Professional Documents
Culture Documents
I. I NTRODUCTION
The Linear Frequency Modulated Continuous Wave
...
(LFMCW) radar is commonly used for range detection
in many applications including automotive driver-assistance,
room-occupancy sensing, and industrial automation. The
signal-processing for range estimation using LFMCW mod-
ulation has been shown in [1]. The estimation of a target’s
Fig. 1. FMCW radar transceiver architecture for range-bearing estimation.
bearing angle from a linear array is well known [2]. To jointly
estimate the range and bearing parameters, an LFMCW array
can be used in which the range information is extracted from
individual elements of the array, and the bearing information
is extracted from the phase variation across the array elements
[3], [4]. However the limitations of this technique of range-
(N
-1
)Δ
bearing processing have not been previously investigated.
R
This paper develops an LFMCW radar-signal model to
2Δ
−5 1 (40)
2
−10 In (40) the larger solution of (39) has been discarded since
−15 it yields a non-physical θ. Note that for small [∆τ ]2 (f0
−20 β∆τ ), we can approximate the bearing-angle from (37) as
−25
−1 c 6 YR (γn )
−30 θ ≈ sin (41)
-3π -2π -π 0 π 2π 3π d P
[k − b1 ] L2 B. Phase-comparison by sum-difference
The target’s bearing can be obtained by applying the sum-
Fig. 3. Sum of the weighted derivatives of sinc([k − b1 ] L
2
): difference phase-comparison method to the receiving elements
PM 1 dm L
m=0 m! dk m sinc([k − b 1 ] 2
) of a linear array [6], [8]. For two adjacent receivers in Fig. 2,
the ratio of their difference and sum can be expressed as
Cn+1 Yn+1 (ω) − Cn Yn (ω)
bearing-angles. We now examine the implications of using Y∆Σ (ω) = (42)
the more complete signal representation in (8) on phase- Cn+1 Yn+1 (ω) + Cn Yn (ω)
comparison technique. Evaluating (42) at ω = γn , and applying the conditions of
(35) yields
A. Phase-comparison by ratio (Cn+1 /Cn )Yn+1 (γn ) − Yn (γn )
Y∆Σ (ω) = (43)
Using the scheme of Fig. 1, the phase-difference between ω=γn (Cn+1 /Cn )Yn+1 (γn ) + Yn (γn )
two adjacent receivers can be obtained from the ratio of their
To remove the influence of channel-to-channel gain variations,
output IF signals. Using (8) we can express the ratio of the IF
we divide the first term in the numerator and denominator of
signals as
(43) by the magnitude of (36), hence arriving at a new quantity
Cn+1 Yn+1 (ω) Cn+1
YR (ω) = = × (Cn+1 /Cn )Yn+1 (γn )|YR (γn )|−1 − Yn (γn )
Cn Yn (ω) Cn Ye∆Σ (γn ) =
−jφn+1 (Cn+1 /Cn )Yn+1 (γn )|YR (γn )|−1 + Yn (γn )
F (ω + γn+1 ) + ejφn+1 F (ω − γn+1 )
e
ejφn+1 − ejφn
(34) φn+1 − φn
e−jφn F (ω + γn ) + ejφn F (ω − γn ) = jφn+1 = j tan (44)
e + ejφn 2
Cn accounts for channel-to-channel gain variations; that is, for and
each channel Cn is constant, but may vary among channels. 2 tan−1 (Re{−j Ye∆Σ (γn )}) = φn+1 − φn (45)
Evaluating (34) at ω = γn , and requiring that
The operation Re{-} preserves the sign of the bearing-angle
T T such that negative angles can be unambiguously determined.
[γn − γn+1 ] ≈ 0, [γn + γn+1 ] π, γn T π (35)
2 2 We can obtain θ by comparing (45) and (37), and replacing
yields 6 YR (γn ) in (40) or (41) with 2 tan−1 (Re{−j Y
e∆Σ (γn )}).
Cn+1 j(φn+1 −φn )
YR (ω) = e (36) C. Phase-comparison by cross-correlation
ω=γn Cn
The bearing-angle can be estimated by cross-correlating
γn is the argument-of-the-maximum of the right-hand side the output of two adjacent receiving channels. The Fourier
of (8). Using (9) the phase-angle of (36) can be shown to be transform of the cross-correlated channels can be used to
obtain the target bearing. Using (8) – (10), the transform of
6 YR (γn ) = φn+1 − φn = −πβ(2n + 1)[∆τ ]2 + P ∆τ (37)
the correlated channels can be written as
with YC (ω) = Cn Cn+1 Yn (−ω)Yn+1 (ω)
2
P = πβT + ω0 − 4πβR0 /c AT
= Cn Cn+1 ×
4
βT 4R0
= 2πf0 1 + 1− = 2πf0 Q (38) h i
2f0 cT e−jφn F (ω − γn ) + ejφn F (ω + γn ) ×
In (38) we again see the influence of the fractional bandwidth
h i
e−jφn+1 F (ω + γn+1 ) + ejφn+1 F (ω − γn+1 ) (46)
βT /f0 . If the quantity 6 YR (γn ) has been determined from
the DFT of (2) for two adjacent channels, ∆τ and hence the Using the first condition of (35) and noting that
target’s bearing can be determined from (37) and (6) as F (ω ∓ γn )F (ω ± γn+1 ) ≈ 0, (46) can be simplified as
1 2 h
P ± [P 2 − 4πβ(2n + 1)6 YR (γn )] 2
AT
∆τ = (39) YC (ω) = Cn Cn+1 ej(φn+1 −φn ) F 2 (ω − γn )+
2πβ(2n + 1) 4
0.4 approx.
0.2 exact
1 Range-bearing-map
0 Sum-diff
10-1
-0.2
Mean-square error
10 -2
-0.4
10-3 0 dB SNR
0 10 20 30 40 50 60 70
10-4
Range (m) 10-5
θ = 80°, 30 dB SNR 10-6 0 dB SNR
40 dB SNR
10-7 ≥ 20 dB SNR
percent angle error
0.4 20 dB SNR
10-8
0.2 80 dB SNR
10-9
0
-120 -100 -80 -60 -40 -20 0
-0.2
Phase noise, dBc/Hz
-0.4
0 10 20 30 40 50 60 70 Fig. 7. Influence of phase noise and SNR on bearing estimation. R0 = 40 m,
Range (m) θ = 80◦ .
Fig. 5. Masking effect of SNR on the benefit of using the exact formula of
(40) in phase-comparison bearing estimation. VII. C ONCLUSION
The simultaneous measurement of range and bearing in
LFMCW radar has been examined. A radar-signal model
1 which establishes a basis for range-bearing estimation in
Mean-square error
10-1 Sum-difference
Ratio, cross-correlation LFMCW radar has been developed. The model is used to
10 -2
10-3 highlight the strengths and limitations of LFMCW modulation
10-4 Range-bearing map for range-bearing estimation. A comparison with the range-
10-5
10-6 velocity model shows that higher-order expansions and their
10-7 associated phenomena are less critical for the range-bearing
10-8
10-9 model. Two variants of the signal model, range-bearing-map
-40 -20 0 20 40 60 80 and phase-comparison, are evaluated for their effectiveness in
SNR (dB) bearing estimation. The range-bearing-map is seen to be more
effective in the presence of noise. More in-depth studies are
Fig. 6. Influence of SNR on bearing estimation. R0 = 40 m, θ = 80◦ . warranted to investigate the noise-performance of the bearing
estimation methods.
R EFERENCES
uniformly distributed over [−ρ, ρ], where |ρ| < π. The random
term was added to the argument of the cosine in (2). ρ was [1] D. Barrick, “FM/CW Radar Signals and Digital Signal Processing,”
National Oceanic and Atmospheric Administration, Tech. Rep. 283-WPL
determined by noting that the power of a zero-mean uniform 26, Jul 1973.
random variable ∈ [−ρ, ρ] is ρ2 /3, which is assumed equal to [2] Constantine A. Balanis, Antenna Theory, 3rd ed. John Wiley & Sons,
the receiver noise power. The noise power PN can be obtained 2005.
[3] D. Langer, “An Integrated MMW Radar System for Outdoor Navigation,”
from the definition of phase noise as PN = LPc B [9], where Ph.D. dissertation, Carnegie Mellon University, January 1997.
L is the phase-noise-power relative to the carrier, Pc is the [4] F. Engels, P. Heidenreich, A. M. Zoubir, F. K. Jondral, and M. Winter-
carrier power, and√B is the receiver bandwidth. ρ can thus be mantel, “Advances in automotive radar: A framework on computationally
efficient high-resolution frequency estimation,” IEEE Signal Processing
specified as ρ = 3LPc B. Magazine, vol. 34, no. 2, pp. 36–46, March 2017.
The mean square of the relative angle-error was simulated [5] P. Asuzu and C. Thompson, “A More Exact Linear FMCW Radar Signal
over phase noise at various SNR levels, and is shown in Model for Simultaneous Range-Velocity Estimation,” in IEEE Radar
Conference, Apr 2018.
Fig. 7. Among the phase-comparison methods, sum-difference [6] Bassem R. Mahafza, Radar Systems Analysis and Design Using MATLAB.
showed the best performance and is hence compared with Chapman & Hall/CRC, 2000.
range-bearing map. Below −50 dBc/Hz, the angle error from [7] P. Asuzu, “Millimeter-wave Radar Investigations of Road Conditions
and their Impact on Wheel Slip,” Ph.D. dissertation, University of
range-bearing-map is largely independent of phase noise, Massachusetts Lowell, May 2018.
whereas phase-comparison requires a more stringent cutoff [8] J. H. Roberts, “Phase comparison monopulse radar: statistics of phase
(−100 dBc/Hz) to become phase-noise independent. At low error,” Electronics Letters, vol. 20, no. 9, pp. 382–383, April 1984.
[9] David M. Pozar, Microwave Engineering, 3rd ed. John Wiley & Sons,
phase noise, the angle error from phase-comparison is a 2005.
steeper function of SNR than range-bearing-map. Thus range-
bearing-map is seen to be more effective against noise than
phase-comparison, since it can tolerate a larger degradation in