UNIT-5

DETECTION OF SIGNALS IN NOISE

Receiver Noise

Noise is the unwanted energy that interferes with the ability of the receiver to detect
the wanted signal. It may enter the receiver through the antenna along with the desired
signal or it may be generated within the receiver. In underwater sonar systems
external acoustic noise is generated by waves and wind on the water surface, by
biological agents (fish, prawns etc) and manmade sources such as engine noise. In
radar and lidar sensors the external electromagnetic noise is generated by various
natural mechanisms such as the sun and lightning amongst others. Manmade sources
of electromagnetic noise are myriad, from car ignition systems and fluorescent lights
through other broadcast signals.
As discussed earlier, noise within the sensor is generated by the thermal motion of the
conduction electrons in the ohmic portions of the receiver input stages. This is known
as Thermal or Johnson Noise.
Noise power PN is expressed in terms of the temperature To of a matched resistor at
the input of the receiver

PN = kTo β W,

where: k – Boltzmann’s Constant (1.38×10-23 J/K),

To – System Temperature (usually 290K),
β – Receiver Noise Bandwidth (Hz).

The noise power in practical receivers is always greater than that which can be
accounted for by thermal noise alone.

The total noise at the output of the receiver, N, can be considered to be equal to the
noise power output from an ideal receiver multiplied by a factor called the Noise
Figure, NF

N = PN FN = kTo β .NF W.
 Noise Probability Density Functions
Consider a typical radar front-end that consists of an antenna followed by a wide band
amplifier, a mixer that down converts the signal to an intermediate frequency (IF)
where it is further amplified and filtered (bandwidth βIF ). This is followed by an
envelope detector and further filtering (bandwidth βV = βIF/2).

The noise entering the IF filter is assumed to be Gaussian (as it is thermal in nature)
with a probability density function (PDF) given by

1 − v2
p (v ) exp
2πψ o 2ψ o

where p(v)dv - probability of finding the noise voltage v between v and v+dv,
ψo - variance of the noise voltage.

If Gaussian noise is passed through a narrow band filter (one whose bandwidth is
small compared to the centre frequency), then the PDF of the post-detection envelope
of the noise voltage output can be shown to be

R − R2
p( R) exp
ψo 2ψ o

where R is the amplitude of the envelope of the filter output. This has the form of the
Rayleigh probability density function

Figure Amplitude distributions of thermal noise pre and post detection

Probability of False Alarm
A false alarm occurs whenever the noise voltage exceeds a defined threshold voltage,
Vt, as illustrated in the figure below. The probability of this occurring is determined
by integrating the PDF as shown
∞
R − R2 −V 2
Pr ob(Vt < R < ∞) = ∫ψ
Vt o
exp
2ψ o
dR = exp t = Pfa
2ψ o

It can be seen that the average time interval between crossings of the threshold, called
the false alarm time, Tfa, can be written as
N
T fa =
N →∞ N
∑T
k =1
k

where Tk – Time between crossings of the threshold Vt by the noise envelope (when
the slope of the crossing is positive).

The false alarm probability could also have been defined as the ratio of the time that
the envelope is above the threshold to the total time as shown graphically in the figure
below
N

∑t k
tk 1
Pfa = k 1
= ave
=
N
Tk T fa β
∑T
k =1
k
ave

where tk and Tk are defined in the figure, and the average duration of a noise pulse is
the reciprocal of the bandwidth β.

Tk Tk+1
Envelope of Noise Voltage

tk tk+2
Threshold
tk+1
Vt
RMS Noise Voltage
√Ψο

Time

Figure : Receiver output voltage illustrating false alarms due to noise

For a bandwidth β = βIF, the false alarm time is just

1 Vt 2
T fa exp
β IF 2ψ o

The false alarm times of practical radars must be very large (usually a couple of
hours), so the probability of false alarm must be very small, typically Pfa < 10-6.
 Probability of Detection
Consider that a sine wave with amplitude, A, is present along with the noise at the
input to the IF filter. The frequency of the sine wave is equal to the centre frequency
of the IF filter. It is shown by Rice that the signal at the output of the envelope
detector will have the following PDF (known as a Rician distribution)

R ⎛ R 2 + A2 ⎞ ⎛ RA ⎞
ps ( R ) = exp⎜⎜ − ⎟Io ⎜ ⎟
ψo ⎝ 2ψ o ⎟⎠ ⎜⎝ ψ o ⎟⎠

Io(Z) modified Bessel function of order zero and argument Z. It can be shown that for
large Z, an asymptotic expansion for Io(Z) is

eZ ⎛ 1 ⎞
I o (Z ) ≈ ⎜1 + + ... ⎟
2πZ ⎝ 8Z ⎠

The probability that the signal will be detected is the same as the probability that the
envelope, R, will exceed the threshold, Vt is
∞ ∞
R ⎛ R 2 + A2 ⎞ ⎛ RA ⎞
pd = ∫ ps ( R)dR = ∫ ψ o ⎜⎜⎝ − 2ψ o ⎟⎟⎠ I o ⎜⎜⎝ ψ o ⎟⎟⎠dR
exp
Vt Vt

Unfortunately, this cannot be evaluated in a closed form and so numerical techniques

or a series approximation must be used. However, this has already been done, and
tables and a series of curves have been produced.

In terms of the PDFs for the noise and the signal plus noise voltages, the detection and
false alarm process is shown graphically in the figure below. The lightly shaded area
represents the Pfa and the dark shaded area (including the tail covered by the light
shading) represents the Pd.

0.8

Rayleigh
Probability Density

Noise
0.6
Detection Ricean
Threshold Signal + Noise
SNR = 15dB
0.4

0.2

4 8 12 14

Amplitude (V)

Figure : PDF’s of noise and signal plus noise for Pfa = 10-2
 Figure : Detection probability as a function of signal to noise ratio with false alarm
probability as a parameter

A typical radar system will operate with a detection probability of 0.9 and a
probability of false alarm of 10-6. The required signal to noise ratio can be read
directly off the graph as 13.2dB. Note that this is for a single pulse of a steady
sinusoidal signal in Gaussian noise with no detection losses.
Detector Loss Relative to an Ideal System
An envelope detector is used by a radar system when the phase of the received pulse
is unknown. This is called non-coherent detection, and it results in a slightly higher
SNR requirement than the curves above show.

Figure : Envelope detector loss as a function of signal to noise ratio

This loss factor Cx is approximately

SNR (1) 2.3

C x (1)
SNR (1)

where Cx(1) – Loss in SNR,

SNR(1) – The pre detector single pulse SNR required to achieve a particular Pd
and Pfa.

The graph shows that for good SNR, the detector loss is very small. In the case where
the Pd = 0.9 and the Pfa = 10-6 it is only about 0.4dB. This effect is known as small
signal suppression as it becomes much more pronounced as the SNR decreases.

One of the advantages of coherent detection is that it has zero response to the
quadrature noise component, whereas, this component is translated into phase
modulation after envelope detection with the result that the effective SNR in the latter
case is degraded. However the phase modulation is very small for high SNR, and few
advantages can be gained by coherent detection for SNR > 10dB (Cx = 1dB) over
envelope detection.
 The Matched Filter
To achieve the best possible SNR, the characteristics of the IF filter must be matched
to those of the signal pulse.

The peak signal to (average) noise power ratio of the output response of the matched
filter is equal to twice the received signal energy, E, divided by the single-sided noise
power per Hz, No

ˆ⎞
⎜ S ⎟ = 2E
⎜N⎟
⎝ ⎠out N o

where Ŝ – Peak instantaneous signal power seen during the matched filter response
to a pulse (W),
N – Average noise power (W),
E – Received signal energy (J),
No – Single sided noise power density (W/Hz).

The received energy is the product of the received power, S, as determined by the
range equation and the pulse duration, τ

E = Sτ ,

And the noise power density is the received noise power, N, divided by the
bandwidth, βIF

N
No
IF

Substituting into (10.13)

ˆ⎞
⎜ S ⎟ = 2Sτ = ⎛⎜ S ⎞⎟ 2 BIFτ
⎜N⎟
⎝ ⎠out N BIF ⎝ N ⎠in

When the bandwidth of the signal at IF is small compared to the centre frequency then
the peak power is approximately twice the average power in the received pulse. So the
output SNR is

⎛S⎞ ⎛S⎞
⎜ ⎟ ≈ ⎜ ⎟ BIFτ
⎝ N ⎠out ⎝ N ⎠in

The matched filter should not be confused with the circuit theory concept of matching
that maximises power transfer rather than SNR.

In practise a matched filter implementation is often hard to achieve exactly, so

compromises are made as shown in the following table:
 Table 1: Efficiency of non-matched filters

Input Signal Matched Filter Optimum B.τ Loss in SNR

Shape Characteristic compared to
Matched Filter (dB)
Rectangular Pulse Rectangular 1.37 0.85
Rectangular Pulse Gaussian 0.72 0.49
Gaussian Pulse Rectangular 0.72 0.39
Gaussian Pulse Gaussian 0.44 0 (matched)
Rectangular Pulse Single tuned circuit 0.4 0.88
Rectangular Pulse Two cascaded tuned 0.613 0.56
circuits
Rectangular Pulse Five cascaded tuned 0.672 0.5
circuits

Integration of Pulse Trains

The relationships developed earlier between SNR, Pd and Pfa apply to a single pulse
only. However, it is possible to improve the radar performance by averaging a number
of returns.

For example, as a search-radar beam scans, the target will remain in the beam
sufficiently long for more than one pulse to hit it. This number, known as hits per
scan, can be calculated as follows

θb f p θb f p
nb = =
θ&s 6ωm

where nb – Hits per scan,

θb – Azimuth beamwidth (deg),
θ&s - Azimuth scan rate (deg/s),
ωm – Azimuth scan rate (rpm).

For a typical ground based radar with an azimuth beamwidth or 1.5°, a scan rate of
5rpm and a pulse repetition frequency of 30Hz, the number of pulses returned from a
single point target is 15.

The process of summing all these hits is called integration, and it can be achieved in
many ways some of which were discussed in Chapter 5. If integration is performed
prior to the envelope detector, it is called pre-detection or coherent integration, while
if integration occurs after the detector, it is called post-detection or non-coherent
integration.

Pre-detection integration requires that the phase of the signal be preserved if the full
benefit of the summing process is to be achieved and because phase information is
destroyed by the envelope detector, post-detection integration, though easier to
achieve, is not as efficient.

If n pulses are perfectly integrated by a coherent integration process, the integrated

SNR will be exactly n times that of a single pulse in white noise. However, in the
non-coherent case, though the integration process is as efficient, there are the detector
losses discussed earlier that reduce the effective SNR at the output of the envelope
detector.

Integration improves the Pd and reduces the Pfa by reducing the noise variance and
thus narrowing the Noise and Signal + Noise PDFs as shown in the figure below

Figure

Effect of integration on signal and noise PDFs before and after integration

For n pulses integrated, the single pulse SNR required to achieve a given Pd and Pfa
will be reduced. However, this results in increased detector losses, and hence a
reduced effective integration efficiency.
The integration efficiency may be defined as

SNR (1)
Ei (n) = ,
nSNR (n)

where: EI(n) – Integration efficiency,

SNR(1) – Single pulse SNR required to produce a specific Pd if there is no
integration,
SNR(n) – Single pulse SNR required to produce a specific Pd if n pulses are
integrated perfectly.

The improvement in SNR if n pulses are integrated, post detection is thus nEI(n). This
is the integration improvement factor, or the effective number of pulses integrated.
 Figure : Integration improvement factor as a function of pulses integrated

The integration loss in dB is defined as follows

⎛ 1 ⎞
L (n) = 10 log10 ⎜⎜ ⎟⎟
⎝ E ( n) ⎠

The integration improvement factor is not a sensitive function of either the Pd or the
Pfa as can be seen by the clustering of the curves in the figure above.

Figure Integration loss as a function of the number of pulses integrated

 Detection of Fluctuating Signals
The discussion in the previous section assumes that the signal amplitude does not vary
from pulse to pulse during the integration period. However, from the discussion of
target cross section in Chapter 9, it is obvious that the RCS of any moving target (with
the exception of a sphere) will fluctuate with time as the target aspect as seen by the
radar changes.

To properly account for these fluctuations, both the probability density function and
the correlation properties with time must be known for a particular target and
trajectory. Ideally, these characteristics should be measured for a target, but this is
often impractical. An alternative is to postulate a reasonable model for the target
fluctuations and to analyse the effects mathematically.

Four fluctuation models proposed by Swerling are used:

• Swerling 1:Echo pulses received from the target on any one scan are of
constant amplitude throughout the scan, but uncorrelated from scan to scan.
The PDF is given by
1 −σ
pσ = exp
σ av σ av

where σav is the average cross section over all target fluctuations.

• Swerling 2: The PDF is as for case 1, but the fluctuations are taken to be
independent from pulse to pulse.
• Swerling 3: The fluctuations are independent from scan to scan, but the PDF
is given by
4σ 2σ
p (σ ) exp
σ av
2
σ av

• Swerling: The PDF is as for case 3, but the fluctuations are independent from
pulse to pulse.
• Swerling 5: Non fluctuating

Figure : Radar returns for different Swerling fluctuations

The PDF for cases 1 and 2 is indicative of a target with many (>5) scatterers of equal
amplitude. These are typical for complicated targets like aircraft, while the PDF for
cases 3 and 4 is indicative of a target with one large scatterer and many small
scatterers.

As one would expect, the single pulse SNR required to achieve a particular Pd
(for Pd > 0.4) will be higher for a fluctuating target than for a constant amplitude
signal. However, for Pd < 0.4, the system takes advantage of the fact that a fluctuating
target will occasionally present echo signals larger than the average, and so the
required SNR is lower.

Figure : Effect of target fluctuation on required signal to noise ratio

To cater for fluctuating targets and integration, a further set of curves describing the
integration improvement factor have been developed.

Figure : Integration improvement factor as a function of the number of pulses integrated for
fluctuating targets
If these curves are examined in isolation, it would appear that the integration
efficiency EI(n) > 1 under certain conditions. One is not getting something for nothing
as the single pulse SNR is much higher in these cases than it would be for the single
pulse case.

The procedure for using the range equation when one of these Swerling targets is as
follows:
1. Find the SNR for the single pulse, non-fluctuating case that corresponds to the
Pd and Pfa required using the curves in Figure.10.4
2. For the specific Swerling target, find the additional SNR needed for the
required Pd. using the curves in Figure 10.10
3. If n pulses are to be integrated, the integration improvement factor II(n)=nEI(n)
is then found using the curves in Figure 10.11
4. The SNR(n) and nEI(n) are substituted into the range equation along with σav
and the detection range found.
 Air Traffic Control Radar Performance

Type: 2D air surveillance radar

Band: L
Frequency: 1250 to 1350MHz
Peak Power: 5MW
Antenna size: 12.8×6.7m
Antenna Gain: 36dB (lower beam)
34.5dB (upper beam)
Beam shape: Cosec2
Elev beamwidth: 4° Cosec2 to 40°
Azim beamwidth: 1.25°
Scan: Mechanical
Scan Rate: 6rpm
PRF: 360pps
PRF Stagger: Quadruple
Pulse width: 2μs
Noise Figure: 4dB

Calculate the theoretical detection range for a 1m2 aircraft target if the detection
probability Pd = 0.9, and the mean time between false alarms is 9 hours.

Matched Filter Assumptions: Rectangular Pulse

Second order bandpass filter
Loss 0.56dB
B.τ = 0.613
 θ b f p 1.25 × 360
Hits per Scan: Use the formula nb = = = 12.5 [use nb = 10]
6ϖ m 6×6

1
False Alarm Probability: Use the formula Pfa =
T fa β

From the matched filter assumptions B.τ = 0.613 and the pulsewidth τ = 2μs, the IF
bandwidth β = 306kHz.

Tfa = 9 hrs = 32400s

1
Pfa = = 10−10
32.4 × 10 × 306 × 103
3

Single Pulse SNR: Use the curves in Figure.10.4 for Pd = 0.9 and Pfa = 10-10

SNR(1) = 15.2dB

Fluctuating Target: Use the curves in Figure.10.10 for Swerling 2 for an aircraft

Additional SNR required 8dB

Pulse Integration: Use the curves in Figure.10.11 for 10 pulses integrated, a

Swerling 2 target and Pd = 90% to obtain an improvement factor of 15dB

Total n Pulse SNR Required: Add up the requirements

SNR(10) = 15.2+8-15 = 8.2dB

Applying the Radar Range Equation

We assume the following losses:

Transmitter Line = Ltx = 2dB (incorporated into Tx power)
Receiver Line = Lrec = 2dB (incorporated into receiver noise figure)

Losses:

1D Scanning Loss = 1.6dB

Matched Filter = 0.56dB
CFAR Loss = 0.7dB
Misc. Loss = 1.3dB
Total L = 1.6+0.56+0.7+1.3 = 4.16dB

Pt G 2λ2σ
Pr =
(4π )3 R 4 L
Transmitter Power: 10log10(5×106) = 67dBW

Less Ltx = 2dB

Radiated peak power Pt = 65dBW

Antenna gain: (Lower beam) 36dB

Radar cross section: σ = 10log10(1) = 0dBm2

λ2
Constant: 10 log10 = -45.7dB
(4π )3
Received Power: Calculated from the range equation

Pr = 65+36+36-45.7+0-4.16-40log10R

= 87.14-40log10R dBW

Receiver Noise: Use the equation N = PN FN = kTsys BFN

K = Boltzmann’s Constant 1.38x10-23 J/deg

Tsys = Temperature 290K

B = 306×103 Hz

We are given the receiver noise figure in dB

10log10(Fn) = 4dB

N = 10 log10 (kTsys B ) + 10 log10 ( Fn ) + Lrec = −149 + 4 + 2 = −143dBW

Received Signal to Noise Ratio

SNRrec = Pr − N = 87.14 − 40 log10 R + 143.0 = 230.1 − 40 log10 R

This must equal the required single pulse SNR if 10 pulses are integrated
SNR(10) to achieve the specified Pd and Pfa.

8.2 = 230.1 - 40log10R

Solving for the Detection Range

Solve for R

R = 352.8km
Atmospheric Attenuation

The atmospheric attenuation αdB at L-and is about 0.003dB/km (one way)

So the total two way attenuation over 353km is 2.1dB which will result in a
significant reduction of the detection range.

These equations are best solved graphically using MATLAB.

The graph below shows two received power curves, one that does not take into
account the atmospheric attenuation which intersects the SNR(10) curve at 353km.

The other graph which takes attenuation into account intersects the SNR(10) curve at
320km.

Figure : Graphical solution to the radar range equation including the atmospheric loss