Ethernet frame
In networking, data is divided into units that are processed independently on a link. In an Ethernet network, which operates at the OSI data-link layer, such a unit is called a frame. A frame carries client data together with addressing and control fields used to forward it within a local segment.
The frame format defines how addresses, the type or length field, and error checks are interpreted. Network interfaces and switches use this structure to process and forward frames. Ethernet networks use two formats: Ethernet II (DIX), where the field identifies an EtherType, and IEEE 802.3, where the field specifies the data length.
Anatomy of an Ethernet II / IEEE 802.3 Frame
An Ethernet transmission has several distinct fields. The preamble and Start Frame Delimiter (SFD) belong to the physical transmission, although diagrams often show them with the frame. The frame itself is 64 to 1518 bytes from Destination MAC through FCS. Adding the 7-byte preamble and 1-byte SFD gives 72 to 1526 transmitted bytes before the interframe gap. Counting the minimum 12-byte gap gives 84 to 1538 byte-times. The interframe gap is a required idle interval on the link.
| Field | Size (Bytes) | Purpose |
|---|---|---|
| Preamble | 7 | Receiver clock synchronization. |
| Start Frame Delimiter (SFD) | 1 | Marks the end of the preamble and the start of the actual frame. |
| Destination MAC Address | 6 | Data-link address of the recipient interface. |
| Source MAC Address | 6 | Data-link address of the sending interface. |
| EtherType / Length | 2 | Identifies the upper-layer protocol (Ethernet II) or specifies the payload length (IEEE 802.3). |
| Data (Payload) and Pad | 46 - 1500 | Contains the actual data being sent (e.g., an IP packet). Padding is added if data is less than 46 bytes. |
| Frame Check Sequence (FCS) | 4 | Error-checking field to detect transmission errors. |
Ethernet Frame Fields
Frame fields identify the recipient, the sender, and the encapsulated data.
1. Preamble and Start Frame Delimiter (SFD)
These first 8 bytes prepare the receiver for the incoming data. The Preamble is a 7-byte sequence of alternating ones and zeros (`10101010...`). In classic 10 Mbps Ethernet, this pattern supports clock recovery and bit synchronization; the 10 MHz square-wave description applies to that speed.
The Start Frame Delimiter (SFD)is a 1-byte field with the pattern `10101011`. It immediately follows the Preamble. The final two '1' bits break the alternating pattern and signal the start of the frame data.
2. Destination and Source MAC Addresses
These two fields contain data-link addresses, known as , of the destination and source interfaces. They are used to forward frames within a network segment. Both fields are 6 bytes (48 bits) long. The Source Address identifies the transmitting interface, while the Destination Address can identify one interface, a group, or all stations. The first transmitted bit identifies a group address.
3. EtherType / Length Field
This 2-byte field has different meanings in Ethernet II and IEEE 802.3. The difference follows from the coexistence of the two frame formats.
The network interface compares this value with the EtherType and length ranges:
- If the value is 1536 or greater (), the field is interpreted as an EtherType. This indicates an Ethernet II frame, and the value tells the receiving device which upper-layer protocol (e.g., IPv4, IPv6, ARP) is encapsulated in the data payload.
- If the value is 1500 or less (), the field is interpreted as the Length of the data payload that follows. This indicates an IEEE 802.3 frame.
Values from 1501 through 1535 fall outside both the IEEE 802.3 length range and the EtherType range.
4. Data (Payload) and Padding
This field carries data from upper layers, such as an IP packet, and may include padding. Its size depends on the carried PDU.
The standard maximum data size is 1500 bytes, the usual . A larger PDU must be divided according to the protocol and path MTU before it is sent over such a link.
Ethernet also has a minimum frame size requirement of 64 bytes (from Destination MAC to FCS). Therefore, the data payload must be at least 46 bytes long. If the actual data is smaller than 46 bytes, the sender adds extra, meaningless bytes called Padding to bring the field up to the minimum required size.
5. Frame Check Sequence (FCS)
The final 4 bytes of the frame contain the Frame Check Sequence. This field is used for error detection. The sending device computes a , over all the fields from the Destination MAC to the Pad field. The 32-bit result is placed in the FCS.
The receiving device computes the CRC over the received fields. A match with the FCS indicates that no error was detected, but does not prove that every bit is correct. The frame can then be accepted for higher-layer processing. A mismatch causes the frame to be discarded. Ethernet detects errors but does not correct them; higher-layer protocols such as TCP may retransmit lost data.
Frame Sizes and Variants
The standard Ethernet frame size from Destination MAC through FCS is 64 to 1518 bytes. Variants can change this size. The most common is , which adds a 4-byte tag to the frame and raises the maximum frame size to 1522 bytes. Some networks also use non-standard Jumbo Frames of 9000 bytes or more; every device along the path must support the same size.