1

The Ethernet preamble, 56 bits of alternating 1 and 0 bits, allowing the receiver to synchronize its clock to the transmitter, followed by a one-octet start frame delimiter byte (10101011) and then the header.

What happens when the header or the payload collides with the SoF? (10101011) Or how is it avoided?

Also, if the mechanism considers the whole pack (preamble + start frame), the question remains, what happens when the header or data contains this sequence?

10101010 10101010 10101010 10101010 10101010 10101010 10101010
|Preamble                                             |Sof    |

Other way to see this question is, how can I send data equals too 10101010 10101010 10101010 10101010 10101010 10101010 10101010 in the ethernet payload?

9
  • 1
    A collision causes the data to be corrupted, so it's never received.
    – Ron Trunk
    May 1, 2020 at 13:50
  • Thanks @RonTrunk so the origin hardware device doesn't let the frame to go out? or it goes out but the receiver hardware will drop it? And if it drops the frame, I suppose, the hardware continues trying to find a new preamble + sof, in this case if I double feed my frame payload with that, is it going to drop again and again util it finds one that has the proper [preamble+sof] [header|payload] [CRC]? May 1, 2020 at 13:58
  • 1
    If the receiver doesn't receive a properly formatted frame, it will not not recognize it as a frame, or drop it as corrupted.
    – Ron Trunk
    May 1, 2020 at 14:01
  • 1
    Once the receiver is synchronized with the data clock (the purpose of the preamble), it collects the bits until the frame stops. It doesn't matter what they are, so long as the receiver can interpret it as a proper frame.
    – Ron Trunk
    May 1, 2020 at 14:10
  • 1
    In that same article, look at End of Frame
    – Ron Trunk
    May 1, 2020 at 14:39

1 Answer 1

1

Any kind of signal that is transmitted concurrently to another signal causes a collision (in half-duplex mode).

You need to understand that the signal on the medium is analog, not binary. So, on shared wire (coax), the signals add to one another, causing illegal voltage levels. On duplex media (twisted pair, fiber), collision detection is simpler: any carrier detection on the receiver side while transmitting is a collision.

A preamble-like sequence somewhere in the middle of a frame doesn't matter. The preamble is used to synchronize the receiver's bit and byte shifters to the carrier. After that, the sequence has no meaning (it's just data) until the current frame has ended (loss of carrier or end-of-frame symbol, depending on the physical layer variant), and the receiver returns to carrier sense mode.

4
  • Thanks, After that, the sequence has no meaning (it's just data) until the current frame has ended @Ron Trunk pointed me the same, and I suppose the frame has a fixed lenght (either by default or by reading its header) and the end-of-frame symbol, therefore the hardware the start to parse the frame so it can sync the clock, read until X lenght. May 1, 2020 at 15:24
  • 1
    Ethernet frames are variable in length. Their end is indicated by loss of carrier or end-of-frame. The line code is decoded in real time and when the EOF is detected, the buffer is passed to the NIC driver.
    – Zac67
    May 1, 2020 at 15:32
  • "A preamble-like sequence somewhere in the middle of a frame doesn't matter" -- actually it does matter...what if a receiver is plugged into a network at the moment that the transmitter is in the middle of transmitting a frame and the data happens to look exactly like a preamble? How do modern ethernet specs address this case? Do they expect the checksum to detect it? Are receivers required to wait for the inter-frame gap before searching for the preamble? Jan 13 at 13:49
  • @LouisSemprini That's actually possible, but only for repeated, obsolete segments (on a switched link, transmission cannot start before link establishment). In that case the sequence would initiate reception, ending at the normal frame end. However, the FCS would not match, causing the partial frame to be dropped by the NIC. The gap isn't used for that because (depending on the L1 variant at hand) it might simply mean silence on the channel.
    – Zac67
    Jan 13 at 14:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.