FMCW Radar Sensors

Frequency Modulated Continuous Wave Radar

Basic operating principles and theory
FMCW (Frequency Modulated Continuous Wave radar differs from pulsed radar in that an electro-
magnetic signal is continuously transmitted. The frequency of this signal changes over time, gener-
ally in a sweep across a set andwidth. The difference in frequency etween the transmitted and
received (reflected! signal is determined y mi"ing the two signals, producing a new signal which
can e measured to determine distance or velocity. # sawtooth function is the simplest, and most
often used, change in frequency pattern for the emitted signal.
FMCW radar differs from classical pulsed radar systems in that an $F signal is continuously output.
Consequently, time of flight to a reflecting o%ect can not e measured directly. &nstead, the FMCW
radar emits an $F signal that is usually swept linearly in frequency. The received signal is then
mi"ed with the emitted signal and due to the delay caused y the time of flight for the reflected
signal, there will e a frequency difference that can e detected as a signal in the low frequency
range. # schematic presentation is shown in Figure '.
Figure '. (chematic presentation showing how a low frequency signal is generated y mi"ing the
received $F signal with the output $F signal. )ue to the delay, *t, caused y emitted signal travel -
Data subject to change without notice. Rev. A 2011 06 - 2011
______________________________________________________________________________________________
Sivers IMA A !e"# $%6-&-'0( 6& 00
o) 12'% *a)# $%6-&-'+1 ,2 '1
S--16% 2, .ista e-/ai"# sa"es0siversi/a.se
Swe1en www.siversi/a.co/
FMCW Radar Sensors
ing the distance to the reflector and ac+ to the receiver, there will e a small difference in signal
frequency etween the two $F signals. This is output as an &F-signal with frequency *f.
# simplified derivation of the intermediate frequency (&F! signal with the frequency *f can e made
in the following way, assume that the $F signal generator will output a frequency that is changing
linearly over time as,
where f
$F-
is the starting frequency, T is the frequency sweep time and +
f
is the slope of the fre-
quency change, i.e. the sweep rate,
where .W is the frequency sweep andwidth. The delay caused y the round-trip of the emitted
signal to the reflector is calculated as,
where d is the distance etween the radar antenna and the reflector and c is the speed of light.
)ue to the delay, the frequency of the received signal compared with the emitted signal will e,

The difference in frequency, *f, etween f
$F $eceived
and f
$F-
is thus,

This is the signal that is output from the detector. The minus sign can e omitted since the real sig-
nal frequency output from the radar detector is wrapped to a positive frequency. Thus the e"pres-
sion can e written as,
Data subject to change without notice. Rev. A 2011 06 - 2011
______________________________________________________________________________________________
Sivers IMA A !e"# $%6-&-'0( 6& 00
o) 12'% *a)# $%6-&-'+1 ,2 '1
S--16% 2, .ista e-/ai"# sa"es0siversi/a.se
Swe1en www.siversi/a.co/
FMCW Radar Sensors

Typical values for the $(/0-- series modules would e a frequency sweep, .W, of '1-- M23 over
T451 ms corresponding to a sweep rate, +
f
, of 6- --- M237s. # distance, d, etween the radar and
a reflector of '1 m would give a delay, *t, of -.' 8s and the &F signal frequency, *f, would then e
6--- 23. This signal is easily sampled with a high resolution #)C in order to e detected. &f several
reflectors are appearing in the measurement setup, the resulting &F signal will contain superposi -
tions of the individual &F-signals from the echoes.
)ifferent echoes are distinguished y their unique &F signal frequency and a Fourier transform of
the sampled signal can e used to e"tract the distances to the different targets. The measurement
range of the system is limited y the sensitivity of the detector and the sampling rate of the #)C.
For the $(/0-- series a sampling rate of 6- +23 gives a ma"imum detectale &F signal frequency of
'- +23, which corresponds to a range of 51 m. 9onger ranges are easily achievale y either in-
creasing the sample rate or lowering the sweep rate. &n addition, antenna gain needs to e fairly
high in order to provide sufficient signal levels for the detector.
Theoretical performance
The fundamental range measurement resolution of the system can e estimated as follows. The
Fourier transform of a time limited signal can only detect an &F signal frequency with a resolution
of '7T, +eeping in mind that *t::'7T; thus the sampling time can e appro"imated y T. <sing
equation =, this gives the minimum change in d, *d, as,
which can e transformed to,

showing that the range measurement resolution is only limited y the sweep andwidth. This is an
important oservation since it is says that resolution is not dependent on the frequency of the $F
signal itself, ut rather only on the sweep andwidth. There are methods of increasing the resolu-
tion of the measurements y a factor of '- to '-- using fitting algorithms. These find a pea+ in the
Data subject to change without notice. Rev. A 2011 06 - 2011
______________________________________________________________________________________________
Sivers IMA A !e"# $%6-&-'0( 6& 00
o) 12'% *a)# $%6-&-'+1 ,2 '1
S--16% 2, .ista e-/ai"# sa"es0siversi/a.se
Swe1en www.siversi/a.co/
FMCW Radar Sensors
&F signal spectrum which is not at an integer frequency point defined y the sampling rate and
sweep andwidth.
The range detection and FMCW radar principle may also e derived using a characteri3ation of the
&F signal phase rather than the frequency. This is recommended in order to understand the possi-
ilities of a discrete system where the frequency sweep really is generated y a discrete set of fre-
quencies. This derivation also lends itself more directly to high resolution range measurements.
For the simplicity of understanding the measurement principle it is however unnecessary and is
thus included as an appendi".
Detailed operating principles and theory
FMCW radar principles using phase measurement (time discrete version).
FMCW radar functions y outputting a continuous $F signal, whose frequency is swept over a spe-
cific frequency and. (ynthesi3ed modules, li+e $(/0--, are in fact not sweeping the frequency
continuously, ut rather step the frequency with a set of discrete frequency points. Thus, these
systems are also called (tepped Frequency Continuous Wave ((FCW! radar. The synthesi3ed signal
source assures very precise frequency control, which is important for the accuracy and repeatail -
ity of measurements.
The $F signal will e radiated and reflected against different o%ects. The echo is then received and
compared (mi"ed! with the radiated $F signal. 2ad the system een measuring time-of-flight for a
pulsed signal, the sensor output could e linear with distance. &n an FMCW, however, the sensor
output corresponds to the cosine of the phase difference etween the echo signal and the radi -
ated signal.

where s is the output signal from the sensor and > denotes the phase difference etween the
echo $F signal and the radiated signal.
&n other words, the measurement signal from the sensor will e a cosine signal indicating the
round-trip electrical distance which the radiated signal has traveled.

Data subject to change without notice. Rev. A 2011 06 - 2011
______________________________________________________________________________________________
Sivers IMA A !e"# $%6-&-'0( 6& 00
o) 12'% *a)# $%6-&-'+1 ,2 '1
S--16% 2, .ista e-/ai"# sa"es0siversi/a.se
Swe1en www.siversi/a.co/
FMCW Radar Sensors
or
where d is the distance to the reflecting o%ect and ? is the electrical wavelength of the $F signal.
The multiplication y 6 accounts for the round-trip.
The e"pression for > can also e written as
where ? has een sustituted with, c, the speed of light and f
$F
is the frequency of the $F signal.
For any measurement (e"cept for those at very short distances! the electrical distance will e"ceed
one wavelength and there will e amiguities aout the measurement result. Fortunately, the $F
signal can e stepped in frequency and several measurements can e performed. From eq. '6, it is
clear that > will increase linearly with f
$F
and thus the detector output will e a cosine shaped sig-
nal. # small value of d, meaning a close echo, will create a slowly varying detector signal and a dis-
tant echo will create a quic+ly varying detector signal. The frequency of the $F output signal, f
$F
, is
stepped over the availale and (.W!. For the $(/0--(, this and is from 0=61 M23 to 1/51 M23,
with .W 4 51- M23. For the $(/0--@, this and is from A61- M23 to '-51- M23 and for the
$(/0--B this and if from 60--- M23 to 611-- M2C, i.e., .W 4 '1-- M23. The e"pression for > is
then,
where n indicates each unique measurement, n 4 -, ', ..., and (D-'! and D is the numer of fre-
quency points used for the measurement sequence. The term f
$F-
denotes the starting frequency.
$ecalling that the detector output is the cosine of >, the equation will e as follows,

Data subject to change without notice. Rev. A 2011 06 - 2011
______________________________________________________________________________________________
Sivers IMA A !e"# $%6-&-'0( 6& 00
o) 12'% *a)# $%6-&-'+1 ,2 '1
S--16% 2, .ista e-/ai"# sa"es0siversi/a.se
Swe1en www.siversi/a.co/
FMCW Radar Sensors
(ince the cosine function only is unique over the range E-, FG, the e"pression can e simplified to,

where >- accounts for the phase value at the starting $F signal frequency of the sweep. &t should
e noted that >- can e limited etween - and F. To e"tract the value of the distance to the re-
flector, d, one needs to estimate how much > changes over the frequency sweep. Hne simple way
is to ta+e the Fourier transform of the signal s(n!.

2ere m denotes the normali3ed inde" in the transformed domain, m4-, ', ..., (D-'!.
With the detector signal eing a time varying signal, m can e seen as the inde" in the frequency
domain for the detector signal. #s an alternative, this domain may e seen as a distance domain.
For simplicity, this domain will e called frequency domain or spectrum in this te"t since in most
cases it is the typical interpretation of a Fourier transformed signal.
$ecalling that the Fourier transform of a cosine yields two )irac-delta functions, the transform of
s(n! ecomes,
The second term on the right hand side refers to a pea+ at a negative value of m, this can e easily
converted to a positive value y adding D, ut is of no further interest here. The first term on the
right hand side will have a pea+ at m4(6dI.W7c!. Conversely, if a pea+ is identified at m4m
-
, the
corresponding distance to the reflector can e calculated as,

Data subject to change without notice. Rev. A 2011 06 - 2011
______________________________________________________________________________________________
Sivers IMA A !e"# $%6-&-'0( 6& 00
o) 12'% *a)# $%6-&-'+1 ,2 '1
S--16% 2, .ista e-/ai"# sa"es0siversi/a.se
Swe1en www.siversi/a.co/
FMCW Radar Sensors
Thus, a ranging function is achieved.
&t is worth noticing that eq. 'J is completely independent of the center frequency of the $F signal.
&n fact, to get an impression of the resolution that is availale from only e"tracting a measurement
of the distance ased on the ma"imum pea+ found in the frequency domain, the difference
etween two integer values of m is c76.W. This shows that the resolution is only depending on the
sweep andwidth of the $F signal and not of its specific frequency. # frequency sweep from A61-
to '-51- M23 will give the same resolution as a frequency sweep from 60--- to 611--M23.
For the specified frequency sweep with .W 4 '1--M 23, the integer range measurement resolu-
tion will e -.'- m.
&n order to achieve higher resolution in range measurements, a weighted average of several fre-
quency points can e used to find a pea+ location that is positioned etween integer points in the
spectrum. #dditionally, when a single echo is availale in a local part of the spectrum, it is possile
to estimate d ased on the slope of the phase angle >. <sing only the first term in the right hand
side of eq. '5, recall that the inverse Fourier transform of a )irac-delta function is a comple" e"po-
nential,
2ere, the right hand side is a comple" series of D points. &n a real-life measurement, the signal will
not e an ideal comple" e"ponential li+e in eq. 'A. 2owever, an inverse Fourier transform of only a
section of the spectrum around a pea+ of interest will give a comple" signal whose phase angle
may e e"tracted. <sing this phase angle, the slope may, for e"ample, e found using a least
squares fit to a linear e"pression.
Data subject to change without notice. Rev. A 2011 06 - 2011
______________________________________________________________________________________________
Sivers IMA A !e"# $%6-&-'0( 6& 00
o) 12'% *a)# $%6-&-'+1 ,2 '1
S--16% 2, .ista e-/ai"# sa"es0siversi/a.se
Swe1en www.siversi/a.co/