METHOD FOR PHASE CALIBRATION OF HIGH-FREQUENCY COMPONENTS OF A RADAR SENSOR

Description

[0001] The invention relates to a method for phase calibration of two high-frequency components of a radar sensor, which has an array of receiving antennas formed by two sub-arrays and an evaluation device configured to perform an angle estimation for located radar targets based on phase differences between the signals received by the receiving antennas, wherein each high-frequency component has parallel receiving paths for the signals of the receiving antennas of one of the sub-arrays.

[0002] In driver assistance systems for motor vehicles, and especially in vehicle guidance systems for autonomous driving, high-performance radar sensors are required that are capable of measuring distances, relative speeds and direction angles (especially in azimuth) of other road users, particularly other vehicles, with high accuracy and reliability.

[0003] Many well-known radar sensors for vehicles operate on the FMCW (Frequency Modulated Continuous Wave) principle, in which the frequency of the transmitted radar signal is ramp-modulated and the received radar echo is mixed with a portion of the signal being transmitted at that moment. This produces an intermediate frequency signal whose frequency corresponds to the frequency difference between the transmitted and received signals. Due to the radar signal's travel time and the frequency modulation itself, this frequency difference contains a distance-dependent component that is proportional to the ramp slope. Furthermore, due to the Doppler effect, the frequency difference also contains a component that depends on the relative velocity of the detected object.By comparing measurement results obtained with different ramp gradients, the two components can be separated, allowing the distance d and the relative velocity v of a located object to be determined.

[0004] FMCW radar sensors are also known in which a single measurement cycle comprises a multitude of steep "fast" frequency ramps (rapid chirps), whose center frequencies are modulated on a "slow" ramp. A higher measurement accuracy in distance and relative velocity measurement can then be achieved through a two-dimensional Fourier transform of the intermediate frequency signal, applied on the one hand via the fast ramps and on the other hand via the slow ramp.

[0005] Estimating the directional angles of located radar targets is made possible by offsetting the receiving antennas within the array relative to each other in the direction in which the angle is to be measured, typically the horizontal direction. The signals received by the individual receiving antennas then exhibit a phase difference that depends on the angle of incidence of the radar echo. The angular resolution can be improved by increasing the array aperture and the number of receiving antennas. However, this also increases the number of receiving channels, thus considerably increasing the complexity of the receiver section of the RF module.

[0006] It is also possible to increase the array aperture without increasing the number of receiving antennas. However, in this case, ambiguities can arise when evaluating the phase differences, making it impossible to reliably determine the actual directional angle of the object. In a MIMO (Multiple Input Multiple Output) radar, the transmitter unit of the radio frequency component has several transmitting antennas that are also horizontally offset from each other. By operating the offset transmitting antennas, for example using time-division multiplexing or code-division multiplexing, the array aperture is virtually increased without having to increase the number of receiving channels.

[0007] The invention relates to radar sensors comprising two or more essentially identical high-frequency components. These components can be used individually in low-power radar sensors, for example, in driver assistance systems, or interconnected to create a radar sensor with higher performance, particularly with higher angular resolution. However, in the latter case, it is necessary to precisely synchronize the various high-frequency components to avoid errors due to phase differences in the receiver and / or transmitter sections of the different high-frequency components.

[0008] In principle, any active electronic component in any high-frequency device can contribute to such phase differences. Since the received signals in the various reception paths are processed separately, phase differences can also occur between different reception paths within the same high-frequency device. It is possible and common practice to calibrate the radar sensor at the factory to largely eliminate these phase differences. However, if the operating conditions change during the operation of the radar sensor, the calibration can be lost.

[0009] This problem arises particularly with radar sensors containing multiple high-frequency components, which must necessarily be arranged at a certain spatial distance from one another and can therefore have different temperatures due to heat generation within the radar sensor. Temperature changes during operation of the radar sensor can thus lead to phase differences due to the thermal behavior of the electronic components involved, which impair the accuracy of the calibration.

[0010] The object of the invention is to provide a method that allows the multiple high-frequency components of a radar sensor to be recalibrated "online", i.e., during the operation of the radar sensor.

[0011] This problem is solved according to the invention with the features specified in claim 1.

[0012] JP 2009-281775 A1 describes a method for calibrating two receiving units of a radar sensor, which has an array of receiving antennas formed by two sub-arrays and an evaluation unit configured to make an angle estimation for located radar targets based on phase differences between the signals received by the receiving antennas, wherein each receiving unit has parallel receiving paths for the signals of the receiving antennas of one of the sub-arrays, with the following steps: analyze the received signals and decide whether a multi-target or single-target scenario is present; in the case of a single-target scenario, measure the phases of the signals received in the sub-arrays and calculate a phase offset between the two sub-arrays, and calibrate the phases in the two receiving units based on the calculated phase offset.

[0013] The basic idea behind this method is to compare the phase differences between signals from receiving antennas belonging to the same sub-array with corresponding phase differences for receiving antennas belonging to different sub-arrays. With correct calibration, the phase differences should depend only on the tracking angle of the located object and the relative positions of the respective receiving antennas, but not on which sub-array the receiving antennas belong to. Significant phase differences between sub-arrays therefore indicate a calibration error, which can be detected and corrected in this way.

[0014] One problem, however, is that in a multi-target scenario, the phase differences become so blurred that no clear phase offset between the two sub-arrays can be determined. A multi-target scenario, in this context, specifically refers to a situation in which two targets are located simultaneously, and the distances and relative velocities of these two targets are such that they result in an intermediate frequency signal with nearly the same frequency. In contrast, a single-target scenario refers to a situation in which only a single object is located, or, if two or more objects are located, their frequencies in the intermediate frequency signal become so distinct that they form two clearly separated and distinguishable peaks in the spectrum of the intermediate frequency signal.

[0015] Methods are known for distinguishing between a single-target scenario and a multi-target scenario, which occurs much less frequently in practice. These methods include, for example, analyzing the spectrum of the intermediate frequency signal and / or, preferably, evaluating the accuracy of an angle estimator function calculated from the measured phase differences. This function provides the probability that the object is located at each possible angle of location. With high accuracy, the angle estimator function has a sharp peak at a specific angle, which then represents the object's location angle. In a multi-target scenario, however, the peak is generally "smeared" over a significantly wider area, making it less reliable to determine the precise location angle.

[0016] Such effects are used to distinguish between a single-target scenario and a multi-target scenario, and recalibration is only performed in the case of a single-target scenario where the phase offset between the two sub-arrays is clearly detectable. To determine this phase offset, the received signals from the two sub-arrays are then compared. In principle, the dependence of the phase difference on the antenna spacing is examined for different pairs of receiving antennas. With correct calibration, the same spacing dependence should be obtained for antennas from different sub-arrays as for antennas from the same sub-array. A faulty calibration of the two receiving units can be recognized by the fact that, in addition to the phase difference expected due to the spacing, there is always an approximately equal additional phase offset for antennas from different sub-arrays.Statistical evaluation of the phase differences for different pairs of receiving antennas allows statistical fluctuations to be smoothed out, providing a meaningful measure of the phase offset to be corrected.

[0017] The invention relates to a method for phase calibration of two transmitting units arranged on two different high-frequency modules of a radar sensor, which has a real array of receiving antennas and an evaluation unit configured to perform an angle estimation for located radar targets based on phase differences between the signals received by the receiving antennas, wherein each transmitting unit feeds at least one transmitting antenna and the transmitting antennas belonging to different transmitting units are offset from each other in the direction of the array such that the real array is extended by a virtual array when both transmitting units are used, comprising the following steps: Analyze the received signals and decide whether a multi-target or single-target scenario is present; if a single-target scenario is present, calibrate the phases in the two transmitting units; the following steps are performed to calibrate the phases: measure phases of the signals received in the real array, determine a straight line that best fits the measured phases of the real array, extrapolate the straight line into the virtual array, calculate a phase offset between the real and the virtual array based on the extrapolated straight line and the phases measured in the virtual array, and calibrate the phases in the two transmitting units based on the calculated phase offset.

[0018] This method is based on the same fundamental principle as the receiver calibration procedure described above, but with the difference that instead of considering two sub-arrays of the same real array, it considers the (complete) real array and a corresponding virtual array. This virtual array is created when the receiving antennas of the real array receive signals from a different transmitting antenna. Due to the altered spatial offset between the receiving and transmitting antennas, different signal paths and thus different phase shifts result. Again, with correct calibration, the phase differences should be independent of whether one considers two antennas from the same (real or virtual) array or two antennas from different arrays. A significant phase offset therefore indicates an error in the phase calibration of the transmitting units involved.

[0019] Advantageous embodiments and further developments of the invention are specified in the dependent claims.

[0020] Statistical noise in phase offset measurements can be reduced by including in the statistical analysis the measurement results obtained for a large number of successive frequency ramps (e.g., rapid chirps) of the transmitted signal. For example, one can consider the median or mean value of the phase offsets across the different frequency ramps.

[0021] Depending on the required accuracy, recalibration can be performed in every measurement cycle or only at certain intervals. Furthermore, depending on the hardware's performance and workload, the phase calibration result can either be used for an accurate angle estimation in the current measurement cycle or initially saved and then used in the next measurement cycle or in several subsequent measurement cycles.

[0022] The methods can also be used analogously for radar sensors with three or more transmitting units.

[0023] The following are examples of implementation explained in more detail with reference to the drawing.

[0024] They show: Fig. 1 a diagram of a radar sensor to which the invention is applicable; Fig. 2 a simplified diagram of the radar sensor to illustrate a method for phase calibration of two receiving units; Fig. 3 a flowchart to illustrate essential steps of a method according to the invention; and Fig. 4 a simplified diagram of the radar sensor according to Fig. 1 , to explain a method for phase calibration of two transmitting units.

[0025] In Fig. 1A schematic diagram shows a radar sensor for motor vehicles, comprising two high-frequency components 10 and 12 on a common circuit board 14. The high-frequency components 10 and 12 are, for example, monolithic microwave integrated circuits (MMICs) and each have a transmitter 16 and a receiver 18. The circuit board 14 has an array 20 of receiver antennas E1–S8, arranged at regular intervals along a horizontal plane. The array 20 is divided into two sub-arrays 22 and 24, each with four receiver antennas. Each receiver antenna is formed by two parallel, vertically oriented columns of antenna patches, which achieve a certain degree of focusing of the received beam in elevation. The receiver antennas E1–E4 of sub-array 22 are connected to the receiver 18 of the high-frequency component 10 via respective conductor tracks 26.Accordingly, the receiving antennas E5 - E8 of the sub-array 24 are connected to the receiving unit 18 of the high-frequency module 12. The lengths of the conductor tracks 26 are tuned so that they differ by no more than integer multiples of the wavelength λ of the microwaves, so that the phases are... relationships between the signals received by the different receiving antennas on the way to the receiving units 18 are not distorted.

[0026] The transmitting units 16 of the two high-frequency modules 10 and 12 are each connected to a transmitting antenna S1 and S2, respectively, by conductor tracks 28. The transmitting antennas S1 and S2 are also each formed by two vertical columns of antenna patches and are arranged symmetrically to the array 20, but vertically offset from it, on the circuit board 14. The conductor tracks 28 are also arranged such that the signal paths from the transmitting unit 16 to the associated transmitting antenna differ by no more than an integer multiple of λ.

[0027] The receiver unit 18 of each high-frequency module 10 has four parallel receive paths, each processing the signal from a receiving antenna. As is known per se, each receive path contains a mixer by which the received signal is mixed with a portion of the transmitted signal supplied to the associated transmitting antenna, so that an intermediate frequency signal is generated for each receive path. The intermediate frequency signals of the two receiver units 18 are fed via signal lines 30 to an evaluation unit 32, where the signals are further evaluated to determine the distances d, the relative velocities v, and the azimuth angles θ of the located radar targets.

[0028] Each high-frequency module 10 is also connected via control lines 34 to a control unit 36, which controls and coordinates the operation of the two high-frequency modules 10, 12.

[0029] As usual, the transmitted signal generated by the transmitting units 16 is ramp-modulated in its frequency. The two transmitting units 16 are operated, for example, in time-division multiplexing, so that only one of the two transmitting antennas S1, S2 is active at any given time.

[0030] The intermediate frequency signals supplied by the receiving units 18 are digitized in the evaluation unit 32, or optionally directly in the high-frequency modules 10 and 12, and recorded over the duration of a frequency ramp. From the resulting eight time signals, a spectrum is generated in the evaluation unit 32 by means of a fast Fourier transform. Each detected object appears in this spectrum as a peak at a specific frequency, which depends on the object's distance and relative velocity. Using known evaluation methods, the distance- and velocity-dependent components are separated, allowing the distance d and relative velocity v of each detected object to be determined.

[0031] In principle, the azimuth angle of each located object can be determined based on the phase relationships between the signals received by the receiving antennas E1-E8 and the corresponding phase relationships between the intermediate frequency signals. An angle estimation algorithm is implemented in the evaluation unit 32 for this purpose. This algorithm calculates an angle estimation function (e.g., a maximum likelihood function) for each object based on the phase relationships, which provides a probability distribution for the location angle of the object in question.

[0032] In a normal measurement cycle, the signals of all eight receiving antennas of array 20 are evaluated on each frequency ramp. The larger the aperture of array 20, the greater the achievable accuracy of the angle estimation.

[0033] However, an accurate angle estimation requires that the phase relationships between the signals from the eight receiving antennas remain unchanged during signal reception and processing. Since each of the two high-frequency modules 10, 12 processes only four of the eight received signals, the two modules must be precisely synchronized. For this purpose, the two modules are connected by a synchronization line 38. This synchronization must ensure, firstly, that the mixers in the two receiving units 18 mix the received signals with phase-matched transmitted signal components.For example, if the transmitting antenna S1 is active, the transmit signal generated in the high-frequency module 10 can be transmitted to the high-frequency module 12 for synchronization purposes, ensuring that the signal propagation time during this transmission does not cause a phase shift. Alternatively, the mixers in the receiving units 18 of the two high-frequency modules can each receive a signal generated by the local transmitting unit 16. In this case, however, the oscillators of the two transmitting units 16 must be synchronized with each other.

[0034] In principle, each of the active electronic components in the eight receiver paths can cause a certain phase shift, which can lead to phase differences between the receiver paths that distort the angle estimation. Provided these phase shifts are stable over time, they can be measured at the factory before the radar sensor is put into operation and eliminated through appropriate calibration measures or compensated for by corresponding corrections during signal processing.

[0035] However, if the operating conditions under which the high-frequency components 10 and 12 operate are not stable during the operation of the radar sensor, the phase differences can change over time, so that even an initial calibration of the radar sensor cannot permanently eliminate measurement errors.

[0036] A major cause of changes in phase differences over time is temperature fluctuations, which affect the operation of the active electronic components in the two high-frequency devices. Phase differences can arise, firstly, because the two high-frequency devices 10 and 12 heat up to different degrees during operation. Secondly, it is also conceivable that the electronic components in the two semiconductor devices react differently to temperature fluctuations.

[0037] Therefore, when high accuracy is required, it is necessary to check and, if necessary, correct the phase calibration of the receiving units 18 and also the transmitting units 16 of the two high-frequency modules 10, 12 from time to time during operation.

[0038] An unclaimed method for such recalibration of the receiving units 18 shall be described with reference to Fig. 2The diagram, which shows a simplified diagram of the radar sensor described above, is further explained. It additionally plots the phases φ of the signals received by the receiving antennas E1-E8 as a function of the position x of these receiving antennas (in the horizontal plane) for a single located radar target whose radar echo is received at a specific azimuth angle θ. It is assumed that any phase differences between the four receiving paths in each receiving unit 18 have been eliminated by an initial calibration, so that, due to temperature fluctuations, only phase differences between the receiving paths in the high-frequency module 10 on the one hand and the receiving paths in the high-frequency module 12 on the other hand are to be expected.

[0039] In the example considered here, the receiving antennas E1 to E8 of array 20 are arranged at equal intervals in the x direction. Therefore, when the radar echo arrives obliquely at a specific azimuth angle, the phases of the signals received by the eight receiving antennas should, with correct calibration, lie on a straight line as a function of the position x, as shown in Fig. 2 The line is indicated by black dots and the graph P1. The slope of this line depends on the angle of incidence of the radar echo. At perpendicular incidence (azimuth angle 0°), the line would be... Fig. 2 run horizontally. If, due to different temperatures of the high-frequency components 10, 12 and the receiver units 18 contained therein, phase differences occur between the phases measured in the different receiver units, this leads to a phase offset Δφ between the phases in array 22 on the one hand and the phases in array 24 on the other, as shown in Fig. 2This is indicated by white dots and the graph P2. The phase differences between antennas within each sub-array, however, remain unchanged. Based on this characteristic pattern, the phase offset can be determined and quantified, and compensated for by appropriate recalibration.

[0040] However, it illustrates Fig. 2 the ideal case in which the measured phases are not noisy and are not distorted by interference effects.

[0041] Interference effects occur particularly in a multi-target scenario where two or more radar targets are simultaneously ordered and the corresponding peaks in the intermediate frequency signal spectrum are so broad and / or so close together that they overlap and can no longer be separated. In such a case, the quality factor of a typical angle estimation based on all eight received signals is significantly lower than in a single-target scenario. Multi-target scenarios, which are relatively rare compared to single-target scenarios, can therefore be identified by the quality factor of the angle estimation function, allowing them to be excluded from the phase offset calculation.

[0042] Statistical noise can initially be suppressed by statistically evaluating the phases and phase differences for different pairs of receiving antennas. For example, one can in Fig. 2First, using the four measurement points for sub-array 22, determine the shape of the straight line (graph P1) that best matches the measurement results. Then, extrapolate this line to sub-array 24 to determine the expected phases for receiving antennas E5 to E8 (black dots). These expected phases can then be compared for each receiving antenna with the actually measured phase (white dots), and by averaging over the four differences obtained, a realistic value for the phase offset Δφ is obtained. This type of evaluation is also possible analogously in cases where the distances between the receiving antennas are not all the same.

[0043] The evaluation described above can, in principle, be performed for every frequency ramp of the transmitted signal. By statistically evaluating the results for several consecutive ramps (for example, by considering the median), the statistical fluctuations can be further suppressed, thus further improving the accuracy of the obtained value for the phase offset Δφ.

[0044] In Fig. 3 The essential steps of a procedure for calibrating the receiving units 18 are shown in a flowchart.

[0045] In step S1, a "normal" measurement cycle is performed to determine the distances d and the relative velocities v, as well as, if applicable, the detection angles of the located radar targets. The transmitted signal generated in the measurement cycle can comprise a multitude of successive frequency ramps with the same slope and, if necessary, different center frequencies, and the calculation of the distances and relative velocities can be based on the entirety of the signals received on these ramps. In principle, evaluating the signals on one or a few frequency ramps is sufficient for angle estimation; however, it is preferable to use the largest possible data set to improve the signal-to-noise ratio.

[0046] In step S2, separate angle estimates are performed for the two sub-arrays 22 and 24. One angle estimator is calculated based on the signals from receiving antennas E1 to E4, and another angle estimator is calculated based on the signals from receiving antennas E5 to E8. Even with faulty calibration of the two receiving units 18, the results of these angle estimates should essentially agree, since the calibration error does not affect the phase differences between receiving antennas within the same sub-array. However, the selectivity of these angle estimates is limited because the aperture of each of the sub-arrays used is only half the aperture of the entire array 20.

[0047] The quality factor is calculated for each of the two angle estimators, and in step S3 it is checked whether the quality factor for both angle estimators is above a certain threshold. This threshold is chosen such that a quality factor above this threshold is normally only obtained in a single-target scenario, in which the received signals, which lead to a peak at a certain frequency in the spectrum of the intermediate frequency signal, originate from only a single reflection center, so that a clear phase offset is detectable between the signals received in the different sub-arrays.

[0048] If this condition is met (J), the phase offset ΔΦ is determined in step S4 using the value based on Fig. 2 The procedure described above is calculated, and in step S5 the phase calibration is corrected based on this phase offset.

[0049] In principle, the calibration correction could be achieved by controlling active electronic components in one of the two receiving units in such a way that the signals processed in the receiving filament are shifted according to the phase offset. Preferably, however, the phase calibration is performed purely computationally by subtracting the phase offset ΔΦ from the phases for the signals from one of the two sub-arrays. This has the advantage that no additional active electronic components are required in the receiving units.

[0050] If a multi-target scenario (N) was detected in step S3, steps S4 and S5 are skipped. In step S6, an angle estimation is then performed—in both single-target and multi-target scenarios—based on all eight received signals of the complete array 20, but now using the phase corrections updated in step S5. Afterwards, the process returns to step S1 so that the next measurement cycle can be executed.

[0051] In a modified embodiment, step S6 can also be integrated into step S1. In this case, the phase corrections updated in step S5 do not yet take effect in the current measurement cycle, but only in step S1 of the subsequent measurement cycle.

[0052] In yet another embodiment, steps S2 to S5 are not performed in every measurement cycle, but only at certain time intervals that are greater than the duration of a single measurement cycle of typically 50 ms.

[0053] The one here in Fig. 1 The radar sensor shown as an example is a MIMO radar sensor in which the two transmitting units 16 operate, for example, in time-division multiplexing, so that at any given time only one of the two transmitting antennas S1 and S2 is active. This achieves a virtual increase in the aperture of the antenna array and thus a greater angular resolution capability.

[0054] The phases of the signals received by the receiving antennas E1 to E8 depend on the total length of the signal path from the transmitting antenna S1 or S2, which is currently sending the signal, to the radar target and from the radar target back to the respective receiving antenna. These phases therefore depend not only on the positions of the receiving antennas E1 to E8 in the x-direction, but also on the x-position of the transmitting antenna currently in use. For example, if the transmitting antenna S2 is active, the phases of the signals received by the receiving antennas E1 to E8 are given by the values ​​shown in Fig. 4These receiving antennas are graphically represented. Switching to transmitting antenna S1 has the same effect on the phases of the received signals as if receiving antennas E1-E8 had been shifted to the right (positive x-direction) by the distance between transmitting antennas S1 and S2. These shifted positions of the receiving antennas thus form a virtual array 38 with virtual receiving antennas V1-V8.

[0055] The positions of the transmitting antennas S1 and S2 are chosen such that the distance between the receiving antenna E8 and the first virtual antenna V1 is the same as the distance between two adjacent receiving antennas. If the received signals during the active phases of the transmitting antennas S1 and S2 are considered together, the phases of the signals in the virtual antennas V1 to V8, with correct calibration, lie on the same straight line (graph Q1) as the phases for the real array 20. However, if there is a phase difference between the transmitted signals generated by the transmitting units 16 of the two high-frequency modules, a corresponding phase offset ΔΦ results for the phases received in the real array 20 and in the virtual array 38 (graph Q2). Fig. 4If one now performs an angle estimation based on the combined arrays 20 and 26 in order to achieve better angle separation by doubling the aperture, the result of the angle estimation would be distorted by the phase offset ΔΦ.

[0056] However, this calibration error between the transmitting units 16 can be corrected according to the invention using the same principle as described above in conjunction with Figures 2 and 3 This was explained for the receiving units 18. If a single-target scenario is detected, the phase offset Δϕ is measured, and this phase offset is then used for a corresponding phase correction in the actual angle estimation.

Claims

1. Method for the phase calibration of two transmitting units (16) which are arranged on two different high-frequency modules (10, 12) of a radar sensor having a real array (20) of receiving antennas (E1 - E8) and an evaluation device (32) which is designed to carry out an angle estimation for located radar targets on the basis of phase differences between the signals received by the receiving antennas, wherein each transmitting unit (16) feeds at least one transmitting antenna (S1, S2) and the transmitting antennas belonging to different transmitting units are offset with respect to each other in the direction of the array (20) in such a way that the real array (20) is extended by a virtual array (38) when using both transmitting units, having the steps of: - analysing the received signals and deciding whether there is a multi-target or a single-target scenario, - if there is a single-target scenario, calibrating the phases in the two transmitting units (16); wherein the following steps are performed in order to calibrate the phases: - measuring phases of the signals received in the real array (20), - determining a straight line that best matches the measured phases of the real array, - extrapolating the straight line into the virtual array (38), - calculating a phase offset between the real array (20) and the virtual array (38) on the basis of the extrapolated straight line and the phases measured in the virtual array, and - calibrating the phases in the two transmitting units (16) on the basis of the calculated phase offset.

2. Method according to Claim 1, in which the quality of an angle estimation function is assessed for the decision as to whether there is a multi-target or a single-target scenario.

3. Method according to Claims 1 and 2, in which separate angle estimation functions are calculated for the real array (20) and the virtual array (38) and the quality of both angle estimation functions is assessed.

4. Method according to one of the preceding claims, for an FMCW radar sensor, in which the transmission signal is modulated in each measurement cycle according to a sequence of a plurality of successive frequency ramps, in which a plurality of phase offsets are determined in each measurement cycle for one of the frequency ramps in each case, and the measured phase offsets are statistically evaluated in order to form a correction value for the phase calibration.

5. Method according to one of the preceding claims, in which a phase calibration is carried out in each measurement cycle.

6. Method according to Claim 5, in which an angle estimation is carried out in each measurement cycle on the basis of phase correction values obtained from the phase calibration in the same measurement cycle.

7. Radar sensor having two transmitting units (16), a real array (20) of receiving antennas (E1 - E8), an evaluation device (32) which is designed to carry out an angle estimation for located radar targets on the basis of phase differences between the signals received by the receiving antennas, and having a control unit (36) for controlling the functions of the radar sensor, wherein each transmitting unit (16) feeds at least one transmitting antenna (S1, S2) and the transmitting antennas belonging to different transmitting units are offset with respect to each other in the direction of the array (20) in such a way that the real array (20) is extended by a virtual array (38) when using both transmitting units, characterized in that the control unit (36) is designed to carry out the method according to Claim 1.