FR2904168A1 - SYSTEM FOR ESTIMATING THE RECEPTION QUALITY OF A DIGITAL TRANSMISSION. - Google Patents

Abstract

System for estimating the quality of a data transmission, (the input of the system is the same as the input signal of a demodulator whose reception quality is to be estimated), the format of the communication being known and signals are known, the signal to be estimated being sampled at at least E times the modulation speed used for the transmission, characterized in that it comprises at least the following elements: o A memory for storing the signal, o a device for correlation (300) in real time with one or more known waveforms depending on the type of communication to be estimated, the number of correlators is equal to the number of patterns of the signal to be detected, o a device (200) for recognizing the number of the recognized sequence, o a correlator (500) between the stored signal and the p <th> recognized reference, o a device for detecting and refining the synchronization position and resampling of the stored signal, if a signal is detected , o a say positive interpolation (600) generating the instant when the correlation is maximum, o a filter (700) providing, in the case of detection of the pth reference, a series of Np samples resampled as a function of the synchronization position dn .

Description

1 System for estimating the reception quality of a transmission

digital The invention relates in particular to a system for estimating the reception quality of a digital transmission. It applies in all cases where at least a portion of the transmitted signal is known or can take one or more known form (s). The invention relates to the problem of estimating the reception quality within the framework of a digital type data transmission, and this independently of the demodulation system. This estimate of the reception quality is a significant help for the operator of the transmission system, because it allows him to choose at any time the most suitable transmission mode: - if the transmission quality deteriorates, he will use a more robust (and therefore at lower speed), - if on the contrary it improves, he can use a mode at higher speed and therefore less robust. This estimate also makes it possible to verify the correct behavior of the transmission system (radio procedures) in real time and in non-simulated operational conditions (in field experiments for example). The input signal is usually the audio output of the receiver, which represents the input of the demodulator. If the receiver has an intermediate output (for example an intermediate frequency or, better still, an output on two quadrature channels, the proposed system will only function better. The binary train received is a priori unknown by definition, especially more than it has undergone various modifications and transcoding in order to make the connection more reliable.

The only information available is the type of transmission used, which, in most cases, is the subject of a standard or internal documentation from the system manufacturer.

2904168 2 In other words, the only known data are the following: - the type of modulation used, - the modulation speed (expressed in bauds), - the detailed description of certain portions of the message which are fixed and systematically present. in any message; these portions are generally used for (re) synchronization of the receiver and / or for estimating the characteristics of the transmission channel in order to make reception possible and / or to specify the characteristics of the communication in progress (part often referred to as d "auto-baud").

Different systems are described in the prior art: - either the demodulator is able of itself to provide an estimate of the reception quality, generally by estimating the signal / noise ratio as a by-product of the process. demodulation, - or it is not (even if internally it does or could do it), in which case the system according to the invention is of great use. Usually, the reception quality is measured by sending a pure carrier and spectral analysis on the receiver side in order to estimate, on the one hand the power of the carrier (signal), on the other hand the noise power ( on either side of the carrier in question).

In addition to the fact that this method requires the introduction into the transmitters / receivers of additional functions, which increases their complexity, it is not necessarily relevant in the sense that during the real transmission, it is a modulated carrier. which is transmitted and not a pure carrier: which undergoes various distortions due to filters, non-linearities, automatic gain controls, - which, for reasons known to those skilled in the art, must be emitted at an average power lower than that of the carrier mentioned above. The proposed system does not require any modification of the existing equipment: it is content to observe their input signal in order to carry out its own evaluation, and to return the measured estimates to the operator. It is positioned, for example, at the output of the transmission channel, before a demodulator or as a bypass allowing the signal to be analyzed to be picked up. The invention relates to a system for estimating the quality of a data transmission, (the input of the system is the same as the input signal of a demodulator whose transmission quality is to be estimated, whether this either the final audio output or an intermediate output if it is accessible), the format of the communication being known and the signals are known, the signal to be estimated being sampled at at least E times the modulation speed used for the transmission characterized in that that it comprises at least the following elements: o a signal storage memory, o a device for real-time correlation with one or more known waveforms depending on the type of communication to be estimated, the number of correlators is equal the number of patterns of the signal to be detected, o a device for recognizing the number of the recognized sequence, o a correlator between the stored signal and the pth recognized reference, o a device for detecting and refining the position of synchronization and resampling of the stored signal, if a signal is detected, o an interpolation device generating the instant when the correlation is maximum, o a filter providing, in the case of detection of the pth reference , a series of Np samples resampled as a function of the synchronization position dn.

The invention also relates to a method making it possible to estimate the quality of a digital data transmission, characterized in that it comprises at least the following steps: 1) correlating in real time the signal received with at least one form of known wave 30 and chosen according to the type of communication in progress, 2) if a signal is received, refine the synchronization position, 3) filter the signal in order to eliminate inter-symbol interference, 2904168 4 4) filter the samples received and calculating the sum of the squares of the modulus of the differences with the original samples, 5) converting the result into a signal-to-noise ratio, 6) comparing the signal-to-noise ratio obtained with a predetermined value. One of the advantages of the present invention is to be able to estimate the reception quality independently of the demodulation system used. Other characteristics and advantages of the invention will appear better on reading the following description of an exemplary embodiment given by way of illustration and in no way limiting, appended to the figures which represent: FIG. 1, a block diagram of the various elements implemented by the method according to the invention, assuming that one places oneself in the most unfavorable case where the only signal available is the LF 15 output o Figure 2, a detail of the structure of the correlators, o Figure 3, the shape of the signal after the correlators, o Figure 4, an example of values of the interpolated signal samples, o Figure 5, an equalized constellation, and o Figure 6, the shape of the signal after equalization.

In summary, the system and method according to the invention are based on the following idea: known parts are detected in the received signal, generally by passive correlation over time. This passive correlation is equivalent to the filtering of the signal received by its "matched filter" (filter whose impulse response is the conjugate of the signal expected after time inversion) of which the modulus of the output is calculated and renormalized at each instant and the result compared. at a carefully chosen threshold. Once the detection has been carried out, the system - which has memorized the past signal 30 - uses this past signal to carry out, for example, the following operations: 5 2904168 5 - refinement of the optimum demodulation position (fine synchronization) by searching, for example example, by interpolation, the position where the correlation is maximum, - resampling (or interpolation) of the received signal to have samples of the best possible quality, - calculation of a so-called "equalizer" filter which improves the quality samples by eliminating what is called inter-symbol interference "generally due to the filters employed and the presence of multiple paths, - filtering of the signal by the equalizer, 10 - estimation of the signal-to-noise ratio from the samples thus obtained, - correction of the value obtained to obtain a more precise value of the true S / N ratio, knowing that the equalizer, being calculated from the received signal is itself "noisy" and introduces a so-called noise "multiplicative" because it is proportional to the received signal. FIG. 1 represents a block diagram of a correlation system according to the invention. The latter is based, in particular, on the use of a set of correlators which provides, where appropriate, the number p of the reference signal 20 detected, and a series of re-synchronized samples which will be used for the actual estimation. known as the S / N ratio, the system comprises, for example, the elements described below which use a signal sampled at E times the modulation speed, E must in practice be at least 4 samples per symbol to lose the least possible information: a device 100 for switching to the baseband known to those skilled in the art carrying out the usual operations of transposing the signal to frequency 0 and low-pass filtering providing the complex signal which is used throughout the continuation of the processing operations (only if the receiver does not itself supply the corresponding output signals), - a memory 200 which makes it possible to store sufficient signals in the past to be able to carry out the processing operations described below, 29 04168 6 - a set 300 of P correlators if the number of patterns to be recognized is P; the outputs of the correlators are of the binary type (no detection or else detection). These correlators denoted 301, 302, ..., 30P are detailed in FIG. 2.

In some cases, several correlators are tuned to frequency-shifted versions of the same pattern. - a selection system 400 of the recognized signal which determines, where appropriate, what is the number p of the recognized sequence, - a generic correlator 500, capable of re-correlating the signal stored in the memory 200 and the pth reference, only at instants close to the instant of detection, - a memory 501 which stores the samples representative of the P distinct references and provides the pth reference on request, - an interpolation device 600 which determines very little about 15 l exact moment at which the correlation would have been maximum. It can for example locally approximate the correlation peak by a parabola. If the calculated maximum is found at sample c (n), the "exact" position of correlation will be equal to n + dn (-1/2 <= dn <= +1/2), dn being a calculated offset. for example by the following equation which assumes that the correlation is a parabolic function of time: dn lc (n + 1) - c (n -1) 2 c (n - 1) - 2 c (n) + c (n A resampling system 700 which, if the pth reference has been detected, provides a series of Np resampled signal samples as a function of the synchronization position dn. These samples have the following numbers, where E is the number of samples per symbol: 30 n + dn n + dn û E 2904168 n + dn - 2 E n + dn - (Np - 1) E For convenience for the following we renumber them in the opposite direction to obtain the sequence ordered according to the increasing times, denoted r (0), r (1), r (Np - 1). r (0) is the sample at position n + dn - (Np - 1) E r (Np - 1) is the sample at position n + dn. The values calculated are by parabolic approximation as a function of dn, from the stored samples m (i). If we set p = n - (Np - 1 - k) E we will calculate the sample for any position k = 0, 1, 2, ... Np - 1: r (k) (m (p - 1 ) - 2 m (p) + m (p - 1)) d22 + (m (p + 1) - m (p - 1)) Zn + m (p) Figure 2 shows the detail of the correlators. The pth correlator 15 associated with the pth reference, numbered 30p, has the particular role of constantly comparing the received signal samples, spaced apart by a symbol (E samples) with the Np theoretical values corresponding to the expected sequence. For this, the correlator comprises a delay line with taps 30p1, the outputs of which are multiplied by the conjugate complexes of the samples 20 of the expected sequence by the complex multipliers 30p2. The outputs of the multipliers are summed by the adder 30p3 from which the square of the modulus of the output is calculated by the system 30p4. The typical shape of the output signal thus obtained appears in FIG. 3, after re-normalization, in a diagram where the time axis is graduated in 25 samples. At the same time, in the 30p5 system, the total energy of the samples present in the 30p1 delay line is evaluated by a conventional system where the square of the modulus of the outgoing sample is subtracted at each sampling instant and the square of the modulus of the incoming sample. The difference obtained makes it possible to update the contents of an accumulator which will contain the desired value.

7 2904168 8 The output of the accumulator is multiplied by a constant C, and the output of the device 30p4 is compared to the result in the comparator 30p6 to decide whether or not to detect the sequence. This is a system for automatically adjusting the detection threshold 5 which guarantees a constant probability of detection whatever the level (a priori unknown) of the signal received. The generic correlator 500 (FIG. 1) does not need a detection threshold since it knows that the signal is present, and at what time the correlation peak appeared.

For the rest of the processing operations, the method preferably uses only the interpolated signal samples denoted r (0) ... r (Np - 1) and the samples z (0) ... z (Np - 1) of the detected reference. FIG. 4 represents an example of values of the samples r (i) in the complex plane at the entry of the last phase of the processing operations.

In the method one seeks, for example, to calculate a finite impulse response filter whose coefficients are complex and: denoted a (i), for i = 10, 10 + 1, ..., 11. IO is negative, 11 is positive. The calculation is done, for example using the least squares method, from the samples calculated previously.

20 We minimize the following mean squared error: n = Np- 1 + IO n = I1 i = I1 z (n) - a (i) r (n - i) i = 10 2 E_ By canceling the derivatives with respect to conjugates of a (i), we obtain the following 25 equations: 2904168 9 aE i = I1 n = Np - 1 + IO aa * ka (i) r (n - i) r * (n - k) () i = IO n = I1 n = Np-1 + I0 z (n) r * (n - k) = 0 n = I1 k = IO ... I l This system of equations is Hermitian and can be solved using the algorithm known as "Choleski decomposition".

5 After the resolution, we use the coefficients a (i) to filter the received samples, according to the simple expression: i = I1 r '(n) = a (i) r (n - i) i = 10 Figure 5 represents the result obtained for IO = -5 and 11 = +9 10 (equalizer with 15 coefficients): the detected sequence was binary and real, with nominal values of -1 or +1, just like in the previous example. The modulus of the equalizer coefficients is given in this figure by way of example (there is generally a much higher value than the others, the one corresponding to a (0), here the 5th coefficient). The method then calculates to calculate the start-squared error E to obtain an amount which is inversely proportional to the signal-to-noise ratio (E measures an error). After calibration in the laboratory, for example, we establish the correspondence between the S / N noise ratio measured by this method and the true S / N noise ratio, for each use.

20 This system can be used for practically all standardized systems, in particular those covered by a STANAG (4285, 4539, MIL110, etc ...) It can be used as a decision aid to adapt the format of transmission (speed, robustness) under current reception conditions.

The method according to the invention applies in particular for checking the radio procedures of modems implemented in the I / Rs (HF, VHF, etc.) in an operational environment. 5

Claims (2)

Claims 1 û System for estimating the quality of a data transmission, (the input of the estimation system is the same as the input signal of a demodulator whose transmission quality is to be estimated), the format of the communication being known, the signal to be estimated being sampled at at least E times the modulation speed used for the transmission, characterized in that it comprises at least the following elements: a memory for storing the signal, a device for correlation (300) in real time with one or more known waveforms depending on the type of communication to be estimated, the number of correlators is equal to the number of patterns of the signal to be detected, o a device (200) for recognizing the number of the sequence recognized by the correlation device (300), o a correlator (500) between the stored signal and the recognized reference sentence, o a device for detecting and refining the synchronization position and resampling of the stored signal é, if a signal is detected, o an interpolation device (600) generating the instant when the correlation is maximum, o a filter (700) providing in the case of detection of the pth reference, a series of Np samples resampled according to the synchronization position dn. 2 û System according to claim 1 characterized in that it comprises a device (100) adapted to shape the input signal such as a device allowing the passage in baseband. 3 û System according to claim 1 characterized in that a correlator p (30p) comprises at least the following elements: 2904168 12 o a tapped delay line (30p1) whose outputs are multiplied by the conjugate complexes of the samples of the sequence expected by complex multipliers (30p2), o an adder (30p3) to sum the outputs of the multipliers of which we calculate the square of the modulus of the output by a system (30p4), o a system (30p5) of evaluation of the total energy of the samples present in the delay line by a conventional system where one subtracts at each sampling instant the square of the modulus of the outgoing sample and adds the square of the modulus of the incoming sample. 10 where the output of the accumulator is multiplied by a constant, and the output of the device (30p4) is compared to the result in the comparator (30p6) to decide whether or not the sequence is detected. 15 4 - Method making it possible to estimate the quality of a digital data transmission characterized in that it comprises at least the following steps: 1) correlating in real time the signal received with at least one known and chosen waveform depending on the type of communication in progress, 2) if a signal is received, refine the synchronization position, 20 3) filter the signal in order to eliminate inter-symbol interference, 4) filter the received samples and calculate the sum of the squares of the modulus of differences with the original samples, 5) convert the result into a signal / noise ratio, 6) compare the signal / noise ratio obtained with a predetermined value. 5. Method according to claim 4, characterized in that the optimum demodulation position is sought by interpolation. 6 -Process according to claim 4 in that, after step 2), the stored signal is re-sampled. The method of claim 4 characterized in that an equalizer filter is determined in order to improve the quality of the samples of the signal by eliminating the inter-symbol interference. 5 8 ù The method of claim 4 characterized in that after calibration in the laboratory, a correspondence is established between the signal / noise ratio measured according to the steps of the method and the signal / noise ratio for each use.