Interframe Spacing and Backoff
The Ethernet standards require a minimum spacing between two non-colliding frames. This gives the media time to stabilize after the transmission of the previous frame and time for the devices to process the frame. Referred to as the interframe spacing, this time is measured from the last bit of the FCS field of one frame to the first bit of the Preamble of the next frame.
After a frame has been sent, all devices on a 10 Mbps Ethernet network are required to wait a minimum of 96 bit times (9.6 microseconds) before any device can transmit its next frame. On faster versions of Ethernet, the spacing remains the same - 96 bit times - but the interframe spacing time period grows correspondingly shorter.
Synchronization delays between devices may result in the loss of some of frame preamble bits. This in turn may cause minor reduction of the interframe spacing when hubs and repeaters regenerate the full 64 bits of timing information (the Preamble and SFD) at the start of every frame forwarded. On higher speed Ethernet some time sensitive devices could potentially fail to recognize individual frames resulting in communication failure.
As you will recall, Ethernet allows all devices to compete for transmitting time. In the event that two devices transmit simultaneously, the network CSMA/CD attempts to resolve the issue. But remember, when a larger number of devices are added to the network, it is possible for the collisions to become increasingly difficult to resolve.
As soon as a collision is detected, the sending devices transmit a 32-bit "jam" signal that will enforce the collision. This ensures all devices in the LAN to detect the collision.
It is important that the jam signal not be detected as a valid frame; otherwise the collision would not be identified. The most commonly observed data pattern for a jam signal is simply a repeating 1, 0, 1, 0 pattern, the same as the Preamble.
The corrupted, partially transmitted messages are often referred to as collision fragments or runts. Normal collisions are less than 64 octets in length and therefore fail both the minimum length and the FCS tests, making them easy to identify.
After a collision occurs and all devices allow the cable to become idle (each waits the full interframe spacing), the devices whose transmissions collided must wait an additional - and potentially progressively longer - period of time before attempting to retransmit the collided frame. The waiting period is intentionally designed to be random so that two stations do not delay for the same amount of time before retransmitting, which would result in more collisions. This is accomplished in part by expanding the interval from which the random retransmission time is selected on each retransmission attempt. The waiting period is measured in increments of the parameter slot time.
If media congestion results in the MAC layer unable to send the frame after 16 attempts, it gives up and generates an error to the Network layer. Such an occurrence is rare in a properly operating network and would happen only under extremely heavy network loads or when a physical problem exists on the network.
The methods described in this section allowed Ethernet to provide greater service in a shared media topology based on the use of hubs. In the coming switching section, we will see how, with the use of switches, the need for CSMA/CD starts to diminish or, in some cases, is removed altogether.