GB2069297A

GB2069297A - Resolving a wanted signal

Info

Publication number: GB2069297A
Application number: GB8101560A
Authority: GB
Original assignee: Dornier GmbH
Current assignee: Dornier GmbH
Priority date: 1980-02-01
Filing date: 1981-01-19
Publication date: 1981-08-19
Also published as: GB2069297B; FR2475323B1; FR2475323A1; DE3003556C2; DE3003556A1; IT8069022A0; IT1165730B

Abstract

In a method of resolving a useful signal from a band limited signal with a superimposed noise signal, a received signal m(t) is converted into digital values m(ti) which are then fed to a microprocessor (3) for processing. The microprocessor (3) generates a simulated digital noise signal <IMAGE> where ai are approximation coefficients which vary slowly with time and ci(t) are formulation functions chosen according to the characteristics of the signals to be processed. The simulated noise signal is subtracted from the digitized input signal to yield the wanted, useful signal n(ti). By allowing the microprocessor (3) to vary the rate at which the values m(ti) are produced, cancelling of the actual noise signal S(t) with the simulated noise signal @(t) can be performed without introducing delay distortion into the useful signal n(ti) and without varying the formulation functions ci(t). <IMAGE>

Description

A method and apparatus for resolving a wanted signal This invention relates to a method of and a device for determining a useful signal from a band limited signal on which a noise signal is superimposed.

To resolve signals with superimposed noise interference, it is known to employ rapid Fourier transformation or analogue or digital model filters. These methods have the disadvantage that the useful signals filtered out of the resulting signal are delayed and distorted. Therefore, in the presence of severe noise interference, delay-free analysis of the signal form of the filtered useful signal cannot be carried out.

Methods in which the complete signal spectrum or received signal is subjected to a rapid Fourier transformation and in which a pattern recognition is carried out on the measured quantity which is so transformed, lead to difficulties if the wanted signals and the noise signals do not originate from reproducible processes.

It is an object of this invention to provide a method of separating a wanted signal from noise signals without introducing delay distortion in the wanted signal.

According to this invention there is provided a method of determining a useful signal from a band-limited signal on which a noise signal is superimposed, wherein an input signal is converted to digital form by an analogue-to-digital converter controlled by a relative or real time clock and wherein the noise signal which is to be compensated can be simulated in a microprocessor by a few defined formulation functions (j (tj) with slowly variable characteristic values (a (to)), the noise signal which is simulated in form being used to cancel out substantially without delay the noise signal in the input signal thereby to provide the required useful signal substantially without delay distortion.

The main advantage of this method is that it allows the detection of wanted signals which occur rapidly and/or once in isolation in a noise signal, provided they are within a certain range. Since the major part of noise signals which have to be compensated contains signals of substantially invarying form, these invarying properties are utilised in order to derive a rapidly variable noise signal of a chosen form from a few slowly variable characteristic values and invarying signal functions. A delay-free noise compensation is feasible by applying this principle. Furthermore, by controlling the moment in time when the input signal is sampled as a function of the signal form, the computing complication is reduced and parameter adaptations no longer apply. Thus, it is possible to use commercially available microprocessors for detecting and processing rapid signals.

A microprocessor connected to an input receiver is controlled by a relative time or real time clock disposed upstream of it and which is arranged to trigger an anlogue-to-digital converter. Each digitized input signal value produced by the analogue-to-digital converter is processed in accordance with the times preset by the relative or real time clock. The microprocessor includes a number of signal generators having fixed centre frequencies, the amplitudes and phases of which vary slowly from computing step to computing step. The sum of the signal generator outputs is compared with the digitized value and the resultant difference is so evaluated that it can be used for very slow correction of the amplitudes and phases of the signal generators.

Since the band limited noise signal is reduced to a few very slowly variable values, once-only short-duration signals (useful signals) have no effect on the slowly variable values. As a result of the subtraction process, signals on which band limited noise signals are superimposed can be separated from these latter without delay.

The invention will now be described by way of example, with reference to the drawing, in which: Figure 1 is a simplified block diagram of a signal processing arrangement in accordance with the invention; and Figure 2 is a flow diagram of signal processing steps carried out by the microprocessor of Figure 1.

Figure 1 illustrates an arrangement for extracting a wanted signal from a signal on which noise signals are superimposed. A signal m(t) from an input receiver or sensor 1 is fed to an analogue-to-digital (A/D) converter 2 which communicates with a microprocessor 3 including a memory and, via the microprocessor 3, with a relative or real time clock 4. The microprocessor is constructed or programmed to control the clock 4 which in turn triggers the AID converter 2 so as to control the points in time at which the input signal m(t) is sampled.In accordance with the incrementing of the clock 4 (responsive to time increment signals AT1), the input signal is periodically converted into a measured value m(tj = t.1 + ti.2... + AT1) and is processed in the microprocessor 3 as a series of individual values m (t1.1), m(ti.) ...

Referring to Figure 2, the measured values are processed in the microprocessor 3 as follows. The signal m(tj) comprises a noise signal S(tj) and a wanted signal n(tj). In orderto separate the noise signal S(tj) from the wanted signal n(tj), signal m(tj) is subjected to a splitting process 5 and for each part of the signal a noise value compensation process is carried out via a demodulator 6, a low-pass filter 7 and a real time modulator 8.In this process, the noise signal S(t) occurring in an approximation interval TAP can be expressed in terms of formulation functions wj(t) and characteristic values (approximation in coefficients) aj:- S(t) = E:aj . wj(t) for t - TAP S 1 S t By a suitable choice of formulation functions i(t), an estimate t(t) of the real interference of any required accuracy can be arrived at:

in which the coefficients a1 are determined by an operation (specifically a demodulation operation) which is dependent upon the formulation function as follows:: a = < S(t) ; (t) > Since the approximation interval TAP, as a result of the real time sampling, is fixed by the point in time t of the last sampled value m(tj), the approximation coefficients a are functions of the time.

When the formulation functions i(t) are chosen such that they are adapted to the form of the noise signal, the approximation coefficients aj(t) are slowly varying functions of the time. The best approximation is achieved when the time variation of the approximation coefficient is at a minimum: T [aj(t)]2 , min Since the formulation functions (t) are related to an approximation interval RAPT, the minimum is found in that, for equal formulation functions, the real approximation interval TAP iS extended or contracted by varying the point in time when the measured signal is picked up. Thus the formulation functions do not need to be altered and the approximation coefficients remain as constant functions.Since departures from an assumed form of noise signal give rise to changes of varying speed in the approximation coefficient, they ~ can be eliminated by a suitable low pass filter 7. When this happens, the mean approximation coefficients a are determined in a regulating process. Depending on the formulation functions, so the noise signal S(tj) is split into parts Sj(t). The variation in the associated approximation coefficient is determined from the difference between Sj(t) and an estimated value Sj(t) related to the formulation functions Ipj(t)::- Aaj(t) = < [Sj(t) - tj(t)]; wj(t) > The mean approximation coefficient is then integrated until such time as the increase disappears: a = a, + K{Aa,(t) - a,1 From the mean approximation coefficient aj and the formulation function qri(t), the corresponding noise signal estimate is determined: Sj(t) a, .

Since the resultant noise signal estimate S can be indicated for any point in time within the approximation interval, comparison with the measured signal or measured value m(t) recorded by the input receiver or sensor 1 (Figure 1) is possible, giving delay-free noise signal compensation or determination of the useful signal.

Claims

1. A method of determining a useful signal from a band-limited signal on which a noise signal is superimposed, wherein an input signal is converted to digital form by an analogue-to-digital converter controlled by a relative or real time clock, and wherein the noise signal which is to be compensated can be simulated in a microprocessor by a few defined formulation functions (1+; (to)) with slowly variable characteristic values (aj (to)), the noise signal which is simulated in form being used to cancel out substantially without delay the noise signal in the input signal thereby to provide the required useful signal substantially without delay distortion.

2. A method according to claim 1, wherein the relative time interval between successive samples of the input signal, is so controlled that the formulation functions (j (to)) remain invariable while the characteristic values (aj(tj)) remain slowly variable.

3. A method according to claim 1 or claim 2, wherein the characteristic values (aj) are corrected by constant comparison of the actual value with the estimated value (S) of the noise signal derived by the formulation functions.

4. Apparatus for carrying out the method of any preceding claim, the apparatus comprising an input signal receiver or sensor, a digital microprocessor with memory, a relative or real time clock and an analogue-to-digital converter, wherein the relative or real time clock determines the data acquisition time of the analogue-to-digital converter in response to the microprocessor, the microprocessor generating the formulation functions (\jim) and determining the slowly variable characteristic values (aj) and also carrying out the compensation.

5. Apparatus according to claim 4, wherein the formulation functions are pre-programmed in the apparatus according to the intended use of the apparatus.

6. Apparatus according to claim 4 or claim 5, including, for each formulation function, an arrangement for generating the characteristic values comprising a demodulator, a low pass filter and a modulator.

7. Apparatus according to any of claims 4 to 6, wherein the said arrangement is implemented by a program controlling the microprocessor.

8. A method of resolving a useful signal substantially as herein described with reference to the drawing.