Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

A Doppler-Tolerant Stepped-Carrier OFDM-Radar Scheme

Based on All-Cell-Doppler-Correction

Benedikt Schweizer∗1 , Daniel Schindler2 , Christina Knill1 , Christian Waldschmidt1

1
Institute of Microwave Engineering, Ulm University
Albert-Einstein-Allee 41, 89081 Ulm, Germany
Email: prename.surname@uni-ulm.de
2
Corporate Sector Research and Advance Engineering of Robert Bosch GmbH
70465 Stuttgart, Germany
Email: daniel.schindler@de.bosch.com

Abstract: Recent publications show that major limitations of orthogonal frequency-


division multiplexing (OFDM) as modulation format for automotive radars are solved.
These include the requirements on analog-to-digital and digital-to-analog converters
whose required sampling rate can be reduced using a stepped-carrier OFDM-scheme
as well as the all-cell-Doppler-correction-based approach that counteracts the loss of
orthogonality due to the Doppler shift. However, so far, the application of both meth-
ods is exclusive to each other. In this paper, a combined evaluation scheme is presented
that allows to use a stepped-carrier scheme in combination with a quasi repeated-symbol
OFDM-waveform and includes the Doppler-correction into the DFT for velocity estima-
tion. This yields a high resolution radar image in both range and velocity at low sam-
pling rate with significantly increased signal-to-noise ratio (SNR) as the orthogonality
of subcarriers is preserved. The scheme is verified with simulations that demonstrate its
benefits.

1. Introduction
Orthogonal frequency-division multiplexing (OFDM) receives growing attention, as its advan-
tages are considered to be beneficial for the development of future radar sensors. This includes
the possibility to realize a true multiple-input multiple-output (MIMO) system where all anten-
nas transmit simultaneously by assigning an exclusive subset of all available subcarriers to each
antenna. It has been shown that this can be done in various ways, i.e., regularly, irregularly, or
randomly [1, 2, 3]. Due to the digital signal generation, it is furthermore possible to achieve
very short symbol durations which lead to very high unambiguously measurable velocities and
ranges at the same time. The state-of-the-art OFDM-radar processing scheme [4] is based on the
evaluation of modulation symbols whose alternation in the channel is directly available at the re-
ceiver, leading to a simple range-Doppler evaluation based on DFT operations. This approach,
however, has the restriction that the Doppler shift is only allowed to be significantly smaller
than the subcarrier spacing Δf [5]. The reason is that the received signal is still sampled on the
frequency grid of the transmitted OFDM-signal but the target reflections experience a Doppler

The 20th International Radar Symposium IRS 2019, June 26-28, 2019, Ulm, Germany 1
978-3-7369-9860-5 ©2019 DGON

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.
shift and are not aligned to that grid anymore. Thus, the orthogonality is lost and inter-carrier-
interference (ICI) occurs which can limit the achievable dynamic range if it exceeds the noise
floor of the system. This problem is addressed in [6], where a different evaluation method is
presented that counteracts the loss of orthogonality by performing the velocity estimation as
the first step, followed by an all-cell-Doppler-correction (ACDC) step. As a consequence, the
received modulation symbols are aligned with the desired grid in frequency domain.

In [7] a major limitation for practical implementations was stated to be the requirement to sam-
ple signals of some GHz, if a high range resolution is desired. Instead, a variant of the stepped-
carrier OFDM-radar scheme [8, 9, 10] that changes the carrier frequency by the baseband signal
bandwidth after each symbol is proposed in combination with a processing scheme that allows
to retrieve a 2D radar image with the same resolution as it is obtained by a constant-carrier
OFDM-radar with equal RF bandwidth. The approach is suitable for an arbitrary number of
targets and does not add computational effort. However, this method is based on the state-of-
the-art processing scheme presented in [4] and suffers from the same limitations induced by the
Doppler shift of target reflections.
In this paper, a new processing method for a stepped-carrier OFDM-radar scheme is presented
that combines the advantages of both methods, yielding high resolution in both range and
Doppler while preserving the orthogonality of the received signal. Furthermore, it simplifies the
ACDC-process and includes the correction step into the DFT for velocity estimation. Thereby
the computational overhead is reduced.

The paper is organized as follows: Section II describes the theoretical background of the pro-
posed method, followed by the scheme itself in Section III. Section IV verifies the scheme
with simulations and shows its benefits for large velocities before the paper is concluded in
Section V.

2. Stepped-Carrier OFDM-Radar Signal Model


Regardless of the processing scheme, an OFDM-modulated waveform, usually called OFDM-
symbol, can be described as

N c −1  
j2πnΔf t t
x(t) = dn e rect , (1)
n=0
T s

where dn are the modulation symbols for each subcarrier n and Nc is the number of subcarriers
that are spaced by Δf in frequency. Ts is the symbol duration, and orthogonality is achieved by
Ts =1/Δf . Such a signal can easily be realized via inverse discrete Fourier transform (IDFT).
To realize the stepped-carrier scheme, as it is visualized in Fig. 1, an agile carrier is required
that changes by the bandwidth of the baseband signal after each OFDM-symbol according to
fm = fc0 + mNc Δf, 0 ≤ m < M, (2)
where M is the number of steps that are used. At the same time Nc is chosen smaller to reduce
the required baseband bandwidth. As M symbols are needed to sample the whole channel band-

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.
TX baseband signal TX RF signal
f IQ f
t DAC
stepped carrier
f t
channel
f
t
f
t ADC t
IQ
RX baseband signal RX RF signal
Figure 1: Visualization of the stepped-carrier OFDM-radar principle. The dots represent modula-
tion symbols and colors visualize gathered channel information according to transmit frequency
[11].

M symbols
=1 block
f f
QPSK codes
f3
OFDM-symbol
f2
OFDM-block
f1

f0
t t
B blocks
Figure 2: Structure of OFDM-signals in frequency domain with all parameters. The colored sym-
bols represent modulation symbols with QPSK codes.
Left: Standard OFDM-scheme. Right: Quasi-repeated symbol stepping-scheme in RF domain.

width, M symbols form a block which is transmitted repeatedly as comprehensible in Fig. 2.


As the carrier frequency is agile, a cyclic prefix of length Np is required before each symbol to
turn the linear convolution of the channel to a cyclic one and to avoid inter-symbol interference
(ISI). Furthermore, it is beneficial if the symbols of one block are coded with unique sets of
modulation symbols dnm depending on subcarrier n and symbol m. The resulting duration of
each symbol is T = Ts + Tp + Tpause , where Ts denotes the duration of the actual symbol, and Tp
is the duration of the prefix. A potentially required time interval to guarantee settling to the next
carrier frequency fm can be included by choosing Tpause accordingly [12]. For the proposed
method that compensates the Doppler shift in time domain, it is required that the waveforms
of all blocks, which are used to evaluate the Doppler frequency, are linearly dependent or in
the most simple case equal [6]. Only if this is fulfilled, the time-domain signal can be used to

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.
estimate the Doppler frequency. Thus, to form the final ICI-free stepped-carrier OFDM trans-
mit waveform, B identical blocks with the same phase codes per symbol dnm are transmitted
consecutively. Since the phase-codes need to be different for each symbol, but identical for
each block, i.e., each repetition of the block, this pattern is referred to as quasi repeated-symbol
OFDM as visualized in Fig. 2.

The transmitted RF waveform at carrier position fm is described by

M
B−1  
−1 N c −1  
j2π(fm +nΔf )t t − bM T − mT
xRF (t) = dnm e rect . (3)
b=0 m=0 n=0
Ts

The received signal is a delayed, attenuated, and Doppler-shifted version of xRF (t), for the
simplified case of one point target. This is sufficient to consider as in the following only linear
operations are performed and the behavior is the same for an arbitrary number of targets. If a
narrow-band assumption is made, the Doppler shift is independent of the carrier frequency and
after down-conversion by fm , the obtained baseband signal is

M
B−1  
−1 N c −1  
j2πnΔf t −j2π
2R0
(fm +nΔf ) j2π 2vt (fm +nΔf ) t − bM T − mT
xRX (t) = dnm e e c0
e c 0 rect .
b=0 m=0 n=0
Ts
(4)

At this point it is noticeable that — compared to the transmit signal— two additional exponential
terms are present, of which one represents the range information and the other one the Doppler
shift. This means that each reflection itself still consists of orthogonal subcarriers, but related
to the sampling grid and also to other reflections with different Doppler shifts they are not
orthogonal anymore. Thus, applying a DFT for frequency-domain evaluation would result in
significant inter-carrier interference (ICI), which cannot be mitigated afterwards. Instead, the
Doppler shift needs to be compensated here.

3. ICI-Free Processing Scheme with Doppler Compensation


To evaluate the signal, sampling at the receiver is performed. In a simplified way with the
sampling rate Ts /Nc and N =Nc +Np this results in


N c −1
2R0
dnm e j2πnΔf ( N +mT +bM T ) e
pT −j2π (fm +nΔf )
xRX [p, m, b] = c0

n=0
 
c ( N
j2π 2v pT
+mT +bM T )(fm +nΔf ) t − bM T − mT
×e 0 rect (5)
Ts

for the discrete time index p∈[Np ,N −1] in each symbol and block, discarding the cyclic prefix.
As it is the case for the classical stepped-carrier OFDM-radar scheme, the velocity estimation
plays a vital role. In [7], it is shown that velocity profiles are required that are consistent for
all steps to be able to perform a joint high resolution range estimation. The underlying problem

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.
is that the channel is measured consecutively. Thus, the symbols of each block incorporate a
phase-shifted version of the Doppler frequency that depends on the speed of the target and also
the time offset between the symbols. Thus, a straightforward compensation based on Doppler
filters would result in a huge computational overhead. However, as stated in [7], it is possible
to mitigate the effect of the phase error by using a modified DFT for velocity estimation that
considers the actual time-frequency position of the symbols in the channel.

As the same block is transmitted repeatedly, it is obvious that by evaluating the series of each ith
sample of all blocks over time, the Doppler frequency can be estimated for each i∈[0,N M −1].
Thus, a similar method can be applied to the time domain samples.

In the Doppler component of (5),

c ( N
j2π 2v pT
+mT +bM T )(fm +nΔf )
xRX,Doppler [p, m, b] = e 0 , (6)

the phase term including bM T is the actual dependency on the blocks, which is the required
phase progression to estimate the Doppler frequency fD . The additional term with pT /N +mT
represent the phase offset due to sampling and stepping the carrier frequency that needs to be
considered in the DFT.

By using a suitable DFT with a modified kernel


B−1
T
X[k] = x[b]e −j2πk(bM N +mN +p) N (7)
b=0

the actual time instance of the samples under consideration is incorporated. When this DFT is
applied on (5), a peak is obtained for the velocity index
2v
k= T (fm + nΔf )M B, (8)
c0
which is independent of m, if the Doppler spread over frequency is neglected. Thereby, the
Doppler component in (5) is compensated and coherent velocity profiles are obtained for all
samples and steps. This means that by applying this modified DFT both goals of the scheme
are realized at once. First, coherent velocity profiles are obtained which are needed for the high
resolution stepped-carrier OFDM-radar processing and at the same time, the Doppler shift is
compensated. Consequently, the additional all-cell-Doppler-correction step described in [6] is
obsolete and can be included into the DFT. Thereby the number of mathematical operations is
reduced.

3.1. Range processing

After the application of the described DFT, velocity profiles are present that still contain the
range (and angle) information of the targets. Additionally, that information is still present in
the same samples of the matrix as before. Thus, to evaluate the range information, the velocity

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.
profiles are treated in the same way as the samples would have been processed in a classical
scheme.

Therefore, a DFT is performed on all samples of each symbol. After that, there are M frequency
domain responses for each velocity index, that correspond to the symbols of each block. As
the incorporated frequency-domain samples consist of the transmitted modulation symbols and
the range information, the transmitted modulation symbols need to be removed by a spectral
division.

Finally, the frequency responses of all symbols of each block are stacked corresponding to the
agile carrier frequency before a final IDFT can be applied on each column of Nc M modulation
symbols to retrieve a high-resolution range profile for each velocity index, which is the desired
2D range-velocity profile.

3.2. Oversampling

The derivation above was done for Nyquist-rate sampling resulting in Nc subcarriers and N time
domain samples per symbol. Although this assumption is barely true in reality, the processing
is similar. The velocity estimation is still performed on each sample of the time domain signal
and also all resulting points of the velocity profile are used for the transformation to frequency
domain. At this point, oversampling leads to a larger bandwidth than required. Hence, before
the spectral division, the frequency-domain signal can be restricted to the actual part of interest,
i.e. to Nc values representing the modulation symbols that were transmitted and contain the
actual channel information.

4. Simulation
To visualize the benefits radar simulations at 77 GHz with parameters according to Table 1 are
used. The newly proposed Doppler-tolerant stepped-carrier OFDM-radar processing is com-
pared to both a standard OFDM-radar scheme with large baseband bandwidth and the conven-
tional stepped-carrier OFDM-radar processing scheme. All simulation and radar parameters are
the same except the coding, which is a random QPSK matrix for the classical methods and a
quasi-repeated symbol scheme with QPSK codes for the new method. To visualize the effect of
ICI on the achievable performance, no additional Gaussian noise is added at the receiver. Two
targets at r1 =10 m and r2 =25 m with velocities v1 =20 m/s and v2 =−35 m/s of the same RCS
are used. All radar images are normalized to target 1.

By comparing Fig. 3 to Fig. 4 and Fig. 5, the benefit of the stepped-carrier schemes is obvious.
Both schemes deliver a range-velocity profile with the same resolution as it is obtained with a
comparable OFDM-radar that uses baseband signals with a high bandwidth. However, Fig. 4
clearly shows the limits of the classical stepped-carrier OFDM-radar scheme as the achiev-
able SNR is limited due to the noise caused by ICI to approximately SNRclassic =60 dB for the
stronger target.

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.
30 0

Normalized amplitude (dB)


20 −50

Range (m)
10 −100

0 −150
−40 −20 0 20 40
Velocity (m/s)

Figure 3:
Simulated radar image for classical OFDM-radar scheme with large baseband band-
width. The achievable SNR is limited by ICI.

30 0

Normalized amplitude (dB)


20 −50
Range (m)

10 −100

0 −150
−40 −20 0 20 40
Velocity (m/s)

Figure 4: Simulated radar image for stepped-carrier OFDM-radar scheme. The resolution is com-
parable to the classical approach, but the achievable SNR is still limited by ICI.

30 0
Normalized amplitude (dB)

20 −50
Range (m)

10 −100

0 −150
−40 −20 0 20 40
Velocity (m/s)

Figure 5:
Simulated radar image for the proposed Doppler-tolerant and ICI-free stepped-carrier
OFDM-radar processing scheme. The resolution is comparable to the classical schemes and the
achievable SNR is drastically increased.

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.
Table 1: Radar Parameters for Stepped-Carrier OFDM-Radar.

Method Classic Classic stepped ICI-free stepped


Symbols/ steps M 1 8 8
Subcarriers Nc 2048 256 256
Blocks B 2048 256 256
RF bandwidth (GHz) W 1.024
Baseband bandwidth (GHz) W 1.024 0.128 0.128
Subcarrier spacing (kHz) Δf 500
Duration of cyclic prefix (μs) Tp 0.6
Unambiguous range (m) Rua 300
Maximum range (m) Rmax 60
Range resolution (m) ΔR 0.146
Unambiguous velocity (m/s) vua ±372.8 ±46.6 ±46.6
Velocity resolution (m/s) Δv 0.36

The result of the proposed method is shown in Fig. 5. Although a small ridge in range di-
rection occurs, it is directly visible that the noise floor is decreased significantly leading to
SNR=150 dB. Thus, it can be concluded that this scheme actually reduces ICI drastically. Fur-
thermore, windowing functions can be chosen differently so that the peak is much sharper,
especially in range direction, as there is no window required.

5. Conclusion
In this paper a new processing technique for stepped-carrier OFDM-radar schemes was pre-
sented. It is based on a quasi-repeated-symbol scheme and motion compensation based on a
modified DFT. Thereby both, ICI is eliminated and the basis for a high resolution range pro-
file is set. The scheme offers a drastically increased maximum SNR compared to the standard
stepped-carrier OFDM-radar processing what is proven with radar simulations.

References
[1] C. Sturm, Y. L. Sit, M. Braun, and T. Zwick, “Spectrally interleaved multi-carrier signals for radar
network applications and multi-input multi-output radar,” IET Radar, Sonar Navig., vol. 7, no. 3,
pp. 261–269, Mar. 2013.
[2] G. Hakobyan and B. Yang, “A novel OFDM-MIMO radar with non-equidistant dynamic subcarrier
interleaving,” in European Radar Conference (EuRAD), 2016, pp. 45–48.

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.
[3] ——, “A novel OFDM-MIMO radar with non-equidistant subcarrier interleaving and compressed
sensing,” 2016, pp. 45–48.
[4] C. Sturm and W. Wiesbeck, “Waveform design and signal processing aspects for fusion of wireless
communications and radar sensing,” Proc. IEEE, vol. 99, no. 7, pp. 1236–1259, Jul. 2011.
[5] G. E. A. Franken, H. Nikookar, and P. V. Genderen, “Doppler tolerance of OFDM-coded radar
signals,” in European Radar Conference, 2006, pp. 108–111.
[6] G. Hakobyan, B. Yang, and S. Member, “A novel inter-carrier-interference free signal processing
scheme for OFDM radar,” IEEE Trans. Vehic. Tech, vol. 9545, no. 6, pp. 5158–5167, Jun. 2018.
[7] B. Schweizer, C. Knill, D. Schindler, and C. Waldschmidt, “Stepped-carrier OFDM-radar process-
ing scheme to retrieve high-resolution range-velocity profile at low sampling rate,” IEEE Trans.
Microw. Theory Tech., vol. 66, no. 3, pp. 1610–1618, Mar. 2018.
[8] G. Lellouch, A. K. Mishra, and M. Inggs, “Stepped OFDM radar technique to resolve range and
doppler simultaneously,” IEEE Trans. Aerosp. Electron. Syst., vol. 51, no. 2, pp. 937–950, Apr.
2015.
[9] K. Huo, B. Deng, Y. Liu, W. Jiang, and J. Mao, “The principle of synthesizing HRRP based on a
new OFDM phase-coded stepped-frequency radar signal,” in IEEE 10th Int. Conf. Signal Process.,
2010, pp. 1994–1998.
[10] C. Pfeffer, R. Feger, and A. Stelzer, “A stepped-carrier 77-GHz OFDM MIMO radar system with
4 GHz bandwidth,” in IEEE Eur. Radar Conf., 2015, pp. 97–100.
[11] B. Schweizer, D. Schindler, C. Knilll, J. Hasch, and C. Waldschmidt, “Expanding the unambiguous
velocity limitation of the stepped-carrier OFDM radar scheme,” in 15th European Radar Confer-
ence (EuRAD), 2018, pp. 22–25.
[12] ——, “On hardware implementations of stepped-carrier OFDM radars,” in IEEE/MTT-S Int. Mi-
crow. Symp. - IMS, 2018, pp. 891–894.

Authorized licensed use limited to: Tsinghua University. Downloaded on August 18,2022 at 08:06:45 UTC from IEEE Xplore. Restrictions apply.

You might also like