A High Resolution Algorithm Based on Chirp

z-Transform for FMCW Radar

Peng-Cheng Gongl, Jie Lil, Weiwei Gu02, Zhenhai Sha02
J School of Electrical and Electronic Engineering,

Hubei University of Technology, Wuhan 43006S, China

2School of Communication and Information Engineering,
University of Electronic Science and Technology of China, Chengdu 611731, China

Abstract-Frequency modulation continuous wave (FMCW)

radar has the merit of simple structure, small volume, high
resolution in space and real-time character. So FMCW radar can r � (/�" -�,,)
F
4 "
(4)
detect and orientate the targets on the ground no mater how far
they are. However, the measure precision is affected by the FFT
method used in the FMCW radar system. In this paper, a new
l c; (fda !,lP)
R
=

S F +

method for increasing the range and velocity measure precision

by the Chirp Z-transform in the FMCW radar is proposed.
where fdo is the down beat frequency and fup is the up beat

Simulation results show that the proposed method can increase

frequency. And we usually take the FFT algorithm to get the
both the calculation efficiency and measure precision. max point of the frequency spectrum, namely, we get the
down beat frequency and up beat frequency using FFT.
I. INTRODUCTION
II. THE PRINCIPLE OF CHIRP Z-TRANSFORM METHOD

The theory of frequency modulation continuous wave In practical application, it is impossible to get the entire
(FMCW) radar is well known [I], [2]. A FMCW radar continuous signal. Digital signal processing techniques deal
transmits series of millimeter waveforms through the antenna with the sampled signal, and then the receive signal can be
and receives the signal reflected by targets. The principle of ()
written as the up time series XUI' n and the down time
FMCW radar is shown in Fig.l. The received and the
transmitted waveforms have the same shape except for the ()
series xdo n . Range and velocity are calculated using the max
delay, and the frequency difference between them is the mixer point of frequency spectrum according to (4). But it is obvious
output. The beat signal is called the intermediate frequency that there must be difference between theoretical data and
(IF). measured data if we take the FFT method. However, if we
If want to reduce this difference, we can increase the points of
Transmitted signal Received signal FFT. The increasing points of FFT can increase frequency
resolution, however, there will increase the computational
r �� , �
······ · ························ ·

,,
complexity. To solve this problem, a new method is proposed
,,
"

4J!..
," ,,
in this paper.
,,
,,,
As we all known, a fast Fourier transform (FFT) is an
""....y.." '

efficient algorithm to compute the discrete Fourier transform

�.................... :Q' .................... �#� ....... .
(DFT). The discrete Fourier transform (DFT) is a specific kind
Fig. 1. The principle of FMCW radar
of discrete transform, used in Fourier analysis. The OFT
According to the Fig.1, we can know usually requires an input sequences is discrete, such inputs are
M
often created by sampling a continuous sequences. The
- --
TI2
(1)
!!F
discrete input sequences must also be a finite set of uniformly
IF spaced time-samples of some signal. And the transformed
M = 2R/c (2) sequences also are a finite set of uniformly spaced samples of
where !:::.t is the time delay, T is the frequency modulation the continuous DTFT and it has the same length as the input
periodicity, !!F is the FM bandwidth and c = 3 x 108 mls. sequences. This is to say, DFT compute z-transform at a
number of points uniformly spaced on the unit circle in the z­
And according to (I) and (2), we can get
plane. And the frequency resolution is 21r / N . But the Chirp
R=�IF (3) Z-transform can compute a more general transform based on
4�F the Z-transform, not have to uniformly or on the unit circle.
If the target is moving, the Doppler shift can cause a range The Z-transform of x(n) can be written as [3]-[6]
error, and the range and velocity can be written as N-J
X(Zk) = L x(n) z
-n
,0:C:;n:C:;N-1 (5)
978-1-4673-65 44-4/15/$31.00 ©2015 IEEE n=O

482
 Let the sample point Zk
of z-plane be targets can't distinguish one from another by FFT method.
However, the calculated frequency can be easily distinguished
Zk =AW k = 0,1,...,M-1
k
- (6) by Chirp Z method, and all center frequency is also accurate.
B. Example 2
where M is the point number of the desired spectrum, and
8
A =AoeJ o , W =woe-.l'Po (7) In this example, assume the perfect beat signal without
Substituting (6) into (7) yields noise and clutter. The frequency transmitted by FMCW radar
k k k is 35GHz. the FM bandwidth is 60MHz. the frequency
Zk =AW- =AoeJBOwo- eJ 'ilJ (8)
modulation periodicity is 10ms, and the sample frequency is
where Ao and Wo are any positive real data, eo is the starting
0.5MHz and O.lMHz. Assume the range and the velocity are
angle and it is arbitrary, f/Jo is the angle between Zk and Zk_J •
4500m, 200mis, respectively. Therefore, the beat frequency is
Substituting (8) into (5), the Chirp Z-transform of N-point 0.23MHz. And the beat frequency is 4.4 kHz when the range
time series x(n) can be written as and the velocity is 50m, 10mis, respectively. The simulation
N-J results are shown as Table I and Table II. Note that the error
X(Zk) = L,x(n)A-nWn k = 0,1,...,M-1
k
(9) between theoretical data and real measured data is fined by
n=O Chirp Z method.
Due to the Chirp Z-transform is more flexible than the FFT, Table I The simulation results with the sample frequency O.5MHz
thus it has additional freedoms. First, the number of time
samples does not equal the number of samples of Z-transform.
Second, the angular spacing f/Jo is also arbitrary.
Consequently, according to the principle of Chirp z­
transform (9), the proposed method can be summarized in the
following steps. FFT method is taken to the beat signal to find
the main lobe of the spectrum, and then Chirp Z-transform is
used to find peak frequency in the adjacent area.
III. ANALYSIS OF SIMULAnON RESULTS

This section uses numerical examples to demonstrate the

superior performance of proposed Chirp Z-transform method
over that of the FFT method.
A. Example 1
The beat signal, for this example, is chosen
as y = sin(2;r J;l) +sin(2;r J;t) +sin(2nI2t) , where /0=999.2
Hz, J1 =1000 Hz, .12=1001Hz. Assume that the sample
Fig.3 to Fig.6 show the trends of errors varied with ranges
frequency is 4000Hz and the simulation time is 5s. To
and velocities. The measure precision by Chirp Z is about as
increase the frequency resolution, the M-point narrowband
three times as FFT method in the long range, and the measure
signal in the frequency dominant can be concentrated to a
precision by Chirp Z is about as much as ten times by FFT in
certain frequency range.
the short range.
Signal spectrum (FFT) Range error with the sam ple frequency O.SMHz
4
r----"-c= -- - � - - - -= =c- - - d
T e err Or U Si ng F F T m e�
h th o� � � 1_- :�_

h
T e error using Chirp Z m e thod 1
---1--
3.5

� I --.l ___ L __
, I
L ___1___ --.l ___
� 0.5 ______
, , ,
______
1
, , ,
«

�---L--�---
2.5 C-- l -'
- � --- � -- __ _ ___ L ___1___ J ___
O
-2000 -1500 -1000 -500 500 10DO 1500 2000 I :\ , y
Frequency(Hz)
(Chirp Z) i 2
' 'I
I
/ "�
'
, , , ,

\ "
Signal spectrum

//
1.5 1 1
fOO 9.2
: :
1

:
1 1 1-1 ----�
1 ------- ________.t�
"
--- -\

--- - - � - - - � - --- -/- - - - - � - --- - - � - --

- - � - - � -- - - -tl �\ - � \-f2-= r1r ---1- - -! - - -

: If1=19p OHZ : :
1 -
1 z 1 1 1 ----J
- 1 1 1

) -f
� __
05

I c:-- \ - - �- � -:- - - - - - - � -
- /-/ --f- - --
,
,
'
y. �'
YNV�I�
O �����LU�LW���UlLU�LU���� A •• , ti ' (\ I \�yvvv. ' o
o
I
1
500
______ ---+---- ./

1 000
1
1
1
1500
1
1
2 000 2500
1
I
3 000
1
1
3500
1
1
4 000 4500
995 996 997 DOD
998 999 1 1001 1002 1003 1004 1005 Range(m)
Frequency(Hz)
Fig.3. The trend of error varied with range

Fig.2. Signal spectrums of FFT method and Chirp Z method

Fig.2 shows that the signal spectrums using FFT method
and Chirp Z method. It can be noticed from the Fig.2 that the

483
 0.1MHz
[--
Range error with the sample frequency

I
Table II The simulation results with the sample frequency 0 IMHz
The error using FFT method
0.9 ! The error using Chirp Z method
! I I I I

0.8 -
I I I

:: , - � � � : � � � � -r - - r - - - - :2 �'\ [ � �
E -
'§' 0.5 I: t
I,\
c
w
0.4 - -
i- - - - � - - - - - � - � - - � - - � - - - - - � \\\-
0.3 I

-
_.J-----
0.1
c =-==- -::.-=-----r-- --
o o���--�----��
c_ __ __c______ __ _
_

10 15 20 25 30 35 40 45 50
Range(m)

Fig.6. The trend of error varied with velocity

IV. CONCLUSION

A new method of increasing the range and velocity measure

precision by the Chirp Z transform for the FMCW radar is
proposed in this paper. Simulation results show that the
proposed method can increase both the calculation efficiency
Velocity error with the sample frequency O.5MHz and measure precision.
1.2
'f
I
,"
I I
I -- The error using FFT method
1 _ __� __� _
� � __ -- The error using Chirp Z method
_
ACKNOWLEDGMENTS

'" This work was supported in part by the Doctoral Starting up

0.8- - --�-- �-----�--�--�--�----- Foundation of Hubei University of Technology under Grant

0.6- -
BSQDl3035, and by the science project in Hubei Province
Vi'

\\
under Grant 2014BAA135.
� I I I
- � -- � -- � -- \-- � -- � -- � -- � -- - --
I j I I
� 0.4-
I j I I I I I
REFERENCES

_I___I___� __
I I I I
�--�--�--�----- [1] M. Bouchard, D. Gingras, Y. Villers, D. Potvin, "High resolution
�-----I -------+--
I I - I
spectrum estimation of FMCW radar signals," IEEE Sventh SP

- - - I - - r- - - r- - - r- - - - --
I � Workshop on Statistical Signal and Array Processing, Canada, 1994, pp,
--'--'-
1 - �-
421-424.
[2] A.G. Stove, "Linear FMCW radar techniques," IEEE Proceedings,
-o .2oc----c2:':co-- c':4oc-� 6c-
0 ----=8c:-0---:1":c00:---1�2c-0 -�
14-c0-c-16'-:0---:1c::80:----:"20-0 vol.139, pp, 343-350, Oct. 1992.
Velocity(m/s) [3] L. R. Rabiner, R. W. Schafer, C. M. Rader, 'The Chirp Z transform
algorithm," IEEE Transactions on Audio and Electroacoustics, vol. 17,
Fig.4. The trend of error varied with velocity

C Wang, C Zong, "A new method to increase the range precision of

pp. 86-92, June 1965.
Range error with the sample frequency O.1MHz [4]
The error using FFT method LFMCW system with Chirp-z transform," Control and Automation
0.9- - - r- - - r- - - - -- The error using Chirp Z method
X Chen, I Liu, "Fast frequency estimation algorithm for FMCW
Publication Group, voUl, pp. 188-192, Dec.2005.
[5]
0.8
microwave liquid level measurement," Chinese Journal Of Sensors and
-----�
0.7-
: - - - ; - - - i
-

0.6- - - r- - - j-- - - r- - - - - - t- -
- - - - I
-
- i�-_/-
-_ �
-
-
_� -_-_ �
r
-_ [6]
Actuators, voLl8, pp. 901-905, Sep. 2001.
Lawrence Rabiner, "The chirp z-transfonn algorithm-a lesson in
I serendipity," IEEE Signal Processing Magazine, vol 21, pp.l18-119
E I
g 0.5 II \ March 2004.

0.4 -1 - - II - - jj - - II - - - - - II - - II - - II - - II - - - "\ -
w \

I I I I I
0.3 , 'I
j I I \
\, I I I \
I I I I
I j I I
1 I I I
�- --�--�--�-----�--�--�--�----
\
�
I j I I I I I
0.1 J___

0------ 10 15 20 25 30 35 40 45 50
Range(m)

Fig.5. The trend of error varied with range

484