1 3.05

“INDUSTRIAL ELECTRONICS
& HIGH-TECH USA ’89”

USA and India: Touching tomor-
row together. That is the theme of
an exhibition of advanced technol-
ogy, electronics and computer sys-
tems. The exhibition^ after suc-
cessful week long exposition in
New Delhi moves to Bombay.
“Industrial Electronics & Hi-Tech
USA ‘89”, organised by the United
States Department of Commerce
and the American Consulate Gen-
eral- sponsor this show in Bombay
from March 8 to 11, at Bangalore
from March 16 to 18 and at
Hyderabad on March 29 and 30.
Computer hardware, peripherals,
specialised software packagers,
process control instruments, elec-
tronic industry production and test
measurements, electronic compo-
nents, CAD/CAM system, laborat-
ory and scientific instruments,
electronic medical equipment, mic-
roprocessor based - pollution con-
trol instruments, electronic desk
top publishing systems, engineer-
ing and design equipment and
other high technology products de-
signed to enhance productivity and
efficiency in office and industry are
being displayed.

Some of the high quality American
companies participating in the
trade show include AT and T, Digi-
tal Electronics, Hewlett Packard,
Tektronix, Texas Instruments,
Tata-Unisysand Citicorp Software.
Mr David Hughes, commercial
consul at the American Consulate,
Bombay, speaking to. “Elektor
'India" outlined the broad features
of the trade show.

Technology import becomes es-
sential for development In the ab-
sence of a hi-tech show like, this,
hundreds of enterpreneurs and in-
dustrialists may have to visit the
the United States on their own to
get the first hand information on
the state of the art products availa-
ble for commercial production. In-
stead of taking the people in large
numbers to tire US, we bring the
products here. This is a cost effec-
tive method ofbilateral trade prom-

“United States is the largest trading
partner of India, apart from being a
big investor. Ready market is avail-

able here from the point of view of
both the countries.”

I laving explained why this exhibi-
tion, Mr Hughes recounts the suc-
cess of a similar exhibition in India
in 1987 which reinforces the imp-
ortance of the hi-tech trade show.
“The 1987 electronics trade show
attracted over 70, 000 visitors. Two
products which were never shown
in any exhibition anywhere in the
world were brought to the Indian
show. Six other products which
had not been shown outside the US
before were also displayed in the
Indian exhibition.

“A company that never sold even a
single product to India before, now
sold more than a million dollars
worth of products, directly as a re-
sult of the 1987 exhibition.
“Interface, a Michigan-based NRI
firm is proud of the fact that the
company built itself from the
scratch through successfid sales of
productss to India”.

Mr Hughes also cites the new joint
ventures by the Digital Electronics
and Tektronix with Indian firms as
a symbol of the emerging trend.

On the query, what new products
are likely to be on display this time,
says Mr Hugest “Products come out
in the US market daily. The pro-
ducts change within a few days.
Thus, always an element of sur-
prise exists in the display of the
products. The equipment and in-
struments to be displayed in Hi-
Tech ‘89 will be worth 12 million
dollars.

Mr Hughes describes India as a
gateway to the International mar-
ket. Of the entire supplies made to
India by the US, electronic pro-
ducts alone account for 60 per
cent Indian export to the US has
also shown an upward trend, by
about 20 per cent Further, the
familiarity of languages between
India and the US makes the people
of both countries feel at home and
this is a significant advantage in the
Indo-U.S. commercial- coopera-

The US computer equipment in-
dustry has entered an era of

heightened foreign competition, inc-
reasing product standardisation, ag-
gressive pricing and shorter product
cycles. The effects of these factors
are evident in the industry’s declin-
ing trade surplus and employment
Mergers ana consolidations have
begun and are likely to increase, be-
cause none of these factors is likely
to abate, according to “U.S. Indust-
rial Outlook", an authentic publica-
tion on the U.S. commerce.

By tire early 1990s, some observers
forecast a growing movement from
silicon to gallium-arsenide based
semiconductors. This develop-
ment will help computer system
performance surpass tire highest
levels now attainable.

The U.S. software industry' must
deal with tire continuing shortage
of skilled programmers and tire es-
calating cost of producing software,
to retain its leadership in tire world
market By 1995, demand for com-
puter programmers and systems
analysts in the United States is ex-
pected to be nearly the 1984 level,
reaching 1 . 1 million professionals,
according to tire Bureau of Labour
Stastics projections.

Electronic components are the
building blocks of all commercial
and military equipment
Technological advancement in
electronic components increase
the performance and realibility of
computers, telecommunications,
consumer electronics, robotics,
aerospace and defence equipment.
The American electronic compo-
nents industry experienced a
shakeout of manufacturers in 1986
and 1987. Joint ventures, mer-
gers, and acquisitions charac-
terised conditions in many product
sectors, including semiconduc-
tors, capacitors, resistors, and
connectors.

U. S. shipments of semiconductor
devices will increase at an average
annual rate of 12 per centuh rough
1992. Future growth will depend
on continued expansion of
semiconductor applications. The
U.S. industry is expected to con-
tinue the lead in design technology.

3.10 ele

ELECTRONICS
ON THE RAILS & W HE ELS

Indian Raiways provide faster
transport than the Indian Airlines.
There is some truth in this state-
ment even if we ignore the frequent
delays in flight schedules. If prop-
ortionate weightage is given to the
railways for the number of passen-
gers transported in a train, with
much less sophisticated technol-
ogy than in an aircraft, rail travel
will turn out to be faster.

The desire for moving fast inspired
the invention of wheel. The urge for
faster movement is fraught with the
danger of derailments and colli-
sions. But, the iron wheel on rails
has been gaining momentum
steadily and safely as well.

The era of high-speed rail travel has
entered the Indian Railways. 'Fite
Shatabdi Express, flagged off in
mid-February, is the fastest train in
the country today. This train run-
ning between Delhi and Bhopal,
overtakes the Rajdhani Express in-
troduced 20 years ago between
Delhi and Calcutta. Rajdhani runs

at a maximum speed of 120 kmph
while Shatabdi runs at 140 kmph.
Speed and safety necessarily in-
volve induction of modem technol-
ogy. No modem technology can be
devoid of electronics. But, the role
of electronics is not widely felt by
the ordinary passenger, except for
the red and green signals or the
time indicators and moving adver-
tisements in the platforms. Of
course, we have closed circuit TV
in the railway stations, nowadays.
Going by the pace of modernisation
elsewhere, one may visualise a sci-
ence fiction like situation where
trains run automatically on the
press of a switch, sense obstacles
on the way, stop and proceed, en-
tirely based on the remote control
computer workstations. It is al-
most a driverless train. Even, the
fully computerised railway systems
abroad have led to collisions, and
hence, Indian Railways are very
cautious in introducing solid state
chips in operating the trains.

Running ofa train calls for the coor-
dination of a variety of services.
The locomotive, the coach forma-
tion, electric traction, mainte-
nance, line clearance, interlocking
of tracking and signals and so on.
When simultaneously dozens of
trains run on the same track, to
control their movement without
any mishap an expert controller is
indispensable. The controller can
be a man or a machine. Tire man
now increasingly uses the machine
to “ control and operate” the mov-
ing trains.

Communication is the life line of
today’s railway network, compris-
ing nine zonal railways covering dif-
ferent parts of the country. Unless
die message of an incoming train or
outgoing train Is communicated
from station to station, there will be
chaos. A controller used to draw a
chart, plotting the movement of
trains manually. This is impossible
now with the increasing number of
trains, moving in quick succes-
sion. The control panel is now

THE SHATABDI EXPRESS

3.14 Ble

modernised with the indicators, ac-
tivated by what are known as “re-
lays” Millions of switches get acti-
vated and de-activated in the pro-
cess of monitoring the train move-
ments on various routes, in-diffe-
rent sections of a railway division.

Communication channels are re-
quired not only for train and track
control but also for administrative
control. Headquarters of the nine
zonal railways and 52 divisions of
die Indian Railways are all linked
with microwave communication
network. From any station in the
country, one can speak to another
station through the decided com-
munication channels of the rail-

The railways have telephone ex-
changes of nearly 90,000 lines.
The trunk phone channel runs
3.30 lakh kilometers. The tele-
graphic and teleprinter lines work
out to 140,000 km. The VI IF mic-
rowave link has about 16,000
route' kilometers. The under-
ground cable communication ac-
counts for about 8,000 km.

Apart from the real time control of
operating trains, similar sub-con-
trols exist for electric traction,
locomotices and wagon formation.
The reliability of railway communi-
cation services is 99. 1 per cent In
communications, die railways
have kept pace with latest in intro-
ducing the optical fibre communi-
cation system. The country’s first
optical fibre railway communica-
tion system was commissioned in
February, 1989, between Chur-
chgate and Virar near Bombay on
the Western Railway. The optical
fibres cover a length of 60 km. The
next major optical fibre communi-
cation systems will be commis-
sioned over a length of 900 km bet-
ween Itarsi and Nagpur of die Cent-
ral Railway.

The railway have chalked out a
perspective plan for establishing an
“Operating Information System”,
incorporating digital data com-
munication. The plan envisages an
oudayofRs. 1,100 crores.

Data communication is an integral
part of the railway system. There
are about 9,500 steam, diesel and
electric locomotives. The coaches,
both conventional and passenger

electromotive units, number about
38,000. There are nearly 200,000
covered wagons and 100,000 open
wagons, not to speak of over
48,000 special type wagons and
about 12,000 department wagons.
These locomotives and wagons are
scattered on railway tracks which
measure 107,000 km, criss-cros-
sing the country.

On an average, daily about 3000
EMUs, 900 mail or express trains
and 3,000 ordinary and mixed
trains run on the Indian railway
network. There is litde doubt drat a
computerised data base and
dynamic exchange of information
through computer networking is
the only scientific method available
for udlising die rolling stock opti-
mally and efficiendv.

The philosophy of safety followed
by the railways aims at reducing die
human element in controlling the
train operations. Modern signalling
systems, panel interlocking, re-
moterelayinterlocking, centralised
traffic control, automatic signalling
and mechanisation of marshalling
yards are some of the steps taken
by the railways in recent years.

Still, the Indian Railways have not
gained the confidence to introduce
solid state electronics in interlock-
ing systems because a chip failure
can mean a major disaster.
Though, such a system has been
introduced on an experimental
scale at Srirangam in Tamil Nadu,
intervention of human element still
continues in die interest of ulti-
mate safety'. Two computers, in
parallel, control the traffic in any
typical modem network. What if
one computer, which compares
notes with its parallel computer
fails. The system will collapse. Ex-
perts now feel a standby third com-
puter can be provided. Why not a
fifth one, some way ask. It is still a
technology in evolution and India,
with its highly-populated rail net-
work cannot hastily jump into the
Futuristic ideas, say’s a seasoned
railway communication engineer.

Yet the railways can speak of many
innovative techniques now in use.
The most popular technique is the
one used to prevent a train from
jumping red signal. A sensor will
switch off the locomotive when the

engine tries to go past the red sig-

Track to train control is a new con-
cept being introduced. An equiva-
lent on the train transmits a signal
at a given frequency. The wheel has
a receiver. When the train runs at a
speed more than the prescribed
limit, the wheel moving faster will
miss the signals. The disturbance
in the frequency will alert the driver
that die train is exceeding its
speed. If diis warning is not
acknowledged within a few sec-
onds, the system gives a command
for activating the breakes and the
train will come to a halt automati-
cally.

An axle counter system blocks
movements of a train in die same
track where another train is already
stabled or moving. An oscillator
coil on the tracks, counts die
number of wheels which pass
through and transmits the number
of “dips” to its counterpart at the
end of the block. The line clear sig-
nal will be available only if the sec-
ond oscillating counter records the
movement of die same number of
wheels which originally entered die
block.

A last Vehicle Check Device has
been developed by the raUways in
collaboration with the Bharat Elec-
tronics Ltd. This system will avoid
accidents caused due to parting of
trains in blocks of tracks.

Similarly, train actuated warning
to level crossings are being intro-
duced. This system uses radio,
solar pow'er panels for signalling
and microprocessor based axle
counters. For detecting flaws in die
tracks and to detect invisible rail
fracture, ultrasonic rail flaw’ detec-
tors are used. To ensure tiiat die
width of die tracks does not alter,
accelerometers are used.

Finally, what affects the people
most direcdy is die reservation of
tickets. The success of the com-
puterised train reservation system,
introduced in the four major cities
of Delhi, Bombay, Madras and Cal-
cutta, has become well known.
Again, a comparison w’ith the air-
lines booking is inevitable. Railway
reservation sy'stems are too com-
plicated to be compared with the

airlines. Still, the computer sys-
tems of the railways withstand the
rigours.

The most perceptible impact of the
computerised reservation system
is that a person normally gets his
ticket in 30 minutes, under man-
ual booking system, the person
might have stood in the queue for
several hours. One can take any
ticket for any destination for
any date in any computer.
Following cancellations, waitlist
positions automatically get up-
dated, reducing the possible tam-
pering with the reservation chart
Besides being one of the largest
railway networks in the world, In-
dian railways offer services which

are nearly unparalleled. There are
seven types of classes like first air-
conditioned class, second air-con-
ditioned class, first class, second
class sleeper and so on. There 60
types of coaches. The railways offer
as many as 90 types concessions to
travellers. The rates of cancellation
vary. The fare structure are mind
boggling. The options of permuta-
tions and combinations available
before deciding on a journey ticket
is stupendously large. Nothing but
a computer can do the job in a trice.

In Bombay, the main Victoria Ter-
minus. Churchgate and Bombay
Central have been linked and tic-
kets for both the Central and West-

ern Railways are available. Soon
the computers of major cities
would be linked in a network and
booking of return tickets will be as
easy as the purchase of the first
journey ticket

While purchasing the computer
printed tickets, in die cozy aircon-
ditioned atmosphere, we also find a
sea change in the persons behind
the counter, tapping keyboard and
watching die console. These were
the same reservation clerks who
exhausted diemselves in the
drudgery of issuing tickets manu-
ally.

CATCHING THE
COMPUTER THIEF

How easy is for anyone to find out
the foreign account holding of an
individual? Going by the informa-
tion available from a series of media
disclosures of Fairfax agencies of
the U.S., it should be easy. And,
pray, how does the agency get the
requisite information? Thereby lies
die whole world of computer code
break-in and the growing insecurty
of stored data.

The villainy of hackers has been
sufficiendy highlighted and how
they squeeze information from a
computer is commonplace. But as

hackers get busy, so do die agen-
cies offering security systems.
Computer security is a big and
growing field. Several locking sys-
tems have been developed to keep
interlopers out and it is a batde of
wits between these who devise
safety and die ones who find ways
ofbreaking them. Be tiiat as it may,
just as the clever policeman may
set a thief to catch a thief, a compu-
ter specialist is often used to
counter a computer fraudster. A
detective agency trying to get the
foreign account holdings, for in-
stance, will work through a ‘plant’
or a ‘Man in Place!. This plant is

nursed for long and is used for
strategic information retrieval. And
costs for services can be pretty high
because the risks are considera-
ble. Whether it be defence infor-
mation, bank transactions, drug
trafficking, the modus operandi is
to use a plant And the security of
the stored information is sought to
be safegarded by a variety of sys-
tems of which locking derices are
only a sample. Some information is
in code so that even a retrieval is
harmless. Certain software pac-
kages are beyond the access of
people. Those entitied to run
specific programmes are frequentiy

changed so that a nexus is never
formed between the operator and
the outsider. Passwords are not
only changed from time to time but
also coded so that the birthday and
anniversary reference are removed.
Such references in password codes
are relatively easy to determine.

A disgruntled employee or a highly
ambitious and get-rich-quick staff
is a hazard for security of informa-
tion. Such people oblige with the
requisite information or, in a more
vicious role, bung a spanner into the
works. They could cause a compu-
ter virus by altering and hacking the
stored data. One of these Smart
Alecs in the U.S. when thrown out
of his job hacked the software in
such a way that each month a
database would be wiped out Re-
cards that thus began to vanish,
the company lost track of its dues
and faced liquidation. The compu-
ter programmers can take it out on
their employers if badly treated. A
disabled programme is hard to
reassemble. The hacker managed
hitherto to go unscathed but not
anymore Some of the states in the
US already regard the offence as a
felony and the judges mete out se-
vere punishment

Over 75 per cent of the security
breaches in a computer is caused
internally. People who have access
to data are in an eminent position to
trade information either for
pecuniary consideration or to av-
enge a severe treatment by a boss.
Disabling a computer or a com-
munication is considered as easy
as snapping a power line. And once
the computer support is lost the
owner-company is virtually
paralysed without computers,
plants cannot be run, bills remain
uncollected, and a planning for pro-
duction and inventory rendered
negatory. The potential for trouble
is even greater in banking industry.
One estimate places the daily
transactions in US through compu-
ter networks- at 8 one trillion, an
amount equal to 25 per cent of the
gross national product of that
country. Of this amount the fraud
is believed to account for some 8
five billion annually. The white col-
lar computer fraud is a big busi-
ness. And the one who is in a posi-

tion to commit it is often a small-
time clerk. The computer related
heist involve vast sums and it is
often too late when the mischief
comes to light

To protect the vast systems there -
fore elaborate security is intro-
duced through extra-thick concrete
walls and ceiling to house the
mechanical facility. The protective
barbed wire is said to have not just
voltage deterrent but a retina scan-
ner that locates the unwanted
guest through sensors that read the
blood vessels in eyeballs.
Protecting the hardware and
software of computers has itselfbe-
come a big business worldwide.
The growth in demand for compu-
ter security has surpassed the
larger computer hardware and
software manufacturers. The cur-
rent series of computers come with
some safety derices but the earlier
ones are still susceptible to unau-
thorised access. The assessment
for market to proriders of computer
security services is an impressive
3 554 million by 1992, 3 340 mill-
ion more than what is currently
spent The leading security
software packages mentioned in
the European press include RACF
from IBM and ACF2 and Topsecret
from Computer Associates of
Slough. Software houses such as
Cap Gemini Sogeti and Logica also
offer systems.

The computer rooms in future will
be like elaborate fortress with sec-
urity guards. And the guard will
both be physical man and mechan-
ical code. The first function of the
software is to check die legitimacy
of each attempt at access. Users
identify themselves by name or ini-
tials and are then authenticated.

The system checks through
passwords, magnetic tokens or
biometric methods which aim at
identification by physical features
including finger prints. In some of
the systems the user has to place a
hand below' a sensor for the compu-
ter to verify the authenticity of the
person planning to sit before a ter-
minal. Once in through the for-
tress, the software must ensure

diat the person’s activity is limited
to appropriate applications. This is
done by careful allocation of the
data for access.

The password itself is in future de-
signed to be based on completely
different lines. Currentiy, these are
determined on the basis of dier
pet’s name or diat of a dear de-
parted. Such associations are vul-
nerable and hacker or more seri-
ously the insider-groper could eas-
ily locate those banal passwords.
Numbered codes provide litde pro-
tection because many use their
birth dates or anniversary refer-
ences. Some of those incharge of
computer software suggest fre-
quent change of passwords and job
changes. The passwords also need
to be of a certain number of words.
One hacker’s programme can crack
a four-letter password in less than a
day while a five-letter one can take
months. Both the stored informa-
tion and the users must be graded
and assigned specific rolls. Some
could have authority to alter and
update data while others could only
scan selectively created files.
Sensitive documents can be indi-
vidually coded so that even if an un-
authorised user gains access to a
file, it remains impossible to read.
Although encryption codes can be
broken down the information could
still remain classified for a period of
time because it takes time to break
the codes. Most information is time
sensitive and die encryption there-
fore has some value and merit for
the computer operator.

Although inadequate or weak sec-
urity had severely undetermined
the business viability of several ser-
vice industry units, a computer-
using company is still not too sure
if tire safeguard systems are worth-
while. Unquestionabfy, there is
value for such security in defence
and other strategic areas but a cor-
porate entity prefers to ponder. The
re-thinking on introducing compu-
ter security is because of the high
cost of tire requisite packages. The
reluctance to introduce the
safeguards stems mainly from the
feeling that these do not add to pro-
fits. At best they may protect profits
but this perception is not too often

realised. Since there is no direct ad-
vantage in using a security pac-
kage, the decision is invariably de-
ferred.

One other fear of a computer sec-
urity safeguard is that it might get
too complex to safely operate on a
day-to-day basis. If the system is
oversecure, the users might have
to put down their passwords in
sheafs of paper. Which, if course, is
too much risk in itself. The main
area of concern is about the sec-
urity of national secrets. Theexten-
sive use of computers in the US has
jeopardised military secrecy. Who
knows if it is hacker upto a harm-
less prank or a an international spy
ring getting access to national sec-
rets. Espionage and counter-es-
pionage is done through breaks
into computers. The U.S. security
agencies receive signals from
Soviet spy satellites, decode infor-
mation and simultaneously ensure
die opposite camp does not do the
same. The Soviets, in their own
ways, set up international elec-
tronic telecommunications, a fact
evidenced by the presence of anten-
nae. Phone conversations and data
transmissions relayed by cellular
radio and microwave links are
picked up routinely. In Cuba, giant
dishes pull in signals beamed down
from satellites to any point in the
lower 48 states. Soviet ships
monitor both coasts along the US.
One estimate places the Soviet in-
terception of American calls at one
half of the latter’s aggregate traffic.

Any traffic can be intercepted and
one method of securing the infor-
mation is to safeguard lines of fib-
reoptic cables burided deep under
tiie ground. There are no connec-
tions to outside phones, so no
hacker can gain access. If a spy
cuts a pipe to tap the cable, the drop
in gas pressure would alert watch-
ers by sounding an alarm. How-
ever, buried cables are no good
when it comes to communicating
with ships and planes. To ensure
secrecy in this area, the US agen-
cies use cryptographic ciphers
which turns English into gib-
berish. It will be impossible to
make sense from a typical English
text with unfamiliar and non-exis-
tent words. The cipher is changed

frequently so that if T stood for TT
now, it might stand for ‘V tomor-
row. The decode requires a
keyboard command which in-
structs the computer to say how it
must recast a message so that it
made sense to the ordinary' mind.
As in tiie private computer, so in
the government even if it involved a
top secret agency'. The worst
enemy to a code is the person in-
side. Turncoats could make a for-
tune selling crypto details only' be-
cause tiie stakes in international
diplomacy are high. Such is the ex-
tent of risk and fear of insecurity
that tiie names used to classify in-
formation are themselves clas-
sified. The possession of crypto
code could mean access to top na-
tional secrets. As more and more is
stored in computers, as more intel-
ligence is gathered through grop-
ers, tiie need for securing tiie sys-
tems and the hardware assumes
urgency. There can never be an
ideal because tiie insider or tiie
know-all could spill the beans and
mar the effect.

The rewards for information are at-
tractive indeed. The knowing per-
sons are vulnerable before tiie
tempting offers. When tiie demand
for information is high and the sup-
ply prospect is restrict ed because of
tiie intensified security, tiie price
offered is attractive. Such offers
trigger perverted minds to work
overtime and find methods of eav-
es dropping data.

What really has made computer
vulnerable to hackers and gropers
is the network diversification. The
PCs have developed enormous
reach as more data is brought
within networks. The hackers,
oriented entirely for mischief and
fast buck, find ways of using their
PCs to break into networks. Some
are at it for pranks but an increasing
number for rewards.

In some places the employees per-
mitted limited access are given
micro-processor-based smart
cards. Each card is designed to pro-
vkie a set information package.
jlic.se cards are hard to duplicate
and safe enough to ensure that the
right person has the right access to
data in the stomach of the compu-

ter. The flaw with the card, how-
ever, is tiie risk of their being sto-
len. It is for this reason that the
biometric access devices are men-
tioned. Machines can scan voice
inflexions, hand prints and even
typing habits!

Why computer frauds are increas-
ing has intrigued people. It may
have little statistical significance if
one went by tiie cases reported
What is relevant is not tiie number
of cases but the huge sums and
stakes behind each of them. An
electronic bank heist runs into
something like tens of millions of
dollars. .Just five of these bring
about tiie closure of a large bank.
Also, the victims are less inclined
to report a white-collar fraud with
tiie use of a computer. If a bank
case were to be reported, it could be
a disaster in terms of public confi-
dence in the institutions. Manage-
ments of banks are known to have
cordially parted company with un-
scrupulous employees, giving
them a terminal gift only to ensure
the matter did not come before the
public. Elsewhere also, even the
detected cases are kept under
wraps, lest it should turn out to be a
public relations disaster.

The most insidious of tiie compu-
ter misuses is tiie spread of virus.
The whole data bank is wiped out
or altered beyond recognition. This
is done by releasing a software to
infect a genuine package. Over
period of time tiie ‘intruder’ pac-
kage undermines tiie original pac-
kage. Some quarter million out-
breaks involving 40 large American
corporations have been reportedly
tak nill with the virus. Some of the
viruses make a passage over wide
distances through tiie globe. Some
versions of a virus are created by
mischief mongers -along tiie way
and they have taken their virus to
such countries as Israel, Europe
and the U.S. That passage itself, it
is suspected, was through tiie
networks.

Among the steps to check tiie
spread of virus are not only the ap-
propriate vaccines but also tighter
law's and also tax incentives for in-
vestment in sophisticated
safeguard programmes. The

American companies have been
spending large sums on the com-
puter chasticity belts with some
software copies selling for as much
as g 35,000. These are also agen-
cies which keep backup in case a
set is lost through fire, flood or any
other ‘Act of God’. These are often
stored in remote places with armed
personnel guarding the tapes.

The ingenuity of the hackers and
gropers and other mischief mon-
gers keeps increasing as the protec-
tors develop defences. It is battle
unlikely to stop. The international
dependence on computers for data
storage is expected to force people
in other countries too to face up to
bad elements in the trade. Since
India has also embarked on a ph-
ased computerised programming,
die devices would now have to be
prepared to prevent the disasters
which western countries faced.
The computers might produce an
environment exacdy the opposite
of die one it set about achieving.
The purpose of die computer was
to guarantee secrecy. That was also
die reason for the popularity of the
gadget with the blackmoney hol-

ders in the country'. With hackers
around, there is a threat to the sec-
recy. Too much of sophisticated
technology is becoming too com-
monplace. There might well be an
open society with the computers
and hackers. Secrets, if any, will
become open. Will dial make a
computer redundant? There is a
growing body of opinion which
feels that computers cannot eter-
nally be relied upon as capability
make it seem there is a rebirth of
Man Friday.

From the dine Alibaba discovered
the ‘Open Sesame’ code, secrets
are hard to keep. There indeed are at-
tempts at keeping secrets under die
hat but diey have die uncanny
habit of getdng away thanks to the
hackers.

Be that as it may, the controversial
retendon of Hersliman as a pro-
vider of vital information on illegal
account holding of Indian politi-
cians appears to strengdien die
view diat secrets arc sought even in
this country. The only access to
such information in developed
countries, and specifically in the
closely-guarded banking systems

of Switzerland is through plants in
the financial institutions. It is be
ieved to be easy to cultivate plants
and secure from them vital tax eva-
sion information. It is not uncom-
mon for die American private in-
vestigation agencies to look for
those opportunities. And the way
to do it is of course to follow the
many methods described hi the
course of this article.

Every account in every bank of
Europe is in the computer. If a way
could be found to retrieve it from
there, die information is common
knowledge. The only ones capable
of getting it out is any employee en-
joying die trust of the employer.
Since the rewards are large, die in-
centive is good.

As die retrieval of data dirough
gropers become easy, die user be-
gins to diink more seriously of in-
stalling security systems. There
are many user worldwide who are
currendy evaluating the various
security packages and wondering
which, if any, oftiiose should be re-
quisitioned. This is giving die sec-
urity package writers a good pros-
pect of business in the times to
come.

95% of TV reception problems are caused by faulty antennas.

_Dont take chances with cheap fakes.

INSTALL A Y AGIN A
TV ANTENNA
TODAY

gpp!

g

Kill

gSSj

Ig

YAGINA Antennas are de-
signed and manufactured
by experienced electronic
engineers to ensure the re-
quired quality, performance
reliability & weather resis
tance.

ft

Manufactured by:

YABABU
GRAPHICS
PVT. LTD.

103, Shanti Industrial Estate, Sarojini Nc
Road, Mulund (W) Bombay-400 080.
Ph. 561 6519

DEALERS WANTED

1 3. 19

MOSFET HI-FI POWER
AMPLIFIER

A quality 160-watt hi-fi output amplifier based on the Siemens
BUZ series MOSFETs.

Until not so long ago, the BUZ series of
MOSFETs from Siemens were hard to
come by and very expensive. That was a
pity, because these devices offer a very
good specification. Fortunately, the situ-
ation has improved considerably,
although the transistors are still only
available as n-p-n types. However, n-p-n
types can be used just as well as com-
plementary pairs as the present circuit
proves.

A power amplifier, whether it uses
bipolar devices or MOSFETs, needs a
drive circuit. When MOSFETs are used,
that circuit can be kept pretty straight-
forward. This means that any modifi-
cations in respect of power handling,
bandwidth and distortion may be
brought in fairly easily.

The device chosen for the present circuit
is recommended by Siemens for use as a
power opamps in control engineering,
which indicates that it is a very stable
component. None the less, to prevent
any mishaps, the amplifier is provided
with protection circuits against short-
circuits and overheating.

The circuit

The circuit diagram in Fig. 2 shows the
highest-power version of the amplifier:
this delivers 160 W into 4 ohms.
Modifications to reduce the power ouput
will be discussed later in the article.
The circuit is based on the two series-
connected MOSFETs, Tis and Tie, being
driven in anti-phase by a differential
amplifier. Since the input resistance of
MOSFETs is of the order of 10 9 ohms,
the drive power needs to be only very
small. The MOSFETs are thus voltage-
driven.

The drive circuit consists essentially of
T1-T2 and T12-T1.1. Negative d.c. feedback
from the output amplifier is provided by
R22 and negative a.c. feedback by Rzj-
Cj. The a.c. voltage gain is about 30 dB.
The lower cut-off frequency depends on
the values of Ci and Cj.

The operating point of the first differen-
tial amplifier, T1-T2, is set by the current
flowing through T3. The collector cur-
rent of T5 determines the reference cur-
rent for current mirror T3-T4. To ensure

that the reference current is stable, the
base voltage of Ts is stabilized by diodes
D4-D5.

The output of T1-T2 drives a second dif-
ferential amplifier, T12-T1J, whose collec-
tor currents generate the gate potential
for the output transistors. The level of
that potential is determined by the
operating point of T12-T13. Current mir-
ror T9-T10 and diodes D2-D3 have the
same function as Tj-Ta and Dj-Ds in the
first differential amplifier. The
magnitude of the reference current
depends on the collector current of Tin,
which in turn is set by P2 in the emitter
circuit of Tn. This arrangement sets the
quiescent (bias) current in the absence of
an input signal.

Stabilization of quiescent
current

The MOSFETs have a positive tempera-
ture coefficient when their drain current
is small, so that the quiescent (bias) cur-
rent is only kept stable by appropriate
compensation. This is provided by Rn

D.C. operating voltage

IPout = maxi 2 ±46 V

IPoui = 01 s ±55 V

Current drawn

(Poui = maxi 3 A

(Pout -01 £0.2 A

(Output short-circuited) < 1 . 5 A

Max. power output

(f - 1 kHz: Ri = 4 ohmsl 160W

Music power output

(Ri - 4 ohms) S240W

Distortion (20 Hz-20 kHz) S0.05%

Intermodulation

(250 Hz: 8 kHz: 4:1) s0.07%

Input resistance s 33 k

Voltage amplification 31 dB

Frequency response ( — 3 dB) a 2 Hz- s 250 kHz
(Rl = 4 ohms; Pout = 15W)

Power bandwidth

(THD = 0.5%; Poui = 80 Wl s5 Hz- 2 70 kHz
Damping factor

(Rl = 4 ohms; f = 40 Hz) £200

Signal-to-noise ratio (unweighted)

(Pout = 50 mWI £73 dB

(Poui = max) >108dB

Output impedance 4 Q

Table 1. Technical specification of the
MOSFET amplifier.

Fig. 2. Circuit diagram of the 160-watt version of the MOSFET power amplifier. Changes for
lower-power versions arc given in Table 2.

across current mirror Tv-T.o, which has Overheating protection Short-circuit protection

a negative temperature coefficient.

When this resistor heats up, it draws a The MOSFETs arc protected against If the output is short-circuited in the

slightly larger portion of the reference overheating by thermistor R12 in the presence of an input signal, the re-
current through T«. This causes a re- base circuit of Ts. When a certain tern- duction in voltage across resistors R33

duction in the collector current of Tu> perature is reached, the potential across and Rsj causes T14 to be switched on.

and this, in turn, causes a decrease in the the thermistor causes T- to switch on. This results in a decrease of the current

gate-source voltage of the MOSFETs, When that happens, Ts draws the larger through T9-T10 and, consequently, of the

which effectively compensates the in- part of the reference current through collector currents of T12 and To. The

crease caused by the PTC of the T*Tn, which effectively limits the out- dynamic range of the MOSFETs is then

MOSFETs. The thermal time constant, put power of the MOSFETs. severely restricted, so that the power dis-

which is dependent on the thermal resist- The temperature threshold is set by Pi sipation is kept low.

ance of the heat sinks, determines the and is equivalent to a heat sink tempera- Since the permissible drain current is de-
time it takes for stabilization to be ef- ture of < 72.5 °C. This assumes a ther- pendent on the drain-source voltage,

fected. The quiescent (bias) current set mal resistance of 0.5 K W and an am- more information is needed for the cor-

by P2 is stable within ±30%. bient temperature of 25 °C. rect setting of the current limiting. This

elektor India march 1989 3.29

information is provided by the voltage
drop across resistors R» and R:-
(positivc and negative output signals re-
spectively). If the load is >4 ohms, the
base-emitter voltage of Tw is reduced to
a value that results in the short-circuit
current being limited to 3.3 A.
Construction

The amplifier is best build on the PCB
shown in Fig. 3. However, before con-
struction is started, it has to be decided
which version is wanted. Fig. 2 and the
parts list of Fig. 3 are for the 160-watt
version. Changes for the 60 W, 80 W,
and 120 W versions are shown in
Table 2.

As shown in Fig. 4, the MOSFETs and
NTCs are mounted on a right-angled.
The pin connections are shown in Fig. 5.
The NTCs are screwed direct into M3-
size, tapped (tapping drill = 2.5 mm),
holes: use plenty of heat conducting
paste.

Resistor R:> and R?i are soldered direct
to the gates of the MOSFETs at the track
side of the board.

Inductor Li is wound on R»: its well-
insulated, pre-tinned terminals are
soldered to the holes adjacent to those
for R«..

Capacitor Ci may be an electrolytic
type, but an MKT type is preferable.
The faces of Ti and T: should be glued

together to ensure that their body tem-
perature remains equal.

Do not forget the wire bridges.

The power supply for the 160-watt ver-
sion is shown in Fig. 6: changes for the
other versions are shown in Table 2. An
artsist’s impression of its construction is
shown in Fig. 7.

Once the power unit has been built, the
open-circuit operating voltages may be
measured. The d.c. voltages should be
not greater than ± 55 V, otherwise there
is a danger that the MOSFETs will give
up the ghost on first power-on. If
suitable loads are available, it is, of
course, preferable that the supply is
tested under load conditions.

When the power supply is found OK, the
aluminium MOSFET assembly is
screwed on to a suitable heat sink. Fig. 8
gives an impression of the size of the
heat sinks and of the complete assembly
of a stereo version of the amplifier. For
clarity, only the position of the compo-
nents of the power supply is shown.
The areas where the heat sink and the
aluminium MOSFET assembly (and,
possibly, the rear panel of the amplifier
enclosure) meet should be given a good
coating of heat conducting paste. Each
of the two assemblies should be screwed
to the associated heat sink with at least
six M4 (4 mm) size screws.

The wiring should follow the guide lines
in Fig. 8 faithfully. It is best to start with
the supply lines (heavy gauge wire).
Next, make the earth connections (star-
shaped) from the power unit earth to the
PCBs and the output earth. Sub-
sequently, make the connections be-
tween PCBs and loudspeaker terminals
and those between the input sockets and
the PCBs. The input earth needs to be
connected only to the earth terminal on
the PCB - no more!

Calibration and testing

Instead of fuses Fi and F2, connect 10-
ohm, 0.25 W, resistors in their position
on the PCB. Preset P2 must be set fully
anticlockwise, while Pi is set to the
centre of its travel. The loudspeaker ter-
minals remain open, and the input must
be short-circuited.

Switch on the mains. If there are any
short-circuits in the amplifier, the 10-
ohm resistors will go up in smoke! If that
happens, switch off immediately, find
the fault, replace the resistors, and
switch on again.

When all appears correct, connect a
voltmeter (3 V or 6 V d.c. range) across
one of the 10-ohm resistors. There
should be no voltage across it. If there is,

P: is not turned fully anticlockwise. The
voltage should rise when P: is slowly
turned clockwise. Set P: for a voltage of
2 V: the current is then 200 mA, i.e.:

100 m A per MOSFET.

Switch off, and replace the 10-ohm re-
sistor by the fuses. Switch on again, and
measure the voltage between earth and
amplifier output: this should be not
greater than ± 20 mV. The amplifier is
then ready for operation.

A final point. As already stated, the
switching threshold of the overheating
protection circuit must be set for about
72.5 °C. This can be ascertained by
heating the heat sink with, e.g., a hair
dryer and measuring its temperature.

However, this is not strictly necessary:

Pi may be left set at the centre of its
travel. Its position should only be ad-
justed if the amplifier switches off too
often. None the less, its position should Fig. 7. Artist's impression of the assembly of a stereo version of the MOSFET power ampli-

never be far from the mid position. H fjer. It also gives an idea of the size of the heat sinks.

3.32 elektor india march 1989

Table 2. Changes and variations for lower-power versions of the MOSFET amplifier.

CAR SERVICE MODULE

A compact unit
that measures
speed of a petrol
engine in
revolutions per
minute, and the
dwell angle of the
ignition.

by. A. Rigby

The car service module is composed of
two units: a circuit for measuring dwell
angle and engine speed on one printed
circuit board, and an associated liquid
crystal display (LCD) read-out on
another board. The units are connected
by a cable terminated in 9-pin D-type
connectors. The compact LCD readout
is purposely kept separate to enable it to
be used in other applications also.

Electronics and the petrol
engine

Engine speed and the ignition dwell
angle are both physical quantities which
are to be converted to a voltage that can
be shown on a display. Figure la shows
the basic elements of an ignition system
in a petrol engine. The primary of the ig-
nition coil is connected between the
positive pole of the car battery and the
contact breaker, which is shunted by a
capacitor and indirectly operated by the
camshaft. When the camshaft revolves,
the contact breaker opens periodically.
A magnetic field is built up in the ig-
nition coil when the contact breaker is
closed. When the contact opens very
briefly as it is pushed open by the ro-
tation of the camshaft, the magnetic
field causes an electrical pulse because
of resonance of the tuned circuit formed
by the ignition coil and the capacitor.
The alternating voltage is boosted to
15,000 to 30,000 volts by the high-
impedance secondary winding of the
coil. The high voltage is then directed,
via the distributor, to one of the 4 spark
plugs (it is assumed here that the service
module is used for 4-cylinder cars). Ob-
viously, the spark rate depends on the
speed at which the engine runs.

The dwell angle is the angular displace-
ment of the contact breaker shaft that
determines how long the contact breaker
remains closed. A correctly adjusted
dwell angle is essential for two reasons:
first, for correct timing of the sparks in
the cylinders, and, with it, the highest
possible engine efficiency; and second,
for enabling the ignition coil to build up

enough energy for the spark-over
voltage.

The timing diagrams in Fig. lb show
how electrical pulses are obtained from
the contact breaker. The top diagram
shows the voltage typically developed
across the contact breaker. This voltage
is clipped and shaped to obtain digital
compatible 5 V pulses that can be pro-
cessed by the service module. The first
negative pulse edge triggers a
monostable multivibrator (MMV),
which pulls its output low for a fixed
period, Tmmv. The output of the MMV
thus supplies a rectangular signal of
which the ‘low’ time, Tl (=Tmmv), is
constant in each period, while the ‘high’
time, Th, is a function of engine speed:
the trigger frequency rises with engine
speed, while Th becomes shorter. The
average voltage, U««, available at the.
output of the MMV is approximated
with the equation
U»v«UbTH/(Tn+TL)

Since the period of the contact breaker,
To, is simply Th+Tl, it follows that

To=l/fo

where fo is the contact breaker fre-
quency, which is a function of engine
speed. From the above, it can be deduced
that U.v is a function of engine speed:

Uov=Ub(To— Tmmv)/To —
Ub[l-(T.MMv/To)] =

Ub(l - foTlUMV)

To understand how the dwell angle, ®, is
measured, it must first be defined as

<t>=T(Tcl/To)(360/n)

where n is the number of cylinders.

A NAND gate is used to combine the
shaped, digital signal (second drawing in
Fig. lb) with the MMV signal (third
drawing). The result is the signal drawn
in the last diagram in Fig. lb. The com-
bining is necessary to get rid of the noise
at the start of each period of the input
signal. The average value of the voltage

at the output of the NAND gate is writ-

Uav = Ub[l - (Ti./To)]

Since, in a four-cyilinder, four-stroke,
engine,

<J> = 90T(Tl/To)

it is evident that Uav is directly pro-
portional to <t>, so that it can be used to
measure the dwell angle.

Circuit description

Figure 2 shows that the circuit of the
meter section of the service module is
fairly simple, and essentially based on
only one integrated circuit, the CMOS
TVpe 4011. The 5 V regulator, IC2, is fed
from the 9 V battery in the display cir-
cuit described below. A zener diode, Di,
and a series resistor, Ri, reduce the am-
plitude of the contact breaker signal to a
value suitable for applying to a CMOS
NAND gate, Ni. Capacitor Ci in the in-
put network shunts any high-frequency
components to ground.

Gate Ni functions as a pulse shaper as
already discussed with reference to
Fig. lb. Parts R2 and C2 form a dif-
ferential network that supplies a very
brief, active low, needle pulse with every
negative pulse transition from Ni. The
monostable multivibrator set up around
N2 and N3 is triggered on the first of
these needle pulses as shown in the
timing diagrams in Fig. 3. In the non-
triggered state, the MMV output (N>
pin 10) as well as the input (N2 pin 5)

Fig. 1. Basic ignition system in a petrol-
engine (la) and timing diagrams of the car
service module (lb).

are logic high. Since the output is con-
nected to pin 6 of N2, pin 4 of this gate
is logic low. This condition is stable with
no voltage across C3. Following a
negative-going needle pulse at the input
of the MMV, the output of N2 toggles
from low to high. The resulting charge
current through C3, shown in Fig. 3c,
causes a quickly rising and a slowly,
logarithmically decreasing, voltage drop
across R> and Pi. Consequently, Ns
toggles: its output, and with it the sec-
ond input (pin 6) of N2, goes low, so
that a stable situation is obtained for as
long as the voltage across R3 and Pi
does not exceed the toggle threshold of
Nj.

When, at a voltage level of about '/iUb,
the input voltage of N3 falls below the
toggle threshold, the gate supplies a high
level again. The monotime Tmmv is over,
and both inputs of N2 are logic high
again. In other words, the stable stand-
by state is restored until the next trigger
pulse occurs.

VMOSFET Ti blocks during the
monotime. As soon as this ceases, how-
ever, the transistor conducts and effec-
tively shunts Pi and R> with a relatively
low resistance, Rj. This causes Ci to be
discharged much faster, so that the
monotime of the MMV remains con-
stant even with relatively high trigger fre-
quencies (= engine speeds). A
VMOSFET Type BS170 is used here
because its - high input impedance en-
sures that N3 is not overloaded.
Moreover, the transistor has a very low
drain-source saturation voltage, so that
it does not affect the operation of in-

Fig. 2. Circuit diagram of the rev counter/dwell
3.34 .Murindl.rn.nh 1989

circuit in the

service module.

tegralor Rx-C-i. This network serves to
convert the rectangular signal at the out-
put of N3 into a direct voltage that is
directly proportional to the average
voltage of the rectangular signal, and,
therefore, to the engine speed. The ca-
pacitor, Gt, is connected to the positive
supply line because Uav is obtained from
the active-low output of the MMV, so
that it is actually an inverse function of
engine speed (refer back to Fig. lb).
With C4 connected to the positive
supply rail, this inversion is inverted
again, since the voltage on the capacitor
increases when U»v decreases.

Dwell angle measurement uses in-
tegrating network Rs-Cs at the output
of Nj. As shown in Fig. lb, this NAND
gate combines the cleaned input signal
with the MMV. signal, so that the voltage
on Cs is directly proportional to the
dwell angle.

Finally, potential dividers R1.-P2-R7 (rev
counter) and R'i-P.i (dwell meter) pro-
vide the drive voltages for the LCD
readout. The presets are used for
calibrating the two functions of the
module. Toggle switch Sm selects be-
tween the revolution counter and the
dwell angle meter functions, while Sis
selects the correct position of the
decimal point on the display (DP 2 for
the rev counter, and DPI for the dwell
meter).

Fig. 4. Circuit diagram of the LCD readout.

1 3.35

A universal LC display
module

The circuit diagram of Fig. 4 shows that
the 3 '/2-digit liquid-crystal display with
the car service module is a standard ap-
plication of the ICL7126 from Intersil
(the ICL7126 is a CMOS version of the
familiar ICL7106 which may also be
used here). Transistor Ti is added to ac-
tuate the LO BAT indicator on the dis-
play when the 9 V battery is exhausted
(Ub<7.2 V). The auto-zero function of
the ICL7126 obviates any null ad-
justments. The display unit is calibrated
by interconnnecting its LO and COM in-
puts, applying a variable voltage be-
tween 0 and 200 mV to LO (-) and HI
( + ), and adjusting preset Pi until the
read-out is in accordance with the actual
value of the applied voltage, which is
measured simultaneously with a digital
voltmeter.

Construction and alignment

Building the two circuits that together
form the car service module on the
PCBs shown in Figs. 5 (meter section)
and 6 (LC display) is straightforward.
Angled 9-pin D-connectors for PCB
mounting are used for interconnecting
the circuits by means of a length of 9-

Completed meter circuit on PCB 886126.

way cable. The size of the PCBs is such
that the units can be housed in identical,
transparent, enclosures, from which the
9-pin connectors protrude. The input to
the meter circuit is made by 2 wander
sockets, a red and a black one, which ac-
cept plugs fitted on heavy-duty, heat-
resistant test wires with insulated croc-
clips at the other end for connecting to
the contact breaker on the car engine.
For the following description of the
alignment of the service meter, it is as-

sumed that the digital read-out has been
calibrated as detailed above.

First, build the 50 Hz source shown in
Fig. 7. The alternating voltage it sup-
plies simulates the contact breaker
pulses, and is applied to the input of the
car service module. Since, in a four-
cylinder, four-stroke, engine, ignition in
a cylinder takes place every fourth revol-
ution of the crankshaft, 50 contact
breaker pulses per second simulate
50x60=3000 sparks per minute, or 750

Fig. 5. Printed-circuit board for the meter circuit of the

module.

LC DISPLAY UNIT. CIRCUIT DIAGRAM: FIG.

Rasistors I ± 5%l:
Ri = 47R
R2;R3;Rt2 = 1M0
R4 = 220K
R6;Ra;R7=470K
Rs;R9=180K
Rio=680R
Rn =390K

■ BC547B

-purpose 354-dlgit LC display

ay unit. dimensioned to give a stable readout, at

the cost of a fairly slow meter response
and adjust Pi for a reading of 0.19 V. to engine speed variations. Also, since
This sets the monotime with sufficient the input signal is combined with the
accuracy. MMV signal, dwell angle measurements

Make sure that the function switch. Si, can only be made at engine speeds lower
is set to rev counter, and adjust P2 until than 3000 rpm.
the LCD readout indicates 1.5, which

•corresponds to 1500 rpm (60 Hz: The meter is also suitable for six-cylinder
1800 rpm). Now switch on the dwell engines. Since these generally run at a
meter function and adjust P3 for a.n, lower speed than 4-cylinder types, no
LCD reading of 45.0°. In practical use of changes are, in principle, required to the
the instrument, it should be borne in previously detailed adjustment of Pi.
mind that integrator R8-C4 is purposely The signal supplied by the test circuit of
Fig. 7 then simply corresponds to
I 1000 rpm (60 Hz: 1200 rpm) and a dwell
angle of 30°. *<

per minute per cylinder. Since one spark
is genereated per two revolutions, the
simulated engine speed is 1500 rpm.
With a 60 Hz mains, this becomes
1800 rpm.

In most cases, the maximum engine
speed will be about 6000 rpm, corre-
sponding to a contact breaker frequency
of 200 Hz. This means that the
monotime of MMV in the circuit should
be set to about 0.8(l/200)=4 ms. Con-
nect a high-impcdance DVM across Cj,

imple signal source for adjusting
ice module.

This compact, ICL7126-based LC display

be used in many applications.

1 3.37

MORE APPLICATIONS
FOR THE 555

There are probably few integrated circuits that have been with us
for as long as timer Type 555. This article does not add to the
seemingly endless list of AMV and MMV applications of this chip,
but discusses some less familar designs derived from these. In
addition, a brief introduction is given to the new CMOS and
LinCMOS versions of the 555.

One explanation of the popularity of the
now 17-year-old timer type 555 may be
that the chip is inexpensive, and contains
a fairly unique combination of sub-
circuits. Looking at the internal struc-
ture shown in Fig. I, these are a bistable
(a digital circuit), two comparators
(analogue circuits) and some discrete
parts, a resistive potential divider and a
transistor. Added to the versatility of
these interesting building blocks come
the abilities of the chip to supply a rela-
tively high output current, and to work
from a wide range of supply voltages.
Pin assignments of the 555 and the dual
version of it, the 556, are given in Fig 2.
Every electronic engineer or student is
bound, at some time, to deal with the
555 in its standard configuration as a
monostable or astable multivibrator.
These applications are so numerous by
now that it is often forgotten, or not
even known, that the 555 can be used in
a number of other, less well known, con-
figurations. To understand how these
work, however, it is useful to first look at
the basic operation of the chip.

Some fundamentals

Judging from the internal diagram of the
555 (Fig. 3), the relatively high number
of components is typical of chip tech-
nology of the early 1970s. Fortunately,
the internal diagram is still fairly simple
to analyse. A trigger comparator (block

Fig. 1. Basic internal structure of the 555.

by T. Wigmore

A) and a threshold comparator, block B,
are clearly recognized as difference
amplifiers. The bistable, block C, is,
perhaps, less conspicuous. In rest, tran-
sistors Qis and Qr are off, while Qis
and Q20 conduct. When the trigger
voltage drops below one-third of the
supply voltage, Q10, Q11 and, therefore,
Qis, also start to conduct. Transistor Qis
removes the base drive of Qis and so
causes this to block. By virtue of Rio
and diode Qis, Qn starts to conduct. As
the trigger voltage rises again, Qis is al-
lowed to turn off again without causing
instability of the new state — Qis is

555-Timer

556-dual Timer

Fig. 2. Pinning of the 555 and the 556

Fig. 3. Detailed internal circuit diagram of the 555.

ICM7555

A - trigger comparator 890001 - 14

C = bistable
D = output amplifier

Fig. 4. Internal structure of a CMOS version

a

Fig. 5. A standard 555 briefly draws a high
current when its output toggles (Fig. 5a;
lower trace shows inverted supply voltage;
decoupling of the supply voltage is a must!).
The new CMOS 555 does not produce this
annoying effect (Fig. 5b)

of the 555, the ICM7555 from Intersil.

then inhibited from conducting via Rn.
The normal procedure is that the
threshold voltage exceeds two-thirds of
the supply voltage. This results in Qi
and Q: starting to conduct. The in-
crease in their collector currents is
amplified by Q= and Q6, so that Qk,
starts to conduct again. This transistor,
in turn, causes Qn to block, but only if
Qis is actually off. If this is not so — in
other words, if the threshold input and
the trigger input are both actuated —
the bistable remains reset. Because the
collector current of Q6 is limited by R’,
Qis pulls the base of Qi<> harder to
ground than Qs can pull it to the
positive supply rail.

An all-overriding method to reset the
bistable is to drive its reset input low.
This results in Q25 conducting, so that
the base drive of Qn is removed. Since
diode Qis creates additional voltage
drop during resetting, the base voltage
of Q17 is sufficiently low to actually
turn this transistor off. When the
bistable is in the reset state, output tran-
sistors Qm and Q24 and, via Rib, dis-
charge transistor Qi«, conduct.

The 555 briefly draws a fairly high cur-
rent when its output changes from low to
high. This is so because Q24 is briefly
driven into saturation, and. takes a while
to actually turn off. As soon as Q21 and
Q22 conduct, a short, non-current
limited, short-circuit of the supply
arises. It is for this reason that the 555
requires particular attention to be paid
to decoupling of the supply voltage (see
Fig. 5a). Output switching from high to
low causes fewer problems because Q21
and Q22 are not driven into saturation;
hence, the switch-off time is short rela-
tive to that of Q24. CMOS versions of
the 555 generally do suffer from this an-
noying effect.

Fig. 6. Standard application of the 555 in
MMV configuration.

Applications

In 9 out of 10 applications of the 555,
the chip is used as a monostable or
astable multivibrator (AMV or MMV re-
spectively). In MMV configuration, the
pulse time is determined by the time
needed to charge the timing capacitor
from 0 V to VSUb, the threshold voltage.
In general, the charge voltage, Uc, on a
capacitor, C, charging through a resistor
R, from a supply voltage, Ub, is equal to
V5Ub when

Uc(t) = Us(l - e' ,/RC )
from which,

T=(-l0ge'/3)RC=l.lRC

The charge voltage also determines the
monotime, provided the trigger pulse is
shorter than the monotime. A longer
trigger pulse also results in a longer out-
put pulse, but this may be prevented by
driving the trigger input with an AC-
coupled signal only (add R2/C3, with
(R 2 C3)<(RlCl)).

The MMV circuit is turned into an AMV
simply by making it self-triggering. Ca-
pacitor Ci, via Ri and R2, is charged to
VSUb in time interval ti: —

n = ( - logc'/3)(R I + R2)C — ( - loge VS)(R 1 + R2 )C
=0.694(Ri + R2 )

and is then discharged again, this time
only via R2. The discharge time, h,
equals

t2=0.694R2C

This means that the voltage on the ca-
pacitor toggles between VSUb and 2 AU\>.
The total period, T, is calculated as

T=ti+t2=0.694(Ri+2R2)C

and the frequency, fo, as

fo=l/T=1.44/(Ri+2R2)C

1 3.41

It should be remembered, however, that
Ci has to be charged from 0 V when
power is first applied, or when the reset
input is made high. The first part of the
first output period, therefore, has a
period of I.IR1C1.

Fig. 7. Standard application of the 555 in
AMV configuration.

One of the nice features of the 555 as an
MMV or AMV is that the pulse time is,
in principle, independent of the supply
voltage, Ub. When this drops, the trig-
ger and threshold voltages, as well as the
charge- and discharge currents, drop ac-
cordingly, resulting in no change overall.
A disadvantage of the AMV circuit
shown in Fig. 7 is its inability to supply
an output signal of duty factor greater
than 0.5: this is because the charge resist-
ance, R1 + R2, is always greater than the
discharge resistance, R2 by itself. The
basic circuit in Fig. 8 shows how this can
be resolved with the aid of a diode, Di.
During charging, it bypasses R2, so that
the charge current can become smaller
than the discharge current. Another di-
ode, D2, is optional if Ri alone is to
determine the charge current. It should
be noted that the above use of diodes
sacrifices, at least partly, the 555’s in-
dependence of the supply voltage level
— when the supply voltage is changed,

Fig. 9. Frequency deviation of a 555 in
AMV configuration is a function of a num-
ber of parameters, including the duty factor.
The effect shown by these oscillograms is
mainly on account of the recovery time of the
trigger comparator and discharge transistor.
Upper trace: output signal; lower trace:
voltage on timing capacitor. The horizontal
traces show the trigger and comparator
threshold levels.

the fixed drop across the diode results in
a non-proportional change of the charge
and discharge current of Ci.

The control voltage input, pin 5, of the
bipolar 555, is normally decoupled to
ground with a 10 nF capacitor for noise
protection. According to the manufac-
turers, this capacitor is no longer re-
quired with the new CMOS versions of
the 555.

a duty factor of about 0.6. The fre-
quency, 29 kHz, already deviates con-
siderably from the calculated 25 kHz.
Fig. 9b shows the output signal of the
same circuit, this time dimensioned for a
much greater duty factor. Since the total
resistance R1 + 2R2 is equal in both
cases, it might be expected that the out-
put frequency remains unchanged. It is
seen, however, that Ci is actually
discharged to below the trigger level
(which, like the threshold level, is
marked by a horizontal trace). This ef-
fect is caused partly by the relatively
quickly falling voltage on Ci, and partly
by the slowness of the trigger compara-
tor in combination with the recovery
time of the discharge transistor. Because
of the excess discharge of Ci, the output
frequency of the 555 will be significantly
lower than calculated: 20 kHz in this

The essence of all this is that the accu-
racy of relatively high output frequencies
depends largely on the duty factor.

When the 555 is configured as an MMV,
due account should be taken of the
saturation voltage of the internal dis-
charge transistor. The level of this
saturation voltage is inversely related to
the value of the charge resistor, and, at
relatively short monotimes, causes the
output pulse to be shorter than
calculated.

At very low output frequencies, factors
such as the leakage current of the timing
capacitor, that of the discharge transis-
tor, and the input current of the
threshold comparator, become increas-
ingly significant.

In general, the lower the frequency, the
higher the values of the charge and dis-
charge resistors. As the charge current
decreases, the importance of various
leakage currents increases. Also
remember that the use of an electrolytic
capacitor with high leakage and toler-
ance in position Ci will cause a much
higher timing error.

Fig. 8. Non-standard AMV configuration
that allows duty factors lower than 0.5 to be
achieved.

3.42 elektorindis march 1989

Timing errors

It is not so simple to express the inac-
curacy of a timing interval produced by
a 555 as a single error-percentage. A
large number of factors should be taken
into account here, but many can be
forestalled by correct dimensioning
and/or selection of the most appropriate
type of 555 for a particular application.
Tolerance on the internally generated
reference voltages, in combination with
input-offset voltages of the trigger- and
threshold comparators, introduces
timing errors of the order of 2%.
Internal reaction and recovery times also
form a factor to be taken into account.
The oscilloscope photographs in Fig. 9
illustrate the behaviour of a 555-based
AMV at a relatively high output fre-
quency. Figure 9a shows the AMV set to

Using the control input

The control voltage input, pin 5, affords
a number of interesting, yet little used,

Fig. 12. Two 555's, or. in this case, a single 556, make an excellent fixed-frequency pulse-
width modulator for low-loss power control systems.

applications, whose background is dis-
cussed below.

The internal diagram shows that pin 5 is
connected to the internal voltage divider.
When not externally loaded, this carries
a voltage of 2 /3Ut>. According to the
manufacturers, this voltage may be
varied between 45% and 90% of the
supply voltage. When the control voltage
is made too high, however, the threshold
comparator will cease to work correctly,
while a too low voltage at the control in-
put upsets the bias point of the trigger
comparator (refer to the internal
diagram in Fig. 3).

The most evident application of the con-
trol voltage input is, of course, the 555 as
a voltage-controlled oscillator (VCO), as
shown in Fig. 10. The 555 itself is con-
figured as an AMV whose output fre-
quency can be varied over about ±50%.
In practice, especially when the supply
voltage is relatively high, a value con-
siderably lower than 0.45Ub, but with a
minimum of about 1.5 V, is permissible
for the control voltage. The frequency so
achieved becomes up to 2fo.

The basic circuit of Fig. 11 shows that
the control voltage input may also be
used for making an MMV with adjust-
able monotime. When, however, the
standard monostable configuration is
chosen, the output pulse can never be-
come too short. Assuming an input
voltage, Ui, at pin 5 of 0.45Uh, the
voltage on Ci will be kept at virtually
0 V by the internal discharge transistor.
When a relatively large control range of
the output pulse is desired, the lowest
voltage on Ci may be raised with the aid
of a zener diode, or a number of series-
connected, forward-biased, diodes, in
the collector line of the discharge tran-
sistor. To obtain a well-defined
minimum voltage on Ci, the quiescent
current through Ri, Im, must be just
high enough to achieve the correct zener
voltage, IL. This current is calculated

Im = (Ub-U/)/Ri

In practice, a few mA will suffice to

achieve the zener effect.

The circuit of Fig. 11 does not provide a
linear relationship between control input
voltage and output pulse-width. Such
linearity can be achieved, however, by re-
placing Ri with a current source. A
practical example and a detailed ex-
planation of this interesting configura-
tion is given in Ref. 1.

It is fairly simple to change the basic
voltage-controlled monostable into a
pulse-width modulated oscillator — see
Fig. 12. All that is required is another
AMV-based oscillator, set up around the
other 555 contained in the 556 chip. The
resulting circuit is an excellent, low-loss,
pulse-width modulator for use with a
power-transistor driver stage".

There are a few more interesting details
in the circuit shown in Fig. 12. The first
has to do with Ci, which is not
discharged to 0 V, but to a level set with
p.d. Ra-Rs plus the base-emitter drop
of Ti. Similar to the previously dis-
cussed ‘zener-trick’, this arrangement
considerably magnifies the span of the
output pulse-width.

The second interesting point of the cir-
cuit entails the simultaneous resetting
and triggering of MMVj to ensure an
accurately defined voltage on Ci at the
start of the each period. In the absence
of the trigger signal, a curious
phenomenon would take place when the
duty factor is, theoretically, as close as
possible to 1. During the first period, the
threshold voltage is not reached, so that
Ci is not discharged. Immediately after
the start of the second period, however,
the threshold level is reached, so that the
output goes low. The result of this se-
quence would be the halving of the out-
put signal frequency, and a reducton of
the duty factor from almost 1 to about
0.5.

As already said, this effect is prevented
by resetting the MMV at the start of

each period. Referring back to the inter-
nal diagram, the bistable is actually set
and reset at the same time. Reliable trig-
gering is, however, still ensured by virtue
of the internal reset circuit switching off
faster than the trigger circuit (Qis has
been driven into saturation, and has a
longer recovery time). Incidentally, the
recovery time of the trigger circuit can be
shortened by using a potential divider
that provides a trigger level just lower
than '/.iUb.

In the concept discussed here, the duty
factor can never become 1, because the
output is invariably low for the duration
of the reset signal of the MMV. This is
why R3 is generally made small relative
to Ri.

The control voltage input of a standard
555 forms a fairly low resistance
(5 kQ//10 kQ = 3.3 kS typ.). CMOS ver-
sions of the 555 have a much higher in-
put resistance thanks to an internal
voltage divider composed of three
100 kQ resistors. In general, tolerance on
these input resistance values is relatively
high, so that a voltage source driving the
control input should be designed to have
a low output impedance. . .

Long-interval timers

As already hinted at in the section on
timing errors, configuring the 555 as a
long-interval timer may pose problems
because of the inevitable role of leakage
currents in the timing components, i.e.,
the high-value resistor(s) and the capaci-
tor. A further aggravating effect is that
the leakage current of an electrolytic ca-
pacitor is age- and temperature-
dependent. In practice, the maximum in-
terval that can be achieved with a 555 in
standard configuration is 10 to
30 minutes long, taking a fairly high tol-
erance for granted.

1 3.43

at about 180 kHz, whereas a 7555 scored
1.1 MHz, and a TLC555 even 2.4 MHz
(test conditions: AMV configuration
with Ri = R2=220 Q and Ci = 100 pF).
As far as output current is concerned,
however, the bipolar 555, with its sink
and source capabilility of 200 mA, is
still superior to the CMOS versions. The
7555 supplies a maximum of 5 to
50 mA, depending on the supply voltage
(10 mA at 10 V). The TLC555 has a
symmetrical output with a source and
sink capability of 10 mA and 100 mA re-
spectively. Ergo , where the replacement
of a standard 555 with a CMOS type is
considered, the current requirement of
the load should be taken into account (a
standard 555s is often used to power a
relay direct). H

Fig. 13. Long-interval timers are best realized with the aid of a ripple-cascade divider.

One solution to obtain better-defined
and longer intervals would be the
cascading of 555s, so that each is trig-
gered by the previous one. This is not a
very neat solution to the problem, how-
ever, since all timing errors of individual
timers in the cascade simply add up (ac-
cumulation effect). Moreover, the dur-
ation of the interval rises only linearly
with the number of 555 stages. The in-
crease can be made exponential by fol-
lowing one 555 in AMV mode with a
divider as shown in Fig. 13. Depending
on the application, the n-th output of
the divider can trigger a further 555, this
time in MMV mode. In this set-up, the
555 in AMV mode is conveniently di-
mensioned for optimum accuracy
(average values for R2 and Rj, and a
low-leakage capacitor for Ci), while
cascaded dividers afford timer intervals
of hours, days or even weeks.

CMOS versions: 7555 and
TLC555

Intersil was the first to introduce the
7555, a CMOS version of the 555. A
little later, Texas Instruments, in line
with its consistent and successful policy
of producing LinCMOS (linear CMOS)
versions of ‘bipolar bestsellers’, came up
with the TLC555. As with a number of
well-established opamps and com-
parators, the TLC555 and TLC556 from
TI were an instant success.

In general, current consumption of the
CMOS versions has been drastically re-
duced with respect to the bipolar 555 —
from 10 mA to 100 n A, while the
minimum supply voltage has been
lowered to 2 V. Obviously, these features
are of great importance for the design of

battery-powered circuits. The CMOS
versions do not suffer the large peak cur-
rent at output switch-over, while the in-
put bias current of the threshold com-
parator, and the leakage current of the
discharge transistor, are also significant-
ly reduced. These features of the new
devices are advantageous because they
allow a higher charge resistance for the
capacitor, bringing longer timing inter-
vals within reach.

Thanks to the virtual absence of satura-
tion effects commonly associated with
bipolar transistors, speed of the new
CMOS 555’s has also increased. In a
laboratory test, a standard 555 gave up

3.44 elektor india march 198 $

COMPUTERS: AN OVERVIEW

by K.A. Roberts, BA

Early machines

It seems a far cry from the first auto-
matic computer, the Automatic Se-
quence Controlled Calculator— ASCC.
Yet, it is not quite half a century ago that
this machine, the result of a collabor-
ation between Dr Aiken of Harvard Uni-
versity and IBM, was presented to Har-
vard University in 1944.

Dr Aiken based much of his design on
the Analytical Engine conceived in 1832
by Charles Babbage. Like Babbage’s
brainchild, the ASCC used sets of
wheels as registers to store numbers. The
machine was composed of no fewer than
nearly 800,000 parts and almost 900 km
of wire.

The first electronic computer

The first digital electronic computer
came close on the heels of the ASCC: it
was in full operation in 1946. Named
ENIAC, acronym for Electronic
Numerical Integrator and Calculator, it
was designed by Dr J. Eckert and Dr J.
Mauchly of the University of Penn-
sylvania.

Where the ASCC was a mechanical
monster, the ENIAC was an electrical
one: it contained some 18,000 electronic
valves and consumed around 150 kW of
electric power.

Not long after the ENIAC had been
taken into operation at the University of
Pennsylvania, a course of lectures was
delivered at the same university that
formed the mould for today’s electronic
computer. The lectures. The Theory and
Techniques of Electronic Digital Com-
puters, contained the principles for the
design of electronic computers that had
been worked out by a group of
mathematicians and electrical engineers
headed by a, now famous, Hungarian
professor of mathematics working at
Princeton, Johann von Neumann.

From then on, computer technology ad-
vanced at an accelerating pace. So much
so that as early as 1956 Sir George
Thomson, the eminent physicist,
declared that

‘the electronic computer has not made
the headlines in the same way as
nuclear energy, but I believe it is com-
parable in importance. The ability to
apply precise reasoning to very large
amounts of data in a reasonable time

is something new, and the introduc-
tion of computers into science may
prove not much less important than
the introduction of mathematics in the
seventeenth century’.

The mainframe era

During the 1950s and 1960s, the elec-
tronic computer evolved into a useful,
but expensive tool. Scientific research,
defence organizations, large accounts
departments, and educational establish-
ments had all begun to use some kind of
computer.

The advent of time-sharing systems saw
an even greater degree of computer
penetration. The obvious advantages of
allowing many users within a single
company or organization to make simul-
taneous use of a central computer were
quickly spotted by commercial en-
trepreneurs who set up commercial time-
sharing services. Time sharing was for
many the only way of making use of
computer power: computers were still
very expensive. Their cost lay not only in
the initial outlay, but also in terms of op-
erational staff, space and power re-
quirements.

The minicomputer

Because of the high cost of computers,
manufacturers realized the need for a
smaller, relatively less complex (and thus
less expensive) machines. The first
manufacturer to bring one of these
machines on the market was Digital (in
1963). In comparison with the main-
frames then current, it was a limited
machine: it ran only one program at a
time, processed data in 12 bit-words and
had only 4 k of memory. None the less,
its advantages were obvious: it was not
much larger than a domestic freezer, did
not require an army of trained support
staff, and its cost was only about
5—10% of the mainframe computers of
the day when it was announced. Since it
sold extremely well, penetrating not only
new, but also existing, markets, its price
rapidly came down, so that even more
customers were attracted. By the early
1970s, no fewer than 70 US Firms were
manufacturing the so-called minicom-

Personal computers

Whereas the minicomputer came about
for sound economic reasons, the
microcomputer, perhaps belter known
as the personal computer, was, like radio
in its early days, developed by amateurs.
It should be noted that the industry at
that time did not think a personal com-
puter would ever cotton on (and this is
only 15 years ago!). It was in 1974 that
the July edition of Radio Electronics , an
American hobbyist magazine, carried an
article for the home construction of a
small computer. The Mark 8, as it was
called, used an Intel 8008 microproces-
sor, had 256 bytes of RAM (expandable
up to 16 k) and had no ROM.

Despite its limitations, interest in the
Mark 8 was phenomenal and sales
of parts for it far exceeded expections.
This interest, coupled with the introduc-
tion of Intel’s 8080 microprocessor,
prompted MITS, a small US electronics
company, to introduce the Altair 8800.
This design was also aimed at the hob-
byist and designed for another American
amateur publication, Popular Elec-
tronics. The project was published as a
series of constructional articles, the first
of which appeared in the January 1975
edition.

The computer was offered to readers of
Popular Electronics for $650 fully
assembled or $395 in kit form.

Apple Computers is born

Interest in the Altair 8800 caused the set-
ting up, all over the USA, of ‘computer
clubs’, run by, and for, amateur en-
thusiasts. A member of one such club in
California, Stephen Wozniak, a self-
taught computer engineer, got the idea
of designing and manufacturing a
similar kind of small computer, based on
the newly-introduced 6502 microproces-
sor.

Wozniak designed a small computer,
which was received enthusiastically by
his fellow club members. However, when
he approached his employers, Hewlett
Packard, to try to interest them in
manufacturing his computer, he met
with a bland refusal. Hewlett-Packard
did not think there was a sufficiently
large market for the machine!

A friend of Wozniak’s, Stephen Jobs,
thought differently. He approached a
number of potential buyers and eventu-

>3.45

ally got a contract for a quantity of the
Wozniak boards. Jobs and Wozniak
thereupon went into business for
themselves and formed what is now one
of the largest computer companies in the
world: Apple Computers. They have
never looked back!

All this happened only 15 years ago.
Today, the personal computer market far
outstrips the mainframe and minicom-
puter markets, and tens of millions of
PCs are in use the world over for a
multitude of applications.

Parallel processing

It was stated earlier on that von
Neumann’s model formed the mould for
today’s computer. That was true until
the arrival of the transputer. The pro-
cessor in traditional computers can
handle only a single instruction at a
time. This is true even in multi-user and
multi-tasking systems such as UNIX and
concurrent MS-DOS, where the pro-
cessor appears to be engaged in several
tasks at a time, but in reality assigns time
slots to portions of the relevant tasks.
Obviously^ the faster the processor, the
less users are aware of the time-sharing
process.

The transputer is a radical departure
from the von Neumann concept. It is
normalized for true concurrency. Paral-
lel processing of data and instructions is
achieved by synchronized very fast
point-to-point communication channels
between processes as well as individual
transputer modules. There is, in prin-
ciple, no limit on the number of
transputer modules that can be connec-
ted to form a computer.

In contrast to other processors,
transputers enable defining the speed of
the system simply by adding as many
modules as required. The IMS T800
transputer from Inmos, the designers
and manufacturers of the transputer, in
its 20 MHz version outperforms all of its
32-bit competitors, including the
National Semiconductor NS32332-32081
and Motorola’s MC68020-68881.

The calculation performance of the IMS
T800 is equal to that of the VAX 8600
scientific computer-from DEC, while a
network of ten IMS T800 modules offers
the speed and processing power of the
Cyber 205 supercomputer from Control
Data Corporation.

Because of their ability to work co-
operatively in parallel on a number of
different but related tasks, transputers
are well suited for use in so-called paral-
lel processing. By designing computers
that work on a number of tasks simul-
taneously, instead of doing everything in
sequence, designers aim to mimic more
closely the workings of the human brain.
Transputers are also being assigned to
less futuristic applications, including
desk top supercomputers, laser printers
and what have been nicknamed turbo-

chargers where the transputer is used as
an add-on unit to an existing system to
upgrade its performance. High-
performance graphics, engineering
workstations, and robotics are other
areas where the transputer has already
begun to make an impact.

The optical computer

Beyond the transputer, research is going
into photonic and molecular-based com-
puters. Basically, the heart of a com-
puter is the transistor (although there
may be thousands of them on one IC).
A transistor is nothing but a switch that
can flip backwards and forwards be-
tween two states. Therefore, computers
are chains of switches. They treat se-
quences of ons and offs to denote
numbers (in which case ons and offs are
read as the ones and zeros of binary
counting) or to denote true or false (in
which case chains of switches may be
used as the building blocks of algebraic
logic). Researchers at AT&T’s Bell Lab-
oratories and at Edinburgh’s Heriot-
Watt University have invented a device
that does for light what the transistor
does for electrons. This switch, known
as a Bistable Optical Device — BOD — or
transphasor, is essentially an optical
transistor. Light emerges from it as a
strong beam (on) or a weak one (off).
Put a bunch of transphasors together,
shine laser beams through them, and
you have the basic ingredients of an op-
tical computers.

The chemical computer

Even more advanced is the chemical
computer that will operate in the same
way as the human brain. The Science
and Research Council — SERC — a few
years ago set up a multi-million pound
research project, called MERI —
Molecular Electronics Research
Initiative — that is intended to keep
Britain in the forefront of advanced
computer technology.

The idea of a molecular computer was
first suggested by an American scientist.
Forest Carter, as a means of overcoming
heat dissipation problems in electronic
computers. Living organisms are made
up of carbon-based compounds, better
known as organic compounds that inter-
react to make possible, among many
other things, such functions as thinking.
Under the MERE biologists and elec-
tronics experts will work side by side to
engineer carbon-based chemicals that
can replace electronic components now
made from silicon. These chemicals will
be able to interact at molecular level and
will, therefore, provide enormous com-
puting power in a very small space. Since
molecules are interconnected in three di-
mensions, the computer based on them
would be able to use parallel processing
(like the transputer), making it very fast.

It would also be better at pattern
recognition than conventional com-

Some newcomers

Back to today, one of the most exciting
PCs to have come on the market in the
past 18 months is undoubtedly the Ar-
chimedes. It is the first PC equipped
with a 32-bit wide bus at a very
reasonable price. Its processor is an
Acorn Rise Machine — ARM — that is
cheap and very fast. The high processing
speed of 4 MIPS (million instructions
per second) is the result of RISC (Re-
duced Instruction Set Computer) tech-
nology. The philosophy behind this tech-
nology is that it is better for the pro-
cessor to work very fast from simple in-
structions than slowly from complex and
often little-used instructions. Already,
some versions of ARM have operated,
under laboratory conditions, at process-
ing speeds approaching 20 MIPS. It is
noteworthy that although the ARM is
comparable to Intel’s 80836 chip in per-
formance, its price is only about 1 /100th
of that of the 80836!

Another interesting introduction just
over a year ago was from the man they
can’t keep down: Sir Clive Sinclair. His
Z88 portable computer is cheap, small
(smaller than a size A4 sheet of paper)
and weighs just about 2 lb (less than
1 kg). All software is in ROM and it is
not compatible with anything. The Z88
is intended as an end-product and
comes, therefore, with all necessary soft-
ware. Its ROM, apart from a number of
tools, also contains a spreadsheet, a
diary, a word processor and the well-
known BBC BASIC. The programs may -
be used simultaneously. There is, of
course, a serial connection for a printer
so that texts from the word processor
may be sent straight to the printer.

Finally

With all the kerfuffle about computers
speeds, peripherals of a thousand kinds,
software of unimaginable variety, it is
sometimes well to reflect on the fact that
a computer can really do only two
things: carry out sequences of relatively
unimportant operations like adding or
copying, and choose between alternative
sequences.

THE DIGITAL MODEL
TRAIN — PART 1

by T. Wigmore

As every railway modeller knows, the control of model railways is
being transferred inexorably from the heavy-duty switches and
relays of yesteryear to the digital computer. In a new series of
articles, we describe a number of units based on the new
technology, culminating in a fully electronic model railway. The
series commences with a description of the Marklin control system
in which all commands to the signals, locomotive and points
(turnouts) are given via the rails.

Although in this and some future ar-
ticles reference will be made to the
Marklin system, it should be noted that
a number of units will be described that
may not only replace the relevant
Marklin circuit, but can be used in a var-
iety of DC railways of other manufac-

Rails: a serial bus

In any model railway, there are a number
of operations that must be under full
control at all times. Points (turnouts)
and signals may be operated in-
dependently of one another in a simple
manner, because they all have their own
power and control connections. The
drawback of this type of parallel control
is the ensuing complexity of the wiring.
It is far more complicated to control
locomotives independently, because
their only contact with the “driver” is
via the rails. There are control systems
that provide a number of high-frequency
command signals. Each locomotive is
then fitted with a special filter that
allows it to be operated on one specific
frequency only. Even these systems are
limited to 10 or 15 independent
locomotives, because the operating fre-
quencies must be spaced fairly widely to
ensure complete freedom from inter-
ference. However, time has already
caught up with these systems.

The Marklin system is unambiguously
based on computer technology. It makes
use of a two-wire bus (communication
channel) that is already present in any
model railway: the rails. Each item to be
controlled is connected to the rails (from
which it is also powered) and given an
address. When a given item, be it signal,
locomotive or point (turnout), is to be

operated, the relevant address is entered
on to the bus followed by a data stream
that contains the operating command. It
is clear that each item needs an address
decoder that will indicate when it is be-
ing addressed. The data stream contains
a certain measure of redundancy to ob-
viate erroneous operations. This is par-
ticularly useful with locomotives,
because the frequently bad contact be-
tween wheels and rails is a real source of
trouble.

The command signal is entered on to the
rails by the central control computer in

packets of nine bits (strictly speaking,
the supply voltage is being modulated).
Of the nine bits, the first four (in
locomotive decoders) or five (in point —
turnout— decoders) are accepted as ad-
dress bits and the remainder as data bits.
It is noteworthy that the so-called trinary
system is used for the address bits. In
this system, a bit can have three states:
logic 0, logic open, and logic 1. The pro-
tocol of these states is shown in serial
format in Fig. 2.

It is because of these three possible states
that a fairly large number of addresses

Fig. 1. Block schematic of a digital model railway as designed by Marklin. The rails are used

1 3.47

J_n n

Fig. 2. The serial data formal. The shorl
pulses serve as markers. The three stales of
the trinary system are: 00 = logic 0; 10 = logic
indeterminate; 11 = logic 1.

may be obtained with relatively few bits
(i.e. connections): 81 for locomotives
and (theoretically) 243 for points (turn-
outs).

Furthermore, the baud rate of the
locomotive control signal is half that of
the signals and point control signal.
Apart from ensuring a more reliable
data transmission to the locomotives,
the use of different baud rates enables
extending the address range. It should be
noted that the decoders for locomotives
and points (turnouts) operate in the
same address range (except, of course.

bit 5 which is a data bit for locomotives
and an address bit for points). The
decoder merely ignores signals with a
baud rate different from that for which
it is designed.

A practical circuit:
point/signals decoder

We have chosen a relatively simple cir-
cuit to describe the Marklin system. The
decoder in Fig. 4 may be used for the
control of up to four points (turnouts)
or signals. The serial data extracted from
the supply voltage via/? 7 and the clamp-
ing diodes on board ICi are decoded by
ICi. The first five bits are accepted as
address bits. However, input As is con-
nected to ground, so that only one third
of the address range is reserved for the
points and signals, i.e., theoretically, 81
decoders may be connected. Each
decoder is given a trinary address with
the aid of shorting plugs or wire bridges
(see also Table 1.).

The total number of points (turnouts) is
restricted to 256, because not more than
16 switching boxes (each with switches
for 16 points) can be connected to the
central computer. Evidently, not all
trinary addresses are used.

In each decoder, three of the four data
bits are used to form a sort of sub-
address that serves to select one of eight
possible magnet coils. This is done with

M-i-l-l

“

Fig. 3. Construction of the 9-bit data words
that control points (turnouts) and signals
(a) and locomotives (b). The address bits
operate with three logic states.

the aid of a 3-to-8 decoder, IC2, which,
on the command of the last data bit,
connects one of the darlington inputs to
the positive supply line via/?2. In the cir-
cuit, use is made of the darlingtons con-
tained in a ULN2001A, because this
device is relatively cheap. It also contains

Fig. 4. Circuit diagram of the decoder for points (turnouts) and signals.

Fig. 5. Block diagram of the MC145027.

a number of indispensable freewheeling
diodes. These diodes prevent the high
voltage peaks generated by the on and
off switching of inductive loads being
superimposed on the supply voltage.

A little more detail about ICi: see Fig. 5.
Network R3-C3 is used to differentiate
between short and long received pulses.
The short ones may be considered as
“markers”; the trinary information is
contained in the intervening long pulses.
Time constant R*-C* serves to separate
sequential data words.

If the received address, i.e., the first five
bits of a data byte, matches its wired-in
address, the decoder will transfer the re-
ceived data to a 4-bit shift register. They
are not yet available at the outputs. Only
when a second, identical, data word is
received are the data transferred to the
output register. This arrangement en-
sures a large degree of freedom from in-
terference.

Price/performance

considerations

It may not be clear what the advantages
are in using points (turnouts) decoders
instead of conventional wiring and
relays. After all, the saving in wire does
not compare with the cost of a decoder.
The main advantage of a decoder is that
it affords the possibility of “intelligent”
control of points (turnouts). The “intel-
ligence” may take the form of pre-
programmed switching of combinations
of points (turnouts) or of computer-
controlled scheduling and protection.

It is, of course, not possible to power
each and every locomotive via separate
wires. The advantage of a decoder is
here, therefore, much clearer.

In principle, a locomotive decoder works
in a similar fashion as that for signals
and points (turnouts). Four bits are used
to address a locomotive (up to 81 may be
used). Bit 5 is used for special functions
and the remaining four bits serve to con-
trol speed and direction.

Since locomotive power is present on the
rails at all times, permanent train
lighting presents no problems. It would
be feasible to power other aspects, such
as station lights, via the rails, but in view
of the maximum current the central unit
can provide, it is wise to power equip-
ment not directly connected with the
rolling stock from a separate supply.

The practical side

Constructing the points (turnouts)/
signals decoder on the printed-circuit
board shown in Fig. 6 should not pres-
ent any difficulties. Connecting it to the
track is no problem either. There are two
connections: red and brown and these
are connected to the corresponding ter-
minals of the Marklin system. Table 1
shows how the short-circuiting jump
wires are to be located for setting the
various addresses.

Each point (turnout) or signal has three
terminals. The central one of these is
used for the common wire of the two

Fig. 6. The printed-circuit board for the decoder.

Fig. 7. The maximum current at an output Fig. 8. Actual data streams. The one at the
may he doubled by connecting the two rcl- top is generated when bit 9 is set (power ap-
cvanl darlingtons in parallel with the aid of plied): the other when the data bits are reset
Iwo short wire bridges. (power removed).

l-'ig. 9. Circuit diagram of an encoder based on Motorola's MCI45026 that enables the
Marklin decoder to he used independently.

solenoids. The darlingtons are capable
of switching up to 500 mA per coil. If
points (turnouts) arc connected in paral-
lel, this maximum current must be borne
in mind. Since not all the darlingtons in
IC3 and IC4 are used, it is possible to in-
crease the current from some outputs by
a factor 2. To do this, two darlingtons
connected to the relevant output (see
Fig. 7) are connected in parallel with the
aid of two short wire bridges.

If Marklin points (turnouts) with lights
are used, these lights are powered by
disconnecting the yellow wire from the
central terminal of the solenoids and
connecting it instead to the central rail.
The same may be done with the signal
lights. This arrangement will cause the
lights to be on permanently. It is worth
considering connecting the wire, perhaps
via a switch, direct to the yellow AC con-
3.50 elektor Mama, ch 1989

nection on the transformer. This has the
added advantage that the load on the
central unit is decreased so that more
power becomes available for the trains.

Testing

For testing at least a Marklin central
unit, a keyboard, and a “control 80” are
needed. When every unit has been con-
nected, a red LED on the central unit
will light (it may be necessary to press
the go key on the “control 80” unit
first). The connected points (turnouts)
may be operated via the keyboard. It is,
of course, essential that all addresses arc
set on the decoder as well as on the dil
switches at the rear of the keyboard (see
Table 1).

Every time a key is depressed, two se-

quential pulse trains are put on to the
rails. Each of these trains contains two
identical data bytes. One pulse train
would, therefore, be sufficient, because
the decoder needs only two data bytes,
but for absolute security of operation
two trains have been arranged. Bits 1 to
4 constitute the relevant decoder ad-
dress; bit 5 is always 0; bit 6 to 8 form the
sub-address of the appropriate solenoid;
and bit 9 is 1 so that power is applied.
When the key is released, four bytes are
again put on to the rails, but this time
with the data bits reset to cause power to
be removed.

Alternative control circuit

The points (turnouts) decoder may also
be used independent of the Marklin sys-
tem, not only with model railways, but
also as a two-wire remote control unit.
This is made possible by Motorola’s
mc145026 encoder, This IC makes it
possible to construct a substitute for the
Marklin control with only a few ad-
ditional components (see Fig. 9). The
encoder has 9 address/data inputs. Input

5 is connected to earth. The trinary
decoder address must be placed on in-
puts 1 to 4 and the solenoid sub-address
(binary) on inputs 6 to 8 . The power is
reinoved or applied by bit 9. In the cir-

cuit, switches are shown for setting the
addresses, but the inputs may just as well
be controlled by the output port of a
computer. Note, however, that these in-
puts are not TTL compatible (not even if
the decoder would operate from a 5 V
supply). Furthermore, for each address
bit a third logic state (high impedance)
must be available. This means that an in-
terface is required between the output
port and the encoder.

A short pulse at the te input (transmit
enable) results in the set byte being sent
twice in succession. If the te input is
kept low permanently, the encoder sends
continuously.

When one central power supply is used
(as in the Marklin system), the data are
superimposed on the supply voltage
which is effected by the boxed section at
the right in Fig. 9. It is, however, also
possible to give each decoder a separate
supply, so that only earth and a signal
line have to be provided. The boxed sec-
tion in Fig. 9 is then not required. The
output of the encoder and earth are then
connected direct to R and the data input
respectively (it may be necessary to
remove/? 7 ). H

MOLECULAR ELECTRONICS

Towards an advanced form of computer technology

by John Delin

Smell sensors, paper-thin tele-
vision screens, moving
holograms and bio-computers
using living organisms arc all
the stuff of fantasy. But so was
the silicon chip or the seem-
ingly incredible notion that a
million pieces of information
could be stored in a grain of
dust, only a few decades ago.
Today's microprocessors are
hardly obsolete but they are
already revealing limitations in
terms of size and other physical
constraints. Where then can
scientists turn' to complement
existing technologies while pro-
viding an exciting springboard
into the fantasies and realities

of the future?

Molecular electronics, designed
to harness the molecule itself as
an information processor, show
considerable promise. Gathered
together from disparate
research over a range of
disciplines, this line of thought
has attracted considerable in-
terest in Britain and has now
been selected as one of the
major areas of the Department
of Trade and Industry's new
Link* 1 * programme of col-
laborative research between
universities and industry.

At least £20 million is to be
allocated to molecular elec-
tronics, half from government

sources and half from industry,
to cover the so-called pre-
competitive stage of develop-
ment, delving into fundamental
principles and the feasibility of
devices. The programme aims
to provide the platform from
which industry and industrially
sponsored research can later de-
velop exploitable products.

Practical application

Molecular electronics uses
organic molecules to process in-
formation. It goes beyond the
digital processing of conven-
tional electronics and adds new
dimensions — for example struc-

tures and shapes — to its
vocabulary. Conventional elec-
tronics are analagous with the
nerves in the body that trigger
when a certain electrical
threshold is reached. Molecular
electronics resemble the white
corpuscles that react to the
shape, density or temperature
of a bacterial invader. They are
conceptual in action rather
than computational.

One familiar example of
molecular electronics in action
is the liquid crystal display seen
in watches and calculators that
respond vigorously to electrical
or heat signals. These are
already in use in a number of

1 3.51

SCIENCE & TECHNOLOGY

Recognizing speech in noise

by Dr William Ainsworth, Department of Communication and Neuroscience, University of Keele

Few people have any experience of communicating verbally with
computers and even fewer have ever done so in a noisy
environment. Yet in a factory or when using a telephone in a busy
office, recognizing and decoding speech is a familiar problem.
But it will take many years of research before the most efficient
form of man-machine interface will be evolved, though the task
has to be tackled if we are to be able to talk to computers
against a background of machinery, in a motor car or on a flight
deck. Headway is already being made in analysing the difficulties
and outlining ways to overcome them.

Speech dominates human communi-
cation. If we want people to do
something, or we need certain infor-
mation from them, we simply speak to
them. If they are far away we may write
them a letter, but most people prefer to
pick up a telephone, perhaps because
reading and writing seem much more
complicated than speaking and listen-
ing. That is hardly surprising, for it takes
years of practice at school to become
proficient in the skills needed to read
and write.

When we want to communicate with a
machine we have to learn new skills. We
need to know how to poke at a keyboard
with our fingers and to watch the effect
it has on a screen. How much easier it
would be if we could simply speak into
a microphone to get the machine to do
what we wanted!

This dream occurred to speech
technologists many years ago, and for
the last 20 years or so they have been try-
ing to devise techniques for getting
machines to respond effectively to
speech signals.

Speech communication appears to be a
simple process. It is learned by every
healthy child with little or no effort. In
reality it is not simple: it is a most com-
plex process. An idea in the mind of the
speaker must first be expressed as a
sentence in a language understood by
both him and the listener. It must then
be articulated. We do it by modulating
the airstream from the lungs by the vocal
cords to produce a sequence of pulses
whose frequency determines the inton-
ation. The pulses excite the resonances
of the vocal tract and then radiate from

the lips as a sound wave. The meaning of
the sentence is coded in this wave by sub-
tle movements of the tongue, jaw and
lips. These complex movements are
known intuitively by everyone who has
learned the language.

But this is only half the story. The sound
wave passes through the outer ear of the
listener and causes the eardrum to
vibrate. These vibrations cause the
ossicles, a series of small bones attached
to the eardrum, to move and pump fluid
in the cochlea, or inner ear. In the
cochlea is the basilar membrane which
oscillates at various places along it
which depend upon the frequencies pres-
ent in the input signal. So, the structure
of the inner ear begins the process of
decoding the speech wave. Attached to
the basilar membrane are a large number
of hair cells, some 30,000 of them,
which actuate nerve cells when they
bend. These cells are the first stage in a
complex system which leads up the
brainstem and eventually to the auditory
cortex.

Automatic recognition

So far, the processes by which the speech
signals are decoded by the brain are not
well understood, so programming a
computer to recognize speech in the
same way that the brain operates is ob-
viously impossible. Nevertheless, for
many practical purposes a machine
which recognizes just a few words can be
very useful. For example, consider a
program that displays the choices
available to the user by means of
numbered menus. If the machine can

just recognize the spoken digits the user
can complete his task by voice.

Most practical speech recognizers work
by pattern matching. The user speaks all
the words in the machine’s vocabulary
and the machine analyses them and
stores the result. These stored patterns
are often known as templates. When an
unknown word is spoken, the machine
compares this new utterance with each
of the stored templates and chooses the
one which gives the best match.

Several techniques have been employed
to analyse speech signals. We know that
speech is encoded in terms of fre-
quencies and that the human auditory
system begins its analysis of sounds by
separating them into their component
frequencies, so spectral analysis is a
popular technique. The upper of the two
illustrations shows a sound spectrogram,
or sonagram, of the word. ‘.recognition’.
Frequency is represented by the
blackness of the picture. The dominant
frequencies, known as the formants, can
be seen as the black horizontal bands.
These reflect the resonances of the vocal
tract.

Problems

Speech recognizers built on these prin-
ciples alone are not very successful for
three reasons:

(1) Every time we utter a word we speak
at a different rate, so some patterns

are spread out in time compared with

(2) Different people have different sized
vocal tracts, so the formants occur at

different frequencies when they say the

Time (seconds)

Time (seconds)

Sonacniins of Hie word 'recognition'. At the top it is spoken against a quiet background, and
above with a signal-to-noise ratio of -6 dB.

same word.

(3) Most speech communication takes
place not in isolation, but against a
background of other noises.

Various techniques have been devised for
dealing with these problems. The first
problem can be dealt with by so-called
dynamic time warping. This enables the
stored templates to be expanded or com-
pressed in such a way that the optimum
match is obtained. Alternatively the
problem can be dealt with by building
statistical models of each word which in-
corporate the variability of the ut-
terances.

Usually the multi-speaker problem has
been circumvented by training the system

with the voice of the user but there have
been some attempts to cope with it by
building transformations for each new
speaker that enable his voice to be
transformed into one like that of the per-
son who originally trained the system.
Here, statistical modelling of the
variability has again been used.

The problem of recognition in noise has
not yet been solved. John Bridle and his
colleagues at the UK Royal Signals and
Radar Research Establishment in
Malvern some years ago showed that a
speech recognizer which worked well in
the quiet recognized only about 50 per
cent of spoken digits correctly when the
signal-to-noise ratio was +3 dB
(decibels). This is far worse than human
performance. It has been known for

many years that spoken digits can be
recognized with almost complete accu-
racy with a signal-to-noise ratio as poor
as -6 dB, which means the intensity of
the speech is much less than that of the
noise. A sonagram of the word ‘recog-
nition’ with a signal-to-noise ratio of
-6 dB is shown in the lower of the two
illustrations.

Auditory modelling

The superior performance of people in
recognizing speech in noise has led to the
suggestion that speech analysers which
operate on the same principles as the
human auditory system might work bet-
ter than those based on conventional
techniques. Preliminary experiments by
Dr Ghitza at the Bell Laboratories in the
USA and others elsewhere have shown
promising results.

Our Department of Communication
and Neuroscience comprises a number
of research groups which investigate the
mechanisms of vision, hearing and
speech. Professor Ted Evans, the head of
the department and leader of the
Auditory Physiology group, has
developed an electronic model of a single
channel of the auditory system. It gives
responses similar to those obtained by
inserting micro-electrodes in the
auditory systems of cats.

Professor Evans' model consists of a
filter with characteristics that simulate
those of the inner ear, a half-wave recti-
fier and logarithmic compressor to
represent the action of the hair cells, and
what is called a probalistic spike gener-
ator to simulate the production of action
potentials in nerve cells.

Our Speech and Auditory Physiology
groups are collaborating with Dr Pat
Wilson of the Auditory Psychophysics
group to produce a computational
model of the auditory system with 100 or
more channels. This work is made poss-
ible by a grant from the UK Science and
Engineering Research Counsil to install
a fast computer that will enable the
model to process signals, especially
speech, in a reasonable time.

The first stage of the model consists of
a bank of bank-pass filters which
simulate the signal processing as far as
the auditory nerve. The characteristics of
these filters are estimated by a process
known as reverse correlation. A random
noise signal is applied to the auditory
system and responses are recorded from
the auditory nerve by means of a
microelectrode. The noise signal causing
the nerve fibre to respond is also re-
corded. By a process similar to cross cor-
relation between the noise input signal
and the response of the nerve fibre, the
impulse response (the response of a filter
to a single impulse) of the auditory filter
is found (in practice the impulse
response is reversed in time; hence the
term reverse correlation). Several ex-

periments have to be done with a num-
ber of nerve cells, each tuned to respond
to different frequencies, to develop the
impulse responses of a bank of filters.
These impulse responses can be pro-
grammed on the computer and used to
simulate the filtering characteristics of
the auditory system. The other stages of
auditory processing, logarithmic com-
pression and rectification by the hair
cells and the generation of spikes accord-
ing to a probability function, can also be
programmed. The result is a compu-
tational model which allows the signals
generated at each level in response to
speech sounds to be studied.

The auditory system is more com-
plicated than I have already outlined.
Recent physiological studies have shown
that there are interactions between the
channels: if there is activity in one chan-
nel, the activity in neighbouring chan-
nels is suppressed. This mechanism
might be responsible for reducing the ef-
fects of noise while enabling speech
signals to be transmitted to the higher
regions of the auditory system. We in-
tend to build lateral suppression into our
model and to investigate what effect it
has on speech processing.

Speech synthesis

Techniques for speech synthesis were
developed about 20 years ago. In a
typical system a sentence is first
translated into a sequence of phonetic
units, which represent the way in which
each sound is pronounced. This can be
done by looking up each word in a
phonetic dictionary or by applying a set
of context-sensitive rules (for example p
followed by A is pronounced f, otherwise
P)-

The phonetic units are then translated
into acoustic parameters which represent
the physical characteristics of the
sounds. The acoustic parameters are the
frequencies of the formants (see left-
hand illustration), their intensities, and
their durations. They are used to control
a speech synthesiser consisting of a set
of resonators excited by a sequence of
pulses.

Although such a system produces in-
telligible speech, the output sounds
rather mechanical. Moreover, it has been
found that when it is heard against a
background of noise it is a great deal less
intelligible than equally loud natural
speech. We are collaborating with the
IBM Scientific Centre in Winchester to
try to discover why this is so.

One possibility is that whereas this sys-
tem faithfully models the resonances of
the vocal tract it does not employ
realistic excitation pulses. A technique
known as inverse filtering is being used
to measure the shapes of the excitation
pulses in human speech. In this tech-
nique the characteristics of the vocal
tract filter are estimated, and then the
characteristic of the filter is inverted. If
speech signals are passed through the
‘inverted’ filter, only the excitation
pulses remain.

Using the technique we are able to study
the variation in shape of the excitation
pulses. This knowledge can be applied to
speech synthesis. We expect that speech
synthesized in this way will be more in-
telligible in the presence of background
noise.

User interface

When the captain of a ship gives a
compass course for the helmsman to
steer, the helmsman repeats it back to
confirm that he has heard it correctly.
When a telephone operator is asked to
obtain a number she repeats the number
back. Communicating with a machine in
a noisy environment is somewhat
similar. The noise may corrupt the
speech signal and cause an error in
recognition. The user will be unaware of
the mistake unless the words are
displayed on a screen or the machine is
equipped with a synthesizer to speak
back to him. If the user is com-
municating over a telephone line, or if
his eyes are busy whith another task, the
latter course may be the only one that is
practicable.

The question arises as to whether the
response of the recognizer should be
checked after each word has been spoken

to it or whether it should be checked
later, for example, at the end of each
sentence. Compass courses always con-
sist of three digits and they are repeated
back as a group. Telephone numbers, on
the other hand, vary quite widely in the
number of digits they contain. They are
often checked after three digits, but on a
bad line digits may be checked one by

Here at Keele we are interested in com-
municating with computers in a noisy
environment where it is likely, in spite of
advances in recognition from auditory
modelling and in synthesis from realistic
excitation pulses, that occasional
mistakes will be made. So we are in-
terested in finding the most efficient
ways of detecting and correcting errors.
We have developed a mathematical
model of the user interface, which
enables us to arrive at the optimum
number of words which should be
spoken before any checking is done. This
model predicts, as might be expected,
that as the noise level rises and the fre-
quency of errors increases, the number
of words spoken before a check is made
should be reduced. Experiments have
shown that the specific predictions of
the model are borne out in practice.

Future plans

Our research is by no means completed.
We have only recently acquired com-
puters powerful enough for us to carry
out the work. When even more powerful
computers come into use, progress will
be faster.

Advances are continually being made in
understanding the physiology of the
auditory system. We intend to incor-
porate these developments in future
auditory models and to test their utility
in automatic speech recognizers. Our
programme in speech synthesis has been
hampered by a lack of kndwledge as to
how the shape of the excitation pulses
varies in natural speech. We are gradu-
ally acquiring this knowledge and in due
course it will be transferred to our
speech synthesis system.

ELECTRONICS NEWS

Japan threatens U.S.

Japanese are rapidly moving ahead of
the U.S. in the development of a cru-
cial new X-ray technology that will be
used to manufacture computer chips
in mid-nineties.

Major American electronic firms and
scientists have cautioned the U. S. ad-
ministration that American Compete-

tiveness is at stake in a number ot vital
areas from military technology to con-
sumer electronics.

The most advanced commercial chips
available now can store on million bits
of information, approximately 62
typed pages. Experts say that the limit
to such chips is 16 million bits.

The developing technology called X-
ray lithography, can make denser
chips that scientists think will ulti-
mately enable storage of 1000 times
more data. Such computing power is

now available only with the largest
IBM mainframe computer, which oc-
cupies several refregirator-sized
cabinets. The new chips would be
about the size of a finger nail.

The Japanese have set up a joint in-
dustry-government venture on X-ray
lithography at a cost of one billion dol-
lars. Since, the technology is too ex-
pensive for any single company to de-
velop, the IBM has approached other
companies to share the equipment
cost for the X-ray lithography studies.

1 3.55

PRACTICAL FILTER DESIGN (2)

by H. Baggott

Each filter has its own typical properties and these can be laid
down in a few parameters. The second part of this series explains
what these parameters are and what they mean

There are a number of parameters that
characterize the properties of a filter.
One of these is the frequency response
characteristic or curve. The designer,
having drawn up a target specification
for the ripple in the pass-band and the
slope of the filter skirt, will have to make
a choice from several possibilities. The
type of filter, whether it is a high-pass or
band-pass, and so on, is not of import-
ance at this stage.

Any type of filter can be converted into
a standard low-pass with a cut-off fre-
quency of 1 Hz. The target requirements
must be translated into a normalized
low-pass filter specification. After that,
they may be compared with available
standard curves with a 1 Hz cut-off
point.

After a choice has been made, the re-
quired filter is simply reconverted and
dimensioned for the required fre-
quencies.

The designer has a choice of the follow-
ing filter types:

• Butterworth

• Bessel

• Chebyshev

• transition

• linear-phase

• synchronous-tuned

• elliptic-function.

Apart from those of elliptic-function
fillers, the frequency characteristics of
all these types are normalized for a
-3 dB cut-off point at 1 Hz. The curves
may be scaled to the desired frequency
with the aid of standard multipliers.

o • o

880176-10

Fig. 6. A simple rc network with its -3 dB
cut-off point at 1 Hz.

3.56 elektor India march 1989

Filter parameters

As an example of the operation of a
filter, we will consider the simplest type:
an rc network as shown in Fig. 6. This
network is terminated into an infinitely
high impedance and powered by a
voltage source that has an internal resist-
ance of zero ohms. The capacitor is the
frequency-dependent element and it in-
troduces a phase shift.

The transfer function of the filter is

r(/co) = 1/(1 +juCR ) [4]

The absolute value of the function is
I T(/«>) I =l/[Vl+(oiCR) : ] [51

Fig. 8. The linear X-axis on the phase shift
curve gives a good idea of the time delay
caused by the filter.

The resulting phase shift is

0 = — arctan(cw/?C) [6]

Equations [5] and [6] enable the gain vs
frequency and the phase shift vs fre-
quency characteristics to be computed
and these are shown in Fig. 7 and Fig. 8
respectively. It should be noted that
these curves are drawn on linear coor-
dinates and that, therefore, particularly
the gain curve is not the nearly straight
line usually encountered. This is because
the curves are normally drawn on a
logarithmic abscissa (X-axis).

None the less, the phase characteristic in
Fig. 8 shows how well the filter func-
tion approaches the condition not to

Fig. 9. The input impedance of the sample
network is not constant but increases at low
frequencies.

Fig. 10. The phase shift and gain
characteristics are normally shown on the
same illustration.

forthcoming articles in this series.

a. gain vs frequency;

b. phase shift vs frequency;

c. time delay vs frequency;

d. step response.

introduce delay distortion (d >//=
constant). On a linear scale, the curve
should be a sloping straight line. This
aspect is difficult to judge when a
logarithmic scale is used.

The input impedance of the filter is, of
course, also a point to be considered. It
is not possible, as many of us have found
by bitter experience, to connect a num-
ber of filters in cascade to obtain a sharp
cut-off response. Since the reactance of
some filter components is frequency-
dependent, the input impedance will
also vary with frequency. This may be
seen from Fig. 9, which illustrates the in-
put impedance of our sample rc net-
work.

Furthermore, a filter is always computed
for a fixed ohmic termination. If that
load is replaced by another filter presen-
ting a frequency-dependent impedance,
neither of the two filters will behave as
originally designed.

As already stated, since the frequency
and phase characteristics are normally
drawn on a logarithmic X-axis (and
quite often shown together as in Fig. 10),
it is difficult to ascertain the time delay
from them. For that reason, the time
delay characteristic (computed from the
frequency and phase characteristics) is
often added on the same illustration.
For some applications, it is important to
known the step response of the filter.
This is a measure of the reaction of the
network to a sudden rise in input
voltage.

The four parameters just discussed give
virtually all the information the designer
normally requires.

Standard curves

The standard curves of our sample rc
network are shown in Fig. 11. Such
curves will also be given for all types of
filter in forthcoming articles in this
series. We will endeavour to give them all
on the same scale so that a direct com-
parison may be made. All curves have
been computed with the aid of a net-
work analysis program to obtain
representative characteristics that are as
accurate as possible. All of them have
been normalized on a cut-off frequency
of 1 Hz.

Reverting to Fig. 11, a and b show the
gain vs frequency and the phase shift vs
frequency respectively. Fig. 11c is the
time delay vs frequency curve computed
from curves a and b. Fig. lid gives the
step response of the network; the upper
part of the illustration shows the sud-
dent increase in input voltage from 0 to
1 V, and the lower part the resulting
change in output voltage. Curves in
future articles in this series will not show
the upper part again, because the rise in
input voltage is always taken as shown
here. The step response of our sample
filter does not mean much, of course,
because the network is so simple. In the

case of more complex networks (of the
second and higher orders), the step
response will show at a glance whether
there is any ringing, how long this lasts,
and the extent of the overshoot.

A sample computation

To end this second part of the series, we
will give a sample computation to show
how a filter is dimensioned in line with
the foregoing discussion. Assume that
we need an rc network as shown in
Fig. 6 that is powered from a low-
impedance voltage source and is ter-
minated into a fairly high im-
pedance^ 1 MS)). The cut-off point is
required to be at 3 kHz (the multiplier,
m, is thus 3,000). We choose a standard
value for*, say 10 k£2. The value of the
capacitor is divided by the value of the
resistor and the multiplier. If the net-
work had contained an inductance in-
stead of a capacitor, the value of the in-
ductor would be multiplied by the value
of the resistor and the result divided by
the multiplier. In the rc network:

C = 0.159/mR =

C = 0.159/3000x10000=

C = 5.3xl0" 9 =5.3 nF

The time delay at a given frequency may
be calculated by reading the delay at that
frequency in Fig. 11c and dividing that
value by m. The same applies to the time
scale of the step response curve. K

VIDEO CARDS FOR PERSONAL
COMPUTERS

by H. Stenhouse

In recent years a bewildering variety of video cards for PCs has
come on to the market. This article attempts to remove much of
the confusion caused by the different specifications and monitor
requirements.

Functionally, the video card in a per-
sonal computer is an output device. Over
the past few years, as PCs grew more
sophisticated and users more deman-
ding, the video card has become more
than the fairly simple text display circuit
of yesteryear. At that time, no provision
was made for displaying, say, a graph on
the screen. Fortunately, this was cor-
rected with the introduction of the
Colour Graphics Adaptor (CGA), which
did allow, at least partly, for integration
of text with simple graphics. The main
disadvantage of the CGA was, however,
its limited resolution for text. The well-
known Hercules card, developed by the
company of the same name, overcame
this shortcoming at least for
monochrome text applications. A few
years later, the EGA card (EGA = En-
hanced Graphics Adaptor) and the PGC
card (Professional Graphics Adaptor)
were introduced to satisfy more deman-
ding users wishing to work with high-
resolution colour screens. But the evol-
ution of the videocard did not stop with
the PGC: the introduction of the new
series of PS/2 computers from IBM
called for even higher resolution and
speed: the answer was provided in the
form of a range of MCGA and VGA

The evolution from the basic video card
to the highly sophisticated graphics
adaptor available now has caused great
confusion among many PC users. This is
mainly because the systems are often in-
compatible as far as the monitor,
horizontal and vertical scanning fre-
quency, and even the interconnecting
cables are concerned. The CGA (8
colours) and the EGA card (16 colours),
for instance, supply output signals at
TTL level, usually combined with an
intensity signal, whereas other
videocards, such as the PGC and VGA
have linear video outputs that allow a
very high number of displayable colours.
Owing to the structure of the on-board
RAM, the VGA and PGC work with
video modes in which a limited number
of colours — say, 256 — can be active
at a time. These cards offer an indirect
choice of nearly a quarter of a million
shades via a palette structure.

Computer display manufacturers have
traditionally supported each new PC
video card with an appropriate display.
An exception to this is formed by the so-
called multisync monitor, which is
available in many types from, for in-
stance, NEC (Multisync-2), Eizo (Flex-
scan 8060S and 9070S) and Taxan. The

electronics in this advanced type of dis-
play is capable of automatic adjustment
to the internal line and raster frequency
detected in the applied video signal. In
addition to this extremely useful feature,
the display often has inputs for linear as
well as digital video signals, so that it
can be used with virtually all current
videocards.

Unfortunately, a CVBS (Chroma,
Video, Blanking and Sync) input to the
PAL (or NTSC) standard is rarely found
on high-resolution colour monitors for
computers. Such an input is, admittedly,
not very useful in the PC environment,
but may, on the other hand, give in-
teresting opportunities for use of the
high-resolution display in conjunction
with cameras, VCRs, video digitizers,
and some types of home computer.

The monochrome scene

The Monochrome Display Adaptor
(MDA) fitted in the earliest of IBM PCs
provided only text display. This card,
which is now obsolete, had a screen
memory of only 4 Kbyte (4,000
characters), and displayed text as 25 lines
of 80 characters. None the less, the resol-
ution of the MDA is relatively high at

720x350 pixels. The character font is
7x11 pixels in a 9x 14 raster, resulting in
a clear text display. The card provides
256 characters, which are stored in an
on-board ROM. No provision is made
for the user to define his own characters.
The Hercules card replaces the MDA,
and adds a graphics option in the form
of a monochrome graphics interface
with a resolution of 720x348 pixels. Text
display is basically the same as with the
MDA, and requires no special software.
Graphics software, however, can only be
run with the aid of software utility
1NT10, since IBM, and, therefore, the
Disk Operating System (DOS), does not
support the Hercules card. After an in-
itial shortage of graphics software for
the Hercules card, this is now supported
in the majority of programs from
leading software companies. The use of
the Hercules card is also boosted by pro-
grams such as MG-2 (MuItiGraph-2)
that allow it to emulate the CGA mode
with the aid of grey shades. Currently,
the Hercules card is probably the widest
used video adaptor for PCs running
word-processing and other text appli-
cations. As a useful boon, the card pro-
vides a parallel printer output port,
LPT1:.

Colour and graphics: the
CGA

The CGA was the first card introduced
by IBM that allowed the connection of a
colour display to a PC. On board the
CGA is a 16 Kbyte memory. The card
can operate in two modes: text and
graphics. In text mode, two sub-modes
are available: 40 or 80 characters per
line, at 25 lines per screen in both cases.
The available memory allows 8 or 4
screens to be stored in 40 and 80 charac-

Fig. 1. Half-length Colour Graphics Adaptor.

ter mode, respectively, so that fast scroll-
ing can be achieved. The graphics mode
also affords a number of sub-modes, in-
cluding one with 640 x 200 pixels at two
colours, and 320 x 200 pixels at four
colours. In graphics mode, characters
with an ASCII value greater than 127
can be shaped by the user. Since
characters are formed in an 8x8 matrix,
the CGA is less suitable for text display.
In many cases, a CGA and a Hercules
card can be used alongside in the com-
puter, but only if the Hercules card is not
used in the so-called full-size mode
(64 Kbytes of screen memory).
Switching between the two cards can be
done in software, so that monochrome
text can be combined with colour
graphics on separate screens.

The CGA double-scan card is an im-
proved version of the standard CGA.
This type of video adaptor is available in
the form of an emulated mode on some
EGA cards, and enables software written
for the CGA to be run on a display with
much higher resolution. This is mainly
by virtue of the double-scan principle,
which provides an interlace function
that effectively doubles the vertical resol-
ution. Unfortunately, this interesting
mode is not available in the form of a
separate card.

Enter the EGA

The cost of an Enhanced Graphics
Adaptor (EGA) was, for a time, pro-
hibitive for the average PC user, but
that, fortunately, changed with the
availability of good-quality products
from the Far East. The EGA has a large,
256 Kbyte, on-board memory, and offers
a graphics resolution of 640x350 pixels,
at 16 possible colours per pixel (a 256-
colour extension for the EGA is de-
scribed in Ref. 1). Pixel colour selection
is from a 64-colour palette. Depending
on the resolution, two or four screens
can be held in the memory.

The character set of the EGA is ROM-
resident, and uses an 8x14 matrix to
guarantee excellent text display
capabilities. Provision has been made
for the user to shape up to 1024
characters at a height of 8 to 32 pixels.
Many manufacturers of EGA cards have
come up with useful extensions to the
basic capabilities, often in the form of
emulation modes. In many cases, soft-
ware is supplied with the card that allows
it to switch to the CGA, MDA and CGA
double-scan mode. The EGA-Wonder
card from ATI takes compatibility even
further by its ability to adapt the outputs
to the display used. Other EGA cards
elektor India march 1983 3.59

can be used in conjunction with a CGA-
compatible monitor, with the obvious
advantage of going round investing in a
new, high-resolution, monitor.

PGC: professional at a
professional price

The Professional Graphics Adaptor
(PGC) was developed and introduced to
convince PC users of the fact that CAD
software heed not necessarily be run on
a professional workstation. Unfortu-
nately, the PGC has remained relatively
expensive, and has, therefore, failed to
become popular. Aimed at the CAD
market, the PGC was designed to pro-
vide an aspect ratio of 4:3, and to
generate up to 256 colours. The
multisync monitor mentioned earlier
makes it possible to run PGC-based
CAD software in the EGA+ mode
available on the latest multi-mode EGA
cards. That the PGC is bound to be
forgotten soon is also caused by the fact
that the Video Grapics Array (VGA), in-
troduced with IBM’s line of PS/2 com-
puters, is in principle capable of taking
over all its functions. . . as a subset!

New standards: MCGA and
VGA

In an attempt to put an end to the
widespread confusion about videocards
in PCs, IBM recently introduced two
new types of display adaptor, the Video
Graphics Array (VGA) and the Multi-
Color Graphics Array (MCGA), for use
in their Series PS/2 computers. Both
adaptors are complete, versatile, and ex-
pected to stay with us for quite some
time. The MCGA is essentially a ‘low-
budget’ version of the VGA. It comes as
standard with the Model 30 computer in
IBM’s PS/2 line, and has 64 Kbyte of
on-board RAM. The maximum resol-
ution of 640 x 480 pixels is achieved in
the two-colour mode. At the lower resol-

ution of 300 x 200 pixels, 256 colours are
available from a total of 262,144 in the
colour palette. The MCGA can display
up to 64 grey shades on a monochrome
monitor. Downwards compatibility is en-
sured at least partly by a CGA emulation
mode. Other cards such as the EGA or
MDA can not be emulated.

The second new card, the VGA, is fitted
in PS/2 Models 50, 60 and 80. It can be
used with colour as well as monochrome
monitors with an analogue (linear) in-
put. Screen memory is 256 kByte for a
standard graphics resolution of 640 x 480
pixels, or 720 x 400 pixels in the text
modes. In the low resolution graphics
mode, each of the 320x200 pixels can be
assigned one of 256 colours. In the high-
resolution mode, this is reduced to 16
colours. The number of available
colours in the palette circuit is equal to
that in the MCGA. Depending on the
selected screen mode, up to 8 screens can
be held in memory. Characters are built
in a matrix of 9x16 pixels in text mode,
or 8x16 pixels in graphics mode. The
VGA is capable of emulating all previous

standards, epsuring software compati-
bility with MDA, CGA, EGA and
MCGA. Characters in these subsets have
a maximum height of 32 pixels.

Wanted: monitors!

Selecting a videocard is one thing, find-
ing a suitable monitor for it is another.
Table 1 summarizes the vertical and
horizontal scanning frequencies of a
number of PC vidcocards. In general,
the requirements of the cathode ray tube
(CRT) used in the monitor rise with sync
frequency. Excellent resolution on a
flicker-free display is achieved thanks to
high raster and line frequencies (up to
80 Hz and 50 kHz respectively), and
non-glare screens.

Obviously, investing in an expensive
videocard is useless if the monitor has
insufficient resolution. In the case of the
monochrome monitor, the maximum
resolution is usually determined by the
bandwidth of the video amplifier. This
means that the design and production of
a monochrome monitor are simple com-
pared with those of a colour monitor
with equal specifications in respect of
resolution. For optimum convergence in
a colour picture tube, the electron beams
must be controlled with great accuracy
to ensure the actuation of only one
phospor element at a time. The size of
these elements varies from 0.62 mm in a
standard colour TV tube to 0.29 mm in
a multisync high-resolution colour
monitor. For most graphics applications,
a dot pitch of 0.31 mm is sufficient.

The large differences in respect of line
and raster frequency between the various
videocards give rise to monitor incom-
patibility. A standard CGA display, for
instance, can not be used in conjunction
with a Hercules card. Some
monochrome, Hercules compatible,
monitors, however, are capable of dual-
frequency operation so that CGA pic-

Fig. 2. Hercules card for combined medium-resolution monochrome graphics and text appli-
cations. The card shown here is a relatively old, full-length model.

Fig. 3. The widely used EGA card affords good colour graphics and text capabilities at a
reasonable price.

tures can be displayed by means of
shades of grey. Similar dual-sync colour
monitors are aimed at users of PCs with
a CGA and/or EGA card. The EGA and
PGC also require their own monitor
type. Apart from a specific line and
raster frequency, some videocards supply
only digital or analogue signals. Poten-
tial buyers of a video card are, therefore,
well advised to take all these different
specifications into account before
deciding on a particular type. Always
remember the monitor!

A great effort is constantly being made
by monitor manufacturers to provide the
widest possible range of products to
meet the requirements of customers as
well as of the videocards they use.

Again, the multisync monitor
(monochrome as well as colour) is the
overall winner here, although VGA and
MCGA compatibility is not always
guaranteed.

Cables and plugs

Combining a videocard with an appro-
priate colour or monochrome monitor is
a problem that is even further com-
plicated by the cables and plugs needed
for each combination. Table 2 shows an
overview of connections for various type
of video card. It will be noted that the

MDA, CGA, EGA and PGC make use
of a 9-pin D-connector, while the new
cards, MCGA and VGA, need more
wires and work with a 15-pin connector.
In priciple, horizontal and vertical sync
signals are sent over separate wires; only
the PGC uses a combined sync line. For-
tunately, this card is of no significance to
today’s PC user. H

ELECTRONICS NEWS

New Status for NIC

The National Informatics Centre
(NIC), responsible for linking the
entire country with a computer net-
work status, be given a special
statusl, according to a high-level ex-
pert panel. NIC is now a cell of the
Planning Commission. It has been
set up at a cost of Rs. 180crores.

'The expert panel observed that most
of the government departments and
ministries are not using the facilities
created by NIC. Even the Department
of Electronics, which originally setup
the centre, does not make use of its
computer and other facilities. The

panel setup under Mr. P.S. Deodhar,
chairman of the Electronics Commis-
sion, submitted its report recendy.
The panel has opined that a number of
ministries like agriculture, home, sci-
ence and technology and others have
dedicated computer systems at dis-
trict level. If NIC is given a special
status and treated as a part of the gov-
ernment, the multiplicity’ of invest-
ment by various ministries can be av-
oided. The annaul budget of about Rs.
40 crores given to NIC will be suffi-
cient to meet all the data collection re-
quirements of the entire government
machinery.

NIC has so far entered into a
memorandum of understanding with
all the stats, excepting Nagaland. So
far, it has created the computer net-
work for 300 districts. Of these, 170
district data network systems are fully
operational and 130 are in the final
stage of testing. It has created infor-
mation system for 27 sectors such as

agriculture, scheduled castes and
tribes, family welfare, health, treasury
accounting, education and so on.

1 3.61

TWEETER PROTECTOR

by K. Baumotte

Fig. 1. Circuit diagram of the proposed tweeter protector. The resistance of the lamp increases
with rising input power. At the same lime, the 2N3055 short-circuits the tweeter during the
high power peaks. The circuit is suitable for all tweeter with a cross-over frequency ol 5 kHz.

Tweeters, the high-frequency drive units
in a loudspeaker system are often dam-
aged by a properly matched and rated
power amplifier. This happens because
many modern power amplifiers are
direct-coupled to the loudspeakers, i.e.,
they do not use a transformer. Such
amplifiers have the nasty property of
producing square-wave signals when
they are (even slightly) overdriven. The
ensuing harmonics lie chiefly in the fre-
quency range of the tweeter. This con-
siderable spectral shift of the audio
signal was, of course, not taken into
account during the design stages. This is
because during the standard (DIN)
testing of the loudspeaker system, the
tweeter is required to handle only 1% of
the total applied power. In other words,
when 100 W of audio power is applied,
to a loudspeaker system, the tweeter
needs to handle only 1 W. Even if the
tweeter is rated well above the standard
test specification requirements, during
loud music passages, when clipping oc-
curs (and square-wave signals are
generated) it may well have to handle too
much power. This may happen before
any distortion of the sound is heard.
Tone controls and equalizers can hasten
the demise of the tweeter: a 6-dB lift at
4 kHz doubles the power applied to the
tweeter, i.e., makes the unit twice as
vulnerable.

Protection circuit

A relatively simple circuit as shown in
Fig. 1 is all that is required to prevent
damage to the tweeter, especially if it is
frequently used at high volume levels. It
can not be a coincidence that most

leading suppliers of disco and public-
address systems fit a similar protection
circuit in their equipment.

A useful side effect of the circuit in
Fig. 1 is that the lamp used as a positive-
temperature coefficient resistor gives a
visible warning if the sound level is too
high. The lamp begins to glow when the
power reaches a certain level. When the
voltage drop across it reaches about
5.5 V, the lamp severely limits the level
of the applied signal. At the same time,
the transistor begins to conduct and
short-circuits the tweeter. If this situ-
ation is allowed to continue, the transis-
tor dissipates enough heat to warrant the
use of a small heat sink.

Distortion

Overloading the tweeter also causes
severe distortion: when the voltage at the
input to the circuit is 12 V, the level of

distortion is likely to be around 10%.
This may be improved very considerably
by the use of the circuit shown in Fig. 2.
Under the same conditions, the level of
distortion is only about 0.2%.

The time constant R3-C1 enables single
pulses to pass through the circuit
unhindered. Only when the overloading
continues do the darlington pair of tran-
sistors begin to conduct and short-circuit
the drive unit. The time constant may be
altered by changing the value of Cl up to
470 n?. Slight alterations in the values of
R2 and R3 allow the toggle time of T2 to
be set to individual needs.

Finally

Although the circuits in Fig. 1 and Fig. 2
are extremely useful, it should be noted
that they are not intended for use with
good hi-fi equipment: they are designed
primarily for use with disco and public-
address equipment.

NEW PRODUCTS

•Oscilloscope Multiplexer

Thurlby Electronic Ltd., UK, have in-
troduced the OM358 Oscilloscope Mul-
tiplexer, a self-contained instrument
which allows up to 8 channels to be dis-
played using just one trace of a conven-
tional oscilloscope. Typical application
areas included microprocessor-based
products, data transmission systems,
analog-digital converters, PLLs, fre-
quency dividers, etc.

M/s. MiniATE System • B-7, Pamposh
Enclave • New Delhi-1 10 048 • Tel:
6418470 •

Digital Temperature Indicator

Instrol’s digital temperature indicators,
single and multipoint types, are measure
temperature in ovens, oil/water baths
and fornaces of ceramic industries, pro-
cess temperatures of chemicals, fertiliser
plastic, cement, food-processing and sol-
vent industries. They are available in
portable or panel mounting forms with
LED/LCD digital displays and for temp-
erature range from ambient to 1999°C
with cold-junction compensation for K/
J/T thermocouples. Digital temperature
Indicators for RTDs are also available.
The accuracy is ± 1% or better and re-
solution is 0.1°C. Thermocouples and

RTD assemblies in standard design or as
per specifications are supplied with the
temperature indicators.

M/s. Instrol (India) Pvt. ,td. • A-37,
G.I.D.C. Electronics Estate • Gan-
dhinagar-382 016 (Gujarat) • Phone:
21551 •

Ultrasonic Hardness Tester

The Ultrasonic Hardness Tester model
8701 for metals working the principle of
ultrasonic contact impedance has a range
of applications due to its independence
of fixed testing locations and testing Of
hard-to-reach points, surfaces, installed
parts eg. teeth, bearings, shafts etc., in-
cluding testing on flat surfaces, curved
surfaces etc. The tester probe can be
used in any position horizontal, slanted,
vertical downwards or vertical upwards
and requiries no correction; reading can
be on huge parts or thin sheets.

The tester is simple to operate. By gently
applying the probe against the test ob-
ject, the hardness of the metal is indi-
cated on the meter. Anyone can measure
the metal hardness accurately with ease
and fast.

Industrial Estate • Saki Vihar Road •
Saki Naka • Bombay-400 072 •

Time Switch

GELCO Time Switch Type-101, based
on latest IC technology, gives more accu-
rate time for switching off the lights of
sign boards, display boards, hoardings^

etc. at set time interval. Pushbutton,
LED and time selection switch are pro-
vided. Time interval can be selected
from 1 to 8 hours. System voltage is 230
V, 50 Hz current capacity 10 A resistive.
The device measure 115 mm x 105 mm x
75 mm and weights 0.95 approx.

M/s. Gujarat Electronics & Controls • 9,
Advani Market • Outside Delhi Gate •
Ahmedabad -300 004 • Phone: 23117 •

Electronic Safety Guard

The Electronic safety Guard from Elec-
tro-Arts works on infrared techniques
and ensures safety in dangerous operat-
ing areas on automatic machines, power
presses, shears, bending machines, etc.,
It consists of a set of parallel beam in-
frared transmitter and receiver which
forms invisible curtain in between the
operator and the machine. When any of
the beam is interupted the same is sensed
and the machine is stopped immediately
preventing possibility of accident.
Various models are available covering a
height of 100 mm to 1000 mm and up to a
range of 4000 mm . The unit works on 220
VAC and is unaffected by vibration and
ambient light conditions. It is designed
for continuous use.

M/s. Electro-Arts • 4, Vaishali • Gan-
gapur Road • Nashik-422 005 • Ph:
0253-78452 •

NEW PRODUCTS

• In-Circuit Microprocessor
Emulator.

SIIMPLETECH Instruments offer a
stand-alone in-circuit microprocessor
Emulator for software debugging and
hardware integration. The CPU in the
circuit can be replaced with the emulator
for the various operations involved in
locating software errors and hardware
faults. The portable, table-top instru-
ment is suitable for field servicing, pro-
duction testing, and R&D testing. It is
totally transparent to the user. A feature
is that the CPU can run at its full speed.
Break-point allows real time program
execution at full speed to a present ad-
dress. When the breakpoint is encocun-
tered the emulator will enter the step
Mode in which the VPU status is dis-
played and makes it possible to examine
or alter memory locations, I/O ports and
registers. The user can step through and
locate the programme, error or circuit
faults. The DMA Read/Write, I/O
Read/Write, Register Read/Write fea-
tures help to see what is happening in the
CPU and in the circuit. Emulators for
8085 and Z 80 are available; emulators
suitable for other processors also can be
considered.

M/s. Simple Tech I-nstruments • M.E.S.
Road • Jallahalli • Bangalore-560 013.

FM Linear Director

THE AK5507B/AK5508B FM linear de-
tectors from Ando Electric- Co. of Japan
are for measuring the modulation
characteristics of FM mobile radio trans-
mitters. They are suitable for a range of
applications, including evaluation of
data or design and prototype analysis of
transmitters and signal generators, as
well as adjustments in the testing of such
equipment. The AK5507B’s frequency
range is from 7 to 520 MHz and that of

AK5508B is from 7 to 1000 MGz. Both
models can read peak voltage or peak-
to-ppeak/2 voltage. The indications ap-
pears in 3 digit decimal LED as well as on
the meter. A minimum range of ± 1 KHz
is provided for the modulation factor
meter to enable simple measurement of
low level modulation. The inductively
tuned local oscillator provides stable
measurement of S/N ratio even in the
presence of external noise and vibration.
The use of an external local oscillator al-
lows the enhancement of S/N measuring
accuracy by using a very stable crystal os-
cillator or other signal source. The
AK5508B features a residual AM mea-
surement function as well.

M/s. Murugappa Electronics Ltd. •
Agency Division • 299, Kamaraj Avenue
• Second Street • Advar • Madras-600
020. • Phone: 41 33 87.

Conductivity Meter.

A modified AC Wheatstone bridge cir-
cuit provides alongwith CMOS devices
an advancement to NCS 3000 series of
industrial conductivity meter from
NAINA. This is a low cost single range
electrical conductivity meter which can
be supplied with a range of 20 micro
siemens, 200 micro siemens, 2 mini
siemens or 20 mili siemens. The reading
is displayed on 3'h digit LED display of
12.5 mm or 20 mm height. Cell constant
compensation is the inbuilt feature how-
ever temperature compesation, 4-20mA
output, printer output, temperature in-
dication, audio-visual alarms with set
points are optionally available. A small

panel mounting cabinet of 96 mm X 96
mm size is used. Selection of dip type
cells, flow type cells of Glass, PVC or SS
304 can be made withg proper cell con-
stant. Equipment can be used by DM/DI
water supply units, boilers, electroplat-
ing units, swimming pools, fisheries
ponds, chemical industry, cooling to-
wers, pollution studies etc.

Naina Electronics P. Ltd. • 181/6, In-
dustrial Area • Chandigarh 160 002.

Automatic Coil-Winding
Machines

M/s. Tekma Kinomat of Italy have been
manufacturing coil-winding machines
since 25 years.

These machines are available in several
models, with or without CNC system,
and having a range of production for
wire dia. from 0.02 mm. to 1.5 mm. The
production of bimetallic coils too is car-
ried out with the relevant models. The
production capacity ranges from 600 to
2000 coils/hour, depending also on the
number of spirals.

The production from these machines has
application in various products and in-
dustries like : Electrical colis; Electroly-
tic capacitors, Resistors; Coils for
Magnetic Cards; Tape-Reading Heads;
Data Processors; Eelectronic Office
Machines; Quartz Watches; Dashboard
Instruments; Magnets for Starter
Motors; Bimetallic coils for Remote-
Controls and overload relays; T.V. and
Radio industry; Washing Machine and
Refrigerator motors and relays; Toy
motors etc. The Import of these
machines is under O.G.L. for Actual -
Users.

Precision International • D-7, Green
Park (Main) • New Delhi- 110 016.

NEW PRODUCTS

.Printed Circuit Board

Besides PCBs for almost every applica-
tion, including flexible PCBs, Grafica
offer PCBs with conductive silver/car-
bon for telephones, remote control
switches, etc. They propose to manufac-
ture touch membrances, keyboard
switches, conductive silicon rubber
keypads and double sided PTH PCBs.

M/s. Grafica Display Co. • 86,

Mathuradas Vasanji Road • Near Dar-
pan Cinema • Andheri (E) • Bombay-
400 093 •

DIGITAL TIMER

The PLA Compatible mounting and
quartz controlled 30 minutes timer for
maximum Demand indicator is available
with 1 C/O output contact, accuracy bet-
ter than 0.1% supply voltage 240/ VAC
and 110/ VAC.

M/s. Sai Electronics • (A Divn. of Starch
& Allied Industries) • Thakor Estate •
Kurla Kirol Road • Vidyavihar (West) •
Bombay-400 086 • Ph: 5136601/51 13094/
5113095 •

Precision Brass Parts

Khanchandani Industries manufacture
Precision brass parts as per specification
and/or drawings/samples. The brass
parts are used in electronics, au-
tomobile, electrical, instrument, compu-

M/s. Khanchandani Industries • 36,
shanti industrial estate • Sarojini naidu
road • Mulund West • Bombay -400 080

Digital Multimeter

The PLA DM-20 AR digital auto/man-
ual multimeter features: 4a digit display,
fast auto ranging audible continuity
tones, frequency/dB measurement facil-
ity (only in manual mode), and Auto bat-
tery test and auto polarity.

M/s. Pla Electro Appliances Pvt. Ltd. •
Thakor Estate • Kurla Kirol Road • Vid-
yavihar (West) • Bombay-400 086. •

Earth Leakage Circuit Breaker

GELCO offer the ELCB (Shock
Guard), a static relay that disconnects
the supply on the detection of lekage cur-
rent. On removal of the fault, current
relay can be reset manually. The device
has fixed leakage current 30mA (adjust-
able leakage current of 7m A-30 m A and
30 mA-100 mA available on request). It

is available in single phase, 16 A, 30 A,
and 60 A. The divice measures 170 x 165
x 70 mm.

M/s. Gujarat Electronics & Controls • 9,
Advani market • Outside Delhi Gate •
Ahmedabad-380 004.

Torque Gauge

Waters Manufacturing Inc. of U.S.A.
Offer the Torque Watch - Torque
Measuring Gauges in Three ranges -
Low, Medium, and High-from 0.2 g-cm
to 14 kg-cm. Torque measurement is
easy requiring no special set-up or tools.
The gauge is attached to the shaft and ro-
tated by hand to give direct readings on a
clear easy-to-read dial. It is useful as a
production or laboratory tool as well as
inspection instrument in application
such as micromotors, small clocks, com-
puter peripherals, small geared drives,
potentiometers, measuring instruments
industry, and research.

M/s. Universal Automation • 47, Mitra
Mandal • Pune 411 009 •

NEW PRODUCTS

Electronic Thermal Wire Stripper

Reliance Electronic have developes an
electronic thermal wire stripper to re-
move the insulation of PTFE, PVC and
any other insulation. The Device leaves
wire free of oxide, nicks or deformations
of any kind, high gauge wire. Diameter
of wire to be stripped should not be more
than 1.6 mm. However cutting of insula-
tion can be done up to 3 mm DIA. The
stripper can be used for single and mul-
tistand wiires.—

A pushbutton switch provided in the
handle enables the special alloy blades to
heat up to 1500°F in less than 10 seconds
allowing the current to pass instantene-
ously only. The compact solid-state
temperature control unit controls the
temperature of the blades ranging from
1000 F to 1500°F. The unit will operate
single phase, 230 V ± 10% with out put
of 3.5 V, 25 W.

LEG operated wire-stripper, electroni-
cally temperature controlled soldering
station, solder bath, soldering repairing
station and 1C remover are also availa-
ble.

M/s. M.R.K. & Brothers Engineers •
310-A, Commerce House • N.M. Road
• Fort • Bombay-400 023 •

Time Switch

The MIL 2008 Q series time switch is fit-
ted with a quartz Electronic drive control
and a step motor. The quartz frequency
is 4.19 million hertz; the quartz stabilisa-
tion ensures the exact running of the
driving mechanism. These time switches
are designed for the accurate control of
oil heating installations, electric heaters,
airconditioning plant, water processing
plant, streetlights, traffic signals, etc.
The MIL 2008 Q is available with contact
rating of 16 A, 250 VAC, and with daily

3.70 eleklo. India rmrch 1989

and weekly programme dial. Operating
on mains supply is, however, continues
to run for 150 hour after power failure on
a battery back-up.

M/s. Sai Electronics • (In association
with cupwud Arts) • Thakkore Estate •
Kurla Kirol Road • Vidyavihar (West) •
Bombay— 400 086 • Ph: 5136601/5113094

Digital Ohmmeter

ECONOMY offer a bench type instru-
ment for measuring the various resistors
in the manufacturing process of electri-
cal heaters, coils and relays. . The instru-
ment can also be used by electronic in-
dustries for inspection of proper values
of resistance.

Features include: 3a digit, 7 segment red
LED, 7 measurement ranges with lowest
range of 2 ohm with 1 millionohm resolu-
tion and highest range of 2 Mohm with
Rohm resolution: other ranges of 20
ohm, 200 ohm, 2 Kohm, 20 kohm, and
200 Kohm; 4 wire measurement; and Ac-
curacy ±0.1% of range ±0.1% of read-
ing ± 1 digit.

Cabiinet size: 240 mm depth x 193 mm
Weight 3 Kg.

M/s. Economy Electronics • 15, Sweet
Home • Plot No. 442 • 2nd floor •
Pitamber Lane • Off. Tulsi Pipe Road •
Mahim • Bombay-400 016 •

PCB TESTING SYSTEM

THE PCB System from Struers of Dan-
mark is a complete system for the prep-
aration of PCB cross sections for micros-
copic examination and testing. The sys-
tem comprises equipment for all prep-
aration steps from sampling and drilling
of reference holes to final grinding and
polishing. The object of the system is to
facilitate and rationalise control with a
single , but frequent source of failure : de-
fects in the through-plated holes in dou-
ble-sided and multilayer printed circuits
PCB System produces 36 or more speci-
mens automatically and reproducibly.
The specimens are sectioned with high
precision along the centre line of the
through-plated holes and mounted in the
sample holder with all the holes
positioned at the same level. The PCB
sampler is a routing/- drilling tool for
highly accurate and deformation-free
sectioning of samples from any given
area of a printed circuit board. At the
same time it provides the two reference
holes ensuring that the inspection holes
will be located at the same level in the
mount. The mounting device is designed
for convenient and troublefree insertion
of disposable positioning pins in the re-
ference holes up to 6 samples. The sam-
ples are automatically spaced evenly,
ready for embedding. The sample holder
fits struers Planopol/Pedemax and the
three abra machines, which give the user
a wide choice of grinding and polishing
equipment. Each sample is held by a clip
and may be removed individually for
microscope inspection between prepara-
tion steps. The illustration below shows
the marks A, B and C for adjustment of
the stop screws for coarse grinding , fine
grinding and polishing, respectively.
M/s. Mukund Iron & Steel Works Ltd •
L B Shastri Marg • Kurla • Bombay-400
070 • Phone:512 0180-88.

NEW PRODUCTS

Programmer Controller

JELTRON Model 814A Programmer
Controller is a self contained microp-
rocessor based set point programmer
and a single loop industrial controller
combined in one compact case. The
814A accepts directly process variable
inputs from TCs. RTDs, transmitter vol-
tages and currents and Optical Radiation
Temperature Detectors. All tempera-
ture inputs are linearized and direct
reading in degress F or C, switch selecta-
ble. All transmitter units are field con-
figurable in engineering units, from 999
to 9999 with full decimal point position-
ing. All program parameters such as
Range Limits, Set points. Ramp Times
and Soak Times are entered in full en-
gineering units. Times are user c.onfigur-
able from 0.0 to 99.9 hours or minutes,
the 814A is a single channel instrument
that can store a maximum of 30 seg-
ments. Individual programs can be fully
independent or linked. A loop instruc-
tion alows the segment or any combina-
tion of segments to be repeated upto 99
times. Complete program configuration
is stored in solid state non volatile
EAROM. The 814A is available in wide
choice of Control outputs. Reverse act-
ing or direct acting control action is
switch selectable. Integral Auto/Manual
Station standard with bumpless transfer
from auto manual. Remote/Local set
point operation is standard. The remote
set point input is automatically scaled to
the field configured range of the control-
ler. Optionally, RS-232Cor RS 422 com-
munication interface is availabe for
supervisory control applications. Set
Point, Control output and Configuration
of parameters are viewed using the main
display. Segment number as well as Set
Point are continuously displayed. Prog-
ram security is assured using a front key

Jeltron Instruments (India) Private
Limited. • 6-3-198/2, Road No. 1 • Ban-
jara Hills • Hyderabad-500 034.

Temperature
Indicator Controller

HOSHAKUN has developed Digital
Temperature with double thumbwheel
type Digital Controller. It is a rugged,
compact panel mountable instrument.
The bright red 12.5 mm LED Display en-.
ables one to read the temperature from a
long distance. Set temperature is all the
time visible on the front panel by the use
of thumbwheel switches. Broken sensor
protection and automatic cold junction
compensation is standard feature for
thermocouple input. Most of the as-
semblies used in this instrument are plug
in which offers simplicity in assembly as
well as dismantling for servicing pur-
pose. This instrument can be used for
furnaces having heaters in delta/star con-
nection such that during the start up the
furnace is heated with heaters in delta
form upto a certain temperature (set
low) and after than it gets connected into
star form up set high around which con-
trol action will take place. In other words
during start up, furnace is in “MORE
HEAT" mode and after exceeding first
control point it goes into “LESS HEAT"
mode which then controls the tempera-
ture of the furnace around the second
control point. This also can be used for
oil fired furnaces (MORE HEAT mode
equivalent to both main and pilot bur-
ners on and less heat mode equivalent to
only pilot burner on and main burner
off).

HOSHAKUN • Vivek Appartments •
Plot No. 15 • Tulshibagwale Colony •
Sahakarnagar No. 2 • PUNE-411 009.

Programmable
Batch Counter Controller

Micronix offers an ideal instrument for
applications where counting, controlling
and sequencing operations are involved.
The Unit is based on 8085 microproces-
sor with battery backed memory to re-
tain data during power failures.

The front panel consists of an interactive
4 digit display and hermatically sealed
membrane keyboard for entry of
parameters. The display is used to indi-
cate event count as well as Jobs done de-
pending on the selected mode. It also in-
dicates te system status on LED indi-
cators. The number of counts per sequ-
ence is 999 and no. of sequences is 8 both
expandable as per user request. The unit
accepts a variety of inputs such as mic-
roswitch , optical or proximity switch etc.
It provides a change over contact for
controller action.

The unit works on 230V AC and is
housed in DIN standard enclosure suita-
ble for panel mounting or bench top
models.

Micromix • D-74, Angol Industrial Es-
tate • Udyambag • Belgaum-590 008.
Karnataka State.

New Open-Type Terminal
Connectors

“IEC" has introduced a new range of
Open Type Terminal Connectors - TBM
Series. They are presently available in 12
ways, 10 ways and 8 ways. The Connec-
tors are rated at 15 Amps, 250 V AC with
a insulation resistance of more than 1000
Mohms and can withstand H.V. test of
2000 V for 1 minute. The Terminals are
of Brass with Nickel plating and the
housing of electrical grade bakelite or
melamine on order.

M/s. Asia Electric Company • Katara'
Mansion • 132A, DR. A. Besant Road •
Worli Naka • Bombay- 400 018.

NEW PRODUCTS

Desoldering Station

Zevac Auslotsysteme GMBH, (WEST
GERMANY) manufacturers desolder-
ing tools, desoldering station and
machines, Zevac Tools are sturdy and
designed for high efficiency, ALS desol-
dering stations consist of vacuum unit
desoldering iron and stand, controlled
by a foot-switch with 220/24 Volts-sup-
piy. PVSG-60 is a desoldering system:
with an integrated vacuum transducer
mounted on the soldering iron handle
with a finger operated control to enable
complete single handed operation with a
supply of 220 or 24 Volts. Zevac Desol-
dering machines are used for desoldering
standard integrated components, mul-
tilead connectors as well as chips,
capacitors, flatpacks etc.

M/s. Arun Electronics Pvt. Ltd. • 2 E,
Court Chambers • 35 New Marine Lines
• Bombay 40 020. Tel: 252160/259207.

10-ohm range for in circuit resistance
measurements etc. It can measure DC/
AC voltage up to 1000 V/750 V, DC/AC
current up to 15 A, resistance upto 20
megohm and audible continuity check by
buzzer. It is housed in compact, portable
high impact plastic case with till stand for
table top as well as field applications.

Ledtron Electronic • 170 Lohar Chawl.
• Bombay- 400 002.

Preset Counters

Micronix offers a range of Digital preset
counters for counting and controlling ap-
plications. The counters find applica-
tions in Machine tools, Pharmaceuticals
and food processing industries, Press op-
erations, process control panels, plastic
and rubber moulding industries, au-
tomobile industries etc.

These counters use CMOS technology.
Modular construction makes servicing
Very easy. The presetting is done
through a set of Thumbwheel switches.
The actual count is displayed on a 0.5
Inch seven segment LED display. It ac-
cepts a variety of input sensors such as
proximity switch, microswitch or optical
sensors.

It gives a set of chageover contacts for
control applications. These counters are

housed in DIN standard panel-mounta-
ble enclosures. Working voltages are
user selectable.

M/s. Micronix • D-74, Angol Industrial
Estate, • Udyam bag • Belgaum • Kar-
nataka- 590 008.

Temperature Controller

The ARTECH series 101 Temperature
Controller is a simple, rugged and accu-
rate instrument which can find applica-
tions in all area of Temperature Control.
The unit is fully solid staate, selectivity
using integrated Circuits. Th€ circuitry
can withstand normal machine vibration
without adverse effects. Automatic cold
junction protection, lead resistance com-
pensation and open sensor protection
are standard features. The dimensions
are suited for flush panel mounting in a
92 x 92 mm cutout as per DIN 43700.

Artech Labs • A-3 Udyog Sadan No. 3
• Central Road • M I D C • Andheri
(East) • Bombay 400 093.

Digital Multimeter

Ledtron Electronics has introduced au-
toranging Digital Multimeter Model DT
860A using latest American LSI
Technology. Its unique features are au-
toranging/manual mode, memory for re-
lative ' measurements, transistor HFE/
diode check, datahold, razor sharp LCD
with automatic range/function display.

3.72

HAVE YOU MISSED OUR

ANNUAL

NUMBER?

The September 88

CIRCUITS SPECIAL

With Over Hundred Exciting Circuits

Special Offer

Send a 3 years subscription

ToeleNto

and get this Annual-88

FREE

This offer is valid tor payment
received between 1st March 1989
to 3.1st August 1989

or Send Rs 16/- by M.O. only to:
eIeIcTOR ElECI RONiCS PVT lid.
52-C Proctor Road, Bombay-400 007

RN No 39881/83

Chemswab
IPA Saturated
Foam Swab

Flux-Off®
C0 2 Propelled
No Scrubbing

Freez-It* -65T
Variable Control
Tracks Intermittents

Konform'
UL Listed
UV Detectable

MH BY WEST - 228
IIC No 9 1

Ultrajet ' Dusters
Precision Filtered
to < 0.2 Microns

Gold Guard"'
Preserves Precious
Metal Contacts

Chempad" Chem-Wik® Lite Twillwipes”'
IPA Saturated Desolders Textured Cotton

Cleaning Pad 50% Faster Lint Free

Performance you can count on.

The electronic service industry’s most popular chemicals and cleaning products for over 30 years.
Call or write today for a free catalog, samples or technical assistance.

Manufactured by:

Chemtronics

NEW YORK. USA

Accra Pac (India) Pvt. Ltd.

917. Raheja Chambers. Nariman Point. Bombay-400 021
Tele*: 011-5771 APIB IN.