Moving Target Indicator Radar (MTI)

Moving Target Indicator Radar (MTI)
Clutter is the term used for radar targets which are not of
interest to the user.
Clutter is usually caused by static objects near the radar
but sometimes far away as well:
n
n
1
Sometimes clutter can be caused by sidelobes in the
antenna pattern or a poorly adjusted antenna
n
n
n
In any case it is desirable to eliminate as much
clutter as possible n
4/15/2019 ELEC4600 Radar and Navigation 2

Engineering
This is done by using the fact that the desired target

is usually moving relative to the radar and thus
causes a Doppler shift in the return signal.
The Doppler shift can also be usedn to determine the
relative speed of the target. n

Engineering
The simple CW example can be modified to
incorporate pulse modulation
n
n
n
Note that the same oscillator is used transmission
and demodulation of the return
Thus the return is processed coherently
i.e. the phase of the signal is preserved
Engineering
The transmitted signal is
and the received signal is
n
these are mixed and the n
difference is extracted
n
This has two components, one is a sinen wave at Doppler
frequency, the other is a phase shift which depends on the
range to the target, R0

Engineering
Note that for stationary targets fd = 0 so Vdiff is constant
Depending on the Doppler frequency onentwo situations will occur

n
fd>1/τ n
fd<1/τ

Engineering
Note that for a pulse width of 1μs, the dividing line is a
Doppler shift of 1 MHz.
If the carrier frequency is 1 GHz, this implies a relative
speed of
n
30,000 m/s n
Thus all terrestrial radars operate in a sampled mode

and thus are subject to the rules of sampled
n signals
n
e.g. Nyquist’s criterion
We shall also see that we can use Discrete Sample
Processing (DSP) to handle some of the problems

Engineering
Looking a successive oscilloscope displays of radar
receiver output we see:
At the ranges of the two
targets, the return ampliude
is varying as the radar n
samples the (relatively) n
slow Doppler signal.
n
n
The bottom trace shows
what would be seen in real
time with the moving targets
indicated by the amplitude
changes
Engineering
We usually want to process the information
automatically. To do this we take advantage of the fact
that the amplitudes of successive pulse returns are
different:
n
n
n
OR
n
-
z-1

Engineering
MTI Radar Block Diagrams
MOPA (master oscillator, power amplifier)
Power amplifiers:
Klystron n
TWT (Travelling wave tube) n
Solid State (Parallel) n

Engineering
One of the problems with MTI in pulsed radars is that
magnetrons are ON/OFF devices.
i.e. When the magnetron is pulsed it starts

n
up with a random
phase and is thus its ouput is not coherent
n with the pulse
before it. Also it can not provide a reference oscillator to mix
with the received signals. n
Therefore some means must be provided to maintain the

coherence (at least during a single pulse period)

Engineering
n
n
n
PLL
n

Engineering
The notes have a section on various types of delay line

cancellers.
This is a bit out of date because almost all radars today
use digitized data and implementing the n required
n
delays is relatively trivial but it also opens up much
more complex processing possibilities
n

Engineering
The Filtering Characteristics of a Delay Line Canceller
The output of a single delay canceller is:
n
n
This will be zero whenever πfdT is 0 or

n
a multiple of π
n
The speeds at which this occurs are where n is
called the blind speeds of the radar 0,1,2,3,…
.

Engineering
If the first blind speed is to be greater than the highest

expected radial speed the λfp must be large
large λ means larger antennas for a given beamwidth

n
n
large fp means that the unambiguous range will be quite small
n
n
So there has to be a compromise in the design of an MTI radar

Engineering
The choice of operating with blind speeds or ambiguous

ranges depends on the application
Two ways to mitigate the problem at the expense

n of increased
complexity are: n
a. operating with multiple prfs n

b. operating with multiple carrier frequencies
n

Engineering
Double Cancellation:
Single cancellers do not have a very sharp cutoff at the nulls

which limits their rejection of clutter
n
(clutter does not have a zero width spectrum)
n
Adding more cancellers sharpens the nulls n

n

Engineering
There are two implementations:
n
n
n
These have the same frequency response: which is the
square of the single canceller response
v  4 sin (f d T )
2

Engineering
n
n

Engineering
Transversal Filters
These are basically a tapped delay line with the taps

n
summed at the output n

Engineering
Transversal Filters
To obtain a frequency response of sinnπfdT,

the taps must be binomially weighted i.e.
n
n

Engineering
Filter Performance Measures
( S / C ) out
MTI Improvement Factor, IC IC 
( S / C ) in
Note that this is averaged over all Doppler frequencies
n
Signal Clutter n
n
Filter
n
0 1/2T 0 1/2T
Engineering
Filter Performance Measures
Clutter Attenuation ,C/A
The ratio of the clutter power at the input

n
of the canceler
n
to the clutter power at the output of the canceler
n
It is normalized, (or adjusted) to the signal attenuation of
n
the canceler.
i.e. the inherent signal attenuation of the canceler is

ignored
23
Transversal Filters with Binomial Weighting with alternating
sign
Advantages:
n
Close to optimum for maximizing thenClutter
improvement factor ( S / C ) out
IC 
( S / C )nin
n
Also close to maximizing the Clutter Attenuation
Cin/Cout

Engineering
Transversal Filters with Binomial Weighting with alternating
sign
Disadvantage:
n
As n increases thesinn
filter cuts off more
n and more of the
spectrum around DC and multiples of PRF
n
This leads to wider blind speed zones and hence loss of
n
legitimate targets.

Engineering
The ideal MTI filter should reject clutter at DC and th PRFs
but give a flat pass band at all other frequencies.
The ability to shape the frequency response depends to a large

degree on the number of pulses used. Then more pulses, the
more flexibility in the filter design. n
Unfortunately the number of pulses is limited

n
by the scan rate
and the antenna beam width.
n
Note that not all pulses are useful:
The first n-1 pulses in an n pulse canceler are not useful

Engineering
Some other transversal filter responses are shown:
n
n
n
(1) 3 pulse canceler
(2) 5 pulse “optimum” filter which maximizes IC
(3) 15 pulse Chebyshev filter

Engineering
Feed forward (finite impulse response or FIR) filters have
only poles (one per delay)
More flexibility in filter design can be obtained if we use

recursive or feedback filters (also known as infinite impulse
response or IIR filters)
These have a zero as well as a pole per delay and thus have
twice as many variables to play with

Engineering
IIR filters can be designed using standard continuous-time
filter techniques and then transformed into the discrete form
using z transforms
Thus almost any kind of frequency response can be obtained
with these filters.
They work very well in the steady state case but unfortunately
their transient response is not very good. A large pulse input
can cause the filter to “ring” and thus miss desired targets.
Since most radars use short pulses, the filters are almost
always in a transient state
Engineering
Multiple PRFs
An alternative is to use multiple PRFs because the blind
speeds (and hence the shape of the filter response) depends on
the PRF and, combining two or more PRFs offers an
opportunity to shape the overall response.

Engineering
Multiple PRFs: Example
Two PRFs
Ratio 4/5
First blind speed is at

5/T1 or 4/T2

Engineering
Four PRFs
Ratio 25/30/27/31
First blind speed is at

about 28.25/TAV
where TAV
is the average period
of the four PRFs
Engineering
Calculation of First Blind Speed

if the relationship between the Pulse Periods is
n1 n2 nN
T1
 T2
T
N
and vB is the first blind speed of a PRF with average period
TAVE 
1
T1  T2  TN 
N
The first blind speed is v1  vB 

n1  n2 nN
N

Engineering
Multiple PRFs can also be used with transversal filters
Example
5 pulse canceler with
4 staggered PRFs n
n

Engineering
Digital (or Discrete) Signal Processing (DSP)
These days almost all radar signal processing is done using DSP.
The down-converted signal is sampled by an A/D converter as
follows:
Note:
The quadrature
channel removes
blind phases

Engineering
•Advantages of DSP
•greater stability
•greater repeatability
•greater precision
•greater reliability
•easy to implement multiple PRFs
•greater flexibility in designing filters
•gives the ability to change the system parameters
dynamically

Engineering
Note that the requirements for the A/D are not very difficult to meet
with today’s technology.
Sampling Rate
Assuming a resolution (Rres) of 150m, the received signal has to be
sampled at intervals of c/2Rres = 1μs or a sampling rate of 1 MHz
Memory Requirement
Assuming an antenna rotation period of 12 s (5rpm) the storage
required would be only 12 Mbytes/scan.

Engineering
Quantization Noise
The A/D introduces noise because it quantizes the signal
The Improvement Factor can be limited by the quantization noise

the limit being: 
I QN  20 log (2 N  1) 0.75 
This is approximately 6 dB per bit
A 10 bit A/D thus gives a limit of 60 dB
In practice one or more extra bits to achieve the desired
performance

Engineering
Dynamic Range
This is the maximum signal to noise ratio that can be handled by
the A/D without saturation 2 N 3
dynamic_ range  2 /k 2
N= number of bits
k=rms noise level divided by the quantization interval
the larger k the lower the dynamic range
but k<1 results in reduction of sensitivity
Note: A 10 bit A/D gives a dynamic range of 45.2dB

Engineering
Blind Phases
If the prf is double the
Doppler frequency then every
othyer pair of samples cane be
the same amplitude and will
thus be filtered out of the
signal.
By using both inphase and

quadrature signals, blind
phases can be eliminated

Engineering
Blind Phases
If the prf is double the
Doppler frequency then every
other pair of samples can be
the same amplitude and will
thus be filtered out of the
signal. (loss of 2.8dB and
13.7dB for Pd of 0.5 and 0.9
respectively)
By using both inphase and
quadrature signals, blind
phases can be eliminated

Engineering
Digital Filter Banks and FFT
Example: Transversal Filter with 8 delay elements:
z-1 z-1 z-1 z-1 z-1 z-1 z-1 z-1
w w w w w w w w
SUMMER
If the weights are set to:

 j 2 (i 1) k / N 
wik  e
i= 1,2,3…N and k is an index from 0 to N-1

Engineering
The impulse response of this filter is:
  t  (i  1)T e  j 2 (i 1)k / N
N
hk (t ) 
i 1
And from the Fourier transform its frequency
response is
 e j 2 (i 1) fT k / N 
N
H k ( f )  e  j 2fT
i 1
The magnitude of the response is

sinN ( fT  k / N 
Hk ( f ) 
sin ( fT  k / N 

Engineering
sinN ( fT  k / N 
Hk ( f ) 
sin ( fT  k / N 
k=0
0 1/T
k=1
0 1/T

Engineering
The response for all eight filters is:

Engineering
The actual response for this filters is:
Which is sin (Nx)/sin(x)
The sidelobes of this filter are quite large (13dB below

peak) and so it is not an ideal implementation

Engineering
Other options are:
Uniform weighting and Chebyshev weights
Clutter Improvement with Uniform weighting

(Filter Bank Only)

Engineering
Filter Banks can also be preceded by cancelers
In the following cases a 3 pulse canceler

Engineering
Filter Banks can also be preceded by cancelers
In the following cases a 3 pulse canceler is
used ahead of an 8 pulse Doppler Filter Bank
Uniform Chebyschev

Engineering
MTD (Moving Target Detector)
Example: ASR (Airport`Surveillance Radar)
Range: 60 NM Antenna Rotation: 12.5 - 15rpm
Frequency:2.7 - 2.9 GHz fp: 700 - 1200 Hz
Beamwidth: 1.4º
Uses a 3 pulse canceller with an 8 pulse doppler filter bank
Measured IC = 45dB
A/D converter uses 10 bits

Engineering
47.5 NM of range (22.5NM - 60NM) divided into 1/16 NM
intervals ( approximately 120m)
760 range bins
Azimuth is divided into 0.75º intervals (approximately 1/2
beamwidth)
480 azimuth bins
The eight doppler filters give 8 Doppler bins for a total of
2,918,00 range/azimuth/doppler cells
each cell has its own adaptive threshold level

Engineering
10 Pulses at one PRF is transmitted for 0.75º of azimuth and 10
pulses at the other PRF for the next 0.75º.
This eliminates second time around clutter

Engineering

Engineering
After the filter bank the doppler filter outputs are weighted to
reduce the effects of the sidelobes in the filter
The outputs of the filters are modified by subtracting 25% of

the outputs of the filters on either side.

Engineering

Engineering
Limitations of MTI Performance
Definitions:
•MTI Improvement Factor (IC) ( S / C ) out
• Defined earlier IC 
( S / C ) in
•Subclutter Visibility(SCV)
•The ratio by which a signal may be weaker than
the coincident clutter and stll be detected with the
specified Pd and Pfa. All radial velocities assumed
equally likely. SCV  (C / S ) IN
•Clutter Visibility Factor (VOC)
•The Signal to Clutter ratio
after filtering that provides the
specified Pd and Pfa.
VOC  ( S / C )OUT
Engineering
Definitions:
•Clutter Attenuation
CA  (CIN / COUT )
• Defined earlier
•Cancellation Ratio
n
•The ratio by which a signal may n be weaker than
the coincident clutter and stll be detected with the
specified Pd and Pfa. All radial velocities
n
assumed
equally likely.
n
•Note
C   S 
IC        (SCV )(VOC )
 S  IN  C OUT
Engineering
Equipment Instabilities:
Changes in signal from pulse to pulse will result in
apparent clutter Doppler shift
These changes can have many sources
n
• pulse to pulse change in amplitude
n
•pulse to pulse change in frequency

n
•pulse to pulse change in phasen

•timing jitter
•Changes in pulse width
•Changes in oscillator frequency between Tx and Rx
Engineering
Changes in signal from pulse to pulse will result in
apparent clutter Doppler shift
These changes can have many sources
n
• pulse to pulse change in amplitude
n
•pulse to pulse change in frequency

n
•pulse to pulse change in phasen

•timing jitter
•Changes in pulse width
•Changes in oscillator frequency between Tx and Rx
Engineering
Example: Phase variation
g1 and g2 are succesive pulses
for a stationary target n
n
The output of the two pulse filter is
n
Resulting in a minimum Improvement

Factor of

Engineering
n
Note that if we need IC=40dB, n
The pulse to pulse phase variation has

to be less than 0.01 rad (0.6º) n

Engineering
Some examples of the effects of instability
n
n

Engineering
Internal Fluctuation of Clutter:
Many sources of clutter are capable of motion of one sort or another
(translation, oscillation)
Examples:
Trees: leaves/branches oscillate withn
n
magnitude/frequency depending on the wind
speed
n than trees)
Vegetation: similar to trees (possibly less
n
Sea: translational motion with variation in phase
and magnitude
Rain: translational motion with oscillation due to
turbulence (thunderstorms)
Chaff: similar to rain with higher magnitude
Engineering
This fluctuation results in a
spreading of the clutter spectrum:
The broader the spreading, the less
effective is the MTI. n
n
Example:
Rain clouds
Chaff n
Sea Echo, high wind n
Wooded Hills,
20 mph wind
Sparse Hills,
Calm wind

Engineering
The spectrum of clutter can be expressed mathematically as a function
of one of three variables:
1. a, a parameter given by Figure 4.29 for various clutter types
2. σC, the RMS clutter frequency spread innHz
n
3. σv, the RMS clutter velocity spread in m/s
n
the clutter power spectrum is representedn by W(f)

Engineering
The power spectra expressed in the three parameters are:
1.
n
n
2.
n
n
3.

Engineering
Clutter Attenuation
The definition of Clutter Attenuation is:
n
n
n
The frequency response of a single delay line
n
filter is
2 sin( f d T )

Engineering
Clutter Attenuation
The definition of Clutter Attenuation is:
n
n
n
The frequency response of a single delay line
n
filter is
2 sin( f d T )
and the spectrum of the clutter in terms of frequency is
exp(  f 2 / 2 C2 )

Engineering
Clutter Attenuation
Then
n
n
Simplifying n

Engineering
Clutter Attenuation
Note that 2 V and c2
C  a 2
 8 V
n
Thus n
In terms of velocity
n
and n
In terms of the parameter a

Engineering
Improvement Factor
Since the average gain of a two pulse canceler is 2.
n
And for the three pulse canceler it is 6 n
n
In general

Engineering
Antenna Scanning Modulation
Since the antenna spends only a short time on the target, the spectrum
of any target is spread even if the target is perfectly stationary:
The two way voltage antenna pattern is
n
n
Dividing numerator and denominator of exponent

n by the scan rate
n

Engineering
Since and (time on target)
n
n
Taking the Fourier Transform n

Engineering
Since this is a Gaussian function the exponent must be of the form
f2
2 2f
n
Thus and n
(voltage
spectrum)
n
or n (power
spectrum)
or

Engineering
Substituting into the equations for IC ,
n
n

Engineering
Tracking with Scanning Radars
(Track While scan)
Automatic Detection and Tracking (ADT)
Pure tracking radars use pencil beam antennas to determine the
azimuth, elevation and range of a single target
n
Digital processing allows a surveillance radar
n to track many targets
and also permits prediction of the target positions in the future
which assists in collision avoidance.
n
Functions performed by ADT are:
n
•Detection •Track smoothing
•Track initiation •Track termination
•Track Association
•Track update
Engineering
(Track While scan)
Target Detection (Range)
As described before, the range is divided into range cells or
“bins” whose dimensions are close to the resolution of the radar.
n
Each cell has a threshold and if a predetermined
n number of
pulses exceed the threshold a target is declared present
Target Detection (Azimuth) n
Resolution better than antenna beamwidth n is obtained by
seeking the central pulse in the n pulses and using the azimuth
associated with that pulse.
Thus theoretically the resolution is fp
360  6  

Engineering
(Track While scan)
Track Initiation
Radar needs enough information to establish direction and speed
of all targets present.
n
If only one target present, only two scansn are required.
Usually require many scans to build up all of the tracks present
e.g. targets for two scans and two targetsn
n
1 2 1 2
2 1 2 1

Engineering
(Track While scan)
Track Association
Radar attempts to associate each detection with an existing track
n
n established track,
-establishes a search region (gate) for each
- if detection falls inside gate, it is assumed to be associated with
that track n

Engineering
(Track While scan)
Track Association
Tradeoffs
Close tracks - small gate
n
n
Small gate: can lose target if it
manoeuvres
Gate size is made variable depending onn
•Accuracy of track n
•Expected acceleration of the target

Engineering
(Track While scan)
Track Smoothing
Predictions are based on track history
predicted measured n
α-β Tracker n
xˆn  x pn   ( xn  x pn ) n
 n
vˆn  vn 1  ( xn  x pn )
TS
Predicted position:
x pn1  xˆn  vˆnTs
Engineering
(Track While scan)
Track Smoothing
The parameters α and β determine the bandwidth of the tracking
filter. Thus these have to be chosen.
n
One approach is to minimize the noise variance
n in the steady state
and the transient response to a ramp function. This gives:
2 And the bandwidth is determined by α
 n
2  n
α lies between 0 and 1
Tradeoff: good smoothing --- narrow bandwidth

rapid responses ---- wide bandwidth

Engineering
(Track While scan)
Track Smoothing
Another approach is the best linear track fitted to the radar data
(least squares) n
2(2n  1) 6 n
  2n((2nn11))
n(n  1) n(n  1)
n
n
n = number of scans

Engineering
(Track While scan)
Track Smoothing
Another approach is the best linear track fitted to the radar data
(least squares) n
2(2n  1) 6 n
 
n(n  1) n(n  1)
n
n
n = number of scans

Engineering
(Track While scan)
Kalman Filter, a maximum likelihood filter
1. Assume an estimate of the position and its error xn 1  x2
and of the measurement (radar) error  e2 n
n
 x2
2. Calculate the Kalman Gain bn `  2
 x  n e2
n
3. Obtain a measurement yn and calculate the estimated xn
xˆ n  xˆ n1  bn ( yn  xˆ n1 )

Engineering
(Track While scan)
Kalman Filter, a maximum likelihood filter
4. Update the variance of the estimate
 x2ˆ n (1  bn ) x2ˆ n1

n
 n
5. Recalculate the Kalman Gain

n
 x2 n
bn `  2
 x   e2

Engineering

Moving Target Indicator Radar (MTI)

Uploaded by

Copyright:

Available Formats

Moving Target Indicator Radar (MTI)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Moving Target Indicator Radar (MTI)

Uploaded by

Copyright:

Available Formats

Moving Target Indicator Radar (MTI)

4/15/2019 ELEC4600 Radar and Navigation 2

This is done by using the fact that the desired target

4/15/2019 ELEC4600 Radar and Navigation 3

The transmitted signal is

and the received signal is

4/15/2019 ELEC4600 Radar and Navigation 5

Note that for stationary targets fd = 0 so Vdiff is constant

Depending on the Doppler frequency onentwo situations will occur

4/15/2019 ELEC4600 Radar and Navigation 6

Thus all terrestrial radars operate in a sampled mode

4/15/2019 ELEC4600 Radar and Navigation 7

4/15/2019 ELEC4600 Radar and Navigation 9

Solid State (Parallel) n

4/15/2019 ELEC4600 Radar and Navigation 10

i.e. When the magnetron is pulsed it starts

Therefore some means must be provided to maintain the

4/15/2019 ELEC4600 Radar and Navigation 11

4/15/2019 ELEC4600 Radar and Navigation 12

The notes have a section on various types of delay line

4/15/2019 ELEC4600 Radar and Navigation 13

The Filtering Characteristics of a Delay Line Canceller

The output of a single delay canceller is:

This will be zero whenever πfdT is 0 or

4/15/2019 ELEC4600 Radar and Navigation 14

If the first blind speed is to be greater than the highest

large λ means larger antennas for a given beamwidth

4/15/2019 ELEC4600 Radar and Navigation 15

The choice of operating with blind speeds or ambiguous

Two ways to mitigate the problem at the expense

a. operating with multiple prfs n

4/15/2019 ELEC4600 Radar and Navigation 16

Single cancellers do not have a very sharp cutoff at the nulls

Adding more cancellers sharpens the nulls n

4/15/2019 ELEC4600 Radar and Navigation 17

There are two implementations:

4/15/2019 ELEC4600 Radar and Navigation 18

4/15/2019 ELEC4600 Radar and Navigation 19

These are basically a tapped delay line with the taps

4/15/2019 ELEC4600 Radar and Navigation 20

To obtain a frequency response of sinnπfdT,

4/15/2019 ELEC4600 Radar and Navigation 21

Clutter Attenuation ,C/A

The ratio of the clutter power at the input

i.e. the inherent signal attenuation of the canceler is

4/15/2019 ELEC4600 Radar and Navigation 24

4/15/2019 ELEC4600 Radar and Navigation 25

The ability to shape the frequency response depends to a large

Unfortunately the number of pulses is limited

Note that not all pulses are useful:

The first n-1 pulses in an n pulse canceler are not useful

4/15/2019 ELEC4600 Radar and Navigation 26

4/15/2019 ELEC4600 Radar and Navigation 27

More flexibility in filter design can be obtained if we use

4/15/2019 ELEC4600 Radar and Navigation 28

4/15/2019 ELEC4600 Radar and Navigation 30

First blind speed is at

4/15/2019 ELEC4600 Radar and Navigation 31

First blind speed is at