Ethernet Protocol

Quick Explanation

ELECTGON
www.electgon.com
[email protected]

06.06.2019
Contents

1 Ethernet Data Link Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Reconciliation Sublayer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Media Independent Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.1 RGMII In Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4 Physical Coding Sublayer (PCS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5 Physical Medium Attachment (PMA) . . . . . . . . . . . . . . . . . . . . . . . . . 8
6 Communication Medium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
6.1 Twisted Pair Cabling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
6.2 Fiber Cabling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
6.2.1 Single‐Mode Fiber Variants . . . . . . . . . . . . . . . . . . . . 11
6.2.2 Multimode Fiber Variants . . . . . . . . . . . . . . . . . . . . . 11
6.3 Twinax Copper Cable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
7 Medium Dependent Interface (MDI) . . . . . . . . . . . . . . . . . . . . . . . . . 12
8 Autonegotiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
 Abstract

Digital Data transmission is standardized and practically used in networks communications

according to OSI layers model. This means that data shall be processed according to speci ications
of that standard in order to comply with this ef icient OSI model. A basic protocol that is
used almost in all network transmissions is the Ethernet protocol which will discussed in
this document along with its main features.
Ethernet Protocol 1. Ethernet Data Link Layer

1 Ethernet Data Link Layer

Data Link layer speci ications can be implemented using many protocols like Token Ring,
Token Bus, Ethernet, TDMA, CDMA, etc. Ethernet was produced as a protocol that can be used
to control transmission of packets within Local Area Networks (LAN). It organizes operation
of Data Link Layer (layer 2) and also Physical Layer (Layer 1) as well. This protocol is now
covered by IEEE 802.3 to describe different families or technologies used to implement that
protocol to support additional network media and higher data rate capabilities. Figure1
shows structure of OSI model with more details about Data Link layer. It can be shown that
this layer is divided into Logic Link Control (LLC) and Media Access Control (MAC).

Figure 1: Structure of OSI Model and Layer 2

LLC sublayer is responsible for creating the frames to be sent over the network. In that
way it handles communication between upper and lower layers. The MAC sublayer communicates
directly with the physical layer. It is responsible for:

• Data encapsulation, including frame assembly before transmission, and frame parsing/error
detection during and after reception.

• Media access control, including initiation of frame transmission and recovery from transmission
failure.

Ethernet standard is de ined to cover MAC sublayer and Layer 1. It ful ills speci ications of
the MAC sublayer by de ining a basic frame format. Figure 2 shows the basic format of the
Ethernet frame.

1
Ethernet Protocol 1. Ethernet Data Link Layer

Figure 2: Ethernet Frame

Preamble is a 7 byte sequence used to synchronize frame sending/receiving between sender

and receiver. These 7 bytes are 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 which are in binary
sequence of 0 and 1.

SFD (Start of Frame Delimiter) is one byte used to indicate frame data is starting after this
byte. For Ethernet it is 0xD5.

Destination Address is the address of the other terminal to which the frame is sent. The
address here is the MAC address of the device.

Source Address is the MAC address of the sender.

Length/Type Ethernet II considered these bytes to represent EtherType while the original
IEEE 802.3 framing considered these octets to represent the size of the payload in bytes.
Thus, values of 1500 and below for this ield indicate that the ield is used as the size of
the payload of the Ethernet frame while values of 1536 and above indicate that the ield
is used to represent EtherType. The interpretation of values 1501–1535, inclusive, is
unde ined.
Below are some values de ined for indicating type of the Ethernet.

2
Ethernet Protocol 1. Ethernet Data Link Layer

EtherType Protocol
0x0800 Internet Protocol version 4 (IPv4)
0x0806 Address Resolution Protocol (ARP)
0x0842 Wake‐on‐LAN
0x22F3 IETF TRILL Protocol
0x6003 DECnet Phase IV
0x8035 Reverse Address Resolution Protocol
0x809B AppleTalk (Ethertalk)
0x80F3 AppleTalk Address Resolution Protocol (AARP)
0x8100 VLAN‐tagged frame (IEEE 802.1Q) and Shortest Path Bridging IEEE 802.1aq
0x8137 IPX
0x86DD Internet Protocol Version 6 (IPv6)
0x8808 Ethernet low control
0x8892 PROFINET Protocol
0x88A2 ATA over Ethernet
0x88A4 EtherCAT Protocol
0x88E5 MAC security (IEEE 802.1AE)
0x88E7 Provider Backbone Bridges (PBB) (IEEE 802.1ah)
0x88F7 Precision Time Protocol (PTP) over Ethernet (IEEE 1588)
0x9100 VLAN‐tagged (IEEE 802.1Q) frame with double tagging

Table 1: Some EtherTypes Codes

What is meant by Ethernet type here is the different protocols that are standardized
based upon Ethernet technology. These extended standards are using Ethernet frame
structure for sending its information and consequently they are using Ethernet physical
medium speci ications for communication. This extended Ethernet classi ication is
different than Ethernet classi ication according to its speed which shall be covered in
physical layer implementation of the Ethernet.

Data: is the Payload obtained from higher layers to be sent over the Ethernet.

FCS: This 4 bytes part is a checksum appended to end of frame to check accuracy of received
data of the frame. For Ethernet, it is using CRC‐32. It worth mentioning here that for
the CRC part, the most signi icant bit is sent irst.

The other part of Ethernet standard is covering Layer 1 speci ications. This layer depends on
physical implementation of the medium. Because of that, Ethernet standard has been evolved
over many years. First Ethernet implementation started with 3 Mbps in 1973. When it started
as industry standard and started with 10 Mbps. Figure 3 shows evolution of Ethernet Data
Rate over time.

3
Ethernet Protocol 2. Reconciliation Sublayer

Figure 3: Ethernet Data Rate Evolution [1]

Because of these different implementations of the Ethernet, Layer 1 is organizing different

implementation techniques by de ining sublayer for the implementation methodology. This
will be discussed shortly in the following lines.

2 Reconciliation Sublayer
This sublayer is medium independent and responsible for reconciling the MII with the 802.3u
MAC sublayer, so that the latter sees the same physical layer service as 802.3. The reason of
having this sublayer is the Ethernet physical interface is implemented with different data rate.
Some data rates are using 4 bits to be transmitted, others are using 8 bits as will be shown
next.

3 Media Independent Interface

Interfaces between MAC component and Physical component are proposed with different
technologies that are listed in table 2

4
Ethernet Protocol 3. Media Independent Interface

Supported Total No. of No. of Data Clock Frequency

Data Rate Pins Pins
MII 10 M 16 Pins TX: 4 pins 2.5 MHz for 10 M
100 M RX: 4 pins 25 MHz for 100 M
RMII 10 M 8 Pins TX: 2 pins 50 MHz for 10/100
100 M RX: 2 pins M
GMII 10M 24 Pins* TX: 8 pins 2.5 MHz for 10 M
100M RX: 8 pins 25 MHz for 100 M
1G 125 MHz for 1 G
RGMII 10M 12 Pins TX: 4 pins 2.5 MHz for 10 M
100M/ RX: 4 pins 25 MHz for 100 M
1G 125 MHz for 1 G
SGMII 10M 8 Pins TX: 2 pins 625 MHz for all
100M RX: 2 pins
1G (each is
Differential
Pair)
XGMII 10 G ** 74 Pins TX: 32 pins 156.25 MHz
RX: 32 pins
Notes * it is mentioned in all online resources that total pins are 24 pins,
Maybe because transmitter signals have 2 different clock signals
for 1 G transmission and another for 10/100 Transmission. I
assume then they count for either one of those clocks.
** Not sure if it is limited to 10 G only or it can cover lower data
rates also.
*** In IEEE standard for Ethernet there are also XLGMII, CGMII.

Table 2: Ethernet Physical Interfaces

3.1 RGMII In Scope

IEEE has de ined standard for the gigabit Ethernet in 1998 (IEEE Std 802.3z). This standard
introduces GMII as the interface between MAC layer and physical layer. Although this interface
provides ef icient description for this accepted data rate, however it suffers from increased
number of pins connections between MAC and PHY layers. Therefore another speci ication
has been produced by Broadcom, HP and Marvell companies to provide RGMII as an adoption
for GMII. This is very useful and ef icient solution in cases when MAC and PHY are implemented
in different hardware.
Originally, the GMII interface depends on the following timing diagram when sending or
receiving Ethernet frames.

Figure 4: GMII TX Timing Diagram [5]

5
Ethernet Protocol 3. Media Independent Interface

Figure 5: GMII RX Timing Diagram [5]

RGMII interface de ined in Speci ications V2.0 contains signals de inition as shown in
table 3

Table 3: RGMII Signals Description

Which means that Transmission path (from MAC to PHY) will have 4 bits for data transmission
(TXD), 1 bit for clock (TXC) and 1 bit for control (TX_CTL). Receiver path has the same structure.
RGMII supports 10/100/1000 Mbps transmission. For the case of 10 or 100 Mbps, transmitted
clock on TXC will 2.5 or 25 MHz respectively and the data will be transmitted on the rising
edge of the clock. In case of 1 Gbps, transmitted clock is 125 MHz and data is transmitted on
both rising and falling edges.
Since data size is reduced (from 8 bits in GMII to 4 bits in RGMII), RGMII speci ies transmitting
lowest 4 bits irst then highest 4 bits.

6
Ethernet Protocol 3. Media Independent Interface

Figure 6: 1G Transmission in RGMII

That is how it is reducing the 8 bits into 4 bits. GMII standard has 2 control signals
gmii_tx_en and gmii_tx_er. RGMII reduces these control signals by sending gmii_tx_en on the
rising edge. And sending a logical derivative of gmii_tx_en and gmii_tx_er in the falling edges.
Accodrding to Speci ications 2.0 this logical derivative is Xoring those signals. Figure 7 shows
how a valid and errored frame shall be sent.

Figure 7: RGMII Frames

RGMII speci ication describes also timing specs of the data, control and clock signals in
terms of setup, hold and skew times between those signals. What is important to point out
here is there should be 2 ns setup time and 2 ns hold times between data and transmitted
clock (the same applies also for setup /hold times between TX_CTL and TXC). Note that it is
meant here that the transmitted data should arrive at the receiver with that setup and hold

7
Ethernet Protocol 4. Physical Coding Sublayer (PCS)

times speci ications.

Figure 8: RGMII Setup Time [6]

4 Physical Coding Sublayer (PCS)

The media‐dependent physical coding sublayer (PCS) provides encoding/decoding of the bit
stream.
Depending on which link speed (Data Rate) is used, the encoding of the serial bits coming
from MII sublayer is carried out in PCS sublayer. For 10 Mbps, Manchester coding is used. For
100 Mbps, 4B/5B coding is used. For 1 G, 8B/10B coding is used in case of iber medium and
Trellis modulation with PAM‐5 is used in case of copper medium. For 10 G, 64B/66B coding
is used.

5 Physical Medium Attachment (PMA)

This sublayer contains the signal transmitters and receivers (transceivers) so it is responsible
for serializing/de‐serializing encoded bits.
PMA sublayer has a mode of operation; either half duplex or full duplex. This shall be
taken into consideration while building the physical layer. Most of Ethernet standard specifying
full duplex implementation. For our scope here 1 Gigabit Ethernet was used in full duplex
operation. Figure 9 shows connection of this standard.

8
Ethernet Protocol 6. Communication Medium

Figure 9: 1G Full Duplex Connection [7]

6 Communication Medium
Physical medium used currently for Ethernet are Copper or Fiber. Together with available
data rates this makes a lot of Ethernet technologies. To organize these different technologies,
igure 10 shows Ethernet nomenclature used to identify each technology

9
Ethernet Protocol 6. Communication Medium

Figure 10: Ethernet Nomenclature

6.1 Twisted Pair Cabling

Twisted Pair Cabling is the most commonly deployed form of Ethernet and includes:

– 10BASE‐T – 10Mb/s to 100 meters

– 100BASE‐T – 100Mb/s to 100 meters

– 1000BASE‐T – 1Gb/s to 100 meters

– 10GBASE‐T – 10Gb/s to 100 meters

It has several categories as shown in igure 11.

Figure 11: Twisted Pair Categories

6.2 Fiber Cabling

There are two main types of iber optic cables

• Single Mode Fiber (SMF)

• Multi‐Mode Fiber (MMF)

10
Ethernet Protocol 6. Communication Medium

The difference is basically in the size of the core.

Optical ibers used in Ethernet come in multiple categories. ISO/IEC 11801 speci ies
Categories OM1, OM2, OM3, OS1 and OS2. OM1, 2, 3 are categories for Multimode iber
while OS1, OS2 are categories for single mode iber. Figure 12 shows applications of these
categories in Ethernet.

Figure 12: Fiber Cables Categories

6.2.1 Single‐Mode Fiber Variants

Single‐mode iber is used for long distance links within large data centers and for links in
campus or metro areas such as:

– 100BASE‐LX – 100Mb/s to at least 5 kilometers

– 1000BASE‐LX – 1Gb/s to at least 5 kilometers

– 10GBASE‐LR – 10Gb/s to at least 10 kilometers

– 10GBASE‐ER – 10Gb/s to at least 40 kilometers

– 40GBASE‐FR – 40Gb/s to at least 2 kilometers

– 40GBASE‐LR – 40Gb/s to at least 10 kilometers

– 100GBASE‐LR – 100Gb/s to at least 10 kilometers

– 100GBASE‐ER – 100Gb/s to at least 40 kilometers

6.2.2 Multimode Fiber Variants

Multimode iber has enabled longer distances at higher speeds within the data center such
as:

– 100BASE‐FX – 100Mb/s up to 2 kilometers

– 1000BASE‐SX – 1Gb/s up to 550 meters

– 10GBASE‐SR – 10Gb/s up to 300 meters

– 40GBASE‐SR4 – 40Gb/s up to 100 meters of OM3

11
Ethernet Protocol 7. Medium Dependent Interface (MDI)

– 100GBASE‐SR10 – 100Gb/s up to 100 meters of OM3

– 40GBASE‐SR4 – 40Gb/s up to 150 meters of OM4

– 100GBASE‐SR10 – 100Gb/s up to 150 meters of OM4

6.3 Twinax Copper Cable

Twinax is a shielded copper cable that has twin conductors with good electrical properties
that enables these short reach applications at high speed:

– 1000BASE‐CX – 1Gb/s up to 25 meters

– 10GBASE‐CX4 – 10Gb/s up to 15 meters

– SFP+ Direct Attach Cable – 10Gb/s to 7 meters

– 40GBASE‐CR4 – 40Gb/s up to 7 meters

– 100GBASE‐CR10 – 100Gb/s up to 7 meters

7 Medium Dependent Interface (MDI)

The medium‐dependent interface (MDI) is the cable connector between the signal transceivers
and the link.
For twisted pairs cabling RJ45 connector is used.

Figure 13: RJ45 Connector

For Fiber cabling there are various connectors that change over time as the technology
progresses. The most common connectors are SC, ST, and LC, although many other connectors
exist. These connectors vary in shape and size, but all are designed to allow easy interconnection
between cables and other devices.

12
Ethernet Protocol 8. Autonegotiation

Figure 14: Fibre Cables Connectors [8]

Each connector needs adopter then to be plugged in.

Figure 15: SFP Connector Adopter [8]

These connectors then can be connected to the physical optical port which is called Optical
Form Factor

Figure 16: Optical Ports Connectors [1]

8 Autonegotiation
As result of having multiple Ethernet speeds, the connected devices have to know at which
speed they are going to talk to each other. This point initially wasn’t needed as the Ethernet
communication was carried out at 10 Mbps. Now we have many Ethernet speed, the connected
device shall setup or prepare the communication. So they have to agree on which speed data
will be exchanged also if it is full or half duplex. This process is called then auto‐negotiation
(also known as NWay algorithm).

13
Ethernet Protocol 8. Autonegotiation

As an optional sublayer in 10 Mbps speed, connected devices detect an active link through
the transmission and reception of Link Integrity Pulses LIP (also known as Normal Link Pulses
NLP). The Autonegotiation sublayer became mandatory for other Ethernet speeds. By producing
100 Mbps speed, the NLP has been replaced by other sequence of pulses called Fast Link
Pulses FLP. The devices are exchanging bursts of FLP to initiate the communication. One FLP
burst consists of series of 33 pulses and it is 2sms long. Total transmission interval is about
16 ms. Figure 17 depicts these pulses.

Figure 17: FLP Pulses [9]

The individual pulses alternate between clock pulses and data pulses with the irst and
all successive odd numbered pulses being clock pulses.

14
Bibliography

[1] http://www.ethernetalliance.org

[2] https://en.wikipedia.org/wiki/Media‐independent_interface

[3] http://appmarkers.tistory.com/entry/MII‐RMII‐GMII‐RGMII

[4] http://humblesblog.blogspot.de/2013/01/rgmiisgmiixaui.html

[5] https://www.xilinx.com/support/documentation/application_notes/xapp209.pdf

[6] http://www.ti.com/product/AM3359/datasheet/peripheral_information_and_timings

[7] http://docwiki.cisco.com/wiki/Ethernet_Technologies

[8] http://www. ibrefab.com/product/lc‐connectors

[9] http://supportsystem.livehelpnow.net/resources/14629/ethernet%20autonegotation%20wp[1].pdf

15