Waveform analysis for high-frequency FMICW radar
R.H. Khan
D.K. Mitchell
Indexing term: High-frequency radar, FMICW waveforms, Ambiguityfunction, Ocean surveillance
Abstract: The paper describes the waveform
analysis for an experimental long-range high-
frequency radar system with a design capability
for over-the-horizon detection of ocean surface
targets up to 400 km from shore with a range
resolution of 400m. Motivated by the need to
maximise the transmitter duty cycle in order to
detect long-range targets, the radar is based on a
frequency modulated interrupted continuous wave
(FMICW) type waveform. The interrupt sequence,
which is necessary for a high power monostatic
operation, complicates the analysis of the radar
performance in terms of the ambiguity function
(AF). This paper analyses the range dependent AF
of FMICW waveforms and proposes practical
interrupt sequences for the synthesis of desired
AFs.
1 Introduction
The authors are involved in the construction of a shore-
based long-range high-frequency (HF) radar facility for
iceberg detection, ship tracking and ocean environmental
measurements. At these frequencies, electromagnetic
waves are capable of over-the-horizon (OTH) propaga-
tion in either the ground wave (GW) or sky wave mode.
This prototype system relies on the GW mode of propa-
gation and has a design capability for OTH detection
and tracking of targets up to 400 km from shore with a
range resolution of 400 m [l]. The main objective is to
develop techniques for the radar waveform analysis and
synthesis to achieve the desired performance specifi-
cations for the prototype radar.
In a conventional single-frequency pulsed radar
system, the desired range resolution determines the pulse
length. Range is computed as the time delay of returned
echoes. The maximum range is determined by the repeti-
tion rate of the pulses and the hardware limit on the
maximum achievable pulse peak power. It is well known
(see for example Reference 2) that pulse compression
techniques can be used to realise high energy pulses with
low peak powers. These waveforms can be compressed,
after reception, using simple hardware and/or software
operations. For example, using a linear frequency modu-
lated (FM) sweep during the radar pulse, large pulse
compression ratios can be achieved. The range measure-
Paper 8091F (ElS), first received 1st June and in revised form 29th
November 1990
Mr. Khan is with the Centre for Cold Ocean Resources Engineering,
Memorial University of Newfoundland, St. John’s, Newfoundland,
Canada A1B 3x5
Mr. Mitchell is with Northern Radar Systems Ltd., PO Box 8323, Stn.
A, Bartlett Building, St. John’s, Newfoundland, Canada A l B 3N7
IEE PROCEEDINGS-F, Vol. 138, NO. 5 , OCTOBER 1991
ment of FM radars is based on the difference in instanta-
neous frequency between the transmitted and received
waveforms. The received waveform from a stationary
target, with a two-way travel time of to seconds, differs in
instantaneous frequency from the received waveform by
(fo = at,), where a is the frequency sweep rate. However,
the linear FM pulse waveform has a range/Doppler
ambiguity. This ambiguity may be resolved by using a
frequency modulated continuous wave (FMCW) sweep;
effectively many realisations of linear FM waveforms are
present and target velocity may be unambiguously esti-
mated as the rate of change in range [3]. FMCW radars
have a range resolution competitive with pulse radars
when the same signal bandwidth is used. A survey of
numerous papers addressing different aspects of FMCW
radars has been presented by Skolnik [4].
The FMCW waveform, with a 100% duty cycle, is
ideal for bistatic or low-power monostatic operation. For
the high-power monostatic application of our prototype
radar, there are two problems with the FMCW
approach. First, the necessary isolation between the
transmitter and receiver for continuous operation at H F
frequencies is very difficult to achieve. Secondly, due to
the propagation losses, the signal levels decrease rapidly
with distance and the dynamic range of the signals
exceeds that of any available receiver hardware.
To overcome both the above mentioned problems the
authors have chosen a frequency modulated interrupted
continuous wave (FMICW) type waveform [SI for the
prototype radar. This is simply an FMCW waveform
which is gated on and off with a well defined sequence,
typically a pseudorandom sequence. The concept is to
transmit for a period of time and, after the transmitter
has been turned off, to receive during the quiet period.
Thus, compared to an FMCW waveform, both the trans-
mit and receive waveforms are subjected to different
interrupt sequences. The introduction of these interrupt
sequences complicates the analysis of the radar per-
formance in terms of the well known ambiguity function
(AF). For example, it can be easily shown that the effec-
tive gating on the received signal varies with the target
range and introduces a range dependence on the AF. The
form of this AF has been discussed in the literature [SI,
however, in Section 3, we will use some properties of the
AF to simplify the AF analysis of FMICW radar wave-
forms with arbitrary transmit and receive interrupt
sequences.
The inverse problem of determining the parameters of
an FMICW waveform to realise a desired AF is more
difficult. Some simplification is possible for a GW-HF
surveillance radar, where many targets of interest are
slow moving and the zero-Doppler ambiguity function is
suffcient to evaluate the performance of the waveform.
This zero-Doppler ambiguity function can be easily
obtained from the general form of the AF and offers us
41 1
an insight into the effect of different interrupt sequences
on the FMICW waveform.
Both random and deterministic interrupt sequences
are considered in Section 4. It is shown that random
interrupt sequences result in an AF with a high noise
floor, which can obscure weaker targets, whereas deter-
ministic sequences cause ambiguity pop-ups. Section 5
details how these ambiguity pop-ups may be successfully
controlled, resulting in an AF with a very low noise floor,
a signal dynamic range with the capability of available
receivers, and a reasonable signal-to-noise ratio (SNR)
with respect to FMCW.
2 Background
Modern radar technolgy relies on the AF to evaluate the
performance characteristics, such as range and Doppler
resolution, of waveforms for different applications [2].
For a radar waveform p(t), the AF is generally defined in
terms of the response function ~ ( z , 4):
where
The response function has two parameters, time delay z
and Doppler frequency 4, and can be interpreted as a
measure of the correlation of the waveform p(t) with a
delayed and/or Doppler shifted version of itself. This is
used to evaluate the capability of a radar waveform to
resolve targets close to each other in range and/or
Doppler. For example, the well known ‘chirp’ waveform
is expressed as
g(t) = ~ e j ( ~ ‘ ~ ) / * all t (2)
where K is a scaling constant and a is the frequency
sweep rate. This ideal ‘chirp’ waveform is an infinite
linear frequency modulated sweep and its response func-
tion can be easily evaluated using eqn. I :
x,(79 4) = I K I22.d(4 - (3)
The response function in eqn. 3, clearly shows the well
known range/Doppler ambiguity of the ‘chirp’ waveform.
The characteristics of the response function for more
complicated radar waveforms are not so obvious. In par-
ticular, the evaluation of the response function for
FMICW waveforms is complicated by the finite extent of
the linear frequency sweep as well as the effect of the
interrupt sequence used. Poole [2] has studied the ‘absol-
ute delay-dependent cross-ambiguity function’ for
FMICW waveforms with pseudorandom interrupt
sequences; an important result reported is that the AF is
not independent of target range. In the next Section, we
present an analysis of the AF of FMICW waveforms in a
manner which will allow the subsequent synthesis of
desired performance characteristics by using different
waveform. Thus some form of interruption is required on
transmission (FMICW waveform) so that the receiver
may be activated during the intervals between transmis-
sions. Our objective here is to evaluate the AF for an
FMICW waveform. The different gating sequences
imposed on the transmitted and received signals compli-
cates the analysis and also introduces a range depen-
dence of the AF. We will use the properties of the AF to
simplify the analysis of some physically realisable
FMICW waveforms.
It is important to note that the receiver must always
be off whenever the transmitter is on, so that the trans-
mitter and receiver interrupt sequences are not identical.
Further differences between the two sequences are intro-
duced by the ‘soft gating’ used to suppress transients
which would be caused if the interrupt sequences were
switched rapidly between the on and off states. Depend-
ing on the target range, the transmitted waveform is not
received in its entirety at the receiver and the radar per-
formance needs to be evaluated by the ‘cross response
function’ between the transmitted and received signals.
The effect of target range on the received signal is
demonstrated in Fig. 1 where, for the same transmitted
signal, the received signals are shown for two different
target ranges. Although the returned signals are similar,
the relative time delays in the receiver gating sequences
significantly change the actual signals gated into the
receiver. The AF will obviously be different for these two
ranges and this effect must be carefully considered when
designing waveforms for long-range applications.
3.1 Evaluation of the response function components
We will now evaluate the AF for a general FMICW
waveform. The soft gating mentioned above is not con-
sidered; the low pass filtering effect due to this gating can
be easily added at any stage of the analysis. It is possible
to write an expression for the AF based on the strict defi-
nition [2] but that approach gives little indication of the
effect of changing the FMICW parameters on the AF.
The derivation presented here evaluates the AF in terms
of a number of component response functions which are
combined using linear operations. Specifically, only the
‘cross response’ function for the interrupt sequence needs
to be computed. The rest of the derivation involves linear
operations and the well known response function of an
FM sweep. For example, a single period of an FMCW
waveform can be described as the time-domain product
of an ideal FM sweep or ‘chirp’ waveform g(t) and a gate
function with width corresponding to the FMCW sweep
period TR. The full FMCW waveform can then be rea-
lised by convolving this single period with an impulse
sequence having a constant period TR. The transmitted
and received FMICW waveforms are obtained by multi-
plying the FMCW waveform with the transmit and
receive interrupt sequence, respectively.
The interrupt sequence corresponding to the transmit-
ted waveform is defined by
-
interrupt sequences. 0 ~ A d t ) ~ l - m < r < m (4)
Similarly, the interrupt sequence corresponding to the
received waveform is defined by
- 00 < t < 00 ( 5 )
3 Ambiguity function for FMICW waveforms
The practical implementation of the ‘chirp’ waveform sets 0 Q AR(t) < 1
a finite limit on the frequency sweep; a periodic repeti-
Of this finite sweep is known as the FMCW wave-
form. In the H F band, for a high-power monostatic
operation, the transmitter and receiver cannot be effec-
A fundamental aspect of FMICW operation constrains
the receiver to be disconnected whenever the transmitter
is operating i.e.
tively isolated for the 100% duty cycle of the FMCW AR(t ) = 0 if AT(t) # 0 (6)
412 I E E PROCEEDINGS-F, Vol. 138, No . 5 , OCTOBER I991
D
fl,
0
d
01
I
I
I
I
I; 1 I
I
11
e
01 t2
f
Fig. 1
(I Generated FMCW signal
b Transmitter gating sequence
c Receiver gating sequence (complement of transmit sequence)
d FMICW signal using the gating sequence of b
e FMICW radar echo from target with two-way travel time off, seconds
.f FMICW radar echo from target with two-way travel time oft, seconds
Effect of target range on FMICW radar signal
It is well known that the response function of the product
of two time-domain signals can be evaluated by convolv-
ing their individual response functions along the Doppler
variable. Thus the ideal response function of the ‘chirp’
waveform in eqn. 3 is broadened by convolution with the
response function of an ideal gate to achieve the response
function of a finite FM sweep pulse. The response func-
tion of an ideal gate is shown in Fig. 2. From the finite
gate G(t) of Fig. 2a, the kernel of eqn. 1 is formed as a
two dimensional function in t and T. The response func-
tion is obtained by computing the Fourier transform in
(t) as per eqn. 1. Note that this linear broadening of the
ideal response of the chirp signal may be implemented
later to retain simplicity in our derivation. The same
comment applies to the operation of convolution with an
impulse sequence to form the repetitive sweeps of the
FMCW waveform.
I
‘ VI\
/ I \ c
response function
x ~ ( r . 4 )
delav
Fig. 2
gate function using eqn. I
Evaluation of the response function corresponding to an ideal
Thus, expressing the transmitted waveform as a gated
version of an ideal ‘chirp’ waveform
fAt) = A A W ) (7)
The radar returns from targets are delayed by the
two-way travel time to and are further modified by the
receiver gating sequence
f a = A J W A r - to)g(t - to )
fR(t + to ) = AR(t + to )AAMt)
(8)
(9)
Eqn. 8 may be simply rewritten as
Examination of eqns. 7 and 9 shows that the response
function for the FMICW waveform can be evaluated by
(a) Evaluating the response function for g(t) .
(b) Evaluating the ‘cross response’ function for AAt)
(c) Convolving the above two response functions along
and AR(t + to)AAt).
the Doppler variable.
413 IEE PROCEEDINGS-F, Vol. 138, NO. 5 , OCTOBER 1991
(d) Convolving the response function of the infinite
sequences, with the impulse response of the finite band-
width.
(e) The response function is periodically repeated, in
the delay variable, due to the periodic nature of the
FMICW sweep.
The response function of the ‘chirp’ waveform is given by
eqn. 3 and contains the effects of any Doppler shift due
to target velocity. The ‘cross response’ function for the
transmitter and receiver gating sequences can be easily
evaluated using:
Note that this two-dimensional response function
depends on the delay time, i.e. the ambiguity function will
vary with target range.
The response function of eqn. 10 can be convolved,
along the Doppler variable, with the response function of
the ideal ‘chirp’ waveform to give the response function
of the FMICW waveform. This ideal response function
can now be broadened to account for the soft gating as
well as the finite bandwidth of the FM sweep.
3.2 Zero Doppler ambiguity function
In many applications of H F radars, the target velocities
are very small and the zero-Doppler ambiguity function
is sufficient to evaluate the performance of the waveform.
This zero-Doppler ambiguity function can be easily
evaluated. It can be seen that the convolution of the
response function of eqn. 10 with the response function of
the ideal chirp of eqn. 3, at each value of the delay
parameter, will shift the response function (eqn. 10) in fre-
quency. This frequency shift is a linear function of the
delay parameter z. A little reflection will show that the
response function needs to be evaluated for a set of delay-
Doppler parameters such that the values will be mapped
to the zero Doppler axis after convolution with the
simple response function (eqn. 6) of the ‘chirp’ waveform.
For this set, the Doppler frequency corresponding to any
delay z is simply -az, where a is the frequency sweep
rate of the ‘chirp’ waveform. Thus the zero-Doppler
response function for an FMICW waveform can be com-
puted using
x/(z? a, = t + i ) R ( + - i)
x A,( , - ;)dzna7‘ dt (1 1)
As the mapping to the zero doppler axis depends on the
frequency sweep rate, a now appears as a parameter in
the evaluation of the response function. From a compu-
tational point of view, eqn. 11 may be simplified by repla-
cing t with ( t + ~ / 2 )
xf(‘9 ‘7 = j:mAdt + + t O ) A T ( t )
(12) e j z n d t + r / 2 ) dt
Eqn. 12 may be interpreted as the correlation of two
sequences P(t , to ) and Q(t, z) given by
P(t , to ) = AR(t + to)Az(t)
Q(t, Z) = Adt + z ) & ~ ~ ~ ~ ( * + ~ ’ ~ ) (13)
The following Section presents some examples of
FMICW waveform ambiguity functions evaluated using
the above technique.
4 Numerical evaluation of ambiguity function
The two examples presented here are evaluated using the
following specifications of the prototype H F long-range
radar:
sweep bandwidth, B 375 kHz
sweep repetition period, TR 0.524288 s
range resolution 400m
409.6 km unambiguous range
For the computer simulations the 0.524288 s FMICW
sweep repetition period was quantised to 8192 samples
corresponding to a sample interval of 6 4 p . Note also
that the AF is plotted as a function of range in contrast
to the usual delay parameter. The main reason for this is
that the AF of FMICW waveforms varies with range; the
sample delay of 64 p s corresponds to a range of 9.6 km.
4.1 FMCW
FMCW may be thought of as a special case of FMICW,
where AT(t) = AR(t) = 1. However, it varies from
FMICW in that AR(t) # 0, when A d t ) # 0. This case is
useful as a benchmark with which to compare FMICW
waveforms. In Fig. 3, it is shown that the design
resolution of 400 m has been obtained.
4.2 FMICW with random interrupt sequence
Random gating sequences are conventionally chosen as a
means to distribute the sidelobe energy of the ambiguity
function. Fig. 4a shows the interrupt sequence for a
simple random gating sequence. The durations T of the
periods within the sequence have a normal distribution,
as shown in Fig. 4b. This distribution was chosen as it
can be shown that, for this case, the ratio of the receive
power to the transmitted power is constant over all
target ranges. This minimises periodicities and produces
a near white noise floor of the ambiguity function as
shown in Fig. 4c. Note that the ambiguity noise floor is
only about 30 dB below the target level. This will seri-
ously limit the radar performance for weak targets in the
presence of stronger targets.
For ground wave H F radar, the radar range equation
[4) is defined as:
P G G oA2W;
( 4 7 ~ ) ~ R ~
p,=
where
P , = transmit power
P , = receive power
G , = transmit antenna gain
G R = receive antenna gain
o = radar cross-section
I = wave length
W, = attenuation function
R = range
Thus, the ratio of receive power from two ranges is
and for a transmit frequency of 6.75 MHz, the variation
in single strength from 50 km to 400 km is
0.7882 400 4 = dB
( T X i i K 3 )
414 1EE PROCEEDINGS-F, Vol. 138, NO. 5, OCTOBER 1991
-100 ' 1
0 50 100 150
I I
200 2 50 300 350 400
range, km
Fig. 3 Computer simulation of AF of FMCW
to = T/2 = 196 km
t
a
1
- I 90
t
b
Fig. 4
interrupt sequence
AF of FMICW with random
J a Transmit interrupt sequence
b Distribution of interrupt periods T with
variance U'
200 300 400 500 600 700 -100; 100
range, km
C
IEE PROCEEDINGS-F, Vol. 138, NO. 5, OCTOBER 1991
c AF corresponding to interrupt sequence in a
415
After acceptable processing gains the ambiguity noise
floor mus