In a parallel Physical layer arrangement, multiple communications paths lie between communicating parties (see the top half of the Image shown below ). Typically, all but one of the paths is used for data transfer; a single path is reserved for the transfer of timing information between the sender and the receiver. As such, multiple data units (typically bits) can be transmitted simultaneously over the interface. As a result of this characteristic, parallel interfaces typically exhibit high transmission rates. Unfortunately, this advantage is somewhat offset by a phenomenon called skew, in which circuits do not respond evenly to a propagated signal. Because each of the physical paths that comprise the parallel channel (typically copper wires) exhibit different electrical characteristics (such as resistance), the bits on each path may behave differently. For example, electrical waves may travel at different relative speeds over each of the paths. When this happens, a set of bits placed on the channel at a single point in time at the transmitter may arrive at the receiver at different times. This effect is exacerbated with distance, such that, after traveling a certain distance from the transmitter, the bits may become unrecoverable at the receiver. Due to this skew phenomenon, parallel interfaces are typically distance-limited. For example, IBM uses a parallel Physical layer interface to interconnect mainframe computers and their peripherals. This so-called bus-and-tag interface runs at 36Mbps but is limited to a mere 400 feet. In a serial Physical layer arrangement, a single communications path exists between communicating parties (see the bottom of below Image).

Thus, bits must follow one another down the channel between the transmitter and the receiver. The inability to send data simultaneously down multiple paths tends to limit data rates in serial environments. This is not to say that serial interfaces are inherently slow, but rather to point out that, all other things being equal, parallel interfaces display a speed advantage over serial ones. Of course, the problem of skew disappears in serial environments, making them more suited to long-distance communications. An issue that emerges in serial environments, however, is that of timing and synchronization between the transmitter and the receiver. With only one functional path available, serial approaches must make use of techniques by which timing information can be “extracted” from the data themselves. We will discuss this issue at length later in this chapter in the section “Timing and Synchronization.”