The Fourier Transform

Author(s): Ronald N. Bracewell

Source: Scientific American , Vol. 260, No. 6 (JUNE 1989), pp. 86-95
Published by: Scientific American, a division of Nature America, Inc.
Stable URL: https://www.jstor.org/stable/10.2307/24987290

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Scientific American, a division of Nature America, Inc. is collaborating with JSTOR to digitize,
preserve and extend access to Scientific American

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
 The Fourier Transform
DNA's double helix, the sunspot cycle and the sawtooth signals
of electronics can be reduced mathematically to a series
. of undulating curves. This idea underlies a powerful analytical tool

by Ronald N. Bracewell

T
o calculate a transform, just lis­ signal, thereby helping to make sense Returning to France relatively un­
ten. The ear automatically per­ of observations in astronomy, medi­ scathed, Fourier focused on math­
forms the calculation, which the cine and chemistry. ematical matters as professor of anal­
intellect can execute only after years ysis at the Polytechnic School, but

T
of mathematical education. The ear he world first learned about the in 1802 he again entered Napoleon's
formulates a transform by converting technique from the mathema­ service. Fourier became the prefect of
sound-the waves of pressure travel­ tician for whom the transform the Isere department. While attempt­
ing through time and the atmos­ is named, Baron Jean-Baptiste-Joseph ing to repair the disruptions remain­
phere-into a spectrum, a description Fourier. Fourier was not merely inter­ ing from the Revolution of 1789, he
of the sound as a series of volumes at ested in heat; he was obsessed by built the French section of the road to
distinct pitches. The brain turns this it. He kept his home in Grenoble so Turin and drained 80,000 square ki­
information into perceived sound. uncomfortably hot that visitors of­ lometers of malarial swamp. During
Similar operations can be done by ten complained. At the same time he this time he derived an equation that
mathematical methods on sound would cloak himself in heavy coats. described the conduction of heat in
waves or virtually any other fluctu­ Perhaps it was the lure of a warm solid bodies. By 1807 Fourier had in­
ating phenomenon, from light waves climate that in 1798 drew Fourier to vented a method for solving the equa­
to ocean tides to solar cycles. These join the retinue of 165 savants that tion: the Fourier transform.
mathematical tools can decompose accompanied Napoleon's expedition

F
functions representing such fluctu­ to Egypt. ourier applied his mathemati­
ations into a set of sinusoidal com­ While Napoleon was fighting Syr­ cal technique to explain many
ponents-undulating curves that vary ians in Palestine, repelling the Turks instances of heat conduction. A
from a maximum to a minimum and from Egypt and hunting the Mameluke particularly instructive example that
back, much like the heights of ocean chief, Murad Bey, the French scientists avoids computational complications
waves. The Fourier transform is a undertook ambitious studies in ge­ is the flow of heat around an anchor
function that describes the amplitude ography, archaeology, medicine, agri­ ring-an iron ring that attaches a
and phase of each sinusoid, which culture and natural history. Fourier ship's anchor to its chain-that has
corresponds to a specific frequency. was appointed secretary of a scientific been thrust halfway into a fire. When
(Amplitude describes the height of body known as the Institute of Egypt. part of the circumference becomes
the sinusoid; phase specifies the start­ He discharged administrative duties red hot, the ring is withdrawn. Before
ing point in the sinusoid's cycle.) with such competence that he re­ much heat is lost to the air, the ring
The Fourier transform has become a ceived many diplomatic assignments. is buried in fine, insulating sand, and
powerful tool in diverse fields of sci­ Yet he was still able to conduct inten­ the temperature is measured around
ence. In some cases, the Fourier trans­ sive research on Egyptian antiquities the outer curve [see illustration on
form can provide a means of solving and contemplate a theory about the page 88).
unwieldy equations that describe dy­ roots of algebraic equations. Initially the temperature distribu­
namic responses to electricity, heat or Shortly before the French were driv­ tion is irregular : part of the ring is
light. In other cases, it can identify the en from Egypt in 1801, Fourier and uniformly cool, and part is uniform­
regular contributions to a fluctuating his colleagues set sail for France. ly hot; in between the temperature
The commander of the British fleet, abruptly shifts. As heat is conducted
Admiral Sir Sidney Smith, promptly from the hot region to the cool re­
RONALD N. BRACEWELL has been on
seized their ship along with its cargo gion, however, the distribution begins
the faculty of the electrical engineering
of Egyptian documents and relics. In to smooth out. Soon the temperature
department at Stanford University since
1955. He was educated at the Univer­ the honorable spirit of the time, Smith distribution of heat around the ring
sity of Sydney and the Cavendish lab­ put the scientists ashore unharmed in reaches a sinusoidal form: a plot of
oratory in Cambridge, England, where Alexandria. The English commander the temperature rises and falls evenly,
he got his Ph.D. His research interests eventually traveled to Paris to return like an S curve, in exactly the way sine
have ranged over microwave radar, ion­ the confiscated material-except for and cosine functions vary. The sinus­
ospheric physics and radio astronomy.
the Rosetta stone (the key to Egyptian oid gradually flattens until the whole
At Stanford he is a member of the Space,
hieroglyphics), which stands today in ring arrives at a constant temperature.
Telecommunications and Radioscience
Laboratory and is professor of comput­
the British Museum as a monument Fourier proposed that the initial,
er science, by courtesy. to Napoleon's military defeat and his irregular distribution could be bro­
contribution to Egyptology. ken down into many simple sinusoids

86 SCIENTIFIC AMERICAN June 1989

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
that had their own maximum temper­ one-cycle variation became known as and position, or phase, for each of the
ature and phase, that is, relative po­ the fundamental harmonic, whereas harmonics is the Fourier transform of
sition around the ring. Furthermore, variations with two, three or more the temperature distribution. Fourier
each sinusoidal component varied cycles in a single rotation became had traded a single distribution that
from a maximum to a minimum and the second, third and higher harmon­ was difficult to describe mathemati­
back an integral number of times in a ics. The mathematical function that cally for a more manageable series of
single rotation around the ring. The describes the maximum temperature full-period sine and cosine functions

SUNBEAM resolved into a spectrum provides a physical anal­ an amplitude at each frequency. Thus, a function of strength
ogy for mathematical transforms (top). The sunlight entering versus time has been transformed into a function of ampli­
the prism varies in strength from moment to moment (bottom). tude versus frequency. The Fourier transform can represent
The light leaving the prism has been separated in space i�to a time-varying signal as a function of frequency and ampli­
pure colors, or frequencies. The intensity of each color implies tude, but the transform also provides information about phase.

SCIENTIFIC AMERICAN June 1989 87

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
 a
that when added together would make
up the original distribution.
In applying this analysis to the con­
duction of heat around the ring, Fouri­
90 270 er reasoned that the greater the num­
ber of periods of a sinusoidal compo­
nent, the more rapidly it will decay.
One can follow his reasoning by ex­
amining the relation between the fun­
180 damental and the second harmonic
b of the temperature distribution. The

(L'M!. ; W ,,,,.,, ','.' L . .,. ..... ""." ;,.." . . .. . . . ....;o-L.

" ...,....) li.u ,. .. •• ".." .....'".". >4,,"'" "".1 ., ", .
... ... · •............,.•.•
w,· ... . .. · ,· ,J temperature of the second harmonic
varies from hot to cool twice around
o 90 180 270 360
POSITION the circumference of the ring, where­
as the fundamental varies only once.
c Therefore, the distance that the heat
must travel from hot peak to cool
trough is only half as far for the sec­

t
UJ
ond harmonic as it is for the fun­
damental. Furthermore, the tempera­
0:::
:::> ture gradient in the second harmonic
�
«
0:::
is twice as steep as it is in the funda­
UJ
Q. mental variation. Because twice the
:::E heat flow occupies half the distance,
UJ
�
the second harmonic will die out four
0 180 360
POSITION times faster than will the fundamental.
Higher harmonics will decay even

d more rapidly. Hence, it is only a single

sinusoidal distribution of the funda­
mental variation that persists as the
temperature of the ring approaches
equilibrium. Fourier believed that the
evolution over time of any initial heat
distribution could be computed by his
technique.

F
ourier's analysis challenged the
mathematical theories to which
his contemporaries adamantly
adhered. In the early 19th century,
many extraordinary Parisian mathema­
ticians, including Lagrange, Laplace,
Legendre, Biot and Poisson, could not
accept Fourier's claim that any ini­
tial temperature distribution could be
decomposed into a simple arithmetic
sum that consisted of a fundamental
variation and its higher-frequency har­
monics. Leonhard Euler also found
fault with Fourier's ideas, although
he had already proposed that some

180 TEMPERATURE OF AN IRON RING was

POSITION
one of the first phenomena analyzed by
Fourier's technique. One distribution of
heat around a ring is shown (a); bright·
er color represents hotter areas. To be­
e gin the analysis, the ring is "uncoiled"
(b), and the temperature is measured at

t
UJ
every point, yielding a temperature dis-
tribution around the circumference (e).
� Then the temperature distribution is de-

�
UJ
composed into many sinusoidal curves
having one, two, three or more cycles (d).
� When 16 of the curves are simply add-
ed together (solid line in e), they yield a
� o�_____________---,180
:-::-::--_____________�
360 good approximation of the original tem-
POSITION perature distribution (broken line in e).

88 SCIENTIFIC AMERICAN June 1989

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
0-

180 180 360 180 360

POSITION POSITION POSITION

CONDUCTION OF HEAT through an iron ring causes the tem­ ic (middle), and the two-cycle distribution, or second harmon­
perature distribution to change over time (left). Just as the ic (right), are shown. Fourier determined that the second har­
temperature distribution at any instant can be described as monic will decay four times faster than the first harmonic
a series of sinusoidal curves, the evolution of a temperature and higher harmonics will decay even faster. Because the first
distribution over time can be described in terms of changes harmonic persists the longest, the overall temperature distri­
in the sinusoids. The one-cycle distribution, or first harmon· bution approaches the sinusoidal shape of the first harmonic.

functions could be represented as a a discontinuous function described as the ebb and flow of the tides. Lord
sum of sine functions. And so when resulting from a combination of or­ Kelvin had invented an analogue com­
Fourier made this claim at a meeting dinary, continuous functions, such as puter for providing information about
of the French Academy of Sciences, linear, quadratic, exponential and si­ the tides to the crews of merchant
Lagrange stood up and held it to be nusoidal functions. If Fourier was cor­ and naval vessels. First sets of am­
impossible. rect, however, a sum of an infinite plitudes and phases were calculat­
Even under these circumstances the number of sinusoids would converge ed manually from a record of tidal
Academy could not ignore the signif­ to represent accurately a function heights and corresponding times that
icance of Fourier's results, and it with jumps, even with many jumps. At had been painstakingly measured dur­
awarded him a prize for his mathe­ the time this seemed patently absurd. ing the course of a year in a particu­
matical theory of the laws of heat In spite of these objections many lar harbor.
propagation and his comparison of workers, including the mathemati­ Each amplitude and phase repre­
the results of his theory with pre­ cian Sophie Germain and the engineer sented a sinusoidal component of the
cise experiments. The award was an­ Claude Navier, began extending Fou­ tidal-height function and revealed one
nounced, however, with the following rier's work beyond the field of heat of the periodiC contributions to the
caveat: "The novelty of the subject, analysis. Yet mathematicians contin­ tide. Then the results were fed into
together with its importance, has de­ ued to be plagued by the question of Lord Kelvin's computer, which synthe­
cided us to award the prize, while whether a sum of sinusoidal functions sized a curve predicting the heights of
nevertheless observing that the man­ would converge to represent a discon­ the tide for the next year. Tidal curves
ner in which the author arrives at his tinuous function accurately. were soon produced for ports all over
equations is not without difficulties, the world.

T
and that his analysiS for integrating he question of convergence aris­ It seemed obvious that a tide­
them still leaves something to be de­ es whenever an infinite series of predicting machine with more parts
sired both as to generality and even as numbers is to be added up. Con­ could process more amplitudes and
to rigor." sider the classic example: will you ever phases and thus would make better
The great uneasiness with which arrive at a wall if with each step you predictions. This turned out not to be
Fourier's colleagues regarded his work travel half of the remaining distance? completely true if the mathematical
caused its publication to be delayed The first step will bring your toe to the function to be processed contained a
until 18 15 . In fact, it was not com­ halfway mark, the second, three quar­ steep jump, that is, it described an
pletely described until the 1822 publi­ ters of the way, and at the end of the essentially discontinuous function.
cation of his book, The Analytical The­ fifth step you are almost 97 percent of Suppose such a function was re­
ory of Heat. the way there. Clearly this is almost duced into a small set of amplitudes
Objections to Fourier's approach as good as reaching the wall, but no and phases-that is, just a few Fourier
focused on the proposition that an matter how many steps you take, you coefficients. The original function can
apparently discontinuous function will never quite reach it. You could then be reconstructed from the sinus­
could be represented by a sum of prove mathematically, however, that oidal components corresponding to
sinusoidal functions, which are con­ you would ultimately get closer to the the coefficients, and the error between
tinuous. Discontinuous functions de­ wall than any distance nominated in the original function and the recon­
scribe broken curves or lines. For in­ advance. (The demonstration is equiv­ structed function can be measured at
stance, a function called the Heaviside alent to showing that the sum of a each point. The error-finding proce­
step function is zero on the left and half, a fourth, an eighth, a 16th and so dure is repeated, each time computing
jumps to one on the right. (Such a on approaches one.) more coefficients and incorporating
function can describe the flow of cur­ The question of the convergence of them into the reconstruction. In every
rent when a switch is turned on.) Fou­ Fourier series emerged again late in case, the value of the maximum error
rier's contemporaries had never seen the 19th century in efforts to predict does not diminish. On the other hand,

SCIENTIFIC AMERICAN June 1989 89

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
 S CHWINGER EIN S T EIN'S
WINFR E E BIOLOGICAL C
GO RDO N STRUCTURES AND
WAT E RM AN ANIMAL NAV I
HOBSON S LEE P
FRIEDMAN SUN AND EAR
S I E VER SA N D
HUB E L E YE , BRAIN, AN
GO U L D
GO U LD.
THE H O N E Y
MENARD ISLANDS
ATK I NS MOLECULES
E AMES
POWER S OF TE
MORRIS ON
This content downloaded from
143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
 Join the most exciting book club in science­
where the only commitment you make
is to reward your own curiosity
This is the Scientific American A one-of-a-kind book club
Library, where the mysteries of the for those who take science seriously.
human and natural worlds unfold
before your eyes in lavishly illustrated Each volume in the Scientific
volumes you'll want to read and American Library was created
refer to for years to come. exclusively for this series, written
by preeminent scientists-including
Here's why 75,000 readers like you many Nobel laureates-who
have already become members of describe their work in an exciting,
the Scientific American Library accessible way.

• 15-day Free Trial

Choose your premier volume
You will receive a new volume to
examine approximately every two
months. (Your first volume is shipped • Extinction,
immediately via UP S at no extra Steven M. Stanley. In
charge.) You have 15 days to look this geologic detective
over each volume. If you decide to story, a renowned
keep it, send in your payment. Other­ paleobiologist ex­
wise, simply return the book in the plores the causes of
same shipping carton and we'll the mass extinctions
immediately credit your account. that have periodi­
cally decimated
• No Prepayment and thousands of thriving species.
No Minimum Purchase "Smack up to date scientif-
Send no money in advance. You pay ically ...the best overview of the
only for the volumes you keep. You changing relationship between the
may choose as many or as few physical environment and life ....
volumes as you like: there is no Enthralling ...."
yearly minimum. -John Gribbon, New Scientist

• Low Members' Price

You may purchase any of these • Molecules, P. w.
handsome volumes for only $24.95- Atkins. A distin-
25% below the bookstore price. . guished chemist
takes us below the
• No Strings, No Risk, surface of visible
No Commitment reality to show how
You may cancel your obligation-free molecules make
membership at any time for any soap "soapy;' deter­
reason, without notice, without mine the taste of
penalty. barbecue, and give rise to the
brilliant colors of fall foliage.
CREDITS (top to bottom)-From Sun and Earth, the "ballerina skirt" "Undoubtedly the most beautiful
model of the solar-imerpianetary current sheet. From The Science of
Structures and Materials, bubble formation due to the surface tension
chemistry book ever written .... It is
0/ a liquid; courtesy Stephen Dalton, NHPA. From The Honey Bee,
drone. queen, and worker; drawing by John Hatzakis. From Einstein's
more than a book of facts;
Legacy, Chaplin and Einstein in Hollywood, 1931; COllrtesy The Molecules is a work of art."
Granger Collection. From Drugs and The Brain. major anatomical
divisions of the brain; painting by Bill Andrews. -John Emsley, New Scientist

Join us. You have nothing to lose. And a world of scientific discovery awaits.
To become a member of the Scientific American Library, simply detach and mail the facing reply card.
Or write to: Scientific American Library, P.O. Box 646, Holmes, PA 19043

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
the error becomes confined to a re­ basis for the Heaviside step function plied when the original function is
gion that gradually shrinks around the and the Dirac delta function; the latter continuous and the other when it con­
discontinuity, so that ultimately at any describes a unit of area concentrat­ sists of many discrete measurements.
given point the error approaches zero. ed at a point. The theory enabled the If the function is made from a list of
Josiah Willard Gibbs of Yale Universi­ Fourier transform to be applied to values at discrete intervals, it can be
ty confirmed this result theoretically solve equations that involved such in­ broken down into a series of sinusoi­
in 1899. tuitively accepted concepts as point dal functions at discrete frequencies,
Fourier analysis is still not appli­ mass, point charge, magnetic dipoles which range from a lowest frequen­
cable to unusual functions, such as and the concentration of a load on cy, the fundamental, through a series

�
those possessing an infinite number a beam. of frequencies that are two, three or
of infinite jumps in a finite interval. more times the fundamental. Such a
By and large, however, a Fourier se­ ter almost two centuries of de­ sum of sinusoids is called the Fou­
ries will converge if its original func­ velopment, the theory behind rier series.
tion represents the measurement of a the Fourier transform is firm­ If the original function provides a
physical quantity. ly established and well understood. value for every real number, then it
Vast areas of new mathematics have As we have seen, Fourier analysis is decomposed into sinusoidal func­
been developed from investigations of breaks down a function in space or tions at all frequencies, which are
whether the Fourier series of a partic­ time into sinusoidal components that combined by means of an operation
ular function converges. One exam­ have varying frequencies, amplitudes called the Fourier integral. The trans­
ple is the theory of generalized func­ and phases. The Fourier transform is form is neither the series nor the inte­
tions, which is associated with George a function that represents the ampli­ gral. In the case of the discrete func­
F. j. Temple of England, Jan G. Mikusin­ tude and phase at each frequency. The tion, it is the frequency-dependent list
ski of Poland and Laurent Schwartz of transform can be derived by two dif­ of amplitudes and phases appearing
France. It established in 1945 a firm ferent mathematical methods, one ap- in the Fourier series; in the case of the

FERREL TIDE PREDICTOR, an analogue computer built in the ic contribution to the tide, such as the gravitational pull of the
late 19th century, performed Fourier synthesis to forecast the moon. The numbers for a specific port could then be fed into
ebb and flow of the tides. Data that were collected on tidal the Ferrel Tide Predictor by twisting knobs on the back of the
heights at a particular harbor could be reduced by hand calcu­ machine (left). When a time was set on the front of the machine
lations into a set of numbers, each one representing a period- (right), the predicted height of the tide could be read off a dial.

92 SCIENTIFIC AMERICAN June 1989

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
unbroken function, it is the function Just how much arithmetic had to back long before Cooley and Tukey's
of frequency that results when the be performed depended on the num­ work to the astronomer Carl Fried­
Fourier integral is evaluated. ber of data points needed to describe rich Gauss. Gauss wanted to calculate
Regardless of the manner in which the wave. The number of additions asteroidal and cometary orbits from
the transform is derived, it is neces­ was comparable to the number of only a few observations. After discov­
sary to specify two numbers at each points, and the number of multiplica­ ering a solution, he found a way to
frequency. These might be the ampli­ tions equaled the number of points reduce the complexity of the calcula­
tude and phase; however, other num­ squared. For example, analyzing a tions based on principles similar to
ber pairs could encode the same infor­ wave specified by 1,000 points taken those of the fast Fourier transform.
mation. These values can be expressed at regular intervals required on the In an 1805 paper describing the work,
as a single complex number. (A com­ order of 1,000 additions and one mil­ Gauss wrote: "Experience will teach
plex number is the sum of one real lion multiplications. the user that this method will greatly
number and another real number mul­ Such calculations became more fea­ lessen the tedium of mechanical cal­
tiplied by the square root of negative sible as computers and programs culation." Thus, the challenge of celes­
one.
) This representation is very popu­ were developed to implement new tial motions not only gave us calculus
lar because it invites the use of com­ methods of Fourier analysis. One was and the three laws of motion but also
plex algebra. The theory of complex al­ developed in 1965 by James W. Cooley stimulated the discovery of a modern
gebra and the Fourier transform have of illM's Thomas]. Watson Research computing tool.
become indispensable in the numeri­ Center and John W. Tukey of the Bell

P
cal calculations needed to design elec­ Telephone Laboratories in Murray Hill, hysicists and engineers, indoc­
trical circuits, analyze mechanical vi­ N.
]. Their work led to the development trinated with complex algebra
brations and study wave propagation. of a program known as the fast Fourier early in their education, have be­
Representing an original function by transform. come comfortable with the represen­
its complex Fourier transform leads The fast Fourier transform saves tation of sinusoids. The convenience
to computational advantages. A typi­ time by decreasing the number of of representing the Fourier transform
cal problem is to ascertain the cur­ multiplications needed to analyze a as a complex function lets us forget
rent that flows when a known volt­ curve. At the time, the amount of mul­ that the underlying sinusoidal com­
age is applied to a circuit. The direct tiplication was emphasized simply be­ ponents are real and not necessarily
method involves solving a differential cause multiplication was slow with complex. This habit of thought has
equation that relates the voltage and respect to other computer operations, obscured the significance of and re­
current functions. The Fourier trans­ such as addition and fetching and tarded the adoption of a transform
forms of the voltage and current func­ storing data. similar to Fourier's that was conceived
tion, in contrast, can be related by an The fast Fourier transform divides a by Ralph V. L . Hartley in 1942.
equation whose solution is trivial. curve into a large number of equally Working in the research laboratory
spaced samples. The number of multi­ of the Western Electric Company, Hart­

T
oday the study of Fourier trans­ plications needed to analyze a curve ley directed the early development of
forms consists largely of acquir­ decreases by one half when the num­ radio receivers for a transatlantic ra­
ing techniques for moving free­ ber of samples is halved. For exam­ diotelephone and invented the Hartley
ly between functions and their trans­ ple, a 16-sample curve would ordinar­ oscillating circuit. During World War I
forms. Analytical methods can be ap­ ily take 16 squared, or 25 6, multipli­ Hartley investigated how a listener,
plied to evaluate the Fourier integral cations. But suppose the curve was through mechanisms in the ear and
and produce the transform. Although halved into two pieces of eight points brain, perceives the direction from
these methods may be difficult for each. The number of multiplications which a sound emanates. Working at
ordinary practitioners, many Fourier needed to analyze each segment is Bell Laboratories after the war, Hartley
integrals have been found and are list­ eight squared, or 64. For the two seg­ was the first to formulate an impor­
ed in tables of reference. These meth­ ments the total is 128, or half the tant principle of information technol­
ods can be supplemented by learning number required before. ogy that states that the total amount
a handful of theorems pertaining to If halving the given sequence yields of information a system can transmit
transforms. With the aid of these the­ a twofold gain, why not continue with is proportional to the product of the
orems more or less complicated wave the strategy? Continued subdivision frequency range the system transmits
forms can be handled by reduction to leaves eight irreducible pieces of two and the time during which the system
simpler components. points each. The Fourier transforms of is available for transmission. In 1929
Fortunately numerical methods are these two-point pieces can be comput- . Hartley gave up the directIon of his
available for computing Fourier trans­ ed without any multiplic;ations, but group because of illness. As his health
forms of functions whose forms are multiplication is required in the proc­ improved he devoted himself to the
based on experimental data or whose ess of combining the two-point trans­ theoretical studies that led to the
Fourier integrals are not easily evalu­ forms to construct the whole trans­ Hartley transform.
.
ated and are not found in tables. Be­ form. First, eight two-point transforms The Hartley transform is an alterna­
fore electronic computers, numerical are combined into four four-point tive means of analyzing a given func­
calculation of a transform was rather transforms, then into two eight-point tion in terms of sinusoids. It differs
tedious, because such a large amount transforms and finally into the desired from the Fourier transform in a rather
of arithmetic had to be performed 16-point transform. These three stag­ simple manner. Whereas the Fourier
with paper and pencil. The time re­ es that combine the pieces each call transform involves real and imaginary
quired could be reduced somewhat by for 16 multiplications, and so the total numbers and a complex sum of sinus­
forms and schedules that guided in­ number of multiplications will be 48, oidal functions, the Hartley transform
vestigators through the calculations, which is 3/ 16 of the original 25 6. involves only real numbers and a real
but the labor involved could still be This strategy for reducing the num­ sum of sinusoidal functions.
daunting. ber of computations can be traced In 1984 I developed an algorithm

SCIENTIFIC AMERICAN June 1989 93

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
 for a fast Hartley transform. The dif­
THE FOURIER AND HARTlEY TRANSFORMS ference in computation time b.etween
the fast Hartley transform and the
The Fourier and Hartley transforms convert functions of time into functions of frequency
that encode phase and amplitude information. The graphs below represent the unbroken fast Fourier transform depends on the
function g(t) and the discrete function g(T), where t is time and T is a number designated computer and the programming lan­
at each data point. guage and style. If these factors are
kept constant and no oversights are
made in the programming, programs
for the fast Hartley transform run
z
o z more quickly than those for the fast
o
(3z (3z Fourier transform. Although both pro­
::;) grams need the same amount of time
w.. ::;)
w..
to retrieve the data, provide for trigo­
z UJ
UJ
nometric functions and perform other
� t;:;
o
cr:
cr:
U
preliminaries, the time spent on the
co Vl
Z stages of the Hartley transform is half
::;) o
that required by the Fourier.
It was not clear at first, however, that
the Hartley transform provided the
same information as the Fourier trans­
Both functions start at zero, jump to a positive value and then decay exponentially. The form. Therefore, when the first pro­
definition of the Fourier transform for the unbroken function is an infinite integral, F(f),
grams were developed for comput­
whereas the definition for the discrete function is a finite sum, F(v).
ing the Hartley transform, an extra
n-l
00 step was provided to convert it to
F(f) = f g(t) (cos 2TTft -I sin 2TTft) dt F( v ) = J.. L g(t) (cos 2TTVT -I sin 2TTVT)
·00 n the more familiar Fourier. Workers
T=O
soon realized, however, that intensi­
Here f is frequency, v is related to frequency, n is the total number of samples and 1 is the ties and phases can be deduced direct­
imaginary number equal to the square root of -1. The integral representation is more
ly from the Hartley transform without
suited to theoretical manipulations, whereas the finite-sum representation is more suited
to computer applications. The Hartley transform and discrete Hartley transform have the need for the additional step. Fur­
similar definitions. ther reflection revealed that either
n-l kind of transform furnishes at each
H(f) = j g(t) (cos 2TTft + sin 2TTft) dt H( v ) = J..
n
L g(t) (cos 2TTVT + sin 2TTVT) frequency a pair of numbers that rep­
-00 T=O resents a physical oscillation in ampli­
Even though the only notational difference between the Fourier and Hartley definitions is tude and phase.
a factor -I in front of the sine function, the fact that the Fourier transform has real and Yet another reservation about the
imaginary parts makes the representations of the Fourier and Hartley transforms quite
Hartley transform was that the Fou­
different. The discrete Fourier and discrete Hartley transforms have essentially the same
shape as their unbroken counterparts. rier transform described physical phe­
nomena more naturally. Many phe­
nomena, such as the response of a
simple system to vibration, are com­
monly described by a complex sum
of sinusoidal functions, which is the
hallmark of the Fourier transform.
It might seem, therefore, that Fourier
transforms are more suitable for de­
scribing the behavior of nature.
Such a conclusion is in fact more
a reflection of our mathematical up­
bringing than it is of nature. After all,
when physical objects are measured,
FREQUENCY FREQUENCY they provide data in real numbers, not
complex ones.
Although the graphs look different, the phase and amplitude information that can be The advent of the fast Hartley trans­
deduced from the Fourier and Hartley transforms is the same, as shown below. form has made obsolete certain adap­
tations of the fast Fourier transform,
such as those used for eliminating
noise from digitally recorded music.
These adaptations require two pro­
grams: one of them transforms real
functions into the complex Fourier
domain, whereas the other converts
complex functions from the Fourier
FREQUENCY domain into real functions. High-fre­
quency noise in digitally recorded mu­
Fourier amplitude is the square root of the sum of the squares of the real and imaginary
sic can be eliminated by filtering out
parts. Hartley amplitude is the square root of the sum of the squares of H(-v) and H(v).
Fourier phase is the arc tangent of the imaginary part divided by the real part, and Hartley portions of the transform produced
phase is 45 degrees added to the arc tangent of H(-v) divided by H(v). by the first program. The second pro­
gram then converts the changed trans-

94 SCIENTIFIC AMERICAN June 1989

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms
FOURIER ANALYSIS can transform X-ray diffraction patterns molecular structure. If the process of transformation is re­
into molecular models_ X rays scatter off the electrons in a versed, the distribution of electrons, and therefore atoms, can
virus, for example, to produce patterns on film (left)- These be deduced (middle)_ From these distributions, models of the
patterns represent part of the Fourier transform of the virus's virus are made (right). Here colors indicate different proteins.

form back into an improved musical jects taken in space by means of Fou­ diffused into general antenna technol­
signal. Although these ingenious pro­ rier analysis. Planetary probes and ogy. NASA commended the maps of
grams run individually at speeds rival­ earth-orbiting satellites transmit imag­ the sun for contributing to the safety
ing the fast Hartley transform, a sin­ es to the earth as a series of radio im­ of the lunar astronauts.
gle Hartley program suffices for both pulses. A computer transforms these I have also applied the Hartley trans­
transforming a real function into a impulses by Fourier techniques. The form to other studies. Recently my
Hartley transform and converting the computer then adjusts various com­ colleague John D. Villasenor and I de­
transform, after the desired filtering, ponents of each transform to enhance scribed an optical method for finding
back to a real function. Therefore, ex­ certain features and remove others­ the Hartley transform, a development
tra computer memory for storing two much as noise can be removed from that enables Fourier phase and ampli­
programs is not required. the Fourier transforms of recorded tude to be encoded in a single real
music. Finally, the altered data are image. We have also developed a de­

I
n the most general terms, Fou­ converted back to reconstruct the im­ vice that constructs the Hartley trans­
rier and Hartley transforms have age. This process can sharpen focus, form using microwaves. I am now writ­
been applied in fields that contend filter out background fog and change ing papers on solar physics in which
with fluctuating phenomena. Their contrast. transform techniques underlie new
field of application is thus very broad The Fourier transform is also valu­ ways of analyzing data from sunspot
indeed. able in plasma phYSiCS, semiconduc­ counts and from the thickness of sedi­
Many applications exist in biolo­ tor phYSiCS, microwave acoustics, seis­ mentary layers on the earth.
gy. In fact, the double-helix form of mography, oceanography, radar map­ The wide use of Fourier's meth­
DNA was discovered in 1962 through ping and medical imaging. Among the od and related analytical techniques
X-ray diffraction techniques and Fou­ many applications in chemistry is the makes what Lord Kelvin said in 1867
rier analysis. A beam of X rays was fo­ use of the Fourier-transform spec­ just as true today: "Fourier's theorem
cused on a crystal of DNA strands, and trometer for chemical analysis. is not only one of the most beautiful
the X rays were diffracted by the mol­ Fourier analysis has proved valuable results of modern analysis, but it may
ecules of the DNA and recorded on in my own work in two-dimension­ be said to furnish an indispensable
film. This diffraction pattern provided al imaging. In 195 6 I stumbled on a instrument in the treatment of nearly
the amplitude information of the crys­ "projection slice" theorem that yield­ every recondite question in modern
tal structure's Fourier transform. The ed a way to reconstruct images from physics."
phase information, which the photo­ strip integrals, a problem now wide­
graphs alone did not provide, was de­ ly known as tomographic reconstruc­
duced by comparing the DNA diffrac­ tion. Later, I hit on the "modified back­ RJRTHER RFADING
tion pattern with patterns produced projection" algorithm, now universally JOSEPH FOURIER: THE MAN AND THE

by similar chemicals. From the X-ray used in computer-assisted X-ray to­ PHYSICIST. John Herviel. Clarendon
Press, 1975.
intensity and phase information in the mography, or CAT scanning.
THE FOURIER TRANSFORM AND ITS AP­
Fourier transform, biologists worked I was also interested in reconstruct­
PLICATIONS, SECOND EDITION, REVISED.
back to a crystal structure-the origi­ ing images based on data from radio Ronald N. Bracewell. McGraw-Hili Book
nal function. In recent years, X-ray dif­ astronomy. I wanted to pinpoint sour­ Company, 1986.
fraction studies combined with such ces of radio waves on the sun's sur­ THE HARTLEY TRANSFORM. Ronald N.
"reverse" Fourier analysis have re­ face, so I applied transform methods Bracewell. Oxford University Press,
vealed the structure of many other to the design of a scanning radio tele­ 1986.
OPTICAL PHASE OBTAINED BY ANALOGUE
biological molecules and more com­ scope that made daily microwave tem­
HARTLEY TRANSFORMATION. John Vil­
plex structures, such as viruses. perature maps of the sun for 11 years.
lasenor and R . N. Bracewell in Nature,
The National Aeronautics and Space The methods led to the first antenna
Vol. 330, No. 6 150, pages 735-737; De­
Administration improves the clarity with a beam sharper than the resolu­ cember 24, 1987.
and detail of pictures of celestial ob- tion of the human eye and have since

SCIENTIFIC AMERICAN June 1989 95

© 1989 SCIENTIFIC AMERICAN, INC

This content downloaded from

143.89.25.192 on Fri, 27 Sep 2024 06:33:51 UTC
All use subject to https://about.jstor.org/terms