DE102004041121B3 - Estimation of signal parameters in radio channels by rotational invariance techniques involves transmission of test signals at intervals and use of test algorithm and complex amplitude values - Google Patents

Abstract

The test signal may be spread and distorted by multipath transmission. The radio receiver has a number of individual antenna elements to receive overlapping test signals. Delay times are noted and a minimum distance search may be instituted. Reflection chains may be detected and approximated using a polynomial. The test procedure may be used with GPS and broadband systems.

Description

The The invention relates to a method for examining a between a transmitter and a receiver radio channel, where the transmitter is a periodic test signal which is subject to multipath propagation and thereby divided into a plurality of path-weighted single test signals and which consists of a plurality of blocks with a block duration, the at least as big as the longest Duration of all to be considered, path-weighted single test signals. This will be on the receiver a plurality of individual antennas of a receiving antenna array superimposed path-weighted test single signals each as a measurement signal of a Receive measuring channels and these measuring signals are the inputs of a Supplied to multiplexers, at each of its inputs connected through at least one block duration to its output from which the multiplexed measurement signals of a signal and data processing means supplying the multiplexed ones Transferred measurement signals of a measurement data set in locally complex impulse responses and as a result, a direction-resolved complex, the properties of the radio channel reflecting impulse response. Further determined the signal and data processing device for this during processing the local complex impulse responses using a direction estimation algorithm, in particular one based on a "maximum likelihood" method ESPRIT algorithm, the directions of incidence and complex amplitude values, i.e. the number of reflections, their delay, power and phase, for every determined direction of incidence at each sampling time of the radio channel like a snapshot.

In one of the communication serving radio channel, in particular in one Mobile radio channel, arise Mehrwegeausbreitungen of the transmitter of the Radio link emitted signal, which is due to the propagation this signal in different ways, giving it to Reflections, shadows, scatters and Doppler shifts this signal comes. At the receiver of The radio link then becomes a plurality of path-weighted signals received, who have spread in various ways and overlay.

to Further development and optimization of mobile radio systems become statistically relevant channel models needed, which describe the behavior of a radio channel realistically. For this purpose must the complex properties of each of the local Conditions dependent Channel model as possible Completely recorded and described. For measuring a real radio channel becomes ordinary a broadband measurement of the complex transfer function of each Time-variant radio channel accomplished, being used as a measuring system so-called channel sounder is used, the transmitter side a test signal radiates whose local, Complex impulse response at the end of the radio channel received and evaluated becomes.

at satellite navigation, for example, happens in the receivers Errors in positioning due to reflections from the satellite transmitted radio signal to objects, e.g. on buildings, in the Propagation channel. These reflections reach the receiver delayed for Direct LOS (Line of Sight) signal. Decisive for the positioning error are the power, the delay (Delay) and the modulation of the reflections. Especially static Reflections with a delay, less than the duration of a chip of the spreading code used is, can are not eliminated today in satellite navigation receivers. They overlap the direct signal and lead to a faulty transit time measurement of the satellite signals. At the known, widely used GPS (Global Positioning System) can this Error some 10 meters. The necessary resolution for Characterization of the channel lies in the direction of deceleration in the area from nanoseconds (ns).

Around Such multipath radio channels to be able to describe exactly are many measurements to perform in the course of which rural and urban canals for pedestrians and Motor vehicles are measured. This does not only apply to satellite navigation to, but also to other, based on a synchronization Systems and above in addition to all radio communication systems affected by multipath problems to. Especially in future Broadband systems, the exact description of the radio channel is necessary.

Out DE 197 41 991 C1 is a channel-sounder-measuring system of the company MEDAV Digital Signal Processing GmbH, Uttenreuth, DE known from which has been assumed in the invention. Here, a fixed method is used for determining a direction-resolved complex impulse response of a radio channel composed of a series of steps. First, a test signal is output at one end of a radio link from a transmitting antenna, which is subsequently subjected to multipath propagation and thereby divided into a plurality of path-weighted test signals which together represent the direction-resolved complex impulse response, and a plurality of sequential equidistant blocks of one block length which is at least as big as the longest term of all too considering path weighted test signals.

At a spatial Antenna arrangement with several each comprising a receiving antenna measurement channels become the superimposed ones path-weighted test signals at the other end of the radio transmission path received, each receiving antenna emits a measurement signal, the then passed to each one input of a multiplexer. The multiplexer is over a control unit for cyclically switching between the inputs is activated, each input at least for the duration of a block length is switched through to the output of the multiplexer.

It Then, a measurement data set is formed, each one block contains each of the measurement signals. It then cyclically more measurement records are formed, the time between successively scanning the same measurement channel so chosen is that on the one hand, the sampling theorem with respect to the Dopplerbandbreite of the radio channel for fulfilled each measuring channel is and on the other hand for every measuring data set all measuring channels can be sampled.

After that become the values of the local complex impulse responses by interpolation all measuring channels determined at a common time. Finally, the interpolated Values the direction-resolved complex impulse response, which provides information about the characteristics of the examined radio channel gives. From the local complex impulse responses are using a direction estimation algorithm, the directions of incidence and with the help of this through a beam images (beamformer) the complex Amplitudes determined.

When Direction estimation algorithm can preferably use a so-called ESPRIT algorithm which is well known and, for example, in the article by Roy Kailath: "ESPRIT - Estimation of Signal Parameters via Rotational Invariance Techniques "in IEEE Transactions on Acoustics, Speech and Signal Processing, Vol. 7, 07.07.1989 is described.

In order to a high resolution in the direction of delay results, one is possible high measurement bandwidth required, and for the determination of Doppler shifts and Doppler bandwidths, a high sampling rate is needed. The Measurement data is now displayed on this sampling rate grid. With the belonging to the channel sounder Software can then be delayed and calculate Doppler spectra from the measured data. However, it is not possible Isolate individual reflections and represent their characteristics independently.

With The known measuring method can reflections in static measurements only approximately, be analyzed very limited in dynamic measurements. In particular about the frequencies of echoes, the duration of echoes, "birth" or "death rates", the number of Reflections, mean powers and delays, variations of power and delay can not statistical statements are made.

Of the Invention is now the object of the known radio channel investigation method so educate that it is possible is to isolate individual reflections in the measured data and their Characteristic independent especially with dynamic channel measurements. moreover it should be possible become, statistical statements about to make a variety of characteristics of the measured channel, which was not possible until now was, but to model the satellite navigation channel, to Design of navigation systems and for the development of solutions is necessary to reduce the influence of errors by reflections in the future on. Similar to the use in these special satellite navigation systems and other synchronization-based radio systems, the method to be created by the invention also on measured data of others Multipath channels as in radio communication and especially in mobile communications occur, be applicable. Especially in future broadband systems is a more detailed description of the channel necessary.

According to the invention, relating to a method for examining a radio channel of relates initially mentioned type, this object is achieved in advantageous Way solved by that the result of this calculation for noise suppression band limited and filtered in both time and delay directions, then using a four-dimensional minimum distance search algorithm in time, delay, Power and direction of movement Detected chains of reflections and subsequently be approximated by polynomials, so that the result is a very exact information about the delay course every reflection in the observation period as well as its beginning and end, and that with the help of this information through interpolation of Measured data of the waveform and the spectrum for each reflection be extracted from the measured data.

After the measurement data has been postprocessed in a known manner firstly with the "Super Resolution Algorithm" ESPRIT, which is based on a maximum likelihood method, in the channel sounder software, the result of which is an estimate of the number of reflections, their delay , Performance and phase to each Is the sampling time (snapshot) of the channel, the result of this calculation according to the invention is now band-limited and filtered in both the time and delay directions to suppress noise in the ESPRIT result.

through of the four-dimensional "minimum Distance Search "algorithm that in time, delay, Power and movement direction then become chains of reflections detected and then approximated by polynomials. The result is a very accurate determination the delay course reflection during the observation period and the determination of their Beginning and end. With this information can subsequently off the measurement data grid by interpolating the waveform and the Spectrum for every single reflection in the measured channel can be calculated.

On Reason of information about the delay course every reflection as well as over whose beginning and end can advantageously statistical statements, in particular about frequencies of reflections, duration of reflections, "birth" or "death rates", number of reflections, mean powers and delays, Variations of power and delay of reflections are taken. This also makes it possible to describe the radio channel very accurately and to model.

What relates to the detection of individual reflections in the measured data, so far only for static measurements, i. For temporally constant delay curves the Characteristics of reflections are determined. For this Case is the delay history constant from reflections. By selecting a delay range in the measured data, these reflections can be isolated and investigated. The inventive method makes it possible for the static and dynamic case every single reflection with their characteristics, so Delay, Performance and phase to extract from the measurement data.

What the analysis of reflections with a time-varying delay curve As far as the detection of the delay curve according to the invention is concerned and from the beginning or end of the reflections each individual reflection for themselves be analyzed exactly. So far, especially in dynamic Measurements only statements about Areas of the radio channel are made, to which typically several reflections posts deliver.

What As far as statistical analysis is concerned, it is only through the Inventive detection of individual reflections possible, statistical statements about frequencies of echoes, their duration, "birth" or "death rate", number of reflections, mean performances and delays, To make variations of power and delay and much more. So far this was not possible.

advantageous Embodiments of the method for examining a radio channel according to the invention are in the direct or indirect to the claim 1 referred back Subclaims specified.

Based of graphs that are in the context of a multipath channel measurement resulting measurement results, is the procedure below for examining radio channels of the invention explained. Show it:

1 4 is a diagrammatic section of an impulse response measurement result of a conventional multipath channel measurement in which a test signal was transmitted from a simulated satellite to a user in a densely built up urban area;

2 in a diagram the result after the well-known "Super Resolution" -Prozessierung with the ESPRIT algorithm,

3 in a diagram the result after the well-known "Super Resolution" processing ESPRIT of a snapshot,

4 in a diagram section of 2 the measurement result according to the known "Super Resolution" processing with implementation of the ESPRIT estimation algorithm,

5 in a diagram showing the clipping of 4 corresponds to the measurement result according to the known "Super Resolution" processing using the ESPRIT estimation algorithm and according to the method according to the invention belonging filtering,

6 in a diagram the result of a reflection detection using snapshot search forward and backwards for the first 50 seconds of a test drive,

7 in a diagram, the result after approximation of the reflection chains by polynomials, which reflect the delay curve, and

8th in a diagram showing the clipping of 4 and 5 corresponds to the reflections detected according to the method of the invention together with the ESPRIT result.

In 1 is an example in diagram form Section of a measurement result of the multipath channel measurement shown. In this example, a test signal was sent from a simulated satellite to a user in densely populated urban area. The abscissa of the diagram coordinate system plots the absolute delay, ie the signal propagation time in ns. The ordinate shows the continuation of the measurement in the first 50 seconds. All 10 ms was emitted a pulse-like test signal, so a snapshot of the channel recorded.

The The measured signal thus represents the impulse response of the channel to the test signal. The direct signal is received first, namely after about 5200 ns at the beginning of the measurement, followed by reflections. The received signal energy is according to the right, vertical Bars coded in dB, with the graduation graded in pure gray scale ranges from about -70dB to -25dB and the dotted ones gray tint covers the range from about -25 dB to 0 dB.

2 shows the result of the measurement following a known "Super Resolution" processing with the ESPRIT algorithm. Clearly, individual reflections can now be distinguished from one another. It should be noted that here in 2 the actual additional delay of interest is already plotted on the abscissa, that is to say the delay which goes beyond the delay of the directly propagating signal (LOS signal), which here represents the zero line.

In 3 FIG. 13 is a detailed timing diagram showing the power of a test signal in dB received at the receiver as a function of the delay in μs for a single snapshot. The values marked with a plus sign are the measured samples a _k whose sampling rate corresponds exactly to the reciprocal of the measuring bandwidth B = 100 MHz. By interpolation of the samples one obtains the complex course of the received signal s (t), which in 3 is shown in the form of a solid curve:

Here in is τ the additional delay the relevant reflection of the test signal, ie the delay, the delay t of the directly propagating signal (LOS signal) is added.

The result of "Super Resolution" processing is in 3 represented by the small circles and provides discrete reflections with delay, power and phase. It should be noted that this is a model which optimally approximates the measurement data in the sense of a "maximum likelihood" method. For example, the phase estimate is far too inaccurate for the calculation of a spectrum.

Since noise is present in the measurement and since the ESPRIT algorithm used is an estimation method, noise is of course also present in the result of the "super-resolution" processing. This means that even in the static case the ESPRIT result varies. The estimated delays can vary by a few ns; also the power varies by a few dB and sometimes a single echo in the mathematical sense is approximated by two reflections "more exactly". In order to facilitate the actual reflection detection, according to the invention in the first step, filtering is performed to eliminate ambiguities. As a method of filtering a two-dimensional averaging in an elliptical section, so in both snapshot and. also implemented in the direction of delay, like those in 4 and 5 illustrate illustrated diagram detail views, wherein 4 the unfiltered impulse response treated with the ESPRIT algorithm shows and 5 the filtered impulse response treated with the ESPRIT algorithm.

• – Ausgehend von lokalen Leistungsmaxima wird jeweils eine Suche in Schnappschussrichtung vorwärts sowie rückwärts durchgeführt.
• – Für alle Reflexionen in einem Abschnitt um den aktuellen Punkt wird nun eine "Distanz" aus den gewichteten Abweichungen von Verzögerung, Ableitung der Verzögerung, Leistung und Zeit (Schnappschussrichtung) berechnet.
• – Jener Punkt mit der niedrigsten Distanz wird nun an die aktuelle Reflexionskette angeknüpft.
• – Eine obere Schranke für diese Distanz führt zum Abbruch der Suche und somit zum Ende bzw. Beginn einer Reflexionskette.

In the next step of the method according to the invention, these data are processed using a kind of "minimum distance search" algorithm, which represents the actual reflection detection. This detection was implemented as follows:

• - Based on local performance maxima, a search in the snapshot direction forward and backward is performed.
• - For all reflections in a section around the current point, a "distance" is calculated from the weighted deviations of delay, derivative of deceleration, power and time (snapshot direction).
• - That point with the lowest distance is now linked to the current reflection chain.
• - An upper bound for this distance leads to the termination of the search and thus to the end or beginning of a reflection chain.

In 6 the result of this search is again displayed for the first 50 seconds of the measurement run, ie for 5000 snapshots at a respective interval of 10 ms, where as in 2 and in 3 to 5 on the abscissa the actually interesting, additional delay is plotted, so that delay, which goes beyond the delay of the directly propagating signal (LOS signal), which maps the zero line.

These reflection chains are according to ei nem further step of the method according to the invention now approximated by polynomials that reflect the delay curve of the individual echoes. The diagram which has emerged after this process step has been carried out in the example measured is shown in FIG 7 shown. The polynomial degree is increased differently for each reflection until the mean square error (MSE) is below a certain limit.

In 8th In a detailed view, the detected reflections are now shown together with the ESPRIT result to determine the performance of the method. to demonstrate according to the invention in conjunction with the four-stage minimum distance search processing algorithm for detection.

With Help of the delay course The individual reflections (echoes) can now be determined by interpolation the measured data the signal of each individual reflection from the measured data extract. One is thus able to very precisely track the signal energy and the Doppler spectrum of the thus detected reflections individually to determine. Above all, you can also in particular about frequencies of echoes whose duration, "birth" or "death rate", the number of Reflections, mean powers and delays, variations of power and delay and much more statistical statements are made.

In order to but you are also able to describe the measured channel and to model. For in the future increasingly broadband communication and navigation systems it is necessary to use the radio transmission channel to know with very high accuracy. The method of the invention allows to extract individual reflections and analyze what represents the most accurate way of characterizing a channel. This is not just for simulation purposes to design signals, but also to improve receiver structures unavoidable.

The end DE 197 41 991 C1 known method of MEDAV for determining a direction-resolved complex impulse response of a radio channel is with its software only able to calculate Doppler spectra for discrete constant absolute delays. However, this means that only for reflections with a constant absolute delay in this way, the spectrum and the waveform can be determined approximately. However, such a situation is only present in static measurements in which neither the transmitter nor the receiver and also the potential reflectors do not move.

Claims (5)

A method for examining a radio channel located between a transmitter and a receiver, wherein the transmitter transmits a periodic test signal which is subjected to multipath propagation and thereby divided into a plurality of path-weighted test single signals and which consists of a plurality of blocks having a blocking duration at least as large as the longest term of all to be considered, path-weighted single test signals is received at the receiver at several individual antennas of a receiving antenna array, the superimposed path-weighted test individual signals each as a measurement signal of a measurement channel and these measurement signals are fed to the inputs of a multiplexer in which each of its inputs is turned on at least one block duration to its output, from which the multiplexed measurement signals are fed to a signal and data processing device which inputs the multiplexed measurement signals it converts measured data sets into locally complex impulse responses and as a result calculates a direction-resolved complex impulse response reflecting the properties of the radio channel and which is used in processing the locally complex impulse responses with the aid of a direction estimation algorithm, in particular an ESPRIT method based on a maximum likelihood method. Algorithm, the directions of incidence and complex amplitude values, ie the number of reflections, their delay, power and phase, determined for each detected arrival direction at each sampling time of the radio channel in the manner of a snapshot, characterized in that the result of this calculation for noise and ambiguity suppression band limited and filtering in both the time and delay directions, then by using a four-dimensional minimum distance search algorithm in time, delay, power, and motion directions, chains of Re Flexionen be detected and then approximated by polynomials, so that the result is a very accurate information about the delay curve of each reflection in the observation period and their beginning and end, and that using this information by interpolation of the measured data, the waveform and the spectrum for each Reflection can be extracted from the measured data. A method according to claim 1, characterized in that on the basis of the information about the delay profile of each reflection as well as the beginning and end statistical statements in particular frequencies of reflections, duration of reflections, "birth" or "death rates", number of reflections, mean performances and delays, variations of performance and delays tion of reflections. Method according to claim 1, characterized in that that for filtering a two-dimensional averaging in one elliptical section, so both in snapshot (time) - as also in the direction of delay, carried out becomes. Method according to claim 1, characterized in that that the detection by means of the four-dimensional "Minimum Distance Search "algorithm in time, delay, Power and movement direction is that first, starting of local performance maxima, one search in each snapshot direction (Time direction) forward and backwards, that then for all reflections in a section around the current point are a "distance" from the weighted ones Deviations from delay, Derivation of the delay, Performance and snapshot direction (time direction) is calculated, that then the point with the lowest "distance" is linked to the current reflection chain and finally an upper barrier for this "distance" to the demolition of Search and thus leads to the end or beginning of a reflection chain. Method according to claim 1, characterized in that that in the approximation of the reflection chains by polynomials, which the delay course reflect the polynomial grade for each Reflection is increased differently until the middle square Mean Square Error (MSE) under a certain bound lies.