Last modified 2/13/95.
Earlier pages have considered signal transmission, data encoding, and framing and synchronization, all of which are needed for multiplexing, the topic of this page. The previous page considered datalink protocols among peers (balanced mode), while this one is primarily concerned with permitting multiple virtual channels to share the same physical channel.
There are three basic forms of multiplexing:
In SDM, separate wave guides (e.g., wires, fiber, etc.) are used so that two physical channels do not interfere with one another. The entire bandwidth of both physical channels may be used simultaneously. A standard example of this is the multiple wires used in a bus, or ribbon cable, or telephones, etc. In this case, there is not a need for a central mux, except to act as a switch. Circuit switching is considered in the next page.
FDM allows multiple stations to transmit on a shared medium (unlike SDM) simultaneously, each transmitter using only a portion of the available bandwidth. There are two main ways of doing this.
Here, a band-limited signal is shifted to a frequency band that is not shared with any other signal. Several such signals may be combined in a channel having larger bandwidth available, with the total spectrum of the channel divided into smaller bands, one per logical channel, with guard bands between them. Each input signal is convolved with the carrier frequency for the band to which is assigned. The receiver runs the received signal through a band-pass filter that eliminates all but the band of interest, and then down- converts the signal to its original form.
CDM convolves the original digital signal with a spreading code, which has the effect of spreading the spectrum of the signal greatly and reducing the power over any one part of the spectrum. The code must have special properties, such as low auto-correlation, in order for the receiver to recover synchronization and the original signal. The receiver must use the same spreading code and convolve it in synchronization with the sender in order to decode the signal. Multiple stations may spread their signals over the same spectrum this way, as long as each has a spreading code that has low cross- correlation with the other stations' spreading codes. A simple way to think of this is that a single bit may be transmitted by modulating a series of signal elements at different frequencies in some particular order. The receiver checks these frequencies in the correct order to see if they are modulated in a way consistent with a 0 or a 1 bit to decode. The number of different frequencies per bit is the chip rate (each partial bit is a chip). If the chip rate is less than one, then one or more bits are sent at the same frequency, and this is called Frequency Hopping. The main advantages that CDM has are as follows:
TDM allows multiple logical channels to share the same physical channel by taking turns using the entire spectrum of the channel. These turns are generally slots within a larger, fixed size frame. Framing is required, since the stations need to synchronize in order to locate a slot for both transmission and reception. If mixing of the signals is done in a distributed manner (i.e., it is not the case that multiple stations attach to one central multiplexer that combines them), then guard periods are needed between slots to account for minor clock variations between stations.
This is the standard version, and is what is usually meant by TDM if no other modifiers are used. Synchronous refers to the fixed allocation of slots to particular logical channels, rather than the way that data are transferred over the communication lines themselves. Here, a station is given a fixed (set of) slot(s) within a larger frame, and transmits its data (or has the mux send its data) during the slot(s in that set). A variety of data rates can be supported easily by allocating a different number of slots to the various stations. No addressing is needed since it is known ahead time which stations use which slots. The sum of the data rates of the stations must be no greater than the net data rate of the shared channel (after guard periods if needed and framing overhead is subtracted). If a central mux is used, then it must buffer the input from each station until the next slot for that station is to be sent.
If the data sources do not occupy a nice round number of slots, then the mux must somehow provide for this. Pulse stuffing is commonly used, and is a method by which a source at one (nasty) data rate has some extra bits added to its stream in fixed locations (so that the receiver can take them back out) in order to raise it to a nice data rate that occupies the desired number of slots.
The fact that for most digital data communications, stations are idle most of the time, and data is transferred in bursts, means that the strict TDM does not provide good efficiency (most slots go idle) or good service (the unused slots cannot be used to speed up transfer for a busy station). STDM does not assign fixed slots, but assigns them on a demand basis. Since a slot number no longer corresponds to a unique station, addresses must be included with each slot to identify the sender or receiver (or both). This extra overhead buys the capability of the STDM to give extra slots to a busy station when it needs them, giving no slots to idle stations.
To handle bursts of data from multiple stations arriving simultaneously, a central STDM must be able to buffer a certain amount of data for later transmission. Since this is a statistical process, it is entirely possible that all the stations could send a burst simultaneously, which could overwhelm the buffering capacity of the mux. The mux in that case either discards the excess data, or it uses some form of flow control to slow down the rate at which the stations send. The sum of the stations' data rates is generally greater than the net data rate of the shared channel (after framing and addressing overhead is taken into account). The sum of the average data rates of all the stations must be less than the net capacity of the shared channel, or else the queuing system is unstable.
In order to provide for lower costs and still maintain flexibility, hybrid techniques are often used. In fact, for wireless communication, federal and international laws state what frequency bands may be used for what applications and by whom. This carves up the spectrum into frequency bands at the top level, so even if TDM is used within one of those bands, it is already part of an overall hybrid scheme in some sense. Even within a band, it may be desirable to have slower electronics (FDM) since price does not scale linearly with speed. A method that can be used within a band is MF/TDM, or multi-frequency TDM. Here, the band is broken into sub-bands, each of which may be multiplexed using TDM. The transmitters do not have to transmit at the full burst rate of the channel, and yet the advantages of TDM are still available.