, which needs only 2 pairs. A 100 Mbps link on a Gigabit port is a typical sign that one pair in the cabling has a fault.
2.3- PoE and the Wire Pairs
PoE (Power over Ethernet) supplies power to devices such as IP phones, cameras or access points over the same Ethernet cable instead of a separate adapter. The side that supplies the power is usually the switch, and the side that receives the power is the connected device.
The statement "2 pairs for data, 2 pairs for power", which is often heard in practice, is true only for 100BASE-TX. 100BASE-TX uses only 4 of the 8 wires, that is, the 1-2 and 3-6 pairs. The remaining 4-5 and 7-8 pairs are idle, so power can be supplied over these idle pairs. In 1000BASE-T there is no idle pair, because all 4 pairs carry data. In this case, power is supplied over the pairs that carry data.
Data and power can be carried on the same pair because the 2 use the wires in different ways. Data is carried as the voltage difference between the 2 wires of the pair. Power is applied to both wires of the pair together, at the same voltage. A voltage that is applied equally to both wires does not change the difference between them, so the data is not affected. This method is called phantom power.
The PoE standard defines 2 different ways to supply power in the cable. In the first way, power is supplied on pins 1-2 and 3-6. The standard calls this way Alternative A. In the second way, power is supplied on pins 4-5 and 7-8, and the name of this way in the standard is Alternative B. The switch that supplies the power decides which way is used, and the connected device is designed to accept both. IEEE 802.3bt, which is defined for cases that need higher power, uses both ways at the same time and supplies power on all 8 pins. The table below shows whether the pins that supply power in each way also carry data at that time.
|
Pins That Supply Power |
Name in the Standard |
These Pins on a 100BASE-TX Link |
These Pins on a 1000BASE-T Link |
1-2 and 3-6 |
Alternative A |
Carry data |
Carry data |
4-5 and 7-8 |
Alternative B |
Idle |
Carry data |
|
All 8 pins |
IEEE 802.3bt (4-Pair PoE) |
Half carry data, half are idle |
Carry data |
The conclusion from the table is as follows. In 100BASE-TX, power can also be supplied over the idle pairs, whereas in 1000BASE-T it is always supplied over pairs that carry data. In both cases, phantom power ensures that data transmission is not affected.
2.4- Gigabit Ethernet over Fibre
The implementations of Gigabit Ethernet over fibre are the 1000BASE-SX and 1000BASE-LX standards. These standards were defined by IEEE 802.3z in 1998, 1 year before 1000BASE-T, and were the first Gigabit Ethernet standards. 1000BASE-SX operates at 850 nm over multimode fibre for short-distance links, and 1000BASE-LX operates at 1310 nm over both single-mode and multimode fibre. The same standard also defined 1000BASE-CX for short-distance shielded copper links.
|
Standard |
Wavelength |
OM1 |
OM2 and Above |
Single-mode |
1000BASE-SX |
850 nm |
275 m |
550 m |
- |
1000BASE-LX |
1310 nm |
550 m |
550 m |
5 km |
Using 1000BASE-LX over OM1 and OM2 multimode fibre requires a mode conditioning patch cord. I explained this subject and the distances of the other Ethernet standards over fibre in detail in my Fibre Optic Cable article.
3- CAT Cable Categories and Supported Ethernet Standards
A cable category and an Ethernet standard are different things, and they are defined by different organisations. Categories such as CAT5e, CAT6 or CAT6a define the performance that the cable provides up to a specific frequency in terms of attenuation, crosstalk and return loss. These definitions are in the TIA-568 and ISO/IEC 11801 standards. Ethernet standards such as 1000BASE-T or 10GBASE-T are defined under IEEE 802.3 and determine how data is carried over this cable.
As the Ethernet speed increases, the frequency range that the signal occupies also becomes wider, and each standard targets a cable category that can cover this range. For this reason, the right question is not "how many Gbps is this cable?" but "which Ethernet standard does this cable category support, and at what distance?". The table below shows the maximum distance at which each category carries each standard. A hyphen means that the category does not support the standard.
|
Category |
Defined Frequency |
100BASE-TX |
1000BASE-T |
2.5GBASE-T
5GBASE-T |
10GBASE-T |
25GBASE-T
40GBASE-T |
|
CAT5 |
100 MHz |
100 m |
- |
- |
- |
- |
|
CAT5e |
100 MHz |
100 m |
100 m |
100 m |
- |
- |
|
CAT6 |
250 MHz |
100 m |
100 m |
100 m |
55 m |
- |
|
CAT6a |
500 MHz |
100 m |
100 m |
100 m |
100 m |
- |
|
CAT7 (ISO Class F) |
600 MHz |
100 m |
100 m |
100 m |
100 m |
- |
|
CAT8 |
2000 MHz |
30 m |
30 m |
30 m |
30 m |
30 m |
2.5GBASE-T and 5GBASE-T in the table are the standards that IEEE 802.3bz defined in 2016 to fill the gap between 1 Gbps and 10 Gbps. They are also known in the industry as Multi-Gigabit. They were developed as a solution for Wi-Fi access points that need a connection faster than 1 Gbps, without replacing the existing cable. They aim to operate at 100 m over CAT5e when the channel conditions are suitable, but in dense cable bundles these speeds must be verified on site.
10GBASE-T can run over CAT6, but because of the alien crosstalk between cables that are installed side by side, the achievable distance varies between 37 and 55 m, depending on the density of the cable bundle. For this reason, CAT6a is preferred in new installations that target 10 Gbps. CAT6a also includes alien crosstalk values in its definition and is produced in both shielded and unshielded versions.
CAT7 corresponds to Class F in ISO/IEC 11801. Each wire pair is shielded with a separate foil, and the overall shield can be a braid (S/FTP) or a foil (F/FTP), depending on the manufacturer. On the Ethernet side, it does not offer any standard beyond CAT6a.
When the channel is not built with components that comply with Class F, for example when it is terminated with RJ45, Class F performance is not guaranteed.
CAT8 was developed for 25GBASE-T and 40GBASE-T, which are defined by IEEE 802.3bq. The signal of these standards operates at a very high frequency, and the attenuation in the cable grows quickly as the frequency increases. To maintain the 25 Gbps and 40 Gbps speeds, the channel length is limited to 30 m.
In CAT8, the channel length is at most 30 m. CAT8 is not designed for cabling inside buildings. It is designed for data centres, where servers and switches are in the same cabinet or in adjacent cabinets, so this distance is sufficient. If 25 Gbps and 40 Gbps are required over a longer distance, fibre is used.
4- Selecting the Right Cable and Standard
The move from Fast Ethernet to Gigabit Ethernet increased the number of wire pairs in use from 2 to 4 and made each pair bidirectional. With 10GBASE-T and later standards, the signal spectrum extends to hundreds of MHz, so the cable category, the termination quality and the effect of the cables on each other now directly determine whether the link is established.
In a new cabling project, I select the category according to the highest speed that the cable will carry during its lifetime, not according to the current port speed. For desktop connections, CAT6 provides enough margin in most scenarios for speeds between 1 Gbps and 5 Gbps. CAT6a is the right choice for server and access point connections where 10 Gbps is expected, and CAT8 or fibre is the right choice for short, very high-speed links inside the data centre. The channel limit of 4-pair BASE-T links is 100 m, so fibre is the common choice for LAN links that exceed this distance.
I hope you find it useful...
Feel free to share your thoughts and questions about this article in the comments below. Every contribution helps the content reach more people and makes for a more useful discussion.