Showing papers on "Fast Fourier transform published in 1970"
••
TL;DR: It is shown that the discrete equivalent of a chirp filter is needed to implement the computation of the discrete Fourier transform (DFT) as a linear filtering process, and that use of the conventional FFT permits the computations in a time proportional to N \log_{2} N for any N.
Abstract: It is shown in this paper that the discrete equivalent of a chirp filter is needed to implement the computation of the discrete Fourier transform (DFT) as a linear filtering process. We show further that the chirp filter should not be realized as a transversal filter in a wide range of cases; use instead of the conventional FFT permits the computation of the DFT in a time proportional to N \log_{2} N for any N, N being the number of points in the array that is transformed. Another proposed implementation of the chirp filter requires N to be a perfect square. The number of operations required for this algorithm is proportional to N^{3/2} .
410 citations
••
TL;DR: A novel structure for a hardwired fast Fourier transform (FFT) signal processor that promises to permit digital spectrum analysis to achieve throughput rates consistent with extremely wide-band radars is described.
Abstract: This paper describes a novel structure for a hardwired fast Fourier transform (FFT) signal processor that promises to permit digital spectrum analysis to achieve throughput rates consistent with extremely wide-band radars. The technique is based on the use of serial storage for data and intermediate results and multiple arithmetic units each of which carries out a sparse Fourier transform. Details of the system are described for data sample sizes that are binary multiples, but the technique is applicable to any composite number.
127 citations
••
TL;DR: It is shown how the Kronecker product can be mathematically defined and efficiently implemented using a matrix factorization method and a generalized spectral analysis is suggested, and a variety of examples are presented displaying various properties of the decompositions possible.
Abstract: A technique is presented to implement a class of orthogonal transformations on the order of pN log p N operations. The technique is due to Good [1] and implements a fast Fourier transform, fast Hadamard transform, and a variety of other orthogonal decompositions. It is shown how the Kronecker product can be mathematically defined and efficiently implemented using a matrix factorization method. A generalized spectral analysis is suggested, and a variety of examples are presented displaying various properties of the decompositions possible. Finally, an eigenvalue presentation is provided as a possible means of characterizing some of the transforms with similar parameters.
123 citations
••
TL;DR: The computed directions are in good agreement with those determined by manual methods with indications that computed speeds are more accurately resolved than those determined manually.
106 citations
••
TL;DR: The fast Fourier transform algorithm provides a mechanism for implementing the sound spectrogram efficiently, and is often useful to generate spectrograms digitally, online.
Abstract: An important aid in the analysis and display of speech is the sound spectrogram, which represents a time-frequency?intensity display of the short-time spectrum.1-3 With many modern speech facilities centering around small or medium-size computers, it is often useful to generate spectrograms digitally, online. The fast Fourier transform algorithm provides a mechanism for implementing this efficiently.
104 citations
••
TL;DR: This paper derives explicit expressions for the mean square error in the FFT when floating-point arithmetics are used, and upper and lower bounds for the total relative meansquare error are given.
Abstract: The fast Fourier transform (FFT) is an algorithm to compute the discrete Fourier coefficients with a substantial time saving over conventional methods. The finite word length used in the computer causes an error in computing the Fourier coefficients. This paper derives explicit expressions for the mean square error in the FFT when floating-point arithmetics are used. Upper and lower bounds for the total relative mean square error are given. The theoretical results are in good agreement with the actual error observed by taking the FFT of data sequences.
89 citations
••
TL;DR: In this article, the authors used the convolution properties of discrete Fourier transforms and fast Fourier transform to solve convolution-type integral equations even when the data sequences consist of hundreds or thousands of points.
Abstract: Radiography is a common diagnostic tool in the field of medicine. The physical process by which a radiograph is formed imposes fundamental limits on the resolution of a radiograph, however. The problem of increasing the resolution of a radiograph can be formulated as the solution of a convolution-type integral equation, but solving this integral equation is extremely difficult if noise is present in the data. A technique for solving integral equation problems was discovered several years ago, but it requires matrix inversion and is unwieldy for data sequences of hundreds or thousands of points. In this article it is shown how this existing technique can be adapted to solving convolution-type integral equations. By using the convolution properties of discrete Fourier transforms and the fast Fourier transform, it is possible to solve convolution-type integral equations even when the data sequences consist of hundreds or thousands of points.
87 citations
••
TL;DR: In this paper, an analysis of roundoff errors occurring in the floating-point computation of the fast Fourier transform is presented, and upper bounds for the ratios of the root-mean-square (RMS) and maximum roundoff error in the output data to the RMS value of the input data for both single and multidimensional transformations are derived.
Abstract: This paper presents an analysis of roundoff errors occurring in the floating-point computation of the fast Fourier transform. Upper bounds are derived for the ratios of the root-mean-square (RMS) and maximum roundoff errors in the output data to the RMS value of the input data for both single and multidimensional transformations. These bounds are compared experimentally with actual roundoff errors.
57 citations
•
28 May 1970
TL;DR: In this paper, the Cooley-Tukey algorithm was used to calculate the Fast Fourier Transform (FFT) in an in-place operation in dual buffers in successive iterations.
Abstract: A real time digital Fourier analyzer using the Cooley-Tukey algorithm for calculating the Fast Fourier Transform. An arithmetic unit simultaneously performs the two complex calculations Ar = Br + Wr Cr and A(r n/2) = Br - Wr Cr. The calculated results are simultaneously stored and retrieved in an in-place operation in dual buffers in successive iterations. Addressing of the buffers uses a single binary address counter and sequential bit complementing for simultaneously addressing both buffers. Parity checking of the binary counter output address controls multiplexing logic to selectively address the buffers to store the calculated results into the desired buffer storage locations.
42 citations
•
21 Aug 1970
TL;DR: In this paper, a generalized convolution of values from two matrices of complex values Ao through Am and Bo through Bn respectively is presented. But this method is not suitable for vector and matrix algebra, linear programming, and transformation weighting and skirting operations.
Abstract: Method and apparatus of computing a generalized convolution of values from two matrices of complex values Ao through Am and Bo through Bn respectively. The formula used in the computation of each complex vector element Ck of the generalized convolution is WHERE P and U specify the increment for each succeeding element involved in a single convolution from each sequence respectively, Q and V specify the increments between first elements of successive convolution coefficients, in each sequence, respectively, and R and W specify the first pair of elements used in forming Co. PC specifies the number of Ck's to be computed. This computation has wide applicability to such allied mathematical operations as vector and matrix algebra, linear programming and a wide variety of transformation weighting and skirting operations such as Bessel function weighting, Hanning windows, complex Kernal transformations, and fast Fourier transforms. In addition, the apparatus described has capability to compute various special cases of the generalized equation involving vectors of real values only.
40 citations
••
TL;DR: The Mellin transform was introduced as a technique for restoring images degraded by a class of spatially-variant degradations as mentioned in this paper, which includes the coma aberration and tilt in a cylindrical lens system.
••
TL;DR: This paper presents the results of the fast Fourier transform in sufficient detail that interested nonexperts can obtain the computer algorithm, and the necessary label permutations, and points out the well known utility of base 2.
Abstract: The fast Fourier transform is usually described as a factorization. Recently this has been done in matrix terms. In this paper we present these results in sufficient detail that interested nonexperts can obtain the computer algorithm, and the necessary label permutations. We also count the number of arithmetic operations required in the calculation and point out the well known utility of base 2, both because of mathematical and machine hardware considerations. A simple FORTRAN program based on these ideas is included.
••
TL;DR: It is concluded that the Blackman-Tukey technique is more effective than the FFT approach in computing power spectra of short historic time series, but for long records the fast Fourier transform is the only feasible approach.
Abstract: Since controversy has arisen as to whether the Blackman-Tukey or the fast Fourier transform (FFT) technique should be used to compute power spectra, single and cross spectra have been computed by each approach for artificial data and real data to provide an empirical means for determining which technique should be used. The spectra were computed for five time series, two sets of which were actual field data. The results show that in general the two approaches give similar estimates. For a spectrum with a large slope, the FFT approach allowed more window leakage than the Blackman-Tukey approach. On the other hand, the Blackman-Tukey approach demonstrated a better window closing capability. From these empirical results it is concluded that the Blackman-Tukey technique is more effective than the FFT approach in computing power spectra of short historic time series, but for long records the fast Fourier transform is the only feasible approach.
•
22 Jun 1970
TL;DR: In this paper, a broad band of signal frequencies, whose power spectrum is to be analyzed by the Fast Fourier Transform (FFT) technique, is sampled by heterodyning with various beat frequencies fb and passage of the modulation products through a low-pass filter of bandwidth b/2 where b represents the width of a subband substantially narrower than the overall band of width B.
Abstract: A broad band of signal frequencies, whose power spectrum is to be analyzed by the Fast Fourier Transform (FFT) technique, is sampled by heterodyning with various beat frequencies fb and passage of the modulation products through a low-pass filter of bandwidth b/2 where b represents the width of a subband substantially narrower than the overall band of width B. By concurrently performing the heterodyning operation in two parallel channels, with introduction of a 90 DEG phase shift between beat and input frequencies in one of the channels, the two sidebands fb +fx and fb -fx (where fx represents any frequency within the selected subband b) can be separated in the outputs of the two channel filters. Frequency limit b/2 is selected in conformity with the capacity of an associated computer to handle the data from the FFT analysis of the subband spectrum. The sampling may be preceded by a transposition of the entire band B to a higher frequency range, in order to prevent any possible cluttering of the spectrum by harmonics of fb .
••
TL;DR: The organization of a special-purpose digital processor for performing nonrecursive digital filtering is described and the cascade organization of the processor allows processing at very high speeds.
Abstract: The organization of a special-purpose digital processor for performing nonrecursive digital filtering is described. The processor uses two complementary cascade fast Fourier transformers. Each transformer can simultaneously transform two independent data blocks of length N words using \log_{2} N arithmetic units and 3/2 N complex words of digital storage. Continuous filtering is achieved by sectioning the input signal, performing a fast transform on each section, multiplying by the frequency characteristics of the desired filter, and inverse transforming. The cascade organization of the processor allows processing at very high speeds. Word rates in excess of 3 MHz are possible with currently available hardware.
••
TL;DR: New and simple derivations for the two basic FFT algorithms are presented that provide an intuitive basis for the manipulations involved and reduce the operation to the calculation of a large number of simple two-data-point transforms.
Abstract: The fast Fourier transform (FFT) provides an effective tool for the calculation of Fourier transforms involving a large number of data points. The paper presents new and simple derivations for the two basic FFT algorithms that provide an intuitive basis for the manipulations involved. The derivation for the "decimation in time" algorithm begins with a crude analysis for the zero frequency and fundamental components using only two data samples, one at the beginning and the second at the midpoint of the period of interest. Successive interpolations of data points midway between those previously used result in a refinement of the amplitudes already determined and a first value for the next higher order coefficients. The derivation of the "decimation in frequency" algorithm begins by resolving the original data set into two new data sets, one whose transform includes only even harmonic terms and a second whose transform includes only odd harmonic terms. Since the first of the two new data sets repeats after the midpoint, it can be transformed using only the first half of the data points. The second of the new data sets is multiplied by the negative fundamental function, thereby reducing its order by one and converting it into a data set that transforms into even harmonics only; in this form it can also be transformed using only the first half of the data set. Successive applications of this procedure result finally in reducing the operation to the calculation of a large number of simple two-data-point transforms.
••
TL;DR: It is shown that in a high-radix cascade processor, significant reduction in memory can often be obtained with a modest increase in complexity of the arithmetic units, as shown by Comparison of the logic organizations of serial and cascade fast Fourier processors.
Abstract: The subject of this paper concerns the analysis of high-radix FFT algorithms appropriate to hardware implementation of nonrecursive filters and the error propagation properties associated with these algorithms. Variations in the structure of high-radix fast Fourier transforms are illustrated by means of their Kronecker product expansions. The implications in hardware design of the different structures are discussed. These implications include tradeoffs between memory size and throughput rate as shown by Comparison of the logic organizations of serial and cascade fast Fourier processors. It is shown that in a high-radix cascade processor, significant reduction in memory can often be obtained with a modest increase in complexity of the arithmetic units. The relationship between the permutation matrices specified in a Kronecker factorization, and the structure and addressing of memory is brought out. As an example, the implementation of a nonrecursive digital filter is included using a radix-4 factorization which provides for an extremely simple memory organization. The permutation matrix is such that both data storage and addressing are provided using serial MOS shift registers. In order to compare the accuracy of high-radix algorithms relative to the base-2 algorithm, a fixed-point simulation study of the error propagation properties was conducted of a full radix-4 and a radix-16 FFT for N= 1024 and 256, respectively. The computational errors were compared with corresponding results obtained from the base-2 algorithm. The results of the simulations were also compared with a model which is a modification of that reported by P. D. Welch. The modification consists of the representation of roundoff error buildup in the radix-2 and radix-4 FFT in closed form by a nonhomogeneous difference equation. The illustrative example chosen is the lower bound discussed by Welch.
••
TL;DR: It has been shown that recursive digital filters can be synthesized using the fast Fourier transform and an algorithm for computer implementation has been developed and used in comparing the computation times and noise figures of filters synthesized in this manner with the computation time and noise figure of filters synthesis by recursion.
Abstract: It has been shown that recursive digital filters can be synthesized using the fast Fourier transform. An algorithm for computer implementation has been developed and used in comparing the computation times and noise figures of filters synthesized in this manner with the computation times and noise figures of filters synthesized by recursion. A model has been proposed for analysis of noise in the two-pole filter. Predictions of this model have been found to be in good agreement with noise measurements.
••
TL;DR: In this article, an application of this method in order to improve the resolution of nuclear and magnetic resonance spectra is presented, an emphasis is laid on the possibility of diminishing the noisy secondary effects.
Abstract: The digital filtering was proven to be useful in processing convolved signals. In this paper an application of this method in order to improve the resolution of nuclear and magnetic resonance spectra is presented, An emphasis is laid on the possibility of diminishing the noisy secondary effects. The processing was achieved using the FFT algorithm. Some practical results obtained are delivered.
•
01 Dec 1970
TL;DR: In this article, a method of synthesizing time domain waveforms from data in the frequency domain with improved resolution in the middle and lower frequencies was proposed, where data is broken into groups and each group processed through an inverse Fourier or fast Fourier transform to obtain time domain data which is then added together in the proper time relationship.
Abstract: A method of synthesizing time domain waveforms from data in the frequency domain with improved resolution in the middle and lower frequencies Frequency data is broken down into groups and each group processed through an inverse Fourier or fast Fourier transform to obtain time domain data which is then added together in the proper time relationship The resulting composite digital information is then passed through a digital to analog converter to provide an analog time domain waveform
•
24 Dec 1970
TL;DR: A fast Fourier transform addressing system using the DanielsonLanczos algorithm was proposed in this paper, which uses a basic relationship between addresses to allow implementation of the address system using a reduced number of address counters and reduced storage requirements for system constants.
Abstract: A fast Fourier transform addressing system using the DanielsonLanczos algorithm. The addressing system uses a basic relationship between addresses to allow implementation of the addressing system using a reduced number of address counters and reduced storage requirements for system constants.
••
TL;DR: A simple fast Fourier transformation (FFT) algorithm has been specifically adapted to calculate the experimental radial distribution function and its greatest advantage is its internal consistency—the ability to exactly transform back to the original domain.
Abstract: A simple fast Fourier transformation (FFT) algorithm has been specifically adapted to calculate the experimental radial distribution function. The number of equi-spaced data points must be a power of two [N = 2n for integer n] and must be greater than the Nyquist frequency [N = 2(rmax) (smax)/2π]. When properly defined, the data set is expanded as an odd function. The greatest advantage of the FFT algorithm is its internal consistency—the ability to exactly transform back to the original domain.
22 Jun 1970
TL;DR: This report serves as documentation for a collection of basic time-series analysis programs written for the CDC 3200 digital computers and is constructed around the fast Fourier transform (FFT) ALGORITHM.
Abstract: : The report serves as documentation for a collection of basic time-series analysis programs written for the CDC 3200 digital computers. These programs are predominantly written in FORTRAN and can be easily adapted to other digital computers. These programs are constructed around the fast Fourier transform (FFT) ALGORITHM. Rather than rediscuss the theory of the FFT algorithm, which is adequately described in the existing literature, this report deals with the practical aspects of the use of the FFT. The problems that can and in many instances do occur in computing spectral estimates are also addressed on the basis of an extensive literature review. (Author)
••
TL;DR: In this article, the performance analysis of a balanced diode modulator is described by an iterative process and the fast Fourier transform is used to obtain the spectrum of the output.
Abstract: Equations describing the performance of a balanced diode modulator are solved by an iterative process and the fast Fourier transform is used to obtain the spectrum of the output. The analysis provides an interesting example of the extension, to large-signal applications, of the method of solution recently described for quasilinear systems.
••
TL;DR: In this paper, a physical interpretation of the BIFORE power spectrum is presented, which is invariant under cyclic shifts of the r-dimensional input data, and the corresponding power spectrum consists of Πi = 1r(1+k) spectral points (Nt = 2k).
Abstract: The BIFORE transform is generalised to r dimensions. The corresponding power spectrum consists of Πi = 1r(1+ki) spectral points (Nt = 2ki), which are invariant under cyclic shifts of the r-dimensional input data. A physical interpretation of the BIFORE power spectrum is presented.
••
[...]
TL;DR: The Fast Field Program (FFP) as discussed by the authors is an efficient algorithm for providing propagation estimations rapidly without undue approximations to the basic hydrodynamical field equations, and it has been shown that the FFP is not only feasible but also efficient.
Abstract: The hydrodynamical field equations yield to three methods of solution. Two of these, the method of characteristics and that of separation of variables, have been widely used for both analytical and numerical investigation, but usually with either physical or mathematical approximations. The third, direct integration of the partial differential equations has always existed in principle but not until the advent of the fast Fourier transform could this approach be exploited in a systematic and economical manner. H. W. Marsh recognized this fact and introduced the concept of the fast field program (FFP). Several examples are presented and discussed which demonstrate that the FFP is not only feasible but a highly efficient algorithm for providing propagation estimations rapidly without undue approximations to the basic field equations. The results are compared with those predicted by both ray theory and normal‐mode theory. [This research was supported by the U. S. Underwater Sound Laboratory.]
•
01 Oct 1970
TL;DR: Digital computer programmes which would overcome the disadvantages of conventional digital spectral analysis were developed, and it was concluded that the tec: inique is more powerful than conventional digital.
Abstract: In the field of control engineering there is
a need
to
study
the dynamic behaviour
of systems which are subjected
to
random
disturbances. A technique
which
is
of great practical use
is to
describe the dynamic
properties as a
function
of
frequency. This
involves determining the frequency content, or spectrum, of
the
disturbances,
and
the frequency
response
function
of
the
system.
There
are many analogue and digital techniques which are designed for this
type
of spectral analysis.
However, digital computer
techniques
are
often avoided because they
are slow, and data must
be
collected
'off-line'.
A
recently
discovered
computational method,
termed the fast-
Fourier-transform (FFT),
enables
digital
spectral analysis
to be
carried-
out in
a much shorter
time than
was previously possible.
In
view of
this discovery it
was
decided to develop digital computer programmes
which would overcome
the disadvantages
of conventional
digital
spectral
analysis. Using these
programmes a computer would
be
connected, via
an analogue
to digital interface, to the
signal source, and would process
the data
as
it
entered
the
computer.
In the jargon
of computing,
the
computer would
be 'on-line'
and analyzing
the
spectra
in 'real-time'.
The first
part of
the
project consisted of an
investigation
of
the FFP
when programmed
for an on-line
digital
computer.
The
results of
this investigation
showed
that
a rapid, accurate, and compact
FFT
could be
programmed
by
using
fixed-point
arithmetic, and coding
in
an assembly language. The
speed of
the transform
was sufficient
to
allow spectral analysis over a
frequency
range useful
in
control
applications.
Two
on-line computer programmes
based upon
the YPP were
then
written; one
for 'real-time'
spectral analysis of a single record, and
another
for the 'real-time'
estimation of
the frequency
response
function
relating
two
signals.
In
order
that the
results of
these
programmes could
be
sensibly
interpreted, a statistical study was made
of
the
spectral estimators used
in the
programmes.
Arising from this
study, several contributions
to the field of
digital
spectra. analysis
were made.
These
were : -
1) A
more general covariance relationship
for cross-spectral
estimators.
2) An
examination of aliasing
in digital
spectral estimators.
3) Some theoretical
results concerning spectral estimators
for
closed loop
systems with random
disturbances inside the loop,
Some
experimental work was conducted with
the
real-time'
spectral analysis programmes, and it
was concluded
that the tec:
inique
is
more powerful
than
conventional
digital.
methods
because it is on-
line,
and can provide estimates with
improved
resolution and
statistical stability. Real-time digital
spectral analysis methods also
have the
advantage
that they
may
be
simply and quickly modified
to suit
specific applications.
••
17 Nov 1970TL;DR: With the advent of high-speed integrated circuits and read-only memories to implement sine and cosine tables for the FFT, actual real-time hardware processing has been accomplished.
Abstract: Digital voice processing has advanced to a relatively high level of sophistication due to the development of the fast Fourier transform (FFT) algorithm. Complete vocoder systems have been developed around the FFT and with the advent of high-speed integrated circuits and read-only memories to implement sine and cosine tables for the FFT, actual real-time hardware processing has been accomplished.
••
01 Jan 1970TL;DR: In this paper, the fast Fourier transform (FFT) algorithm was used as a highly efficient numerical means for the analysis of transient scattering phenomena, which can best be exploited when additional characteristics of the radar system such as the actual transmitted waveform and the receiver transfer function are taken into account.
Abstract: The fast Fourier transform algorithm can be used as a highly efficient numerical means for the analysis of transient scattering phenomena. Examples illustrating the performance of this method are given for an incident pulse of Gaussian shape. It is noted that such a technique can best be exploited when additional characteristics of the radar system such as the actual transmitted waveform and the receiver transfer function are taken into account.
01 Jan 1970
TL;DR: In this paper, a modified fast Fourier transform (FFT) is used in a hybrid computer program to permit processing of tracking data during a run to yield the human operator's describing function almost immediately after the data-taking period.
Abstract: A modified fast Fourier transform (FFT) is used in a hybrid computer program to permit processing of tracking data during a run to yield the human operator's describing function almost immediately after the data-taking period. The computer processing time is substantially reduced at no cost in accuracy.