RADAR DEVICE WITH PHASE NOISE ESTIMATION

Abstract

A method for estimating phase noise of an RF oscillator signal in a frequency-modulated continuous-wave (FMCW) radar system and related radar devices are provided. The method includes applying the RF oscillator signal to an artificial radar target composed of circuitry, which applies a delay and a gain to the RF oscillator signal, to generate an RF radar signal. Furthermore, the method includes down-converting the RF radar signal received from the artificial radar target from an RF frequency band to a base band, digitizing the down-converted RF radar signal to generate a digital radar signal, and calculating a decorrelated phase noise signal from the digital radar signal. A power spectral density of the decorrelated phase noise is then calculated from the decorrelated phase noise signal, and the power spectral density of the decorrelated phase noise is converted into a power spectral density of the phase noise of an RF oscillator signal.

Claims

1. A method for estimating phase noise of radio frequency (RF) oscillator signal in a frequency-modulated continuous-wave (FMCW) radar system, the method comprising: applying the RF oscillator signal to an artificial radar target composed of circuitry, which applies a delay and a gain to the RF oscillator signal, to generate a RF radar signal; down-converting the RF radar signal received from the artificial radar target from a RF frequency band to a base band; digitizing the down-converted RF radar signal to generate a digital radar signal; calculating a decorrelated phase noise signal from the digital radar signal; calculating a power spectral density of the decorrelated phase noise from the decorrelated phase noise signal; and converting the power spectral density of the decorrelated phase noise into a power spectral density of the phase noise of the RF oscillator signal.

2. The method of claim 1, wherein calculating the decorrelated phase noise signal from the digital radar signal is done in accordance with the equation: ${}_O\left[n\right]\frac{\frac{A_O}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right)-\frac{y_O\left[n\right]}{A^2}}{\frac{A_O}{2}.Math.\sin \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right)},$ wherein n is a time index, T.sub.A is a sampling period, .sub.O is the decorrelated phase noise signal, y.sub.O [n] is digital radar signal, A is the amplitude of the RF oscillator signal, A.sub.O is the gain of the artificial radar target, f.sub.BO is a beat frequency and .sub.O is a phase offset.

3. The method of claim 1, wherein converting power spectral density of the decorrelated phase noise into the power spectral density of the phase noise of an RF oscillator signal is done in accordance with the equation: $S_{,}\left(f\right)=\frac{S_{{}_O,{}_O}\left(f\right)}{2.Math.\left(1-\cos \left(2.Math..Math..Math.f.Math..Math.{}_O\right)\right)},$ wherein S.sub.,(f) is the power spectral density of the phase noise, S.sub..sub.O.sub.,.sub.O(f) is the power spectral density of the decorrelated phase noise and .sub.O is the delay of the artificial radar target.

4. The method of claim 1, wherein the RF oscillator signal is supplied to at least one antenna to be radiated as electromagnetic radar signal.

5. The method of claim 1, wherein the RF oscillator signal is a sequence of chirps and signal parameters of the RF oscillator signal include a start frequency, a bandwidth, and a duration of the chirps.

6. The method of claim 1, wherein calculating the decorrelated phase noise signal comprises: calculating an estimation of the decorrelated phase noise signal dependent on the gain and the delay of the artificial radar target and dependent on signal parameters of the RF oscillator signal.

7. The method of claim 6, wherein the signal parameters of the RF oscillator signal include a start frequency, a bandwidth, and a duration of the chirps.

8. A method for estimating phase noise of a radio frequency (RF) oscillator signal in an a frequency-modulated continuous-wave (FMCW) radar system; the method comprising: applying the RF oscillator signal to an artificial radar target composed of circuitry, which applies a delay and a gain to the RF oscillator signal, to generate a RF radar signal; down-converting the RF radar signal received from the artificial radar target from a RF frequency band to a base band; digitizing the down-converted RF radar signal to generate a digital radar signal; calculating a power spectral density of the digital radar signal; calculating a power spectral density of a deterministic summand of the digital radar signal; and calculating a power spectral density of the phase noise of the RF oscillator signal based on the power spectral density of the digital radar signal and the power spectral density of the deterministic summand.

9. The method of claim 8, wherein calculating a power spectral density of a deterministic summand of the digital radar signal comprises using Welch's method.

10. The method of claim 8, wherein calculating the power spectral density of the phase noise of the RF oscillator signal comprises calculating a difference between the power spectral density of the digital radar signal and the power spectral density of a deterministic summand to obtain a power spectral density of a stochastic summand of the digital radar signal.

11. The method of claim 10, further comprising converting the power spectral density of the deterministic summand into the power spectral density of the phase noise of the RF oscillator signal.

12. A radar device comprising: a local oscillator configured to generate a radio frequency (RF) oscillator signal, which includes phase noise; an artificial radar target composed of circuitry, which is configured to apply a delay and a gain to the RF oscillator signal, to generate a RF radar signal; a first frequency conversion circuit configured to down-convert the RF radar signal received from the artificial radar target from a RF frequency band to a base band; an analog-to digital conversion unit configured to digitize the down-converted RF radar signal to generate a digital radar signal; and a signal processing unit configured to: calculate a decorrelated phase noise signal from the digital radar signal, calculate a power spectral density of the decorrelated phase noise from the decorrelated phase noise signal, and convert the power spectral density of the decorrelated phase noise into a power spectral density of the phase noise of an RF oscillator signal.

13. A radar device comprising: a local oscillator configured to generate a radio frequency (RF) oscillator signal, which includes phase noise; an artificial radar target composed of circuitry, which is configured to apply a delay and a gain to the RF oscillator signal, to generate a RF radar signal; a first frequency conversion circuit configured to down-convert the RF radar signal received from the artificial radar target from a RF frequency band to a base band; an analog-to digital conversion unit configured to digitize the down-converted RF radar signal to generate a digital radar signal; and a signal processing unit configured to: calculate a power spectral density of the digital radar signal, calculate a power spectral density of a deterministic summand of the digital radar signal, and calculate a power spectral density of the phase noise of the RF oscillator signal based on the power spectral density of the digital radar signal and the power spectral density of the deterministic summand.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The invention can be better understood with reference to the following drawings and descriptions. The components in the figures are not necessarily to scale; in-stead emphasis is placed upon illustrating the principles of the invention. More-over, in the figures, like reference numerals designate corresponding parts. In the drawings:

[0011] FIG. 1 is a schematic diagram illustrating the operating principle of a FMCW radar sensor with a single radar target in a transmission channel according to one or more embodiments;

[0012] FIG. 2 illustrates the waveform of the transmitted and reflected radar signals in the radar sensor of FIG. 1 according to one or more embodiments;

[0013] FIG. 3 is a block diagram illustrating the function of the radar sensor of FIG. 1 according to one or more embodiments;

[0014] FIG. 4 is a simplified block diagram representing a function of a FMCW radar sensor according to one or more embodiments;

[0015] FIG. 5 is a schematic diagram illustrating a cause and origination of leakage by reflection at a short range target according to one or more embodiments;

[0016] FIG. 6 is a block diagram illustrating a radar sensor with noise cancellation in accordance with one or more embodiments;

[0017] FIG. 7 is a diagram illustrating a decorrelated phase noise for different delay times according to one or more embodiments;

[0018] FIG. 8 is a diagram illustrating a cross-correlation coefficient between a decorrelated phase noise included in short-range leakage and a decorrelated phase noise included in a signal obtained from an artificial on-chip target according to one or more embodiments;

[0019] FIG. 9 is a flow chart illustrating noise cancellation in accordance with one or more embodiments;

[0020] FIG. 10 is a diagram illustrating one example of the calculation of the phase noise of a local oscillator according to one or more embodiment; and

[0021] FIG. 11 is a diagram illustrating another example of the calculation of the phase noise of a local oscillator according to one or more embodiments.

DETAILED DESCRIPTION

[0022] FIG. 1 illustrates a conventional frequency-modulated continuous-wave (FMCW) radar system 100. In the present example, separate transmit (TX) and receive (RX) antennas 101 and 102, respectively, are used (bistatic or pseudo-monostatic radar configuration). However, it shall be understood that a single antenna can be used so that the receive antenna and the transmit antenna are physically the same (monostatic radar configuration). The transmit antenna continuously radiates a sinusoidal RF signal s.sub.RF(t), which is frequency-modulated, for example, by a saw-tooth signal (periodic linear ramp signal, see also FIG. 2). The transmitted signal s.sub.RF(t) is back-scattered at a target T.sub.1, which is located within the measurement range of the radar system and received by receive antenna 102. The received signal is denoted as y.sub.RF(t). In the radar device 100, the received signal y.sub.RF(t) is demodulated by mixing the signal y.sub.RF(t) with a copy of the transmit signal s.sub.RF(t) to effect a down-conversion of the RF signal y.sub.RF(t) into the base band. This down-conversion is illustrated in FIG. 2. The received RF signal y.sub.RF(t) lags behind the transmit signal s.sub.RF(t) due to the time taken for the signal to travel to and from the target T.sub.1. As a consequence, there is a constant frequency difference between the received RF signal y.sub.RF(t) and the reference signal (i.e. the copy of the transmit signal s.sub.RF(t)). When the two signals s.sub.RF(t) and y.sub.RF(t) are mixed (i.e. demodulated), a demodulated signal y(t) of constant frequency (in case of a linear frequency modulation) is obtained (also referred to as beat frequency). The beat frequency of the received and demodulated signal y(t) can be determined (e.g. using Fourier analysis) and used to calculate the distance between the radar device 100 and the target T.sub.1.

[0023] The radar device 100 may include or be implemented in a monolithic microwave integrated circuit (MMIC), which includes circuitry for providing the core functions needed for distance and/or velocity measurement in one chip (also referred to as single chip radar). Thus the chip may include, inter alia, RF oscillators, amplifiers, mixers, filters, analog-to-digital converters, and digital signal processors. FIG. 3 illustrates the transmit path and the receive path of a radar transceiver, which may be used for distance measurement in a radar device 100 shown in FIG. 1. Accordingly, the RF transceiver 1 includes a mixer 110, which is supplied with radar signal y.sub.RF(t) and with RF oscillator signal s.sub.RF(t) used to down-convert the radar signal y.sub.RF(t) into the base band. The radar signal y.sub.RF(t) (i.e. a back scattered portion of the transmit signal s.sub.RF(t)) is received by receive antenna 102 and may be pre-amplified (see RF amplifier 105, e.g. a low noise amplifier LNA) before being supplied to the mixer 110. In the present example, the RF oscillator signal s.sub.RF(t) is generated by a local oscillator (LO) 103, which may include a voltage controlled oscillator (VCO) coupled in a phase locked loop (PLL). However, the RF oscillator signal s.sub.RF(t) may be provided by other circuitry dependent on the actual application. When used in a radar distance measurement device, the RF oscillator signal s.sub.RF(t) may be in the range between approximately 24 GHz and 81 GHz (approximately 77 GHz in the present example). However, higher or lower frequencies may also be applicable. The RF oscillator signal s.sub.RF(t) is also supplied to transmit antenna 101 (e.g. via power amplifier 104) and radiated towards the radar target (see also FIG. 1).

[0024] As mentioned, the mixer 110 down-converts the radar signal (amplified antenna signal A y.sub.RF(t), amplification factor A) into the base band. The respective base band signal (mixer output signal) is denoted by y(t). The base band signal y(t) is then subject to analog filtering (filter 115) to suppress undesired sidebands or image frequencies, which may be a result of the mixing operation. The filter 115 may be a low-pass filter or a band-pass filter. The filtered base band signal (filter output signal) is denoted by y(t). Receivers (e.g. the receiver portions of transceivers) which make use of a mixer to down-convert the received RF signal into the base band are as such known as heterodyne receivers and thus not further discussed in more detail. The filtered base band signal y(t) is then sampled (temporal discretization) and converted to a digital signal y[n] (analog-to-digital converter (ADC) 120), which is then further processed in the digital domain using digital signal processing (n being the time index). The digital signal processing may be performed in a digital signal processing unit 125, which may include, e.g., a digital signal processor (DSP) executing appropriate software instructions.

[0025] FIG. 3 illustrates the receive path of a radar transceiver 100 of a so-called bistatic or pseudo-monostatic radar system, in which the receiver may be separate from the transmitter (as receiver and transmitter use separate antennas). In the present example, the transmitter and the receiver portion of the radar transceiver are, however, integrated in one MMIC. In a monostatic radar system, the same antenna is used to transmit and receive RF radar signals. In such cases, the radar transceiver additionally includes a directional coupler or a circulator (not shown) coupled between the mixer for separating the RF transmit signal s.sub.RF(t) from the received signal y.sub.RF(t).

[0026] The transmission channel 200 represents the signal path from the transmit antenna 101 to the target and back to the receive antenna 102. While passing through the transmission channel the radar signals s.sub.RF(t) (transmitted signal) and y.sub.RF(t) (back-scattered signal) are subject to additive noise w(t), which is usually modelled as additive white Gaussian noise (AWGN). FIG. 4 is a simplified block diagram illustrating the analog frontend of the radar transceiver shown in FIG. 3. To allow a simple and clear illustration, antennas and amplifiers have been omitted. Accordingly, the RF transmit signal s.sub.RF(t), which may be generated by local oscillator 103, is sent through transmission channel 200 and finally arrives (as received radar signal y.sub.RF(t)) at the RF input of mixer 110, which is configured to down-convert the radar signal y.sub.RF(t) into the base band. The resulting base band signal y(t) (beat signal) is low-pass filtered (low-pass filter 115), and the filtered base band signal y(t) is then digitized using analog-to-digital converter 120. Band-pass filtering may also be applicable instead of low-pass filtering. The digitized base band signal y[n] is then further processed digitally to estimate the distance between the transceiver 100 and the target. As mentioned additive white Gaussian noise is added to the radar signal while passing through the transmission channel 200.

[0027] FIG. 5 is basically the same illustration as shown in FIG. 1 but with an additional object Ts located in the transmission channel 200 comparably close to the antennas (e.g., a fixture or a cover mounted in front of the radar antennas). Such objects are herein referred to as short-range targets. A short-range targets is usually located a few centimeters (e.g. less than 50 cm) in front of the radar device (which is less than the lower margin of the measurement range of the radar system) and reflects a portion of the transmit signal s.sub.RF(t) back to the receive antenna 102. As mentioned above, such reflections at short-range targets give rise to a phenomenon referred to as short-range leakage. In the example of FIG. 5, the transmitted RF signal s.sub.RF(t) is back-scattered at target T.sub.1 (which is within the normal measurement range of the radar transceiver) as well as reflected at the short-range target Ts. The signal back-scattered from target T.sub.1 is denoted as y.sub.RF,1(t) and the signal reflected at the short-range target Ts is denoted as y.sub.RF,S(t). Both signals y.sub.RF,S(t) and y.sub.RF,S(t) superpose and the resulting sum signal y.sub.RF(t) is received by the antenna 102. Considering the fact that the received signal power decreases with the fourth power of the distance, the signal amplitude of the radar signal y.sub.RF,S(t) due to short-range leakage is significant. Furthermore, the phase noise of the transmitted radar signal s.sub.RF(t) is the dominant cause of noise in the received radar signal y.sub.RF(t) as a result of the short-range leakage.

[0028] FIG. 6 is a block diagram of a radar transceiver in accordance with one exemplary embodiment, which is configured to cancel short-range leakage and thus the mentioned phase noise from the received radar signal using digital signal processing in the base band and an artificial radar target 300 (further referred to as on-chip target or OCT). Again, antennas and amplifiers have been omitted in the illustration for the sake of simplicity and clarity. The transmit signal s.sub.RF(t) is a frequency-modulated continuous-wave (FMCW) signal (chirp-signal), also referred to as chirp signal. Accordingly, the signal s.sub.RF (t) can be written as:

s.sub.RF(t)=cos(2f.sub.Ot+kt.sup.2+(t)+),(1)

wherein f.sub.0 is the start frequency of the chirp signal, k (k=B/T) denotes the slope of the chirp with bandwidth B and duration T, is a constant phase offset and (t) is the introduced phase noise (PN) due to imperfections of the local oscillator (see FIG. 3).

[0029] The transmission channel 200 (see FIGS. 5 and 6) comprises of two types of signal reflections. Firstly, reflections (back-scattering) at targets T.sub.i, whose distances from the radar transceiver are to be measured. These targets T.sub.i are modeled by a delay .sub.Ti and gain A.sub.Ti, wherein i=1, . . . , N.sub.T, and N.sub.T denotes the number of targets T.sub.i (not including the short-range target). Secondly, the reflection at a short-range target T.sub.S, which represents the undesired near target causing reflections (short-range leakage) which are to be cancelled. Analogously to a normal target the short-range target may be modeled by a delay .sub.S and gain A.sub.S. In practice, the gain A.sub.S will be significantly larger than any of the gains A.sub.Ti. This model of the transmission channel 200 is depicted in the upper signal path of the block diagram of FIG. 6. At the receiver side, additive white Gaussian noise (AWGN) w(t) is added before down-conversion to the base band is done. Consequently, the received RF radar signal y.sub.RF(t) may be written as:

y.sub.RF(t)=A.sub.S.Math.S.sub.RF(t.sub.S)+.sub.i=1.sup.N.sup.TA.sub.Ti.Math.s.sub.RF(t.sub.Ti)+w(t),(2)

wherein the first summand represents the signal component due to the short-range leakage, the second summand represents the signal components due to reflections at the normal radar target(s) and the last summand represents AWGN. The delays .sub.S and .sub.Ti are also referred to as round trip delay times (RTDT) associated with the short-range target T.sub.S and the targets T.sub.i, respectively. It should be noted that, in the present disclosure, the previously mentioned on-chip leakage is not considered as several concepts for cancelling on-chip leakage exist.

[0030] As can be seen from FIG. 6, the received radar signal is subject to a down-conversion using the mixer 110 and a subsequent band-pass or low-pass filtering using the filter 115, which has a filter impulse response h.sub.F(t). As in the previous illustrations, the down-converted and filtered signal is denoted as y(t), which can be modelled as follows (assuming =0 for the sake of simplicity):

[00001]$\begin{array}{cc}\begin{array}{c}{y}^{}\left(t\right)=.Math.\left(s_{\mathrm{RF}}\left(t\right).Math.y_{\mathrm{RF}}\left(t\right)\right)*h_F\left(t\right)=\\ =.Math.\frac{A_S}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BS}}.Math.t+{}_S+\left(t\right)-\left(t-{}_S\right)\right)+\\ .Math.\underset{i=1}{\overset{N_T}{.Math.}}.Math.\frac{A_{\mathrm{Ti}}}{2}.Math.\cos \left(2.Math..Math..Math.f_{{\mathrm{BT}}_i}.Math.t+{}_{T_i}+\left(t\right)-\left(t-{}_{T_i}\right)\right)+w\left(t\right).\end{array}& \left(3\right)\end{array}$

The beat frequencies resulting from the short-range leakage and the reflections at the normal targets are denoted as f.sub.BS and f.sub.BT.sub.i (target T.sub.i), respectively, and can be represented by the following equations:

f.sub.BS=k.sub.S, and f.sub.BT.sub.i=k.sub.T.sub.i.(4)

Furthermore, the constant phase .sub.S and .sub.T.sub.i can be computed as:

.sub.S=2f.sub.0.sub.Sk.sub.S.sup.2, and .sub.T.sub.i=2f.sub.O.sub.T.sub.ik.sub.T.sub.i.sup.2.(5)

The beat frequencies (equations 4) and constant phases (equations 5) depend only on given system parameters (such as the start frequency f.sub.0 of the chirp as well as its bandwidth and duration as represented by the variable k=B/T) and the RTDTs .sub.S and .sub.Ti associated with the short-range leakage and the radar targets T.sub.i to be detected, respectively. It follows from equations 3, 4 and 5 that the signal component of y(t), which results from the short-range leakage (i.e. the first summand in equation 3), is zero when the RTDT .sub.S is zero (.sub.S=0). Even the term (t)(t.sub.S) becomes zero when the delay time .sub.S is zero. With increasing values of the RTDT .sub.S (i.e. with increasing distance of the short-range target) the correlation of the phase noise components (t) and (t.sub.S) decreases. This effect is called range correlation effect and the phase difference (t)(t.sub.S) is referred to as decorrelated phase noise DPN. It is noted that DPN is usually not an issue in the context of on-chip leakage as the associated delay is negligibly small.

[0031] In the following, the first summand of equation 3, i.e. the short-range leakage signal

[00002]$\begin{array}{cc}{y}_S^{}\left(t\right)=\frac{A_S}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BS}}.Math.t+{}_S+\left(t\right)-\left(t-{}_S\right)\right)& \left(6\right)\end{array}$

is analyzed in more detail (see FIG. 6). In equation 6, the gain A.sub.S/2 is primarily determined by the radar cross section (RCS) of the short-range target. Generally, the RCS may depend on the shape and the material of the short-range target. The beat frequency Gs (see equation 4) depends on the RTDT .sub.S associated with the short-range target. The RTDT .sub.S depends on the distance d.sub.S between the radar device and the short-range target. Accordingly, the distance d.sub.S can be calculated as d.sub.S=c.Math..sub.S/2, wherein c denotes the speed of light. In equation 6, the DPN (t)(t.sub.S) represents noise in addition to the mentioned AWGN. To analyze how the DPN affects the spectrum of the received radar signal, the power spectrum S.sub.S,S (f) of the DPN is calculated:

S.sub.S,S(f)=S.sub.,(f).Math.2(1cos(2.sub.Sf)),(7)

wherein S.sub.,(f) is the power spectrum of the phase noise signal (t) included in the RF transmit signal s.sub.RF(t). Further analysis of a realistic example (.sub.S=800 ps, d.sub.S12 cm) shows that, for frequencies higher than 100 kHz, the noise level of the DPN is 140 dBm/Hz, assuming a transmit power of 10 dBm and an AWGN noise floor of 140 dBm/Hz. The presence of DPN entails an increase of the noise floor and results in a 10 dB reduction of sensitivity for the detection of radar targets. As a result, the total noise floor increases, which is equivalent to a loss of sensitivity of 10 dB for the detection of radar targets.

[0032] To at least reduce the effect of the DPN due to (unavoidable) short-range targets an (artificial) on-chip target (OCT) is included in the radar device and incorporated in the signal processing chain as illustrated in FIG. 6. The OCT is used to obtain an estimation of the DPN and to (at least partially) cancel the DPN from the received radar signal in the base band. As can be seen from FIG. 6, the RF transmit signal s.sub.RF(t) is (in addition to being radiated to the radar channel 200) supplied to OCT 300 that is basically composed of a gain A.sub.O (A.sub.O<1) and a delay .sub.O, which can be seen as an on-chip RTDT. The RF signal received from OCT 300 is denoted as y.sub.RF,O(t). This signal y.sub.RF,O(t) is down-converted into the base band (mixer 110) and band-pass filtered (filter 115) in the same manner as the RF signal y.sub.RF(t) received from the radar channel 200. The down-converted signal received from OCT 300 is denoted as y.sub.O(t) and the respective band-pass (or low-pass) filtered signal is denoted as y.sub.O(t). Both, the filtered base band signal y(t) received from radar channel 200 and the filtered base band signal y.sub.O(t) received from OCT 300 are digitized using analog-to-digital converters 120 and 120, respectively, for further digital signal processing. In another embodiment a single analog-to-digital converter and a multiplexer may be used to provide the same function. The respective digital signals are denoted as y[n] and y.sub.O[n].

[0033] Theoretically, it would be desirable that the delay .sub.O of OCT 300 equals the RTDT .sub.S of the short-range target Ts present in radar channel 200. In realistic examples the RTDT .sub.S of the short-range target T.sub.S is in the range of a few hundreds of picoseconds up to a few nanoseconds, whereas the delay .sub.O of an on-chip target is practically limited to a few picoseconds when implementing the radar device on a single MMIC. In a single-chip radar higher values of delay .sub.O (which would be needed in case of .sub.O=.sub.S) would result in an undesired (or even unrealistic) increase in chip area and power consumption and are thus only economically feasible when using discrete circuit components. Therefore, the delay .sub.O of OCT 300 is limited to values that are significantly lower than the RTDT .sub.S of any practically relevant short-range target T.sub.S.

[0034] Further analysis of the properties of the cross-correlation coefficient of the decorrelated phase noise (DPN) signals

.sub.S(t)=(t)(t.sub.S),(8)

i.e. the DPN included in the RF signal received from the short-range target Ts (see FIGS. 5 and 6), and

.sub.O(t)=(t)(t.sub.O)(9)

i.e. the DPN included in the RF signal received from OCT 300, shows that the cross-correlation coefficient

[00003]$\begin{array}{cc}{}_{.Math..Math.{}_O,{}_S}\left(l\right)=\frac{E.Math.\left\{.Math..Math.{}_O\left(t\right).Math.{}_S\left(t-l\right)\right\}}{\sqrt{{}_{.Math..Math.{}_O}^2}.Math.\sqrt{{}_{.Math..Math.{}_S}^2}}& \left(10\right)\end{array}$

is very similar for different values of OCT delay .sub.O (the operator E denoting the expected value and .sub..sub.O.sup.2 and .sub..sub.S.sup.2 are the respective variances). Note that the DPN terms are assumed to have a mean value of zero. For an OCT delay .sub.O equal to the RTDT r.sub.S, the cross-correlation coefficient assumes a maximum for a time lag l of zero (l=0). For smaller values of .sub.O (i.e. .sub.O<.sub.S) the cross-correlation coefficient is scaled and shifted as compared to the case when .sub.O=.sub.S. This result is illustrated in the diagrams of FIGS. 7 and 8.

[0035] FIG. 7 illustrates exemplary realizations of a DPN signal (t)=(t)(t) for different delay times . The DPN signals (t) shown in FIG. 7 (for =40 ps, =160 ps, =400 ps, and =800 ps) have been obtained by simulating the phase noise (t) using a stochastic model, which models the phase noise of the local oscillator (see FIG. 3, LO 103). It can be seen from FIG. 7 that the waveforms of the resulting DPN signals are very similar, even when the delay time r is different. In this context similar means that one waveform (e.g. for =40 ps) can be transformed into any other waveform (e.g. the waveform for =800 ps) by applying a gain and a time-shift (equivalent to phase-shift). This fact can also be observed in the cross-correlation coefficient shown in FIG. 8. Equation 10 has been estimated with a discrete-time simulation, wherein the expected value (operator E) has been approximated over a representative length of the random signals (obtained using the mentioned stochastic model) representing phase noise signal (t).

[0036] As the DPN .sub.O(t) included in the down-converted RF signal

[00004]$\begin{array}{cc}{y}_O\left(t\right)=\frac{A_O}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.t+{}_O+\left(t\right)-\left(t-{}_O\right)\right),& \left(11\right)\end{array}$

which is received from OCT 300, and the DPN .sub.S(t) included in the baseband signal y.sub.S(t) which is received from the short-range target (see equation 6), are highly correlated, the DPN included in the baseband signal y.sub.O(t) obtained from OCT 300 can be used to estimate the DPN caused by the short-range leakage. In equation 11, f.sub.BO denotes the beat frequency caused by OCT 300 and is calculated analogously to f.sub.BS (see equation 4). Also the constant phase .sub.O is computed in an analogous manner as constant phase .sub.S (see equations 5 and 14). In a practical example, the RTDT .sub.S associated with the short-range target Ts is approximately 800 ps (corresponds to d.sub.S=12 cm), whereas the OCT delay time .sub.O is only 40 ps. Therewith, the beat frequency f.sub.BS is 20 times higher than beat frequency f.sub.BO.

[0037] As can be seen from FIG. 6, the sampling clock signal, which triggers the sampling of the upper signal path (i.e. the sampling of signal y(t) received from channel 200), is delayed by a time offset T.sub.A. This time offset of the sampling clock signal may be chosen equal to the time lag l, at which the cross-correlation coefficient (see equation 10 and FIG. 8) has its maximum for a specific RTDT .sub.O, wherein .sub.O<.sub.S. Further analysis of the cross-correlation coefficient shows that the optimum sampling time offset T.sub.A is equal to half of the difference .sub.S.sub.O, that is

[00005]$\begin{array}{cc}.Math..Math.T_A=\frac{{}_S-{}_O}{2}.& \left(12\right)\end{array}$

Using the mentioned sampling time offset for maximization of the correlation coefficient results in a high correlation coefficient .sub..sub.O.sub.,.sub.S(0) of, for example, 0.9 for .sub.S=800 ps and .sub.O=40 ps (see diagram of FIG. 8).

[0038] As the DPN signals included in the discrete time signals y[n] and y.sub.O[n] (provided by analog-to-digital converters 120 and 120, respectively) are highly correlated (particularly when using the mentioned sampling time offset), an estimation of the discrete-time DPN signal .sub.O[n] may be calculated from the down-converted signal y.sub.O[n] obtained from OCT 300. This estimation and the subsequent calculation of a corresponding cancellation signal is performed by the function block 130 labelled LC (leakage cancellation). Therefore, the LC function block basically provides the two functions of estimating the DPN from signal y.sub.O[n] and generating a cancellation signal .sub.S [n] to be subtracted from the down-converted and digitized radar signal y[n] in order to eliminate the short-range leakage (see also equation 6) included in the radar signal y[n].

[0039] The discrete-time version of equation 11 is

[00006]$\begin{array}{cc}{y}_O\left[n\right]=\frac{A_O}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O+.Math..Math.{}_O\left[n\right]\right)& \left(13\right)\\ \mathrm{with}& \\ f_{\mathrm{BO}}=k_O,\mathrm{and}.Math..Math.{}_O=2.Math..Math..Math.f_0.Math.{}_O-k.Math..Math.{}_O^2& \left(14\right)\end{array}$

wherein f.sub.A is the sampling rate determined by the period T.sub.A of the sampling clock signal (f.sub.A=T.sub.A.sup.1). Applying the trigonometric identity

cos(a+b)=cos(a)cos(b)+sin(a)sin(b)(15)

and the approximations (since .sub.O[n] is sufficiently small)

cos(.sub.O[n])1 and(16)

sin(.sub.O[n]).sub.O[n](17)

to equation 13 simplifies it to

[00007]$\begin{array}{cc}{y}_O\left[n\right]\frac{A_O}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right)-\frac{A_O}{2}.Math.\sin \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right).Math.{}_O\left[n\right]& \left(18\right)\end{array}$

[0040] As the gain A.sub.O and the beat frequency f.sub.BO are a-priori known system parameters of the radar system the DPN .sub.O[n] can be approximated based on the down-converted signal y.sub.O [n], which is received from the OCT, in accordance with the following equation:

[00008]$\begin{array}{cc}{}_O\left[n\right]\frac{\frac{A_O}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right)-y_O\left[n\right]}{\frac{A_O}{2}.Math.\sin \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right)}& \left(19\right)\end{array}$

[0041] Beat frequency f.sub.BO and phase .sub.O may be measured after production of the radar device as a part of a system test and calibration procedure. These parameters can be computed in the same manner as for the short-range leakage signal y.sub.S[n] (see equations 4 and 5 and equation 14). In order to account for parameter variations of OCT 300 (e.g. due to temperature changes) beat frequency f.sub.BO and phase .sub.O may be estimated repeatedly and updated regularly.

[0042] As the DPN signals .sub.O[n] and .sub.S[n] are highly correlated, the short-range leakage signal (cf. equation 6)

[00009]$\begin{array}{cc}{y}_S\left[n\right]=\frac{A_S}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BS}}.Math.{\mathrm{nT}}_A+{}_S+{}_S\left[n\right]\right)& \left(20\right)\end{array}$

[0043] can be approximated as

[00010]$\begin{array}{cc}{\hat{y}}_S\left[n\right]=\frac{{\hat{A}}_S}{2}.Math.\cos \left(2.Math..Math..Math.{\hat{f}}_{\mathrm{BS}}.Math.{\mathrm{nT}}_A+{\hat{}}_S+{}_L.Math.{}_O\left[n\right]\right),& \left(21\right)\end{array}$

[0044] where .sub.L is referred to as DPN gain. Gain .sub.L can be determined with the help of the auto-covariance function

c.sub..sub.OL.sub.,.sub.OL(l)=E{.sub.OL(t).sub.OL(tl)},(22)

[0045] where .sub.OL(t)=.sub.O(t)*h.sub.L(t), i.e. the convolution with the impulse response of the lowpass filter 115, and the cross-covariance function

c.sub..sub.OL.sub.,.sub.SL(l)=E{.sub.OL(t).sub.SL(tl)}(23)

with .sub.SL(t)=.sub.S(t)*h.sub.L(t). The DPN gain .sub.L can then be determined as

[00011]$\begin{array}{cc}{}_L=\frac{c_{{}_{\mathrm{OL}},{}_{\mathrm{SL}}}\left(-T_A\right)}{c_{{}_{\mathrm{OL}},{}_{\mathrm{OL}}}\left(0\right)}.& \left(24\right)\end{array}$

Note that the numerator equals equation 23 (resulting in .sub.L=1) when .sub.O=.sub.S (see also FIG. 8, in which the cross-correlation coefficient has a maximum of 1 for .sub.O=.sub.S and maxima lower than 1 for .sub.O<.sub.S). Therewith, .sub.L is a measure of how much the DPN of the OCT needs to be amplified such that it approximates the DPN of the SR leakage. For example, with a typical phase noise power spectrum, .sub.S=800 ps and .sub.O=40 ps results in a DPN gain of .sub.L=19.8. The parameters {circumflex over ()}.sub.S and .sub.S are specific to the radar system (i.e. dependent on the short-range target) and can be calculated or determined by calibration.

[0046] The estimated short-range leakage signal .sub.S[n] is generated by the LC function block 130 illustrated in FIG. 6. The actual noise cancellation is accomplished by subtracting the estimated short-range leakage signal .sub.S[n] from the signal y[n] received from the radar channel. The DPN compensated signal is denoted as z[n] and is calculated as:

z[n]=y[n].sub.S[n].(25)

The cancellation method is summarized in the flow-chart of FIG. 9. As compared to a known radar system the RF transmit signal s.sub.RF(t) is transmitted to an on-chip target (OCT) 300 (see step 701). The signal y.sub.RF,O(t) received from OCT 300 down-converted to the base band (base band signal y.sub.O(t), step 702) and digitized (digital base band signal y.sub.O[n], step 703). The decorrelated phase noise (DPN) signal .sub.O[n] is estimated from digitized signal y.sub.O(t), and a corresponding cancellation signal .sub.S [n] is generated based on the estimated DPN signal .sub.O[n] (step 704). Finally, the cancellation signal is subtracted from the (down-converted and digitized) radar echo signal y[n] in order to compensate for the short-range leakage included therein.

[0047] In some radar systems it may be of interest to measure the phase noise (PN) (t) of the local oscillator 103 (see FIG. 1 and equation 1). As shown below, the power spectrum of the phase noise (t) can be derived from the DPN .sub.O[n] (see equation 19). In equation 1, the amplitude A of the oscillator signal has been assumed to be one (A=1). For an arbitrary signal amplitude A of the oscillator signal s.sub.RF(t) equation 19 can be written as:

[00012]$\begin{array}{cc}.Math..Math.{}_O\left[n\right]\frac{\frac{A_O}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right)-\frac{y_O\left[n\right]}{A^2}}{\frac{A_O}{2}.Math.\sin \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right)}.& \left(26\right)\end{array}$

Accordingly, the power spectral density (PSD) can be estimated from the time domain DPN signal .sub.O[n], which is extracted from the signal y.sub.O[n] received from the OCT 300. For an arbitrary signal amplitude A equation 13 can be written as:

[00013]$\begin{array}{cc}{y}_O\left[n\right]=\frac{A^2.Math.A_O}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O+{}_O\left[n\right]\right).& \left(27\right)\end{array}$

It should be noted that this signal is readily available in an FMCW radar transceiver, which implements the short-range leakage cancelation concept as shown, for example, in FIG. 6. In order to determine the PSD of the phase noise (PN) (t) the spectral properties of the DPN are investigated more closely.

[0048] The auto-covariance function of the DPN is:

c.sub..sub.O.sub.,.sub.O(u)=E{.sub.O(t).sub.O(t+u)}(28)

which can be expanded to:

c.sub..sub.O.sub.,.sub.O(u)=E{(t)(t+u)}E{(t)(t+u.sub.O)}E{(t.sub.O)(t+u)}+E{(t.sub.O)(t+u.sub.O)}.(29)

The term E{(t)(t+u)} is the auto-covariance c.sub.,(u) of the phase noise (t) and thus equation 29 can be written as:

c.sub..sub.O.sub.,.sub.O(u)=2c.sub.,(u)c.sub.,(u.sub.O)c.sub.,(u+.sub.O).(30)

Finally, the PSD S.sub..sub.O.sub.,.sub.O(f) of the DPN can be calculated from equation 30 using the Wiener-Khinchine theorem as follows:

[00014]$\begin{array}{cc}\begin{array}{c}{S}_{{}_O,{}_O}\left(f\right)=.Math.\left(c_{{}_O,{}_O}\left(u\right)\right)=\\ =2.Math.S_{,}\left(f\right)-S_{,}\left(f\right).Math.\left({}^{j.Math.2.Math..Math..Math.f.Math..Math.{}_O}+{}^{-j.Math.2.Math..Math..Math.f.Math..Math.{}_O}\right)=\\ =2.Math.S_{,}\left(f\right)-2.Math..Math.S_{,}\left(f\right).Math.\left(1-\cos \left(2.Math..Math..Math.f.Math..Math.{}_O\right)\right),\end{array}& \left(31\right)\end{array}$

wherein denotes the Fourier transform. Rearranging equation 31 results in

[00015]$\begin{array}{cc}{S}_{,}\left(f\right)=\frac{S_{{}_O,{}_O}\left(f\right)}{2.Math.\left(1-\cos \left(2.Math..Math..Math.f.Math..Math.{}_O\right)\right)},& \left(32\right)\end{array}$

wherein S.sub.,(f) is the desired PSD of the phase noise (t).

[0049] The PSD S.sub..sub.O.sub.,.sub.O(f) of the DPN can be estimated from the DPN discrete time domain signal .sub.O[n] as given in equation 26. To obtain the desired PSD of the phase noise (t) equation 8 is evaluated. Therefore only the known system parameter .sub.O (the delay time of the on-chip target 300) as well as A and A.sub.O are needed. It is noted that the resulting PSD S.sub.,(f) of the phase noise (t) is evaluated over the whole chirp bandwidth B rather than at a fixed frequency. The present approach is summarized by the diagram of FIG. 10, which shows a part of the receive signal path of the (down-converted) radar signal y.sub.O(t) received from OCT 300 (see also FIG. 6). As already discussed the signal y.sub.O(t) is filtered, e.g. by band-pass 115, and the filtered signal y.sub.O(t) is digitized, e.g. by ADC 120 to obtain the digital signal y.sub.O[n]. The digital signal y.sub.O[n] can be used for noise cancellation as explained above with reference to FIGS. 6 to 9. In the present example, the DPN .sub.O [n] is calculated (approximated) from the digital signal y.sub.O[n] (function block 401), e.g., in accordance with equation 26. For this, no additional calculations are required if the DPN is calculated in the leakage calculation block 103. From the DPN signal .sub.O[n] the PSD S.sub..sub.O.sub.,.sub.O(f) is calculated by known algorithms (e.g. Welch's method, function block 402), and finally the desired PSD S.sub.,(f) can be calculated from S.sub..sub.O.sub.,.sub.O(f) using equation 32 (function block 403).

[0050] Alternatively, the desired PSD of the phase noise (t) can be derived directly from the digital signal y.sub.O[n] without the need to calculate the DPN as in the previous example. Therefore the digital signal y.sub.O[n] is approximated as shown before in equation 18, which results in:

[00016]$\begin{array}{cc}\begin{array}{c}{y}_O\left[n\right].Math.\frac{A^2.Math.A_O}{2}.Math.\cos \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right)-\frac{a^2.Math.A_O}{2}.Math.\sin \left(2.Math..Math..Math.f_{\mathrm{BO}}.Math.{\mathrm{nT}}_A+{}_O\right).Math.{}_O\left[n\right]\\ =.Math.y_{O.Math..Math.1}\left[n\right]-y_{O.Math..Math.2}\left[n\right].\end{array}& \left(33\right)\end{array}$

The generally time-dependent PSD of y.sub.O[n] can be calculated as the difference:

S.sub.y.sub.O2.sub.,y.sub.O2(f,t)=S.sub.y.sub.O.sub.,y.sub.O(f,t)S.sub.y.sub.O1.sub.,y.sub.O1(f,t)(34)

wherein S.sub.y.sub.O.sub.,y.sub.O(f,t) and S.sub.y.sub.O1.sub.,y.sub.O1(f,t) are the PSDs of the digital signals y.sub.O [n] and y.sub.O1 [n], respectively, which can be approximated by known methods (e.g. Welch's method). By analyzing the time-dependent auto-covariance c.sub.y.sub.O2.sub.,y.sub.O2(t,u) of the second summand y.sub.O2[n], and using the approximation of equation 33 it can be shown that the PSD S.sub.,(f) can be expressed as:

[00017]$\begin{array}{cc}{S}_{,}\left(f\right)\frac{4.Math.\left(S_{y_O,y_O}\left(f,t\right)-S_{y_{O.Math..Math.1},y_{O.Math..Math.1}}\left(f,t\right)\right)}{{\left(A^2.Math.A_O\right)}^2.Math.\left(1-\cos \left(2.Math..Math..Math.f.Math..Math.{}_O\right)\right).Math.\left(1-\cos \left(4.Math..Math..Math.f_{\mathrm{BO}}.Math.t+2.Math.{}_O\right)\right)},& \left(35\right)\end{array}$

wherein the nominator of the fraction is calculated in accordance with equation 34. Similar as mentioned above with regard to equation 32, it is noted that the resulting PSD S.sub.,(f) of the phase noise (t) is evaluated over the whole chirp bandwidth B rather than at a fixed frequency.

[0051] The present approach is summarized with reference of FIG. 11, which is mainly identical with FIG. 10 except that the function blocks 401, 402 and 403 are replaced by function blocks 501 and 502. Function block 501 represents the calculation of the PSD S.sub.y.sub.O.sub.,y.sub.O(f,t) and function block 502 the calculation of the desired PSD S.sub.,(f) of the phase noise (t) in accordance with equations 34 and 35. The PSD S.sub.,(f) can be provided to any internal or external controller, which may control the function of the FMCW radar dependent on the current values of the PSD of the phase noise (t).

[0052] In accordance with a further exemplary embodiment the radar device includes a noise cancellation function. Accordingly, the radar device includes an RF transceiver configured to transmit an RF oscillator signal to a radar channel and receive a respective first RF radar signal from the radar channel, and an artificial radar target composed of circuitry that provides a gain and a delay to the RF oscillator signal to generate a second RF radar signal. A first frequency conversion circuit includes a first mixer configured to down-convert the first RF radar signal; a second frequency conversion circuit includes a second mixer configured to down-convert the second RF radar signal. An analog-to digital conversion unit is configured to digitize the down-converted first RF radar signal and the down converted second RF radar signal to generate a first digital signal and a second digital signal, respectively. A digital signal processing unit receives the first and second digital signals and is configured to: calculate a decorrelated phase noise signal included in the second digital signal, to generate a cancellation signal based on the estimated decorrelated phase noise signal, and to subtract the cancellation signal from the first digital radar signal to obtain a noise compensated digital radar signal. Additionally, the digital signal processing unit is configured to calculate a power spectral density of the decorrelated phase noise from the decorrelated phase noise signal, and to calculate the power spectral density of the decorrelated phase noise into a power spectral density of the phase noise of an RF oscillator signal.

[0053] Moreover, a method for cancelling noise in a radar signal is described. IN accordance with one embodiment, the method comprises transmitting an RF oscillator signal to a radar channel and receiving a respective first RF radar signal from the radar channel, applying the RF oscillator signal to an artificial radar target composed of circuitry, which applies a delay and a gain to the RF oscillator signal, to generate a second RF radar signal. The method further comprises down-converting the first RF radar signal and the second RF radar signal from a RF frequency band to a base band, digitizing the down-converted first RF radar signal and the down-converted second RF radar signal to generate a first digital signal and a second digital signal, respectively, and calculating a decorrelated phase noise signal included in the second digital signal. A cancellation signal is generated based on the decorrelated phase noise signal, and the cancellation signal is subtracted from the first digital radar signal to obtain a noise compensated digital radar signal. Additionally, the method includes calculating a power spectral density of the decorrelated phase noise from the decorrelated phase noise signal, and converting the power spectral density of the decorrelated phase noise into a power spectral density of the phase noise of an RF oscillator signal.

[0054] In a further embodiment, a radar device includes an RF transceiver configured to transmit an RF oscillator signal to a radar channel and receive a respective first RF radar signal from the radar channel and further includes an artificial radar target composed of circuitry that provides a gain and a delay to the RF oscillator signal to generate a second RF radar signal. A first frequency conversion circuit includes a first mixer configured to down-convert the first RF radar signal, and a second frequency conversion circuit includes a second mixer configured to down-convert the second RF radar signal. An analog-to digital conversion unit is configured to digitize the down-converted first RF radar signal and the down converted second RF radar signal to generate a first digital signal and a second digital signal, respectively. Furthermore, a digital signal processing unit of the radar device receives the first and second digital signals and is configured to calculate a decorrelated phase noise signal included in the second digital signal, to generate a cancellation signal based on the estimated decorrelated phase noise signal, and to subtract the cancellation signal from the first digital radar signal to obtain a noise compensated digital radar signal. Additionally, the digital signal processing unit is configured to calculate a power spectral density of the digital radar signal, to calculate a power spectral density of a deterministic summand of the digital radar signal, and to calculate a power spectral density of the phase noise of the RF oscillator signal based on the power spectral density of the digital radar signal and the power spectral density of a deterministic summand.

[0055] Another exemplary method for cancelling noise in a radar signal comprises transmitting an RF oscillator signal to a radar channel and receiving a respective first RF radar signal from the radar channel, and applying the RF oscillator signal to an artificial radar target composed of circuitry, which applies a delay and a gain to the RF oscillator signal, to generate a second RF radar signal. The first RF radar signal and the second RF radar signal are down-converted from a RF frequency band to a base band, and the down-converted first RF radar signal and the down-converted second RF radar signal are digitized to generate a first digital signal and a second digital signal, respectively. The method further comprises calculating a decorrelated phase noise signal included in the second digital signal, generating a cancellation signal based on the decorrelated phase noise signal, and subtracting the cancellation signal from the first digital radar signal to obtain a noise compensated digital radar signal. Additionally a power spectral density of the digital radar signal is calculated, a power spectral density of a deterministic summand of the digital radar signal is calculated, and a power spectral density of the phase noise of the RF oscillator signal is then calculated based on the power spectral density of the digital radar signal and the power spectral density of a deterministic summand.

[0056] Although the invention has been illustrated and described with respect to one or more implementations, alterations and/or modifications may be made to the illustrated examples without departing from the spirit and scope of the appended claims. In particular regard to the various functions performed by the above described components or structures (units, assemblies, devices, circuits, systems, etc.), the terms (including a reference to a means) used to describe such components are intended to correspondunless otherwise indicatedto any component or structure, which performs the specified function of the described component (e.g., that is functionally equivalent), even though not structurally equivalent to the disclosed structure, which performs the function in the herein illustrated exemplary implementations of the invention.

[0057] In addition, while a particular feature of the invention may have been disclosed with respect to only one of several implementations, such feature may be combined with one or more other features of the other implementations as may be desired and advantageous for any given or particular application. Furthermore, to the extent that the terms including, includes, having, has, with, or variants thereof are used in either the detailed description and the claims, such terms are intended to be inclusive in a manner similar to the term comprising.