DE10142538A1 - Signal runtime measurement method for electric, electromagnetic or acoustic signals measures a signal between a transmitter and a receiver or a transmitter used simultaneously as a receiver - Google Patents

Abstract

A transmitter emits pulsed signals that reach a receiver as a group of waves rising and falling over multiple fluctuations or as several groups of waves of this type. A cross-correlation function for a received signal is generated with a reference signal. A parameterized function is digitally adapted directly to a cross-correlation function that exists in a digitalized form.

Description

The invention relates to a method to measure the runtime of an electrical, electromagnetic or acoustic Signal between a transmitter and a receiver or a transmitter that at the same time as a recipient is used, the transmitter emitting pulse-shaped signals that the receiver than over several vibrations rising and falling wave group to reach.

Out EP 0797105 respectively. EP-PS 0452531 such a method is known. The transit time measurement, for example of an ultrasound signal, is based on the fact that a pulse-shaped sound signal is injected into the measurement medium by a transmitter transducer and is detected by a receiver transducer after passing through the measurement section. The sound propagation time is the time difference between the transmission process and the arrival of the ultrasound signal at the receiving location. In the case of sound reflection, the transmitted sound signal is reflected at a boundary layer between the measuring medium and an adjacent medium before it reaches the receiving transducer. In this case, a single sound transducer can also be operated alternately as a transmit and receive transducer.

The ultrasonic transit time measurement is for diverse measuring tasks used. Which includes for example distance measurement, flow measurement or Concentration measurement in gaseous and fluid Media.

For the determination of the term is in EP 0797105 described a method for determining the maximum of the envelope of the cross-correlation function. This method is based on the fact that a received and digitized signal is correlated with a reference signal previously recorded in the same way and the shift in time, ie the transit time between the two signals, is determined from the position of the maximum of the envelope of the cross-correlation function.

A disadvantage of this method is its inapplicability to signals that are unstable in time and signals with a low signal / noise ratio. If signal deformations occur in the received signal with respect to the reference signal, the envelope of the cross-correlation function, which now becomes asymmetrical, sometimes delivers plateaus or even several local maxima. The height of these local maxima shifted on the time axis relative to the original maximum can reach and exceed that of the actual maximum. Apparent runtime changes occur, sometimes even runtime jumps, and an evaluation of the method using the maximum of the envelopes of the cross-correlation function becomes impossible. The accuracy of the transit time measurement according to this known method thus depends sensitively on the temporal stability of the received signal in its signal form. Signal deformations of the in 1 The type shown can arise, for example, from variable noise coupling in from the environment, or thermal drifts in the running path or the transducer elements, or can occur in the case of a nonlinear sound field, for example in the case of focusing transducers.

The invention is based on the object a method for measuring the transit time of an electrical, electromagnetic or acoustic signal, even with long-term signal deformation of the received signal of the reference signal with high precision Delivers measured values of the runtime.

To solve this task, the new methods of the type mentioned in the characterizing Part of claim 1 specified features. In the subclaims are advantageous developments of the method described.

The analog received signal is sampled, digitized, stored and correlated with a second received or transmitted signal that has already been digitized and stored beforehand, or with a synthetic signal provided. In contrast to EP0797105 Furthermore, the envelope of the cross-correlation function found in this way is not formed, but a function (parameterized appropriately (see application example 2)) is numerically adapted directly to the cross-correlation function itself. As a result, all available data points for the parameters of the adaptation function to be determined are used and not only the local extremes of the cross-correlation function forming the envelope, which in the Fourier space only corresponds to the use of the low frequencies determining the envelope and thus to a strong loss of information. The method described thus enables the use of the entire data set of the cross-correlation function.

At the same time, this procedure in the case of plateaus, local minima and secondary maxima of the envelope no longer usable for determining the runtime, as in the application examples 1 (determining the center of gravity of the cross-correlation function) and 2 (adaptation of a Gaussian double peak cosine function) for example is shown.

wobei R(t) für das Referenzsignal und S(t) für das gemessene Empfangssignal steht.The cross correlation function K is defined via a convolution integral in the spatial area

where R (t) stands for the reference signal and S (t) for the measured received signal.

The running parameter Δ of the cross correlation function has the unit of a time (seconds) and stands for the considered delay between R and S. In the Fourier space the folding integral transforms to the product of the Fourier transforms of R and S. The reference signal R (t) can, as in claims 4-6 also described a using parameter-dependent model formulas that the expected signal curve if possible reproduce well, be calculated synthetic signal.

For example, it can be composed of three temporally successive functions: one with A ₁ (1-exp (-t /? 1)) increasing exponentially and carried with the carrier frequency, then one for a certain time (t _h - the holding time) one Wave group of constant amplitude A ₂ , and then a signal decaying exponentially with A ₃ exp (-t / τ ₂ ) ( 2 ). This synthetic signal is thus by ten parameters: the carrier frequency ω, the individual phase angles φ ₁ ,, φ ₂ and φ _{3 of} the carrier frequency for the three curve sections, the individual amplitudes A ₁ , A ₂ , A ₃ , the time constant τ _{1 of} the increasing and τ _{2 of} the decaying section, and the holding time t _{h are} determined. These parameters can also be selected (claims 5 and 6) in such a way that the curve adaptation to the resulting cross-correlation function K is optimized, for example via a self-regulating artificial neural network which contains the parameter set {ω, φ ₁ , φ ₂ , φ ₃ , A ₁ , A ₂ , A ₃ , τ ₁ , τ ₂ , t _h } in such a way that the correlation coefficient of the adaptation is maximum, or the sum of the Gaussian squares of the adaptation (cf.

Application example 2) minimal and therefore the parameter set of the adaptation parameters ({B ₁ , B ₂ , μ ₁ , μ ₂ , σ ₁ , σ ₂ , ω, ϕ} in the application example, which is then determined from the curve fitting to the cross-correlation function 2 ) can be specified as precisely as possible in order to then calculate the desired transit time t _d from these.

The use of a synthetic Reference signal has opposite A measured signal has the advantage that it is noise-free is present. The noise goes with a measured reference signal twice in the calculation of the cross correlation function, and through the received signal and the reference signal itself. When using a synthetic reference signal goes against it the noise only due to the measured received signal, so that halving the error in the values of the cross-correlation function results.

There is also a measured reference signal just a temporal "snapshot" and represents hence actually the underlying signal to which correlation is to be carried out is insufficient. By appropriate modeling and choice of parameters, the actual Signal properties can only be worked out. Furthermore has the use of a synthetic reference signal for cross-correlation the advantage that there are any number of support points with any accuracy additionally calculate and the Nyquist frequency of the synthetic reference signal no limit is subject to a real measurement.

The particular advantage of using it of a synthetic signal as a reference signal is that it over a digital-to-analog converter transferred to the transmitter converter can become and an artificial Neural network depending on the parameters of the synthetic function the conditions of the measuring section, for example the thermal Drift of the transmitter / receiver converter (s) or the occurring Noise can be adjusted in such a way that the measured and digitized received signal the cross-correlation function in the sense of the curve fitting described, for example with regard to the minimum of the sum of the Gaussian Error squares or the maximum of the correlation coefficient of Curve fitting can be optimized.

This "adaptive signal formation" can on the one hand for the measurement setup optimal signal can be found, on the other hand reacts the system self-regulating to external thermal Drift and is able to compensate for noise to be.

Below are two application examples for one Calculation of the runtime from the entire data set of the cross correlation function specified. Application example 2 gives a variant for curve fitting to the entire cross-correlation function again.

1. Application example: focus the cross-correlation function

The center of gravity of the cross correlation function K (Δ) is defined as follows:

For a discrete data set, this equation is written in the form:

Weighted with the associated (positive) squares of the cross-correlation function, it is summed over all delays Δ _i and the center of gravity S of the cross-correlation function is thus calculated. On the one hand, this method uses all N available data points and thus the entire information store, on the other hand it is insensitive to a plateau of the envelope of the cross-correlation function that may occur, or several local extremes. In this case, the runtime difference you are looking for is simply given by the following formula:

K _{ref, emp stands} for the cross-correlation function between the received and reference signal, K _{emp, emp} for the cross-correlation of the reference signal with itself, for determining the zero point of the transit time.

2. Application example (adaptation a Gaussian double peak cosine function)

mit dem Parametersatz {B₁, B₂, μ₁, μ₂, σ₁, σ₂, ω, ?} numerisch unter Ausnutzung des gesamten Datenvorrates nach dem Minimum der Summe der Gaußschen Fehlerquadrate angepasst.In this example, the following function is applied to the determined cross correlation K:

with the parameter set {B ₁ , B ₂ , μ ₁ , μ ₂ , σ ₁ , σ ₂ , ω,?} numerically adjusted using the entire data set according to the minimum of the sum of the Gaussian squares.

angegeben werden. Hier dienen die Minenwerte μ₁, μ₂ der beiden Gaußpeaks zur groben, der Phasenwert φ zur Feinbestimmung der Laufzeit. Dieses Verfahren ermöglich schon eine stark verbesserte Präzision und Stabilität gegenüber EP 0797 105 .The running time t _d to be determined can then, for example, according to the formula:

can be specified. Here the lead values μ ₁ , μ _{2 of} the two Gaussian peaks are used for the coarse, the phase value φ for the fine determination of the transit time. This procedure already enables a greatly improved precision and stability compared to EP 0797 105 ,

The relationship given here between runtime and the parameters determined from curve fitting can also be artificial Neural network to be replaced so that in the learning phase one enough size Number of samples of known transit time delays in the measurement section introduced, and the artificial Neural network these runtimes are communicated so that it is self-reliant Recognizes the relationship between all fit parameters and the runtime.

The invention is therefore distinguished by the following advantages:
Avoiding the process becoming unusable in the event of plateau formation or formation of further maxima in the envelope of the cross-correlation function,
- Nutrition of the entire available data set of the cross correlation function
- Simplify measurement, reduce noise and improve cross-correlation function by using a synthetic reference signal
- Possibility of self-regulating signal optimization by using a synthetic reference signal and an automatic control mechanism (e.g. an artificial neural network)
- Higher accuracy and significantly improved stability of the determined transit time with a low signal-to-noise ratio and with respect to thermal drifts in the transmit / receive converters or the measurement section.

List of pictures:

1 : The procedure becomes unusable after EP 0797105 in the event of a signal deterioration, the same duration is required;

2 : Example of a parameterized synthetic reference function,

Claims (12)

Method for time-of-flight measurement by means of a cross-correlation function of an electrical, electromagnetic or acoustic signal between a transmitter and a receiver, or a single transceiver, the transmitter emitting pulse-shaped signals that identify the receiver as a group of waves rising or falling over several vibrations or several groups of this type achieve, characterized in (a) that a cross-correlation function of the received signal is generated with a reference signal and (6) that a parameterized function is numerically fitted directly to the digitized cross-correlation function and (c) the transit time depending on the parameters determined in this way is highly precise and stable is specified. A method according to claim 1, characterized in that the Reference signal one with undisturbed transmission digitized received signal corresponds. A method according to claim 1, characterized in that the Reference signal corresponds to a digitized transmission signal. A method according to claim 1, characterized in that the Reference signal to a synthetic, i.e. not obtained by measurement Signal corresponds. Process according to the claims 1 and 4, characterized in that an artificial neural network for Determination of the parameters of the synthetic reference signal used becomes. Process according to the claims 1, 4 and 5, characterized in that the synthetic reference signal via a Digital / analog converter is transmitted to the transmitter converter and an artificial one Neural network depending on the parameters of the synthetic function the conditions of the measuring section, for example the thermal Drift of the transmitter / receiver converter (s) or the noise that occurs variable in time in such a way that with the measured and digitized received signal the cross-correlation function optimized in the sense of the adjustment described in claim 1 (b) becomes. A method according to claim 1, characterized in that a artificial Neural network the runtime from the from the cross correlation function certain parameters in the sense of claim 1 (b) is derived. A method according to claim 1, characterized in that a Sample known measurement properties in the measurement section for calibration is introduced. A method according to claim 1, characterized in that a Another measurement signal when the direction of the electrical electromagnetic or acoustic signal is obtained and that the difference between the run times measured in both directions is determined. A method according to claim 1 and 9, characterized in that the flow rate or the mass flow in a liquid or gaseous medium is measured. A method according to claim 1, characterized in that the Receive signal electronically or in its digitized form by a bandpass with the frequency of the carrier oscillation as the pass frequency is filtered. A method according to claim 1, characterized in that with an ultrasonic measuring device Distances or Velocity of sound measured in transmission or reflection become.