2

Target Detection

2.1 INTRODUCTION
This chapter addresses an important aspect of radar design: the detection of tar-
gets in real-world environments. The theory of radar detection is covered thor-
oughly in many text books, including references [2] through [5]. The chapter’s
main objective is to describe the practical application of target detection con-
cepts to radar design and analysis.
First, target detection is described as the determining the presence of the “de-
sired” object in different environments, including:

• “Undesired” thermal noise

• Reflections from natural and man-made objects (i.e., clutter).

The commonly used Swerling target models are described, and their applica-
bility is discussed. Next, target detection in both surface and volume clutter is
described. Last, multiple-pulse detection methods are described, including co-
herent and non-coherent integration.

33
34 Chapter 2

2.2 TARGET RADIO FREQUENCY (RF) SCATTERING

MODELS
Real targets are not usually “point scatterers” of incident radio frequency (RF)
energy (i.e., point targets), but are a complex combination of individual scatter-
ers dependent on the radar viewing geometry, RF operating frequency, and
bandwidth. For very wide modulations bandwidth radars (e.g., B ≥ 500 MHz),
the radar can usually resolve individual RF scattering centers on the target’s
physical configuration. As long as these resolved scatterers are not obscured
from the radar (e.g., via blockage or geometry), they will generally appear to the
radar as nonfluctuating echoes. Narrowband radars, however, will not resolve
individual scatterers. This situation leads to the constructive and destructive
combination of scattering sources, and ultimately results in target fluctuations.
Figure 2.1 illustrates these concepts pictorially for an aircraft target.
Figure 2.1 addresses the concept of RF scattering from air targets encountered
by air defense radars. Scattering centers arise due to specular points and discon-
tinuities on a target’s surface when illuminated by a short radar pulse. Returns
from an air target may result from reflections of the nose, cockpit, wings, engine
cavities, and engine turbine blades of an aircraft as depicted in Figure 2.2. As
stated previously, the composite target response will be a function of whether or
not these scatterers are resolved by the radar’s waveform.
The appropriate target scattering models are a function of physical structure
(e.g., size, shape, scattering centers, and reflected polarizations), radar operating
frequency, and waveform and processing parameters (e.g., bandwidth, integra-
tion time).

Figure 2.1 Wideband Target Scattering versus Radar Resolution

 Target Detection 35

Figure 2.2 RF Scattering from Air Targets

Air targets are in general complex scattering sources that consist of many indi-
vidual scattering centers as indicated in Figure 2.2. Alternatively, ballistic missile
target objects encountered by missile defense radars are typically simpler tar-
gets. Figure 2.3 depicts two such targets, a re-entry vehicle (RV) or warhead and
an attitude control module (ACM) or post-boost vehicle (PBV).

Figure 2.3 Representative Ballistic Missile-Shaped Targets

36 Chapter 2

Simulated radar cross sections (RCSs) for two tumbling ballistic missile-shaped
objects are shown in Figure 2.4 as a function of time. As can be seen, there is a pe-
riodic structure to the amplitudes of the radar returns due to differences in target
scattering as the objects tumble relative to the radar line-of-sight (LOS).

2.3 TARGET DETECTION IN NOISE

This section reviews some fundamental detection theory. For a nonfluctuating
target (i.e., constant amplitude), the return radar echo is of amplitude A and du-
ration T. The return in each radar (resolution) cell is compared to a fixed thresh-
old voltage set above background thermal noise. When the return amplitude
exceeds the threshold a “target” is declared. Otherwise, “no target” is declared
per the test:

H1
>
r (t ) V , (2.1)
< T
H0

Figure 2.4 Simulated Return Amplitudes for Tumbling RV and Booster-Shaped Objects
 Target Detection 37

where r(t), VT, H0, and H1 are the radar return from the target, the detection
threshold, and the hypotheses for no target (i.e., the null hypothesis) and for tar-
get a target present, respectively.
The probability of detection (PD) for the nonfluctuating target is a function of
the signal-to-noise ratio (SNR):

 A2 
SNR =   ( k TS B ) =
 2 
−1
( )
A2 N
2 0 B =
 A2 T 
 2  N 0
, (2.2)

where Ts, N0, and B are the system noise temperature, noise power, and noise
bandwidth, respectively. The target detection problem is depicted graphically in
Figure 2.5.
Prior to detection, at the quadrature detector, the target voltage is:

Vt ( t ) = A ( t ) cos (ω c t ) + A ( t ) sin
I Q (ω c t ) , (2.3)

Figure 2.5 Illustration of the Target Detection Problem

38 Chapter 2

where AI , AQ, and ωc are the in-phase and quadrature waveform envelopes and
radian carrier frequency, respectively. The noise voltage at the quadrature detec-
tor is:

Vn ( t ) = X I ( t ) cos (ω ct ) + YQ ( t ) sin (ω ct ) . (2.4)

The signal-plus-noise complex-envelope is then:

{  A (t ) + X (t )  }
1
r (t ) = +  AQ ( t ) + YQ ( t ) 
2 2 2
I I
. (2.5)
1
=  X ( t ) + Y ( t ) 
2 2 2

If the noise statistics are bi-variable Gaussian (with zero-mean), the joint prob-
ability density in r and θ (electrical phase angle) is:

( )
f ( r,θ ) = r exp Ap2 2σ 2 exp  − r 2 − 2r AX cosθ + Ay sinθ
 ( ( )) 2σ 2  , (2.6)


where:

2 2 2
A p = A x+ A y . (2.7)
2σ = Total Noise Power
2

By integrating over θ, the density for r becomes:

( ) (
fr ( r ) = r exp − Ap2 2σ 2 exp − r 2 2σ 2 I o r Ap σ 2 r σ 2 , ) ( ) (2.8)

where I0 is the Bessel function of the first kind. The PD is given by the integral of
(1) from VT (threshold) to ∞:

( )
∞
PD = ∫ − VT
fr (r ) dr = Q Ap σ , VT σ , (2.9)
 Target Detection 39

where Q is the Marcum Q-function. Equation (2.9) can be evaluated for a specific
SNR and detection threshold, VT. Usually, an acceptable probability of false
alarm, PFA, is specified. For the Rayleigh noise distribution, the PFA is given by:

α VT 2
∞ −  22 −

∫
σ 
PFA = α
σ2
e dα = e 2σ 2 . (2.10)
− VT

Solving for the detection threshold yields:

VT 2 = ( − n P ) 2σ FA
2 . (2.11)

For a Swerling I radar cross section fluctuation model:

1  γ ,
f γ (γ ) = exp  − (2.12)
SNR  SNR 

where: SNR = Average SNR .

1
∞ ∞
PD = ∫
VT
fγ (γ ) dγ = ∫ fγ (γ ) dγ = ( PFA ) 1+ SNR . (2.13)
SWI

The Swerling I fluctuating target model assumes an exponential RCS (or

power) probability density; or equivalently, a Raleigh voltage probability den-
sity. Another possible target model exhibits RCS fluctuations that follow a chi-
square distribution with 4 degrees-of-freedom:

 γ    2γ   .
fγ (γ ) = 4  2  exp  −   (2.14)
 SNR    SNR  

This is the Swerling III model. As to the applicability of these models, note that:

• Swerling I is appropriate for slowly fluctuating (e.g., scan-to-scan for search)

RCS dominant point targets
• Swerling III is more appropriate for slowly fluctuating dominant multiple-
scattering center targets.
40 Chapter 2

Integrating the Swerling III density from VT to ∞ yields:

 
PD
SW III
= 
(
2
)
 2 + SNR 2   ((
T ) )
  2 + V σ 2 / 2 SNR + SNR 2 2 

  . (2.15)
 V
( ) ( 
)
2
exp  − T σ / 2 / 1 + SNR / 2 
 

The benefit of Swerling III (SW III) over Swerling I (SW I) is primarily when non-
coherently integrating pulses at higher SNRs. SW III is superior to SW I for de-
tection purposes when SNR ≥ 10 dB.
Some detailed Swerling model definitions and characteristics are:

• For two N-pulse bursts, 1 per scan:

– Return amplitudes from N pulses on Scan A are statistically independent of
amplitudes from N pulses on Scan B (i.e., {A1, A2, ... , AN}, {B1, B2, ... , BN})
• Swerling I & III assume Set of A amplitudes have identical value, and Set of B
amplitudes have identical value: A1 = A2 = ... = AN and B1 = B2 = ... = BN
• Swerling I assumes fγ (γ) from equation (2.12) and Swerling III assumes equa-
tion (2.14) applies
• Swerling II and IV assume each reflection in Sets A & B are statistically inde-
pendent, and that equation (2.12) and (2.14) apply, respectively: A1 not equal
to A2, etc., and B1 not equal to B2, etc.

In summary, Swerling I & III are slowly fluctuating targets (i.e., scan-to-scan,
search frame-to-frame). Swerling II & IV assume fast RCS fluctuations (i.e.,
pulse-to-pulse). Swerling I & III can be converted to Swerling II and IV, respec-
tively, if pulse-to-pulse RF changes are sufficient to de-correlate the target of in-
terest (a function of target shape, etc.).

2.4 TARGET DETECTION IN CLUTTER

The previous section discussed target detection in thermal noise. In this section
the effects of backscatter from natural and man-made objects will be modeled as
 Target Detection 41

an augmentation to a thermal noise background. Doppler processing is a key

technique for mitigating clutter, since in general, the targets of interest are mov-
ing faster than the apparent clutter velocity. Two basic classes of waveforms,
with different matched filter (MF) processors, are:

• Moving target indicator (MTI) waveforms and processing (cancellers)

• Pulse-Doppler (PD) waveforms and processing (Doppler filter banks).

Clutter discriminants are features that allow separation of targets from clutter
returns. Candidate clutter discriminants include:

• Elevation

• Velocity

• Polarization

• Carrier frequency sensitivity

• Azimuth angle sensitivity

• Signal bandwidth

Each of these is discussed next.

Target elevation: Most often, target returns are from above “ground level” and
land clutter returns are not; therefore, elevation is useful for separating targets
from ground clutter. Target velocity: The most commonly used feature for sepa-
rating targets from ground, sea, and weather clutter. These clutter types are gen-
erally stationary to slowly moving. Since most targets of interest exhibit higher
radial speeds, velocity can be a useful clutter discriminating feature. Target po-
larization sensitivity: Certain RF polarizations reflect from targets differently
than from undesired objects (e.g., rain backscatter). An example is the use of cir-
cular polarization, which can suppress rain clutter relative to target objects by
up to 10 dB. Carrier frequency sensitivity: Certain RF frequencies maximize tar-
get-to-clutter responses, and use of frequency diversity can maximize target-to-
clutter ratio. RF bandwidth: Higher RF bandwidth reduces range cell dimension
(and increases range resolution) and therefore reduces area (sea, land) and vol-
ume (rain, chaff) clutter backscatter.
42 Chapter 2

The primary quantity of interest in characterizing clutter magnitude is the

equivalent radar cross section of the clutter, since the signal-to-clutter ratio
(SCR) (for main beam clutter at the target range) is given by:

(S C ) =
RCSt arg et .
RCSclutter
(2.16)

Two basic types of clutter must be addressed: (i) area clutter (land, sea), and
(ii) volume clutter (rain, chaff). The defining relationships for these two clutter
types are provided by the effective radar cross section of the clutter:


 σ AC ,
σC =   (2.17)
 σ VC

where σ°, AC, and VC are the normalized clutter coefficient, area of the surface
clutter, and volume of the volumetric clutter, respectively. The units of σ° are
(m2/m2) for surface clutter and (m2/m3) for volume clutter to result in the clutter
RCS having units of m2.
Clutter area and volume are defined as:

 Rcτ 
AC =  tan φ θ AZ (2.18)
 2 

and:

( )
VC = R 2 θ AZ θ EL cτ 2 , (2.19)

where R, c, τ, and φ are the clutter range, speed of light, pulse length, and de-
pression angle, respectively. θAZ and θEL are the antenna azimuth and elevation
beamwidths.
In many cases of interest, clutter is neither in the antenna main beam nor at the
target range. For this more general case, the SCR is derived from the radar range
equation for received power defined as:
 Target Detection 43

t tσ Ar
PG
Prec = , (2.20)
(4 π R ) 2 2
Lt Lr

where σ is the effective RCS of either the target or clutter. For surface clutter, the
signal-to-clutter ratio is defined as the ratio of the received target power to the
received clutter power:

(4 π R ) 2 2
t tσ A r
2
PG c Lt Lr
SCR =
(4 π R )
2 2
Lt Lr t tσ clutter Ar
PG '2

 , (2.21)
2
 Ar   R  3  σ 
SCR = c
 A'   R   
 σ   c τ  tan φ θ 
4
r
  2  AZ 

where Rc , R, Ar , and A' r are the range to the clutter, the range to the target, the
full antenna aperture, and the effective antenna aperture of the sidelobe upon
which the clutter impinges, respectively.
Since velocity (or more precisely range-rate or radial speed) is a key clutter
discriminant, the Doppler frequency shift of the RF carrier frequency is usually
exploited to mitigate clutter returns’ effect on target detection. Two major classes
of Doppler techniques are:

• MTI waveforms and MTI delay line cancellers

• Pulse-Doppler (PD) waveforms and Doppler filter banks.

Both types will be discussed. However, in brief, the difference between the two
methods is:

• MTI:
– Excellent performance against stationary narrowband clutter; simple pro-
cessing that uses two to three pulses per detection
• Pulse Doppler:

– Excellent performance against stationary or moving clutter. Uses coherent

burst (N-pulse) waveform and a bank of N-Doppler filters for detection.
44 Chapter 2

Moving target indicator waveforms and processing exploit the low-speed na-
ture of land, and sea and weather clutter. A typical MTI application uses three
pulses (two delays), and the processing shown in Figure 2.6.
The response of the three-pulse (two-delay) MTI canceller shown is given by
Figure 2.7.
Clutter attenuation (CA) of an MTI canceller is defined as:

Cin
CA = , (2.22)
Cout

where Cin and Cout are the input and output clutter powers, respectively.
The Doppler-frequency dependent improvement factor, I(fD), is given by:

( Signal − to − Clutter ) out

I ( fD ) = . (2.23)
( Signal − to − Clutter ) in

Figure 2.6 Three-Pulse MTI Processor

Figure 2.7 MTI Canceller Frequency Response

 Target Detection 45

and:

( Signal ) out
I ( fD ) =
( Signal ) in
CA ( for a linear receiver ) . (2.24)

The average improvement factor (averaged over frequency) for a two-pulse

canceller is defined as the clutter attenuation and is given by:

−1 2
 
(
 2 πσ T ) 
2
I2 = CA ≈ =  PRF σ  19.75 , (2.25)
 f
  f

where σf , T, and PRF are the clutter spectral width, pulse repetition interval
(PRI), and pulse repetition frequency, respectively. For the three-pulse canceller,
the CA is given by:

4
 PRF 
I3 ≈   780 . (2.26)
 σf 

The generalized MTI canceller transfer function is given by:

H M ( f ) = gM 1 − exp ( − j 2π fD T )

M
(2.27)

or:

m  M  M
= gM Σ ( −1)   exp ( − jm 2π fD T );
M
HM ( f )  m  = binomial coefficient
m=0  m
(2.28)

The weights for M = 1, 2, 3, and 4 are:

M = 1 ⇒ (1, −1)
M = 2 ⇒ (1, −2, 1)
. (2.29)
M = 3 ⇒ (1, −3, 3, − 1)
M = 4 ⇒ (1, −4, 6, − 4, 1)
46 Chapter 2

These correspond to M = 1: single-delay or two-pulse canceller; M = 2: double-

delay or three-pulse canceller; M = 3: triple-delay or four-pulse canceller; and M
= 4: quadruple-delay or five-pulse canceller. For all cases, the sum of weights
equal to zero ensures a null in the response at zero Doppler. For many applica-
tions, a three-pulse canceller with PRF stagger will satisfy clutter cancellation re-
quirements.
When clutter is nonstationary, such as sea clutter at high sea states, or rain
clutter in medium to high wind, the mean Doppler shift is well above zero fre-
quency. For these types of clutter, MTI cancellers are not effective since the clut-
ter spectrum is not centered in the canceller notch.
An alternative waveform and processing to MTI cancellers is the pulse-Dop-
pler (PD) waveform (an N-pulse coherent burst) processed using a bank of
Doppler filters after the range matched filter. PD waveforms have several ad-
vantages over the simpler MTI waveforms, and associated cancellers, includ-
ing:

• Targets receive close to the theoretical optimum SNR gain relative to noise (N2
versus N)
• PD waveform processing using a bank of Doppler filters can deal with non-
zero or frequency-offset Doppler clutter
• More degrees of freedom are available to tailor the effective transfer function,
H(f), when a bank of Doppler filters are used.

The simplest implementation of a Doppler filter bank is the output of FFTs

(fast Fourier transforms [FFTs]) in the Doppler domain, computed from the
range-cell-sampled, matched-filter output sub-pulses of the pulse train. A nar-
rowband target will “integrate up” in one (or two or three) Doppler filters. Typi-
cally, one Doppler filter will contain the majority of the target energy. Figure 2.8
illustrates the PD waveform and its processing.
The advantages of PD waveforms and Doppler processing have been de-
scribed (i.e., coherent target gain, mitigates nonzero Doppler clutter). However,
the specific “post processing” associated with Doppler processing needs to be
described.
First, some form of automatic detection processing is desirable. A constant
false alarm rate (CFAR) processor is usually employed for this purpose. The con-
 Target Detection 47

Figure 2.8 Pulse-Doppler Waveforms and Processing

figuration typically used is a CFAR processor subsequent to each Doppler filter.

Next, the filters dominated by clutter must be identified and ignored for subse-
quent targets return candidacy. Automatic methods to detect clutter are based
upon CFAR background estimation, as well as a prior knowledge of likely filters
(frequency or “speed” likelihood) that might contain clutter (e.g., using a clutter
map).
Typically, the zero and N – 1th filters are dominated by stationary clutter, espe-
cially at low altitudes. Moving clutter (e.g., sea clutter at higher sea states, blow-
ing rain, etc.) will usually occur in the lower and (symmetrically) higher
Doppler filters.

2.5 MULTIPLE-PULSE DETECTION

The previous section addresses the detection of targets in clutter using multiple
pulses and either the MTI or pulse-Doppler signal processing approach. How-
ever, it is often necessary to integrate multiple pulses to enhance the SNR to an
adequate level for detecting weak targets or targets at long slant ranges. This sec-
tion addresses three multiple-pulse detection techniques: (i) binary (or M-out-of-
N) integration, (ii) non-coherent integration, and (iii) coherent integration.
48 Chapter 2

2.5.1 Binary Integration

This technique uses multiple pulses to increase detection probabilities by defin-
ing a detection criterion of detecting at least M times out of N opportunities. This
technique enhances detectability at the slight cost of increased probability of
false alarm. When this technique is employed, the effective probability of detec-
tion is given by:

N  N! 
∑  ( N − m )! m!  ( P ) (1 − PD )
m N −m
PD M − out − of − N = D . (2.30)
m=M

For the special case of M = 1, equation (2.30) reduces to:

PD M − out − of − N = 1 − ( 1 − PD ) .
N
(2.31)

Since there are N opportunities for a false alarm when using this technique, the
probability of false alarm for M-out-of-N detection is given by:

PFA M − out − of − N = N PFA . (2.32)

2.5.2 Non-Coherent Integration

A second type of multiple-pulse detection is non-coherent integration. This tech-
nique adds pulses in a root-mean-square (RMS) sense before applying the
threshold test in equation (2.1). This approach can be beneficial when adding
pulses for fluctuating targets at reasonable SNRs as compared to the coherent in-
tegration approach described in the next section. In addition, since no phase in-
formation is used in this technique, non-coherent integration can be used in
conjunction with frequency diversity which, as discussed in Section 3, can im-
prove detectability for Swerling target models.
Due to the RMS addition of target returns in non-coherent integration, the ef-
fective SNR of the integrated return is approximately given by:

N
SNR NCI ≈ ∑
i =1
SNR i , (2.33)
 Target Detection 49

where is the SNR of the ith return from the transmitted pulse train. When the
SNRs of each return are approximately equal, equation (2.34) becomes:

SNR NCI ≈ N SNR 1 , (2.34)

where SNR1 is the SNR of each return. Equation (2.34) is the rule of thumb for
the expected benefit of non-coherent integration relative to single pulse detec-
tion. Now, since the noise components of each return are added in an RMS sense,
the effective PFA is given by:

PFA NCI = N PFA 1 , (2.35)

where PFA 1 is the probability of false alarm for each return.

2.5.3 Coherent Integration

A third type of multiple-pulse detection is coherent integration. This technique
adds pulses as voltages with knowledge of phase before applying the threshold
test in equation (2.1). Equivalently, coherent integration can be considered to be
vector addition of the multiple returns. However, since phase information is
used in this technique, coherent integration cannot be used in conjunc-
tion with frequency diversity, which destroys coherency from return-to-
return. In addition, this type of integration is limited by the coherency
time which is the shorter of the times that the radar hardware and tar-
get remain coherent. In most cases, the target correlation time con-
stant is the limiting component.
Due to the vector addition of target returns in coherent integration, the effec-
tive SNR of the integrated return is approximately given by:

N
SNR CI ≈ ∑
i =1
SNR i , (2.36)

where SNRi is the SNR of the ith return from the transmitted pulse train. When
the SNRs of each return are approximately equal, equation (2.34) becomes:

SNR CI ≈ N SNR 1 , (2.37)

50 Chapter 2

where SNR1 is the SNR of each return. Equation (2.37) is the rule of thumb for
the expected benefit of coherent integration relative to single pulse detection.
Since the noise components of each return are again added in an RMS sense for
coherent integration, the effective PFA is given by:

PFA NCI = N PFA 1 , (2.38)

where PFA is the probability of false alarm for each return. This is the same as for
non-coherent integration.

2.6 REFERENCES
[1] A. Papoulis, Probability, Random Variables, and Stochastic Processes, McGraw-Hill,
1965
[2] D. K. Barton, Modern Radar System Analysis, Artech House, 1988
[3] D. K. Barton, Radar System Analysis and Modeling, Artech House, 2004
[4] H. Van Trees, Detection, Estimation and Modulation Theory, Part 1, Wiley-Interscience,
2001
[5] J. DiFranco & W. Rubin, Radar Detection, SciTech, 2004
[6] E. Brookner, Aspects of Modern Radars, Artech House, 1988
[7] N. Levanon, Radar Principles, Wiley-Interscience, 1988
[8] R. Nitzberg, Radar Signal Processing and Adaptive Systems, 2nd Edition, Artech House,
1999
[9] M. Skolnik, Introduction to Radar Systems, 3rd Edition, McGraw-Hill, 2002
[10] M. Skolnik, Radar Handbook, 2nd Edition, McGraw-Hill, 1990
[11] S. Haykin & A. Steinhardt, Adaptive Radar Detection and Estimation, Wiley, 1992

2.7 PROBLEMS
1. Consider a radar that employs a fixed-noise threshold to achieve a specified
probability of false alarm. Calculate the threshold-to-noise ratio to achieve
false alarm probabilities of 10–M where M = 3, 4, and 5. Note that:

 VT 2 
 2σ 2  = − n PFA , (2.39)
 

where 2σ2 is the noise power at the output of the matched filter.
 Target Detection 51

2. Consider a radar being degraded by barrage noise jamming. For a radar re-
ceiver with system noise (kTB) of –143 dBm, a jammer effective radiated
power (PJ GJ) of 10 W, a 0.01 square meter radar antenna aperture (equiva-
lent to a –30 dB sidelobe), with the jammer at 1,000 km, calculate the neces-
sary thresholds for the probabilities of false alarm from problem 1. If the
radar has 15 dB SNR in the clear (i.e., no jamming) at a specified target slant
range, calculate the detection probabilities with and without jamming for a
Swerling I target model for these values of PFA. Remember that:

PJ GJ Ar '
N effective = kTs Br + .
( 4π R )J
2

3. Calculate the probability of detection for a Swerling III target with nominal
SNRs of 10, 15, and 20 dB, respectively, for PFA = 10–6. Note that for Swerling
III RCS fluctuations:

 
P =  2
(( ) )
  2 + V σ 2 / 2 SNR + SNR 2 2 
(
 2 + SNR  
) T 
d 2
SW III
 
 V
( ) ( 
)
2
exp  − T σ / 2 / 1 + SNR / 2 
 

Compare these results with the Swerling I calculations for the same SNRs
and PFA.

4. Consider a radar with a peak power of 100 kW, 45 dB transmit antenna gain,
a 10 square meters antenna aperture, a receiver with system noise (kTB) of
–143 dBm, transmit and receive losses of 3 dB each, with a 0 dBsm target at a
slant range of 500 km. Using the received power at the radar aperture from
the relationship below along with:

t tσ Ar
PG
Prec =
(4 π R ) 2 2
Lt Lr
52 Chapter 2

and

 
 Ar   R   
2 3
σ
SCR = c
 A'   R   
 σ   c τ  tan φ θ 
4
r
  2  AZ 


the expression for signal-to-clutter ratio for surface clutter, calculate the SNR
and SCR for values of σ° of 10–M where M = 4, 5, and 6, a clutter range of 20
km, a 10 microsecond pulse length, a depression angle (φ) of 6 degrees (coin-
cident with a –20 dB antenna sidelobe), and a 4-degree 3 dB beamwidth in
azimuth.

5. For clutter with 3 dB spectral widths of 5, 10, and 15 Hz, calculate the clutter
attenuation achieved by two- and three-pulse MTI cancellers with a PRF of
750 Hz, where:

I = CA 4
 PRF 
−1
 
2
I3 ≈ 
(
I 2 ≈  2 πσ f T  ) 
2
=  PRF σ   σf  780
   f 19.75

and where:

σf = 1-sigma spectral width of clutter (Hz)

PRF = pulse-repetition frequency waveform (Hz)
T = pulse repetition interval = 1/PRF (sec).

6. For the SCRs calculated in problem 1, apply the three-pulse MTI canceller
assuming a clutter spectral width of 5 and 10 Hz and a PRF of 750 Hz to cal-
culate the improved SCRs.

Next, calculate the detection probabilities assuming a PFA of 10–6 and a

Swerling I target model for the resulting signal-to-clutter-plus-noise ratios
after using the MTI canceller (use the SNR calculated for problem 1).

Compare these results to the Pd in the clear (i.e., noise-only).

 Target Detection 53

Note the relationship:

S 1 SNR SCR
= =
C+N 1
+
1 SNR + SCR
SCR SNR

Comment on the results. Which cases are clutter-limited? Noise-limited?

7. If each return from a three-pulse transmission follows a Swerling I target
model and the desired probability of false alarm is 10–6, calculate the proba-
bilities of detection if coherent integration and 1-out-of-3 detection ap-
proaches are employed for SNRs of 5 and 15 dB. Comment on the results.