Exploiting The Delay Properties Of Fiber Optic Cable For LAN Extension:Brief History of Local Area Networks

<< Exploiting The Bandwidth Of Fiber Optic Cable-Employment by Multiple Users:Sharing the Transmission Medium

Exploiting The Advantages Of Fiber Optic Cable In the Industrial Environment:The Problem of Interference >>

Page 45 of 97

place time slot boundaries accurately. Protection against TDM cross-talk is achieved by putting guard

times in the slots. Data is not packed end-to-end in a time slot. Rather, there is either a dead space, or

dummy bits or some other mechanism built into the TDM protocol so that if data slides from one slot to

another its impact on BER is minimal.

With WDM cross-talk arises because the optical signal spectrum for a given link placed upon one

particular (center) wavelength is not bounded in wavelength (equivalently frequency). This is a

consequence of it being a physical signal that can actually be generated. The optical signal spectrum will

spill over onto the optical signal spectrum for another link placed at another (center) wavelength. The

amount of spillage depends upon how close the wavelengths are and how much optical filtering is built

into the WDM to buffer it. The protection against cross-talk here is measured by a parameter called

isolation. This is the attenuation (dB) of the optical signal placed at one (center) wavelength as measured

at another (center) wavelength. The greater the attenuation the less effective spillage and the less impact

on BER.

At the present time, clock stability for digital circuitry is such that TDM cross-talk presents no real

impact on BER in the context of premises data communications and at the composite link speeds that

can be accommodated. The TDM cross-talk situation may be different when considering WANs.

However, this is the case in the premise environment. The situation is not as good for WDM. Here,

depending upon the specific WDM design, the amount of isolation may vary from a low value of 16 dB

all the way to 50 dB. A low value of isolation means that the impact upon BER could be significant. In

such situations WDM is limited to communications applications that can tolerate a high BER. Digital

voice and video would be in this group. However, LAN traffic would not be in this group. From the

perspective of BER generated by cross-talk TDM is more favorable than WDM.

CHAPTER 4

EXPLOITING THE DELAY PROPERTIES OF FIBER OPTIC CABLE FOR

LOCAL AREA NETWORK (LAN) EXTENSION

Fiber optic cable provides a way for extending reach of Local Area Networks (LANs). If you are well

versed on the subject of LANs you are welcome to jump right into this subject and skip the next two

subchapters. However, if you have not been initiated into LAN technology then you will find the

subjects covered in these next two subchapters worthwhile reading.

4.1 Brief History of Local Area Networks

Two full generations ago, in the early days of the data revolution, each computer served only a single

user. In the computer room (or at that time 'the building') of an installation there was 1 CPU, 1 keyboard,

1 card reader, (maybe) 1 magnetic tape reader, 1 printer, 1 keypunch machine etc. From a usage point of

view this was highly inefficient. Most data processing managers were concerned that this highly

expensive equipment spent most of the time waiting for users to employ it. Most data processing

managers knew this looked bad to the Controllers of their organizations. This led to the pioneering

development of time-sharing operating systems by MIT with Project MAC.

Time sharing opened up computational equipment to more than 1 user. Whole departments, companies,

schools etc. began making use of the expensive computational equipment. A key element in time sharing

Page 46 of 97

systems concerned the keyboard. A computer terminal replaced it. The multiple terminals were

connected to the CPU by data communications links. There was a marriage of computation and data

communications. In particular, the data communications was mostly (though not exclusively) premises

data communications.

Throughout the years time sharing led to distributed computation. The idea of distributed computation

being that applications programs would reside on one central computer called the Server. Applications

users would reside at PCs. When an applications user wanted to run a program a copy of it would be

downloaded to him/her. In this way multiple users could work with the same program simultaneously.

This was much more efficient than the original time sharing. Distributed computation required a data

communications network to tie the Server to the PCs and peripherals. This network was called a Local

Area Network (LAN). This network had to have high bandwidth. In fact, it had to accommodate speeds

that were orders of magnitude greater than the original time sharing networks. Entire applications

programs had to be downloaded to multiple users. Files, the results of running applications programs,

had to be uploaded to be stored in central memory.

LANs first came on the scene in a noticeable sense in the late 1970's. From that time until the present

many flavors of LANs have been offered in the marketplace. There are still a number of different flavors

each with its group of advocates and cult following. However, some time around the late 1980's the

market place began to recognize Ethernet as the flavor of choice. All of the discussion in the sequel will

concern only Ethernet.

The Ethernet LAN architecture had its origins in work done at Xerox Palo Alto Research Center

(PARC) by Robert Metcalf in the early 1970's. Metcalf later went on to become the founder of 3COM.

Xerox was later joined by DEC and Intel in promoting Ethernet as the coming LAN standard. In the

development of the Ethernet LAN architecture Metcalf built upon previous research funded by the

Advanced Research Projects Agency (ARPA) at the University of Hawaii. This ARPA program was

concerned with an asynchronous multiple access data communications technique called ALOHA.

The basic operation of an Ethernet LAN can be briefly explained with the aid of Figure 4-1. This

illustration indicates various data equipment that all need to communicate with each other. The data

equipment constitute the users of the LAN. Each is a Source and User within the context of the

discussion of Chapter 1. The location on the LAN of each data equipment unit is termed the station.

Figure 4-1: Ethernet Bus architecture

The communication between the data equipment is accomplished by having all the data equipment tap

onto a Transmission Medium. Each station taps onto the Transmission Medium. The Transmission

Medium is typically some type of cable. As shown in Figure 4-1 it is labeled Broadcast Channel - The

Page 47 of 97

Ethernet Bus. The Bus Interface Units (BIUs) provide the essential interfacing at a station between the

data equipment and the Broadcast Channel. That is, they provide the transmit/receive capability and all

needed intelligence.

It is an essential feature of the Ethernet LAN architecture that any data equipment can transmit to any

other data equipment and any data equipment can listen to all transmissions on the Broadcast Channel,

whether intended for it or for some other data equipment user. Implicitly, the Ethernet architecture

assumes that there is no coordination in the transmissions of the different data equipment. This is quite a

bit different from the sharing of a Transmission Medium by TDM where coordination is essential.

Transmitted data only goes in its assigned slot.

Now how does an Ethernet LAN operate? It operates by making use of three essential items. First, it

employs a Carrier Sense Multiple Access/Collision Detection (CSMA/CD) protocol. Secondly, data to

be communicated is enveloped in packets that have the addresses of the data equipment units

communicating. The packet has the address of the equipment sending data (the origin) and the data

equipment that is the intended recipient (the destination). Thirdly, the Ethernet Bus - the Transmission

Medium - is taken as passive and supports broadcast type transmissions. The way in which the Ethernet

LAN architecture uses these items is explained briefly below.

Consider a specific data equipment unit at its station. This will be our data equipment unit, station and

BIU of interest. For the sake of an example, suppose it is a PC wanting to communicate with the

Computer with File Server at its station. Before attempting to transmit a data packet onto the Ethernet

Bus our terminal's BIU first listens to determine if the Bus is idle. That is, it listens to determine if there

are any other packets from other data equipment already on the Bus. It attempts to sense the presence of

a communication signal representing a packet, a carrier, on the Bus. Our BIU and any BIU have

circuitry to perform this Carrier Sensing. An active BIU transmits its packet on the Bus only if the Bus

has been sensed as idle. In other words, it only transmits its packet if it has determined that no other

packet is already on the Bus - carrier is absent. If the Bus is sensed as busy- carrier is present- then the

BIU defers its transmission until the Bus is sensed as idle again. This procedure allows the various data

equipment to operate asynchronously yet avoid interfering with one another's communications.

However, it may be that a carrier has not sensed an existing packet is already on the Bus. Transmission

of a packet by the BIU of interest begins but there are still problems. There are propagation delays and

carrier detection processing delays. Because of these, it may be that the packet from our PC's BIU still

interferes with, or collides with, a packet transmitted by another equipment's BIU. This interfering

packet is one that has not yet reached our BIU by the end of the interval in which it had performed the

carrier sensing. A BIU monitors the transmission of the packet it is sending out to determine if it does

collide with another packet. To do this it makes use of the broadcast nature of the transmission medium.

A BIU can monitor what it has put on the Ethernet Bus and also any other traffic on the Ethernet Bus.

Our BIU and any BIU has circuitry to perform Collision Detection. The BIU that transmitted the

interfering packet also has circuitry to perform Collision Detection.

When both BIUs sense a collision they cease transmitting. Each BIU then waits a random amount of

time before re-transmitting - that is sensing for carrier and transmitting the packet onto the bus. If

another collision occurs then this random time wait is repeated but increased. In fact, it is increased at an

exponential rate until the collision event disappears. This approach to getting out of collisions is called

exponential back off.

4.2 Transmission Media Used To Implement An Ethernet LAN

Let us direct attention now to the Transmission Medium that is used to implement the Broadcast

Page 48 of 97

Channel, the Ethernet Bus.

Early implementations of Ethernet LANs employed thick coaxial cable. Actually, it was thick yellow

coaxial cable - the original recipe Ethernet cable. The cable was defined by the 10Base-5 standard. This

implementation was called Thicknet. It could deliver a BER of 10-8. It supported a data rate of 10

MBPS. The maximum LAN cable segment length was 500 meters. The segment length is the maximum

distance between data terminal equipment stations. These are attractive features.

Unfortunately, the thick coaxial cable was difficult to work with. As a result, second wave

implementations of Ethernet LANs employed thin coaxial cable. The cable was RG58 A/U coaxial cable

- sometimes called Cheapernet. This cable was defined by the 10Base-2 standard. The implementation

was called Thinnet. It supported a data rate of 10 MBPS. But, it had a BER somewhat degraded relative

to Thicknet. The LAN cable segment length was reduced to the order of 185 meters.

Thinnet ultimately gave way to the replacement of coaxial cable with Unshielded Twisted Pair cable

(UTP). This came about through an interesting merging of the Ethernet LAN architecture with another

LAN flavor called StarLAN, an AT & T idea.

StarLAN was based upon what a Telco, a phone company, normally does for businesses that is, provide

voice communications. The Transmission Medium a Telco uses within a facility for voice

communications is Unshielded Twisted Pair cable (UTP). It provides voice communications within a

facility and to the outside world by connecting all of the phones, the handsets, through a telephone closet

or wiring closet. The distance from handset to telephone closet is relatively limited, maybe 250 meters.

The StarLAN idea was to take this basic approach for voice and use it for a LAN. The LAN stations

would be connected through a closet. The existing UTP cable present in a facility for voice would be

used for the LAN data traffic. There would be no need to install a new and separate Transmission

Medium. Installation costs would be contained. Unfortunately, StarLAN only supported 1 MBPS. It

never got off the ground.

However, in 1990 aspects of StarLAN were taken and merged with the Ethernet LAN architecture. This

resulted in a new Ethernet LAN based upon UTP and defined by the 10Base-T standard. It was with this

UTP approach that Ethernet really took off in the market place.

Ethernet under the 10Base-T standard has a hub and spoke architecture. This is illustrated in Figure 4-2.

The various data equipment units, the stations, are all connected to a central point called a Multipoint

Repeater or Hub. The connections are by UTP cable. This architecture does support the Broadcast

Channel-Ethernet Bus. This occurs because all data equipment units can broadcast to all other data

equipment units through the Hub. Likewise, all data equipment units can listen to the transmissions from

all other data equipment units as they are received via the UTP cable connection to the Hub. The Hub

takes the place of the telephone closet. The Hub may be strictly passive or it may perform signal

restoration functions.

Page 49 of 97

Figure 4-2: 10Base-T hub-and-spoke architecture

The illustration Figure 4-3 indicates how the 10Base-T topology may actually look in an office set-up at

some facility. Here the data equipment units are all PCs. One serves as the file server. The illustration

shows what is usually referred to as a 10Base-T Work Group. It may serve one specific department in a

company. By connecting together these work groups Ethernet LANs may be extended. This is

accomplished by connecting hubs using LAN network elements called bridges, routers and switches. A

description of their operation is beyond the focus of the present discussion.

Figure 4-3: Ethernet operating as a 10Base-T work group

But, let us get back to 10Base-T. It supports a data rate of 10 MBPS. It has a BER comparable to

Thinnet. However, the LAN segment length is reduced even further. With 10Base-T the LAN segment

length is only 100 m - a short distance but a distance that is tolerable for many data equipment stations

in a typical business. However, it may be too short for others. This is a place where fiber optic cable can

come to the rescue.

Page 50 of 97

For the LAN market place 10Base-T was far from the last word. It led to the development of 100Base-T

- Fast Ethernet. It is also based on using UTP cable for transmission medium. However, it supports a

data rate of 100 MBPS over cable segments of 100 m.

Fast Ethernet, itself, is not the end of the road. Vendors are starting to promote Giga Bit Ethernet which

is capable of supporting 1 GBPS. However, we will stop at Fast Ethernet and the problem that both it

and 10Base-T have - the short cable segment of 100 m.

Before continuing it will be worthwhile to define two terms that come up in discussing Ethernet

characteristics. These are 1) Network Diameter and 2) Slot Time.

The Network Diameter is simply the maximum end-to-end distance between data equipment users,

stations, in an Ethernet network. It is really what has been referred to above as the cable segment. The

Network Diameter is the same for both 10Base-T and 100Base-T, 100 m.

After a BIU has begun the transmission of a packet the Slot Time is the time interval that a BIU listens

for the presence of a collision with an interfering packet. The Slot Time cannot be infinite. It is set for

both the 10Base-T and 100Base-T Ethernet architectures. It is defined for both standards as the time

duration of 512 bits. With a 10Base-T Ethernet network operating at 10 MBPS the Slot Time translates

to 51.2µsec. With a 100Base-T Ethernet network operating at 100 MBPS the Slot Time translates to

5.12µsec.

4.3 Examining the Distance Constraint

The distance constraint of an Ethernet LAN is the Network Diameter. As noted above this is 100 m for

both the 10Base-T and 100Base-T implementations. This may not be enough for all potential users of an

Ethernet LAN. Now how do you support LAN users that are separated by more than this 100-m

constraint? To deal with this question it is important to understand where this constraint comes from and

what is driving it.

Many people believe that the Network Diameter is set strictly by the attenuation properties of the UTP

copper cable connecting data equipment to the Hub. This is erroneous. Attenuation does affect the

Network Diameter, but it is not the dominant influence. However, if it were, you would be able to see

the immediate possibilities of improving it by using fiber optic cable rather than UTP copper cable. The

significantly less attenuation of fiber optic cable would boost the Network Diameter. No, it is not

attenuation but instead the Slot Time that really sets the Network Diameter.

The Slot Time is related to the amount of time delay between a transmitting BIU and the furthermost

receiving BIU. The diagram showed in Figure 4-3 illustrates the Slot Time issues to be discussed now.

Here we show two data equipment users of an Ethernet LAN - either 10Base-T or 100Base-T - it doesn't

matter. These are labeled as Data Terminal Equipment Unit A and Data Terminal Equipment Unit B.

For brevity they will be referred to as Unit A and Unit B. The BIU's are taken as subsumed in the ovals.

Page 51 of 97

Figure 4-4: 2 Stations communicating on an Ethernet Bus. Delays shown.

Suppose Unit A transmits a data packet over the Ethernet Bus to Unit B. The transmitted data packet

travels along the Ethernet Bus. It takes a time interval of TA seconds to reach Unit B. In the meantime,

Unit B has performed carrier sensing and has determined, from its perspective, that the Ethernet Bus is

not busy and so it also begins to transmit a data packet. From a collision detection point of view the

worst case occurs when Unit B begins to transmit its data packet just before the data packet from Unit A

arrives in front of it. Why is this worst case? When the Unit A data packet arrives at Unit B, Unit B

immediately knows that a collision has occurred and can begin recovery operations. However, Unit A

will not know that there has been any collision problem until the data packet from Unit B arrives in front

of it. This packet from Unit B takes a time interval of TB seconds to arrive at Unit A. Putting this

together Unit A has to wait at least TA + TB seconds before it can detect the presence/absence of a

collision. There is some additional time needed to sense the presence/absence of a collision at both Unit

A and Unit B. The collision detection processing time is denoted as TC. For 100Base-T networks a

typical value for this is 1.12 µsec. The Slot Time is the sum TA + TB + TC.

TA and TB usually can be taken as equal and denoted as t. Putting these together brings:

τ = (Slot Time- TC) / 2

The one way delay, τ, is equal to the distance between Unit A and Unit B divided by the velocity of

transmission between Units A and B. The maximum distance is of course the Network Diameter. The

velocity of transmission will be denoted by 'V.' This is the speed of an electromagnetic wave on the

Ethernet Bus. Applying these brings:

Network Diameter = (V/2)(Slot Time- TC)

The Slot Time is fixed by the 10Base-T and 100Base-T Ethernet standards. TC is a function of BIU

design. It is evident then that it is the value of V that really drives the Network Diameter. In

characterizing the Ethernet Bus you usually deal with the inverse of V. For UTP copper cable V-1 is

approximately 8 nsec/m. Consider a 100Base-T Ethernet LAN. Applying this value for V-1 above brings

a value of 250 m for the Network Diameter. On the face of it this is quite a bit better than the 100-m

allotted for the Network Diameter by the standard. The difference is accounted for by a number of delay

items that were excluded from the example. These were excluded in order to bring out the principle

point - the dependence of Network Diameter on V-1. This difference is taken up by margin allotted for

other processing functions. These functions include the delay through the Hub. They include processing

delays in software at the interface between the data equipment and its BIU. The margin is also allotted

for deleterious properties of cable.

Page 52 of 97

However, the essential point remains. The achievable Network Diameter is determined by the delay

through the transmission medium. The speed of V-1 through UTP copper cable results in a Network

Diameter of 100 m.

Consider a fiber optic cable. Typically, the value of V-1 is 5 nsec/m for multi-mode fiber optic cable.

This is almost 50% lower than for UTP copper cable. Applying this value in the above example would

bring a Network Diameter of 400 m, quite a bit more than 250 m.

By using a fiber optic cable you can connect data equipment stations to the LAN that are much further

apart than the 100 m distance allowed for by the assumed UTP copper cable in 10Base-T or 100Base-T

LANs. You can do this because the velocity of light through a fiber optic cable is much faster than the

group velocity of electromagnetic waves in copper cable- the speed of current in copper cable. You can

do this because the transmission delay, V-1, of a packet traversing a fiber optic cable is about 50% lower

than it is for UTP copper cable.

How would you do it? How would you exploit a fiber optic cable to bring distant users into a UTP

copper cable based Ethernet LAN? How would you accommodate really distant stations to a 10Base-T

or 100Base-T Ethernet LAN, stations much further than the Network Diameter?

In order to do this you need to connect them to the Hub using a fiber optic cable. This may be either a

multi-mode or single-mode fiber optic cable. However, neither the Ethernet Hub nor the BIU at the

distant data equipment user knows anything about signaling on a fiber optic cable Transmission

Medium. So, at the Hub you need some type of equipment that will take the 10Base-T or 100Base-T

packets, in their electrical format, and convert it to light to propagate down a fiber optic cable. You need

the same equipment at the distant data equipment's BIU for transmission toward the Hub. Similarly, you

need this device to be able to take the light wave representations of a packet coming out of the fiber

optic cable and convert it to an electrical format recognizable by the Hub or the BIU. This is called a

LAN Extender.

By using a LAN Extender you get a distance benefit. In addition, on the particular LAN link you get the

other benefits available with fiber optic cable. These include protection from ground loops, power surges

and lightning.

4.4 Examples of LAN Extenders Shown In Typical Applications

Telebyte offers a variety of LAN Extenders. These are now described.

Model 373 10Base-T to Multimode Fiber Optic Transceiver

This unit is pictured in Figure 4-5. It extends the distance of a 10Base-T Ethernet LAN to over 2 km.

The Model 373 10Base-T to Multi-Mode Fiber Optic Transceiver takes 10Base-T Ethernet signals and

converts them to/from optical signals that are transmitted/received from multi-mode fiber optic cable.

Page 53 of 97

Figure 4-5: Model 373 10Base-T to Multi-Mode Fiber Optic Converter

The Model 373 has a group of five LED's. These indicate the presence of the fiber optic link, traffic

going back and forth in both directions, the presence of a collision and power. The unit even includes a

Link Test switch. This assures compatibility between older and newer Ethernet adapters. It allows the

enabling/disabling of the Link Test heart beat option.

The Model 373 uses ST connectors for the fiber optic cable. It is designed for transmission/reception

over 62.5/125 multi-mode fiber optic cable. On the 10Base-T port side, it is in full compliance with the

IEEE 802.3 specification. The Model 373 is also in full compliance with the Ethernet 10Base-FL

standard. This is the standard for using multi-mode fiber optic cable to extend the Network Diameter of

a 10Base-T Ethernet LAN.

The Model 373 is illustrated in a typical application in Figure 4-6. This shows the stations of a 10Base-T

Ethernet LAN in a typical business environment. Most of the stations of the LAN are located near one

another in the same building. This is Building A. All of the stations in Building A are within 100 m of

one another. For purposes of this example, these people at these stations may all be in the company's

Accounting Department. They can all be connected to the LAN through the Hub located in Building

using the UTP copper cable - the ordinary building block of a 10Base-T LAN. They are all within the

100-m Network Diameter for a UTP copper cable based 10Base-T network. However, there is one

remote station of this LAN that is not in Building A. This may be the station of the manufacturing

manager. His office is in Building B- the production facility. Building B is located some distance away

from the front office of Building A. In fact, Building B is about 1 km away from Building A. The

manufacturing manager needs to be tied into the Accounting Department LAN so that he can update the

Controller with inventory and purchasing information.

As Figure 4-6 indicates the manufacturing manager in Building B can easily be tied into the LAN. This

is accomplished by placing a Model 373 at the Hub in Building A. A multi-mode fiber optic cable to

another Model 373 in Building B then connects the Model 373. The second Model 373 is connected to

the manufacturing manager's work station. The pair of Model 373's and the fiber optic cable will be

Page 54 of 97

completely transparent to all stations of the LAN, both the Accounting Department stations in Building

A and the remote station of the manufacturing manager in Building B.

Figure 4-6: Model 373 shown in a typical application

Model 374 10Base-T to Single Mode Fiber Optic Transceiver

This unit is pictured in Figure 4-7. It is the almost the same as the Model 373 except that its fiber optic

components are adapted for single-mode transmission. Because single mode fiber optic cable has much

lower attenuation this allows a significant extension of distance. In fact, the Model 374 10Base-T to

Single-Mode Fiber Optic Transceiver extends the distance of a 10Base-T Ethernet LAN to over 14 km.

The ability to achieve the extended distance is due to full duplex transmission. Full-Duplex* has one

important advantage. Since there are separate transmit and receive paths, DTE's can transmit and receive

at the same time. Collisions are therefore eliminated. Full Duplex Ethernet is a collision free

environment.

For single-mode fiber optic cable transmission there is no standard comparable to 10Base-FL.

*Duplex operation - Transmission on a data link in both directions. Half duplex refers to such

transmission, but in a time-shared mode- only one direction can transmit at a time. With full duplex

there can be transmission in both direction simultaneously.

Page 55 of 97

Figure 4-7: Model 374 10Base-T to Single Mode Fiber Optic Transceiver.

The application illustrated in Figure 4-6 also applies to the Model 374. However, now our

manufacturing manager can be located in a building as far as 14 km away from the Accounting

Department and still be tied into their 10Base-T Ethernet LAN.

The Model 375 100Base-T to Multimode Fiber Optic Transceiver allows any two 100Base-TX

compliant ports to be connected by multimode, 62.5/125 micron fiber optic cable, while assuring that

collision information is preserved and translated from one segment to the other. The operation of the

device is transparent to the network and is offered in two versions. The Model 375ST is equipped with

ST fiber connectors and offers 2 Km performance.

The 100 BASE-T adapters allow full duplex, simultaneous transfer of data with a minimum of

collisions. The Model 375 extends this full duplex capability using dual fibers, while offering flawless

data transmission at 100 MBPS. The Model 375 incorporates three LED's that report if the 100 Base-T

and fiber are active and powered. The fiber optic connector is a duplex as ordered, designed for

operation at 100 MBPS for FDDI, ATM or Fast Ethernet. Power for the Model 375 is via a supplied

power pack.

Table of Contents: