SYNCHRONOUS TIME DIVISION MULTIPLEXING

SYNCHRONOUS TIME DIVISION MULTIPLEXING

Characteristics

Synchronous time-division multiplexing is possible when the achievable data rate

(sometimes, unfortunately, called bandwidth) of the medium exceeds the data rate

of digital signals to be transmitted. Multiple digital signals (or analog signals carrying

digital data) can be carried on a single transmission path by interleaving portions

of each signal in time. The interleaving can be at the bit level or in blocks of

bytes or larger quantities. For example, the multiplexer in Figure 7.2b has six inputs

which might each be, say, 9.6 kbps. A single line with a capacity of at least 57.6 kbps

(plus overhead capacity) could accommodate all six sources.

A generic depiction of a synchronous TDM system is provided in Figure 7.6.

A number of signals [mt(t), i = 1, N] are to be multiplexed onto the same transmission

medium. The signals carry digital data and are generally digital signals. The

incoming data from each source are briefly buffered. Each buffer is typically one bit

or one character in length. The buffers are scanned sequentially to form a composite

digital data stream mc(t). The scan operation is sufficiently rapid so that each

buffer is emptied before more data can arrive. Thus, the data rate of mc(t) must at

least equal the sum of the data rates of the mi(t). The digital signal mc(t) may be

transmitted directly or passed through a modem so that an analog signal is transmitted.

In either case, transmission is typically synchronous.

The transmitted data may have a format something like Figure 7.6b. The data

are organized into frames. Each frame contains a cycle of time slots. In each frame,

one or more slots is dedicated to each data source. The sequence of slots dedicated

to one source, from frame to frame, is called a channel. The slot length equals the

transmitter buffer length, typically a bit or a character.

The character-interleaving technique is used with asynchronous sources. Each

time slot contains one character of data. Typically, the start and stop bits of each

character are eliminated before transmission and reinserted by the receiver, thus

improving efficiency. The bit-interleaving technique is used with synchronous

sources and may also be used with asynchronous sources. Each time slot contains

just one bit.

At the receiver, the interleaved data are demultiplexed and routed to the appropriate

destination buffer. For each input source mi(t), there is an identical

output source which will receive the input data at the same rate at which it was

generated.

Synchronous TDM is called synchronous not because synchronous transmission

is used, but because the time slots are preassigned to sources and fixed. The

time slots for each source are transmitted whether or not the source has data to

send; this is, of course, also the case with FDM. In both cases, capacity is wasted to

achieve simplicity of implementation. Even when fixed assignment is used, however,

it is possible for a synchronous TDM device to handle sources of different data

rates. For example, the slowest input device could be assigned one slot per cycle,

while faster devices are assigned multiple slots per cycle.

TDM Link Control

The reader will note that the transmitted data stream depicted in Figure 7.6 does not

contain the headers and trailers that we have come to associate with synchronous

transmission. The reason is that the control mechanisms provided by a data link protocol

are not needed. It is instructive to ponder this point, and we do so by considering

two key data link control mechanisms: flow control and error control. It should

be clear that, as far as the multiplexer and demultiplexer (Figure 7.1) are concerned,

flow control is not needed. The data rate on the multiplexed line is fixed, and the

multiplexer and demultiplexer are designed to operate at that rate. But suppose that

one of the individual output lines attaches to a device that is temporarily unable to

accept data? Should the transmission of TDM frames cease? Clearly not, as the

remaining output lines are expecting to receive data at predetermined times. The

solution is for the saturated output device to cause the flow of data from the corresponding

input device to cease. Thus, for a while, the channel in question will carry

empty slots, but the frames as a whole will maintain the same transmission rate.

The reasoning for error control is the same. It would not do to request retransmission

of an entire TDM frame because an error occurs on one channel. The

devices using the other channels do not want a retransmission nor would they know

that a retransmission has been requested by some other device on another channel.

Again, the solution is to apply error control on a per-channel basis.

How are flow control, error control, and other good things to be provided on

a per-channel basis? The answer is simple: Use a data link control protocol such as

HDLC on a per-channel basis. A simplified example is shown in Figure 7.7. We

assume two data sources, each using HDLC. One is transmitting a stream of HDLC

frames containing three octets of data; the other is transmitting HDLC frames containing

four octets of data. For clarity, we assume that character-interleaved multiplexing

is used, although bit interleaving is more typical. Notice what is happening.

The octets of the HDLC frames from the two sources are shuffled together for

transmission over the multiplexed line. The reader may initially be uncomfortable

with this diagram, as the HDLC frames have lost their integrity in some sense. For

example, each frame check sequence (FCS) on the line applies to a disjointed set of

bits. Even the FCS is not in one piece! However, the pieces are reassembled correctly

before they are seen by the device on the other end of the HDLC protocol.

In this sense, the multiplexing/demultiplexing operation is transparent to the

attached stations; to each communicating pair of stations, it appears that they have

a dedicated link.

One refinement is needed in Figure 7.7. Both ends of the line need to be a

combination multiplexer/demultiplexer with a full-duplex line in between. Then

each channel consists of two sets of slots, one traveling in each direction. The individual

devices attached at each end can, in pairs, use HDLC to control their own

channel. The multiplexer/demultiplexers need not be concerned with these matters.

Framing

So we have seen that a link control protocol is not needed to manage the overall

TDM link. There is, however, a basic requirement for framing. Because we are not

providing flag or SYNC characters to bracket TDM frames, some means is needed

to assure frame synchronization. It is clearly important to maintain framing synchronization

because, if the source and destination are out of step, data on all channels

are lost.

Perhaps the most common mechanism for framing is known as added-digit

framing. In this scheme, typically, one control bit is added to each TDM frame. An

identifiable pattern of bits, from frame to frame, is used on this "control channel."

A typical example is the alternating bit pattern, 101010 . . . . This is a pattern

unlikely to be sustained on a data channel. Thus, to synchronize, a receiver compares

the incoming bits of one frame position to the expected pattern. If the pattern

does not match, successive bit positions are searched until the pattern persists over

multiple frames. Once framing synchronization is established, the receiver continues

to monitor the framing bit channel. If the pattern breaks down, the receiver

must again enter a framing search mode.

Pulse Stuffing

Perhaps the most difficult problem in the design of a synchronous time-division

multiplexer is that of synchronizing the various data sources. If each source has a

separate clock, any variation among clocks could cause loss of synchronization.

Also, in some cases, the data rates of the input data streams are not related by a simple

rational number. For both these problems, a technique known as pulse stuffing

is an effective remedy. With pulse stuffing, the outgoing data rate of the multiplexer,

excluding framing bits, is higher than the sum of the maximum instantaneous

incoming rates. The extra capacity is used by stuffing extra dummy bits or

pulses into each incoming signal until its rate is raised to that of a locally-generated

clock signal. The stuffed pulses are inserted at fixed locations in the multiplexer

frame format so that they may be identified and removed at the demultiplexer.

Example

An example, from [COUC95], illustrates the use of synchronous TDM to multiplex

digital and analog sources. Consider that there are 11 sources to be multiplexed on

a single link:

Source 1: Analog, 2-kHz bandwidth.

Source 2: Analog, 4-kHz bandwidth.

Source 3: Analog, 2-kHz bandwidth.

Sources 4-11: Digital, 7200 bps synchronous.

As a first step, the analog sources are converted to digital using PCM. Recall

from Lesson 4 that PCM is based on the sampling theorem, which dictates that a

signal be sampled at a rate equal to twice its bandwidth. Thus, the required sampling

rate is 4000 samples per second for sources 1 and 3, and 8000 samples per second

for source 2. These samples, which are analog (PAM), must then be quantized

or digitized. Let us assume that 4 bits are used for each analog sample. For convenience,

these three sources will be multiplexed first, as a unit. At a scan rate of

4 kHz, one PAM sample each is taken from sources 1 and 3, and two PAM samples

are taken from source 2 per scan. These four samples are interleaved and converted

to 4-bit PCM samples. Thus, a total of 16 bits is generated at a rate of 4000 times

per second, for a composite bit rate of 64 kbps.

For the digital sources, pulse stuffing is used to raise each source to a rate of

8 kbps, for an aggregate data rate of 64 kbps. A frame can consist of multiple cycles

of 32 bits, each containing 16 PCM bits and two bits from each of the eight digital

sources. Figure 7.8 depicts the result.

Digital Carrier Systems

The long-distance carrier system provided in the United States and throughout the

world was designed to transmit voice signals over high-capacity transmission links,

such as optical fiber, coaxial cable, and microwave. Part of the evolution of these

telecommunications networks toward digital technology has been the adoption of

synchronous TDM transmission structures. In the United States, AT&T developed

a hierarchy of TDM structures of various capacities; this structure is used in Canada

and Japan as well as in the United States. A similar, but unfortunately not identical,

hierarchy has been adopted internationally under the auspices of ITU-T

(Table 7.3).

The basis of the TDM hierarchy (in North America and Japan) is the DS-1

transmission format (Figure 7.9), which multiplexes 24 channels. Each frame contains

8 bits per channel plus a framing bit for 24 X 8 + 1 = 193 bits. For voice transmission,

the following rules apply. Each channel contains one word of digitized

voice data. The original analog voice signal is digitized using pulse code modulation

(PCM) at a rate of 8000 samples per second. Therefore, each channel slot and,

hence, each frame must repeat 8000 times per second. With a frame length of

193 bits, we have a data rate of 8000 X 193 = 1.544 Mbps. For five of every six

frames, 8-bit PCM samples are used. For every sixth frame, each channel contains

a 7-bit PCM word plus a signaling bit. The signaling bits form a stream for each

voice channel that contains network control and routing information. For example,

control signals are used to establish a connection or to terminate a call.

The same DS-1 format is used to provide digital data service. For compatibility

with voice, the same 1.544-Mbps data rate is used. In this case, 23 channels of

data

are provided. The twenty-fourth channel position is reserved for a special sync

byte, which allows faster and more reliable reframing following a framing error.

Within each channel, seven bits per frame are used for data, with the eighth bit used

to indicate whether the channel, for that frame, contains user data or system control

data. With seven bits per channel, and because each frame is repeated

8000 times per second, a data rate of 56 kbps can be provided per channel. Lower

data rates are provided using a technique known as subrate multiplexing. For this

technique, an additional bit is robbed from each channel to indicate which subrate

multiplexing rate is being provided; this leaves a total capacity per channel of

6 X 8000 = 48 kbps. This capacity is used to multiplex five 9.6-kbps channels, ten

4.8-kbps channels, or twenty 2.4-kbps channels. For example, if channel 2 is used to

provide 9.6-kbps service, then up to five data subchannels share this channel. The

data for each subchannel appear as six bits in channel 2 every fifth frame.

Finally, the DS-1 format can be used to carry a mixture of voice and data

channels. In this case, all 24 channels are utilized; no sync byte is provided.

Above this basic data rate of 1.544 Mbps, higher-level multiplexing is achieved

by interleaving bits from DS-1 inputs. For example, the DS-2 transmission system

combines four DS-1 inputs into a 6.312-Mbps stream. Data from the four sources

are interleaved 12 bits at a time. Note that 1.544 X 4 = 6.176 Mbps. The remaining

capacity is used for framing and control bits.

ISDN User-Network Interface

ISDN enables the user to multiplex traffic from a number of devices on the user's

premises over a single line into an ISDN (Integrated Services Digital Network).

Two interfaces are defined: a basic interface and a primary interface.

Basic ISDN Interface

At the interface between the subscriber and the network terminating equipment,

digital data are exchanged using full-duplex transmission. A separate physical line

is used for the transmission in each direction. The line coding specification for the

interface dictates the use of a pseudoternary coding scheme.' Binary one is represented

by the absence of voltage; binary zero is represented by a positive or negative

pulse of 750 mV +lo%. The data rate is 192 kbps.

The basic access structure consists of two 64-kbps B channels and one 16-kbps

D channel. These channels, which produce a load of 144 kbps, are multiplexed over

a 192-kbps interface at the S or T reference point. The remaining capacity is used

for various framing and synchronization purposes.

The B channel is the basic user channel. It can be used to carry digital data

(e.g., a personal computer connection), PCM-encoded digital voice (e.g., a telephone

connection), or any other traffic that can fit into a 64-kbps channel. At any

given time, a logical connection can be set up separately for each B channel to separate

ISDN destinations. The D channel can be used for a data-transmission connection

at a lower data rate. It is also used to carry control information needed to

set up and terminate the B-channel connections. Transmission on the D channel

consists of a sequence of LAPD frames.

As with any synchronous time-division multiplexed (TDM) scheme, basic

access transmission is structured into repetitive, fixed-length frames. In this case,

each frame is 48 bits long; at 192 kbps, frames must repeat at a rate of one frame

every 250 psec. Figure 7.10 shows the frame structure; the upper frame is transmitted

by the subscriber's terminal equipment (TE) to the network (NT); the lower

frame is transmitted from the NT to the TE.

Each frame of 48 bits includes 16 bits from each of the two B channels and

4 bits from the D channel. The remaining bits have the following interpretation. Let

us first consider the frame structure in the TE-to-NT direction. Each frame begins

with a framing bit (F) that is always transmitted as a positive pulse. This is followed

by a dc balancing bit (L) that is set to a negative pulse to balance the voltage. The

F-L pattern thus acts to synchronize the receiver on the beginning of the frame. The

specification dictates that, following these first two bit positions, the first occurrence

of a zero bit will be encoded as a negative pulse. After that, the pseudoternary rules

are observed. The next eight bits (Bl) are from the first B channel; this is followed

by another dc balancing bit (L). Next comes a bit from the D channel, followed by

its balancing bit. This is followed by the auxiliary framing bit (FA), which is set to

zero unless it is to be used in a multiframe structure. There follows another balancing

bit (L), eight bits (B2) from the second B channel, and another balancing bit (L);

this is followed by bits from the D channel, first B channel, D channel again, second

B channel, and the D channel yet again, with each group of channel bits followed

by a balancing bit.

The frame structure in the NT-to-TE direction is similar to the frame structure

for transmission in the TE-to-NT direction. The following new bits replace some of

the dc balancing bits. The D-channel echo bit (E) is a retransmission by the NT of

the most recently received D bit from the TE; the purpose of this echo is explained

below. The activation bit (A) is used to activate or deactivate a TE, allowing the

device to come on line or, when there is no activity, to be placed in low-powerconsumption

mode. The N bit is normally set to binary one. The N and M bits may

be used for multiframing. The S bit is reserved for other future standardization

requirements.

The E bit in the TE-to-NT direction comes into play to support a contention

resolution function, which is required when multiple TE1 terminals share a single

physical line (i.e., a multipoint line). There are three types of traffic to consider:

B-channel traffic. No additional functionality is needed to control access to

the two B channels, as each channel is dedicated to a particular TE at any

given time.

D-channel traffic. The D channel is available for use by all the subscriber

devices for both control signaling and packet transmission, so the potential for

contention exists. There are two subcases:

a Incoming traffic: The LAPD addressing scheme is sufficient to sort out the

proper destination for each data unit.

0 Outgoing traffic: Access must be regulated so that only one device at a time

transmits. This is the purpose of the contention-resolution algorithm.

The D-channel contention-resolution algorithm has the following elements:

1. When a subscriber device has no LAPD frames to transmit, it transmits a

series of binary ones on the D channel; using the pseudoternary encoding

scheme, this corresponds to the absence of line signal.

2. The NT, on receipt of a D-channel bit, reflects back the binary value as a

D-channel echo bit.

3. When a terminal is ready to transmit an LAPD frame, it listens to the stream

of incoming D-channel echo bits. If it detects a string of 1-bits equal in length

to a threshold value Xi, it may transmit; otherwise, the terminal must assume

that some other terminal is transmitting, and wait.

4. It may happen that several terminals are monitoring the echo stream and

begin to transmit at the same time, causing a collision. To overcome this condition,

a transmitting TE monitors the E bits and compares them to its transmitted

D bits. If a discrepancy is detected, the terminal ceases to transmit and

returns to a listen state.

The electrical characteristics of the interface (i.e., 1-bit = absence of signal)

are such that any user equipment transmitting a 0-bit will override user equipment

transmitting a 1-bit at the same instant. This arrangement ensures that one device

will be guaranteed successful completion of its transmission.

The algorithm includes a primitive priority mechanism based on the threshold

value Xi. Control information is given priority over user data. Within each of these

two priority classes, a station begins at normal priority and then is reduced to lower

priority after a transmission. It remains at the lower priority until all other terminals

have had an opportunity to transmit. The values of Xi are as follows:

Control Information

Normal priority XI = 8

Lower priority XI = 9

User Data

Normal priority X2 = 10

Lower priority X2 = 11

Primary ISDN Interface

The primary interface, like the basic interface, multiplexes multiple channels across

a single transmission medium. In the case of the primary interface, only a point-topoint

configuration is allowed. Typically, the interface supports a digital PBX or

other concentration device controlling multiple TEs and providing a synchronous

TDM facility for access to ISDN. Two data rates are defined for the primary interface:

1.544 Mbps and 2.048 Mbps.

The ISDN interface at 1.544 Mbps is based on the North American DS-1

transmission structure, which is used on the T1 transmission service. Figure 7.11a

illustrates the frame format for this data rate. The bit stream is structured into

repetitive 193-bit frames. Each frame consists of 24 8-bit time slots and a framing

bit, which is used for synchronization and other management purposes. The same

time slot repeated over multiple frames constitutes a channel. At a data rate of

1.544 Mbps, frames repeat at a rate of one every 125 psec, or 8000 frames per second.

Thus, each channel supports 64 kbps. Typically, the transmission structure is

used to support 23 B channels and 1 64-kbps D channel.

The line coding for the 1.544-Mbps interface is AM1 (Alternate Mark Inversion)

using B8ZS.

The ISDN interface at 2.048 Mbps is based on the European transmission

structure of the same data rate. Figure 7.11b illustrates the frame format for this

data rate. The bit stream is structured into repetitive 256-bit frames. Each frame

consists of 32 &bit time slots. The first time slot is used for framing and synchronization

purposes; the remaining 31 time slots support user channels. At a data rate

of 2.048 Mbps, frames repeat at a rate of one every 125 psec, or 8000 frames per second.

Thus, each channel supports 64 kbps. Typically, the transmission structure is

used to support 30 B channels and 1 D channel.

The line coding for the 2.048-Mbps interface is AM1 using HDB3.

SONET/SDH

SONET (Synchronous Optical Network) is an optical transmission interface originally

proposed by BellCore and standardized by ANSI. A compatible version,

referred to as Synchronous Digital Hierarchy (SDH), has been published by ITU-T

in Recommendations G.707, G.708, and G.709.~S ONET is intended to provide a

specification for taking advantage of the high-speed digital transmission capability

of optical fiber.

Signal Hierarchy

The SONET specification defines a hierarchy of standardized digital data rates

(Table 7.4). The lowest level, referred to as STS-1 (Synchronous Transport Signal,

level 1) or OC-1 (Optical Carrier level I ) , ~is 51.84 Mbps. This rate can be used to

carry a single DS-3 signal or a group of lower-rate signals, such as DS1, DSlC, DS2,

plus ITU-T rates (e.g., 2.048 Mbps).

Multiple STS-1 signals can be combined to form an STS-N signal. The signal

is created by interleaving bytes from N STS-1 signals that are mutually synchronized.

For the ITU-T Synchronous Digital Hierarchy, the lowest rate is 155.52 Mbps,

which is designated STM-1. This corresponds to SONET STS-3. The reason for the

discrepancy is that STM-1 is the lowest-rate signal that can accommodate an

ITU-T level 4 signal (139.264 Mbps).

Frame Format

The basic SONET building block is the STS-1 frame, which consists of 810 octets

and is transmitted once every 125 ps, for an overall data rate of 51.84 Mbps (Figure

7.12a). The frame can logically be viewed as a matrix of 9 rows of 90 octets each,

with transmission being one row at a time, from left to right and top to bottom.

The first three columns (3 octets X 9 rows = 27 octets) of the frame are

devoted to overhead octets. Nine octets are devoted to section-related overhead

and 18 octets are devoted to line overhead. Figure 7.13a shows the arrangement of

overhead octets, and Table 7.5 defines the various fields.

The remainder of the frame is payload, which is provided by the path layer.

The payload includes a column of path overhead, which is not necessarily in the first

available column position; the line overhead contains a pointer that indicates where

the path overhead starts. Figure 7.13b shows the arrangement of path overhead

octets, and Table 7.5 defines these.

Figure 7.12b shows the general format for higher-rate frames, using the

ITU-T designation.