U.S. patent number 3,760,172 [Application Number 05/046,248] was granted by the patent office on 1973-09-18 for method of and apparatus for signal processing.
This patent grant is currently assigned to The Board of Trustees of Leland Stanford Junior University. Invention is credited to Calvin F. Quate.
United States Patent |
3,760,172 |
Quate |
September 18, 1973 |
METHOD OF AND APPARATUS FOR SIGNAL PROCESSING
Abstract
Method of and apparatus for convolution and correlation of
electromagnetic signals by application of two signals to a
piezoelectric medium to establish two acoustic waves which are
propagated in a fashion such that phase matching and frequency
conservation conditions are met whereby parametric coupling
occurs.
Inventors: |
Quate; Calvin F. (Los Altos
Hills, CA) |
Assignee: |
The Board of Trustees of Leland
Stanford Junior University (Stanford, CA)
|
Family
ID: |
21942424 |
Appl.
No.: |
05/046,248 |
Filed: |
June 15, 1970 |
Current U.S.
Class: |
708/815; 307/424;
333/133; 333/187; 310/334; 359/330 |
Current CPC
Class: |
G06G
7/195 (20130101); G01S 13/282 (20130101) |
Current International
Class: |
G06G
7/00 (20060101); G01S 13/28 (20060101); G01S
13/00 (20060101); G06G 7/195 (20060101); G06g
007/19 (); H01v 007/00 () |
Field of
Search: |
;235/181 ;307/308,88.3
;330/5.5 ;333/30 ;310/8.1 ;332/26 |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
Tien : Parametric Simplification and Frequency Mixing in
Propagating Circuits. Journal of Applied Physics Sept. 1958 Vol. 29
No. 9. p. 1347-1357. .
Tseng : Surface Ultrasonic Wave Parametric Amplifier. IBM Tech.
Discl. Bull. Vol. 12 No. 10 3/1970 p. 1699-1700..
|
Primary Examiner: Gruber; Felix D.
Claims
What is claimed is:
1. The method of signal processing in a piezoelectric medium which
comprises the steps of
propagating a first modulated acoustic wave through the medium at a
predetermined phase velocity and propagation vector,
propagating a second modulated acoustic wave through the medium at
a different phase velocity and propagation vector whereby
translation of the acoustic waves occurs,
said waves being propagated in a fashion such that phase matching
and frequency conservation conditions are simultaneously met during
the wave propagation through the medium whereby parametric coupling
results, thus to provide the modulation product of the modulated
wave energy, and
extracting the wave energy of the modulation product to provide the
output signal.
2. The method of claim 1 wherein
the step of propagating the first acoustic wave is carried out by
propagating the first wave in one direction through the
piezo-electric medium, and
the step of propagating the second acoustic wave is carried out by
propagating the second wave in the opposite direction through the
piezoelectric medium.
3. The method of claim 2 wherein,
the propagation vectors of said first and second acoustic waves are
equal but opposite whereby the parametric coupling produces an
output wave with no spatial variation.
4. The method of claim 2 wherein,
the propagation vectors of said first and second acoustic waves are
of different values whereby the parametric coupling produces an
output wave of finite velocity.
5. The method of claim 1 which comprises the steps preceding the
wave propagating steps of
generating said first and second acoustic waves by application of
modulated electromagnetic signals to the piezoelectric medium.
6. The method of claim 5 wherein
the step of generating the first acoustic wave is carried out by
applying a modulated electromagnetic signal to one end of the
piezoelectric medium, and
the step of generating the second acoustic wave is carried out by
applying a modulated electromagnetic signal to the opposite end of
the piezoelectric medium.
7. The method of claim 6 which comprises
varying the frequencies of the electromagnetic signals so that one
constitutes the frequency mirror image of the other.
8. The method of claim 6 which comprises the additional steps of,
generating a swept-frequency pulse to provide one signal and,
inverting the pulse to form the mirror-image signal.
9. The method of claim 1 wherein
the step of propagating the first acoustic wave is carried out by
propagating the first wave through the piezoelectric medium in a
predetermined direction at a predetermined phase velocity, and
the step of propagating the second acoustic wave is carried out by
propagating the second wave through the piezoelectric medium in the
same direction but at a different phase velocity than that of the
first acoustic wave.
10. The method of claim 9 which comprises the steps of,
generating a shear acoustic wave by application of one
electromagnetic signal to the piezoelectric medium at a given time,
and
generating a longitudinal acoustic wave by application of another
electro-magnetic signal to the piezoelectric medium at a later
time.
11. The method of claim 1 wherein,
the energy extraction step is achieved by coupling to an external
circuit tuned to the combined frequencies of the
parametrically-coupled acoustic waves.
12. The method of claim 1 which comprises the steps preceding the
wave propagating steps of
generating the first acoustic wave directly by coupling
electromagnetic energy to the piezoelectric medium, and
generating the second acoustic wave indirectly by coupling
electromagnetic energy to the medium in a fashion such that an
electrical polarization exists which parametrically couples with
the first acoustic wave to generate the second acoustic wave.
13. The method of claim 1 wherein,
said first acoustic wave has an unknown modulation pattern, and
said second acoustic wave has a known modulation pattern consisting
of a series of short pulses spaced at predetermined intervals and
which comprises the additional step of coupling electromagnetic
signals generated by parametric coupling of the waves from the
medium at intervals equivalent to the predetermined pulse intervals
of said second acoustic wave.
14. The method of claim 1 which comprises the steps preceding the
wave propagation steps of
generating said first acoustic wave with a modulation pattern
consisting of a finite determined sequence of separate and
distinctly shaped signals, all of which simultaneously exist in the
piezoelectric medium, and
generating said second acoustic wave with a modulation pattern
consisting of unknown signals corresponding to one of the finite
sequence of signals of said first wave.
15. The method of claim 1 which comprises the steps preceding the
wave propagation steps of
generating said first acoustic wave by applying an electromagnetic
signal of unknown frequency to the piezoelectric medium,
generating said second acoustic wave by applying an electromagnetic
signal of known but variable frequency to the piezoelectric medium,
and
varying the frequency of said variable signal until the conditions
for parametric coupling are established.
16. The method of claim 1 which comprises the steps of,
generating a pulse-modulated electromagnetic signal,
applying such signal to the piezoelectric medium at a controlled
but variable time to generate said first acoustic wave transmitting
said pulse-modulated signal,
receiving a pulse-modulated signal constituting a reflected version
of the transmitted signal and applying said reflected signal to the
piezoelectric medium to generate said second acoustic wave whereby
parametric coupling occurs if phase coherence of the two signals
exists in the medium.
17. The method of claim 16 wherein,
said pulse-modulated electromagnetic signal is frequency-modulated,
and which comprises the additional step of inverting one of the
signals applied to the medium to generate said first and second
acoustic waves.
18. Signal processing apparatus which comprises
a piezoelectric medium,
a pair of transducers adjacent said medium to permit coupling of
electromagnetic signals into said piezoelectric medium whereby
acoustic waves are propagated therethrough,
means for applying modulated signals to said transducers in a
manner such that different phase velocities of the acoustic waves
exist within said piezoelectric medium and wave translation occurs
so that the conditions of parametric coupling, phase matching and
frequency conservation exist during propagation of the acoustic
waves through the medium, thus to provide the modulation product of
the modulated wave energy, and
external circuit means for coupling the product modulation wave
energy generated by the parametric interaction of the modulated
acoustic waves.
19. Signal processing apparatus according to claim 18 wherein,
said external circuit means constitutes a resonant cavity
encompassing said piezoelectric medium.
20. Signal processing apparatus according to claim 18 wherein,
said external circuit means constitutes a folded strip line
adjacent said piezoelectric medium.
21. Signal processing apparatus according to claim 18 wherein,
said transducers are disposed at opposite ends of said
piezoelectric medium wherefore the acoustic waves propagate in
opposite directions.
22. Signal processing apparatus according to claim 18 wherein,
said transducers are disposed at the same end of said piezoelectric
medium.
23. Signal processing apparatus according to claim 18 which
comprises,
means for varying the frequency of one of the applied
electromagnetic signals.
24. Signal processing apparatus according to claim 18 which
comprises,
means for pulse modulating one of the applied electromagnetic
signals.
25. Signal processing apparatus according to claim 24 which
comprises,
means for frequency modulating each pulse of the pulse modulated
signals.
26. Signal processing apparatus according to claim 18 which
comprises,
means for modulating one of the applied electromagnetic signals to
provide a sequence of patterns, each having a predetermined
distinct shape.
Description
FIELD OF THE INVENTION
The present invention relates generally to signal processing and
more particularly, to a method of and apparatus for obtaining the
convolution and correlation of signals. The invention described
herein was made in the course of work under a grant or award from
the United States Air Force.
BACKGROUND OF THE INVENTION
Mathematically, as explained in detail in Chapter 3 of Bracewell,
The Fourier Transform And Its Applications (1965), the correlation
of two time functions f(t) and g(t) is defined by
Cr = .intg. f(.tau.) g(.tau.-t)d.tau. (1)
and convolution, in turn, is the time reversed relationship defined
by
Cn = .intg. f(.tau.) g(t-.tau.)d.tau. (2)
t, in each equation being representative of time displacement of
one function relative to the other. Accordingly, the mathematical
evaluation of either operation can be considered as a process
whereby first, the two functions are translated in time with
respect to one another by a specified amount, t, secondly, the
product of the translated functions is taken, and finally, this
product is integrated. Quite obviously, the mathematical process
can be physically realized if two signals can be made to undergo
the three steps of the process, translation, multiplication, and
integration.
In a rather complex fashion, the correlation function has been
physically realized in a number of instances, one well established
application being found in the so-called "correlation radars"
wherein the cross-correlation of the transmitted signal with the
reflected signal, a time-delayed replica of the transmitted signal,
is specifically defined by the equation
Cc = .intg. f(.tau.) f(.tau.-t)d.tau. (3)
The value of t now constituting the time taken by the radar wave in
traveling to the target and back. It has been found that the
cross-correlation process constitutes a measure of the coherence
between the transmitted and received signals and thus has provided
far reaching significance in solving a major problem encountered in
any radar operation, that of detecting the received signal against
a background of noise.
A number of cross-correlation radar receivers have been designed
and for the most part employ analog and digital techniques. For
example, one well known technique couples the input signal to a
digital computer by means of an analog-to-digital converter and the
computer is programmed to cross-correlate, point by point, the two
signals and to extract them from the noisy background. This
requires a sizeable computer memory and a relatively large amount
of computer time. Thus in addition to the complexity of such
installations, the cross-correlation function is not immediately
available, that is, the function is not performed substantially
simultaneously, or in other words, in real time.
More commonly, "correlation radars" employ a "matched filter" as
mentioned in Chapter 9 of Introduction To Radar Systems, by Skolnik
(1962) along with the cross-correlation detectors mentioned
hereinabove and as discussed in detail in the "matched filter"
issue of IRE Transactions on Information Theory, Volume IT-6 (June,
1960), the matched filter basically constituting a filter designed
so that its output is proportional to the "correlation" of the
signal with itself. Obvious problems in matched filter design are
encountered with complex modulation waveforms such as the noise
waveform utilized to preclude electronic countermeasure techniques.
Yet other sophisticated modulation waveforms also present problems
in matched filter utilization. For example, if as commonly
provided, two radar pulses are emitted sequentially by the radar
transmitter with different modulation characteristics, two matched
filters are obviously necessary for appropriate detection, one
filter being matched to the first pulse and the other matched to
the second pulse.
In spite of the fact that complexity and other practical
difficulties of the type mentioned hereinabove do exist in
correlation radars, the convolution and correlation processes are
recognized as powerful tools in the processing and analysis of
signals and in addition to the mentioned radar utilizations, Lee
and Weisner have employed cross-correlation in the characterization
of linear systems as reported in their Statistical Theory of
Communication, and other have applied cross-correlation to the
analysis of brain waves, vibration analysis and any number of
applications where the comparision of two signals provides useful
data, particularly, if such data can be extracted and presented
substantially simultaneously to the investigator.
SUMMARY OF THE PRESENT INVENTION
Generally, it is the objective of the present invention to provide
a method of and apparatus for convolution and/or correlation of
signals in real time and in a relatively simple fashion through
utilization of the nonlinear acoustic interaction of the signals in
a piezoelectric medium. In accordance with the invention the three
steps of translation, multiplication and integration are achieved
by introducing the two signals into the piezoelectric medium in a
particular and precisely defined fashion by application of suitable
electromagnetic signals thereto. The phase velocities of the two
signals which are propagated through the piezoelectric medium in
the form of modulated acoustic waves are different thus to achieve
the requisite translation step. For example, in one specific case
of correlation, the first signal is introduced at one time with a
predetermined phase velocity and the second signal is introduced at
a later time but with a higher phase velocity so as to overtake the
first signal during the period of signal propagation through the
medium. In one specific case of convolution, in turn, the two
signals are introduced at opposite ends of the medium so as to meet
during their propagation therethrough again to provide the
translation step in the form of a time reversal as requisite to
meet the conditions of the convolution definition.
The second step of multiplication is provided through the
well-established mechanism of parametric coupling which briefly
provides that if the phases of the signals are matched and
simultaneously the principle of frequency conservation is observed,
a nonlinear interaction of the acoustic waves results which will
induce an electric polarization that is proportional to the product
of the two modulation functions of the acoustic waves.
The third step, integration, is readily achieved by known
techniques for driving an external circuit with the induced
polarization. For example, if microwave frequencies are employed in
a fashion such that no spatial variation of the polarization
occurs, the induced polarization can be utilized to drive an
appropriate mode of a microwave cavity. In the case of convolution,
this cavity would be tuned to the sum of the frequencies of the two
signals whereas to the contrary, in the case of correlation, the
cavity would be tuned to the correlation signal which would
constitute the difference frequency in accordance with the
frequency conservation principle of parametric interaction, as
mentioned hereinabove.
If, as an alternative, the acoustic waves are parametrically
coupled so that the induced polarization, D, has a resultant
propagation vector, the external circuit can then be arranged, for
example in the form of a suitably shaped strip line, to enable
coupling to the propagating polarized output electromagnetic
signal.
The external output circuit in either the cavity or travelling wave
form as described can be reversed in its function, serving as the
input for one signal which through appropriate parametric coupling
in the piezoelectric medium with a second signal (acoustic) can
produce essentially a reserved operation of the
convolution/correlation process to produce an output signal, for
example, at one end of the piezoelectric medium.
Generally then, the method of signal processing in accordance with
the present invention involves the steps of propagating a first
acoustic wave through a piezoelectric medium at a predetermined
phase velocity, propagating a second acoustic wave through the same
medium at a different phase velocity, the two waves being
introduced and propagated in a fashion such that the conditions of
phase matching and frequency conservation are met to provide a
nonlinear parametric interaction whereupon the induced energy can
be extracted to provide the output data. The acoustic waves can be
of any known type such as volume waves, surface waves, flexural
waves, torsional waves, Love waves or the like.
The general characteristics of the described signal processing
method involving the nonlinear interaction of acoustic waves in a
piezoelectric medium lend themselves to a variety of applications.
By way of example, and as will be explained in detail hereinafter,
no output will be observed unless the conditions of parametric
coupling are met. Because of the frequency conservation conditions,
if a known signal is introduced into a piezoelectric medium, a
second unknown signal will cause the process to occur only if it,
in turn, is of a precisely delineated frequency. Thus, the
immediate application of the mechanism to a narrow bandpass filter
suggests itself.
Furthermore, because of the requirement for phase matching,
regardless of the complexity of the introduced signals, an output
will be obtained only when this condition is also met unambigously,
thus allowing for example, the precise detection of a reflected
radar signal otherwise obscured by attendant noise.
Additionally, with respect to radar systems wherein pulses having
different modulation patterns are utilized, the necessity for
utilizing a plurality of "matched filters" is obviated since a
number of known signals can be supplied through the piezoelectric
medium for comparison with the reflected signals, thus simplifying
the detection of sophisticated radar signals.
Extrapolating, the possibility of introducing a number of known
signals into the piezoelectric medium for comparison with unknown
signals leads one to the obvious application of the present method
to a pattern recognition scheme wherein, in essence, a number of
known patterns can be compared with an incoming unknown pattern so
that, through the use of the correlation mechanism, one can
establish the unambigous recognition of the unknown signal
pattern.
Additionally, if one introduces a number of spaced output circuits
along a piezoelectric medium, a series of equivalently-spaced short
pulses can be introduced to one end of the medium to provide
sampling of the waveform of an unknown signal of longer duration
introduced to the opposite end of the medium.
Any number of additional applications will suggest themselves
immediately but it is to be observed that, in each case, the
relative slow velocity of acoustic waves makes it possible to carry
out the convolution or correlation operations in a piezoelectric
crystal of very convenient size with signals several microseconds
in duration and the results of the operations are immediately
available to the investigator. Not only are the results available
in real time, but are available through utilization of a relatively
simple and inexpensive structure as compared to either the
mentioned matched filter or other correlation detection
equipment.
BRIEF DESCRIPTION OF THE DRAWINGS
The stated objective of the invention and the manner in which it
may be achieved as summarized hereinabove will be more readily
understood by reference to the following detailed description of
the exemplary structures shown and explained in the accompanying
drawings wherein:
FIG. 1 is a diagrammatic perspective view of a structure embodying
the present invention for carrying out a convolution operation,
FIG. 2 is a frequency-propagation diagram explanatory of the
operation of the FIG. 1 structure.
FIG. 3 is a view similar to FIG. 1 diagrammatically depicting a
structure arranged to perform a correlation operation,
FIG. 4 is a frequency-propagation diagram similar to FIG. 2
explanatory of the principles of the FIG. 3 structure,
FIG. 5 is a diagrammatic perspective view of a structure arranged
to enable output coupling of a traveling wave,
FIG. 6 is a frequency-propagation diagram similar to FIGS. 1 and 3
explanatory of the operation of the FIG. 5 structure,
FIG. 7 is a diagrammatic side elevational view of a structure
enabling sampling of the waveform of an acoustic wave,
FIG. 8 illustrates a structure employing the convolution mechanism
as a bandpass filter,
FIG. 9 is a diagram illustrating the bandpass characteristics of
the FIG. 8 filter,
FIG. 10 is a block diagram illustrating application of the
invention to a radar system, and
FIG. 11 is another block diagram illustrating the invention as
applied to a pattern recognition system.
DETAILED DESCRIPTION OF THE EXEMPLARY EMBODIMENTS OF THE
INVENTION
The method of signal processing can be specifically utilized to
carry out a convolution operation in the manner diagrammatically
illustrated in the FIG. 1 structure where two input electromagnetic
signals, which may constitute modulated microwave frequency
signals, are applied through suitable transducers diagrammatically
indicated at 10 and 12 to opposite ends of a piezoelectric crystal
14 so as to generate two oppositely propagating fast shear acoustic
waves indicated at A.sub.1 and A.sub.2 along what may be
denominated the X-axis of the crystal. It will be immediately
obvious that the opposite propagation of the two acoustic waves
A.sub.1, A.sub.2 provides signal translation, requisite as a first
step in the evaluation of the convolution equation (2).
In order to carry out the second evaluation step, that of
multiplication, the frequency and phase relationships of the two
acoustic waves A.sub.1 and A.sub.2 are chosen so that the
conditions for parametric coupling, phase matching (k.sub.1 +
k.sub.2 = k.sub.p) where k, the propagation vector, is equal to the
frequency, .omega., divided by the acoustic velocity, v, and
frequency conservation (.omega..sub.1 + .omega..sub.2 =
.omega..sub.p) are met in the manner explained in detail in Chapter
5 of W.H. Louisell, Coupled Mode and Parametric Electronics, 1960.
More particularly, with reference to the frequency-propagation
diagram of FIG. 2, for the convolution process, the first acoustic
wave A.sub.1, constitutes a forward traveling wave (k.sub.s) with a
frequency, .omega..sub.s, whereas the second acoustic wave A.sub.2
is a backward traveling wave (-k.sub.s) at the same frequency,
.omega..sub.s. Even though the phase velocities of the two waves as
represented by the angles of the vectors of the waves A.sub.1 and
A.sub.2 in FIG. 2 have the same absolute value, they are
"different" in the sence that one is negative and the other
positive so that the required translation occurs. Introducing these
values into the parametric equation for phase matching we have,
k.sub.s + (-k.sub.s) = o and, in turn, .omega..sub.s +
.omega..sub.s = 2 .omega..sub.s provides the specific values of
frequency conservation. Since k.sub.p = o, there is no spatial
variation in the output electric displacement or polarization, D,
of the parametrically-combined waves (i.e., it has zero velocity)
and it has a frequency which is the sum of the input wave
frequencies, that is, .omega..sub.p = 2 .omega..sub.s, as shown in
FIG. 2.
The mentioned multiplication function is obtained through the
nonlinear interaction of the acoustic waves, A.sub.1 and A.sub.2,
so that the displacement or polarization, D, is proportional to the
product of the strain amplitudes, S.sub.1 and S.sub.2, of the
amplitude modulated acoustic waves and is directed along the Z-axis
of the crystal throughout its entire volume as indicated by the
arrows D in FIG. 1. This result stems from the well established
effect that the polarization or displacement, D, is a function of
both the electric field, E, and the strain, S, and in simplified
(non-tensor) notation can be represented by the following
equation
D = B(S.sub.1 + S.sub.2) + C(E.sub.1 + E.sub.2) + F(E.sub.1 +
E.sub.2).sup.2 + G(E.sub.1 + E.sub.2)(S.sub.1 + S.sub.2)+
(K/2)(S.sub.1 + S.sub.2).sup.2 (4)
wherein B, C, F, G and K are constants E.sub.1 and E.sub.2 are the
electric fields and S.sub.1 and S.sub.2 the strain amplitudes of
the two acoustic waves A.sub.1 and A.sub.2. The first two terms
will be recognized as linear, the third as the electro-optic
coefficient which relates to the change of dielectric constant with
electric field, and the fourth as the photoelastic constant which
relates to the change in dielectric with strain. The fifth term,
which is that critical to the operation of the present invention,
is related to the change in the velocity of sound with electric
field as explained in detail in Piezoelectric Crystals And Their
Applications To Ultrasonics by W.P. Manson, 1950 (page 463). In the
case under consideration here E.sub.1 and E.sub.2 are zero, so that
only the fifth term is of interest and the foregoing equation can
be reduced to the simple phenomenological equation
D = KS.sub.1 S.sub.2 (5)
we see that, D, is proportional to the product of the strain
amplitudes S.sub.1 and S.sub.2 and in the case, for example, of
fast shear acoutic waves propagating along the X-axis of lithium
niobate, the constant, K, has a value of 10 coulombs/meter.sup.2 so
that detectable electric polarization, D, is provided with input
strain amplitudes of the order of 10.sup..sup.-6 (1
watt/cm..sup.2).
In order to carry out the third step of the process, integration,
the product of the two strain amplitudes, S.sub.1 and S.sub.2, are
summed as the two signals S.sub.1 and S.sub.2 are translated
relative to one another. If one chooses to operate at microwave
frequencies and D has no spatial variation as explained hereinabove
the output resonant circuit can constitute a microwave cavity 16
coupled to an output waveguide 18. As explained by Harrington in
Chapter 8 of Time-Harmonic Electromagnetic Fields (1961). The
output power, P, radiated into a coupled wave guide, is given
by
P = [2.omega..sub.s .beta.Q.sub.o /(1 + .beta.).sup.2
][.intg.D.sub.2 .omega..sub.s .sup.. E.sub.i dV].sup.2 (6)
where .beta. is the coupling of the cavity to the guide, Q.sub.o,
the unloaded Q of the cavity, D, the driving polarization, E.sub.i,
the normalized mode amplitude, and the integration is performed
over the volume, V, of the cavity. In the present instance, the
spatial variation of E.sub.i and transverse variations of D can be
neglected to first order and subtitution of equation (5) into
equation (6) under such conditions yields the result ##SPC1##
where L is the length of the crystal.
It, in turn, one transforms variables so that .tau. = t - x/v,f(t)
= S.sub.1 (o,t), g(t) = S.sub.2 (L,t - L/v), and v is the acoustic
velocity, the equation can be rewritten as ##SPC2##
which can be immediately recognized as the convolution of the
amplitude modulation functions, S.sub.1 and S.sub.2, of the
acoustic waves.
Whereas in FIG. 1, the acoustic waves A.sub.1, A.sub.2 are
propagated as a result of the introduction of input signals to the
opposite ends of the crystal 14, and the output signal is extracted
through the cavity 16, the process can obviously be reversed so
long as the conditions for parametric coupling are retained. Thus,
an input electromagnetic signal at a frequency 2.omega..sub.s and
propagation vector of zero can be introduced through the cavity 16
into the crystal 14 and a second input signal at frequency
.omega..sub.s and propagation vector, + k.sub.s can be introduced
to one end of the crystal, to provide an output signal at this same
end of the crystal resultant from a backward wave of frequency
.omega..sub.s (2 .omega..sub.s - .omega..sub.s - .omega..sub.s) and
propagation vector -k.sub.s (o - k.sub.s = -k.sub.s) in accordance
with the frequency conservation and phase matching conditions
requisite for the parametric coupling.
In the described cases, the acoustic waves have travelled in
opposite directions to provide the translation necessary for
acoustic interaction. As an alternative, translation can occur if
two acoustic waves travel in the same direction but at different
velocities. The result is a correlation operation as depicted in
FIG. 3. More particularly, two input electromagnetic signals are
introduced to the same end of a piezoelectric crystal 20 through
suitable transducers 22, 24 such that one traveling acoustic wave,
A.sub.4, is propagated in the form of a shear wave with a
predetermined acoustic velocity and a second wave A.sub.3 is
launched at a delayed time but in the form of a longitudinal wave
having a slightly greater velocity as can be seen by the vector
angles in FIG. 4 so that during the transit of the crystal 20, the
second wave A.sub.3 overtakes the first wave A.sub.4 to provide the
translation function requisite for either convolution or
correlation. More particularly, the transducer 22 for the acoustic
shear wave A.sub.4 can, in a conventional fashion, take the form of
a crystal disposed between two electrodes across which the one
signal is applied to one end of the piezoelectric crystal, this
transducer crystal being oriented so that the requisite shear wave
A.sub.4 is generated. In turn, the second transducer 24 can be
stacked adjacent the first transducer between two electrodes to
which the second signal can be applied, this transducer crystal
being oriented so as to generate the desired longitudinal wave
A.sub.3 at the higher velocity.
With additional reference to FIG. 4, the multiplication function
resultant from the nonlinear interaction of the parametrically
coupled waves A.sub.3 and A.sub.4 will be readily understood. The
acoustic wave A.sub.4 first launched in time in the form of an
acoustic shear wave has a frequency, .omega..sub.4, and a
propagation vector, k.sub.4, while the second acoustic wave A.sub.3
launched subsequently in time at a greater velocity in the form of
a longitudinal acoustic wave has a frequency .omega..sub.3 and a
propagation vector k.sub.3. When the second launched wave A.sub.3
overtakes the first launched wave A.sub.4, the conditions requisite
for parametric coupling are established so that (k.sub.3 - k.sub.4)
is equated to zero so that the resultant electric polarization, D,
has no spatial variation (zero velocity) and, in turn, is at the
difference frequency (.omega..sub.3 - .omega..sub.4), to which a
resonant cavity 26 or other power extracting mechanism can be
coupled to provide the final step of integration. In this instance,
since no time reversal occurs, the output signal represents the
correlation function of the strain amplitudes S.sub.3 and S.sub.4
of the acoustic waves. 0 In both the FIG. 1 convolution structure
and the FIG. 3 correlation unit, the propagation vector k.sub.p of
the output wave was zero, thus enabling coupling to a resonant
cavity 16 or 26. However, it is to be expressly understood that the
conditions for parametric coupling, .omega..sub.1 = .omega..sub.2 =
.omega..sub.p, and k.sub.1 + k.sub.2 = k.sub.p in no wise require
that k.sub.p = o. If k.sub.p .noteq. o, spatial variation of the
output signal output does occur, but only a change in the output
coupling mechanism is requisite to match the finite velocity of
such signal. For example, if as shown in FIG. 5 and explained
through the frequency propagation diagram of FIG. 6, one signal
S.sub.5 is introduced at one end of a piezoelectric crystal 11 at
frequency .omega..sub.5 and propagation vector, k.sub.5, and a
second signal S.sub.6 at frequency .omega..sub.6, and propagation
vector, -k.sub.6, is introduced to the opposite end of the crystal
11, a convolution of the signals S.sub.5 and S.sub.6 will provide
an output, D, whose output frequency is the sum of .omega..sub.5
and .omega..sub.6 and whose propagation vector is k.sub.5 -
k.sub.6. If, as shown in the FIG. 6 diagram, k.sub.5 and k.sub.6
have different values, k.sub.5 - k.sub.6 .noteq. o, the output
polarization, D, will shift along the length of the crystal 11 at a
predetermined velocity. To enable output coupling then, instead of
a resonant cavity, a folded strip line 13 is formed on the crystal
11 with dimensions such that it is matched to the velocity of
D.
The foregoing discussions of the method of signal processing have
been limited to the convolution of correlation of two input signals
to provide a third output signal but it is obvious that the
inventive concept is not so limited. If for example, the amplitude
modulation pattern of a complex wave of known frequency is to be
analyzed, such signal S.sub.7 can be introduced at one end of a
crystal 15 as shown in FIG. 7 and a plurality of short pulses as
indicated at S.sub.8, S.sub.9, S.sub.10, S.sub.11 at the same
frequency can be introduced at the opposite end of the crystal at
predetermined intervals corresponding to the spacing between output
electrodes 17 along the crystal. When parametric coupling occurs
with the pulses S.sub.8, S.sub.9, S.sub.10, S.sub.11 adjacent the
respective electrodes 17, as shown in FIG. 7, the individual output
signals will constitute the products of the individual pulse signal
amplitudes and the amplitude of the analyzed signal at each
particular position, in accordance with the general convolution
principles discussed hereinabove, thus providing a sampled analysis
of the waveform of the signal 7.
By way of specific example, the convolution operation discussed in
general terms hereinabove with respect to FIGS. 1 and 2, has been
carried out at microwave frequencies with the structure shown in
FIG. 8. A piezoelectric crystal 14 having the precise configuration
shown in FIG. 1 and an overall length of approximately 3.5 cm. is
housed within walls of suitable conducting material defining the
microwave cavity 16, one wall of the cavity being provided with an
adjustment screw 28 to enable fine tuning. To opposite ends of the
piezoelectric crystal, the input signals which constitute microwave
signals each having a frequency of 1,440MHz and modulated by a
rectangular pulse are delivered through coaxial cables 23, 25 to
opposite ends of the crystal 14, the center conductors of the
coaxial cables being disposed against the opposite ends of the
crystal at the positions of the transducers 10, 12 illustrated in
FIG. 1 and the outer cable conductors being, in turn, connected to
the conducting cavity defining walls, thus to generate electric
fields which launch acoustic shear waves of identical frequency but
precisely opposite phase propagation vectors in the manner
explained in connection with FIGS. 1 and 2. In accordance with such
explanation, an electric polarization D is generated in the crystal
14 at a frequency of 2,880 MHz and is delivered through a suitable
coupling aperture to an output wave guide 18. Specifically, the
acoustic power injected was approximately 2 watts/cm..sup.2 which
corresponds to a strain amplitude of 2 .times. 10.sup..sup.-6 (the
Q of this particular resonator is 300 and the value of .beta. is
0.2), and at the peak output power level of -70 dBm the
signal-to-noise ratio was 20 dB which value is in good agreement
with the theoretical considerations of the convolution operation
discussed hereinabove with respect to FIGS. 1 and 2.
Further in accord with such considerations, it will be intuitively
obvious from the discussion of the requisite conditions of
parametric interaction that any variations in frequency of the
input signals will cause a sharp reduction in the output signal as
represented by the electrical polarization, D. More particularly,
with reference to FIG. 9 wherein the output amplitude, D, is
plotted against the variable frequency f.sub.2 of one signal in
relationship to another signal of designated frequency, f.sub.1 the
output response curve having the general form of the function of
sin x/x where, x = .pi.[(f.sub.1 /v.sub.1)-(f.sub.2 /v.sub.2)] L
and L is the crystal length, includes but one major central
response lobe whose width is indicative of the narrow bandpass of
the structure. For example, in the case of the crystal described in
connection with FIG. 8 whose overall length, L, was 3.5 cm., the
operating passband at the signal frequency of 1,440 MHz is no more
than approximately 100KHz. Thus it is apparent that this structure
can be used as an excellent bandpass filter and that such passband
can readily be narrowed by the simple process of lengthening the
crystal. Furthermore, it is apparent that an electronically tunable
filter is provided, it being merely necessary to vary f, to tune
the filter.
In addition, since the convolution and correlation functions merely
constitute time reversed operations of one another, as explained
hereinabove, a structure such as shown in FIG. 8 can be readily
converted from the performance of the convolution operation to
performance of a correlation operation by the simple time reversal
of one of the input signals. This statement can be more readily
explained by way of specific example in the form of a
pulse-compression radar system illustrated in block diagram in in
FIG. 10. As shown, a pulse generator 30 is arranged to supply a
frequency-swept pulse P to a frequency modulator 32 that modulates
the carrier wave from a radar transmitter 34 so that a
frequency-swept pulse is delivered to the radiating antenna 36, the
frequency of the pulse P increasing with time. The reflected signal
received by a receiving antenna 38 from the object to be detected
is delivered to a mixer 40 in a conventional fashion wherein the
return signal is mixed with the output of a local oscillator 42 for
delivery to a conventional intermediate frequency amplifier 44
whereupon the reflected signal is delivered in the form of a pulse
P.sub.1 whose frequency increases with time to one end of a
piezoelectric crystal 46 which may be, for example, precisely of
the type shown in FIG. 8 and whose theory of operation was
described in connection with FIGS. 1 and 2.
A frequency-swept reference pulse P.sub.2 is delivered to the
opposite end of the crystal 46 and in order to enable the
correlation operation to proceed, the time reversal necessary to
enable such operation is achieved by passing the same pulse P from
the pulse generator 30 to a pulse inverter 48 so that the reference
pulse P.sub.2 delivered to the opposite end of the crystal 46 has a
frequency which decreases with time, thus constituting a mirror
frequency image of the shape of the reflected pulse P.sub.1.
However, since the reflected singal pulse P.sub.1 is delivered to
the right end of the crystal 46 and the reference pulse P.sub.2 is
delivered to the left end and so are propagated in opposite
directions, the corresponding acoustic signal wave forms S.sub.t,
S.sub.r are identical, thus enabling correlation when phase
coherence of the two occurs. To provide time correspondence between
the two signals S.sub.t, S.sub.r propagated through the crystal 46,
the reference pulse P.sub.2 may be passed through a conventional
variable delay line 50 so that both the reference signal and the
return reflected signal will be propagated through the crystal
simultaneously to provide the phase matching conditions of
parametric coupling. The crystal output in the form of the electric
polarization D is delivered through a suitable wave guide for
detection and ultimate display, for example, on a scope for visual
indication of the target position in a conventional fashion forming
no part of the present invention.
Several points should be observed in operation of the described
radar system utilizing the principles of the present invention. In
the first place, the duration of the transmitted pulse can be
relatively long since the acoustic velocity through the crystal 46
is relatively slow (v = 3 .times. 10.sup.5 cm./sec.), thus allowing
processing of a signal 10 microseconds duration in a crystal but a
few centimeters in length. The longer pulse duration, in turn,
allows more energy to be transmitted in each pulse thus ultimately
improving object detection, without the requirement that the
individual pulse amplitude be at an excessive level. In order to
obtain good range resolution, however, short pulses are requisite
and various techniques such as described in U.S. Pat. No. 2,624,876
have been utilized to provide "pulse compression."
In the present case, an output signal appears in the crystal only
when phase coherence of the transmitted and reflected signals,
S.sub.t and S.sub.r, exists and because of the short time duration
of such coherence, the output signal resultant from signals S.sub.t
and S.sub.r having a pulse duration of 10 microseconds constitutes
a narrow output spike approximating 10 nanoseconds, thus
representing a pulse compression factor of approximately 1,000.
Additionally, it will be obvious that in view of the explained
operation of the convolution mechanism that the output in the form
of the electric polarization D is proportional to the product of
the strain amplitudes of the signals propagated through the crystal
that, in effect, a substantial amplification is derived, thus
allowing the entire system to operate at relatively lower power
levels to obtain the same amplitude of output signals.
Finally, because of the rapid dimunition in output as represented
by the electric polarization D with variance in the coherence of
the reference and reflected signals, a very high signal-to-noise
ratio, as mentioned herein-above, is achieved so that ambiguity in
target signals is removed and indirectly, the effects of
countermeansure "jamming" techniques are rendered less effective in
obscuring a target detected by this radar installation.
In view of the mentioned fact that the acoustic waves travel at a
relatively slow velocity of approximately 3 .times. 10.sup.5
cm./sec., a piezoelectric crystal of reasonable dimensions can be
utilized to enable the comparison of a number of known signal
patterns having individual waveform characteristics with incoming
unknown signals thus to enable the realization of a simple yet
practical method of pattern recognition. By way of example, there
exist 26 letters in the English alphabet and each of these can be
presented in the form of distinct electrical signals which can be
sequentially delivered as reference signals to a piezoelectric
crystal whose length is no greater than five inches. This follows
from the fact that distinctive electrical signals can have a time
duration of no more than 1 microsecond and at an average acoustic
velocity of 3 .times. 10.sup..sup.-5 cm./sec., a sequence of 26
sequential signals can exist simultaneously within a crystal of
this length. With reference to FIG. 11, these reference signals can
be suitably shaped and continuously recycled from a reference unit
70 so as to constitute a sequential reference input to one end of a
piezoelectric crystal as indicated at 72 in FIG. 11 for comparison,
in accordance with the basic theory discussed hereinabove, with an
unknown input signal which can be derived from an optical scanner
74 whose output resultant from scanning of a particular visual
letter display is converted, for example, by a converter 76 in the
manner described in U. S. Pat. No. 3,453,494 to provide distinct
electrical pulses which can, in turn, be inverted by an inverter 78
in the fashion discussed hereinabove in connection with the
description of the radar system shown in FIG. 10 thus to provide
the time reversal requisite for correlation. The inverted signal is
delivered to the opposite end of the crystal 72 and the output
resultant from the electrical polarization, D, existent when
coherence between the input signal and one of the 26 reference
signals exists can be suitably displayed. Since the reference
singals are delivered in time sequence, the precise timing of the
output signal will provide immediate identification of the unknown
"letter" of the alphabet. The output data is accordingly precise
and unambiguous since the mechanism provides a high signal-to-noise
ratio, as discussed hereinabove.
While but a few applications of the convolution and correlation
signal processing method have been described, immediate application
to many other correlation techniques will be apparent and can be
carried out with relatively simple and inexpensive units.
Furthermore, the acoustic waves can take varied forms, as mentioned
hereinabove, which are convenient for each particular application.
Accordingly, the foregoing explanation of the invention and its
application to several specific utilizations is to be considered as
purely exemplary and not in a limiting sense and the actual scope
of the invention is to be indicated only by reference to the
appended claims.
* * * * *