Make your Micro the life and soul of the party
By MIKE COOK
WITH the festive season once more upon us, here is a little add-on
that will make your BBC Micro a welcome guest at any party. As
well as being available in kit form, it can also be purchased
ready-built for the less intrepid among you.
It is a little-known fact that I used to be a disc jockey in
my early youth. The truth of the matter was that I enjoyed wiring
up all the amplifiers, turntables and tape decks.
Such was the mess of wiring spaghetti that I was the only one
who would dare operate it. It also fell to my lot to wire up the
disco lights and to flash them on and off. This was in the Dark
Ages when such a thing was a novelty.
However, the control of disco lights was one of the first things
I automated. A natural progression was to make the music control
the switching of the lights.
When lights are fixed, the results produced do not get very
complicated, but the overall effect adds great atmosphere to any
party.
I have always had the feeling that something better could be
done and for a long time have wanted to do something similar with
a computer.
However, not until the advent of the BBC Micro has a computer
incorporated sufficient facilities to make this project worthwhile.
This month's exercise is therefore an add-on which produces
patterns related to music or, if you have built the solid state
relay (see the September issue of the Micro User), controls conventional
spot lights.
If you look at music on an oscilloscope you will see a complex
ever-changing waveform. In fact it looks a mess. This is because
when you use an oscilloscope you are looking at the signal as
a plot of amplitude against time, known as a "time domain"
view of the signal.
If, however, you take a short section of music (say half a second)
and make a plot of the amplitude at each specific pitch you have
what is known as a "frequency domain" view of the signal.
This can be displayed using an instrument known as a spectrum
analyser. (No, it doesn't work out what is wrong with rubber computers!)
The frequency domain view of a signal (or, as we say, its spectrum)
shows, at any instant, the composition of the sound in terms of
pitch. For most music, this means which instruments were playing
at the instant the spectrum was taken.
So all we need for a picture of the sound is to feed our hi-fi
into a spectrum analyser. As these cost around £3,000 it
is not a very practical proposition.
We can make a crude simulation of a spectrum analyser by filtering
the sound so that only a restricted band of frequencies is produced.
This can be done with a band-pass filter.
If several band-pass filters are used, all at different frequencies,
we can build up an approximation of the spectrum of the sound.
Figure I illustrates the principle: the sound is split up into
four frequency-bands and the output of each filter is measured
in order to build up the spectrum.
Figure I:
Block diagram of our spectrum analyser
In terms of our ultimate goal, the samples can then be used
as parameters to create patterns on the computer's display. The
problem is to build four filters which will cover the same range
of frequencies as the music.
A simple filter can be made with a capacitor and resistor. (See
Figure II.) A capacitor does not allow DC to flow through it,
only AC. The higher the frequency of this current, the more easily
it flows.
Figure II:
A low pass filter
In jargon we say the capacitor has a high impedance for low
frequencies and a low impedance for high frequencies. Therefore
a capacitor can be thought of as a frequency-dependent resistor.
So you will see in Figure II that if the input is a high frequency
the output will be small, because the impedance of the capacitor
is much smaller than the impedance of the resistor.
Conversely, for a low frequency input, the capacitor has such
a high impedance it can be considered effectively removed. So
the output will be the same as the input. Because low frequencies
are passed through unchanged and high frequencies are attenuated
(reduced) this is known as a low-pass filter.
See if you can work out why the circuit in Figure III is a high-pass
filter.
Figure III:
A high pass filter
By combining a high-pass and a low-pass filter we can produce
a band-pass filter. Unfortunately, the circuits in Figures II
and III are not usable in practice, due to the impedances of the
input and output devices. We have to use amplifiers to reduce
the effect of "loading impedances" as they are called.
When an amplifier is included in a filter it is called an active
filter (not "Active Fitter" as one Ministry of Defence
typist once titled a report of mine).
The simplest way of making an active filter is to use an operational
amplifier (or op-amp).
Unfortunately, most op-amps need a split power supply, that
is, one with a positive and negative voltage as well as a zero
volt rail. This is available from the auxiliary power supply socket
of the BBC Micro.
However, if you have a negative supply rail there is a danger
that the output of the op-amp will go negative. This in itself
is no problem, but if it is connected to the analogue input of
the computer it will damage the A/D converter chip.
A better solution is to use an op-amp which does not need a
negative supply. These little-known amplifiers work with input
currents (as opposed to input voltages as used by the more conventional
op-amps). They are known as Norton amplifiers and the symbol is
shown in Figure IV.
Figure IV:
A Norton amplifier
These current-differencing amplifiers have been designed to
work off single-supply voltages over a wide range (4 to 36 volts)
and as such are ideal for use with digital electronics.
Like any op-amp they have two inputs. The first is known as
the positive input, because any current flowing into it will be
amplified and cause the output to rise. The other input is known
as the negative or inverting input - when current flows into it,
the output falls (goes in a negative direction).
The trick of using op-amps is to arrange the feedback circuit
(connecting the output to the input) so that these currents balance
out and a stable output is achieved for a stable input. This is
not nearly as difficult as it sounds.
Consider Figure V. When no voltage is applied at point P, the
output voltage will rise due to the current flowing into the positive
input. However, as it rises it makes more current flow into the
negative input, which tends to force it down. This reaches equilibrium
when the same amount of current is flowing in both inputs of the
op-amp.
Figure V:
An inverting amplifier
When the voltage at point P starts to rise, more current flows
into the negative input causing the output to fall, thus reducing
the amount of feedback current, until again an equilibrium point
is reached.
But this time, some of the current in the negative input will
come from the signal, and some from the output (feedback). As
the output falls when the input signal increases, this configuration
is known as an inverting amplifier. You should be able to see
that the gain of such an amplifier is controlled by the values
of the resistors in the circuit.
If you understood that on the first reading you are doing very
well. If not, go back and go through it again and sketch yourself
little graphs of the input and output - I find it often helps.
The Norton amplifier is available in a chip called the LM3900N.
There are four of them, all in the same package.
Now, promise me you won't stop reading this article, and go
and have a peep at Figure VI. Pretty impressive hey? Well actually
it's not that complicated, because as we need four filters, the
basic circuit is just repeated four times. If you can bear it,
have a closer look and you will see that the same circuit really
is repeated. Only the component values differ.
Each strip of circuit corresponds to a separate filter/input
circuit. The sound signal comes from the input socket. This is
a 7-pin DIN socket wired up in exactly the same way as the cassette
input socket of the BBC Micro. In this way you can use your cassette
recorder to play the music (let's face it, it never was much good
for programs!).
If you have a stereo input to this circuit, one channel is controlled
by VR1 and the other by VR2. The input socket is wired up to allow
this but if you only have a mono system the two halves should
be joined together by making link LI.
If you are feeding the unit from the output of a hi-fi amplifier
you can wire it up in parallel with the loudspeakers. If you are
using stereo, make sure the common lead goes to earth. To be on
the safe side, only wire up one speaker wire on one side and two
on the other. If both channels do not work you can then safely
swap wires until they do.
As the unit has a very high input impedance it will not place
any loading on the sound source. A signal of about two volts RMS
will be needed to drive the unit, but the sensitivity can be increased
by software as we shall see later.
Construction can be done in a number of ways. I made the original
prototype on Vero board, with flying leads connecting it to the
computer. However, as it is a circuit which quite a lot of people
might be interested in, I have produced a printed circuit board.
This is available as a kit, with all the components.
The board uses a 15-way D-type plug and just plugs into the
analogue input socket. Two pillars are used at the other end of
the board to support it. If you want, however, it may be made
in a box to make it look neater, although there is nothing nicer
than a well-laid-out circuit.
Construction of the circuit board involves finding the component
value from the components list and inserting the appropriate component
in the correct place on the printed circuit board. For example,
C7 is a 0.1 uF capacitor and should be placed at the component
position marked C7 on the printed circuit board.
Note that the resistors only have their number on them, they
don't start with 'R'. The diodes need to be placed the right way
round. This is shown by the band at one end. Again, match it up
with the markings on the printed circuit board.
Apart from the ICs the only other components that need to be
the correct way round are the electrolytic capacitors C1 to C4.
A small white stripe and a + should be at one end. Put this end
to the hole marked + on the board.
The four band-pass filters have their frequencies evenly distributed
over the range used for the fundamental notes in most music. The
exact frequencies are l00Hz, 220Hz, 560Hz and l000Hz. Now, hands
up all those who think I have made a mistake. Are those frequencies
evenly distributed? Well in fact they are, as far as the ear is
concerned.
This is because the response of the ear is very nearly logarithmic
and so these frequencies sound evenly spaced. If you plot them
out on a logarithmic scale you will see there is the same space
between them.
Now - hold onto your hats - here is the simple version of how
the circuit works. The signal is fed into the input of IC la through
Rl and R2 resistors and the capacitor C10. The feedback is provided
by R3 and C9.
As we saw earlier, capacitors act as frequency-dependent resistors,
so the amount of signal flowing into the amplifier both from the
source and the feedback are frequency-dependent.
At low frequencies, the capacitors can be considered as open
circuits. So no signal gets to the imput of the amplifier.
At high frequencies, the capacitors are a very low impedance.
So although the signal gets through to the input,
there is lots of negative feedback and the output does not move
very much.
Only at some intermediate frequency is there any output. The
resistor R4 provides a steady trickle of current, known as the
bias current. This bias supply is provided from a potential divider
formed by R41 and R42 and is used to prevent the resistor values
becoming too high.
Once the signal has been filtered it is passed to the next stage
(1C 1b) which is known as a peak detector.
Capacitor C5 removes any DC component from the signal and feeds
it into the input of the next op-amp. Current is also being supplied
from R5 and is acting as the bias for this stage. The output of
the op-amp is then passed through a diode.
A diode acts as a one-way valve, allowing current to flow from
the output of the op-amp into the capacitor, but not allowing
the capacitor to discharge back through the amplifier.
Therefore the diode only conducts when the voltage at the op-amp
output is higher than that on the capacitor.
Thus the voltage on the capacitor represents the peak voltage
from the amplifier. Resistor R8 controls how fast this peak voltage
builds up and R9 and R10 allow this voltage to decay. So the voltage
fed into the analogue input socket represents the loudness of
the music at that instant. This is known as following the envelope
of the music.
All we have to do now is to feed our music signal into this
circuit and use the values from the analogue input port to draw
our patterns.
As usual I have written a program to illustrate how this is
done, which gives very good results. But also, as usual, it is
only a starting point for you to experiment with.
The signal being fed into the analogue input port is larger
than I required, so it is cut down to size with the variable SCALE%.
There is also a DC offset voltage on the input. This is removed
by the variable OSET%. These values can be changed to increase
the sensitivity of the circuit if the input signal is low.
To test the circuit, you only need to type in lines 10 to 180
(omitting line 150) and lines 950 to 1140.
This displays a bar graph of each input channel so you can make
sure that it is moving over a good range. If not, adjust the volume
into the filters either from the source, or by adjusting the variable
resistors. Alternatively, alter the two variables already mentioned.
The procedure SAMP samples the appropriate input channel and
scales the result so that it is between 0 and 40. This value is
returned in the variable V%. The procedure BAR then prints a line
of stars proportional to the size of the input.
The display looks very much like the rows of flashing lights
you get on some modern hi-fi systems.
Having got that working, you can now type in the rest of the
program (not forgetting to add line 150). This uses the sound
in three ways to produce moving patterns which seem to be evocative
of the music being played.
Each different procedure is ended by pressing a key. If this
were being used at a party you might like to end each display
procedure after a certain time has elapsed - not forgetting to
zero the time at the start of each procedure . The lines that
need changing are 670, 930 and 1360.
In order to speed up the patterns, use is made of a lookup table,
X%(N) and Y%(N), to convert the sound input values to screen co-ordinates.
These are pre-calculated in the procedure LOOKUP, along with the
initialisation of most constants used in the program. Wherever
the program is sensitive to speed I have used integer variables.
The procedure POLYGON uses the value of each sound channel to
control the position of each corner of a quadrilateral. This changes
shape and distorts in sympathy with the music.
A record of each shape is stored in the array OX% and OY% (old
X and old Y) and when there are N% shapes displayed the oldest
of them is removed to make way for a new one. This prevents the
screen becoming cluttered with remnants of old patterns.
The display responds rapidly to the change in music and is very
effective, especially if the music contains a lot of variation
and sudden chords.
The second display is more subtle. The pattern woven is much
more intricate and changes more gradually. The patterns are very
beautiful and seem to follow the mood of the music. This type
of display is best suited to contemplative music.
The pattern is formed by marking a point on each axis which
corresponds to each channel of the sound. Then every point is
joined up to every other point. This is repeated for each quadrant
to give a pleasing symmetrical effect. Because the pattern is
so complex, only two are displayed at any one time.
The final display is a rapid response display. Each quadrant
of the screen is devoted to one channel. A triangle is drawn whose
size and displacement from the centre are controlled by the sound-level
on each channel.
Again, like the first procedure, the variable N% controls the
number displayed at any one time.
In all the above procedures, the colour of the display has been
chosen at random, as this seemed to give the most pleasing effect.
But you could experiment with the choice of colour being made
by the level on one of the channels.
I have found that, for a vigorous display, the patterns should
be roughly predictable. I seem to get the greatest enjoyment from
this when I can use my knowledge of the music to predict when
the patterns will change.
I am more than pleased with the results given by the circuit.
It has a very hpynotic effect - you could just stare at it all
day. My wife made the point that it helps you listen to the music
and gives you a focal point to concentrate on.
School music teachers might find it stops a class getting restless
and helps the children pick out individual instruments when listening
to music.
Unfortunately I cannot tell you how it fares under actual party
conditions for two reasons. Firstly, this is being written before
Christmas, so there are not many about, but mainly because I never
seem to be asked to any. It might be the soap. I must start using
some!
Next month we tell you how to cope with a slipped disc. Merry
Christmas!
Body Build Pack No. 8:
C1-C4 luF Tantalum capacitor,
C5-C8, C17 0.luF paper capacitor,
C9 C10 4700 pF polystyrene capacitor,
C11 C12 3300 pF polystyrene capacitor,
C13 C14 1000 pF polystyrene capacitor,
C15 C16 470 pF polystyrene capacitor,
D1-D4 1N4148 Diode,
IC1 IC2 LM3900N Norton op-amp,
VR1 VR2 10 Skeleton preset pot,
7 pin DIN socket,
2 14 pin DIL IC sockets,
15 way D-Type plug,
2 16 mm plastic pillars,
2 self-tapping screws,
printed circuit board.
RESISTORS 5%
R1,R31 1M5
R2,R7,R27,R17,R32, R37 33K
R3,R4,R21,R33,R34 3M3
R5,R24,R25,R15,R35, R23 6M8
R6, R9, R10, R16, R19, R20, R26 1M
R29,R30,R36,R39,R40 1M
R8, R38 5K6
R28,R18,R41,R42 10K
R11 2M2
R13,R14 4M7
R22 18K
R12 11K
Figure
VI: The sound-to-pattern converter
What the Beeb Body Building packs contain
Upgrade your BBC Micro with these Beeb Body Building Packs.
• Packs 1 and 2: Really use your User Port! These two packs
let you safely and simply join your micro to the outside world.
The best way to learn about interfacing.
• Pack 3: Give your BBC Micro real data-processing muscle by
building yourself a dual cassette system - the way Acorn said
you couldn't.
• Packs 4 and 5: These packs allow you to control mains equipment
in complete electrical isolation from your micro - the only safe
way to do it. Pack 4 is rated at 4 amps, pack 5 at 10 amps.
• Pack 6: All the electronic components you'll need to construct
the Micro User light pen, with its special screen-sensing circuit.
• Pack 7: Two interesting experiments that should appeal both
to schools and the amateur scientist: investigate the behaviour
of a pendulum and study capacitor discharge.
• Pack 8: Construct this versatile sound to pattern converter.
Its many applications include running a light show with your BBC
Micro. It will also make the micro a resident fixture in many
music classes.
