GB1565797A - Creation of Inventions and Patents with Artificial Intelligence

GB1565797A – Signal processing circuits
– Google Patents

GB1565797A – Signal processing circuits
– Google Patents
Signal processing circuits

Download PDF
Info

Publication number
GB1565797A

GB1565797A
GB28657/72A
GB2865772A
GB1565797A
GB 1565797 A
GB1565797 A
GB 1565797A
GB 28657/72 A
GB28657/72 A
GB 28657/72A
GB 2865772 A
GB2865772 A
GB 2865772A
GB 1565797 A
GB1565797 A
GB 1565797A
Authority
GB
United Kingdom
Prior art keywords
signal
noise
input
inputs
signals
Prior art date
1971-06-18
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)

Expired

Application number
GB28657/72A
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)

CA MINISTER NAT DEFENCE

Minister of National Defence of Canada

Original Assignee
CA MINISTER NAT DEFENCE
Minister of National Defence of Canada
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
1971-06-18
Filing date
1972-06-19
Publication date
1980-04-23

1972-06-19
Application filed by CA MINISTER NAT DEFENCE, Minister of National Defence of Canada
filed
Critical
CA MINISTER NAT DEFENCE

1980-04-23
Publication of GB1565797A
publication
Critical
patent/GB1565797A/en

Status
Expired
legal-status
Critical
Current

Links

Espacenet

Global Dossier

Discuss

Classifications

G—PHYSICS

G01—MEASURING; TESTING

G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES

G01S3/00—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received

G01S3/80—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using ultrasonic, sonic or infrasonic waves

G01S3/86—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using ultrasonic, sonic or infrasonic waves with means for eliminating undesired waves, e.g. disturbing noises

G—PHYSICS

G01—MEASURING; TESTING

G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS

G01V1/00—Seismology; Seismic or acoustic prospecting or detecting

G01V1/001—Acoustic presence detection

Description

(54) IMROVEMENTS IN OR RELATING TO SIGNAL
PROCESSING CIRCUITS
(71) I, MINISTER OF NATIONAL DEFENCE OF HER MAJESTY’S CANADIAN
GOVERNMENT, Ottawa, Canada, do hereby declare the invention, for which I pray that a patent may be granted to me, and the method by which it is to be performed, to be particularly described in and by the following statement:- This invention relates to signal processing circuits for the detection of relatively weak signals against a background of noise, and typical uses are in the acoustic detection of aircraft, submarines, and land vehicles.
Two problems are involved in target detection, firstly the actual detection of the presence of a target, and secondly the location or direction-finding of that target. These two requirements are contradictory, since a narrow beam or sector of maximum sensitivity, which is desirable for direction-finding, leads to lack of sensitivitv in neighbouring directions. Thus an acoustic detection system using a narrow beam requires constant scanning by that beam, at a suitably low speed for detection to take place, if detection (as distinct from direction-finding) is important.
The present invention is directed to detection, rather than direction-finding, of a target.
I have already devised a direction-finding system utilizing two microphones and a phase-difference detector which receives the outputs from the two microphones and combines those two outputs in a particular manner in a discriminator which enables the system to be used as a highly-directional device for direction-finding, and a greater number of pairs of microphones can be combined in a similar manner. Such an arrangement provides a compact microphone system for a given narrowness of the beam produced.
It is possible to provide a beam of the same narrowness merely by using a sufficiently large array of microphones with their outputs combined in an additive manner, and for that given beam, the ability to detect a weak signal against background noise is better with the additive arrangement than with the more compact arrangement using discriminators.
An object of the present invention is the provision of a signal processing circuit for improved detection of a selected signal against a background of noise for any given number of signal-receiving transducers.
According to the present invention, a signal processing circuit for use in the detection of a selected signal against background noise comprises separate inputs which comprise at least first, second and third inputs which conduct respectively the selected signal plus a first noise signal; the selected signal plus a second noise signal; and the selected signal plus a third noise signal (the three noise signals being uncorrelated): first combining means arranged to produce from the first and second inputs a fourth signal from which the selected signal has been eliminated and which comprises the difference between the first and second noise signals; second combining means arranged to produce from the first and third inputs a fifth signal from which the selected signal has been eliminated and which comprises the difference between the first and third noise signals: discriminator means arranged to compare the fourth and fifth signals. and when the signals are of the same polarity to select the signal which has an instantaneous valve closer to zero, and to present that signal as an instantaneous intermediate output; and when the two signals are of opposite polarity to present zero as the said instantaneous intermediate output; and third combining means by which a selected one of the said inputs is combined with the said intermediate output; whereby the signal-to-noise ratio in the output of the third combining means is higher than the signal-to-noise ratio in any of the said inputs.
According to a further feature of the invention the signal processing circuit comprises more than three separate inputs each of which conducts the selected signal plus a noise signal, the noise signals of the various inputs being uncorrelated; a plurality of combining means arranged severally to serve different pairs of said inputs and to produce respectively noise difference signals from which the said selected signal is absent; means for combining different pairs of these noise difference signals in discriminators each of which includes; (a) signal comparison means arranged to compare the pair of signals, to select the noise signal which has an instantaneous value closer to zero, and to present that signal as the instantaneous output;
(b) when the signals are of opposite polarity to present zero as the instantaneous output; and further combining means by which a selected one of the said inputs is combined with an intermediate output derived from the outputs of the discriminators, whereby the signal-to-noise ratio in the output of the said further combining means is higher than the signal-to-noise ratio in any of the said separate inputs.
The invention will now be described, by way of example, with reference to the accompanying largely diagrammatic drawings, in which: Figure I is a block diagram of a three-input signal processing circuit utilized in an aural target detection system; Figures 2, 3 and 4 are graphical representations of two input signals and an output signal, for different relative phases of the former, of a discriminator shown as forming part of the circuit shown in Figure 1;
Figure 5 is a block diagram of one embodiment of the discriminator used in the circuit of
Figure l; Figure 6 is a more detailed circuit diagram of the discriminator of Figure 5;
Figure 7 is a graphical representation of possible noise components in the output from the circuit of Figure 1 before filtering.
Figure 8 is a circuit diagram of a second embodiment of the invention in which four input signals are utilized:
Figure 9 is a graphical representation illustrating the beam-forming properties of 3-input, 4-input and 6-input circuits according to the present invention;
Figure 10 is a block diagram of a multistage filter utilizing a signal processing circuit shown in Figure 1;
Figure 11 is a graphical representation of observed gains of the three input signal processing circuit used in the filter mode; Figure 12 is a graphical indication of accumulated gain for a multistage filter using the three input signal processing circuit in the filter mode; and Figure 13 is a block diagram of a multistage filter utilizing a four-input signal processing circuit such as that shown in Figure 8.
Referring first to Figure 1, three microphones 1, 3 and 5 are equally spaced along a straight line 7, and each of these is sensitive to sound approaching the microphone from the left in Figure 1.
Considering the sound received by the microphones from a single sound source, typically two or three miles from the microphones. each microphone will receive a common signal S and a noise signal which will be different for each microphone, and will be designated by X,
U and Y respectively for the three microphones. Thus the three outputs from the microphones will be referred to as follows:
From microphone 1 Output is S + X
From microphone 3 Output is S + U
From microphone 5 Output is S + Y
It is important to realize that in the arrangement described the three noise signals X, U and
Y, although possessing identical statistical characteristics. should be uncorrelated.
In Figure 1, the output S + U from microphone 3 is passed through an inverter INV. I the output of which is thus -S-U, and this output is applied as one input to an adder ADD 1. The other input to adder ADD. 1 is the output from microphone 1. i.e. the signal S + X.
Thus the output from adder ADD. 1 is X-U Similarly, the output S + Y from microphone 5 is applied to an inverter INV. 2, the output of which is thus -S-Y, and this output is applied as one input to an adder ADD. 2.
The other input to adder ADD. 2 is the output from microphone 1, i.e. the signal S + X.
Thus the output from adder ADD. 2 is X-Y.
The outputs from the two adders are applied respectively as first and second inputs to a discriminator D.1.
The discriminator used is a circuit which obeys the following two rules:
(i) when input signals 1 and 2 are both of the same polarity, the circuit selects the input closer to zero and presents it at the output;
(ii) when the input signals are of opposite polarity, the circuit presents zero at the output.
Figures 2, 3 and 4 show this effect pictorially: In Figure 2, the two inputs 1 and 2 are equal and in phase: since they are never of opposite polarity, rule (ii) does not apply, and the output is identically similar to each input.
In Figure 3, the two inputs 1 and 2 are equal but input 2 lags input 1 by 90 degrees: during period 0 to 90 degrees of input 1, since the two inputs are of opposite polarity, the output is zero; from 90 degrees to about 135 degrees, input 1 is larger than input 2, so that the output follows input 2; from about 135 degrees to 180 degrees, input 1 is smaller than input 2, and therefore the output follows input 1: from 180 degrees to 270 degrees, the two inputs are of opposite polarity, and therefore the output is zero; from 270 degrees to about 315 degrees, input 1 is larger than input 2, and the output follows input 2; from about 315 degrees to 360 degrees, input 1 is less than input 2, so that the output follows input 1.
In Figure 4, where input 2 lags input 1 by 180 degrees, at all time (except the crossover points) input 1 and input 2 are of opposite polarity, so that the output remains at zero.
It will be seen that the change in the output from the discriminator is from maximum to zero over a phase difference in the inputs of zero to 180 degrees.
Figure 5 shows in block diagram form one form of the discriminator D1 of Figure 1, and
Figure 6 shows the complete circuit diagram for this embodiment. The blocks shown in
Figure 5 are all well known and commercially available logic circuit components. Thus each of the comparators C1, C2 and C3 is a form of analog to digital converter. It receives two analog inputs, and compares them with one another. if the algebraic sum of the two analog inputs is positive, then the logic output is HIGH (i.e. 1). If the algebraic sum of the two analog inputs is zero or negative, then the logic output is LOW (0). Thus the output is binary in nature, and will always be either “1” or ‘0″.Each of the two modulus circuits
MOD. 1 and MOD. 2 is in effect a full wave rectifier circuit without any smoothing, so that the shape of both positive and negative parts of the output waveform are similar to those of the input waveform, but the negative parts of the waveform are rendered as positive parts.
Each GATE GT. 1 and GT. 2 acts as a high speed relay which controls the passage of an analogue signal according to the binary logic level applied to its driving input. As usual, each gate is a transistor switching circuit, rather than an electromagnetic relay. The AND gates AND. 1 and AND. 2 each provide a “0” output unless both of their inputs carry a “1” input. The “exclusive OR” circuit EX. OR has two inputs, and provides a “0” output except when both of its inputs have the same input signal, i.e. when both are “1” or when both are “0”. When the inputs are the same, the output is a “1”. The inverter INV. 3 provides as its output a binary signal opposite to its input binary signal, i.e. a “1” input signal produces a “0” output signal and a “0” input signal produces a “1” output signal.The amplifier AMP. 1 is an operationa! amplifier used as a summing amplifier.
In Figure 5, several points on the diagram are denoted by the various letters A through F, and the following “Truth Table” indicates the signals at these various points for varying inputs 1 and 2.
Case Input Amplitudes Logic Signals Output
Polarity (inputs) A B C D E F I/Ps I/P2 1 + + |I/P1| > |I/P2| 1 1 1 1 0 1 I/P2 2 + + |I/P2| > |I/P1| 1 1 1 0 1 O I/PI 3 + – PII > IIIP2I 1 0 0 1 0 0 0 4 + – |I/P2| > |I/P1| 1 0 0 0 0 0 0 5 – + |I/PI| > |UP2l 0 1 0 1 0 0 0 6 – + |I/P2| > |I/P1| 0 1 0 0 0 0 0 7 – – |I/P1| > |I/P2| 0 0 1 1 0 1 I/PZ 8 – – |I/P2| > |I/P1| 0 0 1 0 1 O I/P,
9 + + |I/P2|=|I/P1| 1 1 1 1 0 1 10 – – |I/P2|=|I/P1| 0 0 1 1 0 1 In practice, the input 1 may not have exactly the same amplitude as the input 2, but the modification of the output by practical differences in these inputs is small, and for simplicity it will be assumed that the amplitudes are the same.
In operation, comparator 1 will provide a digital signal at A which will be “1” while input 1 is positive, and will otherwise be “0”. Comparator C2 will provide a similar output at B depending upon the polarity of input 2. The exclusive OR circuit EX. OR will then provide at C a binary signal which will be “1” when both inputs 1 and 2 are positive, or negative, and a signal “0” when the inputs are of opposite sign. As long as the signal at C is “0”, neither of the AND gates AND. 1 and AND. 2 will be enabled, so that both gate GT. 1 and GT. 2 will be non-conducting, so that the output applied to amplifier AMP. 1 will be zero and the output from the whole circuit will be zero.
When the signal at C is “1”, then for each of the AND gates AND. I and AND. 2 one input is provided.
The comparator C3 receives at all times a full-wave rectified but unsmoothed version of the input 1 as a first input, and a full-wave rectified but unsmoothed version of the input 2 as a second input, If the instantaneous numerical value of input 1 is greater than the instantaneous numerical value of input 2, then the output of comparator C3, i.e. at the point D, is “1”. On the other hand, if the opposite is true, the signal at point D is “0”. The signal at D is applied directly as the second input to AND gate AND.2, while an inversion of the signal is applied directly as the signal to AND gate AND. 1. Thus if signal input 1 is greater than signal input 2 (and they are of the same polarity) at point F appears a signal “1” so that gate GT. 2 is enabled and the analog signal on input 2 is applied to the amplifier
AMP. 1.On the other hand, if signal input 2 is greater than input signal 1 (and they are of the same polarity), at point E appears a “l” signal and gate GT. 1 is enabled so that the analog signal input 1 is applied to the amplifier AMP. 1.
It will be seen that the smaller of the two input signals is applied to the amplifier AMP. 1 as long as the signals are of the same polarity. In the case of Figure 2, when the two signals are equal and of the same polarity, the output is input 2. This result is necessary to avoid the occurrence of zero output when the sound locator is precisely directed at the target, and is achieved by setting comparator C3 to give a “1” output when its input 1 is equal to input 2.
Referring now to Figure 6. it will be seen that the Exclusive Or circuit EX. OR includes three AND gates AND. 3, AND. 4 and AND. 5 and five inverters INV. 4, INV. 5, INV. 6,
INV. 7 and INV. 8. The modulus circuit MOD. 1 makes use of an operational amplifier
AMP. 2 used as a differential amplifier to the non-inverting input of which is applied input 1, the inverting input being supplied with the same signal which however is inverted in an amplifier AMP. 3 and then gated by gate GT. 3 which is driven by the output on lead A.
Modulus circuit MOD. 2 includes differential operational amplifier AMP. 4, inverting amplifier AMP. 5, and gate GT. 4.
Returning now to Figure 1, it will be appreciated that only “noise” is processed through the discriminator D.1. The output from the discriminator is passed through a further inverter INV. 9 the output of which is applied as one of the two inputs to an adder ADD.3.
The output from microphone 1 is applied directly to the second input of adder ADD. 3. The output from adder ADD. 3 is applied to a band-pass filter F. 1 the pass-band of which is centered on the frequency of desired signalS, and the output from that filter is the useful output which is monitored to ascertain the presence of signal S.
Certain assumptions are made regarding the operation of the signal processing circuit shown in Figure 1, and naturally circuit components and values are selected which will make these assumptions tenable. Thus the signal component S is assumed to have the same phase and amplitude in all three inputs. This will be true as long as the wavefront of the incoming sound wave is substantially parallel to the line 7, and as long as the three microphones have equal responses. Small deviations from parallelism will have little effect since the phase difference varies inversely as the wavelength of the sound signal S; and as regards amplitudes, microphones can be suitably matched.Further, the noises are assumed to have the same RMS amplitude at each input, but to be mutually incoherent, i.e. on the assumption that the “noise” comes from a different bearing than that of the target, “noise” signals X, U and Y will have equal RMS amplitudes but will have different phases. Since the signal is processed linearly in the adder ADD. 3 and in the filter F. 1 and since the noises are fixed in RMS amplitude, the output signal-to-noise gain will not vary with input signal-to-noise ratio. This “noise” is then inverted and added to signal S + X in adder
ADD. 3. The resultant complex wave form is then filtered in Filter F1. Figure 7 shows the three original “noise” inputs at A, B and C in Figure 1 and the resultant “noise” component of the output at E in Figure 1, either U, Y or X.The effect of “adding” an inversion of the selected “noise” signal is to minimise the noise signal passing to filter F. 1.
One of the problems in the processing of signals is that the amplitude statistics at the output tend to be decidedly non-gaussian, but tests show that with the signal processing circuit of Figure 1 the output amplitude statistics for noise are very close to gaussian so that measured RMS gains in signal-to-noise ratio are not degraded in the detection process. The measured RMS gains depend on input bandwidth, the number of inputs, and the nature of the processed signal, as the table below indicates.
Octave 1% No. of
SIGNAL Band Bandwidth Inputs
Sinewave 4.6 db 6.2 db 3
Noise (at stated bandwidth) 4.1 5.7 3
Sinewave 5.5 7.2 4
Noise (at stated bandwidth) 5.0 6.7 4
Sinewave 7.3 8.6 6
Noise (at stated bandwidth) 6.8 8.1 6
As indicated by the above table, the signal processing circuit of Figure 1 can be modified to accept 4 inputs or 6 inputs from a corresponding number of microphones. Figure 8 is a diagrammatic representation of a signal processing circuit having four inputs 101, 103, 105 and 107 derived respectively from four microphones arranged on a common line in the manner indicated in Figure 1 for the three microphone arrangement. The input 101 is applied to first and second operational amplifiers AMP. 101 and AMP. 103. The input 103 is applied to first, second and third operational amplifiers
AMP. 105, AMP. 107 and AMP. 108.The input 105 is applied to a single operational amplifier AMP 109. The input 107 is applied to a single operation amplifier AMP. 111. In
Figure 8, each amplifier shown has associated with it. for these inputs, a series input resistor indicated by RXXX.S (where XXX is the number of the amplifier) and a feedback resistor
RXXX.F where XXX is the number of the amplifier).
Amplifier AMP. 105 serves merely as no-loss inverter and its output is applied through a series resistor R1O1X to amplifier AMP. 101. This amplifier AMP. 101 serves as an adder and receives signal S + P (where S is the desired signal and P is the noise) from input 101 and the inverted signal – S – Q (where Q is the noise on input 103) and its output P- Q is applied to one input of a discriminator D.12.
Amplifier AMP. 109 serves merely as a no-loss inverter and its output is applied through a first series resistor R103X to amplifier AMP. 1U3 and through a second series resistor
R107X to amplifier AMP. 107. The input 105 consists of the signal S and noise “R”, and the output from amplifier AMP. 109 is thus – S – R. Amplifier AMP. 103 acts as an adder, and its output is thus P-R. This is applied to discriminator D12. Amplifier AMP. 107 also acts as an adder, and its output is thus Q-R, and this is applied to a discriminator D23.
Amplifier AMP. 111 serves as a no-loss inverter of the signal S + T (T is the noise) from input 107 and its output – S – T is applied through series resistor R108X to amplifier AMP.
108. Amplifier AMP. 108 acts as an adder, and its output is thus Q – T, and this is applied to the discriminator D23.
Each of the discriminators D i2 and D,3 operates in the manner described above in connection with discriminator D1 of Figure 1. The output from discriminator D 12 is applied through a series resistor R113X to an operational amplifier AMP. 113, and the input 101 is also applied to this amplifier through a series resistor R1 13S. The output from discriminator
D23 is applied through a series resistor R115X to an operational amplifier AMP. 115, and the input 103 is applied through a series resistor R115S to this amplifier.
Each of the amplifiers AMP. 113 and AMP. 115 acts as an adder, and their outputs are applied respectively through series resistors R1 17S and R117X to an operational amplifier
AMP. 117. The output from that amplifier is passed through a band-pass filter F. 2 to provide the useful output from the circuit.
In the circuit of Figure 8, all the operational amplifiers are integrated circuits sold under the type number AMELCO 809 CE, and all the resistors shown have a resistance of 100,000 ohms.
The manner of operation of the circuit of Figure 8 will be seen to be similar to that of the circuit of Figure 1. First, pairs of inputs are combined to eliminate the target signal S, and then the “noise”difference signals are applied in pairs to discriminators.
However, although the three input signal processing circuit of Figure 1 is effective for target detection, the four-input signal processing circuit of Figure 8 is required for beam forming, since the phase response curve for the three-input signal processing circuit of
Figure 1 is at a maximum at 180 degrees relative phase difference between the three inputs.
Referring to Figure 1, when the line 7 is mis-oriented so that the incoming wavefront is not parallel to the line, then there will be a phase-difference between the signals arriving from a target at the various microphones. The phase difference will depend upon the angle by which line 7 is mis-oriented, and the distance between the microphones in terms of wavelengths of the sound from the target. As far as the circuits of Figures 1 and 8 are concerned, it is the phase difference between the various inputs which is critical, and the curves in Figure 9 show how the output, expressed as voltages for 3-input signal processing circuits according to Figure 1. for 4-input signal processing circuits according to Figure 8, and a 6-input signal processing circuit using the same method of detection, vary with the phase difference between the inputs.It will be clear to those skilled in the art that from the curves of Figure 9 it is possible to draw beam patterns for the three signal processing circuits concerned. It can be shown that all three signal processing circuits produce wider beams than would be produced by a simple additive arrangement of the same number of inputs.
On the other hand, the side lobes produced are smaller than with such additive arrays. Such an arrangement provides a wide beam useful for the initial detection of a target.
It has been found that the signal processing circuits described are particularly efficient when dealing with sinewave-like signals processed in a narrow band, when the gain obtained is compared with that given by additive processing (on an equal-beam-width basis). In a multi-microphone arrangement it is found that a slightly larger number of microphones is needed to produce a given beam width. compared with an additive system.
In an experimental array with a signal processing circuit according to Figure 8, band widths of 41 degrees with side lobes 15 and 19 dB down have been achieved.
The three-input signal processing circuit of Figure 1 and the 4- and 6-input modifications described in connection with Figure 8 provide arrangements in which the signal-to-noise gain is independent of input-signal-to-noise ratio the output amplitude statistics are essentially gaussian; and the signal to noise gain is superior to that of an adder for equal beam widths for sinewave signals processed in narrow band where the noise background is random or flow noise.
Of course, distant in-beam noise is processed as if it is a part of the useful signal S, and for that type of noise the performance is very much like that of a simple additive array. As in an additive array, in-beam noise which differs considerably in frequency from the target frequency can be reduced considerably by use of the various band-pass filters.
One useful feature of apparatus using a signal processing circuit as described is its ability to operate properly in a windy environment, in which there is much wind-induced noise to contend with.
Non-acoustic system applications can include radio monitoring, and low frequency radar.
It has been found that, in the absence of any bandpass filters, the noise output bandwidth is appreciably greater than the noise input bandwidth. It follows that if a final bandwidth filter has the same bandwidth as an input band@@@@@@@@@@@@equencies have been band spread in the non-linear pro@@@@@@@@@@@@@@@@@@@@@@@@@@@ iaes a distinct gain in signal-to-noise ratio.
The signal processing circuit which has been @@@@@@@@@@@@@ica@@on in the filtering of noise from a single input signal
Linear filtering is an extremely common ope@@@@@@@@@@@@@@@@and the basie idea of limiting the bandwidth of a system so as to @@@@@@clude unwanted wide band noise has proven extremely @@@@@@@@@@@@@@
An observed phenomenon in the outp@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@nal processing circuit provides the key @@@@ sup@@@@@@@@@@@@@@@@@@@@@@@@@@@@ bandwidth limited in the normal way@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ is reduced with respect to a singal. Signal here @@@@@@@@@@@@@@@@@@@@@@@@@ bandwidth, that is of long duration. wh@@@@@@@@@@@@@@@@@@@@@@@@@enilally constant. Real signals may have slight frequen@@ and amplitude variations.However, the signal bandwidth is assumed to be small comp@@@@@@@@@@@@@@@ of the noise within the output bandwidth of the filter @@ mg co@@@@@@@@@@@@@@@@@@@@@@@@@@@ill set the ultimate limit to the bandwidth of @@@s p@op@@@@@@@@@@@@@@@@@@@@@@@@in one @@@@@@@ type of filter.
As previously discussed. the ou@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@effecd@@@@@ bandspread the noise whereas the @@@@@@@@@@@@@@@@@@@@@@@@@@@@@spread. it was alse shown that this effect became prop@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@dwidths. The basic mechanism is related to the @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ginares in the non-linear discriminating process itsell@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@pr@@ed to be highly efficient especially in @@@@@@@@@@@@@@@@@@@@@@@@@@@@ined by sign@ @@
noise gain for a given beamwidth in arr@@@@@@@@
These characteristics and the following @@@@@@@@@@@@@@@@@@@@@@@@@@@@@ea@@no@@@ear
processing has a useful application in filt@@@@@@@@@@@@@@@@@@@@@@@@@@@@@is gi@en
by: Rx(t) = N (SinW@T – SinW # T where::
W1 = 2# f1, f1 is the lower frequen@@@@@@ W2 = 2# f2, f2 is the upper frequen@@@@@@ @@@
N = # Rx (O) = # W2 – W1 W2 – W1
T = delay time A solution for this equation show that lo@@@@@@@@@@@@@@@@@@@of one@@@@@@@@@@@@@@@@@
fo, the auto-correlation function is near zero for @@ bandwi@@@@ e@ual to the center treauen@y
That is the auto-correlation @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
that a octave band (i.e BW @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
output of a delay line of length T @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@es
of equal length connected in series @ill pro@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@put
processing circuit.Assuming the propet re@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@the
noises at all three inputs will be @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
Referring now to Figure @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@three-input
circuit in three basically sir@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ to
receive the noise-containing @gnal @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@an o@thodo@
band-pass filter 303 with a p@e@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@[email protected]
output of this filter is applic @o two @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@vide
three inputs as shown to a discimin@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@time T
equal to the period of one cyde at the @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@re
three input circuit of inverters. adde@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
the discriminator circuit 309 is app@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
approximately equal to the bandwidth @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@rcuit. i.e.
approximately 70@ fo. The output@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ne inpu
to stage 2, being upplied to two se@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ing
delay time 2T (i@e cqual to twice the dela @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
to a further discriminator circiut 31@ An @@@@@@@@@@@@@@@@@@@@@@@@@@@of om@@@@@@@@@@@@@
is connected between circuit 317 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@and @@@ each producing a time delay of 4T (i.e. equal to four times the delay time T). This arrangement provides three inputs for discriminator circuit 325, the output of which is connected through an output filter 327 having a bandwidth of only 18% fro to an output terminal 329.
It will be seen that a one octave bandwidth is used at the input with unit T rather than the calculated 100% f0 bandwidth. The reason for this is the phase response characteristic of the three-input device, which peaks at 1800 phase difference. By limiting the noise bandwidth to one octave for this value of T, this peak of the phase response curve cannot occur, but as the above formula shows the auto-correlation function is now greater than zero.
The effective filtering per stage can be calculated from the phase response curve for the three-input signal processing circuit and extensive measurements with a swept sine wave at the input confirm the calculations. However, these apparent bandwidths cannot be used in calculating the signal-to-noise gain of each stage because of the bandspreading effect of the nonlinear processing of noise. It was found that the filter mode arrangement of Figure 10 yields the best result. Great care must be used in measuring the input and output noise spectra, together with the RMS change involved, to determine the correct gain per stage. A measurement of the auto-correlation function of the noise at the output of a linear-nonlinear filter stage checked with the measured noise bandwidth. In addition the amplitude distribution was virtually gaussian.There were no measurable differences between the output noise and an equivalent bandwidth of noise after linear filtering. This allows the calculation of multistage performance of a linear-non-linear filter design.
In linear filtering the correct gain is given by G = 10 log BWVBwoin dB, where signal gain is assumed to be unity.
where Bwi = input noise bandwidth
Bwo = output noise bandwidth
In this linear-non-linear type of filtering the gain per stage was measured directly using the RMS measuring circuits. A value of X was calculated from X = ~ G Log BWi
Bwo using measured values of Bwi, Bwo and G, and this value is plotted on Figure 11 (curve 331) as a function of bandwidth for five different input bandwidths; 70sic, 35%, 30%, 11% and 1% of fO. It will be seen that the value of X increases with decreasing bandwidth, For comparison purposes the standard value of X = 10 for linear filtering is also shown (curve 333) on Figure 11.
Clearly there is a measurable gain using linear-non-linear processing when compared to standard filtering. This gain derives partly from the normal out-of-band rejection based on the phase response curves for the 3-input device. it also derives, in part, from the bandspreading of noise from within the pass band of the filter stage itself to outside of the final pass band which the terminal filter eliminates. This latter part of the gain is very small at wide bandwidths but becomes increasingly important at narrow bandwidths. Measurements of improvements were easy but unimpressive at octave bandwidths. The gains recorded at 1% bandwidths were difficult to measure but were entirely convincing. A special very narrow band pass measurement filter which could be swept across the output spectra had to be constructed.A linear regenerative delay line filter was used whose band pass was held constant at 0.3% fO.
Using the measured values of X at various bandwidths and the calculated number of stages required to achieve narrow bandwidths. the accumulated gain of this filter as a function of bandwidth is plotted on Figure 12. The bottom curve 335 is the result of standard linear filtering over the three decades of bandwidth change shown. Curve I shows the result of linear-non-linear filtering and is supported by measurements to the 1% bandwidth point, and is an extrapolation beyond that point. Curves II and III are the results of starting with initially narrower bandwidths and are largely extrapolations of the measured data. Since it is clear that this type of filtering works best in narrow bands, it is
always possible to heterodyne wide band low frequency energy into relatively narrow band higher frequency energy for more efficient processing.
Standard linear filtering is one dimensional filtering in that the transfer function is related to frequency only. In the linear-non-linear type of filtering just described, the circuit is both phase sensitive and amplitude sensitive. As a result. the transfer function involves both frequency or phase differences across a delay line and amplitude differences across the
same delay line. The mechanism of noise bandspreading from within the output pass band
can and does occur because the noise is not fully correlated across the delay line even within
the output pass band. This is a fundamentallv different situation from the in-beam distant
noise case which arises in spatial processing. Such in-beam noise is almost fully correlated and is processed as signal.For this reason, the gains achievable vis-a-vis the additive processing are limited to sources of noise which are uncorrelated at each microphone.
When filtering no such restriction applies as the degree of noise correlation across the delay line is determined by the auto-correlation function or by the relationship of the noise bandwidth and the value of delay involved.
In Figure 10, the 3-input signal processing circuit is employed. It has the disadvantage of using expensive and, in higher stages, very long delay lines. The terminal filter in each stage has to be carefully set to achieve a proper balance between the apparent gain of a given stage and the correct input bandwidth to the next stage. A 4-input signal processing circuit arrangement such as that shown in Figure 13 using the circuit of Figure 8 requires an additional delay line per stage but would be less critical in its terminal filter requirements and would require fewer stages to achieve a narrow band output. Measurements on the 4-input circuit showed a lower efficiency, i.e. gain in signal-to-noise ratio per unit change in bandwidth, than the 3-input circuit. The values of T must be increased by a factor of 3 rather than 2 per stage.The lower efficiency was more apparent at wide bandwidths than at narrow bandwidths.
A final point concerns the rise time of this type of filtering. For a multistage linear-non-linear filter the rise time is approximately twice as long as the rise time for a standard linear filter of equal bandwidth. This means that for signals of finite length, any post detector time averaging gain will be 1.5 dB less for the linear-non-linear filter than for the standard filter. This loss is relatively small when compared to potential improvement of 10 to 30 dB.
The phase response for the three-input circuit or multiples thereof produces wide main beams when used in an array. Side lobes are 18 dB down which is lower than the corresponding additive processing for four-inputs. The wide beam is useful in search and hence this is called the search mode of operation.
The signal-to-noise gain is independent of input signal-to-noise ratio in the 3-input circuit but depends on the input bandwidth. It is particularly efficient in narrow bands and is superior to the additive array on an equal bandwidth basis. This superiority is restricted to operation against an inter-element noise background which is uncorrelated and does not apply to in-beam distant noise.
When used in the filter mode configuration, the three-input processing circuit bandspreads the noise within the pass band of the filter which noise is subsequently removed. This effect does not occur for narrowband signal which is processed linearly. The result is a high gain, two dimensional filtering process. Measurements indicate large improvements over standard linear filtering. This improvement lies between 10 and 30 dB better than ordinary filtering for a three decade change in bandwidth.
WHAT I CLAIM IS:
1. A signal processing circuit for use in the detection of a selected signal against background noise, comprising separate inputs which comprise at least first, second and third inputs which conduct respectively the selected signal plus a first noise signal; the selected signal plus a second noise signal; and the selected signal plus a third noise signal (the three noise signals being uncorrelated); first combining means arranged to produce from the first and second inputs a fourth signal from which the selected signal has been eliminated and which comprises the difference between the first and second noise signals; second combining means arranged to produce from the first and third inputs a fifth signal from which the selected signal has been eliminated and which comprises the difference between the first and third noise signals; discriminator means arranged to compare the fourth and fifth signals, and when the two signals are of the same polarity to select the signal which has an instantaneous value closer to zero, and to present that signal as an instantaneous intermediate output; and when the two signals are of opposite polarity, to present zero as the said instantaneous intermediate output; and third combining means by which a selected one of the said inputs is combined with the said intermediate output; whereby the signal-to-noise ratio in the output of the third combining means is higher than the signal-to-noise ratio at any of the said inputs.
2. A signal processing circuit as claimed in claim 1, in which the fourth and fifth signals are applied respectively to first and second gates as analog signals, each gate is arranged, when enabled, to pass the said analog signal on as an output from the discriminator and the first and second gates can be selectively enabled to provide the said intermediate output.
3. A signal processing circuit as claimed in claim 2, where first comparator means ascertain whether the fourth and fifths signals are of different polarity, and when that condition exists, disable both said gates.
4. A signal processing circuit as claimed in claim 2 or claim 3. wherein second comparator means ascertain which of the fourth and fifth signals has instantaneously the smaller value, and are arranged to enable the appropriate first or second gate to permit that
**WARNING** end of DESC field may overlap start of CLMS **.

Claims (12)

**WARNING** start of CLMS field may overlap end of DESC **. and is processed as signal. For this reason, the gains achievable vis-a-vis the additive processing are limited to sources of noise which are uncorrelated at each microphone. When filtering no such restriction applies as the degree of noise correlation across the delay line is determined by the auto-correlation function or by the relationship of the noise bandwidth and the value of delay involved. In Figure 10, the 3-input signal processing circuit is employed. It has the disadvantage of using expensive and, in higher stages, very long delay lines. The terminal filter in each stage has to be carefully set to achieve a proper balance between the apparent gain of a given stage and the correct input bandwidth to the next stage. A 4-input signal processing circuit arrangement such as that shown in Figure 13 using the circuit of Figure 8 requires an additional delay line per stage but would be less critical in its terminal filter requirements and would require fewer stages to achieve a narrow band output. Measurements on the 4-input circuit showed a lower efficiency, i.e. gain in signal-to-noise ratio per unit change in bandwidth, than the 3-input circuit. The values of T must be increased by a factor of 3 rather than 2 per stage.The lower efficiency was more apparent at wide bandwidths than at narrow bandwidths. A final point concerns the rise time of this type of filtering. For a multistage linear-non-linear filter the rise time is approximately twice as long as the rise time for a standard linear filter of equal bandwidth. This means that for signals of finite length, any post detector time averaging gain will be 1.5 dB less for the linear-non-linear filter than for the standard filter. This loss is relatively small when compared to potential improvement of 10 to 30 dB. The phase response for the three-input circuit or multiples thereof produces wide main beams when used in an array. Side lobes are 18 dB down which is lower than the corresponding additive processing for four-inputs. The wide beam is useful in search and hence this is called the search mode of operation. The signal-to-noise gain is independent of input signal-to-noise ratio in the 3-input circuit but depends on the input bandwidth. It is particularly efficient in narrow bands and is superior to the additive array on an equal bandwidth basis. This superiority is restricted to operation against an inter-element noise background which is uncorrelated and does not apply to in-beam distant noise. When used in the filter mode configuration, the three-input processing circuit bandspreads the noise within the pass band of the filter which noise is subsequently removed. This effect does not occur for narrowband signal which is processed linearly. The result is a high gain, two dimensional filtering process. Measurements indicate large improvements over standard linear filtering. This improvement lies between 10 and 30 dB better than ordinary filtering for a three decade change in bandwidth. WHAT I CLAIM IS:

1. A signal processing circuit for use in the detection of a selected signal against background noise, comprising separate inputs which comprise at least first, second and third inputs which conduct respectively the selected signal plus a first noise signal; the selected signal plus a second noise signal; and the selected signal plus a third noise signal (the three noise signals being uncorrelated); first combining means arranged to produce from the first and second inputs a fourth signal from which the selected signal has been eliminated and which comprises the difference between the first and second noise signals; second combining means arranged to produce from the first and third inputs a fifth signal from which the selected signal has been eliminated and which comprises the difference between the first and third noise signals; discriminator means arranged to compare the fourth and fifth signals, and when the two signals are of the same polarity to select the signal which has an instantaneous value closer to zero, and to present that signal as an instantaneous intermediate output; and when the two signals are of opposite polarity, to present zero as the said instantaneous intermediate output; and third combining means by which a selected one of the said inputs is combined with the said intermediate output; whereby the signal-to-noise ratio in the output of the third combining means is higher than the signal-to-noise ratio at any of the said inputs.

2. A signal processing circuit as claimed in claim 1, in which the fourth and fifth signals are applied respectively to first and second gates as analog signals, each gate is arranged, when enabled, to pass the said analog signal on as an output from the discriminator and the first and second gates can be selectively enabled to provide the said intermediate output.

3. A signal processing circuit as claimed in claim 2, where first comparator means ascertain whether the fourth and fifths signals are of different polarity, and when that condition exists, disable both said gates.

4. A signal processing circuit as claimed in claim 2 or claim 3. wherein second comparator means ascertain which of the fourth and fifth signals has instantaneously the smaller value, and are arranged to enable the appropriate first or second gate to permit that
signal to pass to an output, unless the two signals are of opposite polarity.

5. A signal processing circuit as claimed in claim 2 or claim 3 wherein second comparator means ascertain whether the fourth and fifth signals are instantaneously equal or whether one is smaller than the other, and are arranged to enable the appropriate first or second gate to permit that smaller signal to pass to the output, unless the two signals are of opposite polarity, and are arranged when the two signals are equal to enable a preselected one of the two gates to connect a preselected input to the said output.

6. A signal processing circuit as claimed in claim 1, and in which the first and second inputs are combined by passing one of these inputs through a sign inverting circuit, and then passing the inverted input and the other input through an adding circuit to produce the desired fourth signal; and in which the first and third inputs are combined by passing one of these inputs through a sign inverting circuit, and then passing the inverted input and the other input through an adding circuit to produce the desired fifth signal.

7. A signal processing circuit as claimed in claim 1, and in which the first and second inputs are combined by passing the second input through a sign inverting circuit, and then passing the inverted input and the first input through an adding circuit to produce the desired fourth signal; and in which the first and third inputs are combined by passing the third input through a sign inverting circuit, and then passing this inverted input and the first input through an adding circuit to produce the desired fifth signal.

8. A signal processing circuit as claimed in claim 1, and in which said intermediate output is inverted and passed through an adding circuit with said selected one of said inputs, whereby the signal-to-noise ratio in the output from the adding circuit is considerably reduced in relation to the signal-to-noise ratio at any input.

9. A signal processing circuit as claimed in claim 1, and in which an output from the third combining means is processed by a band-pass filter to reduce or eliminate noise signals having a frequency appreciably different from a preselected frequency chosen to be representative of the said selected signal.

10. A signal processing circuit for use in the detection of a selected signal against background noise, comprising more than three separate inputs each of which conducts the selected signal plus a noise signal, the noise signals of the various inputs being uncorrelated; a plurality of combining means arranged severally to serve different pairs of said inputs and to produce respectively noise difference signals from which the said selected signal is absent; means for combining different pairs of these noise difference signals in discriminators each of which includes:
(a) signal comparison means arranged to compare the pair of signals, to select the noise signal which has an instantaneous value closer to zero. and to present that signal as the instantaneous output;
b and when the signals are of opposite polarity to present zero as the instantaneous output; and further combining means by which a selected one of the said inputs is combined with an intermediate output derived from the outputs of the discriminators, whereby the signal-to-noise ratio in the output of the said further combining means is higher than the signal-to-noise ratio at any of the said separate inputs.

11. A signal processing circuit according to claim 1, in which the inputs are obtained from microphones and the selected input signal is an acoustic signal coming from the object whose presence the circuit is required to ascertain.

12. A signal processing circuit substantially as hereinbefore described with reference to
Figures 1, 5, 6, 8, 10 or 13 of the drawings filed herewith.

GB28657/72A
1971-06-18
1972-06-19
Signal processing circuits

Expired

GB1565797A
(en)

Applications Claiming Priority (1)

Application Number
Priority Date
Filing Date
Title

CA115,983A

CA1080339A
(en)

1971-06-18
1971-06-18
Signal processing for detection

Publications (1)

Publication Number
Publication Date

GB1565797A
true

GB1565797A
(en)

1980-04-23

Family
ID=4090048
Family Applications (1)

Application Number
Title
Priority Date
Filing Date

GB28657/72A
Expired

GB1565797A
(en)

1971-06-18
1972-06-19
Signal processing circuits

Country Status (2)

Country
Link

CA
(1)

CA1080339A
(en)

GB
(1)

GB1565797A
(en)

1971

1971-06-18
CA
CA115,983A
patent/CA1080339A/en
not_active
Expired

1972

1972-06-19
GB
GB28657/72A
patent/GB1565797A/en
not_active
Expired

Also Published As

Publication number
Publication date

CA1080339A
(en)

1980-06-24

Contact information