NICs: (Network Interface Cards)

• Handle most of the "housework" of network transmission, and thus relieve the load on the CPU.
• Receives data from CPU
• Transmits at rate proper for the network
• Packages data in correct frame format for LAN protocol (Ethernet, Token Ring, etc.)
• Checks destination address on incoming frames; if not directed to its CPUdiscards. If meant for its CPU, makes a copy, send it to memory, and interrupts CPU
• Computes checksum for error detection

• Convert bits into electrical signals
• Convert electrical signals into bits
• Detect carrier waves

AUI (Attachment Unit Interface) cable

Types of Ethernet cabling

To increase the number of computers which can be attached to a bus, a connection multiplexor is used. It acts as a regular transceiver would (reports collisions, etc.), but allows multiple computers to be linked to the bus through it. 
 

Thinnet wiring (10Base2) is also coax: it's cheaper than thicknet (10Base5); it's more flexible; it doesn't need a transceiver because those functions are built into the NIC.

 Twisted Pair Ethernet (10Base-T) differs from thicknet and thinnet because its bus topology is realized through a hub. 

This is an apparent contradiction; a TP ethernet LAN looks suspiciously like a star topology. Physically, TP Ethernet is a star topology; logically (how the protocol functions) it is a bus. Hence IEEE standard 802.3 doesn't care if the wiring is a true bus like thicknet or thinnet; thicknet, thinnet, and TP can all interact transparently. 

Ethernet can use analog, but it most often uses digital transmission, which means that digital data is transformed into electrical impulses when it's transmitted. We recall there are two problems: first, if 0 is represented by a lack of voltage, how do we know when data is being transmitted or the line is just dead? Second, how do we get all the stations exactly synchronized one one clock?

Two methods of encoding address these problems, Manchester Encoding and Differential Manchester Encoding. Both divide the representation of binary digits into bit periods, where there is a change in voltage in the middle of the period.

In Manchester Encoding, binary 1s are designated when voltage is high during the first half of the bit period, and then drops down. Binary 0s are designated when voltage is low during the first half of the bit period and then jumps up. Unfortunately this requires twice as much bandwidth as straight binary encoding, but it does solve the problem of not being able to distinguish a 0 from a lack of voltage.

Differential Manchester Encoding is, well, different. Here 0s are indicated by a transition in voltage level at the start of the bit period. 1s are indicated by a lack of transition in voltage at the start of the bit period. There's still a transition in the middle of the bit period. This diagram will help a lot:
 
 
 

High voltage is +0.85 volts, and low voltage is -0.85 volts, which gives a DC value of 0.