Time-Division Multiplexing:

Time-division multiplexing (TDM) is a digital process that allows several connections to share
the high bandwidth of a linle Instead of sharing a portion of the bandwidth as in FDM, time is
shared. Each connection occupies a portion of time in the link. The following figure gives a
conceptual view of TDM.

We can divide TDM into two different schemes: synchronous and statistical.
Synchronous Time-Division Multiplexing: In synchronous TDM, each input connection has an
allotment in the output even if it is not sending data. In synchronous TDM, the data flow of
each input connection is divided into units, where each input occupies one input time slot.
A unit can be 1 bit, one character, or one block of data. Each input unit becomes one output
unit and occupies one output time slot. However, the duration of an output time slot is n times
shorter than the duration of an input time slot. If an input time slot is T s, the output time slot is
T/n s, where n is the number of connections. In other words, a unit in the output connection
has a shorter duration; it travels faster. The following figure shows an example of synchronous
TDM where n is 3.

In synchronous TDM, a round of data units from each input connection is collected into a
frame. If we have n connections, a frame is divided into n time slots and one slot is allocated for
each unit, one for each input line. If the duration of the input unit is T, the duration of each slot
is Tin and the duration of each frame is T.

Time slots are grouped into frames. A frame consists of one complete cycle of time slots, with
one slot dedicated to each sending device. In a system with n input lines, each frame has n
slots, with each slot allocated to carrying data from a specific input line.
Interleaving
TDM can be visualized as two fast-rotating switches, one on the multiplexing side and the other
on the demultiplexing side. The switches are synchronized and rotate at the same speed, but in
opposite directions. On the multiplexing side, as the switch opens in front of a connection, that
connection has the opportunity to send a unit onto the path. This process is called interleaving.
On the demultiplexing side, as the switch opens in front of a connection, that connection has
the opportunity to receive a unit from the path. The following figure shows the interleaving
process.

Empty Slots
Synchronous TDM is not as efficient as it could be. If a source does not have data to send, the
corresponding slot in the output frame is empty. The following figure shows a case in which one
of the input lines has no data to send and one slot in another input line has discontinuous data.

The first output frame has three slots filled, the second frame has two slots filled, and the third
frame has three slots filled. No frame is full. We learn in the next section that statistical TDM
can improve the efficiency by removing the empty slots from the frame.
Data Rate Management
One problem with TDM is how to handle a disparity in the input data rates. If data rates are not
the same, three strategies, or a combination of them, can be used. The three different
strategies are multilevel multiplexing, multiple-slot allocation, and pulse stuffing.
Multilevel Multiplexing: Multilevel multiplexing is a technique used when the data rate of an
input line is a multiple of others. For example, in the following figure, we have two inputs of 20
kbps and three inputs of 40 kbps. The first two input lines can be multiplexed together to
provide a data rate equal to the last three. A second level of multiplexing can create an output
of 160 kbps.

Multiple-Slot Allocation: Sometimes it is more efficient to allot more than one slot in a frame
to a single input line. For example, we might have an input line that has a data rate that is a
multiple of another input. In the following figure, the input line with a SO-kbps data rate can be
given two slots in the output. We insert a serial-to-parallel converter in the line to make two
inputs out of one.

Pulse Stuffing: Sometimes the bit rates of sources are not multiple integers of each other.
Therefore, neither of the above two techniques can be applied. One solution is to make the
highest input data rate the dominant data rate and then add dummy bits to the input lines with
lower rates. This will increase their rates. This technique is called pulse stuffing, bit padding, or
bit stuffing as shown in the following figure. The input with a data rate of 46 is pulse-stuffed to
increase the rate to 50 kbps. Now multiplexing can take place.

Frame Synchronizing
The implementation of TDM is not as simple as that of FDM. Synchronization between the
multiplexer and demultiplexer is a major issue. If the, multiplexer and the demultiplexer are not
synchronized, a bit belonging to one channel may be received by the wrong channel.
For this reason, one or more synchronization bits are usually added to the beginning of each
frame. These bits, called framing bits, follow a pattern, frame to frame, that allows the
demultiplexer to synchronize with the incoming stream so that it can separate the time slots
accurately. In most cases, this synchronization information consists of 1 bit per frame,
alternating between 0 and I, as shown in the following figure.

Statistical Time-Division Multiplexing:

In synchronous TDM, each input has a reserved slot in the output frame. This can be inefficient
if some input lines have no data to send. In statistical time-division multiplexing, slots are
dynamically allocated to improve bandwidth efficiency. Only when an input line has a slot's
worth of data to send is it given a slot in the output frame. In statistical multiplexing, the
number of slots in each frame is less than the number of input lines. The multiplexer checks
each input line in round robin fashion. It allocates a slot for an input line if the line has data to
send otherwise it skips the line and checks the next line.
The following figure shows a synchronous and a statistical TDM example. In the former, some
slots are empty because the corresponding line does not have data to send. In the latter,
however, no slot is left empty as long as there are data to be sent by any input line.

Addressing:
The above figure also shows a major difference between slots in synchronous TDM and
statistical TDM. An output slot in synchronous TDM is totally occupied by data, in statistical
TDM, a slot needs to carry data as well as the address of the destination.
In synchronous TDM, there is no need for addressing. Synchronization and preassigned
relationships between the inputs and outputs serve as an address. We know, for example, that
input 1 always goes to input 1. If the multiplexer and the demultiplexer are synchronized, this is
guaranteed. In statistical multiplexing, there is no fixed relationship between the inputs and

outputs because there are no preassigned or reserved slots. We need to include the address of
the receiver inside each slot to show where it is to be delivered.
The addressing in its simplest form can be n bits to define N different output lines with
n =log 2 n. For example, for eight different output lines, we need a 3-bit address.
Slot Size
Since a slot carries both data and an address in statistical TDM, the ratio of the data size to
address size must be reasonable to make transmission efficient. For example, it would be
inefficient to send 1 bit per slot as data when the address is 3 bits. This would mean an
overhead of 300 percent. In statistical TDM, a block of data is usually many bytes while the
address is just a few bytes.
No Synchronization Bit
There is another difference between synchronous and statistical TDM, but this time it is at the
frame level. The frames in statistical TDM need not be synchronized, so we do not need
synchronization bits.
Bandwidth
In statistical TDM, the capacity of the link is normally less than the sum of the capacities of each
channel. The designers of statistical TDM define the capacity of the link based on the statistics
of the load for each channel. If on average only x percent of the input slots are filled, the
capacity of the link reflects this. Of course, during peak times, some slots need to wait.