Operation Manual

VigorSwitch G1280 User’s Guide
104
After the frame is assembled, when transmitting the frame, the preamble (PRE) bytes are
inserted and sent first, then the next, Start of frame Delimiter (SFD), DA, SA and through the
data field and FCS field in turn. The followings summarize what a MAC does before
transmitting a frame.
1. MAC will assemble the frame. First, the preamble and Start-of-Frame delimiter will be
put in the fields of PRE and SFD, followed DA, SA, tag ID if tagged VLAN is applied,
Ethertype or the value of the data length, and payload data field, and finally put the FCS
data in order into the responded fields.
2. Listen if there is any traffic running over the medium. If yes, wait.
3. If the medium is quiet, and no longer senses any carrier, the MAC waits for a period of
time, i.e. inter-frame gap time to have the MAC ready with enough time and then start
transmitting the frame.
4. During the transmission, MAC keeps monitoring the status of the medium. If no collision
happens until the end of the frame, it transmits successfully. If there is a collision
happened, the MAC will send the patterned jamming bit to guarantee the collision event
propagated to all involved network devices, then wait for a random period of time, i.e.
backoff time. When backoff time expires, the MAC goes back to the beginning state and
attempts to transmit again. After a collision happens, MAC increases the transmission
attempts. If the count of the transmission attempt reaches 16 times, the frame in MAC’s
queue will be discarded.
Ethernet MAC transmits frames in half-duplex and full-duplex ways. In halfduplex operation
mode, the MAC can either transmit or receive frame at a moment, but cannot do both jobs at
the same time.
As the transmission of a MAC frame with the half-duplex operation exists only in the same
collision domain, the carrier signal needs to spend time to travel to reach the targeted device.
For two most-distant devices in the same collision domain, when one sends the frame first,
and the second sends the frame, in worstcase, just before the frame from the first device
arrives. The collision happens and will be detected by the second device immediately.
Because of the medium delay, this corrupted signal needs to spend some time to propagate
back to the first device. The maximum time to detect a collision is approximately twice the
signal propagation time between the two most-distant devices. This maximum time is
traded-off by the collision recovery time and the diameter of the LAN.
In the original 802.3 specification, Ethernet operates in half duplex only. Under this condition,
when in 10Mbps LAN, it’s 2500 meters, in 100Mbps LAN, it’s approximately 200 meters and in
1000Mbps, 200 meters. According to the theory, it should be 20 meters. But it’s not practical,
so the LAN diameter is kept by using to increase the minimum frame size with a
variable-length non-data extension bit field which is removed at the receiving MAC. The
following tables are the frame format suitable for 10M, 100M and 1000M Ethernet, and some
parameter values that shall be applied to all of these three types of Ethernet.
Actually, the practice Gigabit Ethernet chips do not feature this so far. They all have their
chips supported full-duplex mode only, as well as all network vendors’ devices. So this
criterion should not exist at the present time and in the future. The switch’s Gigabit module
supports only full-duplex mode.