Ethernet Frame Format Explained: Preamble, SFD, Field Sizes and Worked Calculations

Learn exactly what the 64-byte Ethernet minimum includes, how preamble and SFD appear on the wire, and how to calculate padding and transmission time.

KnowledgeGate Team

Exam prep & CS education

Updated 19 Nov 20265 min read
Ethernet Frame Format Explained: Preamble, SFD, Field Sizes and Worked Calculations

The counting mistake is what makes many Ethernet questions look harder than they are. A diagram may put the preamble, SFD, MAC frame and interpacket gap in one row, but the 64-byte minimum and 1,518-byte maximum for untagged Ethernet II do not count every item in that row. The 64-byte minimum and 1,518-byte maximum for untagged Ethernet II apply to the MAC frame, which runs from Destination Address through FCS; preamble, SFD and interpacket gap are separate. For the wider subject route, use CS Fundamentals for Exams & Placements.

Ethernet frame format: first decide what is being counted

Keep three nested quantities separate:

  1. The MAC frame runs from Destination Address through FCS.

  2. The transmitted sequence has 7 preamble bytes and 1 SFD byte before the MAC frame.

  3. A 96-bit-time interpacket gap follows. It is required idle time, not a frame field and not 12 transmitted data bytes.

For untagged Ethernet II, Destination through FCS is 64 to 1,518 bytes, with 46 to 1,500 bytes of Data plus Pad. Preamble through FCS therefore takes 72 to 1,526 byte-times. The minimum start-to-next-start interval, including the gap, is 84 byte-times.

An IP packet sits inside the Ethernet Data field. The common 1,500-byte MTU limits that payload, not the complete frame. Preamble, SFD and the gap are link overhead. IP Addressing and Subnetting Explained continues at the Network Layer, while TCP vs UDP: Transport Layer Explained covers the layer carried inside IP.

Ethernet preamble and SFD: how the receiver finds the first frame bit

In ordinary hexadecimal and MSB-first binary, each preamble octet is 0x55 = 01010101; SFD is 0xD5 = 11010101. Ethernet sends each octet least-significant bit first, making their time-ordered wire patterns 10101010 and 10101011.

Preamble's repeated transitions let the receiver recover and align its bit clock. SFD means Start Frame Delimiter and is one byte. It ends in consecutive 11 bits on the wire, breaks the alternation, and marks Destination Address as the next octet.

Both are transmitted, but neither enters the 64-byte minimum or FCS calculation. They are not an 8-byte MAC header.

Ethernet frame fields from Destination Address to FCS

Field

Size in the untagged example

Purpose

Included in 64-1,518 bytes?

Preamble

7 bytes

Clock synchronisation

No

SFD

1 byte

Frame boundary

No

Destination Address

6 bytes

Receiver or group address

Yes

Source Address

6 bytes

Sender address

Yes

EtherType

2 bytes

Upper-layer protocol

Yes

Data plus Pad

46-1,500 bytes

Carried packet plus any minimum-length padding

Yes

FCS

4 bytes

CRC-based error detection

Yes

Interpacket Gap

12 byte-times

Recovery and spacing before the next frame

No

Our first example uses Destination 00:1A:2B:3C:4D:5E, Source 02:AA:BB:CC:DD:EE, and EtherType 0x0800 for IPv4. FCS covers Destination, Source, EtherType, Data and Pad. It detects corruption but neither repairs the frame nor guarantees retransmission.

An IEEE 802.1Q VLAN tag adds 4 bytes between Source and the original EtherType. Our untagged examples retain the 14-byte header and 1,518-byte maximum.

Ethernet minimum-frame worked example: padding a 28-byte IP packet

The frame carries a 28-byte IPv4/UDP packet: a 20-byte IPv4 header and 8-byte UDP header, with no UDP data. It uses the values above. Required padding is:

46 - 28 = 18 bytes

Now sum the MAC frame:

6 + 6 + 2 + 28 + 18 + 4 = 64 bytes

Preamble and SFD give 64 + 7 + 1 = 72 bytes = 576 bits through FCS. The gap gives 72 + 12 = 84 byte-times = 672 bit-times start to start.

At 100 Mb/s, through-FCS time is 576 / (100 x 10^6) = 5.76 microseconds. The gap is 96 / (100 x 10^6) = 0.96 microseconds, so the earliest next start is 6.72 microseconds after the first preamble bit. Thus 64 bytes, 72 bytes and 84 byte-times answer three different boundaries.

Left-to-right Ethernet frame strip showing a 64-byte MAC frame, 72 bytes through FCS, 84 byte-times start-to-next-start, and 100 Mb/s timings.

Ethernet on-wire-time worked example: a 1,000-byte payload at 1 Gb/s

An untagged Ethernet II frame now carries a 1,000-byte IP packet. It needs no padding.

  • MAC frame: 6 + 6 + 2 + 1,000 + 4 = 1,018 bytes

  • Preamble through FCS: 1,018 + 8 = 1,026 transmitted bytes

  • Start to earliest next start: 1,026 + 12 = 1,038 byte-times

At 1 Gb/s, through-FCS time is 1,026 x 8 / 10^9 = 8.208 microseconds. With the gap, start-to-next-start time is 1,038 x 8 / 10^9 = 8.304 microseconds. The gap remains link time, not frame bytes.

Payload divided by MAC-frame bytes is 1,000 / 1,018 = 98.23%. Payload divided by the complete start-to-next-start budget is 1,000 / 1,038 = 96.34%. Always name the denominator.

Three aligned timing bars compare a 1,000-byte payload at 1 Gb/s with a 1,018-byte MAC frame, 1,026 transmitted bytes, and 1,038 byte-times including the gap.

Ethernet frame questions: a reliable exam-solving order

Use this sequence:

  1. Identify whether the frame is untagged or VLAN-tagged.

  2. Underline whether “frame” starts at Destination, or whether the question asks for actual wire time.

  3. For the untagged case, calculate Data plus Pad as max(given payload, 46).

  4. Add the 14-byte Destination-through-EtherType header and 4-byte FCS.

  5. Add 8 preamble/SFD bytes only for transmitted-through-FCS time.

  6. Add 96 bit-times only for start-to-next-start timing.

Typical tasks include labelling fields, distinguishing preamble from SFD, decoding 0x0800, finding padding, checking FCS coverage, calculating serialisation time, and comparing payload ratios.

Check the boundaries: a 46-byte payload gives 14 + 46 + 4 = 64 bytes with no pad. A 1,500-byte payload gives 1,518 bytes, or 1,526 transmitted bytes with preamble/SFD. For the broader Ethernet map, Ethernet in the Data Link Layer: Frame Format, CSMA/CD, and Worked GATE Examples covers CSMA/CD, backoff, switching and collision domains; frame-format problems instead turn on boundaries, padding and on-wire timing.

Ethernet frame-format traps and their corrections

  • Preamble/SFD inside the minimum: This makes 72 bytes the MAC-frame size. Bracket Destination through FCS first.

  • Gap counted as bytes: It is 96 bit-times of idle spacing, not transmitted frame data.

  • Short packet accepted as a short frame: 6 + 6 + 2 + 28 + 4 = 46 bytes omits padding. Raise Data plus Pad to 46 bytes first.

  • SFD credited with clock recovery: Preamble supplies the transitions; SFD identifies the boundary.

  • FCS said to cover preamble or repair errors: It covers Destination through Data/Pad. A failed frame is rejected, not repaired.

  • Untagged maximum used with a VLAN tag: Write and sum the named fields.

Ethernet frame format: the short version and next step

The durable count is 7 + 1 setup bytes, a 6 + 6 + 2 untagged Ethernet II header, 46 to 1,500 bytes of Data plus Pad, and 4 bytes of FCS. The gap is a separate 96 bit-times. The 64-to-1,518-byte range runs from Destination through FCS.

For a self-test, replace 28 bytes with 40. Pad becomes 46 - 40 = 6 bytes; the MAC frame remains 64 bytes; through-FCS time at 100 Mb/s remains 5.76 microseconds; and start-to-next-start time remains 6.72 microseconds. For structured Computer Networks study including Data Link Layer Ethernet, continue with GATE Guidance by Sanchit Sir.

Keep learning

Discussion