EP1807711A2

EP1807711A2 - Method and device for receiving a degraded radionavigation signal

Info

Publication number: EP1807711A2
Application number: EP05797254A
Authority: EP
Inventors: Jean-Pierre Raffegeau
Original assignee: Thales SA
Current assignee: Mitac International Corp
Priority date: 2004-10-15
Filing date: 2005-10-11
Publication date: 2007-07-18
Also published as: WO2006040325A2; FR2876845B1; FR2876845A1; US20090052505A1; US7826518B2; WO2006040325A3

Abstract

The invention relates to a method for receiving a radionavigation signal, comprising a received sampled signal correlation calculation stage for a given first time period with a T period code which is locally generated. The received signal and the code are offset in relation to each other from one correlation to the other. The inventive method is characterized in that during the time period thus determined, the signal is offset from one correlation to another, the code being fixed.

Description

METHOD AND DEVICE FOR RECEIVING A DEGRADE RADIONAVIGATION SIGNAL

The invention relates to the reception of a radionavigation signal from a satellite positioning system such as the GPS system (acronym for the English expression "Global Positioning System").

To function well, current receivers usually require direct satellite reception. The positioning deteriorates rapidly in particular in terms of accuracy and acquisition time, when the reception is disturbed as is the case inside a building or more generally in a degraded environment.

It briefly recalls the operation of the GPS system. It consists of a constellation of 28 satellites and a terrestrial network of ground-based reference stations. Each satellite gravitates 20 000 km from the earth with a period of revolution of 12 hours. They emit two signals, one at 1575.452 MHz for civil applications and the other at 1227.6 MHz for reserved access applications. In the following, we only consider the civil frequency. The signal emitted by a satellite consists of a 1575.452 MHz frequency carrier, modulated by a known spreading code and possibly by unknown data also known as databits. The satellites all transmit on the same frequency and the emitted signals are differentiated by their code.

These codes have a period T, for example 1 ms and typically consist of 1023 chips.

The positioning of the receiver is obtained by measuring the distance between a satellite and the receiver from the propagation time of the signal between this satellite and the receiver. In the receiver, a replica of the transmitted code is generated locally; the offset between the received signal and the local signal (i.e. the replica) corresponds to the desired propagation time. This offset is measured by phasing the received signal and the local signal; the phasing criterion corresponds to the maximization of the correlation function of the two signals, that is to say to the search for a peak of correlation.

This correlation calculation is generally carried out from half-code chip to half-code chip over an integration interval that can be vary. For an integration interval of 1 ms, a correlation computation time of about 2s is obtained. (2 x 1023 x 1 ms = 2s). This calculation is multiplied by a factor K depending on the drift of the local clock (or driver) of the receiver and the number of assumptions on the frequency of the signal to be considered to compensate for the Doppler effect. For a clock uncertainty of about ± 10 kHz, K = 21; we then obtain a computation time of approximately 2s x 21, ie 42s per code, that is to say by satellite.

In the case of a reception in degraded medium, a solution to obtain a better signal-to-noise ratio is to increase this interval. Taking an interval of 10ms, we gain about 10 dB, but the calculation then lasts about 20s multiplied by a factor K ≡201, or about 4020 s per satellite. We finally won only 10 dB for a positioning time that has increased considerably.

One solution for reducing this computation time is to perform 2 x 1023 multiplications in parallel instead of calculating them in series as before. We then obtain a computation time on an integration interval of 10 ms of about 2s (201 x 10 ms = 2s).

This result is obtained by using for example a correlator structure known as the "Matched Filter": the local code is moved in front of the signal received at each correlation. This structure actually comprises two identical substructures in parallel, one for the real component of the sampled signal, the other for the imaginary component.

In the following, the sampled signal is the received signal converted into baseband and sampled at N kHz, where N is the number of code chips considered over the period of the code. Here, N = 2046 = 2 x 1023. The operation of one of these substructures will be described in relation with FIG.

The received signal whose frequency is compensated to take into account the Doppler effect, is loaded into a non-rotating shift register 1 and the code is loaded into a rotating shift register 2. The length N of the shift register represents the number code chips considered over the period of the code, ie 1 ms for example for the GPS. To fully load the signal into its shift register 1, it is necessary to wait N clock cycles. This first step of loading register is represented on the timing diagram of Figure 2, by the designation "shift register".

Then the signal remains fixed in its shift register 1 and the code performs N rotations on itself in the rotating shift register 2. A new correlation is calculated for each rotation, by means of N multipliers 3 and an adder 4 at N entries. Full code rotation requires N rotations, ie N clock cycles. This second step is represented on the chronogram by the designation "Calculation of correlations". The two steps therefore require 2N clock cycles. They have the same duration. The timing diagram represents the treatment over 4 code periods, for a component, a satellite and a frequency hypothesis to compensate for the Doppler effect.

To obtain a real-time result, the code must be rotated about 2046 times the sampling frequency, which is difficult to implement and requires a large number of logic gates.

An important object of the invention is therefore to optimize the signal processing time to allow to increase the integration times and therefore the sensitivity of the receiver, especially when the reception is degraded.

To achieve this object, the invention proposes a method of receiving a radionavigation signal which comprises, at a first determined time, a correlation calculation step of the sampled received signal with a locally generated period code T, the received signal and the code being shifted with respect to each other from a correlation to the next correlation; the method is mainly characterized in that, over the determined time, the signal is shifted from one correlation to the next, the code being fixed. This determined time is typically equal to T. One of the advantages of this method is that it makes it possible to parallel two phases of the conventional solutions (input of the signal in a shift register and calculation of the correlations). The time of the signal processing, therefore search satellites is divided by 2.

This method has another advantage which appears when the signal includes data (also called data bits) in addition to the code. These data are transmitted by the satellite for example every 20 ms, that is to say at 50 Hz. When these data change their value (for example, from +1 to -1 or the opposite), this translates into by a phase inversion in the signal. This phase inversion can occur during the integration interval, when the signal is fixed and the code is rotating on this interval. In this case, calculating a correlation peak over this interval is difficult if not impossible. When according to the invention, the signal is sliding and the code is fixed on this interval, the phase inversion is necessarily synchronous from the beginning of the integration interval because it is synchronous with the code which is fixed: the inversion of phase then no longer affects the correlations calculated during a code period.

According to one characteristic of the invention, the sampled signal comprising a real component I and an imaginary component Q, the samples of the two components I and Q are put in series alternately in the same memory and the correlation calculations of the code with I and with Q are performed alternately.

According to another characteristic of the invention, the results of correlation calculations of the same component and having the same offset n between the signal and the code, and obtained over Tc code periods, are added so as to obtain for each component a coherent integration respectively designated Iτc (n) and Qτc (n).

Preferably, a sum of the amplitudes or powers of the coherent integrations having the same offset n, is computed over a time Tnc (Tnc = K × Tc) so as to obtain for each shift n, a non-coherent integration C (n).

According to one embodiment of the invention, prior to the correlation calculation step, the method comprises a step of storing the signal in another memory during a second determined time, according to a write frequency fe and the signal is read at a reading frequency fl such that fl> fe.

With this embodiment designated post-processing mode, the signal is sampled and stored in memory preferably during the noncoherent time Tnc which can reach 16s; then the same samples are read into memory for each satellite and each frequency assumption. The samples are then read at a much higher frequency fl at their write frequency in memory fe. This post-processing mode makes it possible to reduce the overall search time of the satellites.

The subject of the invention is also a receiver of a radionavigation signal comprising means for implementing the method.

According to one characteristic of the invention, it is provided with a correlator capable of correlating the received signal with a code, and the correlator comprises a non-rotating shift register for the code and at least one non-rotating shift register for the signal. . Preferably, the non-rotating shift register for the code comprises N cells, where N is the length of the code, the non-rotating shift register for the signal comprises 2N cells and it comprises N multipliers each connected on the one hand to a cell of the register of the code and on the other hand to a cell of the signal register and an adder with N inputs, each input being connected to a multiplier.

Thus, the number of gates for the signal and code shift registers remains unchanged. On the other hand, the structure has only one adder and one set of elementary multipliers. The silicon surface is practically divided by a factor of 2 and the cost of the circuit is reduced.

According to one embodiment of the invention, it further comprises connected to the correlator means capable of storing the 2N results from the correlator and incrementing them respectively as and when the results provided by the correlator for a duration Tc to achieve 2N consistent integrations.

Advantageously, it comprises means able to memorize the N amplitudes or powers of coherent integrations, and to increment them respectively during a duration Tnc, as the amplitudes or the powers of coherent integrations are obtained, to realize N inconsistent integrations.

Other features and advantages of the invention will appear on reading the detailed description which follows, given by way of non-limiting example and with reference to the appended drawings in which: FIG. 1, already described, schematically represents a known correlator of the matched filter type for an I or Q component of the signal, FIG. 2 already described schematically represents a timing diagram of the processing of the signal received with rotation of the code, FIG. example of a correlator according to the invention of the Matched Filter type for an I or Q component of the signal, FIG. 4 schematically represents a timing diagram of the processing of the signal received with signal rotation, FIG. 5 schematically represents another example of a correlator according to FIG. the invention of "matched filter" type in which the two substructures are combined, FIG. 6 schematically represents an optimization example of the example of FIG. 5, FIG. 7 schematically represents an example of a correlator according to FIG. invention associated with means for mixing coherent integrations and non-coherent integrations your.

The principle of finding a correlation peak lies in a series of correlation tests between the signal and the code by modifying the offset of one with respect to the other. Conventional solutions have opted for code shifting, as shown in the preamble. This technique proceeds in 2 phases. The first phase loads the shift register with the signal over a period of code. The second phase calculates a correlation with each rotation of the code on itself.

The invention is based on the same principle of correlation but adopts the shift of the signal. From one correlation to another, the code is now fixed and only the signal is shifted: it slides in its shift register at each clock cycle. The slice of signal is always equal to a period of code but this slice is slippery.

This correlation method is for example obtained by means of an improved "Matched Filter" type structure.

According to a first embodiment, it is in fact two identical substructures in parallel, one for the real component of the signal, the other for the imaginary component. We will describe the operation of one of these substructures in relation to the example of Figure 3. From one figure to another, the same elements have the same references.

The code is entered in a non-rotating shift register 2 'and the component of the compensated signal of its Doppler is continuously fed into a non-rotating shift register 1. The length N of the shift register represents the number of code chips considered over the period of the code, ie 1 ms for example for the GPS. To calculate all the correlations, it is necessary to wait N clock cycles. At each clock pulse, the component of the signal is shifted into shift register 1 and a new correlation is calculated. After N offsets the signal, all the correlations over the period of the code are calculated.

One of the advantages of this correlator lies in the parallelization of the two phases of the conventional solutions (loading of the signal in the shift register and calculation of the correlations). The signal processing time, and thus the search time of the satellites, is thus divided by 2, as illustrated in the timing diagram of FIG. 4 which represents the processing of 8 code periods for a component, a frequency hypothesis to compensate the Doppler effect. and a satellite. According to another embodiment, the two structures for the real and imaginary components are combined as shown in the correlator example of FIG. 5.

This correlator 20 is provided with a non-rotating shift register 2 'for the code, common to the two components, of length N. The other elements were already present in the example of FIG. 3, namely two shift registers 1 of length N (one for each real and imaginary component of the signal), 2N multipliers 3 (N for each component) and 2 adders 4 to N entered each (1 adder for each component). An adder requires a very large number of logic gates and constitutes the main silicon surface of the invention.

According to a preferred embodiment, the samples of the two real and imaginary components of the signal are serially stored alternately in order to limit the number of doors required. The number of gates for the signal and code shift registers remains unchanged. On the other hand, the structure has only one adder and one set of elementary multipliers. The silicon surface is practically divided by a factor of 2 and the cost of the circuit is reduced. An example of such a correlator 20 is shown in FIG. 6. It comprises a non-rotating shift register 2 ', of length N for the code, a non-rotating shift register 1', of length 2N for the signal, N multipliers 3 and an adder 4 with N inputs.

The adder 4 makes it possible to obtain the correlations for a component, real or imaginary. Then, at the next clock cycle, the signal slides in the register 1 by shifting one cell and the adder 4 then makes it possible to obtain the correlations for the other component and so on as the slip progresses. of the signal.

We will now detail on an example the calculations for N = 2046 = 2 x 1023 because the correlation calculation is performed half-chip code in half-chip code. The code chips can be divided by 2 as in this example or by 3, 4, ..., even 3.3, etc.

The two real and imaginary components of the signal are alternately put in series in the shift register and can for example be expressed in the following way:

SR (2i, tk, _n ) are the samples of the real component of the signal at the instant ix 1 ms / 2046 + t _k , _n ,

SR (2i + 1, tk, _n ) are the samples of the imaginary component of the signal at the instant ix 1 ms / 2046 + t _k , _n ; i the index of the shift register varies from 0 to 2045, tk, n = t ₀ + kx 1ms + nx 1ms / 2046. t _k , _n represents the (k + 1) th code period, n varying from 0 to 2045 and represents the delay index between the signal and the code.

At the k + 1th code period, the result of the correlation on the real component, provided by the adder for an offset n, is:

Code (0) * SR (O, t _k J + Code (l) * SR (2, t _kn ) +

... + Code (2044) * SR (4088, t _kn ) + Code (2045) * SR (4090, t _kn ) Similarly, at the k + 1th code period, the result of the correlation on the imaginary component, provided by the adder at the following clock cycle, for an offset n is:

Code (0) * SR (l, t _kn ) + Code (l) * SR (3, t _{k; n} ) +

... + Code (2044) * SR (4089, t _kn ) + Code (2045) * SR (4091, t _kn )

The signal generally includes data (also called data bits) in addition to the code. These data are transmitted by the satellite for example every 20 ms, that is to say at 50 Hz. When these data change their value (for example from +1 to -1, or vice versa), this is translated by a phase inversion in the signal. This phase inversion can occur during the integration interval, when the signal is fixed and the code is rotating on this interval. In this case, calculating a correlation peak over this interval is difficult if not impossible. When according to the invention, the signal is sliding and the code is fixed on this interval, the phase inversion is necessarily synchronous from the beginning of the integration interval because it is synchronous with the code which is fixed: the inversion of phase then no longer affects the correlations calculated during a code period. When the reception is degraded, the computation of a peak of correlation over a single code period is not possible: this computation is carried out by integrating over several periods of code, the correlations of the same component (real or imaginary) and of same offset between the signal and the code. It is possible to carry out so-called coherent integration over a coherent time Tc or a so-called non-coherent integration over a non-coherent time Tnc.

If a phase inversion due to a data bit change occurs during the coherent time, it may be difficult or impossible to detect the correlation peak. The coherent integration requires taking into account the phase of the data bits by means of a known device 5 represented in FIG. 7: the result of each correlation is multiplied by ± 1 according to the value of the databit on the code period during which the correlations are made. The characteristic of coherent integration is to keep the phase information of the signal. Integration consistent with the consideration of the phase of the data bits is the most efficient signal processing in terms of detection sensitivity. An important coherent time makes it possible to have a narrow bandwidth and to bring out the signal of the noise. On the other hand, it is not possible to greatly increase the coherent time. This one is very quickly limited by: the number of assumptions of frequencies to be considered to compensate the Doppler effect, the potential acceleration of the mobile, the characteristics and the price of the pilot oscillator in the receiver.

Because of these limitations, consistent integration is less than

100ms and more generally set at 20ms, that is to say at the time of the databits. The reception sensitivity for a coherent time of 20 ms is located around -144 dBm according to the probabilities of absence and false detections of the radionavigation signal.

According to a preferred embodiment of the invention, a coherent integration is mixed over a time Tc and a non-coherent integration over a time Tnc. This mixing is implemented for example by means shown in FIG.

The process according to the invention proceeds in 3 steps. The first step calculates a coherent integration over a time Tc with its 2 real and imaginary components. The second step calculates an amplitude or a power from the 2 components. The third step calculates over a time Tnc the sum of the amplitudes or powers having the same offset. The 3 steps are repeated N times. In our example N = 2046.

The coherent integration (1 ^st step) is performed during a time Tc, multiple of the period of the code, in the multiple occurrence of 1 ms; it is performed separately on both the real and imaginary components. For an offset n between the signal and the code, it follows from the coherent integration, l (n, Tc) for the real component and Q (n, Tc) for the imaginary component: _{l2045 =} ^ | Code (0) * SR (0, t _M ) + Code (l) * SR (2, t _M ) +

"= ⁰ ^{(n> c)} Ij ^ _{+ Code} (2044) * SR (4088, t _kn) + Code (2045) * SR (4090, t _k J

₂₀₄₅ _ ^τ ^ ( ^Cod <°) * ^8R (M _M ) + Code (l) * SR (3, t _k , _n ) +

Q _n ^{= O} ( ⁿ , Tc) - Σ ... + Code (2044) * SR (4089, t _kn ) + Code (2045) * SR (4091, t _kn )

with tk, _n = to + kx 1 ms + nx 1 ms / 2046.

This coherent integration can be obtained by means of a microprocessor or by means of an adder 6 associated with a memory 7 to 2N boxes able to be read and written such as a RAM memory. In known manner, the memory 7 is read and written as and when the results provided by the correlator 20. An amplitude or a power is then calculated from the

2 components of coherent integrations (2nd step):

Vl (n, Tc) ² + Q (n, Tc) ² or [l (n, Tc) ² + Q (n, Tc) ² ]

This second step is obtained in known manner by software means or a device 8.

In the prior art, the non-coherent integration consists of calculating over the period of the code, the amplitude or the power of the correlations on the two components, and then summing them. According to the invention, the non-coherent integration (3rd step) is calculated by performing the sum over a time Tnc multiple of Tc (Tnc = K × Tc), coherent amplitudes or powers of integrations having the same offset. We obtain :

[l (n, kTc) ² + Q (n, kTc) ² ]

This non-coherent integration can be obtained by means of a microprocessor or by means of an adder 9 associated with a memory 10 with N boxes able to be read and written such as a RAM. In known manner, the memory 10 is read and then written as the results provided by the memory 7. These three steps are repeated for each shift n, that is to say N times. In our example N = 2046.

According to one embodiment of the invention, designated real-time mode, the signal is sampled and processed for a satellite and a frequency assumption (considered to compensate for the Doppler effect) at a time. There is no memorization of the signal: the samples are directly used for the processing of the signal which lasts Tnc.

According to a preferred mode designated post-processing mode, the signal is sampled and stored in memory 11 preferably during the noncoherent time Tnc which can reach 16 s. Then the samples are read into memory for each satellite and each Doppler frequency hypothesis; that is, the I and Q signals from the memory 11 are converted to baseband and compensated for a Doppler frequency by the element 12 before being inserted into the shift register 1 of the correlator 20. The samples are read at a frequency f1 much higher than their write frequency in memory fe. This post-processing mode makes it possible to reduce the overall search time of the satellites as it appears on the following comparison. We design :

NBSAT: the number of satellites

BPDOPPLER: the frequency band to consider to compensate for the effect

Doppler for satellite search

Tnc: the non-coherent integration time Tc: the coherent integration time

K: the ratio of writing and reading frequencies in post-processing

T tr: Real time global processing time

Tpt: Total treatment time post-treatment We have:

Ttr = NBSAT x BPDOPPLER x Tc x Tnc

Tpt = NBSAT x BPDOPPLER x Tc x Tnc / K + Tnc,

Tpt = Ttr / K + Tnc Tnc is negligible compared to Ttr / K; typically K = 25 or more. In the example described, two-state signals (± 1) were considered; of course, the invention applies in the same way to signals with more than two states. So far it has been considered that only one code is generated locally. When it is desired to consider other codes, these can be considered in series by reprogramming the local code and the associated Doppler frequencies, as one changes code.

They can be considered in parallel: it is then sufficient, according to a method known to those skilled in the art, to provide the receiver with parallel structures, that is to say to multiply the structure comprising the correlator (as well as the memory for post-processing if necessary, and / or integrators if necessary), by the number of local codes to consider.

The received signal may be a satellite and / or pseudolite radionavigation signal.

This is for example a positioning signal type GPS (acronym for the English expression "Global Positioning System"), or GALILEO or other positioning system.

Claims

A method of receiving a radionavigation signal which comprises at a first determined time a correlation calculation step of the sampled received signal with a locally generated period code T, the received signal and the code being shifted relative to one another. to the other of a correlation to the following correlation, characterized in that, on the determined time, the signal is shifted from one correlation to the next, the code being fixed.

2. Method according to the preceding claim, characterized in that the sampled signal comprising a real component I and an imaginary component Q, the samples of the two components I and Q are alternately placed in series in the same memory and in that the calculations of Correlation of the code with I and with Q are performed alternately.

3. Method according to one of the preceding claims, characterized in that the determined first time is equal to the period of the code T.

4. Method according to one of the preceding claims, characterized in that the sampled signal comprising a real component I and an imaginary component Q, the results of correlation calculations of the same component and having the same offset n between the signal and the code , and obtained over Tc code periods, are added so as to obtain for each component, a coherent integration respectively designated Iτc (n) and Qτc (n).

5. Method according to the preceding claim, characterized in that Iτc (n) ² + Qτc (n) ² is calculated.

6. Method according to the preceding claim, characterized in that a sum of the amplitudes or powers of coherent integrations having the same offset n is computed over a time Tnc (Tnc = K x Tc) of to obtain for each shift n, a non-coherent integration C (n).

7. Method according to one of the preceding claims, characterized in that prior to the correlation calculation step, it comprises a step of storing the signal in another memory during a second determined time and according to a writing frequency fe and in that the signal is read at a read frequency f1 such that f1 for the correlation calculation step.

8. Method according to the preceding claim, characterized in that the correlation calculation step is performed for several codes and in that the signal is read for each code.

9. Method according to one of claims 7 or 8, characterized in that the correlation calculation step is performed for several frequency hypotheses considered to compensate for the Doppler effect and in that the signal is read for each of these hypotheses.

10. Method according to one of claims 7 to 9, characterized in that the second determined time is equal to Tnc.

11. Method according to one of the preceding claims, characterized in that the signal received is a satellite navigation signal and / or pseudolites.

12. Receiver of a radionavigation signal comprising means for implementing the method according to one of the preceding claims.

Receiver according to the preceding claim, characterized in that it is provided with a correlator (20) capable of correlating the received signal with a code, and in that the correlator comprises a non-rotating shift register (2 '). for the code and at least one non-rotating shift register (1 or the) for the signal.

Receiver according to the preceding claim, characterized in that the non-rotating shift register (2 ') for the code comprises N cells, N being the length of the code, the non-rotating shift register (1') for the signal comprises 2N cells and in that it comprises N multipliers (3) each connected on the one hand to a cell of the register (2 ') of the code and on the other hand to a cell of the register (1') of the signal and an adder (4) N inputs, each input being connected to a multiplier.

15. Receiver according to one of claims 13 or 14, characterized in that it further comprises connected to the correlator (20), means (7, 6) able to store the 2N results from the correlator and increment respectively as the results provided by the correlator (20), for a duration Tc, to achieve 2N coherent integrations.

16. Receiver according to the preceding claim, characterized in that it comprises means (8) for obtaining N amplitudes or coherent integrating powers.

17. Receiver according to the preceding claim, characterized in that it comprises means (10, 9) capable of storing the N amplitudes or powers of coherent integrations, and to increment them respectively for a duration Tnc, as and when that the amplitudes or the powers of coherent integrations are obtained, to realize N non-coherent integrations.

18. Receiver according to one of claims 13 to 17, characterized in that it further comprises connected to the correlator (20) another memory (11) adapted to store the received signal at a write frequency fe and the restore to the correlator according to a reading frequency fl such that fl> fe.