THE NOISE GENERATOR

COOKBOOK
A complete guide for the the do-it-yourselfer showing
how to design and build noise generators for
analog synthesizers, electronic percussion,
audio testing and more

By Thomas Henry
 The Noise Generator Cookbook
by Thomas Henry

©2003, 2009 Thomas Henry

The entire contents of this book is copyright, and may not be reproduced in any form
without written permission of the copyright owner. All rights reserved. The infomation
contained herein is believed to be accurate but neither the author nor the copyright
owner assumes any responsibility or liability for any inaccuracies or printing errors. The
information contained herein is presented for illustrative purposes only. Neither the
author nor the copyright owner assumes any responsibility or liability for any
consequences resulting from the use or misuse of any information contained within this
book.

Printed in the United States of America

Second Edition

Cover image by Dora Mitsonia

This and other Thomas Henry books are published by

Magic Smoke Electronics

www.magsmoke.com
 About the Author

Thomas Henry is the author of over 100 articles and several books on the subjects of electronic
music, microcomputers, astronomy, and caves. While in school he helped form the East Side
Pharaohs (the Midwest's zaniest band) in which he played guitar and bass for nearly 18 years. After
attaining an M.A. in mathematics, he taught at the collegiate level for ten years. His leisure time
activities include bird watching, caving, amateur astronomy, magic, road trips to national parks, and
flower gardening. The bat is his favorite mammal.

Editors Note: In the winter of 2005, I was approached by Scott Stites. Scott is an avid electronics
hobbyist who had been working with Thomas Henry testing and prototyping several circuits, some
of which were slated to appear in a new book. Unfortunately, Thomas had recently decided to pull
the plug on his business, “Midwest Analog.” The decision was a good one for Mr. Henry, and it
allowed him to pursue one of his other loves, teaching, full-time. Unfortunately, this left no
distribution outlet for the Thomas Henry “Cookbook” series. These books had reached near-
legendary status among synthesizer hobbyists, and many people wondered if they were going to
join the ranks of the myriad other out-of-print books treasured by synth fans. Scott had followed my
posts on the Synthesizer DIY mailing list during a particularly interesting discussion about copyright
laws and whether copying was okay in cases where publications (like the Thomas Henry books)
were no longer available. Despite all logic, Scott thought I might be a good choice to carry on
distributing the books. I contacted John Mahoney, with whom I had worked before, and asked if he
wanted to join in on a possible venture to publish Thomas' work. The reply was an immediate and
enthusiastic “yes!” And so, Magic Smoke Electronics was born. Although we have plans to create
additional books and circuit kits, if we do nothing other than provide an outlet for Thomas' existing
body of work, we'll consider Magic Smoke a success. This book has been published with an
absolute minimum of changes from the original. A new font, and references to defunct web links
have been removed or updated. Otherwise, these books remain true to the original works, and are a
marvelous source of knowledge and inspiration. John and I would like to thank Mr. Henry for
allowing us to continue publishing these books, and we hope you, the reader, will enjoy them as
much as we have.

Tim (Servo) Parkhurst

April, 2009
Contents

Preface v

Resources vi

Design Collection 1

1 - General Principles 1
2 - Transistor Noise Sources 3
3 - Using Negative Feedback 6
4 - Oscillator Bank Noise Source 8
5 - CMOS Digital White Noise Source 10
6 - Pink Noise Filters 13
7 - Adding Voltage Control 18
8 - Tone and Equalization Options 23

Bibliography 25
Preface

While working on this guide, two conflicting ideas kept running through my mind. First, I felt that
these were some of the neatest circuits I had ever stumbled upon. Some were outlandishly original,
like using white noise to frequency modulate a VCO, while others were updated and improved
versions of older circuits. In both cases, though, I was very happy with what was falling together. As I
continued to experiment at the work bench (many rewarding hours!) it was becoming clear to me
that this would end up being one of the most useful handbooks I had ever written.
At the same time, though, I knew that it would be a hard sell. After all, isn’t noise boring? And weren't
noise generator circuits all worked out years ago? Convincing others that the material herein would
indeed be lively, unique and musically useful was going be a real challenge.
Well, the fact you're reading this Preface now indicates that I got through to at least one other
person! As a matter of fact, noise generators are anything but boring, and there are still tons of new,
unique circuits just waiting to be invented. If nothing else, the fact I've devoted over twenty years to
the subject suggests there might be something intriguing about it after all. (Or perhaps it indicates
I'm easily amused - see what you think). Ever since 1980 or so when I first discovered how to unlock
the power of the noise generator inside the marvelous SN76477 complex noise chip, I’ve been
fascinated with the field. My first experiments led me to come up with unusual methods of adding
voltage control, and then later I became interested in percussion synthesis. Neither one of these
topics had been very well developed yet, leaving things wide open for original research. And so the
greater portion of my electronic music career, such as it is, has been devoted to noise generator
circuits.
Noise is distinctly non-boring, if you keep an open mind. Many exotic effects are possible with more
advanced circuits, and the good news is most of these are very easy to build. Contained within these
pages, then, are more than one dozen of the best circuits kicking around. You’ll find uses for them in
modular analog synthesizers, drum units, audio testing and more.
So, yes, the hours and hours of researching, designing, prototyping, and testing have sometimes
been a tedious struggle. But that's nothing compared to the hard part remaining: convincing others
that there's more to noise generators than “hiss." Despite the long hours, I've had fun writing these
results up. I hope I've gotten the message across and that you find it an eye opening guide.

Thomas Henry
North Mankato, Minnesota
November 2003

-v-
Resources

The literature of noise generators is enormous, as one peek at the Bibliography beginning on page
25 will convince you. The trouble is, most of these references are from the “old days" of audio and
analog synthesis. A lot has happened in the electronics world since then. On the one hand, a
number of wonderful chips for noise generators are no longer easily available. This would include
LSI wonders like the MM5437 and SN76477. And the fantastic 2N2712 transistor, unequaled in its
ability to spit out broadband noise, seems to have faded away as well.
But on the bright side of things, we now have access to a wide range of better op-amps, capacitors
and so forth, and at bargain rates. The modules we build nowadays generally have improved audio
characteristics compared to the circuits of yesteryear.
As you begin your journey into the realm of noise generators, you may want to take a look at what
other people have done in the past. The Bibliography will point you in the right direction. Electronic
music circuit design has progressed quite a bit in the past three decades, so some of the earlier
books and articles may show their age. Nonetheless, you`ll frequently pick up a neat wrinkle or two
by looking over another person’s creation.
One of the more exciting developments is the use of personal computers to aid in the design of new
circuits. For example, an ordinary spreadsheet program can really streamline the calculations
involved in bandwidth limiting (p. 4), oscillator frequencies (p. 8), and more.
Another modern contribution to the design process is the availability of inexpensive or even free
SPICE simulation packages. I've found the free student edition of CircuitMaker to be especially
useful when writing this handbook. For example, the Bode plots on pp. 14 - 17 go a long way toward
explaining how the filters perform.

- vi -
General Principles 1

What is noise?
The type of sound we have in mind here is
akin to the static produced by an FM radio
when parked between stations. (With the
advent of digital tuning, this description is
apt to become obsolete in the near future!)
In very loose terms, noise is a complex
waveform with components of every
frequency, each attaining a completely
random amplitude at any give instant.
Figure 1 shows how it might appear on an
oscilloscope. Noise can span all Figure 1
frequencies from subsonic to supersonic, White noise is a complex waveform containing com-
but since readers of this guide will be most ponents of every frequency.
interested in musical applications, we’ll
restrict our attention to audio frequencies
only.
As homely as it may seem, defining noise precisely is exceedingly difficult. The trouble arises when
trying to get a handle on the term “random." This could easily take us deeply into the realm of
mathematics, statistics and philosophy, so let`s just assume we all have an intuitive understanding
of what it means when it’s said noise is a signal of random frequency and amplitude.
There are several types of noise, even though each is a mixture of random frequencies. The
simplest to comprehend is white noise. In this case, every span of frequencies of the same
bandwidth yields the same electrical power.
A chart will make this clear. The graph in
Figure 2 is of the white noise spectrum. As
plain jane as it may seem, it tells us quite a
lot. The horizontal axis spans all audio
frequencies from 20Hz to 20kHz, while the
vertical axis measures the amplitude of
Amplitude

each spectral component (normalized to

some convenient value). Notice that the
resultant graph is a horizontal straight line.
In other words, each component frequency
of white noise attains the same average
amplitude, hence the same average power
1kHz 10kHz 20kHz level as its neighbor. To put it succinctly,
Frequency white noise is characterized by equal
power per unit bandwidth - garden variety
linear relationship if ever there was one.
Figure 2 For example, the amount of electrical
Power spectrum graph of white noise. power between 10Hz and 20Hz will be
equal to that in the range 20Hz to 30Hz,
30Hz to 40Hz and so on. Incidentally, the
notion of “summing” power across a band of continuously varying frequencies is standard fare in the
integral calculus. That's beyond the scope of this guide, so we'll just assume that it is indeed possible
to compute power in the fashion just mentioned.

-1-
The term white, of course, was originally applied to this type of noise by analogy to light. Just as white
light is a mixture of all visible colors, so is white noise an amalgam of all possible audio frequencies. But
when you bring the human brain and ears into the picture, things get interesting.
Our sense of hearing behaves in a logarithmic fashion (and in more ways than one). Every musician
knows this, at least intuitively. If the frequency of a tone jumps from 100Hz to 200Hz, we perceive a
change of one octave. But a jump of identical width, from 200Hz to 300Hz is not an octave. Instead, our
ears (or brains or whatever) require it to go from 200Hz to 400Hz. Thus, a linearly increasing sequence
of octaves is actually an exponentially increasing sequence of frequencies. Recalling that the logarithm
function is the inverse of the exponential, we can also say that a linearly increasing sequence of
frequencies represents a logarithmically increasing sequence of octaves. In short, our sense of pitch
perception is logarithmic.
Another way of thinking of this is that exponentials represent an expansion of scale, while logarithms
represent a contraction. Thus, our sense of hearing contracts the entire audio range of all frequencies
from 20Hz to 20kHz into the more manageable span of about ten octaves.
This has immediate consequences with regard to white noise: white
noise doesn't sound white! To understand why, consider this Frequency
simple example. Refer to Figure 3. Notice that the lowest (most
bassy) octave runs from 25Hz to 50Hz, a mere 25Hz wide. But eight 25
octaves up, the span is from 3200Hz to 6400Hz; this treble region is 50
100
3200Hz wide now. Recalling that the amount of electrical power is 200
evenly distributed in white noise, we see that there is great deal more 400
energy in the treble band than in the bass band. The upshot is that 800
white noise sounds very shrill or hissy to our ears; the high end 1600
seems unnaturally emphasized. 3200
6400
Since the width of each successive band doubles (from the definition 12800
of octave, of course), the amount of power in each band doubles for
white noise. A doubling of power is represented by +3dB. Hence, we
can say that the normal audio response of white noise slopes upward Figure 3
at a rate of +3dB/octave. In order to create noise which sounds flat, Note how the width of
that is with equal power in each octave, we need to pass white noise each band doubles as you
through a low pass filter with a -3db/octave characteristic. The move from octave to
negative slope undoes the normally rising response of white noise. octave.

This result is referred to as pink noise. As we’ve just seen, it has a rather surprising property: pink noise
sounds white! The key here, of course, is the verb “sounds." White noise is white in a laboratory sense
of the word, while pink noise sounds white due to the way our sense organs behave. Now the study of
psychoacoustics is a fairly deep affair, and a number of details have been glossed over here.
Nonetheless, for the purposes of designing useful electronic music circuits we can summarize things
intuitively by stating pink noise sounds the smoothest or most even. Incidentally, in a number of
instruments like snare drums you might still want to go with white noise, to deliberately give extra punch
to the high end. But for testing hi-fi equipment or the acoustics of a room, pink noise is the best choice.
And in case you're wondering: in the visible light spectrum, red has the lowest frequency of all the
colors. Mixing red with white gives pink. In a similar fashion, if white noise is processed to give more
emphasis to the low end (or equivalently, less emphasis to the high end as described above), then it's
not unreasonable to label it pink. Some books call this 1/f noise. This refers to the fact that the power
level falls off inversely with respect to increasing frequency. As we saw above, it does so with a slope of
-3dB/octave.
To wrap things up, note that there is such a thing as brown noise. In this case, white noise is processed
by a -6dB/octave lowpass filter. The name is a bit of a mathematical joke, not referring to color, but to
Brownian motion which is a concept arising in the study of probability theory. As it turns out, this type of
random noise can be described by Brownian motion principles.

-2-
Transistor Noise Sources 2

Noise can be generated by either digital or analog means. A digital approach is explained in Chapter
5. Here we want to look into analog methods for creating white noise. One obvious (but not very
elegant) attack is simply to exploit the inherent noisy nature of inexpensive op-amps. Merely gang a
number of these together in series; each stage amplifies the noise of the previous and adds in some
of its own. Six or more stages might be needed to obtain the desired output level.
Another approach is to mix the outputs of a half dozen or more oscillators, each operating at a
different frequency. It's important that the oscillators not be phase-locked if a truly random result is
expected. Since the sound is somewhat pitched, this technique is perhaps not suitable for
generating white noise (in which all frequencies are represented equally). But it is fantastic for
electronic percussion circuits, among other things. You’ll find a practical design in Chapter 4.
By far the most elegant, inexpensive and highest quality method is to take advantage of an
interesting property of transistors. In particular, if the base-emitter junction of an NPN transistor is
deliberately reverse-biased - making it behave as a Zener diode - then an amazingly broad range of
frequency components is generated at the emitter. This "shot noise" is created as soon as you
exceed the reverse breakdown base-emitter voltage. Many writers refer to this as avalanche mode,
which does in fact conjure up the image of millions of electrons tumbling down a hillside, bumping
into each other noisily.
Figure 4 shows the general idea. Q1 is
reverse-biased from the +15V power
supply. R1 not only restricts the current
flow to a safe level but also acts as a load
resistor for the audio output. Incidentally,
the transistor really is acting like a Zener
diode now, and attempts to clamp the DC
voltage at a fixed level. For common
transistors like the 2N3904, this is around
9V or so. And obviously, we're using Ql as
a diode here, so the collector lead is simply
ignored.
The noise that's generated in the process
is low level, on the order of 10 to 25mV pp
for most transistors. It rides on a DC bias,
mentioned above. So the output must be
AC coupled with capacitor C1 before
passing it on to any remaining audio
circuitry. Incidentally, the signal may be
asyrnrnetrical about the base line. This
might matter if you were using the noise
source to generate random numbers for an
encryption circuit, but is usually of little
consequence in music applications.
And now a word about Q1. Most any NPN transistor will produce reasonable white noise when used
in this fashion. The venerable 2N3904 works well, for example. But it is true that some transistors
are inherently “better" than others in this regard. Often, those with poorer specs make the best
noise sources. For example, the punk 2N2712 has a lower base-emitter reverse breakdown
voltage than more modern parts and that's exactly what we want here. It's a lousy component by
today's standards, but it sure makes great noise! In any event, when getting ready

-3-
to build a noise generator circuit, you may want to breadboard it first and try a number of transistors
to see which produces the “smoothest" or the greatest quantity of noise.
In this type of circuit, the amplitude of the output is far too low for most uses. Figure 5 shows some
practical designs which boost the white noise up to a usable level for synthesizer and other audio
applications.
In part (a), the noise transistor is followed by a common emitter amplifier configured around Q2.
Note how R3 is used to implement feedback; this tends to keep the biasing more stable. The input
impedance of the amplifier is low, but the gain of the circuit is quite good at about 200. This is a
reasonable compromise, and the entire white noise generator gives great results. The output
amplitude will differ depending on noise transistor Q1. For a 2N3904, it might be 4Vpp, while the
trusty 2N2712 could give as much as 7Vpp.
One possible modification would be to roll off the low end should you be after a more shrill hiss.
Simply decrease C1 and C2 in value and lower frequencies won't be passed as easily as higher
ones.
The real advantage to this circuit is the fact it only requires a single voltage. But if you have the luxury
of a bipolar power supply, then the circuit of part (b) is a good choice. In this case, the noise transistor
is buffered by a non-inverting amplifier. The nice thing here is that such an op-amp circuit has an
extremely high input impedance. But for best stability, we override this by paralleling R3 with the
input; this still leaves us with a decent 100K impedance
R1 and R4 set the gain, with the usual formula of:
R4
Gain = +1
R1

With the values shown, the noise output averages 3Vpp for the 2N3904, and 4Vpp for the 2N2712. If
you need more output, it's perhaps better to follow this stage with an additional one, rather than
increasing the gain of IC1 up to some ridiculous value.
One nice thing about using an op-amp with a bipolar power supply is the increased headroom it
affords. For example, it is possible that the transistor amplifier of Figure 5(a) might clip with a
2N2712, since this part has such a hot noise output. That won't be the case with the op-amp version.
Incidentally, since the purpose of the circuit is to create white noise, the noise figure of the op-amp
really doesn't matter. A cheesy one like the ever-popular 741 makes sense. On the other hand,
better op-amps do have a wider bandwidth which may be a concern here. A TL071, for example,
would probably give a crisper high end response should that be an important design criterion.
Along these lines, with the op-amp circuit of Figure 5(b), it's easy to tailor the frequency response for
special purposes. A capacitor in parallel with feedback resistor R4 will roll off the high end, while a
capacitor in series with R1 will attenuate the low end. Just employ the usual formula:
1
f=
2 •R•C

to set the -3dB points as desired.

-4-
-5-
Using Negative Feedback 3

Here's a neat technique for boosting the output of`a noise transistor. Even though the idea dates
back to at least 1972, it seems to have slipped through the cracks and isn't so well known nowadays.
However it did resurface recently in the November 2003 issue of Nuts & Volts Magazine. (See the
Bibliography for details.)
The general idea is that we place the noise source in the negative feedback loop of a transistor
amplifier. This has the twin advantages of being very economical (low parts count) as well as
yielding a very stable performance. Best of all, the frequency response is extremely smooth and
even throughout the entire audio band. Let's see how it works.
Refer to Figure 6(a) which shows just how simple it can be. Q2 is configured as a commonemitter
amplifier. But notice how the noise transistor Q1 is wired within a negative feedback loop from the
collector to the base. This arrangement establishes a fairly stable base bias, and R2 does double
duty by limiting the current flow through the reverse biased noise transistor. Finally, C1 bypasses
the feedback resistance, eliminating AC feedback at signal frequencies. You really can't get any
more elegant than this!
The output of the circuit will be around 3Vpp for a typical 2N3904, and perhaps 4Vpp for a good
2N2712. The DC bias of the output will be in the neighborhood of +8.5V, so you'll want to level shift
things or block the offset with a capacitor. Depending on any following circuitry, you may also need
to buffer the output so that it isn't loaded down. (Loading could change the frequency response or
overall volume.)
This idea of using negative feedback can be extended to a more complex but more versatile noise
generator, as shown in part (b). The nice thing here is that the entire circuit is DC coupled, meaning
that the response extends way down into the bass frequencies decently. For that reason, this could
be a good front end for a random control voltage generator. Incidentally, the essential concept here
is due to Walter G. Jung, and appeared in his article “Generating Complex Waveforms," Electronic
Engineering, May 1972. I haven't actually seen the original article, but caught it on the rebop when it
was reprinted in Thomas Wells` The Technique of Electronic Music, (New York: Schirmer Books,
1981), p. 62.
Q1 and Q2 are arranged in a fashion similar to that in part (a). But notice that we take advantage of a
bipolar power supply now with the common-emitter amplifier straddling both lines. Then the output
on the collector of Q2 is further buffered by the common-collector amplifier configured around Q3.
This type of amplifier adds no voltage gain (in fact, it reduces it a trifle), but has a lower output
impedance which minimizes the effects of loading.
Finally, the output is further processed by IC1 which level shifts the white noise appropriately. With
R7 near the center of its rotation, you can trim up the bias to 0V quite nicely. With a 2N3904 as the
noise transistor, the output is about 1Vpp, while the 2N2712 gives around double that. Further
observe that changing noise transistor Q1 may alter the DC bias slightly. But there's plenty of range
in the trimmer to take up the slack.
And obviously you can attenuate or boost the output as desired, simply by changing op-amp
feedback resistor R5.
In both of these circuits, a time constant is involved due to the electrolytic capacitors. What this
means is that there is an impulse when power is applied or removed, after which the output
stabilizes. The time for this is less than a second or so.

-6-
-7-
Oscillator Bank Noise Source 4

Normally we think of noise as being a mixture of all frequencies. But in fact there are a number of
noise sources (both natural and manmade) that do exhibit a certain tonality. A good example of this
is the whistle of a vacuum cleaner. Yes, it's a complex signal and yes, it's a mixture of many
component frequencies. But there's no denying it displays a sort of “pitch.” Another example would
be when pieces of junk are dropped into a dumpster; this essentially produces a random noise, but
due to the resonant frequency of the container, you'd still hear a tone that might even be classified as
musical. The point is, white noise isn't the only type useful to us. Other kinds of complex mixtures
might find their way into musical situations.
The circuit in this chapter is just such an example. It`s a weird one and not so well known, but of
tremendous utility when trying to synthesize clangorous sounds - things like cymbals, chimes,
gongs and so forth. It might also prove useful in recreating steam calliopes and other “noisy" musical
instruments. Incidentally, this oscillator bank approach was exploited in my popular Clangora hi-hat
project which appeared in the November 2003 issue of Nuts & Volts Magazine.
The essential concept is to take a half-dozen or so oscillators, tune them to non-harmonic
frequencies and mix them together. The result is a mess, but in fact the human ear can still detect a
number of individual frequencies more predominantly than others. Words can't describe this effect
adequately, so I urge to try it and hear for yourself just how unique it really is.
Due to the repetitive nature of the circuit, it could
easily become a nightmare without a simple
oscillator building block. Fortunately, an inexpensive
CMOS chip comes to the rescue. The 40106 is a hex
Schmitt trigger and each unit within it is readily
coaxed into becoming an oscillator. Figure 7 shows
how this is accomplished with 1/6th of the chip. The
output keeps trying to go low or high, as any good
digital gate should. If the output swings high, then the
capacitor begins to charge. Once the upper trip point
is hit, the Schmitt trigger goes low and the cap
discharges. When the voltage drops below a lower
trip point, the whole process starts over.
You choose appropriate values for R and C to set the
rate of oscillation. If you'd like to automate this
process, be sure to see the free design sheet posted
on the Midwest Analog Products Web site; refer to
page vi for details. But note that the frequency
depends not only upon the component values, but
the temperature of the chip, the power supply
voltage, and the internal trip points of the Schmitt
triggers (which might vary from unit to unit). Keeping
this in mind, Figure 7 shows a formula which yields
the approximate output frequency; in my experiments, the results were off by as much as a 3:1 ratio,
which truly doesn't matter in this situation.
In Figure 8, we see six of these circuits mixed to create a noise source. The oscillators are tuned to a
variety of nonharmonic frequencies, some close to each other and some not, and the result is the
strange tonal effect described earlier. The output of each oscillator is summed through a 100K
resistor, but feedback resistor R1 attenuates the mix by a factor of 10 so that the signal doesn’t peg
out. The resulting waveform is about 10Vpp, biased at -5V. In other words, it is a quasi-digital
waveform swinging from 0V to -10V and back up again.

-8-
-9-
CMOS Digital White Noise Source 5

A digital noise source is based upon the concept of pseudo-random number generators. The idea
here is that we configure an arithmetic circuit of some sort to spit out a string of binary numbers.
Each number will appear sooner or later, but shouldn't be obviously related to the one preceding it or
following it in the sequence. An analogy would be flipping a handful of pennies repeatedly and
counting the number of heads which appear each time; with fair coins what happened on the
previous trial should have no impact on the next trial. But where the electronic version differs from
actually tossing coins is in the fact it's completely deterministic. That is, while the pattern of numbers
may appear random over the short term, eventually they will repeat in the same order they originally
came up. Hence the prefix "pseudo." The reason for this, of course, is that the sequence is
completely programmed into the logic gates making up the circuit.
The digital ICs which go into a pseudo-random number generator will have a fixed length or “word
size," and it's this (among other things) that determines the period of repetition. The larger the word
size and hence the longer the period, the less likely the repeating pattern of numbers will become
noticeable. And the rate at which the numbers are spit out, determined by an internal clock, also has
a bearing on how the noise sounds.
But back to music. Once we have a source of pseudo-random numbers, we couple its output to a
filter circuit which converts the digital patterns into analog audio form. What comes out is an
extremely reliable white noise source composed of all frequencies, as required. If the period is
sufficiently long enough, then the repetitive nature of the noise won't be apparent. Incidentally, the
binary numbers can be processed into analog form one bit at a time (serial) or all at once (parallel).
The former approach is simpler and less expensive, requiring fewer electronic components.
Several integrated digital noise sources were available in the past, but are now very hard to find.
First was the MM5837 from National Semiconductor. With a word size of only 17 bits and a fixed
clock rate of 130kHz, the pseudo-random numbers repeat once a second, and this is indeed audibly
noticeable as a sort of “whirring." The MM5437 was slightly better with a 23 bit word size. My all time
favorite was the Texas Instruments SN76477. What made this one so neat was the ability to get
access to the noise clock and filter - wow, what a chip! Its less powerful cousin, the SN94281, was
also a hoot to play with. These complex noise ICs are gone now but still sorely missed.
The only reasonable option nowadays is to build a digital noise source from scratch. Thanks to the
ready availability of CMOS, this isn't too onerous of a task. Figure 9 shows a practical design. This
implements a 31-bit word size, using four 4015 8-bit shift registers wired in series. (Bit 32 at pin 2 of
IC4 is used only as an output, and doesn't figure into the numerical manipulations). With 31 bits, this
gives a range of over two-billion pseudo-random numbers! At a clock rate of 200Khz, the sequence
will repeat every three hours or so - an astoundingly good performance for the most demanding of
situations!
As long as we've mentioned the clock, let's continue there. XOR gates IC5a and IC5b are configured
as inverters, and then patched together to make a standard CMOS oscillator. R1 and C1 determine
the rate of oscillation, which as just mentioned is approximately 200kHz. The oscillator output at pin
3 of IC5a is applied to the clock inputs of all the shift registers, IC1 through IC4. Each time the clock is
pulsed, whatever bits are currently in the registers are passed down the line one notch.
But a new "random" bit has to be created at each instant. Well, an exclusive NOR gate, comprising
IC5c and IC5d, monitors the states of the 18th and 31st bits and applies the result to the input of the
entire shift register at pin 7 of IC1. Exactly why this creates a decent “random" bit is beyond the
scope of this guide, requiring a fairly deep understanding of abstract algebra. The interested reader
is referred to some of the sources listed in the Bibliography. For our purposes we'll simply mention
that several different choices of feedback taps can be used to generate a maximal length sequence

- 10 -
of random numbers. In this case, 18 and 31 have been used as the magic values. Just intuitively, try
to imagine a string of 31 bits stepping down the shift register one at a time, with a new one selected
at random to be pumped in at the start.
By the way, if the bit pattern ever becomes all ones, then the shift register would stall. The exclusive
NOR gate composed of IC5c and IC5d would just keep creating more ones and nothing would ever
change. This can never happen once the circuit is in operation and there's at least one zero
occurring somewhere. But when you first power up the circuit, you never know for sure what state
the internal gates of the shift register will assume. It is possible at the outset, quite by chance, that all
ones could be latched in, freezing the circuit up. Well, IC6 is included to prevent this occurrence.
This is the CMOS version of the trusty old 555 timer and is pressed into service here as a power-up
one-shot. In particular, when a voltage is first applied to the circuit, IC6 automatically goes through a
timing cycle of about 5 milliseconds; the duration is determined by R5 and C3. The output of IC6 is
applied to the reset inputs of IC1 through IC4, and this clears the internal registers to all zeros. After
the one-shot times out, the circuit commences normal operation.
So, we`ve got some random bits marching down the shift register. They will continually spill out of
pin 2 of IC4 in a serial stream. But remember, CMOS is a digital family. What comes out here is a
nothing more than a bunch of pulses swinging between 0V and +15V (but admittedly of many
different durations). In other words, we definitely have a wide variety of frequencies completely
random in the short term, but unfortunately the voltage levels are entirely predictable: 0V or +15V.
Well, not to worry. By passing the digital bit stream through a low pass filter, we in effect integrate the
rectangular signal - there comes that mathematics again! Even if the integral calculus is not your
bag, you'll at least intuitively appreciate that integration smoothes out the pulses, creating a signal
which fluctuates unpredictably and continuously in amplitude.
The question is, what cutoff frequency should the filter have? The usual rule of thumb is that the
-3dB point should be at 1/20th the clock frequency to give the output noise a so-called Gaussian
characteristic. If in fact you decide to use that figure, then go with R3 at 100k and C2 at 150pF. This
gives a cutoff of about 10kHz which is in fact 1/20th the clock rate of 200kHz. But if you look over
some of the resources listed in the Bibliography, you might notice that experimenters have toyed
with this a bit. You may want to try different values of C2 on the breadboard and see what you think.
For example, to knock things down an octave, double C2 to make it 300pF.
There is in fact another way to alter the frequency response of the circuit, and that's by changing the
clock speed. Here's a neat way to think about it. In the paragraph above, it was suggested that
doubling C2 would halve the cutoff frequency of the output filter. But if the filter's cutoff remains fixed
and instead you double the clock frequency, you're no longer filtering things at 1/20th the clock rate
but at a factor of 1/40. In other words, fewer treble components will make it through. And the reverse
would occur if you brought the rate downward. Since repetition isn't a problem with a deluxe circuit
like this featuring 31-bit numbers (remember - at 200kHz, the pattern repeats only after 3 hours or
so!), feel free to play around in either direction. And whether you're into the calculations or not, I can
assure you that varying the clock rate of a digital noise source produces some wonderful effects. In
fact, I wrote a bunch of articles on the value of doing this with the SN76477. You can find these in my
Electronic Music Circuits: The Reprints. (Again, see the Bibliography for details.).
Finally, keep in mind that you can adjust feedback resistor R3 to change the gain of op-amp IC7 and
hence the level of the noise output. One of the nice things here is that the circuit is very repeatable.
With the transistor noise sources of Chapters 2 and 3, you never know for sure what output level to
expect, since it depends so dramatically on the noise transistor itself. But in the digital noise
generator of Figure 9, you can always be assured of getting a 0 to +15V swing on pin 2 of IC4, and
hence the output volume is completely determined by R3, R4, and the frequency response by C2
and the clock. Nice!

- 11 -
- 12 -
Pink Noise Filters 6

As explained back in Chapter 1, white and pink noise both contain components of random
frequencies and amplitudes. They differ in how the power is distributed across the spectrum. We've
seen that white noise has equal power per unit bandwidth (a linear relationship), while pink noise
exhibits equal power per octave (an exponential relationship). This implies that the power of white
noise increases at a rate of +3dB/octave. So, to convert it to pink noise, we “simply" pass it through a
-3dB/octave low pass filter to “undo" the power rise. And here's where things get interesting; there's
more to that adverb "simply" than meets the eye!
For you see, garden variety filters built from resistors and capacitors have a minimum slope of
-6dB/octave due to the nature of capacitive reactance. And yet what we're really looking for here is a
1/2 pole filter! So, yes, the idea is theoretically simple, but a real, live working circuit is anything but
trivial. Not to keep you in suspense, but in fact it is possible to approximate the required slope using
a number of fairly sophisticated design methods. This can get pretty tricky in a hurry, involving such
concepts as phase lag, complex plane analysis and so forth. So, we'll just summarize things here by
presenting some circuits which have proven themselves over time. And to give you a feel for how
well these filters behave, each is accompamed by a graph showing the typical frequency response.
These were created in the CircuitMaker SPICE simulation software package.
To begin, you need to ask yourself what the intended purpose is for your pink noise source. For
example, in John Simonton's “Chatter Jammer” project, the only goal was to create some pleasant
sounding noise which when pumped into a set of earphones would block out distracting sounds. In
this case, he strapped a single cap across the output to dump some of the high frequencies - a crude
approach, but perfectly suitable for the intended application. See the Bibliography for a complete
reference.
A more sophisticated attack is that of Figure 10 which shows the pink noise filter recommended by
National Semiconductor for use with their MM5837 digital noise chip. It appeared in the 1977 Audio
Handbook, among other places; again refer to the Bibliography for details. But note, the circuit will
work well with just about any other noise source. A sequence of three RC combinations work against
each other, as it were, to create a very reasonable approximation to a -3dB/octave slope. Good
quality polystyrene caps should be used since close values are important here. Speaking of that,
notice how C2 and C3 are paralleled to create a 0.094pF capacitance. The response curve
demonstrates that the filter performs very well. Using elementary algebra, you can confirm that the
slope of the graph is indeed close to -3dB/octave.
For more stringent applications, the similar passive filter of Figure 11 can be used. Coming from the
Hill and Horowitz classic, The Art of Electronics, it features four stages and more precise
components. (Need I mention it again...see the Bibliography...) In either of these circuits, you will
probably want to buffer the output with a voltage follower. Or you could use a non-inverting amplifier
configuration which will let you modify the amplitude while providing good buffering simultaneously.
Figure 12 illustrates a simple active filter for less demanding situations. You'll notice that the
resulting curve isn't perfect by any means, showing a couple bumps and waves. But ignoring these,
it does in fact fall off in a general fashion at the required rate. Because of its simplicity and non-critical
parts, it would be ideal for a synthesizer noise source where great precision isn't required. That is,
you're making music not laboratory tests, so specsmanship doesn't enter into it. Incidentally, I
believe the choice of values for this circuit is due to Terry Mikulic, one of the early pioneers of
homebrew synthesizer design.
But if a more precise response in an active filter really is required, then check out the circuit of Figure
13. This has been kicking around for several decades now and recently resurfaced in the November
2003 issue of Nuts & Volts Magazine, pp. 28, 29, in response to a reader's question. It gives superior
results and yet uses pretty much run-of-the-mill components. Like the preceding design, you can
bump the amplitude up or down, by changing the input resistor.
- 13 -
- 14 -
- 15 -
- 16 -
- 17 -
Adding Voltage Control 7

If you conduct a survey of modular analog synthesizer noise sources, I think you'll be dismayed at how
plain vanilla they are. The great bulk sport a single output jack and that's that. Among the “elite"
models you might find two jacks - wow - one for white and one for pink noise. There's no call for this
excess of mediocrity. As we'll see, it's quite easy to add voltage control to a noise source, opening up a
wealth of new and exotic sonic effects. If you've never experienced this type of module before, you're
in for a real treat. All manner of weird, ghostly and even explosive sounds are possible.
Adding voltage control to the digital noise source of Chapter 5 is very straightforward. Recall that a
clock circuit is used to step binary bits down a shift register. All we do is rip out the fixed clock and
substitute a VCO (Voltage Controlled Oscillator). Let's see how. Refer back to Figure 9 first. Start by
removing IC5a, IC5b, C1, R1 and R2. Next, refer to Figure 14 which shows the replacement.
Here we've used the popular 566 function generator chip as the heart of a VCO. But a number of
special tricks have been employed to really squeeze top performance out of it. The general idea is that
the voltage controlled current source inside the 566 can be greatly improved by putting it within the
feedback loop of an op-amp. In loose terms, the benefit here is that the source is now more or less
unloaded, and so the linear response is extended. In particular, the performance is much more
predictable, especially at the two extremes of high and low control voltage inputs.
And by putting the 566's current source in the negative feedback loop of an op-amp, the unit now
responds directly to a control voltage input. That is, an increase in voltage yields an increase in output
frequency. (For the typical configuration recommended in the data sheets, this is annoyingly
backwards). And best of all, by powering the 566 off of a negative supply, the control voltage is now
referenced to ground or 0V, not the awkward +15V as in the usual arrangement. If you`d like to
understand in more detail how this magic works, be sure to see the thorough treatment in my
handbook, Making Music With the 566, (Available through Magic Smoke Electronics -
www.magsmoke.com)
Okay, we've got -15V applied to pin 1 and ground to pin 8. As far as the 566 is concerned, there`s still a
+15V potential straddling the two pins. But, as mentioned, this has the effect of referencing the control
voltage input at pin 5 to ground - a much more convenient scenario. And if you chase the op-amp
connections around you can confirm that the current source (between pins 5 and 6 of the 566) is
indeed within a negative feedback loop.
C1 is the timing capacitor, and you'll note that it now connects to the negative supply instead of
ground. C3 and C4 are decoupling caps used simply to stabilize the chip. I've shown a simple pot, R4,
as providing the control voltage input. But really, this can be any voltage source you desire, like the
output of an envelope generator, an LFO and so on.
The square wave output of the 566 is a bit nonstandard in any event, and with the negative power
supply trick it becomes even weirder. It swings between -8.25V and -0.9V, an overall amplitude of
7.35Vpp. This is hardly what would make the shift registers in Figure 9 happy. But with a clever trick
we can convert it to a decent +15V pulse wave.
I've shown XOR gate IC3a configured as an inverter here; feel free to use any other leftover CMOS
gate in its place. (Tying pin 2 high makes it behave as an inverter). Notice how the other input, pin 1, is
biased up to +7.5V by resistors R5 and R6. Now C2, while blocking the heavy negative DC
component of the 566, supplies a low impedance pulse which easily overcomes the high impedance
of the 1M resistors. In effect, the 7.35Vpp square is shifted up to the halfway point, meaning that pin 1
of IC3a sees a signal swinging from +3.825V to +11.175V. Owing to the wonderful noise immunity of
CMOS and the fact that these values are sufficiently far away from the center line, they'll be sensed as
a zero and one, respectively. The inverter responds appropriately, with the output at pin 3 cleanly
moving from 0V to +15V. The pulse width is of no consequence here since the shift registers simply
respond to a rising edge. By the way, I first saw this stunt of capacitively coupling an

- 18 -
- 19 -
unsuitable signal to a biased CMOS input in Craig Anderton’s combo organ circuit, Organtua which
PAiA Electronics put out a couple decades ago. It's a neat idea.
Tack this VCO on to the circuit of Figure 9 and away you go. As mentioned in Chapter 5, altering the
clock frequency of a digital noise source is equivalent to passing the noise through a filter. So, with
just a handful of common parts, you in effect wind up with a noise source and a VCF (Voltage
Controlled Filter)! Finally, if you'd like an exponential response instead, refer to Making Music With
the 566 handbook mentioned above, which explains a fantastic alternative.
But what if you`re going with one of the analog noise sources from Chapters 2 through 4; is it still
possible to add on voltage control? Rather surprisingly, it is, using an idea which I came up with for
the Clangora analog hi-hat cymbal project.
What we'll do is create a voltage controlled analog noise generator using FM (Frequency
Modulation) instead of a VCF, to obtain a broad range of sounds. What comes out of it is noise, of
course, but it's possible to emphasize certain bands of frequencies. When sweeping it over its range
the effect is quite ethereal, a sort of “swooshing" not unlike the sound of wind whistling through
forest trees. Let's investigate the general procedure first before turning to the circuit itself.

Figure 15 shows the block diagram of the tunable noise generator. First white noise, created using
any of the techniques in this guide, is allowed to modulate the frequency of a VCO configured
around a 566. The depth control is used to set the amount of modulation desired. A separate control
voltage input (shown as a simple pot here) adjusts the center frequency of the VCO. The modulating
white noise, then, causes the VCO to stray randomly from this center frequency. In this way a certain
tonality is imparted to the sound thus created; it's still noise, but it seems to be focused around a
certain band of frequencies. By twisting the tuning control it is possible to sweep the sound across
the entire audio spectrum.
Incidentally a triangle wave seems to work best here, giving a fairly smooth sound. Square waves,
on the other hand, lead to a rather harsh and gritty effect.
So we have arrived at a unique method for “tuning" the response of a noise generator without
requiring the use of active filters. Let's check out a practical circuit which implements the block
diagram of Figure 15.
Refer to Figure 16 which shows the schematic for the tunable noise generator. The noise source
applied to the input is apt to be fairly weak, perhaps on the order of 1Vpp or so. So, we send it to a
preamplifier configured around IC2b, AC coupling it in the process by means of capacitor C6. Notice
that C6 is purposely kept fairly small so that the bass frequencies will be attenuated a tad. This helps
prevent rumbles in the sub-sonic region.

- 20 -
- 21 -
R2 and R9 set the gain of op-amp IC2b to a factor of 10. Depending on the noise source you've used
at the input, this may not be enough "oomph" to properly modulate the VCO. So if needed, adjust R9
upward and/or R2 downward to boost things a bit.
The full strength signal is applied across potentiometer R6, which lets you manually adjust how
deeply the noise will modulate the VCO.
The standard linear control voltage input at pin 5 of the 566 is too seedy for our purposes here, so
we'll use an exponential approach instead. If you'd like to learn how this is done, I'll refer you once
again to my 566 handbook - mentioned above - which gives a complete design analysis. R10 is
driven by the wiper of R7, the tuning potentiometer. The top and bottom ranges of this control have
been limited appropriately by R4 and R5 respectively. With this pot you can sweep the VCO's basic
frequency from about 20Hz to 20KHz in one fell swoop; there's lots of usable range here! And of
course, you can easily convert this into a general purpose control voltage input since IC2a is in fact a
summing amplifier.
But here is the key to the whole circuit. Recall that the white noise generator output can be picked off
the wiper of R6, the depth control. This signal feeds R3, which then sums into IC2a. Thus the noise
modulates the basic frequency of the VCO, completely at random. Potentiometer R6 lets you adjust
how much modulation you want. Naturally, when set to zero the circuit performs as a normal VCO.
And that brings us up to the output of the entire circuit. There are two waveforms available from the
566 chip, a square wave (pin 3) and a triangle wave (pin 4). Feel free to use either waveform, but in
general the triangle output sounds quite a bit smoother. But remember that pin 4 rides on a rather
heavy DC bias which needs to be blocked, and the amplitude of the triangle there is a bit feeble.
Capacitor C9 AC couples the signal to op-amp IC3, which then boosts it up to about 10Vpp.
When using the tunable noise generator, be careful not to blow out either your loudspeakers or your
ears! The output of this device is pretty hefty (up to a max of 10Vpp). This higher value was selected
so that the circuit would be compatible with analog synthesis equipment. On the other hand, if using
the tunable noise generator with standard hi-fi gear, you can reduce R12 to lower the gain of the
output stage.

- 22 -
Tone And Equalization Options 8

I suppose for critical audio testing purposes one would always want the flattest response possible
from a white or pink noise source. But for composing electronic music, being able to color the sound
is important. It's a simple matter to add tone and volume controls to any basic noise generator
circuit. For best "feel,” you will want the tone controls to have a linear taper, while the volume
controls should have an audio taper. And to maximize stability, be sure to use polystyrene or mylar
capacitors throughout (apart from the electrolytics).
Figure 17 shows one easy approach. Notice that C1 and R1 form a highpass circuit, while R2 and
C2 comprise a lowpass circuit. R3 allows you to continuously attenuate either the low or high
frequencies. The lowpass portion cuts off at about 159Hz, while the highpass is set for 723Hz; the
slope in either case, of course, is -6dB/octave. With R3 in the center, you'll get a notch of sorts with a
center at around 498Hz. While simple to implement, one drawback is that the volume control loads
the filter down somewhat and might change the amplitude or frequency response. This may not
matter for many simple applications. In any event, the output should be buffered by some sort of
active circuit.
Figure 18(a) shows how to attain a broader range of sounds while avoiding the loading problem. The
circuit has been implemented to give +20 dB of boost or cut at 72Hz for the bass control, and the
same response at 4.8kHz for the treble control. For the design details, refer to National
Semiconductor's Audio Handbook, (Dems Bohn, editor, Santa Clara: National Semiconductor,
1977, pp. 2-44 through 2-47).
For even better tone shaping, Figure 18(b) demonstrates how to add on a simple 3-band equalizer.
This arrangement is ideally suited for most electronic music applications where the noise will be
acting as a basic sound source. The critical frequencies for the Low, Mid and High controls are about
30Hz, 1kHz and 10kHz, respectively. Notice that C1 plays no real role in the tone shaping but is
added simply to attenuate any undesired RF (radio frequency) interference. If desired, refer to the
Audio Handbook, mentioned above, pages 2-48 and 2-49 for the design equations.

- 23 -
- 24 -
 Bibliography

(Books and Articles Discussing Noise and Noise Generators)

anonymous, "White Noise Generator," Tech Notes, (Newton, New Hampshire: CFR Associates,
no date), pp. 1, 2
Barbarello, James, “Build Cabonga: An Electronic Percussion Synthesizer," Popular Electronics,
August 1977, pp. 39 - 42 and September 1977, pp. 76 - 80.
Barbarello, James, "Build a Unity Gain Indicator," Popular Electronics, November 1978, pp. 106
- 108.
Bohn, Denis, “Build a Pink Noise Generator for Audio Testing," Popular Electronics, July 1977,
p. 66.
Bohn, Denis, editor, “Noise Generator," Audio Handbook, (Santa Clara, California: National
Semiconductor, 1977), pp. 2-56 through 2-59.
Brown, Robert and Olsen, Mark, Experimenting with Electronic Music, (Blue Ridge Summit,
Pennsylvania: Tab Books, 1974), pp. 114 - 116.
Buchsbaum, Walter, “Noise Fundamentals,” Popular Electronics, January 1982, pp. 72, 73.
Byers, T. J., “Pink Noise Filter,” Nuts & Volts Magazine, August 2003, pp. 28, 29.
Chamberlain, Hal, Musical Applications of Microprocessors, (Rochelle Park, New Jersey:
Hayden Book Company, 1980), pp. 30, 93, 483 - 489.
Damashek, Marc, “Shift Register with Feedback Generates White Noise,” Electronics, May 27,
1976, pp. 107, 109.
Davis, Gary and Jones, Ralph, Sound Reinforcement Handbook, (Milwaukee: Hal Leonard
Corporation, 1987), pp. 75 - 80.
Douglas, Alan, Electronic Music Production, (Blue Ridge Summit, Pennsylvania: Tab Books,
1982), p. 61 - 63.
Henry, Thomas, “The Mini-Controller (Part 1)," Electronic Musician, October 1986, pp. 66 - 68.
Henry, Thomas, and Orman, Jack, "Build the ADV-Snare," Nuts & Volts Magazine, July 1996,
pp. 71 - 75.
Henry, Thomas, and Orman, Jack, “Build the ADV-Bass,” Nuts & Volts Magazine, November
1996, pp. 56 - 59.
Henry, Thomas, The Electronic Drum Cookbook, (San Jose, California: Magic Smoke
Electronics, 2006), pp. 21, 37, 38.
Henry, Thomas, Electronic Music Circuits: The Reprints, (San Jose, California: Magic Smoke
Electronics, 2009), pp. 45 - 49, 65 - 74, 92 - 95, 130 - 135.
Henry, Thomas, "Clangora Percussion Synth," Nuts & Volts Magazine, November 2003, pp. 44 -
51.
Henry, Thomas, Making Music with the 566, (San Jose, California: Magic Smoke Electronics,
2007), pp. 23 - 25.
Henry, Thomas, “Build a Tunable Noise Generator," Nuts & Volts Magazine, November 1999,
pp. 25 - 27.

- 25 -
Horn, Delton T., The Beginner's Book of Electronic Music, (Blue Ridge Summit, Pennsylvania: Tab
Books, 1982), p. 41 - 43.
Horowitz, Paul and Hill, Winfield, The Art of Electronics, (Cambridge: Cambridge University Press,
1980), pp. 655 - 664.
Hutchins, Bernie, Musical Engineer’s Handbook, (Ithaca, New York: Electronotes, 1975), pp. 5-1
through 5-4.
Hutchins, Bernie, “Theory and Application of Noise Generators in Electronic Music," Electronotes,
Volume 8, Number 64, April 1976, pp. 3 - 63.
Hutchins, Bernie, “Simple Noise Sources," Electronotes Application Note #143, August 15, 1979.
Hutchins, Bernie, Builder’s Guide and Preferred Circuits Collection, (Ithaca, New York:
Electronotes, 1980), no page numbers.
Hutchins, Bernie, "The ENS-76 Home-Built Synthesizer System - Part 8, Random Sources,"
Electronotes, Volume 9, Number 76, April 1977, pp. 3 - 16.
Klein, Barry, Electronic Music Circuits, (Indianapolis: Howard W. Sams and Company, 1982), pp.
66 - 69.
Jung, Walter G., IC Array Cookbook, (Rochelle Park, New Jersey: Hayden Book Company, 1980),
pp. 164 - 165.
Lancaster, Don, The CMOS Cookbook, (Indianapolis: Howard W. Sams and Company, 1977), pp.
318 - 323.
Lancaster, Don, The TTL Cookbook, (Indianapolis: Howard W. Sams and Company, 1974), pp.
277 - 282.
Marston, Ray, “Bipolar Transistor Cookbook - Part 5," Nuts & Volts Magazine, November 2003,
pp. 68 - 72.
Mazur, Jeffrey G., "Audio Testing With Pink Noise," Radio-Electronics, September 1978, pp. 44,
45.
Meyer, Chris, “White Noise is Not a Cliche," Polyphony, December 1983, pp. 40, 41.
Mikulic, Terry, “Four Analog Circuits from Terry Mikulic," Electronotes, Volume 5, Number 34,
January 25, 1974, pp. 16, 17.
Pryor, Lany, “Sample/Hold and Noise Source," Polyphony, May/June 1979, pp. 30 - 35.
Simonton, John S., "Build a Pink Noise Generator," Popular Electronics, December 1970, pp. 61 -
63.
Simonton, John S., “Programmable Drums," Radio-Electronics, February 1978, pp. 41 - 44 and
March 1978, pp. 45 - 48, 107.
Rossum, Dave, "EN-73, Noise Source: Option 1,“ Electronotes, Volume 4, Number 30, November
5, 1973, p. 4.
Ward, Brice, Electronic Music Circuit Guidebook, (Blue Ridge Summit, Pennsylvania: Tab Books,
1975).
Wells, Thomas, The Technique of Electronic Music, (New York: Schirmer Books, 1981), pp. 60 -
62.
Wood, Steve, “Professional Drum Synthesizer," Radio-Electronics, May 1980, pp. 49 - 52, and
June 1980, pp. 59 - 62.
Wright, Harold, “Model Railroad Sound Synthesizer," Popular Electronics, December 1977, pp. 80
- 83.

- 27 -