Patent Grant
7205935
The present invention is related to a method and apparatus for receiving and processing signals, essentially signals of the spread spectrum type, such as are used in the Global Positioning System (GPS).
Today, satellite based positioning systems are widely in use. Most known are the Global Positioning System (GPS) developed by the United States Government or the Global Navigation Satellite System (GLONASS) developed by the Russian Government. The primary purpose of these systems is to provide position, velocity and time to any user on or near the Earth's surface. The user determines his position in space and time by measuring his range to at least 4 satellites whose position and time are accurately determined, the range being defined as the distance between the user and a satellite.
The GPS satellites transmit signals on two carrier frequencies called L1 at 1575.42 MHz and L2 at 1227.6 MHz. The carriers are modulated by pseudo-random (PRN) spreading codes that are unique to each satellite, and by a navigation message. All satellites transmit at the same frequencies. The separation between the satellites is possible because all the PRN codes are chosen to be orthogonal.
The L1 carrier is modulated by the so-called coarse/acquisition code (“C/A-code”) and the precision code (“P-code”). The C/A-code has a chipping rate of 1.023 MHz and a length of 1 ms, thus it contains 1023 chips (the term “chip” is used for the code symbols). The same 1023 chips are repeated every millisecond. The P-code has a chipping rate of 10.23 MHz and a length of 1 week. The same pattern is thus repeated every week. The L2 carrier is modulated only by the P-code. The signal bandwidth at L1 and L2 is limited in the satellite to the main lobe of the P-code spectrum, i.e. 20.46 MHz.
In addition to the PRN code modulation, each of the carriers is also modulated by a 50-Hz navigation message, conveying all the necessary parameters to compute the satellite position and time.
Signal tracking involves synchronising local replicas of the carrier and PRN codes of the satellite to be tracked with the incoming carrier and PRN codes from that satellite. This means that the receiver must have the ability to generate the PRN codes of all the satellites. Signal tracking is achieved by continuously modifying the phase and frequency of the carrier replica and delaying or advancing the code replica in order to maintain them locked on the incoming signal. Carrier tracking is accomplished in a phase lock loop (PLL), while code tracking is performed by a delay lock loop (DLL). By definition, when synchronisation is achieved, the phase of the carrier replica is the same as the incoming carrier phase, and the delay that had to be applied to the code replica is the same as that of the incoming code.
The range measurement by a GPS receiver is based on the measurement of that delay, which is a direct measure of the signal travel time from the satellite to the receiver. By multiplying the delay by the speed of light, the receiver computes its range to the satellite.
Several error sources influence the accuracy of the satellite range measurement. One of the most cumbersome of them is the multipath error. Multipath is a special type of interference where the received signal is composed of the desired line-of-sight signal, and one or more constituents which have traversed slightly different paths due to reflections on surfaces or objects in the antenna surroundings. Multipath signals arrive at the receiver with a different delay, phase and power than the line-of-sight signal.
The ranging error due to multipath depends on the delay, phase and power of the multipath signal with respect to the line-of-sight signal, and on the type of signal processing the receiver uses. Several digital processing techniques have been devised to reduce the effect of multipath on the ranging measurements.
However, a common drawback of the available multipath mitigation methods is that they are not able to reduce multipath having a short delay. Even for the most recent multipath error mitigation techniques, multipath arriving at the receiver with a delay of less than around 20 m affect the range measurement as if no mitigation technique were used at all. This is a serious limitation of the current techniques because in real life situations, most of the multipath signals are of short delay type.
Document U.S. Pat. No. 5,390,207 is related to a receiver for pseudorandom noise encoded signals. The operation of the device is based on narrowing the early-late spacing, to a fraction of a PRN code chip time, in order to mitigate multipath errors. The use of early-late correlators is a known technique for locking onto an incoming PRN code, and the early-late spacing, i.e. the time spacing between early and late versions of the local PRN code is an important parameter when it comes to multipath error mitigation. The exact definition of this term is given in the next paragraph.
Document U.S. Pat. No. 5,734,674 is equally related to a receiver for PRN encoded signals; the receiver described in this document has the capability of dynamically adjusting the early-late spacing between correlation signals.
Document U.S. Pat. No. 5,809,064 is related to improvements to the same type of receiver as described in U.S. Pat. No. 5,734,674, and equally based on dynamically adjusting early-late spacing.
All techniques related to adjusting and essentially reducing the early-late spacing are linked to a modification of the DLL discriminator. A general name for these techniques is ‘narrow spacing’, as opposed to techniques wherein no multipath error mitigation takes place (‘wide spacing’ techniques). It is found that prior art ‘narrow spacing’ techniques reduce the maximum range error with a factor 10 compared to ‘wide spacing’ techniques. However, an important multipath error remains for a wide range of multipath delays.
Document U.S. Pat. No. 5,414,729 is related to a receiver for PRN encoded signals, wherein a plurality of correlation signals are fed to a parameter estimator, from which the delay, amplitude and phase parameters of the direct path signal, as well as any multipath signals, may be estimated. This estimation however takes place by solving a system of equations, for example through a least-squares calculation, which requires complex and expensive hardware and software capabilities.
Document U.S. Pat. No. 5,953,367 is related to a receiver for PRN encoded signals, comprising a plurality of DLL correlators provided in each of the receiver's multiple processing channels. These correlators are combined in such a way as to build a DLL discriminator which is not affected by most of the multipath errors. This improves the tracking in the presence of multipath, but requires a significant modification of the DLL architecture, in comparison to wide spacing and narrow spacing DLL's.
Document U.S. Pat. No. 5,901,183 is related to a receiver for PRN encoded signals, wherein both the receiver DLL code and PLL carrier loops include a loop component that senses an error in its main loop caused by the presence of a multipath signal. A compound correlator loop is formed of a primary correlator and a secondary correlator that senses and corrects for multipath induced tracking error of the primary correlator.
Document U.S. Pat. No. 5,781,152 is related to a method wherein four pseudo random code replicas are used, namely a first set of early and late codes, with a first early-late spacing 2 d, and a second set with a second early-late spacing 2 kd. The multipath error is estimated from an extrapolation on the basis of correlation values measured with these two sets of codes.
Document U.S. Pat. No. 5,966,403 is related to a method and apparatus for minimizing the residual multipath distortion signal present in a received signal, by using a non-uniform weighting function for said estimation/minimization.
An overview of the capabilities and limitations of prior art methods is given in
The present invention aims to provide a method and a device for receiving and processing signals of the spread spectrum type, allowing a reduction of multipath errors in the lowest range of multipath delays, said reduction being in comparison to the closest prior art.
The present invention further aims to propose a method and device which are technically straightforward and inexpensive compared to existing methods and devices.
The present invention is related to a method for processing at least one signal sent by a transmitter, said signal preferably being used for measuring the range, i.e. the distance between said transmitter and a receiver, said signal comprising a carrier signal modulated by a pseudo random noise (PRN) code, said method comprising the steps of:
According to the preferred embodiment, the method of the invention comprises the steps of:
In the method of the invention, said linear combination preferably has of the following form:
wherein MP represents the multipath error, d represents the early-late spacing, I0 represents the correlation value of said punctual replica, I−M, . . . , I+N represent the correlation values, αi represent M+N+1 fixed values.
Preferably, every one of said M+N+1 replica's is shifted over the same time delay with respect to the next and/or previous replica.
Further according to a preferred embodiment, said αi values are calculated according to the method comprising the steps of:
In this method, all except two of said αi values may be set to zero.
According to a further preferred embodiment, said αi values are calculated according to the method comprising the steps of:
The vector α may be overdetermined by said system of equations in which case said vector α is obtained by an optimization technique.
According to a particular embodiment, two replica's are used in the estimation of the multipath error, and the early-late spacing d is 1/15 of a chip length, and the second replica P+2 is 1/15 of a chip length later than said punctual replica P0, and said multipath error estimation MP is calculated as:
The invention is equally related to a receiver for ranging applications, said receiver comprising a plurality of channels for detecting and locking onto a plurality of PRN encoded signals, each channel comprising:
In a receiver according to the invention, said formula preferably has the form:
wherein MP represents the multipath error, d represents the early-late spacing, I0 represents the correlation value of said punctual replica, I−M, . . . , I+N represent the correlation values, αi represent M+N+1 fixed values.
In a receiver according to the invention, said αi values are preferably calculated according to the method comprising the steps of:
In this method, all except two of said αi values may be set to zero.
According to a further preferred embodiment of the receiver of the invention, said αi values are calculated according to the method comprising the steps of:
The vector α may be overdetermined by said system of equations, in which case said vector α is obtained by an optimization technique.
According to a preferred embodiment of the receiver of the invention, two replica's are used in the estimation of the multipath error, and the early-late spacing d is 1/15 of a chip length, and the second replica P+2 is 1/15 of a chip length later than said punctual replica P0, and said multipath error estimation MP is calculated as:
In a receiver according to the invention, said multipath estimator module may comprise software means for calculating the multipath error estimate on the basis of a predefined formula.
Alternatively, in a receiver according to the invention, said multipath estimator module may comprise hardware means for calculating the multipath error estimate on the basis of a predefined formula.
The invention is equally related to a method for estimating a ranging error due to multipath in a receiver, said receiver comprising:
wherein MP represents the multipath error, d represents the early-late spacing, I0 represents the correlation value of said punctual replica, I−M, . . . , I+N represent the correlation values, αi represent M+N+1 fixed values, said method comprising the steps of:
In this method, all except two of said αi values may be set to zero.
According to a further embodiment, said method comprises the steps of:
a and 3b illustrate the effect of multipath errors on the tracking point in devices of the prior art.
a and 6b illustrate the principle of the multipath estimator module.
A typical implementation of the signal processing technique according to the prior art for ranging applications is represented in
The delay lock loop is represented by the group of elements 4. As known in the art, an acquisition step is first performed in order to acquire a ‘coarse’ match between the incoming signal and a local version of the PRN code. After that, the local code needs to be locked onto the incoming signal. The present invention is related to this locking step.
Referring again to
For the GPS C/A-code, the code chip duration is close to 1 μs, and the chip length close to 293 m. The delay between the early and late taps is an important design parameter, which is referred to as the early-late spacing, and noted ‘d’. Many receivers use a so-called “Wide spacing” of d=1 code chip. The spacing is said to be narrow if d is lower than 1 chip.
Returning to
The role of the DLL 4 is to keep the punctual code P0 aligned with the incoming code. It is apparent from the
This implementation gives the correct tracking point in absence of multipath. However, when multipath is present, the correlations represented in
It can be seen that the tracking point, defined by the equality I+1=I−1, is no longer positioned at Δτ=0. This means that the DLL is not able to correctly align the punctual code and the incoming code. There is a tracking error, which in
The early-late spacing has a large influence on the sensitivity of the DLL to multipath signals. For example, the envelope noted “Narrow spacing” (curve 101) in
The state-of-the-art techniques to reduce the range errors caused by multipath rely on a modification of the DLL so that the deviation of the tracking point with respect to the Δτ=0 condition is kept as small as possible in presence of multipath. As discussed above, many prior art techniques are based on a reduction and/or a dynamic adjustment of early-late spacing d, in order to mitigate multipath errors. Others employ elaborate calculation techniques.
The present invention uses a different approach, based on a known fact, namely that the signal amplitude measurement (or equivalently the carrier-to-noise ratio measurement) reported by a receiver is highly correlated with the range error caused by multipath. This is disclosed in the document “Multipath Mitigation, Benefits from Using the Signal-to-Noise Ratio”, J. M. Sleewaegen, Proceedings of the ION GPS-97 Meeting, pp. 531–540, 1997. One of the key properties of the signal amplitude measurement is that it is most sensitive to short multipath delays. The present invention makes use of this property to derive a multipath error estimator that is operating even for short multipath delays.
In the invention, a conventional narrow spacing DLL is used for the tracking, i.e. a DLL with a low early-late spacing, for example d= 1/15 of a chip length. Tracking is thus performed in the classic way, i.e. by keeping the early and late correlation values equal to each other (see
According to the general case presented in
As in the prior art receiver, a carrier generator 2 and code generator 3 are used, and a plurality of mixers 7 and accumulators 8. All these elements in themselves are identical to the ones used in the prior art receivers. For the tracking, a conventional DLL discriminator and filter 10 is used, to perform tracking in the classic way, namely based on one early (P−1), one punctual (P0) and one late PRN version (P+1). Independently, all the M+N+1 correlation values (I−M, . . . , I+N) are fed into the multipath estimator module 23 at each calculation of the range. This module derives M+N+1 independent estimates of the signal amplitude by scaling each of the M+N+1 correlation values by 1/(1−|i|d/2), i=(−M, . . . , N). When multipath is present, each of these signal amplitude estimates exhibits an error which is highly correlated with the range error, although it is different for each estimate. The invention lies in taking advantage of these differences to build a multipath range error estimator. More specifically, it will be shown that an appropriate linear combination of the M+N+1 signal amplitude estimates, normalized by the punctual correlation value I0 closely matches the ranging error due to multipath. In other words, the range error due to multipath may be closely approximated by the following formula:
This estimation of the multipath error MP is characteristic to the invention. The coefficients αi are constant. They are computed only once, during the design of the receiver. After the calculation according to the predefined formula, the noise on the estimation (MP) is filtered out by a low pass filter 24. The result is then subtracted from the range derived from the early-late tracking by the DLL, resulting in largely removing the multipath error from the range measurement.
This predefined formula allows a very fast estimation of the multipath error, compared to existing techniques, in particular compared to the technique presented in U.S. Pat. No. 5,414,729 wherein the multipath error is found at each ranging step, by solving a system of equations. In the method of the invention, no estimation is made of multipath parameters such as delay, phase and amplitude. Only the multipath error itself is estimated.
The coefficients αi are obtained by a calculation described in hereafter. Multipath range errors and M+N+1 correlation values are simulated for a predefined signal-to-multipath amplitude ratio (SMR), and for different values of multipath delays. Formulas to perform this computation are well documented, for instance in the document “Multipath Mitigation, Benefits from Using the Signal-to-Noise Ratio”, J. M. Sleewaegen, Proceedings of the ION GPS-97 Meeting, pp. 531–540, 1997. This simulation can be easily performed using a known calculation software such as Matlab.
The αi's are found by equating the ‘known’ simulated multipath range error (expressed in code chips) to a linear combination of the normalized correlation values. The coefficients αi then become unknowns in a system of equations. It is also possible to set one or more αi's to zero, thereby excluding the corresponding correlation(s) from the linear combination. At least two αi values however must not be put to zero.
According to the preferred embodiment, the αi coefficients are computed using the following method: The multipath error profile is simulated for one particular value of the signal-to-multipath amplitude ratio (SMR), for example 10, and for the worst case conditions where multipath is in phase with the direct signal, and 180 degrees out of phase.
To cover the whole range of multipath delays, it is sufficient to sample the multipath profile at a given number of D points linearly equally spaced at multipath delays between for example 0 and 1.2 chips. When D=50, this yields 100 samples of the multipath profile: 50 for the in-phase case, and 50 for the out-of-phase case. This is shown in
Let y be the vector of 100 multipath error samples. For all the corresponding 100 multipath conditions, the correlation values at taps i (i=−M..N) are computed, and normalized by the factor 1/I0/(1−|i|d/2). This yields M+N+1 vectors of 100 normalized correlation values, noted cj (j=1..M+N+1). Let C=[c1 c2 . . . cM+N+1] be the matrix [100×(M+N+1)] of correlation samples.
The αi coefficients are chosen to best fit the vector of multipath estimates
to the vector of multipath samples y.
In other words, the coefficients αi are the solution of the system Cα=y. As α is overdetermined by this set of equations, given that N+M+1 is normally smaller than 100, the solution is obtained by an optimization technique, preferably a least-squares optimization. By removing one or more columns from the C matrix, it is possible to exclude one or more correlations from the linear combination, or equivalently to set the corresponding αi coefficient to zero.
The curve 27 in the
The optimisation of the αi coefficients has been performed for a particular value of the SMR, e.g. 10. The invention lies in the finding that computing αi for other values of SMR yields very similar values. In other words, the αi coefficients differ little for a wide-range of SMR. This is essential to the invention as it is impossible to predict the SMR without concrete knowledge of the reflecting obstacles.
For instance, taking the example of M=0, N=3 and d= 1/15, the α coefficients are shown in the following table for different values of the SMR ranging from 10 to 100. It is assumed that this range of SMR is the most prevailing in a real-life multipath environment:
The table shows that the absolute variation of the αi's over the selected SMR range is small.
To be specific, the following linear combination is obtained for M=0, N=3, d= 1/15, and based on a SMR of 10:
The case of M=0 is equivalent with the situation wherein M=1 and α−1 is set to zero. In a receiver of the invention, there will always be at least one early correlator, one late correlator and one punctual correlator (and at least one additional correlator), so that M and N are at least equal to one, when indicating the number of physical early and late correlators in the receiver. The fact that one or more of the αi values may be set to zero however allows in the calculation of the range error, to disregard one or more correlation values, in this case I−1, so that M=0 for the purpose of calculating MP.
a and 6b compare the ranging error due to multipath and the estimation of it from the above formula, as a function of the multipath delay, for two different multipath signal amplitudes: 10 and 100 respectively. In
We can therefore summarize the inventive effect of the invention as follows:
The invention is equally related to a receiver, comprising a multipath estimator which calculates the range error based on the linear combination with fixed αi values as disclosed above.
The design parameters that have to be optimised in the design of the receiver are the number of correlators and αi coefficients, M+N+1. They result from a trade-off between receiver complexity, accuracy of the multipath error estimation, and noise on the estimation. Generally speaking, using more correlators (needing to increase the number of taps of the delay line) results in being able to better estimate the multipath error, at the expense of receiver complexity, and noise.
As a result, a receiver according to the invention comprises a plurality of channels, each channel being able to detect and lock onto a different PRN encoded signal (from a different satellite). Each channel comprises the elements shown in
The invention is equally related to the method per se, already described, of calculating the αi-values on the basis of a simulated signal.
Simulations have shown that a good compromise between accuracy of the multipath error estimation and noise on the estimation can be achieved by using only one additional correlator, at position +2 in the delay line (M=1, N=2), i.e. based on a version of the PRN code that is delayed over a time equal to d with respect to the punctual version. The preferred design is represented in
The coefficient −0.42 is the result of the optimisation of the αi parameters described above so that the resulting multipath estimation best fits the actual multipath error in the least-squares sense, for a SMR of 10. In the preferred design, the parameter d is set to 1/15. The values α−1 and α+1 have been set to zero in the calculation to obtain the above formula.
The estimation is then fed to a low pass filter 24. In the preferred design, the low pass filter has a noise equivalent bandwidth Bn of 0.1 to 1 Hz.
Also, contrary to prior art techniques, the multipath estimation is filtered independently from the range measurement in the low-pass filter 24, allowing to use a low noise equivalent bandwidth Bn, and hence to keep the noise on the estimate low. This makes sense because multipath errors typically only contain very low frequency components. As an illustration, it can be demonstrated that the standard deviation of the noise on the preferred multipath error estimator is given by:
where λc=293 m for the GPS C/A-code. As an example, σMP is only 0.1 m for d= 1/15, Bn=0.1 Hz and a nominal C/N0 of 45 dB-Hz.
Finally, some real measurement results are shown in
A complete analysis of the 24-hour data demonstrated that the method of the invention achieved a reduction of the multipath error by 20% with respect to the narrow-correlator technique, and 5% with respect to the strobe correlator.
| Number | Date | Country | Kind |
|---|---|---|---|
| 01870171 | Aug 2001 | EP | regional |
This application is a Continuation-in-Part of PCT application PCT/BE02/00136, filed Aug. 7, 2002, which application(s) are incorporated herein by reference.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5390207 | Fenton et al. | Feb 1995 | A |
| 5414729 | Fenton | May 1995 | A |
| 5615232 | Van Nee | Mar 1997 | A |
| 5734674 | Fenton et al. | Mar 1998 | A |
| 5781152 | Renard et al. | Jul 1998 | A |
| 5809064 | Fenton et al. | Sep 1998 | A |
| 5901183 | Garin et al. | May 1999 | A |
| 5918161 | Kumar et al. | Jun 1999 | A |
| 5953367 | Zhodzicshsky et al. | Sep 1999 | A |
| 5966403 | Pon | Oct 1999 | A |
| 6031882 | Enge et al. | Feb 2000 | A |
| 6252863 | Raby et al. | Jun 2001 | B1 |
| 6658048 | Valio | Dec 2003 | B1 |
| 6917644 | Cahn et al. | Jul 2005 | B2 |
| 6987820 | Brenner | Jan 2006 | B1 |
| Number | Date | Country |
|---|---|---|
| WO 9514937 | Jun 1995 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 20060133461 A1 | Jun 2006 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | PCT/BE02/00136 | Aug 2002 | US |
| Child | 10772722 | US |
