Communication Networks/Channels

Channels

A channel is a communication medium, the path that data takes from source to destination. A channel can be comprised of so many different things: wires, free space, and entire networks. Signals can be routed from one type of network to another network with completely different characteristics. In the Internet, a packet may be sent over a wireless WiFi network to an ethernet lan, to a DSL modem, to a fiber-optic backbone, et cetera. The many unique physical characteristics of different channels determine the three characteristics of interest in communication: the latency, the data rate, and the reliability of the channel.

Bandwidth and Bitrate

Bandwidth is the difference between the upper and lower cutoff frequencies of, for example, a filter, a communication channel, or a signal spectrum. Bandwidth, like frequency, is measured in hertz (Hz). The bandwidth can be physically measured using a spectrum analyzer.

Bandwidth, given by the variables B_w or W is closely related to the amount of digital bits that can be reliably sent over a given channel:

$r_b=2W$

where r_b is the bitrate. If we have an M-ary signaling scheme with m levels, we can expand the previous equation to find the maximum bit rate for the given bandwidth.

$r_b=2Wm$

Channel Capacity

The "capacity" of a channel is the theoretical upper-limit to the bit rate over a given channel that will result in negligible errors. Channel capacity is measured in bits/s.

Shannon's channel capacity is an equation that determines the information capacity of a channel from a few physical characteristics of the channel. A communication systems can attempt to exceed the Shannon's capacity of a given channel, but there will be many errors in transmission, and the expense is generally not worth the effort. Shannon's capacity, therefore, is the theoretical maximum bit rate below which information can be transmitted with negligible errors.

The Shannon channel capacity, C, is measured in units of bits/sec and is given by the equation:

$C=W{\log }_2(1+SNR)$

C is the maximum capacity of the channel, W is the available bandwidth in the channel, and SNR is the signal to noise ratio, not in DB.

Because channel capacity is proportional to analog bandwidth, some people call it "digital bandwidth".

Acknowledgement

Digital information packets have a number of overhead bits known as a header. This is because most digital systems use statistical TDM (as discussed in the Time-Division Multiplexing chapter). The total amount of bits sent in a transmission must be at least the sum of the data bits and the header bits. The total number of bits transmitted per second (the "throughput") is always less than the theoretical capacity. Because some of this throughput is used for these header bits, the number of data bits transmitted per second (the "goodput") is always less than the throughput.

In addition, since we all want our information to be transmitted reliably, it makes good sense for an intelligent transmitter and an intelligent receiver to communicate directly to each other, to ensure reliable transmission. This is called acknowledgement, and the process is called hand-shaking.

An essential part of acknowledgement is forward error correction, a subject that we will talk about more in depth later. Forward error correction (FEC) is the process of embedding some sort of checksum (called a CRC sum in IP communications) into the packet header. The receiver then reads this checksum, determines if there is an error in the transmission, and then sends back an acknowledgement packet.

ARQ: ACK and NAK

In an acknowledgement request (ARQ) scheme, the transmitter sends out data packets, and the receiver will then send back an acknowledgement. A positive acknowledgement (called "ACK") means that the packet was received without any detectable errors. A negative acknowledgement (called "NAK") means that the packet was received in error. Generally, when a NAK is received by the transmitter, the transmitter will send the packet again.

Streaming Packets

In some streaming protocols, such as RTP, the transmitter is sending time-sensitive data, and it can therefore not afford to wait for acknowledgement packets and checksum calculations. In these types of systems, the receiver will attempt to detect errors in the received packets, and if an error is found, the bad packet is simply deleted.