elektor

84

April 1982
65p.

up-to-date electronics for lab and leisure

" KEYBOARD!

T - v - audio output.

SAlE SOUND;

®ASIC for the Junior

elektor april 1982 - 4-03

Selektor

Mini EPROM card

This is an elegant, low cost solution for Junior Computer owners seeking a
suitable RAM extension to accommodate the Junior BASIC, or large as-
sembler.

100 W amplifier 4-18

A design following a well-beaten, reliable track, without compromising
on output power or performance, 100 W into 4 fi. A versatile amplifier to
suit everybody.

Tuning aid

(S. Akkal)

Readers interested in music, and musicians in particular, should find this
article interesting and useful. Tuning an instrument can sometimes be a
chore. The circuit described provides a worthwhile solution using only
3 ICs.

T.V. sound interface

A straightforward circuit to enhance the sound of your television set
safely.

Dynamic RAM card

Dynamic RAMs are so economical these days that it is worthwhile using
them in place of static RAMs.

TECHNICAL EDITORIAL STAFF |
J. Barendrecht

E. Krempelsauar
G. Nachbar
A. Nachtmann

When is an OTA notan OTA?

Recently, an new improved version of the OTA, the 13600, was an-
nounced. In this article we take a close look at this new circuit.

2114 RAM tester

A highly popular 1C now used in many computers, is the 21 14 RAM. The
circuit presented in this article allows readers to test each circuit, simply
and quickly with a go or no go indication.

Connection tester

An excellent aid to examine the quality of solder joints and connections in
any electronic circuit. The tester indicates a 'good' connection by emitting
an acoustic signal, which makes testing quicker and easier.

Toroidal transformers

The article discusses the excellent electrical qualities and advantages of the
toroidal transformer over the conventional types.

Capacitive keyboard

The choice and price range of available computer keyboards are nearly
infinite. The use of 'mechanical keys', although simple, is relatively expens-
ive. An economical alternative is introduced here; a capacitive, touch
activated keyboard, which has proved reliable in practice.

The 13600, a new OTA

The article deals with several practical applications for this useful 1C.

:or-

4-40 } f

M

i KcvaoARa

BASIC on the Junior Computer 4-52

It is about time, the Junior Computer extended its linguistic ability to
include BASIC. This article introduces and explains how to implement the
new facility.

selektor

Holographies and crime

Civil and government authorities in the
United States are getting seriously
concerned with the criminally abusive
use of the very latest technological
advance in opto electronics. A recently
published report states that the realis-
ation of holographic projections (HP)
is creating havoc in law enforcement
agencies throughout the USA. It would
appear that the legal aspect regarding a
holographically projected image has yet
to be defined. This state of affairs has
been brought to light by the felonous
activities in the use of HP with regard to
fraud.

Several large companies involved in the
sale of second-hand cars initially used
HP for the purpose of defrauding the
Inland Revenue Service. In order to
qualify incorrect sales figures on tax
returns the companies used HP to pro-
ject the images of a far larger stock of
vehicles than actually existed in their
showrooms. However, one car distribu-
tor extended the use of HP a little too
far with the result that its entire stock
consisted of vehicles that did not in
fact actually exist. This action went
undetected until an unfortunate buyer
purchased a vehicle that was an HP.
The driver took the car home and found
in the morning to his horror that the
car had disappeared. Unknown to the
car sales company, the current state-of-
the-art in holographic projection can
only maintain an image for about
20 hours.

It is known that prominent figures in
criminal organisations are using HP in
order to confirm their appearance in
public when they are likely to be
suspected of felonous activities else-
where. Certain unscrupulous politicians
have also been known to maintain an
HP image of themselves at social and
public functions. When questioned
they maintained that this directly
effected a gain in popularity. The
authorities fear that this use of HP may
spread into vice, armed robbery and
bodily assault.

One personnel officer in a multinational
corporation has expressed concern at
the high rate of possible absenteeism at
senior executive level. The legal position
of a holographically projected image of
an employee is rather uncertain with
regard to pay and working conditions.
This has far-reaching effect when acci-
dent insurance claims are submitted for
an injury sustained to an HP image on
employers' premises.

The disappearance of large amounts of
cash from a number of clearing banks
in the Mid West is now being attributed

to the use of HP. In view of this the FBI
have been assigned the task of verifying
the existence of the Federal gold reserve
at Fort Knox. Fears are increasing for
the long-term stability of the American
economy.

Real estate has not been overlooked in
the illegal use of HP. In a recent court
case in Redwood City L.A., a fraudulent
financier was indicted for selling a chain
of hotels that did not exist. Apparently
the buyer discovered the plot when the
particular hotel in which he was staying
disappeared during a local power failure.
The missing persons branch of the FBI
discovered another fraudulent use of HP
when attempting to trace a number of
families who apparently disappeared
while on holiday. The common factor
in a large proportion of the cases was a
travel agency. Moonlight Travel, which,
following investigations, was found not
to exist. It is thought that the company
used holographic projections of offices,
aircraft, staff and holiday resorts to fly
unsuspecting holidaymakers to Jackson
Heights, a suburb of New York.

The worry expressed by the authorities
cannot be understated since it has been
suggested in some quarters that the
Statue of Liberty in New York harbour
is an HP. The original was reputedly
sold to a wealthy Arab at some time
during 1968 in order to aid the failing
economy of the City of New York.
However, HP also has its positive side
since sources suggest that the US
government have been experimenting
with HP power stations for the past two
years. It is hoped that this may be an
answer to the energy crisis.

So far the abusive uses of HP have not
reached the UK to any great degree.
A spokesman for Scotland Yard
speaking from the Metropolitan Police
Headquarters in Hyde Park stated that,
■Reports that certain members of the
conservative party are engaged in H.P.
have not been substantiated by our
staff'.

APR Inc. USA (S 758 .

alaktor april 1982 - 4-13

Flawless micro-machining by
beams of ions

Ions, several thousand times as heavy as
free electrons, cause a great deal of
damage on an atomic scale when they
are accelerated by high electric potential
into a beam striking the surface of a
workpiece. In this way they can be used
to knock off fragments, atom by atom,
in a highly precise machining or etching
process. Machining tools based on the
ion-beam technique are now being used
by the micro-electronics manufacturing
industry, where it has great potential for
the production of chips with very high
packing densities. Ion beams are capable
of cutting a neat groove as little as one
micrometre deep, with a repeatable
precision of the order called for when,
as will almost certainly happen in the
not-too-distant future, a million 'bits'
may be incorporated on a single minute
slice of semiconductor material.

Dr. Roy Clampitt, of Oxford Applied
Research at Witney, near Oxford,
believes that industry could find a large
number of diverse applications for the
technique. His company, formed three
years ago, provides an independent
centre for research and is able to call
upon the vast knowledge and expertise
to be found in the Oxford area. Estab-
lished applications of the ion-beam
technology already include cutting fine
lines on optical flass for use in spectro-
photometers, and in smoothing titanium
steel for joints in replacement-part
surgery. The picture shows an engineer
making final adjustments to ion-beam
equipment prior to its despatch from
Witney.

Spectrum No. 175

(757 S)

4-14 - eleklor april 1982

The silent noise of a gas turbine

The low-pitched rumble from gas tur-
bines can be felt as much as heard over a
radius of one kilometre or more from
aircraft engine test beds and similar
installations. It is a background noise
which many people find objectionable
and even distressing when continuous,
and hitherto it has proved almost
impossible to suppress. Now, in what is
believed to be the first successful appli-
cation of electronic 'active sound con-
trol' to this kind of problem, noise
from one such machine has been cut to
one-twentieth of its original level. Such
vibration-cancelling devices promise to
become a common feature of machines
and vehicles of the future.

All who have examined high-fidelity
music systems know that it is the low-
frequency, bass sounds that are most
difficult and expensive to control. To
reproduce them faithfully calls for large
loudspeakers, which give out a sound
that penetrates the walls and ceilings of
houses and is heard far away as a
booming or rumbling noise; the treble
elements lack such persistence and are
easily absorbed.

So it is, too, with the noise of powerful
machinery. Though the high-frequency
elements are easily muffled, suppressing
the bass rumble is extremely difficult
and costly, for it means investing in
massive acoustic enclosures. The gas
turbine, our most compact source of
power, is also man's noisiest machine.
Little wonder that it is on the gas
turbine that most noise-control devel-
opment is concentrated and that the
first full-scale demonstration of anti-
noise techniques is now reported.

Noise is a vibration of the air, the
pattern of aerial vibrations travelling in
waves from source to receiver and
beyond. And different noises can be
superposed without distortion or loss.
Incoherent sounds do not interfere with
the background chatter at a cocktail
party and in no way deform any par-
ticular conversation; only raised voices
are heard - but the tonal quality and
information in the successive undu-
lations of the sound wave are not in the
least distorted by the different level.
Nor would two sounds, one with peaks
and troughs corresponding exactly to
the other's troughs and peaks, interfere.
But the combination of the two would
be silence, for one would be the nega-
tive of the other. That is the principle of
anti-sound, a principle that has pro-
gressed from its science-fiction base.

F igure 1 . Blockdiagram of the ges-turbine compressor plant that has bean silenced by

frequency noise by anti-sound from actively-
controlled loudspeakers.

through numerous small-scale and lab-
oratory experiments to this full-scale
application on an 11 -MW gas turbine.
The UK National Research Development
Corporation (NRDC) has stimulated the
growth of this novel method of noise
control for more than ten years. It has
sponsored projects to perfect the tech-
nique and bring it out of the laboratory
to a practical application. The technique
calls for fast electronic signal processing
and precise superposition of the noise
and anti-noise fields, so it is most easily
applied to the low-frequency, long-
wavelength elements of noise, the sort
that cause the window-rattling rumble
of large engines. Because it is this very
rumble that has always been most diffi-
cult to suppress, the new technique is a
welcome complement to existing muf-
flers, which work best at high fre-
quencies.

Two years ago, laboratory demon-
strations had reached the stage where
the time was ripe for field trials. A
worth-while trial to suppress loud
rumble would mean running powerful
machines, for they alone are capable of
generating rumble at a really annoying
level. The British Gas Corporation
agreed to cooperate by allowing one of
their compressors powered by a Rolls-
Royce Avon engine to be fitted with
'anti-sound', and the job was taken on

by Topexpress, a small, high-technology
research company based close to Cam-
bridge University. Within six months
it was demonstrated that control was
feasible, and the anti-noise suppressor
was built, fitted and switched on within
two years. It instantly silenced a rumble
that would otherwise be heard more
than a mile away — and annihilated it
everywhere. NRDC's £300000 invest-
ment in the principle had brought a
new hush to the landbased gas turbine;
it would take more than ten silenced
units to make as much noise as the
standard, unsilenced one. Moreover, the
silencing carries no performance penalty
and is attained at less than half the cost
of conventional muffling.

Gas turbines operate by compressing air
which is then heated at high pressure by
burning gas in it. The hot products of
combustion flow through the turbine,
which extracts most of the energy. The
engine uses that energy to drive its own
compressor, leaving a balance over for
useful work. With the Avon, the balance
is 1 1 MW. After flowing through the
turbine the hot, spent gases are ex-
hausted through a stack, which they
enter in a highly turbulent and noise
state. The airflows are large and the
exhaust is eventually vented to the
atmosphere at a speed of some 50 m /
sec and a temperature of 200°C from a
duct with a diameter of three metres.
The inlet to the engine is carefully
treated to prevent noise escaping from
the site, and so is the exhaust, but that
poses a much more challenging prob-
lem. It is not easy to absorb and prevent
the escape of noise when one end of the
pipe is a gaping three-metre hole! In
practice the noise control is extremely
good, but inevitably the low-frequency
rumble escapes. That rumble is a broad-
band sound in the frequency range 20
to 50 Hz, with the peak of the spectrum

at just over 20 Hz. Of course, this noise
too can be controlled by conventional 3
means, but only by constructing massive
extensions to the exhaust stack, an
expensive operation that makes the 1
installation more visually obtrusive. The *
suppression of the rumble without any ^
modification to the stack was the task sj
set to the T opexpress team. “j

The first step was to monitor the sound
with advanced, highly accurate equip-
ment to find out its level and waveform,
how it varied from time to time and
point to point and to discover in par-
ticular whether a control signal could be
obtained to drive the active control
gear. The entire technique depends on
obtaining enough information about the
precise characteristics of the noise and
activating the anti-noise sources in time
to prevent the sound escaping from the
chimney. The Topexpress team had to
be sure both that the control technol-
ogy existed and that the acoustical
equipment was good enough for the job
before deciding to embark on manu-
facture. It was — but there was little
room for manoeuvre. Amplifiers would
have to handle 12kW of peak power,
enough to drive 72 of the loudest bass
loudspeakers known to discos.

The diameter of the exhaust stack is
about half the wavelength of the
highest-frequency elements to be con-
trolled and is just small enough for the
noise field to have a relatively simple
geometry. A good control signal was
obtained by using. four microphones.

FREQUENCY (Hz)

FREQUENCY (Hz)

FREQUENCY (Hz)

The loudspeakers were tightly packed
around the stack exit and early tests
which recorded signals showed that the
audio system could accurately mimic
the noise. Conditions were expected to
change when the gas turbine was started
up and the exhaust duct filled with
moving, turbulent, hot gas, but the
change was anticipated satisfactorily.
So, the team has reached the critical
stage in producing a control system.

It would be necessary to accept the
signals from the four microphones and
generate an electrical input to the loud-
speakers that would induce them to
cancel exactly the exhaust noise. There
are powerful mathematical and compu-

4-16 - elektor gpril 1982

tational techniques for establishing the
optimal control strategy in that type of
problem, and they were used to the full.
Some of the controllers examined in the
design process turned out to be un-
stable. If they had been installed, the
loudspeakers would have generated
noise that the microphones would have
heard, causing the controller to drive
the speakers harder. They would then
have made more noise, and so on to an
acoustical catastrophe!

First test

Other control strategies, though effec-
tive in the appropriate frequency range
caused extra noise at frequencies out-
side the control band; they would have
produced an unacceptable emphasis of
high-frequency noise as the price for
low-frequency suppression. The team
examined many controllers systemati-
cally before finding the ideal solution.
Once they knew what they had to do,
they turned to the task of constructing
the electronics system and programming
the microprocessor to become a com-
pact, packaged controller with the
characteristics they wanted. In the end
they had a small 'black box' that could
produce exactly the right signal for the
72 speakers when it was fed from the
four microphones. The system was
complete and ready for test. Their first
test took place at Duxford, near
Cambridge, on 27 January 1981 . At the
flick of a switch that activated the con-
troller, the gas turbine's rumble dis-
appeared.

Noise levels are measured in decibels, a
logarithmic scale that compactly rep-
resents the enormous range of com-
mon sounds. More than one million of
the sort of sounds that are barely
audible must be superimposed to equal
the noise level at which the ear begins to
feel pain. The difference between the
two is 130 dB, a number that is a lot
easier to manage. And, because the
pitch or note of a sound is such a
distinguishing feature, it is natural to
represent a noise in terms of its spec-
trum, the strength of each constituent
note. Gas turbine exhaust noise covers a
wide and continuous spectrum and the
active silencer attained a reduction of
about 10 dB across the lowest octave,
with a peak attenuation of 13 dB, which
represents a factor' of 20. In other words
it would have taken 20 silenced com-
pressors, all working simultaneously,
to make as much noise as the unsilenced

Work is going on to develop the system
to even better standards. Meanwhile, it
will be possible to duplicate this anti-
noise system at less than half the cost of
silencing a gas turbine's rumble by
conventional means. Moreover, the new
silencer is not visible from the outside
and has no detrimental effect on the
engine's performance, which cannot be
said of the conventional techniques that
call for massive enclosures and partial
throttling of the engine's exhaust.

This means that gas turbine rumble can
now be cured economically. The same
principles apply to the noise of air
conditioning systems, which are far
simpler to deal with using this anti-noise
technology. Simplest of all are the sort
of periodic, repetitive sounds that we
hear from diesel engines, for example;
there is no fundamental reason why
diesel exhaust noise should not be
eliminated by anti-sound. Other forms
of vibration and waves can also be
tackled on the same basis and it is
likely that active vibration-cancelling
devices will become a common feature
of machines and vehicles where smooth
operation is important. Quiet havens
could be built — there is no obvious end
to the development that is possible. We
can look forward to a future quiet that
will be shattered only by a power cut!
Professor J.E. Ft owes Williams,

Rank Professor of Engineering
(Acoustics),

Cambridge University
Spectrum, No. 175

New reading aid

As 1981, the Year of the Disabled, drew
to a close, ideas to make life easier for
the handicapped were only just begin-
ning to dawn. It is interesting to note
the favourable repercussions that the
campaign has had on the manufacturing
industry, the design described here
being just one of many devoted to
helping the partially sighted read and
thus providing them with better com-
munication facilities.

The electronic reading aid incorporating
the latest in optical technology has been
developed by engineers at Wormald
International, a New Zealand company
specialising in aids for the disabled.

The aid, called Viewscan, is portable
and can be used by the partially sighted
at work, school or home. It is designed
to produce a bright, magnified image of
reading material when the material is
scanned by a hand-held camera.

A 0.5 W micro-miniature bulb illumi-
nates the page via a high resolution fibre
optic ribbon in the solid state camera.
The image is transmitted to a photo
diode array and the signal from this is
fed directly to two high speed micro-
processors in the display unit. A buffer
memory then reassembles the infor-
mation into moving lines of print.

The display screen, developed specially
for Viewscan, is a flat neon matrix
panel. Special attention was paid to
colour. The orange shade for the display
has been selected for reasons of maxi-
mum visibility and clarity and can be
switched easily from a positive to a
negative image. Other display colours
were extensively researched but orange
was finally chosen because it can also
provide higher levels of brightness than
any other alternatives. Magnification is
achieved electronically and the eight
settings from X4 to X64 are easily
selected by means of a switch on the
front panel.

Viewscan is efficient and economical
in power consumption so that a port-
able version can be powered using
rechargeable NiCad batteries.

Wormald International

mini EPROM card

a miniature memory

extension for

the Junior Computer

This is an elegant, low-cost
solution for Junior Computer
owners seeking a suitable RAM
extension (to accommodate the
Junior BASIC or a large assembler,
for instance). At the same time,
the board takes up a minimum of
space.

Book 3 in the series on the Junior
Computer explained how to fetch the
three NMI, RES and IRQ vectors from
EPROM connected to the bus board
(and page FF). Appendix 3 gave an
example of this using a 82S23 PROM.
A 2716 EPROM, however, provides a
better solution, as it is much more
straightforward to program (see the
'EPROM programmer' in the January
issue, E-81). The miniature board
described here is designed to cater for
2k of EPROM.

The circuit diagram for the miniature
EPROM card is shown in figure 1. The
EPROM stored in IC2 is addressed by
way of IC1. The memory range com-
prises addresses $F800 . . . $FFFF. Two
wire links enable the range to be ac -
cessed either by way of pin OE (O utput
Enable ) or by way of pin 51 (Chip
Enable ). On the one hand the Output
Enable method speeds up EPROM
operation, but consumes a fair amount
of stand-by c urrent. On the other hand,
Chip Enable addressing saves up to
300% current and is slightly slower.
Readers may choose either method,
provided they remember to ground the
enable pin and connect the other to the
output of IC1. This is carried out with
the aid of the links shown in the circuit
diagram.

Finally, the following data must be
programmed into the final six locations
in the SF800. . .$FFFF memory range:
address SFFFA should store 2F;
address SFFFB should store 1 F;
address SFFFC should store ID;
address SFFFD should store 1C;
address SFFFE should store 32;
address SFFFF should store IF.

The remaining 2018 memory locations
are entirely at the disposal of the
operator. H

Parts List Semiconductors:

IC1 = 74LS30
IC2 " 2716

Capacitors: Miscellaneous:

Cl ,C2 = 1 p/16 V tantalum 1 64-pin male connector ('a' and ’c'l 4161 2

elektor april 1982

HM)\V |H«V(T

amplifier

a welcome boost to stereo

This 100 W power amplifier design follows a well-beaten, reliable track,
without compromising the output power or distortion. It can be used
as a power amplifier, or slave, and will deliver 100 W into a 4 fi load.

Table 1

Technical specifications:
output power:
(continuous sinewave sign

input impedance:
output impedance:

current consumption:
transistor quiescent currei

100 W (R|_ * 4 fl, k = 0.1%)

70W(R L = 8fi,k = 0.1%)

< 10 Hz. . . >20 kHz at 100 W.

<10 Hz... >100 kHz (—3 dBI
<0.1% at 20 Hz . . . 20 kHz and 100 W
0.28% measured at 40 Hz and at 10 kHz;

70 dB 3

0.775 V at full modulation
100 kfl

0.052 n (at 1 kHz)

4fi

80 V symmetrical (+40 V. 0. -40 V)
2.25 A max. where R L = 4 n
50 mA

Output power amplifiers (or slaves as
they are known in the music world) are
not only fun to build, but can be
adapted to suit anyone's personal
requirements. With the aid of numerous
interesting suggestions made by readers,
a set of parameters for the 'ideal' power
amplifier were drawn up:

• It must deliver 100W (into 4 £2 load).

• The distortion must not exceed 0.1%
at 100W (even at 20kHzl).

• The power bandwidth should be ex-
tensive.

• It must be 'short' proof, for the pro-
tection of the output transistors.

• The power supply must be symmetri-
cal, so that no electrolytic capacitors
are needed in the output stage.

• Only standard, easily available com-
ponents should be used.

• Construction and calibration must be
straightforward.

• The amplifier should be economically
viable, and reliable.

Some readers are bound to think that to
build an amplifier that complies with
most (if not all) of the above parameters
is practically impossible. However, they
should reserve judgement until they
have taken a closer look at this com-
pact circuit, designed around modern
darlington transistors.

How does this particular circuit com-
pare to other Elektor amplifiers
published in the past? Well, the highly
popular EQUA and EQUIN amplifiers
were not intended to deliver anything
like as much power. The Elektornado
can only manage 100W when it is con-
structed in the form of a bridge circuit.
At the other extreme we have the 'disco
power amplifier' which was described in
the January 1981 issue. This does
provide 200 W (into 4 S2), but suffers
from a restricted (low) frequency re-
sponse and has the disadvantage that
electrolytic capacitors need to be used
in the output stage. Taking everything
into consideration the need arose for an
inexpensive, good quality, medium
rated power amplifier design.

The circuit

Figure 1 shows the complete amplifier
circuit diagram. The input consists of a
discrete differential amplifier, built up
around transistors T1 and T2. This is
followed by the driver stage T4, the
collector of which is connected to
transistor T3. This acts as an 'adjustable
zener diode' and sets the quiescent
current level. A fully complementary
output stage (using darlington transis-
tors T7 and T8) is connected to the
driver. One advantage of using a sym-
metrical power supply is that the mid-
point between T7 and T8 has a zero
voltage potential, avoiding the need for
an electrolytic capacitor at the output.
The amplifier has a fairly high input
impedance of 100 kO, since C4 is
'bootstrapped' by R2; the input im-
pedance of T1 is also quite high, of

2

Negative feedback (both DC and AC) is
fed from the output to the base of T2
(the other input of the differential
amplifier) by way of R6. The DC
component of the negative feedback
keeps the output at a zero voltage
potential. The AC feedback determines
the gain as set by R6, (C4) and R3.
Using the values in the circuit (figure 1)
the gain can be seen to be:

U 0

U?

R3 + R6 _ 3420
R3 = 120

28. 5x

The driver T4 is coupled to the collector
of T1. T4 boosts the signal; it must be
capable of supplying the output transis-
tors (T7 and T8) with sufficient base
current. Fortunately, the darlington
transistors (T7 and T8) require very
little base current, since they have a
high current gain. As a result T4 dissi-
pates a negligible amount of heat, and
therefore does not need to be 'cooled'.
Transistor T3 determines, and resistors
R18 and R19 stabilise the quiescent
current passing through the elements of
the output stage. The voltage drop
across the emitter resistors R18 and
R19 is determined by the setting of PI,
since this sets the collector-emitter volt-

age of T3. Resistor R1 1 is 'bootstrapped'
by C5, which increases the AC im-
pedance of R11, and so boosts the gain
of the driver.

In the heart of the output stage are the
'darlington' transistors BDX66 and
BDX67. The B-series (as used in this
circuit) have the following character-
istics at a case temperature of approxi-
mately 25°C:

• maximum collector-emitter voltage =
100V

• peak collector current = 16 A

• continuous device dissipation = 150W
Not bad! At a collector current of 10 A
the collector-emitter saturation voltage
is 2 V and the DC gain lOOOx. At a
collector current of 5 A, the gain is
4000x, with a saturation voltage of
between 0.4 and 0.5 V. These character-
istics make the BDX66 and BDX67
ideal for this type of circuit!

No matter how 'sturdy' the output
transistors are, they still need protection
against overload. The voltage drop
across the emitter resistors R18 and

R19 is a measure of the output current.
This means it can be used for current
limiting. When the voltage drop across
R18 and R19 is sufficiently high it will
turn on T5 and T6 by way of the base
voltage dividers R 16/R14 and R17/R15.
T5 and T6 will draw current via D2 and
D3, thus limiting the base drive to T7
and T8.

The various little capacitors each serve
an important purpose. Cl limits the
input bandwidth which, as any audio
enthusiast will tell you, is a good
idea . . . . C3 causes the frequency
response to roll off — it is 3 dB down
at 100 kHz. C6, C7 and C8 ('Miller
Capacitors') and R 20 and C9 at the out-
put all help to stabilise the circuit.
C10/R21, Cl 2, C11/R22 and C13 serve
to suppress spikes and RF from the
power supply.

Technical specification

The circuit is a straightforward, reliable
design, and very few difficulties should

be experienced during construction.

With a bit of luck the amplifier will
deliver 120W into a 4ft load, but
unfortunately the distortion is approxi-
mately 1%. At 100W (again into 4 S2),
the distortion is less than 0.1%. This
power rating and distortion factor is
more suitable for HiFi applications ....
Table 1 shows the technical specifi-
cations of the amplifier. As figure 5
illustrates, the distortion factor remains
virtually constant over the complete
frequency range from 40 Hz to 20 kHz.
It is less than 0.1% all the wayl
For full modulation a minimum input
of 0.775 V is required. Most modern
preamplifiers supply approximately this
output level. When using equipment
that can produce a higher output, it is
advisable to add a 10 k preset poten-
tiometer at the input of the power
amplifier.

The power supply

It is only a slight exaggeration to say

that a power amplifier is only as good as
its power supply. For this amplifier, a
nominal supply voltage of ± 40 V is
required. At full drive, 2.25 A is needed
(100 W into 4(2) - or 1,1 A for 70W
into 8(2. For reasons of cost and
simplicity, an unstabilised supply is
normally used. By its very nature, if this
delivers ±40V at full load, it will run
up to a higher voltage as the current
demand is reduced. However, the out-
put devices are only 'safe' up to 1 00 V -
corresponding to ±50V. To allow a
safety margin, an off-load supply volt-
age of ± 46 V is preferred. This leaves
only 6 V lee-way between off-load and
full drive (over 2 A), so that a very low-
impedance power supply is a must.
A good way to achieve this is by using
a good quality transformer . , .

Given a good mains transformer, a
bridge rectifier and a few electrolytics
are all that is needed to complete the
power supply circuit. Fuses in both
supply lines are a good idea, since the
current limiting circuits in the output

stage cannot guarantee 'permanent short
circuit’ protection. They only increase
the chances of survival until the fuses
'blow'!

Two units are necessary for stereo,
because the power supply will only
cater for one power amplifier.

Construction

The printed circuit board is shown in
figure 6. The emitter resistors R18 and
R19, should be mounted at least 5 mm
'off' the board. This will allow good
ventilation, so that heat can be dissi-
pated freely.

The output transistors (T7 and T8) and

the capacitors C7 and C8 are mounted
'off-board'. As shown in figures 3 and 4,
each transistor (T7 and T8)' and its
capacitor (C7 and C8) should be
mounted on a separate 1.2°C/W heat
sink, such as a black SK84 (see photo).
Alternatively, if heat conductive paste is
applied to both sides of the mica
washer, then a 1.8°C/W, black SK03
(100 mm) type should be sufficient. It
is worthwhile noting that, when several
transistors are mounted on a single heat-
sink, the thermal resistance must be
divided by the total number of transis-
tors. Therefore if T7 and T8 (with or
without heat conductive paste) are on

one heatsink, it must be either a
0.6°C/W (SK84) or a 0.9°C/W (SK03)
type.

Under no circumstances should the case
or the connections of the power transis-
tors come into contact with the heat
sinks, as this would cause a 'short'. Be
reminded that the case of the transistor
is the collector. As a matter of interest,
insulating caps for power transistors are
available.

When connecting C7 and C8 as shown in
figure 4, ensure that their leads are in-
sulated. The connections to the printed
circuit board are made with thin copper
wire and should be kept as short as
possible. It is common knowledge that
DIN sockets are unreliable, so the use of
good Japanese HiFi spring terminal
blocks is advised for the output.

Use screened AF cable to connect the
input socket to the printed circuit board
(see figure 4). The most convenient
point to connect the ground of the
amplifier and earth of the case, is the
ground connection of the input socket.
The cable and the input socket should
be mounted as far away as possible from
the other components and wiring. This
is necessary, in order to reduce the

supply, and the other two to the re-
maining connections. Should OV be Figure 7 .The copper treck pattern and component layout of the power supply board,
registered, then the wires should be
swapped over.

The electrolytic capacitors C2 and C3 ,h e fuse F2 from the power supply cir- 50 mA quiescent current should then be
should be securely attached to the cu j t Then set a multimeter to the flowing through the output transistors,
board by means of plastic cable ia DC current measurement range and This completes the setting-up pro-
fasteners. This will prevent the terminal connect it to the terminals of the fuse cedure. Replace the fuse F2 (switch off
leads from snapping off (see photo). holder (+ terminal of the meter, to C2 / the supply voltage first!). If there are
The connection wires between the fuse junction). any problems, readers will be able to

printed circuit board and the loud- Turn PI fully anti-clockwise. Check all track down the error quickly by corn-
speaker socket or terminal are laid along the connections and switch on the paring their voltage readings to the
the sides of the case and away from mains supply. The multimeter needle DC voltage values indicated in the
everything else, to reduce the chance of should hover around 0A. If a higher circuit diagram. In the prototype all the
feedback. current is indicated switch off immedi- voltages were measured with the input

ately, as there must be something shorted out and the circuit connected to
wrong! Provided only a few milliamps a loudspeaker.

Setting up the amplifier are flowing through the circuit, the

Make sure there is nothing connected to meter may be reset to the 100 mA range
the output, 'short' the input and remove and PI adjusted to give 80 mA. About

ril 1982

electronic tuning aid

This article will be of special
interest to those readers who 1

enjoy music, particularly
musicians. Tuning any instrument
quickly and efficiently can
sometimes be a problem, at the
very least, it can be laborious.

This article provides a quick and
easy method with a circuit using
the minimum of components, in
fact, just three ICs. The use of
digital technology ensures that
simplicity is not at the expense
of accuracy. The circuit easily
lends itself to be modified to suit
any particular purpose.

S. Akkal

Simple is beautiful it is said. In this case
it may not be exactly beautiful, but this
tuning aid will provide what a great
many musicians have been looking for.

There are two main problems associated
with tone generators — which is all that
a tuning fork really is. The first is that
of stability. It is obvious that the instru-
ment being tuned can only be as accu-
rate as the tuning source and therefore
the circuit must produce the same F# in
a month's time as it does today. A large
number of components increases this
problem. of an octave without any need for All that remains of the circuit are the

The second difficulty arises when the external components. gates N3 and N4 and the surrounding

tuning source provides a number of A 1 MHz crystal will provide a fre- components. The two gates are connec-

notes. The relationship between them quency at pin 16 of 2092.0502 Hz for ted in parallel and act as buffers for the

must of course be fixed and they must C 8 . It is worth considering the com- output. Potentiometer PI is used to

also remain stable. A look at the circuit plexity of a circuit required to do this adjust the output level,

diagram in figure 1 will show that the little job before the appearance of LSI For precise calibration a frequency

number of components could hardly be ICs. counter is needed. This is connected to

any less. All the available tones are It is a simple matter to select any one the output of N2 and Cl is adjusted

derived from a 'master' oscillator (or of the outputs with the aid of a switch, for a reading of 1.00012MHz, How-

frequency generator). This is formed SI in this case. However, we still finish ever, in reality the difference between

by two gates, N1 and N2, and is crystal up with just any one note of one octave, this figure and 1 MHz is so small that it

controlled for greater accuracy. This This may satisfy many requirements, can be ignored. Although originally

effectively takes care of the stability but it would be very useful to be able intended as a tuning aid, the circuit can

and long term accuracy problems. The to select any one of the number of have many other uses. For specific

use of a crystal keeps drift to an absol- octaves as well. purposes one or even both switches may

ute minimum. The oscillator frequency Fortunately this can be accomplished be dispensed with, and one or a few

can be trimmed by means of the variable quite simply. The wiper of switch SI is specific tones can be 'hard wired'. For

capacitorCI. fed directly to the clock input of a instance guitarists will require E, A, D,

The oscillator frequency is fed to pin 1 7 stage counter, IC2. The 7 outputs of G, B and E. A six way switch with

of the master tone generator 1C 1. This this 1C provide us with any one of seven 'tappings' at the right outputs will

will provide the complete set of 1 3 notes outputs - seven octaves in fact. provide this very easily. H

id interface

elektor april 1982 - 4-25

Some readers may be fortunate enough
to possess one of the newer TV sets
already equipped with headphone and /
or tape recorder sockets. If that is the
case, the connection of a screened lead
to the auxiliary input of a stereo system
is sufficient. This will allow JR Ewing's
intrigues and dulcet tones to achieve a
reasonable HiFi quality. If no improve-
ment in listening pleasure results, then
the script writers are probably to blame.
Readers who are ardent followers of
heavy metal or other kinds of loud
head-bashing music would probably
make good use of a headphone socket
if it is made available on the TV set.
Irrespective of the quality, members of
the family and certainly some neigh-
bours may complain about the quantity.
TV viewers may question the need for a

TV sound
interface

In certain countries TV sets capable of producing 'stereo' sound are
fast becoming available. The majority of TV sets already in home use
(there are exceptions) produce a sound which may require
improvement. With this in mind our designers found one way of solving
the problem. They have produced a simple, relatively low cost device
which helps to enhance the original sound. By doing so, it is hoped that
even the most critical of viewers will be pleased.

circuit. To many of them a straight-
forward method is to connect a socket
in parallel with the speaker. At first
sight this may seem plausible. But
unfortunately, as the chassis of a TV
set is normally connected to the mains
supply, the application of such a
solution would be hazardous to the
health of tape recorders or HiFi equip-

The fitting of an isolation transformer
can prove to be rather costly (especially
if fitted by an expert). Another dis-
advantage is that the use of a trans-
former would add noise and distortion

to the output signal.

An electronic solution like the one
described here would seem to be a
worthwhile compromise: By means of
an optocoupler, a safe separation is
achieved, between the TV set and any
external equipment to be used. From an
electronic point of view, figure 1 shows
the TV sound adapter to be straight-
forward, two separate symmetrical
power supplies and an optocoupler
complete the circuit.

Transistor T1 is the input stage to
trigger the internal LED of the opto-
coupler. The current through the LED is
set at a 'quiescent' value of 18 mA via
R1...R3 and D1. In this way the
optocoupler is made to function within
its linear transmission range. With an
input voltage level of 1 Vpp, the LED
current will fluctuate between 16 and
20 mA. In contrast to the cheaper,
slower types, the optocoupler used,
contains a photo diode: This is a good
way to transmit high frequencies
without incurring too many problems.
The transistor in IC3 together with T2
form the output stage with negative
feedback via R6 and R7. The base bias
for T2 is set with R4. The value of
resistor R5 is selected so that the output
stage can be driven to near maximum,
thereby ensuring the circuit is not over-
loaded when the output voltage of the
TV final stage is applied to the input.
The amplitude drop for the high fre-
quencies is determined by C6, in other
words, this capacitor limits the trans-
mission bandwidth. Any reader wishing
to do so may experiment with this value
in order to achieve a preferred fre-
quency characteristics. Amplification of
the TV sound is set at a gain of 1 .
Although the use of two power supplies
may seem a little extravagant, they are
found to be necessary, because the
input of the circuit must be separated
from the output. Two mains trans-
formers or one transformer with two
completely separated secondary wind-
ings are therefore required. Finally, to
achieve the object of the exercise, a
good quality transformer should be
used.

Construction and application

The layout for the printed circuit board
of the TV sound interface is shown in
figure 2. With the exception of the
transformer, the mains switch and fuse,
all components will fit on the board.
Before mounting the circuit into the TV
set, a few checks to ensure correct con-
struction should be made. Prior to
mounting the optocoupler measure the
operating voltage. Mount the opto-
coupler. Connect an audio signal to the
input, and check that the output signal
is of the same value: The circuit is now
ready and can be inserted into the TV

The circuit as described has an input
impedance of 1 kfi and was specifically
designed for connection in parallel to a

1

Resistors:
R1 = 2k7
R2 = 2k2
R3 = 270 «
R4 = 4k7
R5- 1k8
R6 = 18 k
R7 = 220 n

Capacitors:

Cl, C2 = 220 p/25 V
C3,C4 = lOOn MKT
C5,C7- 10 m/ 1 6 V
C6 - 2n2 MKT

Semiconductors:

B1.B2- B40C500

D1 ■= 1N4148

T1 ,T2 = BC 547B

IC1.IC2- 78L12

IC3 = 6N135 (Hewlett-Packard)

Miscellaneous:

SI = dp mains switch

FI - 1 00 mA slow miniature fuse

Tr = 2 x 1 2 V/50 mA mains transformer

TV loudspeaker. If 220 k resistors are
substituted for R1 and R2 and a BC 51 7
transistor for T1, an input impedance
of 100 kfl is achieved. Together with
the further addition of a potentiometer
(50 k log) to the output (to vary the
output level), the circuit becomes quite
versatile.

With a few minor modifications (if
necessary), it can be applied to any
situation where isolation from the mains
is required. It can be applied easily
and successfully to AF circuits to
control strobes and optical/lighting
effects.

In practice there are two ways of
connecting the circuit to the TV. The
first is shown in figure 3a, in parallel
to the speaker (Rj_). When switched off
(by means of S2) the connection is
made across R (R = R|_).

Most loudspeakers have their resistance
values indicated on the chassis. Should
no markings be found, then a TV circuit
diagram would be useful. Fortunately
most loudspeakers fitted fall into the
4 to 16 n region, but occasionally a
high impedance of 25 S2 or more may
be found. Should this be the case, then
the TV volume control will have to be
set low to counteract the rather high
output level. The problem of the
control setting, although not a major
one, could prove to be inconvenient.
A simple remedy is to connect a voltage
divider or preset before the input of the
circuit.

The second possibility (figure 3b) is to
eliminate the audio power amplification
stage of the TV altogether, by connec-
ting the circuit directly to the de-
modulator output. In other words the
'live' and earthed terminals of the TV
volume potentiometer are connected to
the input and ground of the circuit.
Normally, setting the volume control
to zero will silence the TV loudspeaker,
but obviously if it continues to emit
any unwanted sound then it can be
switched off by using S2 (see figure 3b).
One side effect of connecting the circuit
to the demodulator output is that the
TV volume potentiometer will now not
be able to control the level of the
external speaker. As there are many
variations in TV circuit design, a brief
study of the TV circuit diagram (if
possible) would be advisable, certainly
before contemplating any further
modifications.

With the use of a multimeter it will not
be possible to ascertain accurately
whether the earth (ground) of a TV set
is connected to the mains, because
irrespective of how unsophisticated the
TV power supply may be, it will always
contain a few diodes and rectifiers. The
only sure way to find out is to take a
close look at the circuit diagram. But
as a generale rule; if the TV set has no
headphone/tape-recorder or auxiliary
speaker socket then it should be
assumed that the ground is connected
to the mains supply. H

1-28 - elektor april 1982

Static and dynamic RAMs are playing
an increasingly important role in home
computers these days. Data stored in
static memory can be preserved for
relatively long periods, (providing the
power supply is not switched off).
Where dynamic RAMs are concerned,
long-term storage is a little more com-
plicated, as all the data has to be
renewed ('refreshed') at regular intervals
to prevent it from being lost Now that
prices have dropped dramatically, dy-
namic RAMs are gaining the upper hand.

dynamic RAM card

and an FET switch. The voltage across
the capacitor determines whether the
data contents are 'high' or 'low'. A dy-
namic RAM memory cell takes up far
less room than its static counterpart.
Theoretically speaking, this would mean
that the former is able to provide a
much greater memory capacity on an
identical chip surface area. But there is
more to dynamic RAMs than meets the
eye, for one of their main disadvantages
is that a great deal more is involved in
making a memory cell fully operative

dynamic RAM card

16 Kin 8 ICs

Dynamic RAMs are so economical
these days that it is worthwhile to
use them instead of static RAMs,
despite the additional control
electronics required. Eight ICs can
store up to 16 K and still leave
plenty of room on the Eurocard
for the control logic. Further
advantages include low current
consumption and high-speed
access times. Computer owners
who are running out of memory
and space will welcome this
opportunity to extend their RAM
facilities.

1

even though static RAMs have always
been preferred in the past.

Dynamic vs static
Static RAMs have an advantage in that
they are very easy to operate. The re-
quired circuitry is already incorporated
inside the 1C, so very few external com-
ponents are called for. Life is also made
easier for the operator by the fact that
no timing problems are involved as long
as 1C types are selected for the right
speed to cope with the application in

A static RAM memory cell consists of a
sort of set/reset flipflop, which contains
at least 5 or 6 transistors. As readers can
imagine, a complete RAM 1C has an
immensely complex structure.

A dynamic RAM, on the other hand, is
based on capacitances rather than flip-
flops. Each cell consists of a capacitor

2

than a capacitor and an FET switch.

As a result of a slight leakage current in
each capacitive unit, the voltage level
across the capacitor slowly drops in
value. Thus, in order to prevent the data
stored in the capacitors from being lost,
their charge must be refreshed from
time to time. This calls for an additional
control circuit and very precise timing
for the operation to pass off smoothly.
That is not the only problem. An awful
lot of memory cells can be integrated on
a single chip and so addressing is rather
complicated. Dynamic RAM manufac-
turers have tried to solve this by using a
multiplexed address bus, (yet another
addition to the circuitry).

Nevertheless, dynamic RAMs are so
cheap nowadays, (compared to their
static rivals), that even the extra com-
ponents required do not affect the over-
all cost. Although they consume very

dynamic RAM ■

ril 1982 - 4-29

little current, the dynamic RAMs used
here do need three separate supply volt- 3

ages.

All things considered, if the same results
can be obtained for less money, there is
no reason why readers shouldn't use
dynamic RAMs.

The structure and operation of a
dynamic RAM chip

The design described in this article is
centred around the inexpensive
4116 1C, which is available from various
manufacturers. The 1C encompasses
16384 x 1 bits. 8 ICs therefore provide
an 8 bit wide 16K memory. The 1 C
series has access times, ranging from
150 .. . 300 ns according to the figure
indicated after the type number.

The 4116 memory is arranged in an
array of 128 columns and 128 rows
(128x128=16,384). To decode 1 of
the 16,384 cell locations within the
41 16, 14 address bits are required, seven
per column and seven per row. An inte-
grated clock, the Row Address Strobe
(RAS) latches the 7 row address bits
into the chip and a second clo ck, the
Column Address Strobe (CAS), sub-
sequently latches the 7 column address
bits into the chip. In other words, the
7 address inputs are multiplexed. The
pin assignment for the 4116 is s hown
in figure 7. A negative pulse at the RAS
input 'reads in' data in the form of a
row address into the a ddres s inputs and
a negative pulse at the CAS input reads
in the data as a column address. As the
memory is only one bit wide, only one
data input and one data output are re-
quired (Din an d D ou t)- The logic state
of the WRITE input determines whether
a bit is to be read out or written in. The
remaining four pins constitute the
supply connections: Vqo. v CC. v BB
and Vss (+12 V, +5 V, —5 V and 0 V,
respectively).

To come back to the internal structure
of the 1C, 128 sense amplifiers are
situated in the middle of the 128 rows
with the task of topping up the capaci-
tors during a 'refresh cycle'. In addition,
they transfer data to and from the
memory locations. A sense amplifier is
a flipflop, each input of which is con-
nected to half a column. Each column
has its own sense amplifier which
detects the charge passing through an
addressed row and amplifies the signal
produced. The boosted signal is a full
logic level, either 'high' or 'low' and is
fed back to the column line, causing the
original (amplified) logic level to be
restored in the capacitor. The sense
amplifier now contains the same data as
the read (and immediately rewritten)
capacitor. Thus, as soon as the row is
accessed, all the logic levels stored in the
capacitors belonging to that row are
refreshed. To give you an idea of the
capacitance level involved: a 4116
storage capacitor has a value of about
0.04 pF I

The order in which the different signals
have to be applied is as follows:

Figure 3. Time sequence charts for reading, writing and refreshing the 4116. No specific times
are given here, because they depend on the speed of the host processor. All times are in the
nanosecond range.

consumption rate for th« various RAS and CAS
is fairly low, although brief peaks of around 100 mA

Data is read out of a memory location,
a seven bit address being stored at the
address inputs befo reha nd. Then a pulse
is generated at the RAS input. The row
address must be available for a certain
amount of time, after which the seven
bit column address can be prod uced.
This is followed by a pulse at the CAS
input. The column address must also be
present for a certain minimum period.
An internal output buffer then sends
the logic level of the selected address bit
to the da ta outp ut. During this pro-
cedure the WRITE input must be high.
Virtually the same principle applies to
write operations, only now the data
input is initia lly prov ided with a logic
level and the WRITE input goes low.
The time sequence chart in figure 3
illustrates these events in the form of

The refresh cycle

As mentioned earlier, using capacitors
to store digital information has a num-
ber of advantages, but there is also
another side to the coin. Slowly but
surely, this type of capacitor loses its
charge and the stored logic level. This
is why it is necessary to refresh the level
from time to time. In the case of the
41 16 this must be done every two milli-
seconds, which is quite reasonable,
considering its low cell capacitance.
Fortunately, the refresh operation is
relatively straightforward, thanks to the
structure of the 1C, in which the 'sense'
amplifiers are situated in the direct
vicinity of the cells. The sense ampli-
fiers, as has already been seen, boost the
logic levels in the memory cells. When a
row address is read in after the com-
puter has generated a RAS pulse the
entire row of 128 bits is read into the
sense amplifiers. At the same time, the
logic levels are amplified and written
back into the 128 row capacitances. In
other words, once a row address and an
RAS pulse have been produced, the
128 bits are refreshed. As long as this
method ensures that the whole oper-
ation is executed within 2 milliseconds,
data stored inside the 1C will remain
intact.

Of course, the refresh cycle may be
shorter than 2 milliseconds, if necessary.
This particular RAM card was designed
to be used with the Junior Computer, or
a similar microprocessor, with a clock
frequency of 1 MHz. This means the
refresh cycle for 128 rows takes 128 ps.

The timing

Figure 3 contains the time sequence
charts for the read, write and refresh
cycles, respectively. The diagrams
clearly show the order in which the
various signals must be provided.
Different times are involved and this
will have to be taken into account. No
specific values are indicated, as they
vary somewhat per 1C type and manu-
facturer.

The power supply

Special attention should be paid to the
power supply of the dynamic RAM
card. The average current consumption
rate for the three supply voltages is
fairly low. The highest pe ak is reached
upon either edge of the RAS and/or
CAS input. An example of this is shown
in figure 4, where relatively high current
peaks occur during the rising and falling
edges of the signal. Up to 100 mA may
be attained (per 1C)!

Obviously, this calls for certain protec-
tive measures. Rather than provide the
power supply with a high current capa-
bility it is best to buffer the power
supply by placing capacitors around the
RAMs.

The circuit diagram

Figure 5 shows the complete circuit
diagram of a 16K dynamic RAM card.
IC12...IC19 constitute the 16K x
8 bit dynamic memory. The data inputs
of the ICs are directly connected to the
data pins of the connector (on the left-
hand side of the drawing). The data out-
puts are connected to the data lines by
way of tri-state buffers. Address lines
A0 . . . A13 are linked to IC9and IC10,
which each contain four multiplexers
(with two inputs and one output). These
multiplex the fourteen address lines in
two groups of seven. The address lines
are linked to the address inputs of the
RAMs by way of the tri-state buffers
Nil . . . N17.

Address lines A12. . . A15 are connec-
ted to the address decoder IC11. This
enables data to be stored in any address
range by mounting wire links between
the outputs of IC11 and gates N27 and
N28.

IC6 serves to refresh the memory blocks
regularly, as it acts as a seven-bit coun-
ter. The outputs of the 1C are also
linked to the address inputs of the
RAMs by way of tri-state buffers
(N20. . . N26). The refresh cycle takes
place during the period that the pro-
cessor is not using the address bus. The
clock input of the counter IC6 and the

elektor april 1982 — 4-31

control inputs of the tri-state buffers,
N20 . . . N26, are connected to the
clock 01 of the processor system by
way of gates N1 . . . N3. During a cer-
tain part of the clock signal, memory
is not accessed. The wire links shown
are needed if the circuit is used with the
Junior Computer. For the sake of
clarity we will describe the rest of the
circuit diagram with reference to the
Junior Computer and then explain how
it may be modified for use with other
microprocessor systems.

In the case of the Junior Computer,
memory is not accessed with the
positive-going transition of 01 and so
this can be used to refresh the stored
information. The pulse diagram in
figure 6 illustrates this. During each
positive edge of 01 the contents of the
counter are incremented by one. Buffers
Nil . . . N17 are disabled, as they are
controlled by the output of N2 (which
is inverted with respect to the output
of N3). The buffers N20 . . . N26 then
send an address to the address inputs of
the RAMs. A delay is enforced, with the
aid of MMV1 and MMV2, to allow a
negat ive pulse to be provided at the
RAS inputs of the RAMs shortly after
the rising edge of the clock signal. That
is sufficient to refresh a complete row.
Since one row is refreshed per positive-
going clock pulse, the counter is reset
after 128 clock periods. After this
period all the rows will have been
refreshed. Thus, a full refresh cycle lasts
128 (at a clock frequency of 1 MHz).
Addresses are read in and out on the
negative edge of the 01 clock. This re-
quires a certain amount of 'timing logic'
with carefully calculated values (in
nanoseconds) to be sure that the posi-
tive and negative edges reach the RAMs
(and the multiplexer) in the right order.
Three pulse 'delays' consisting of
N4 . . . N10, R1 . . . R3 and C3 . . . C5
are included for the purpose.

When an address is accessed in the RAM
address range, the output of NAND gate
N29 is pulled low by the address
decoder. The clock signal is then sent to
N7 and N9 by way of N31 , which is also
connected to 01 . The falling edge of the
clock is delayed by the R1/C3 combi-
nation and is fed to the RAS inputs via
a couple of gates (see figure 6). This
means that the first seven address bits
are read into the RAMs. After this, the
multiplexer must be activated, which is
achieved by delaying the falling RAS
edge through R3 and C5. Once the fol-
lowing seven address bits have been
accessed, a fal ling edge may be pro-
duced at the CAS inputs. The latter
edge is derived from the falling clock
edge by way of the R2/C4 delay unit.
The wfe inputs are directly linked to
the corresponding connector pin.

That covers the main signals. A couple
of gates and a flipflop are shown in the
top left-hand corner of the circuit dia-
gram. These simply serve to adapt the
various signals to make them 'digestible'
for processors other than the 6502.

dynamic RAM card

irapril 1982 - 4-33

R1 ... R3 = 2700
R4.R5 - 2k2
R6 . . . R8 ■ 390 O

Capacitors:

Cl = 33 p
C2>- 100 p
C3 - 68 p
C4 = 470 p
C5 = 1 20 p

C6 . . . C21 - 1 p/16 V tantalum

Semiconductors:

IC1.IC2 « 74LS14

IC3 = 74LS221

IC4,IC5,IC20 - 74LS244

IC6 = 74LS393

IC7 = 74LS08

IC8 - 74LS32

IC9,IC10 = 74LS157

IC11 = 74LS154

IC12. . . IC19 = 4116 (260 ns)

IC21 = 74LS15

IC22 - 74LS74 (see text)

Miscellaneous:

1 x 64-pin DIN 41612. male connector

Construction

The printed circuit board for the
dynamic RAM card is shown in figure 7.
Since timing is very important, care
must be taken when mounting the
components. Readers are advised to
abide by the indicated component
values, especially with regard to the
resistors and capacitors. The easiest
method is to use the Elektor printed
circuit board.

How the wire links are positioned de-
pends on which processor is in use.
Table 1 shows the requirements for the
Junior Computer, the Z80 and the
8085. IC22 can be omitted for the 6502
and the Z80, as flipflop FF1 has also
been left out. As far as the 8085 is con-
cerned, things are different again.
Unlike the Z80, the 8085 does not
produce a refresh signal. Instead, this is
generated by S0, SI and I NT A (which
indicate the opcode fetch status).
During the period that the processor
needs to detect the code, the RAM is
not being used and so a refresh cycle
may take place. In the 8085 (multi-
plexed) addresses are accessed by way
of a data bus. Since the dynamic RAM
card is only suitable for a non-mu Iti-
plexed bus, however, the data bus will
have to be demultiplexed elsewhere in
the 8085 system.

The connections between points V, W,
X and Y and the outputs of IC1 1 define
the address range. Each output of the
1C represents an address range of 4K.
The memory array is shown in table 2.
A total of 16K therefore requires four
outputs of IC11 to be linked to points
V . . . Y. This enables data to be stored
in blocks of 4K practically anywhere
within the memory range.

Operators must remember one import-
ant aspect: the same code may not be
used twice for A12 and A13 (see the
last column in table 2), because the two
address lines are both linked to the
address inputs of the RAMs. This means
that the wire links must be mounted in
such a manner that the following combi-
nations of A13 are stored in consecutive
memory blocks:

A13 A12

0 0

0 1

1 0

1 1

It can easily be deduced which combi-
nations are feasible. A valid combination
would, for instance, be blocks 8000,
9000, A000 and B000. But 0000, 4000,
8000 and C000 are totally out of the
question, because A13 and A12 would
be 00 for all blocks.

If the card is used in combination with
the Junior Computer, the required
supply voltages will already be available.
The other processor systems will have to
produce the required voltages using
integrated voltage regulators. Plenty of
power supplies meeting the require-
ments can be found in previous Elektor
issues.

Testing the circuit
Before connecting up the supply volt-
ages, it is a good idea to check all the
solder joints thoroughly. Then the card
may be plugged into the bus of the /iP
system.

It doesn't really matter in which order
the supply voltages are connected,
although the manufacturer recommends
constructors to start with the — 5V line.
This provides an extra safety margin in
the event of an overload (which is
unlikely to happen if a good power
supply is used).

If all is well the memory should func-
tion normally as soon as the power
supply is switched on. As the memory
locations are invisible to the naked eye,
the best way to test the system is to
read data in and out and compare the
results. A special test program has been
written for the purpose and is shown in
table 3. This can also be used to test
other types of RAM. Once the program
has been entered, the start address and
the end address of the memory range
being tested must be stored at locations
0000 (= ADL) and 0001 (= ADH) and
at locations 0002 (= ADL) and 0003
(= ADH), respectively. The program is
then initialised at address 0004 and 00
is written into the memory range. The
program checks whether 00 is in fact
stored at the first address of the range-
under-test. If so, 01, 02. 04, 08, 10. 20,
40 and 80 are written into the address
in succession and read out again at once.
As a result, every bit in the address will
have been high once. Subsequently, FF
is stored at this address to track down
any addressing errors. For if there is an
found in a different address. The mis-

take is detected when FF is read out
somewhere along the line.

The above procedure is applied to every
single address until the program reaches
the end of the test range. Then the
entire test program is repeated (it also
tests the operator's patience!) starting
with the storage of 00s. This time the
range is examined back to front. Again,
this is necessary to be able to trace any
addressing errors that might have
cropped up.

If everything passed off without a hitch,
address 0000 will appear on the display
at the end of the program, followed by
the low order address byte of the
entered start address. If on the other
hand an error was detected, the address
at which it was found is shown on the
display together with its (erroneous)
contents. Restart the program at address
000A in order to carry on with the test.

Tab)# 1

6502 Z-80 8085

A-B 2-2’ 2-2'

C-D J2 3-3'

E-F J3 4-4'

G-H J4 5-5'

J8 J5 J1

J9 J6 J2

J9 J4

IC22 is omitted J6

IC22 is omitted J9

J10

Table 1 . This indicates which links are
required on the printed circuit board when
using the 6502, the Z80 or the 8085.

4 Kbyte-block
. . 0FFF

15 A14A13 A12

1000

2000

3000

4000

5000

6000

C 000
D000
E000
F000

2FFF

3FFF

4FFF

5FFF

6FFF

7FFF

8FFF

9FFF

AFFF

BFFF

CFFF

DFFF

EFFF

FFFF

Table 2. The address range can be defined by
linking the outputs of IC5 to points V, W, X
and Y. Each connection provides 4K bytes,
so that four connections are needed for a total
of 16K.

OTA i

OTA?

ril 1982 -4-35

The difference between an OTA and a
normal opamp can be summed up in a
few words. An opamp is voltage-driven:
the differential input voltage is mul-
tiplied by a fixed gain (100,000 times,
or so), so that a much larger voltage
appears at the output. In other words:
it's a voltage amplifier with fixed
gain.

The input to an OTA is also a differen-
tial voltage, but the output is a current.

when is an OTA
not an OTA?

obvious reasons, there is no feedback
around this circuit — the gain must be
set by lABC. not by the values in a
feedback loop!

Applications are not restricted to 'pure'
audio, as illustrated in figure 2. This is
an oscillator, with triangular and square-
wave outputs. The output frequency is
voltage controlled, over an extremely
wide range: 2 Hz to 200 kHz! This
corresponds to control currents from
10 nA to 1 mA. This is by no means
the only alternative application for
OTAs: we have seen them used (and
used them!) in AM modulators, mul-
tipliers, true RMS converters, automatic
level controls, voltage-controlled re-
sistors, filters, sine-wave generators,
timer circuits, phase-locked loops,
sample-and-hold circuits, logarithmic
amplifiers, and so on . . . The new DNR
circuit also uses an OTA, to construct
a filter with a variable cut-off fre-
quency.

. . . when it's a 13600!

Many readers will be familiar with the 3080 and 3094 Operational
Transconductance Amplifiers, or OTAs. Since their introduction
by RCA in the early seventies, this type of device has been used
extensively in the most varied applications. Recently, a new and
improved version has been announced — the 13600. It includes
linearizing diodes at the input, to allow for higher input levels and
a greater linear control range (over six decades!), as well as controlled
impedance buffers at the output.

In this article, we will take a closer look at the device. The conclusion
is surprising: when used in the 'ideal' circuit arrangement, this OTA
isn't an OTA!

This means that the relationship be-
tween input and output signals is not a
simple (voltage) gain: it is the 'forward
transconductance', g m , expressed in
'mho' or mA/V. The output current can
be converted back into a voltage by the
simple expedient of passing it through a
load resistor, R|_. This leaves us with
a voltage amplifier, with a gain of
gm x Rl. This is not so spectacular,
until you discover that g m can be
controlled by a DC bias current dABC),
over an extremely wide range. In other
words, an OTA is a voltage-driven cur-
rent source (or, with an external load
resistor: a voltage amplifier) with a
'gain' that can be varied over a wide
range by means of a control current.

As we have seen in recent years, this
leads to a wide range of interesting
applications. Take figure 1, for example:
a voltage-controlled volume control!
The control voltage, Uq, determines the
bias current lABC- The higher the vol-
tage, the higher the overall gain; when
Uc is zero, the output is also zero. For

The 13600

The 13600 contains two current-con-
trolled transconductance amplifiers,
each with differential inputs, linearizing
diodes and controlled output buffers.
The internal circuit for one OTA is
given in figure 3.

At the input, T4 and T5 are a straight-
forward differential amplifier ('long-
tailed pair'). The current source in the
'tail' (T1, T2 and D1) is actually a cur-
rent 'mirror': the collector current of
T2 is equal to the bias current lABC- At
this point, we are faced with the choice:
dive deeply into the theory or skip
it? . . . We will attempt a compromise.
For small differential input signals, the
collector currents of T4 and T5 (l 4 and
l s ) are almost identical; together, they
are equal to the sum of their emitter
currents, so each is equal to approxi-
mately ’/4IABC- At the same time, the
ratio between these two currents is
determined by the differential input volt-
age. For small signals, it can be shown
that the difference between the collec-
tor currents, l 5 -l 4 , is equal to the
input voltage times the bias current
(Ujn x lABC), multiplied by a constant
factor:

Uin x lABC = K (l s -l 4 ).

So far, so good. The next step is to add
three more current mirrors (T6-T7-D4,
T10-T1 1-D6 and T8-T9-D5), in such a
way that the collector current of T1 1 is
equal to l s and that of T9 is equal to l 4 .
This means that the output current is
equal to the difference between l 5 and

Uj n x >ABC= K* l 0 ut-

In other words, l 0 ut over Uj n (the
transconductance) is equal to I ABC over
K. It is determined by the bias current!
And that is precisely why an OTA is

OTA i

OTA?

Bktor april 1982-4-37

comes in: as stated above, it ensures
that the current setting of T12 is
determined by lABC- Neat! T12 and
T13 now provide an almost ideal buffer
stage between the high-impedance out-
put of the OTA and a following (low-
impedance) input.

Add a few diodes . . .

In most applications, OTAs are used
without feedback. As mentioned above,
this is unavoidable when the device is
to set the overall gain in the circuit.
However, there is a major drawback:
any non-linearity in the transfer charac-
teristic will give rise to distortion . . .

For this reason, it is essential for the
total circuit to be as linear as possible.
This is no mean trick, when you con-
sider that the bias current through the
input stage may vary over an extremely
wide range — from nano-amps to milli-
amps! To make matters worse, the
input stage of an OTA is inherently
non-linear — the transconductance is
determined by the 'diode'-characteristic
of the input transistors. For the older
type of OTA, this meant that the input
voltage should not be more than some
50 mV peak-to-peak, in other words,
the dynamic range is rather limited.

If we could eliminate the non-linearity
at the input, it would be possible to
apply much higher input levels — im-
proving the signal-to-noise ratio, for the
same maximum distortion. An alterna-
tive solution is to pre-distort the input
signal in such a way that the distortion
caused by the input stage will result in a
'clean' signal. In other words, distort the
signal 'in the opposite direction' before
applying it to the input.

This, effectively, is what the two diodes
at the input are for (D2 and D3 in
figure 3). To see how these work, it is
easier to re-draw the input circuit as
shown in figure 4. The two diodes are
now shown as transistors, with base con-
nected to collector (which is precisely
what they are, on the chip!), and all
the currents in the circuit are assumed
to be controlled by current sources.

The common emitter current for the
long-tailed pair (I ABC) is set by the
current source Ib; the total bias current
for the two diodes is Iq. To ensure that
the same current flows through both
diodes, under static conditions, a
further current source (141 d) is con-
nected between D2 and the negative
supply rail. Finally, and rather surpris-
ingly, the input signal is also assumed
to be a current, 1$. For the present, we
will not consider what happens when a
voltage is applied to the input! For the
purpose of this explanation, we will
also assume that the base currents of
T4 and T5 are so small that they can be
ignored.

When the input current is zero, the
currents through D2 and D3 must be
identical (both equal to KIq). Since all
transistors are identical, this means that

3

Figure 3. The circuit diagram of the new OTA. It consists of little more than four current
mirrors and an ordinary differential amplifier. Diodes D2 and D3 are the linearizing diodes,
which allow for larger input signal amplitudes.

the voltages across the two diodes must
also be the same. This, in turn, leads to
identical base-emitter voltages for T4
and T5; therefore l 4 = Is , and so
l s — l 4 = 0. No signal in means no
signal out, as you would expect . . .

Now, when a current 1$ is supplied to
the input, the current through D2 will
be reduced: l s plus the current through
D2 must be equal to '/zId, as determined
by the lower left-hand current source.
But when the current through D2 is
reduced, that through D3 must increase
by the same amount: the sum of the
two currents must remain equal to ID-
Less current through D2 means that the
voltage across this diode must also be
reduced; similarly, the greater current
through D3 corresponds to a larger
voltage across this diode. As a result, a
voltage difference appears between the

base of T4 and that of T5. The differen-
tial amplifier converts this voltage
difference into a differential output.

In effect, the input current is first
converted into a (distorted) voltage, by
means of D2 and D3; when this voltage
is applied to T4 and T5, the distortion
effects cancel out and the differential
output current is undistorted! This is
further illustrated in figure 5. In the
upper half of the drawing, the input
signal is zero. The currents through
the two diodes are identical (= V4I D).
so the voltages across them are also the
same. In the lower half of the drawing,
an input current (Is) has reduced the
current through D2 to V4 Id — Is; as
explained above, the current through
D3 must then increase to ’/2ID + Is-
When we 'bounce' these values off the

4-38 - alektor april 1982

OTA i

OTA?

5

Figure 5. In the absence of an input current,
there will be no voltage across the input of
the differential amplifier, (figure 5a). In
figure 5b an input current causes a voltage
U s -U 4 to appear, in turn this leads to a
current at the output of the differential
amplifier.

When is an

curve that represents the characteristic
of the diodes (figures 5a & b), we find
the voltages Uq 3 and Ud 4 as shown.
The difference between these voltages
is the differential input voltage,
U 5 — U 4 .

In the special case when the bias current
(IB) is equal to the diode bias current
Iq, the next step is very easy. The
base-emitter voltages of T4 and T5 are
then identical to Uq 2 end Ud 3 , respect-
ively. The collector currents are there-
fore found by 'bouncing' these values
back off the same curve; as can be seen,
the difference between the two currents
is equal to 2 l s .

As the amplifier bias current, lABC. is
reduced, the base voltages of T4 and T5
'slide down the curve'. This leads to a
smaller difference between l 4 and ls,as
can be seen; however, the linear relation-
ship between 1$ and the differential
output current is maintained. This not
only 'seems reasonable', when you look
at the plot; it can also be proved math-
ematically. In fact, the calculation is
given elsewhere;the less mathematically-
minded can forget the proof, however,
and believe the result:

ffirmil mirrors

A current mirror is nothing new: it is
simply a circuit that ensures that two
currents are identical. However, the idea
is extremely useful in 1C technology,
since excellent accuracy can be achieved
without the need for any external cali-
bration or on-chip trimming. The trick
is to make good use of the identical
characteristics of transistors — certainly
when they are on the same chip, as in
normal 1C manufacture.

As with all the best ideas, the basic
principle of a current mirror is quite
simple. When current is passed through
a diode, a voltage will appear across it.
The converse is also true: if exactly the
same voltage is applied across the diode
as that which appeared in the previous
case, a current will flow through the
diode that is also identical. A current
will define an exact voltage, and that
voltage will define the same current.
Taking this idea one step further: if the
same voltage is applied to two identical
diodes, the same current will flow in
each! The same applies to transistors:
provided they are identical, the same
base-emitter voltage will lead to equal
base currents and, since the transistors
are identical, the collector currents will
also be the same. It should be noted
that this is true over the entire range of
permissable currents — in spite of the
fact that the voltage/current character-
istic of a diode (or transistor) is anything
but linear.

From this point, it is only a small step
to a current mirror. Figure 1 shows the
simplest version, using only two transis-

tors. I, is the input current and l 2 is
the 'mirrored' output current — which
should be identical to l! .

When current is 'forced' into T1, this
transistor will conduct. It adjusts the
collector-emitter voltage (and, with it,
the base voltage) in such a way that the
base-emitter voltage exactly corresponds
to the desired collector current — ig-
noring the base current, for the moment.
For any given input current, I , , the
corresponding base-emitter voltage will
be set up by the transistor itself.

In this circuit, exactly the same base-
emitter voltage is also present across T2.
Since this transistor is identical to T1,
the collector current l 2 must be the
same as the collector current of T1 !

To sum it up: when a current (1 2 ) is
applied to T1, this transistor will set up
a corresponding voltage between base
and emitter. This voltage also appears
between base and emitter of T2, so the
collector current of T2 (l 2 ) must be
identical to I , (once again: ignoring
the base currents!).

Basically, that is all there is to a current
mirror. There is nothing mysterious or
'earth-shattering' about itl In fact, you
can easily build one, as shown in figure
2. For best results, the transistors
should be identical. An easy way to
make sure of this is to use a CA3046
(or CA3086). This 1C contains five
identical transistors, two of which can
be used for the current mirror. By the
way, although two meters are shown in
the circuit, it is quite possible to use
only one. First set the current through
T1, by adjusting PI; then move the
meter to the other 'leg' (and substitute
a wire link in the original position): the
collector current through T2 will prove

to be identical.

While we're at it, we can add a further
transistor (T3) in parallel with T2.
Obviously, it will also draw the same
collector current as the other tran-
sistors. In other words, the total col-
lector current of T2 and T3 must be
exactly equal to twice the collector
current of T1 ! A precision current
multiplier . . . which might prove useful
for a cheap-and-reliable D/A converter?
Don't bother trying to patent it: Philips
already have . . . it's what they use in
the 'compact disc' players!

So far, we have consistently ignored the
base current. In the basic circuit, how-
ever, they must all come out of 1 2 . This
means that the collector current of T1 is
actually slightly less than I,, so all
'mirrored' currents will also be slightly
smaller. Going back to figure 1 : if the
current gain of the transistors (8) is
100, the base currents will be 1% of the
collector currents. If we assume that the
base currents are both '1', the collector
currents will both be '100'. As can be
seen, this means that 1 1 = 1 02 and
l 2 = 100 - a 2% error!

Obviously, this error will be reduced as
the current gain of the transistors is
increased. However, for a precision
current mirror the gain would need to
be almost infinite . . . which is not so
easy to achieve. A better solution is to
add one more transistor, as shown in
figure 3. This reduces the error by a
factor that is equal to the gain of the
additional transistor!

At first sight, this circuit may seem
'reversed' — with l 2 as the input current
— but it's not that bad. Let's assume
that a current, 1 1 , is forced down at the
left. If T1 is blocked, initially, the cur-

When is an OTA not an OTA?

'out ■ Is - I. - 2 I, (|||.

From this formula, and from the circuit,
several things are apparent. In the first
place, the current gain is proportional to
the ratio between Ib and Iq. With one
proviso: the current through the diodes
can only flow in one direction! This
means that l s must always be less than
'/alD- If too small a value is chosen for
Iq, the distortion can become extremely
high . . ,

A further conclusion is rather surprising:
this OTA isn't an OTA - it's a current
amplifier! In the ideal case, it requires
an input current; it produces an output
current; and the ratio between the two
(the current gain) is set by the am-
plifier bias current, lABC. and the diode
bias current (Ip). 'Voltage' doesn't
come into the story! In fact, if a (differ-
ential) voltage is applied to the input,
the two diodes cannot pre-distort it.
The total circuit will then behave like
any other OTA - with all the distortion
associated with that operation mode!

In practice, an intermediate mode will

1

Figure 1. The basic version of a current mirror.

3

Figure 3. An extended version of the current

elektor april 1982 - 4-39

normally be used. Instead of current
sources, resistors are used to 'convert'
voltages into the necessary currents. The
results obtained in this way will ob-
viously be less than ideal, but they are
still surprisingly good when compared
with the older type of OTA. In particu-
lar, the input voltage can easily be ten
times as large for the same distortion.
This can be used to obtain a 20 dB im-
provement in signal-to-noise ratio. Not
bad — and when you consider the out-
put buffer and the fact that you have
two of these devices on one chip, it all
adds up to an interesting Id

Literature:

EXAR and National Semiconductor
datasheets (for the XR 13600 and the
LM 13600, respectively).

Linear relationship between l s and
•out

For the differential input stage, T4 and
T5, the ratio of the collector currents
is determined by the voltage difference
between the bases:

U 5 -U,

The factor KT/q depends on tempera-
ture, among other things; at room
temperature (25°C) the value is ap-
proximately 25 mV.

When we consider that the difference
between I; and I 4 is equal to the output
current, lout, and that their sum is
equal to Ig, it is apparent that the above
formula can be converted into:

U -U -KT. K'B+mout
5 4 q V4lB - ttlout

However, the same voltage (U 5 - U 4 )
also appears across 02 and 03. This

_ KJ KID + Is
5 4 “ q n 'A Id - Is

With a bit of shuffling, the relationship
between l 0 ut and 1 $ is found as:

•out * 2 Is ([^).

rent must flow through the base of T3
and down through T2. In effect, T1 and
T2 form a current mirror: T1 will now
draw a current that is equal to the
collector current of T2. This, in turn, is
almost identical to the collector current
of T3. Now, let's look at the base cur-
rents. All transistors are set to virtually
identical collector currents, so their base
currents must also be the same. Two
base currents (for T1 and T2) are de-
rived from the emitter of T3; half of
this current must correspond to the
base current of T3 (which comes from
1 1 ) and the other half comes from l 2 .
This means that I, and l 2 are almost
identical: the only difference is the
slightly higher base current needed by
T3 to pass the marginally higher col-
lector current. As stated above, this
error is reduced in proportion to the
current gain of T3.

This three-transistor current mirror is
almost perfect. It is used in the OTA.
Note that T2 is actually connected as
a 'diode'; it is drawn as such in the
circuit of the OTA. Needless to say,
awkward things like temperature fluc-
tuations have no adverse effect, since
the transistors are all on the same
chip. H

: igure 2. The

4-40 - ele

>r april 1982

2114 RAM tester

’III

RAM

tester

A memory 1C is a tiny 'black box'
in that none of its inner activities
can be seen from the outside. We
have to rely on the facts and
figures published in data sheets. It
is extremely difficult to know
whether the 1C is functioning
properly, since about as much is
going on inside as at the London
stock -exchange!

Where digital data is concerned,
however, operators can find out
how the 1C will react to a certain
logic input signal, using for
instance the 21 14 1C tester
introduced here.

The 2114 RAM is a highly popular 1C
and is used in just about every type of
personal computer. Consequently it
has dropped in price considerably in
recent years. Nevertheless, quite a few
RAMs are required for a computer
to have a reasonable memory capacity,
so that it would be nice if they could
be tested there and then in the shop -
with the dealer's permission, of course!
— to avoid buying any 'duds'.

Before we consider the circuit, let's take
a look at the test program in figure 1 .
As mentioned earlier, a digital 1C must
react to a specific input level. If the
1C is 'out of order' a red LED lights by
way of warning. Initially, every memory
location is loaded with a logic 1 . Should
the tester detect a low logic level any-
where, something is bound to be wrong
and the indicator will light. If, on the
other hand, everything is perfectly O.K.,
the next section in the test program is
run. This time low logic levels are stored
throughout the memory range. Again,
as soon as the tester encounters a logic
one, the LED lights. Otherwise, the test
program simply starts all over again.
More details about the test cycle will
follow later.

At the same time, the tester checks the
current consumption of the RAM. An
additional red LED lights whenever the
power supply is 'shorted' or the 1C
consumes more than 100 mA.

The circuit

When SI, a pushbutton switch, is de-

Figure 1 . The test program flowchart. A red
LED lights whenever the output data does not
correspond to the input information.

pressed the output of N2 is 'high' to
start with. This resets IC1 and the flip-
flop N3/N4 (by way of N5) and the Q
outputs of IC1 are ail 'low'. After about
100 milliseconds the capacitor C2 is
charged by way of R6 until its level
reaches the switching threshold of the
Schmitt trigger N2. This makes the out-
put of N2 go 'low'. As a result, the
12-bit binary counter (IC1) is started.
During the first 1024 (= 1 k) pulses
produced by the clock generator N1 (at
a clock frequency of about 10 kHz),
outputs Q10 and Q11 remain 'low'.
This means that the WE input will also
be logic 0. Since the inputs of gates N6
and N7 are 'low' and their outputs are
therefore 'high', inputs 1/01 . . . 1/04 of
the RAM will be 'high'. In other words,
one nibble (4 bits) per clock pulse is
stored.

After 1024 clock pulses, Q10 goes high,
which prepares the memory 1C for the
output (READ) of data by way of the
WE input. Seeing as Q11 continues to
remain 'low', the logic ones which were
entered previously .may now be read
out during the next 1024 clock pulses.
Gates N9...N12 act as (EXOR)
comparators. Their outputs will always
be high. Provided there is a logic 1
at only one input. In this particular
case, the outputs will be low. Diodes
D1 . . . D4 and resistor R11 together
form an OR gate. None of the diodes
conduct, so that the input of N8 will be
low. Gate N8 acts as an inverter and
under the conditions described here its
output will be high.

Since the inverter N5 also produces a
high logic level at its output, which
reaches the RS flipflop N3/N4, the Q
output and the base of transistor T1

4-42 - ele

ril 1982 2114 RAM'

connection

tester

The connection tester is an excellent aid to examine the quality of
soldered joints and connections in an electronic circuit. The tester will
indicate a 'good' connection with an acoustic signal. With a normal
multimeter one must keep at least one eye on the pointer, so an
acoustic indication makes testing that much quicker and easier: both
eyes are free to check the circuit. The tester gives a tone when there
is a connection, and remains silent when there is an open circuit or
when the resistance across the connection exceeds 1 . To prevent

any damage to sensitive components, and for good battery life, it
injects only a weak signal.

stay 'low'. As a result, T1 does not
conduct and the LEO D5 remains unlit.
Because of the OR circuit around
D1 . . . D4, a single low level at the
RAM outputs sets the flipflop N3/N4
and makes the transistor conduct. D5
is then provided with current and lights
-something is wrong. The flipflop
will not respond to any further error
messages. SI has to be depressed again
to initiate a new test cycle once N5 has
produced a reset pulse.

Now for the actual test cycle. Supposing
the first test (the writing and reading of
logic ones), was successful and the LED
D5 did not light. At the end of another
1 024 clock pulses, Q1 1 goes high and
Q10 goes low. The RAM is now in the
writing mode and low signals are read in
by way of N7. Once Q10 has gone high
as well, the RAM will be in the reading
mode and the logic zeros are output. If
the comparators N9 . . . N12 are unable
to detect a single logic '1' at their
inputs, nothing happens. Otherwise, the
flipflop is set and the LED lights.

With regard to the test circuit it should
be noted that the RC network R7/C3
disables the pulses caused by varying
gate times when the logic state of the
RAM inputs and outputs changes.

The second test measures the current
consumption. Pushbutton SI needs to
be depressed for about 2 ... 3 seconds
to allow the test cycle to be run a few
times. The RAM will be in perfect
working order if D5 does not light. The
current consumption is measured at the
start of the cycle. The 'current' tester
consists of R9, RIO, T2 and D6. If the
RAM consumes more than 100 mA,
because of an internal 'short' for in-
stance, (or if the RAM has been incor-
rectly connected to the supply!) the
drop in voltage across RIO will cause T2
to conduct and D6 to light. Normally
speaking, the L-type RAM draws 25 mA
(40 mA maximum), whereas the normal
types are rated at 50 mA typical and
70 mA maximum. The maximum short
circuit current is 140 mA and is con-
trolled by IC5.

Note that it is also possible to self-test
the RAM tester. Simply depress SI
when no RAM is inserted and if all is
well LED D5will light.

Construction

As can be seen from the printed circuit
board in figure 3, the construction is
quite straightforward. The photograph
shows what the finished product should
look like. The RAM may. of course, be
mounted in a zero insertion force (ZIP)
socket if required, but these do tend to
be rather expensive. The circuit must
be powered with a 9 V battery. To
'wrap up' the project, insert the tester
circuit into a small plastic case. Make
sure there is plenty of room around the
RAM on the test socket and that it
doesn't come into contact with other
components. K

When testing connections there is a fair
chance that resistors, semiconductors
and other components are involved in
the measurement. Moreover it is poss-
ible that certain components cannot
cope with the current and/or voltage the
tester injects. For this reason, a good
tester is one that will not react to low
impedance PN junctions (diodes, tran-
sistors) and resistors. Furthermore the
device must be sensitive enough to oper-
ate with a weak test signal. The circuit
shown in figure 1 meets all these re-
quirements. Thanks to the high gain of
the opamp (type 741) used in this cir-
cuit, the current and voltage for the test
signal can be limited to 200 /xA and
I 2 mV, respectively.

The opamp is connected as a differential
amplifier; the voltage difference be-
tween the inverting (pin 2) and non-
inverting input (pin 3) is increased con-
siderably. The voltage drop across R2
ensures that the output of the opamp
becomes negative, since the inverting
input has a higher potential than the
non-inverting input. The potential at the
non-inverting input can be increased by
turning PI, so that this input becomes
more positive than the inverting one
(as soon as the voltage across R2 drops).
The result is a positive voltage at the
output of the opamp. The oscillator
constructed around N1 will then pro-
duce a tone via the buzzer. The voltage
drop across R2 is caused by a good

elektc

ril 1982-4-43

contact between the probes of the
tester. PI is used to calibrate the tester.
Compared to an optical indication an
acoustic indication is not only more
convenient, but its current consump-
tion can be lower as well. The buzzer is
loudest when its resonance frequency is
about 4.6 kHz. The current consump-
tion will then be about 3 mA. The fre-
quency, and therefore the volume, can
be set with the aid of P2.

Calibration

After correct calibration, only resist-
ances of up to 1 ohm (in a connection)
are tolerated. A value lower than 1 ohm
either indicates a good contact or a
short circuit. The calibration procedure
is as follows: Place a resistor of 1 ohm
(5 or 10%) between the probes and set
PI in a way that the buzzer is about to
give a tone. Remove the 1 S2 resistor
and cause a short circuit between the
probes; again the buzzer will make itself
noticeable. Volume can now be set with
P2. When the short circuit is removed
the buzzer must remain silent. To be
certain, correct operation can be
checked once more, by placing a resistor
of a few ohms between the probes. If
the buzzer sounds now the calibration
will have to be repeated.

One final remark: The supply voltage of
the circuit under test must be switched
off when being examined with the tester
described in this article. The supply
voltage could have a negative influence
on the tester or even damage it. H

vil 1982

toroidal transformer

loroklnl

transformer

the best transformers . . . around!

Ring core or toroidal transformers are becoming fashionable. Thin
is beautiful? As their name implies, they are 'round' and low in
profile, allowing the home constructor and manufacturer, to build
highly compact circuits. This seems to be necessary in order to satiate
the public's appetite for any equipment that looks like a permanent
'Weight-Watcher'. Seriously, they do have excellent 'electrical' qualities,
and advantages over the conventional transformer, other than looks.
Unfortunately good taste is always relatively expensive.

The toroidal transformer has a ring core
formed by a tightly bound metal lami-
nated band. Copper windings are simply
placed on the core without the use of
bobbins.

The wire is wound over the complete
surface of the core and this considerably
aids the dissipation of heat. Due to the
round shape, there is good 'concen-
tration' of the magnetic flux lines in the
core, thereby reducing the 'stray'
fields.

It requires less wire than the conven-
tional transformer for the equivalent
number of windings, thus reducing the
ohmic resistance, and the chance of
overheating. So far so good. But why is
the ring core transformer in most cases
more expensive to buy than the conven-
tional type? After all, they use less
copper wire, no bobbins etcetera! Good
question. The answer is that they take a
lot longer to manufacture than conven-
tional transformers, and today more
than ever, time is money.

The core is formed as a complete ring
without an air gap. It is made from a
strip of high grade sheet steel, which is
rolled up very tightly. The end of the
strip is then welded, to prevent it un-
winding. This form of construction
helps to concentrate the lines of flux
within the core and keeps losses to a
minimum. An added advantage is its
lack of buzz; due to the very tight 'lami-
nations', which are completely enclosed
by the winding. The result: an inbuilt
disability for the production of noise.
Mains toroidal transformers are readily
available in the 15 to 680 VA range, and
up to 5000 VA types are supplied by
some manufacturers. Most are available
with two secondary windings, of be-
tween 6-60 V.

Winding toroidal transformers

The manufacture of toroidal trans-
formers may present something of a
question mark to the inquisitive reader.
As in most things of this nature, the
answer is quite simple; once you know
howl Figure 1 illustrates, what, in
simple terms, actually occurs.

The complete core is loaded onto a
machine that is able to hold and, rotate
it. A ring, that is about three times the
diameter of the core, is linked onto the
core in much the same way as two links
of a chain. This ring is called, not un-
reasonably, a shuttle and can also be
rotated. While doing so, an amount of
wire equalling one complete winding is
fed onto it. And now we come to the
'trick' that makes it all so simple. The
end of the wire is turned through
180° around a guide wheel on the
shuttle, and held onto the outer edge
of the core. The shuttle then reverses
direction and lays the wire onto the
core as one winding. The core is of
course rotated slowly as this happens,
so that the winding is evenly spread
around it. Tension of the wire is easily

controlled. Mechanically this method
is both simple and quick, and in fact
takes just three minutes per winding.

The Lord of the Rings
(Transformers)

The equivalent conventional trans-
former is in most cases 2 to 3 times
heavier. The same ratio in size also
holds true.

The ring core transformer's 'iron losses',
(when compared with the 'standard'
conventional type), are only 10%. The
advantage of the ring type are clearly
noticeable when comparing 'stray
fields'.

In a no-load situation the conventional
field is at a maximum and the ring core
at a minimum. With an increasing load
the 'stray field' of the conventional
decreases and the ring core's field in-
creases in strength.

No matter what the situation, the stray
field of the ring core type is always
considerably smaller. Therefore using a

toroidal transformer reduces the risk of
unwanted noise being generated in any
power supply circuit.

Quality costs money?

Toroidal transformers up to power
ratings of 200 VA are more expensive
to buy than conventional types. Above
200 VA and up to 500 VA this situation
is reversed. A reasonably priced, com-
pact, transformer above 200 VA is cer-
tainly useful, especially when building
high power amplifiers.

Final remarks

Compared to the ordinary standard
transformer, the high-grade core ma-
terial of the toroidal type will cause a
higher initial surge current; A slow-blow
fuse on the primary winding is there-
fore necessary. A fuse having approxi-
mately double the value normally used
with an equivalent conventional trans-

former should do the trick.

Even so, do not be alarmed if the whole
neighbourhood is 'blacked out' the
moment you switch on your new 2 x
50000 . . . W amplifier (with multiple
toroidal transformers). This is a normal
occurance! H

4-46 - elektor april 1982

: keyboard

capaddw

keyboard

Small, single-board microprocessor sys-
tems require a keyboard consisting of
between 10 and 20 keys. Keyboards of
this kind look very simple, but are
surprisingly expensive. Normal use sub-
jects the keys to considerable mechan-
ical wear, making frequent replacement
necessary. The use of conductive rubber
and 'hall' effect elements in one way to
overcome this problem. But a more
effective solution is a capacitive
keyboard using touch activated keys,
thereby dispensing with mechanical
devices altogether.

In principle, normal keys are substituted
by copper squares, arranged in a matrix,
which alter in capacitance when
touched. Such a system although
sounding complicated is actually quite
straightforward.

Operation

a solid state 'keyless keyboard'

The choice and price range of available computer keyboards is today
swiftly approaching a level of infinite proportions. As a result, many
readers prefer to build their own.

A system using mechanical keys, although being simple, is relatively
expensive. Capacitive touch activated keyboards are an economical
alternative. They achieve a high standard of reliability, without the
need to use expensive conventional mechanical equipment.

The drawing in figure 1 shows, in a
nutshell, how a capacitant keyboard
works. Two capacitors are connected
in series between the wires. The
junction between them forms the touch
contact. One of the capacitors (Ct>) is
linked to the input of a monoflop,
whereas a pulse is fed to the other
(C a ). Without the contact being
touched, the pulse will be differentiated
by the combination of C a , Cb and R.
The monoflop will switch on the
leading edge of the differentiated
waveform and will produce an output
pulse with a specific length.

Touching the contact plate causes a
leakage capacitant and/or resistance
path to earth, which reduces the ampli-
tude of the pulse to the monoflop to
below its input trigger threshold. Con-
sequently the monoflop does not
generate an output pulse.

A complete keyboard can be operated
along these lines. A short program
makes sure pulses are generated and
enables the system to detect any inter-
ruptions in the pulse flow (i.e. when one

1

Figure 1. The basic principle behind a
capacitive keyboard.

4-48 - elektor apr

little further, to the point at which the
monoflop will be triggered when a 'key'
is firmly pressed. The same procedure is
carried out for each column in turn.

Control and detection

Figure 4 shows how the column pulses
are generated, usually by the micro-
processor used. After the positive-
going edge of every pulse, the micro-
processor checks whether one of the
row outputs has gone 'high' during the
monoflop activity period. Any row
output found to have gone high means
that the key situated between the
column where the pulse was produced
and the row where the pulse was found
to be missing must have been touched.
Using the component values as indicated
the monoflop activity time period is
about 50 as- This can be modified by
replacing the 470 pf capacitor with one
of a different value.

The keyboard must also include a de-
bouncing unit. Again, this is under the
control of the host microprocessor.
Figure 5 contains a flow chart for a
hypothetical program. Unless a key is
depressed, the program will remain in
the B loop, where the computer waits
for 10 ms. During this time it can
perform other tasks, such as driving
the display. When a pulse is generated
at column 1, the rows are scanned,
followed by a pulse at column 2, and
so on, until all the rows have been
scanned. Once all five columns have
been dealt with without the computer
encountering a logic 1 (= no pulse)
anywhere, a return is made to the be-
ginning of the B loop, as obviously no
keys were touched.

On the other hand, if a high logic level
is detected, the B loop is left and the
processor waits for 10 more milli-
seconds before verifying whether the
same key is still being touched. If that
is the case then the program is exited at
the point marked by the arrow and a
new program is called to process this
information. Subsequently, the
computer returns to the KEY label and
if the key is still being activated, the A
loop is run until the key is released. The
B loop is then reinitialised and the
computer waits for a key to be operated.
This procedure enables the keys to be
debounced for at least ten milliseconds.
In practice, this seems to work very
well. As can be seen from the circuit
diagram (figure 2), five outputs for
the columns and four inputs for the
rows are required making a total of nine
I/O lines from the ; iP .

The 'keys'

The keyboard is made using a piece of
double-sided printed circuit board.
Twenty copper squares (each 1.5 x
1.5cm) arranged in a 4x5 matrix are
etched onto the upper side, leaving a
5 mm space between each (see figure 6).
Identical copper squares are then etched
on the lower side of the board, in the
same corresponding position, only this

3

un«i

1 after the second pulse.

col i n

COL 2 1

n » n

col 3 :

COL 4

COL 5

n I

--d

Figure 4. How the various columns are controlled.

time a narrow strip is scratched out
through the centre of each one. In
addition, the upper halves are all inter-
connected by means of a narrow copper
track whereas the lower half of each
square is provided with a soldering
point.

Great care should be taken when
etching and preparing the printed circuit
board, as the successful performance of
the keyboard depends entirely on the
fact that the keys all have the same
capacitance. In order to achieve this,
they must all have the same surface
area, and they must be situated in
identical relative positions on both
sides of the printed circuit board.

The 'Keyboard' is shown in figure 6.
Readers may feel that they wish to in-
clude the electronics on it. Once the
board has been etched, the column
soldering points are interlinked in the
manner shown in figure 7. Again, a

fair amount of care is called for. Thin
copper wire can be used for this
purpose.

The 'Keyboard' can be covered with
a piece of transparant cellophane.
Readers are invited to experiment
with rub-on lettering.

It would be wise to recalibrate the
Keyboard after it has been fitted in a

To ensure maximum protection, fit a
metal plate about 2 cm below the
keyboard. The plate should be parallel
to the printed circuit board, otherwise
some 'keys' may turn out to be more
sensitive than others. The metal plate
should be earthed to the case. Finally,
keep the column and row connection
wires as short, and as symmetrical as
possible.

Readers are of course free to select any
number of key they like. The perfor-
mance of the circuit is largely depend-

1

4-50 -elektor april 1982

the 13600,
a

new OTA

The new OTA, type 13600, is
an extremely useful 1C. This
article deals with some practical
applications; the theory is
discussed elsewhere in this issue.

Practically nothing can go wrong when ex-
perimenting with the new OTA, as long as the

table 1 are not exceeded.

According to available information at least
three companies presently manufacture the
13600; EXAR (XR13600), National Semi-
conductor (LM13600), and Philips (NE5517).
The Philips version has a slightly different
internal layout, but it is pin-compatible to
the other two. Each 1C contains two com-
pletely separate OT As.

shown in figure 1 . The circuit diagram for the

shown in the article describing the theory
of this OTA.

The applications for the 13600 have been
selected with the thought in mind, that there
should be something for everybody. The
circuit diagram in figure 2 may look like the
volume control from the theoretical article,
but it is, in fact, an AGC (Automatic Gain
Control) amplifier, that tries to keep the
amplitude of the output signal constant, in-
dependent of the input signal. Its operation
is as follows; As soon as the output voltage is
high enough, (more than three times the
voltage at a PN junction), the darlington stage
and linearising diodes will start to conduct.
This reduces the gain, stabilising the output
level. The offset voltage at the output can be

Figures 3 and 4 show a low pass and a high
pass filter, respectively. These filters have
unity gain inside the pass-band; beyond the
turnover frequency the gain drops at 6d6
per octave. The cutoff frequency can be
calculated with the aid of the formula in-
dicated in both figures 3 and 4. The relation-
ship between R and Ra determines the gain
of the OTA.

A completely different kind of circuit is
shown in figure 5. This is a timer {monostable
multivibrator) that does not consume any

nection between the bias input and the out-
put of the opamp. When the output voltage
drops to zero, the bias current also becomes
zero and all stages in the OTA are prevented

with a positive pulse at the input. Provided
the pulse is higher than 2 V, the OTA will be
switched on, and its output will swing high.

4-52 - elektor april 1982

BASIC i

BASIC still remains the number one
computer language. Although it may
not be as grammatical and efficient as
other languages (such as COMAL or
PASCAL), its popularity shows that it
meets the essential requirements of
computer users all over the world.
Thanks to Microsoft, who developed an
excellent version of BASIC for the KIM
computer some time ago, the Junior
Computer can now be made bilingual,
its 'mother tongue' being machine
language of course. Even with the
addition of a BASIC vocabulary, ma-
chine language still plays an important
role in various routines and timing
processes etc., so there is no question

1SASK' cm die

•Junior

Computer

. . . puts the microprocessor
in touch with the world

Although the Junior Computer is quite fluent in machine language, its
linguistic skills cannot lead to full 'adult' communication until the
machine has learned a 'high level' language, such as BASIC. A specially
adapted version of BASIC is now available on cassette from the
Microsoft/ Johnson Computer Corporation, which will enable Junior
Computer operators to fulfil their dreams at long last.

This article introduces the Microsoft cassette and describes how to
implement the new facility on the Junior Computer. Anyone who feels
that their BASIC has become rather 'rusty' will welcome the oppor-
tunity to brush up their knowledge. As for beginners, it is never too late
to learn!

of it being completely replaced.

The KB-9 BASIC by Microsoft is a nine
digit 8k BASIC on cassette. Since this
was originally developed for the KIM, it
will have to be modified before it will
run on the Junior Computer. Contrary
to what might be expected, this is a
straightforward operation that takes a
mere fifteen minutes or so. This is
nothing compared to the thousands of
man-hours involved in developing the
Microsoft BASIC. Only 31 of the eight
thousand memory locations need to
have their contents altered. Now to
discover what ingredients are required
to 'cook up' a BASIC on the Junior
Computer.

The ingredients

First of all, what about the hardware?
Obviously, the computer will have
to be a fully extended version, that is,
equipped with an interface board and
extended memory. How this is ac-
complished is fully described in Junior

Computer Book 3. In addition, 16k of
RAM has to be located in the address
range $2000 . . . S5FFF. This can be
made up from two RAM/EPROM cards
each containing 8 k of RAM, or the 16k
dynamic RAM card which is described
elsewhere in the present issue.

Although the extensions were fully
described in Book 3, it may be an idea
to briefly recap on a few main points
here, as this is really a very basic part of
the BASIC facility! The extra bus board
memory should also contain the three
jump vectors situated in the address
range SFFFA ... SFFFF. Appendix 3
in Book 3 mentions two ways in which
to include these vectors without the
need for an expensive RAM/EPROM
card. This issue of Elektor also describes
a mini EPROM card which provides yet
another option.

As far as the software requirements
are concerned, both the printer monitor
(PM) and the tape monitor (TM) rou-
tines must be available. The former
contains the input/output subroutines
RECCHA ($1 2AE), PRCHA ($1334)
and RE3TTY ($14BC) which serve to
start the Junior Computer BA3IC. The
latter contains the main cassette rou-
tines DUMP ($09DF) and RDTAPE
($0B02). Then, of course, the KB-9
BA3IC cassette (not KB-6 nor KB-81!)
will have to be acquired, together with
all the necessary documentation. Other
requirements include a cassette recorder,
an A3CII keyboard, a printer and/or a
video display terminal and an under-
standing of programming in the BASIC
language. Anyone who wishes to brush
up on their BASIC knowledge should
read the crash course published in the
March . . . June 1979 issues of Elektor
or SC/MP Book 2.

The recipe

• Switch on the Junior Computer and
start up the PM routine. Place the

KB-9 cassette in the tape recorder:

RST 1 0 00 GO RES
G1 (CR)

• Start the recorder in the play mode
at the beginning of the tape. The

program number (ID) of KB-9 is 01.
Reading in the instructions etc. takes
several minutes, after which the com-
puter reports 'READY'. Remove the
cassette from the recorder as it is
advisable to store the Junior BASIC on
a separate cassette and preserve the
KB-9 version in its original form.

• Using the PM routine, alter the
contents of 31 memory locations, as

indicated in the first section of the
accompanying table. Start by checking
the 'old' data at the locations concerned.
Any discrepancies will mean that you
have been landed with the wrong
version of BASIC!

• Place a new cassette in the recorder.
Start at the beginning, reset the

counter and depress the record and play
buttons. After about ten seconds enter:

BASIC

1982 - 4-53

The KB-9 to Junior BASIC conversion teble

(based on the KB-9 cassette. # 4065©1977 by Microsoft Co.; version VI .1 1.

1 . The interpreter

a. ID ■ B1 instead of 01.

b. 1. Address S2457 should contain AE instead of 5A;

2. Address $2458 should contain 12 instead of IE;

3. Address S26DD should contain 80 instead of 40;

4. Address S26DE should contain 1 A instead of 1 7;

5. Address $2746 should contain 79 instead of F9;

6. Address $2747 should contain 1 A instead of 17;

7. Address $274D should contain 70 instead of F5,

8. Address $274E should contain 1A instead of 17;

9. Address $2750 should contain 71 instead of F6,

10. Address $2751 should contain 1 A instead of 17;

1 1 . Address $2757 should contain 72 instead of F7;

12. Address $2758 should contain 1 A instead of 17;

13. Address $275A should contain 73 instead of F8;

14. Address $275B should contain 1 A instead of 17;

15. Address $275E should contain 1 A instead of 18,

16. Address $2791 should contain 70 instead of F5;

17. Address $2792 should contain 1 A instead of 17;

18. Address $2794 should contain 71 instead of F6;

19. Address $2795 should contain 1 A instead of 17;

20. Address $2799 should contain 79 instead of F9;

21. Address $279A should contain 1 A instead of 17;

22. Address $27A4 should contain 09 instead of 73;

23. Address $27 A5 should contain 1 A instead of 18,

24. Address $27B9 should contain FA instead of ED;

25. Address $27BA should contain 00 instead of 17;

26. Address $27BC should contain FB instead of EE;

27. Address $27BD should contain 00 instead of 1 7 ,

28. Address $2A52 should contain 34 instead of A0;

29. Address $2A53 should contain 1 3 instead of 1 E ;

30. Address $2AE6 should contain AE instead of 5A;

31. Address $2AE7 should contain 12 instead of IE.

2. Additional instructions on page 1A

a. ID = B2.

b. 1 . Address $1 A00 contains 20;

2. Address $1A01 contains DF;

3. Address $1 A02 contains 09.

4. Address $1A03 contains 20;

5. Address $1A04 contains BC;

6. Address 81A05 contains 14.

7. Address $1 A06 contains 4C;

8. Address $1 A07 contains 48;

9. Address $1 A08 contains 23;

10. Address $1A09 contains 20;

1 1 . Address $1 A0A contains 02;

12. Address $1A0B contains 0B;

13. Address SI A0C contains 20;

14. Address $1 A0D contains BC,

15. Address $1 A0E contains 14,

16. Address S1A0F contains 4C;

17. Address SI A 10 contains A6;

18. Address $1 All contains 27.

SB1, 2000,4261 (CR)

It only takes a matter of minutes for the
Junior BASIC to be recorded. The
program number will now be B1.

• As soon as the Junior BASIC is
stored on cassette, the message

'READY' will appear on the printer or
the video screen. Let the tape continue
for a further ten seconds before de-
pressing the stop key.

• 18 locations on page 1A (PIA RAM)
need to be loaded with six LOAD

and SAVE instructions. The address
area concerned is $1 A00 . . . $1A1 1 ; the
details of the contents of these locations
can be found in the second half of the
table. This data is given the program
number B2 and is again stored on
cassette.

• Depress the record and play keys
once more and enter:

SB2, 1 A00, 1 A12 (CR).

• After the 'READY' message, the
cassette recorder can be stopped.

Now it is time to check whether the
Junior BASIC is correctly stored in
memory. This can be done with the aid
of the 'question and answer' game
following the BASIC start address
($4065). It is always a good idea to
enter a test program. The cassette com-
mands can be verified by writing a
BASIC program, storing it on cassette
(SAVE), erasing the program area (NEW)
and then reading the program in again
from cassette (LOAD). Once the Juniqr
BASIC has met with approval, the same
procedure can be used to test the Junior
BASIC cassette. For this, the Junior
Computer is switched off for a while
and then on again, after which the two
programs (B 1 and B2) are loaded from
cassette.

Ready to serve

By now the operator is all set to dish up
the Junior Computer BASIC. Do not
forget to read the manual supplied with
the cassette. This consists of the 'Micro-
soft Introduction', 'Dictionary' and

'Usage Notes'. Although the contents
are rather concise, to the point of being
cryptic, all the necessary information is
provided. As far as the software is con-
cerned, only one or two actual addresses
are mentioned.

The following remarks, however, should
make things a bit clearer;

1 ) After entering:

RST 1 0 0 0 GO RES (RUBOUT)
GB1 (CR)

READY (depress stop key)

GB2 (CR) (depress stop key again)
READY

the Junior BASIC can now be started. A
cold start entry takes place at address
S4065.

4065 (SP)R

The program must be started by way of
PM and not by way of the original
monitor routine, as otherwise the
input/output parameters will be in-
correctly defined. In any case, PM is
indispensable for reading in data.

4-54 - elektor april 1982

BASIC on the .

lior Cor

2) The Junior BASIC utilises the follow-
ing memory range on page zero:

$0000 . . . $00 DC and $00FF. Thus one
of the locations (MODE) belonging to
the original monitor is used. This merely
serves to start PM.

3) The start address for a warm start
entry is $0000. In the KIM the warm

start entry allows the computer to
return to BASIC after writing or reading
a BASIC program to or from cassette.
In the case of the Junior Computer
things are slightly different (see point 9).
Flere, the warm start entry may be used
to return to BASIC from PM. The jump
from BASIC to PM occurs either as a
result of a non-maskable interrupt (NMI)
or because the BREAK key on the
ASCII keyboard was depressed. The
BRK jump vector points to the label
LABJUN ($105F) of the PM routine.
After printing the text 'JUNIOR', the
computer jumps to the central label
RESALLofPM (see Book 4 chapter 14).
In the event of an NMI, RESALL is
reached at the end of the STEP in-
itialisation routine ($14CF).

4) The ST key may be used during PM
to examine the contents of various

memory locations, such as the ones on
page zero (see point 2) for instance. A
warm start entry heralds the return to
the BASIC program.

5) Supposing the operator is executing a
BASIC program (RUN) making use

of the Elekterminal (up to 16 lines on
the display) and the BASIC program
turns out to contain more than 16 lines.
This is what should be done:

RUN (CR)

BRK (while 16th line is being printed)

examine result

(SP) R The computer prints

OK Start the program again: RUN (CR)

enter the 16th and following 14 lines,

6) When starting the Junior BASIC
by way of a cold start entry, the

operator will be requested to state the
•TERMINAL WIDTH'. If the Elek-
terminal is being used, this is set at 64
(CR).

7) The ASCII keyboard does not
provide a 't' nor a '"' key necessary

for power functions, where At4 corre-
sponds to A 4 . What is required is an
ASCII key which generates the code
$5E. This can be improvised by sacrific-
ing another key. One contact is connec-
ted to row x7 and the other to column
y9 in the keyboard matrix (pins 32 and
22 of IC1). Only two keys are suitable:
the 'PAGEt' key at the far right in the
top row and the 'ESC' key at the far left
in the second row. The latter affords the
most elegant solution, as the ESC
function is preserved (a matter of
combining it with the Shift key).
Interrupt the two connections x5 and
Y10 (without actually cutting the wires!)
and link the ESC key to pins 22 and 32
of IC1. Further details are provided in
the article concerning the ASCII key-
board (Elektor November 1978), in

Book 3 of the Junior Computer series
and in SC/MP Book 2.

8) In order to start the Junior BASIC
by way of a fresh cold start entry

during a computer session, the program
(file B1) will have to be loaded from
cassette all over again. This is necessary
as a relatively large section of data block
B1 is reserved as the first section of the
BASIC work area if any trigonometric
functions are required. After the cold
start entry, the computer will request
the operator to specify its task. Whether
trig functions are to performed or not,
the computer must be informed by way
of a cold start entry, (once B1 has been
loaded again).

N.B. In file B1 ($2000 ... $4260),
locations $4041 . . $4260 are added to
the user work space if the operator
wishes to utilise the trigonometric
functions (depress the Y key); locations
S3F1F . . . $4260 are added to the
user work space if the trig functions
are not required (N key); locations
$3FD3 . . $4260 are added if the ATN
function (A key) is cancelled.

The first memory location is loaded
with 00 (BOF: Beginning Of File). Now
that 16k of RAM has been added, the
user work space will cover the following

$4042 . . . $5FFF (8126 bytes) when Y
is depressed;

$3F20 . . . 85FFF (8416 bytes) when N
is depressed,

and 83FD4 . . . 85FFF (8236 bytes)
when A is depressed.

9) Thanks to the Junior Computer
subroutine system, reading and

writing BASIC programs to and from
cassette (SAVE and LOAD) is much
easier than with the KIM BASIC. The
only snag is that this occupies the
second file, B2. After SAVE has been
entered, the BASIC program is stored
on tape (where ID = FE). After a while,
the 'OK' message will appear followed
by an empty line. After LOAD (CR) is
entered, a BASIC program is read from
tape (where ID = FF, so make sure the
required BASIC program is stored
before this!). A little later 'LOADED' is
printed. This is not followed by the
message OK and the computer does not
start a new line. In other words, the
video screen will display 'LOADEDLIST'
if the entered program is to be checked.

Any questions?

Here are the answers to one or two
questions which are likely to be asked:

1) Elektor cannot comply with requests
for a copy of the notes accompanying

the Microsoft/Johnson BASIC cassette
as this would be an infringement of
copyright.

2) The source listing of the KB-9 costs a
few thousand dollars. Not surpris-
ingly, Elektor is not in a position to sell
it to readers.

3) The KB-9 BASIC cassette should be
available from Calist Computers Ltd.

119 John Bright Street Birmingham
B1 1 BE Tel. 021/6326458. H

coming

soon...

Many new projects and designs are on
the way, with subjects ranging from
audio to microprocessors. In keeping
with the policy of Elektor every reader
will find something to his liking. Here
are some of the projected articles:

For the musician the guitar prempli-
fier would be of interest. This is a
sophisticated circuit including such
facilities as active tone controls, equal-
isation, reverb, fuzz, and many more.

Part two of the polyformant enables
readers to start building. Provides
constructional details together with the
printed circuit board designs.

The motorist is also catered for: the
auto burglar alarm. An alarm with an
automatic reset facility. It does not
matter how many attempts are made to
break in, the alarm is always ready.

A versatile counter that can be used for
many different applications as well as a
lap counter for Blotcar racing.

Readers whose prime interest is R.F
may breathe a sigh of relief. A 20 metre
S.S.B. receiver comparable in perform-
ance with professional equipment,
without costing the earth.

The ever-increasing number of readers
having home computers will find plenty
to keep them busy. As regular readers
know, Elektor stays one step ahead
where electronics is concerned, and we
will continue to produce articles and
designs keeping in step with the latest
technological advances.

1982

Low cost computer graphics with
Robocom BIT STIK

Designed and produced by the robotics re-
search and development company, Robocom
Ltd of London N4, this new hardware/soft-
ware package enables microcomputer users to
create multi-colour graphics, sketches, tech-
nical drawings, electronic circuit diagrams,
typography, etc.

Apart from opening up new creative possi-
bilities, the precision of the BIT STIK hard-
ware and speed of its machine-coded, user-
friendly software, facilitates drawing and the
inputting of information for any personal
or business user of a microcomputer. The
software package included in the system also

A secondary MENU can also be selected in
order to SAVE and LOAD drawings onto
floppy disc. In this way a file of drawings
can be built up with easy access for use as
components of other drawings. Selections are
also available to TEXT IN GRAPHICS (pos-
itions text at any scale, angle and colour in a
drawing) DIMENSION (for dimensioning
technical drawings) and ANIMATE which
allows for the dynamic recording and replay

Two PAGES are available at any time, either
can be used for direct drawing or to hold
components or scaled parts of the work in
hand. Finally a memory COUNTER is pro-
vided to indicate the available workspace
and a valuable COMPRESS function allows
drawings to be squeezed into the minimum
memory space.

An integral function of the system is the
ability to retain hidden detail. A single page
can in fact hold up to 1 6,000 pages of infor-
mation which can be viewed by zooming in
on specific areas, the additional detail being
called from the disc as required.

PAPER COPIES of drawings created with the
BIT STIK graphics system can be produced

Once the software has been loaded from the
DOS 3.3 disc provided, the user can draw
directly on the video screen using the BIT
STIK as the only input. A comprehensive

and accessed simply by 'pointing' to the
required items with the drawing cursor. The
menu provides PALLETTE facilities for
automatic LINES, ARCS and CIRCLES, six
COLOURS, four LINE TYPES, plus a variable
NIB for colouring and lettering. The MENU
mode selections are DRAW, ERASE, ZOOM
(for detail drawing and viewing), COPY (for
reducing and compiling), FIND (locks onto a
particular point) and WIPE (for a clean page).
In addition, an automatic paint facility allows

drawings is limited only by the capabilities of

the plotter, not the original video page.

A fully illustrated user-friendly MANUAL is
provided. As well as containing easy-to-follow
instructions and examples of how to get the
best out of the BIT STIK system, information
on writing programs and using drawings in

Robocom Ltd.,

CIL Trading Estate,

Fonthill Road,

London N4.

Telephone: 01 2633388

colours. To a

:o 16

drawing of tl

grams, etc., a lock LCD multimeter
i angle and two grid The Model 3T is a battery operated 3!4 digit

trols allow fast and easy operation whilst

small size, robust construction and long bat-
tery life make the 3T truly portable. Supplied
complete with battery, test leads and instruc-

62, Curtis Road,

Miniature UHF tuned preamplifier

The TE/435P is a tiny UHF preamplifier that
employs a tuned line on the input, and a twin
chamber helical filter on the output. In view

h VHF

idwidths

SERVICES

TO

READERS

I EPS print set vice

Many Elektor circuits are accompanied by printed circuit
designs. Some of these designs, but not all. are also available
as ready-etched and pre-drilled boards, which can be ordered
from any of our offices. A complete list of the available
boards is published under the heading 'EPS print service’ in
every issue. Delivery time is approximately three weeks.

It should be noted however that only boards which have at
some time been published in the EPS list are available; the
fact that a design for a board is published in a particular
article does not necessarily imply that it can be supplied by

j He^Rys

/-—x /'■“"'I f~\ / — Nl I UNITED KINGDOM

ELECTRONIC COMPONENTS

600 MHz Digital Frequency Counter

| Technical gucric/

Please enclose a stamped, self-addressed envelope, readers
outside UK please enclose an IRC instead of stamps.

Letters should be addressed to the department concerned
TQE (Technical Queries). Although we feel that this is an
essential service to readers, we regret that certain restrictions
are necessary :

1. Questions that are not related to articles published in
Elektor cannot be answered.

2. Questions concerning the connection of Elektor designs
to other units (e.g, existing equipment) cannot normally
be answered, owing to a lack of practical experience
with those other units. An answer can only be based
on a comparison of our design specifications with those
of the other equipment.

3. Questions about suppliers for components are usually
answered on the basis of advertisements, and readers can
usually check these themselves.

4. As far as possible, answers will be on standard reply

^^AT8D|

| .88881

[EDA SPARKRITE LIMITED 82 Bath Street. Walsall. West Midlands. WS1 3DE England Tel (0922)614791

NAME

ADDRESS

I ENCLOSE CHEQUE(S)/POSTAL ORDERS FOR

£ KIT REF

CHEQUE NO.

24 hr Answerphone

PHCN E YOUR ORDE R WIT H ACCESS/ BARCLAYCARD A
SEND ONLY SAE IF BROCHURE IS REQUIRED

VOYAGER £59.95 £119.90

| MAG I DICE | £9.95 | £19.90 |

PRICES INC. VAT. POSTAGE & PACKING

i i \ a . j j ;

STARRING: Resi, the irresistible resistor Transi, the (not very) active component

AND CO-STARRING : Cappy the capable capacitor

BANISH THE MYSTERIES OF ELECTRONICS !

Excitement, entertainment, circuits. Complete with printed circuit board and Resimeter.

-from Elektor

Elektor Publishers Ltd., Elektor House, 10 Longport, Canterbury CT1 1PE, Kent, U.K.

Tel.: Canterbury (0227) 54430. Telex: 965504. Office hours: 8.30 - 1 2.30 and 1 3.30 - 1 6.30.

mopuri

Maplin Electronic Supplies Ltd
P.O. Box 3, Rayleigh, Essex.

Tel: Southend (0702)
552911/554155.

Atari 400 with 16K RAM( AF36P) £345
Atari 400 with 32K RAM(AF37S)£395
Atari 800 with 16K RAM (AF02C) £599

THE CHOICEST GEMS OF ATARI SOFTWARE FROM MAPLIN