DE102008009596A1 - Radar system for detecting surrounding of motor vehicle, has oscillator modulating frequency of power, and digital-analog-converter producing discrete control signal values, such that discretization errors are concentrated on blocking zone - Google Patents

Abstract

The system has a transmission unit for directed transmission of a transmission power, and a receiving unit for receiving the transmission power reflected at objects. A voltage-controlled oscillator (VCO) modulates the frequency of the transmission power, and a digital-analog-converter (DAC) produces a sequence of discrete control signal values, where a control signal of the oscillator is generated from the signal values with analog low-pass filtering. The signal values are produced in such a manner that discretization errors are concentrated on a blocking zone of a V-tune-filter. An independent claim is also included for a method for controlling frequency ramps in a radar system for detecting surrounding.

Description

The The invention relates to an apparatus and a method for frequency modulation of the transmission signal in a radar system. Radar systems are z. B. used for environment detection in motor vehicles.

motor vehicles are increasingly being equipped with driver assistance systems, which with the help of sensor systems to capture the environment and out of the sun recognized traffic situation automatic reactions of the vehicle derive and / or instruct the driver, in particular warn. there one distinguishes between comfort and safety functions.

in the The range of comfort functions currently plays FSRA (Full Speed Range Adaptive Cruise Control) the most important role. The vehicle regulates the airspeed to the desired speed specified by the driver if the traffic situation allows, otherwise the airspeed will be automatically adapted to the traffic situation.

at The safety functions include reducing the braking distance Emergency situations in the center. The spectrum of the corresponding driver assistance functions ranges from an automatic incident of the brake to the reduction the brake latency (prefill), about an improved brake assistant (BAS +) up to the autonomous Emergency braking.

For driver assistance systems The types described above are now predominantly radar sensors used. These work well in bad weather conditions reliable and can in addition to the distance of objects also directly their relative velocity over the Measure Doppler effect. Because radar sensors are still relatively expensive are found, such driver assistance systems mainly in high-priced Vehicles.

Around with radar sensors to be able to measure the distance of objects is a modulation of the transmission signal necessary. While in the past mainly the transmission amplitude modulated in a pulse shape was, move Today method with modulation of the transmission frequency in the foreground. They have the advantage of a higher one Range and a lower susceptibility to interference, the latter at an increasing use of radar-based driver assistance systems becomes an increasingly important criterion.

The Frequency modulation today is typically using phase locked loops (PLL) realized, which, however, for a relatively complex and are expensive, and on the other hand, even with very fast modulation problems with full engagement can have. Existing approaches without phase locked loops either place high demands on the linearity and stability the frequency dependence of the oscillator of its control size or result in reasonable Cost in a reduced performance in terms of detection quality and safety of objects, what for many driver assistance functions but is not acceptable.

task The invention is an improved method and an improved Specify device for frequency modulation, whereby a frequency modulation simple, inexpensive and with big ones precision is reached.

These The object is achieved by the features of the independent claims.

It is a radar system for Umfeldfassung with transmitting means to the directed Transmission of transmission power, receiving means for directional reception reflected by objects transmitted power and signal processing means for processing the received power and for controlling the Sender specified. The frequency of the transmission power is below Using a controllable oscillator modulated, wherein for the frequency of the Transmit power a target course is specified. The control signal of Oscillator, which drives the oscillator, becomes a sequence of discrete control signal values followed by analog low-pass filtering generated. There is only a finite set of discrete control signal values to disposal, which leads to discretization errors and thus to a deviation between Actual and target course of the frequency of the transmission power leads. The Sequence of discrete control signal values and the analogue lowpass are designed so that the before the analog filtering resulting discretization error spectrally on the stopband of the analog low pass are concentrated, resulting in a significant Reduction of discretization errors after analog filtering and thus to a reduction of the deviation between actual and Target course of the frequency leads. In particular, the analog low pass has such a wide passband, that the control of the desired function substantially comes through.

In an alternative formulation, the control signal values and the analogue low-pass filter are designed in such a way that the discretization errors resulting before the analogue filtering are spectrally minimal on the passband of the analog low-pass filter or partial sections thereof. This is z. B. achieved in that the sum of the discretization error is minimized to a plurality of control signal values. In particular, the analog low-pass filter has such a wide passband that the control of the desired function essentially comes through The invention will be explained in more detail with reference to figures and embodiments.

1 : Block diagram of a radar system with means for frequency modulation

2 : Transmission signal as a result of linear frequency ramps

3a : Choice of DAC values so that the discretization error becomes minimal at all times

3b : Choice of DAC values so that the sum of the discretization error becomes minimal at all times

4 : Discrete signal flow graph for determining the DAC values so that the discretization error becomes minimal at each time point

5 : general discrete signal flow graph for determining the DAC values

6 : Amount of error transfer function H _e (z) = - (1 - z ^-1 ) ²

n 1 a block diagram of a radar system is shown. By means of a voltage-controlled oscillator (VCO), a frequency-modulated transmission signal is generated, which is supplied via a coupler structure of an antenna. At the same time, the oscillator signal is fed to a mixer, where it is mixed with the received signal. The transmission signal used is a sequence of linear frequency ramps (see 2 , solid line), the frequency deviation per unit time is so large that the difference frequency between the transmit signal and received signal almost exclusively on the duration and thus the distance to the object at which the reflection takes place, and only to a much lesser part of the relative speed , Thus, the distance information is obtained from a single ramp, the frequency of the baseband signal being proportional to the distance. By evaluating the phase change between the ramps, the Doppler frequency and thus the relative velocity result. The calculation of the distance and the relative speed can thus be carried out with the aid of a two-dimensional Fourier transformation, which is realized as a two-stage FFT.

The Frequency ramps need very precise, so especially highly linear, otherwise it will lead to blurred images in the distance and thus also to cover smaller targets, to interference lines in Distance spectrum and thus misdetections and increased noise can come.

Because of the extremely short duration of the frequency ramps of z. B. 18 μs would be the use of a Phase locked loop very critical, since this then problems with the Snapping into place. Instead, the oscillator is over a digital-to-analog converter (DAC) with subsequent analogue Low pass filtering (V-tune filter) activated. The V-tune filter has about a bandwidth of z. B. 1 MHz and leaves the frequency ramps - from one tolerable delay apart - approximately undistorted by.

The However, the invention is also unlimited for frequency ramps with a longer one Duration applicable.

In addition to The main path described above, a Rückmesspfad is provided with whose help is dependency the oscillator frequency is measured by the control voltage. To a frequency divider, a bandpass filter and a multiplexer are provided. This path is during the actual environment measurement deactivated to a coupling to avoid the reception path. He will be in every sensor cycle (eg 66 ms) are activated for a few milliseconds. there will over the DAC gradually increases the VCO drive voltage and the associated frequency the oscillator signal determined. It will be the oscillator frequencies all DAC values measured, which for the generation of the frequency ramps used for environmental measurement (a measurement of only individual DAC values with following Obtaining the intervening DAC values by interpolation would not lead to acceptable errors, since the DAC non-linearities has, in particular differential type). To the frequency measurement accuracy to increase be the over successive cycles won oscillator frequencies to the individual DAC values filtered.

The following describes the measurement of the oscillator frequencies versus the DAC values used. The oscillator signal is first divided down, then filtered with a bandpass and sampled with an analog-to-digital converter. The digitized signal is multiplied by a suitable window function and then an FFT is calculated. Here, the same FFT as in the target processing can be used, which offers a great advantage especially when implemented on an FPGA. The return measurement signal is formed according to its frequency as a peak power at a specific line position in the spectrum. The shape of this peak power is given by the spectrum of the window function. The line position of the highest power peak alone only leads to a very inaccurate frequency indication. Therefore, the power of the highest line as well as the power of its left and right neighbors is used to under Considering the window function to interpolate the exact line position. This can be done either by means of a lookup table or an approximation function. Likewise, the interpolation can first be roughly calculated on an FPGA and then refined on an MCU by means of a correction function. The fact that the feedback signal is measured via a separate path within the sensor, gives a very good signal-to-noise ratio, whereby the interpolation of the line position can be done very accurately.

The Bandpass filter in the feedback path is particularly designed so that the harmonics, which in the Frequency division arise, are very much attenuated, as they follow the sampling in the same frequency range as the return measurement signal itself can be located what a falsification of lead measured frequency would.

(k = 0, 1, 2 ..) liegt (Übertastung für k > 0). Zum anderen ist die Fensterfunktion geeignet zu wählen, so dass für die sich ergebenden relevanten Frequenzbereiche, in denen die Leistungsspitzen zu liegen kommen können, das Spektrum des Fensters schon so weit abgeklungen ist, dass sich die Leistungsspitzen gegenseitig nicht mehr beeinflussen.Since the return measurement signal is a real signal, it forms symmetrically in the spectrum. In addition, the spectrum of a sampled signal repeats periodically with the sampling frequency. If the envelopes of the power peaks, which are given by the spectrum of the window function, are too close to each other, so that power components of one power peak are still to be found at the maximum of the other power peak, this likewise leads to a falsification of the measured frequency. Therefore, on the one hand, the divider factor K ₁ and the sampling frequency f _AOC are to be matched to one another in such a way that the frequency of the return measurement signal is in particular at

(k = 0, 1, 2 ..) (over-sampling for k> 0). On the other hand, the window function is suitable to choose, so that for the resulting relevant frequency ranges in which the power peaks can come to lie, the spectrum of the window has already subsided so far that the power peaks no longer affect each other.

With the oscillator frequencies thus measured, the linear frequency ramps with the desired frequency f _Soll (t) can now be generated by a corresponding drive sequence of the DAC. The DAC is z. B. clocked at the frequency f _A = 80 MHz, ie, it is all T _A = 12.5 ns loaded with a new digital value, the associated voltage then he sets the output.

The simplest approach for generating the frequency ramps would be to set at each instant nT _A that DAC value whose oscillator frequency f _{dis is} closest to the _setpoint frequency f _Soll (n) at this time nT _A ; this is in 3a shown. Because the desired frequency f _Soll (n) can not be set exactly, a discretization error e (n) arises. This method, which minimizes the discretization error at all times, would produce a white power density spectrum of the error if consecutive values of f _set (n) were uncorrelated; The background is that this method does not connect the error at one time with errors at other times, and thus implicitly does not prefer a frequency of the error. For the resulting control voltage of the oscillator, only the frequencies of the error e (n), which are passed by the low-pass filter (V-tune filter) after the DAC, play a role; the bandwidth of this low pass is z. At 1 MHz. Although this suppresses most of the discretization error, using low-cost DACs typically still does not suffice to produce enough good quality of the frequency ramps-DACs with high numbers of bits would be needed, but they are expensive.

In order to avoid this disadvantage, one can determine the sequence of DAC values such that the discretization error in the passband of the low-pass filter has as small as possible spectral components. One approach is not to minimize the discretization error e (n) at all times, but to do so as in 3B represents the sum over the discretization errors up to a given time or a predetermined measurement interval. In 3b is the sum of the discretization error that is minimized at each measurement time n (n = 1, 2, ..N). Due to the summation, the effect is greatly reduced that repeated errors of the same sign can be repeated over several consecutive times, which are reflected in the power density spectrum of the error at low frequencies.

• a) Bestimmung des Frequenzsollwerts f_Soll(n) zum Zeitpunkt nT_a
• b) Bestimmung der Größe f₂(n) als die diskrete Frequenz f_dis(n – 1) zum zurückliegenden DAC-Werts DAC(n – 1), wobei die Anfangsbedingung f_dis(0) = 0 ist (f_dis bezeichnet jeweils die stationäre diskrete Oszillatorfrequenz, welche sich im eingeschwungenen Zustand bei dem jeweiligen DAC-Wert ergibt),
• c) Bestimmung der Größe f₃(n), indem vom Frequenzsollwert f_Soll(n) die Größe f₂(n) subtrahiert wird,
• d) Bestimmung der Größe f₄(n) durch Addition der Größe f₃(n) zum vorhergehenden Wert f₄(n – 1), wobei die Anfangsbedingung f4(0) = 0 ist; f₄(n) ist damit die Summe über alle vorausliegenden Diskretisierungsfehler e(n – k) = f_Soll(n – k) – f_dis(n – k) plus der aktuellen Sollfrequenz,
• e) Bestimmung des DAC-Werts DAC(n) so, dass die zugehörige Frequenz f_dis(n) möglichst wenig von der Größe f₄(n) abweicht; der Diskretisierungsfehler e(n) = f₄(n) – f_dis(n) ist im Signalflussgraphen nach 4 im Block ,Diskretisierung' als subtrahiertes Signal berücksichtigt. 13.02.2008

For this, high-frequency errors can form stronger in this method, which are suppressed by the low-pass filter. The sequence of DAC values DAC (n), n = 1... N, is then z. B. iteratively to form the following procedure - the corresponding discrete signal flow graph is in 4 shown:

• a) Determination of the frequency _setpoint f _Soll (n) at time nT _a
• b) determining the quantity f ₂ (n) as the discrete frequency f _dis (n-1) with respect to the past DAC value DAC (n-1), the initial condition f _dis (0) = 0 (f _dis denotes the respective stationary discrete oscillator frequency, which results in the steady state at the respective DAC value),
• c) determining the quantity f ₃ (n) by subtracting the quantity f ₂ (n) from the frequency _setpoint f _Soll (n),
• d) determining the quantity f ₄ (n) by adding the quantity f ₃ (n) to the previous value f ₄ (n-1), wherein the initial condition f4 (0) = 0; f is ₄ (n) so that the sum over all preceding discretization errors e (n-k) = f _setpoint (n-k) _{-f dis} (n-k) plus the current setpoint frequency,
• e) determining the DAC value DAC (n) such that the associated frequency f _dis (n) differs as little as possible from the size f ₄ (n); the discretization error e (n) = f ₄ (n) _{-f dis} (n) is in the signal flow graph 4 considered in the block, discretization 'as a subtracted signal. 13/02/2008

The transfer function from the error e (n) to the output f _dis (n) is calculated as H _e (z) = - (1 - z ^-1 ), which corresponds to a high pass, the so-called first difference. If the discretization error e (n) represents white noise (same power density for all frequencies), then the error in the output signal f _dis (n) is not white, but mainly has high frequency components which are suppressed by the subsequent low pass.

The idea underlying this approach of shifting the discretization error spectrally to higher frequencies by taking into account earlier discretization errors in the discretization can now be generalized; the associated discrete signal flow graph is in 5 shown. There are in each case a filter with the transfer function H _R (z) and H _v (z) in the backward and forward paths. For the transfer function H _e (z) from the error e (n) to the output f _dis (n) a certain high-pass function is selected. Together with the requirement that the transfer function from the input f _Soll (n) to the output f _dis (n) should have the value 1 without the discretization, so that the shape of the frequency ramp is not distorted, the transfer functions of the two filters to H _R (z) = 1 + H _e (z) and H _v (z) = - 1 / H _e (z). The boundary condition for the choice of the error transfer function H _e (z) is that no loop without a delay element may occur in the signal flow graph.

This boundary condition is z. B. is always satisfied if H _e (z) represents a non-recursive filter of the form H _e (z) = - 1 + a ₁ z ^-1 + ... + a _K z ^-K .

In the exemplary embodiment, the second error sum was minimized, which corresponds to a choice of the error transfer function H _e (z) = - (1-z ^-1 ) ² - this is the so-called second difference. The amount of their transfer function is in 6 shown. As a result, the discretization error is for the most part 'pushed' into the stop band of the low pass (the stop band of the low pass starts at approximately f = 1 MHz, which corresponds to the normalized frequency f / f _A = 0.0125 with f _A = 80 MHz) ,

Thereby needed the DAC is about 6 bits less than the simplest approach, which minimizes the discretization error at any time.

Around to ensure that the discretization error is in any case white add a white noise before discretization (thereby For example, artifacts through periodically recurring conditions avoided).

It It should be emphasized that this approach requires that the clock frequency of the DAC significantly higher than the bandwidth of the analog low-pass filter (V-tune filter) is.

Claims (13)

Radar system for environment detection with - transmitting means for directional emission of transmission power, - receiving means for the directional reception of object-reflected transmission power and - signal processing means for processing the received power and for controlling the transmission means, wherein a) the frequency of the transmission power is modulated using a controllable oscillator , wherein a desired course is predetermined for the frequency and b) a control signal of the oscillator is generated from a sequence of discrete control signal values with subsequent analog low-pass filtering, whereby only a finite set of discrete control signal values is available, resulting in discretization errors and thus in a deviation between a tasächlichen actual and the desired course of the frequency leads, characterized in that the sequence of discrete control signal values and the analog low-pass filter are designed so - that arise before the analog filtering the discretization error is spectrally minimized to the passband of the analog low pass or portions thereof. Radar system according to claim 1, wherein a sum of the Discretization error to a plurality of control signals in Passband of the analog low pass is minimized. A radar system according to claim 1 or 2, wherein the Sequence of discrete control signal values through a digital-to-analog converter is generated, with its clock frequency for a significantly higher than the bandwidth of the analog low-pass filter and on the other hand significantly higher than the bandwidth of that oscillator control signal which is exactly generate the desired course of the frequency. Radar system according to one of the above claims, wherein the sequence of discrete control signal values s (n), n = 1 ... N, is iteratively calculated according to the following procedure: a) Determining the frequency _setpoint f _Soll (n) for Time nT _a , where T _{a is} the clock time of the digital-to-analog converter, b) determination of the size f ₂ (n) from the stationary discrete frequencies f _dis (n-1), f _dis (n-2), ... to the previous discrete control signal values s (n-1), s (n-2),... through a filter with the transfer function H _R (z), which consists of a high-pass filter with the transfer function H _e (z) according to the instruction H _R (z) = 1 + H _e (z) is formed (f _dis (.) So is the oscillator frequency, which results in the steady state when applying the respective discrete control signal value s (.)), c) determining the size f ₃ ( n) by subtracting the quantity f ₂ (n) from the _desired frequency value f _Soll (n), d) determining the quantity f ₄ (n) from the quantity f ₃ (n) by a filter having the transfer function H _v (z) which is formed from the high pass with the transfer function H _e (z) according to the instruction H _v (z) = - 1 / H _e (z), e) determination of the discrete control signal value s ( n) so that the associated stationary oscillator frequency f _dis (n) deviates as little as possible from the size f ₄ (n); the state variables of the filters are suitable to initialize, z. For example, all to zero. Radar system according to claim 4, wherein before the last step e) a stochastic signal, for. B. from a white noise process, the size f ₄ (n) is added so that the discretization error in step e) is as uncorrelated as possible, so that, for example, artifacts are avoided by periodically recurring conditions. A radar system according to any one of claims 3-4, wherein the high pass filter H _e (z) is non-recursive and has the form H _e (z) = -1 + a ₁ z ^-1 + ... + a _K z ^-K . Radar system according to one of the above claims, wherein the sequence of discrete control signal values s (n) is chosen so that the M-th error sum (M ≥ 1) of the associated stationary oscillator frequency sequence f _the (n) with respect to the frequency _setpoint sequence f _soll (n) is minimal , which corresponds to a choice of the high pass H _e (z) = - (1 - z ^-1 ) ^M. Radar system according to one of the above claims, which when determining the sequence of discrete control signal values s (n) the Distortions and in particular the delay due to the analogue low-pass filtering considered. Radar system according to one of the above claims, at wherein - the Sequence of the discrete control signal values s (n) through a digital-to-analogue converter is generated and - at the determination of the sequence s (n) delays in switching behavior of the digital-to-analog converter (eg different switching sides when setting and resetting single bits). Radar system according to one of the above claims, at which the modulation of the transmission frequency very short linear ramps having. Radar system according to one of the above claims, at which for each discrete control signal value used for frequency modulation the associated stationary oscillator frequency is measured and this survey at least the following steps includes: - Scanning the oscillator signal or a frequency division obtained therefrom Signal, optionally after appropriate preprocessing, during the respective discrete control signal value is applied and the oscillator is in the steady state, - Windowing of the scanned Signals and - Frequency determination for the windowed signal through a high-resolution spectral analysis. Method for controlling frequency ramps in one Radar system for environment detection with - Sending means to the directed Emission of transmission power, - Receiving means to the directed Reception of transmitted power reflected on objects and - Signal processing means for processing the received power and for controlling the transmission means, in which c) the frequency of the transmit power using a controllable Oscillator is modulated, wherein for the frequency of a desired course is given and d) a control signal from the oscillator a sequence of discrete control signal values followed by analogue Low pass filtering is generated, with only a finite set of discrete Control signal values available stands, what to Diskretisierungsfehlern and thus to a deviation between a real thing Actual and the nominal course of the frequency leads, characterized, that the sequence of discrete control signal values and the analog Low pass are designed, - that is in front of the analog Filtering discretization error spectrally the passband range of the analogue lowpass or subranges of which is minimized. The method of claim 12, wherein a sum of the Discretization error to a plurality of control signal values is minimized.