 |
Description  |
|
|
This invention relates to radar systems generally, and more specifically to
arrangements for reducing range sidelobes in radar systems using Doppler
processing of received echoes.
The high speed and long range of modern airborne vehicles places increasing
range demands on radar systems used for tracking. The long-range
requirement also requires the use of relatively high transmitted power to
reliably detect small targets. High transmitted power implies a relatively
higher peak transmitter power or a longer duration transmitter pulse (also
known as a "wider" pulse). Assuming a maximum available peak power, longer
range implies a longer duration transmitted pulse. A longer duration pulse
tends to reduce range resolution, which is the ability to distinguish
among targets which are at similar ranges. Pulse compression techniques
can be used to improve range resolution in spite of the longer pulse
duration, thus eliminating the need for high peak power short pulses, but
the minimum range at which a target can be detected increases with the
transmitted pulse length. Thus, if the transmitter pulse duration is 100
microseconds (.mu.s), the minimum distance at which a target may be
detected is about 8 nautical miles (nm). Clearly, a radar using pulses of
such a duration cannot be used to detect close-in targets, as for example
aircraft which are landing or taking off from an airport at which the
radar is situated. An additional problem associated with pulse compression
is the appearance of range sidelobes (as distinguished from antenna
sidelobes) in addition to the main range lobe. The time position, or
range, of the main lobe is the position that is tested for the presence of
a target and for estimating the parameters of that target (reflected
energy or power, closing speed, fluctuations in echo power and closing
speed, etc.). The presence of range sidelobes on the compressed pulse
results in interfering echoes which originate at ranges other than the
range of the main lobe. This interference, known as "flooding," can cause
erroneous estimates of the echo characteristics in the range cell (i.e.,
range increment) covered by the main lobe. Prior art techniques for
suppressing range sidelobes include the "zero-Doppler" technique, in which
the range sidelobe suppression scheme is based in part upon the assumption
that the interfering echoes, as well as the desired echo, have a closing
velocity that has no significant Doppler phase change or shift over the
duration of the uncompressed pulse. The Doppler phase shift .phi..sub.DV
across the uncompressed pulse is 2.pi. times the product of the Doppler
frequency shift and the uncompressed pulse duration (i.e. .phi..sub.DV
=2.pi. f.sub.d T.sub.O radians). When this Doppler phase shift is actually
zero or very small, moderate sidelobe suppression is achievable with the
zero Doppler design. However, the zero Doppler design is very sensitive to
small Doppler frequency shifts, making deep sidelobe suppression
impossible for radar applications in which very deep sidelobe suppression
is desired, as for example in weather mapping, clear air turbulence
detection, and microburst detection.
Copending U.S. patent application Ser. No. 07/685,792, filed Apr. 16, 1991
in the name of Urkowitz, describes a pulse radar system in which Doppler
processing is used to separate returns into frequency bins representative
of radial speed. Interference from scatterers at other ranges is reduced
by range sidelobe suppression filtering applied to the signal in each
frequency bin. A radar with an improved range sidelobe suppression
arrangement is desired.
SUMMARY OF THE INVENTION
In a radar system, first and second pulse sets are recurrently transmitted.
The first set of pulses is dispersed in time pursuant to a first phase
code, and the second set of pulses is dispersed in time pursuant to a
second phase code which is complementary to the first. The echoes from the
target are received to form received first and second pulse sets. The
echoes are processed by separation into frequency bins, ordinarily
referred to as Doppler filtering. Thus, the received pulse sets are
separated by frequency, and also by incremental time of receipt, which
corresponds to range. Within each frequency band, the received first pulse
set is filtered by a first-code-matched filter, and the second pulse set
is filtered by a second-code-matched filter. The matched-filtered received
first and second pulse sets have range sidelobes which are of mutually
opposite polarity. The matched-filtered received first and second pulse
sets, after suitable delay of the matched filtered received first set, are
summed together, whereby the range main lobes add and the range sidelobes
cancel.
In a particular embodiment of the invention, the Doppler-filtered returned
pulses are received sequentially. A first code-matched filter filters the
first pulse sequence, and a switch is operated between the end of the
first pulse sequence and before the beginning of the second pulse
sequence, to decouple the first code-matched filter, and to couple in-line
a second code-matched filter. The second code-matched filter then filters
the second pulse sequence. A delay associated with the first code-matched
filter delays the matched-filtered first pulse sequence until matched
filtering of the second pulse set is accomplished, whereupon the delayed
first set is summed with the first set.
DESCRIPTION OF THE DRAWING
FIG. 1 is a simplified block diagram of a radar system as described in the
abovementioned Urkowitz patent application;
FIG. 2 is a simplified block diagram of a portion of the arrangement of
FIG. 1 illustrating a prior art range sidelobe suppression arrangement;
FIG. 3 is a simplified block diagram of an improved range sidelobe
suppression arrangement which may be used in the radar of FIG. 1, as
described in the abovementioned Urkowitz application;
FIG. 4 is a simplified block diagram of a portion of the arrangement of
FIG. 3;
FIG. 5 is a simplified block diagram of another improved range sidelobe
suppression arrangement which may be used in the radar of FIG. 1, as
described in the above mentioned Urkowitz application;
FIG. 6 is a simplified block diagram of a pulse compression and range
sidelobe suppression filtering portion of the arrangement of FIG. 5;
FIG. 7 is a simplified block diagram of another embodiment of a pulse
compression and sidelobe suppression filter which may be used in the
arrangement of FIG. 5, in which range-sample-rate and chip-rate filters
are separated;
FIG. 8 is a simplified block diagram of the arrangement of FIG. 5 using the
principle of the filter of FIG. 7, rearranged to have a lower parts count,
as described in the abovementioned Urkowitz application:
FIG. 9 represents amplitude-frequency spectra and the way they are
rearranged in FIGS. 3, 5 and 8;
FIGS. 10a-10i, collectively referred to as FIG. 10, are amplitude-time
representations useful in explaining autocorrelation of the subpulses of a
pulse set;
FIGS. 10i-10s are amplitude-time representations of the results of the
autocorrelations of FIGS. 10a-10i, respectively;
FIGS. 11a and 11b illustrate a pulse set which is complementary to the
pulse set of FIG. 10, and the result of its autocorrelation, respectively;
FIG. 12 represents the summing of the autocorrelated waveforms of FIGS. 10
and 11;
FIG. 13 represents a pulse transmitted by the arrangement of FIG. 1 in
order to allow processing by the arrangement of FIG. 14;
FIG. 14 is a simplified block diagram of a signal processor according to
the invention, for performing matched filtering of first and second
complementary pulse sets; and
FIG. 15a is a simplified block diagram of a signal processor according to
the invention, for performing matched filtering of first and second
complementary pulse sets, and FIG. 15b is a simplified block diagram of a
structure useful in performing matched filtering with differing weights in
the arrangement of FIG. 15a.
DESCRIPTION OF THE INVENTION
FIG. 1 is a simplified block diagram of a radar system as described in the
abovementioned Urkowitz application. In FIG. 1, an antenna 18 is connected
by way of a transmit-receive (T/R) duplexing or multiplexing system 50 to
a transmit controller (TX)3. Controller 3 establishes system the pulse
duration, PRF, frequency and the like, and provides other control
functions including generation of local oscillator and tuning signals.
Antenna 18, controller 3 and T/R 50 together cause transmission of
electromagnetic signals, illustrated as 7, and couple echoes of the
electromagnetic signals received by antenna 18 by a path 9 to a receiver
and analog signal processor (ASP) 52 for low-noise amplification,
frequency downconversion, and the like, with the aid of local oscillator
(L.O.) signals. In their broadest concept, there are conventional radar
techniques. The resulting baseband signals may, in general, include
orthogonal inphase (I) and quadrature (Q) components. The baseband signals
are applied from receiver/ASP 52 to an analog-to-digital converter (ADC)
associated with a block 62, which converts the analog baseband signals to
digital form with the aid of system timing signals. The "range clock"
portion of the timing signals establishes the smallest time interval into
which the received signals are quantized, and therefore establishes the
smallest discernible target range increment.
As described in the abovementioned Urkowitz application, a buffer may be
associated with ADC 62 of FIG. 1 for purposes unrelated to the present
application. The digital signals are coupled from ADC 62 (or its buffers,
if used) to a digital signal processor (DSP) 68.
FIG. 2 is a simplified block diagram of a portion of the processing which
might be included in DSP block 68 of FIG. 1 for prior-art range sidelobe
reduction. In FIG. 2, an I+jQ signal from the complex analog-to-digital
converter in block 62 is applied by way of an input port 210 to a pulse
compressor illustrated as a block 212. The input I+jQ signal is desirably
in digital form, but may be analog, and represents a sequence of pulses
reflected from the target at a particular beam position. Pulse compressors
are known in the prior art and may be implemented, for example, by a
surface acoustic wave (SAW) filter matched to the transmitted pulse code
in an analog system before the downconversion to baseband I+jQ, or as a
processor in a digital system. The output of pulse compressor 212 is a
relatively short-duration pulse with unwanted range sidelobes. A range
sidelobe suppressor 214 acts on the compressed pulse to reduce the range
sidelobes. Range sidelobe suppressor 214 may be implemented as a further
processor operating upon digital I+jQ baseband signal. Such a processor is
designed on the assumption of zero Doppler shift. As in the case of the
pulse compressor, range sidelobe suppression, based upon the same
assumption of zero Doppler shift, may instead be applied by a further SAW
filter in the analog portion of the radar receiver before conversion to
digitized baseband I+jQ. However, such an approach is not Doppler tolerant
and represents prior art over which the arrangement of the abovementioned
Urkowitz application is an improvement. The compressed, sidelobe reduced
pulses are applied from suppressor 214 to a bank of narrow-band Doppler
filters illustrated together as a filter bank 216. Each filter element of
bank 216 responds to a particular narrow frequency band f.sub.0, f.sub.1,
f.sub.2 . . . f.sub.M-1, thereby separating the incoming signal into a
plurality of frequency bins, the frequencies of which depend upon the
Doppler frequency attributable to the radial velocity of the target. FIG.
9 illustrates a baseband spectrum f.sub.0 and additional spectra f.sub.1 ,
f.sub.2, f.sub.3 . . . f.sub.M-1, which together represent the output
signals from filter bank 216. An echo having a given Doppler shift
produces a substantial output from only one filter output. For best
velocity selectivity, the bandwidths of filter elements f.sub.0, f.sub.1,
f.sub.2 . . . f.sub.M-1 of filter bank 216 of FIG. 2 are narrow, in the
range of a few Hertz or less. The bank of Doppler filters represented as
block 216 may be implemented by a signal processor performing a discrete
Fourier transform (DFT) by means of a fast Fourier transform (FFT)
algorithm. The output of each filter is a range trace which is the sum of
a sequence of Doppler filtered range traces. A particular filter output,
therefore, represents target echoes having the particular Doppler
frequency shift corresponding to its center frequency, and a small range
of Doppler shifts about that center frequency, which depends upon the
bandwidth of the filter. The output of each filter is coupled to a
corresponding amplitude detector 218a, 218b, 218c . . . 218m, to generate
signals which, when arrayed, can be sorted according to the velocity of
the target by selecting the appropriate detector output. Thus, the
presence of a target signal at the output of a Doppler filter indicates
that the target has a particular radial velocity. Within each Doppler
frequency bin, the target range is known from the time of arrival of the
signal. The signals produced by detectors 218 are coupled to threshold
circuits in DSP block 68, to allow separation of significant returns from
noise, and thence for further processing. The circuits fed by the various
Doppler filter elements f.sub.0, f.sub.1, f.sub.2, . . . f.sub.M-1, may
each be considered a "Doppler channel." Thus, filter element f.sub.0 and
detector 218a constitute a Doppler channel relating to targets with a low
radial velocity, while filter element f.sub.2 and detector 218b together
constitute another Doppler channel relating to targets with a larger
radial velocity, corresponding to f.sub.2.
In the context of the Urkowitz application, DSP block 68 of FIG. 1 may
perform the functions of (a) pulse-to-pulse Doppler filtering by means of
a Fast Fourier Transform (FFT) algorithm, with data weighting to control
signal leakage from neighboring Doppler shifts (frequency leakage); (b)
digital pulse compression; (c) range sidelobe suppression; and (d) further
signal processing including CFAR (constant false alarm rate) processing,
thresholding for target detection, spectral processing for weather
mapping, etc. Items (a) and (d) are performed in ways well understood in
the art, and form no part of the invention. The range sidelobe suppression
(c) is advantageously Doppler tolerant as described in the abovementioned
Urkowitz application, and as described below in conjunction with FIGS.
3-9. The results of the processing done in block 68 may include (a) target
detection reports (aircraft); (b) radar track detection reports; (c)
weather components for each resolvable volume of space, including (c1)
echo intensity; (c2) echo closing speed, and (c3) spectral spread of the
echo, and these components of information may be included in Digitized
Radar Detection Reports (DRDR). The DRDR reports may also include data
relating processing. A person skilled in the art of pulse compression will
know that the radar pulse must be coded in some manner that allows DSP
block 68 to correlate received signals with the known transmitted pulse
code. The correlation process simultaneously improves the signal-to-noise
ratio and the range resolution of target echoes. A person skilled in the
art knows that a variety of satisfactory pulse coding techniques are
available in the prior art. Such techniques include the well known Barker
Codes, pseudorandom noise codes, and linear FM coding techniques. DSP
block 68 therefore also performs digital pulse compression on the received
signals.
As mentioned, the pulse compression in block 312 of FIG. 2 gives rise to
range sidelobes, which are in the form of amplitude responses representing
times other than the actual time of the return from the target in
question, and thus represent other possible ranges. This introduces an
ambiguity or error in the apparent echo from a specific resolvable range
interval because the echoes from other range intervals "flood" into the
range interval of interest from the range sidelobes. The reason is that
the total echo at any instant of time is the RF sum of all the echoes from
all ranges covered by the compressed pulse with its range sidelobes or
suppressed range sidelobes. As described in the aforementioned Urkowitz
application, the range sidelobe suppression represented by block 214 may
provide substantial range sidelobe suppression for certain phase shifts
attributable to the Doppler frequency shift, and less suppression at other
phase shifts. Thus, the prior art range sidelobe suppression can be
optimized for a particular value of radial velocity of a target, but
provides less suppression at other velocities.
The quantity which controls the sensitivity of the range sidelobe
suppression is the product of the uncompressed transmitter pulse duration
and the Doppler frequency shift. The product may be measured as the
Doppler phase shift over the uncompressed pulse duration.
In accordance with an aspect of the improvement described in the
abovementioned Urkowitz application, range sidelobes are suppressed by a
technique which includes separating the sequence of target echoes or
pulses into a plurality of Doppler or frequency "bins", and applying range
sidelobe suppression to each bin separately.
This is illustrated in the simplified block diagram of FIG. 3. Elements of
FIG. 3 corresponding to those of FIG. 2 are designated by like reference
numerals. The processor of FIG. 3 uses a plurality of range sidelobe
suppressors 328a, 328b, 328c . . . 328m, one of which is associated with
each Doppler filter element f.sub.0, f.sub.1, f.sub.2, . . . f.sub.M-1 of
Doppler filter bank 216, i.e. with each Doppler channel. It would be
possible to make each range sidelobe suppressor with different filtering
parameters to optimize the range sidelobe suppression for the center
frequency of the associated Doppler filter element. This would
substantially improve the overall range sidelobe suppression, because the
range of frequencies at the output of each filter is small, on the order
of a few Hertz. This may represent a small percentage of the center
frequency of the filter. Thus, each range sidelobe suppressor may be
optimized at one frequency, and its performance will not be excessively
degraded by the small phase shifts attributable to a range of frequencies
which is a small percentage of the optimized frequency. To avoid the need
for different suppression parameters in each of the range sidelobe
suppressors so that identical suppressors may be used for cost reasons,
the filtered output signal from each filter element of filter bank 1216
(except the lowest-frequency filter element f.sub.0) is converted to a
common frequency range. A suitable range is the "baseband" range of filter
element f.sub.0, which may for example be the range extending from zero
Hertz to a few Hertz. In FIG. 3, the output from filter element f.sub.0 of
filter bank 216 is applied directly to a Zero Doppler Sidelobe Suppressor
(ZDSS) 328a, because the output frequency range of filter element f.sub.0
is already at baseband, and therefore no frequency conversion is
necessary. The outputs from all the other filter elements f.sub.1, f.sub.2
. . . f.sub.M-1 are individually applied to multipliers 320 for converting
each filter output to baseband. For example, filter element f.sub.1 of
filter bank 216 has its output connected to a first input port of a
multiplier 320b. Multiplier 320b has a second input port coupled to an
oscillation source (not illustrated in FIG. 3) of signal
exp(-j2.pi.f.sub.1 k.tau..sub.0), k=0, 1, . . .
where
f.sub.1 is the center frequency of the corresponding filter element of
filter bank 216,
.tau..sub.0 is the range sampling period, and
k is the integer time index.
The oscillator frequency is thus the negative (i.e., same absolute
frequency but 180.degree. out-of-phase) of the center Doppler frequency at
which the corresponding filter element of filter bank 216 is centered. For
example, the oscillator signal exp(-j2.pi.f.sub.2 k.tau..sub.0) applied to
multiplier 320c is the negative of frequency f.sub.2 at which filter
element f.sub.2 of filter bank 216 is centered. Any initial phase shift
associated with the oscillator signal is unimportant, because eventually
only the magnitudes of the Doppler channel signals are used. Essentially,
the output signals of the individual elements f.sub.1, f.sub.2 . . .
f.sub.M-1 of Doppler filter bank 216 are heterodyned by multipliers 220 to
be centered at zero frequency, whereupon identical zero frequency Doppler
range sidelobe suppressors (ZDSS) 328 may be used in each Doppler channel.
For example, ZDSS 328a is coupled to filter element f.sub.1, and provides
baseband range sidelobe reduction; ZDSS 328b is coupled to the output of
multiplier 320b for receiving therefrom filtered signals originally at
f.sub.1 but downconverted to baseband, and suppresses sidelobes in the
baseband signal. The process of downconversion is illustrated generally in
FIG. 9, in which filtered signals at frequencies f.sub.1 . . . f.sub.M-1
are converted to baseband by the multiplying processes represented by
arrows 912, 913, 914, . . . 91m. Each of the other ZDSS 328c . . . 328m of
FIG. 3 also receives signals downconverted to baseband. Thus, all ZDSS are
identical. The outputs of ZDSS 328a . . . 328m are applied to detectors
218a . . . 218m, respectively. The detected signals in each channel are
coupled for thresholding and further processing, in known manner.
FIG. 4 illustrates a tapped delay line or transversal filter of the type
known as a "finite impulse response" (FIR) filter, because a change in the
input causes a change in the output which extends over a finite time. The
FIR filter of FIG. 4 may be used as any range sidelobe suppressor 328 in
the arrangement of FIG. 3. For definiteness, the structure of FIG. 4
represents zero Doppler sidelobe suppressor (ZDSS) 328b of FIG. 3. As
illustrated, ZDSS 328b of FIG. 4 includes a delay structure 440 which
receives signal at its input port 442 and causes the signal to propagate
to the right, past taps illustrated as nodes 444a, 444b . . . 444n. The
temporal spacing (delay) between adjacent taps equals range sampling
period .tau..sub.0. The delay structure, if in digital form, may be a
shift register. Each node 444 is coupled to a tap weight multiplier
illustrated by a triangular symbol 446a, 446b . . . 446n. The weighted,
delayed signals from multipliers 446 are applied to a combinatorial summer
(.SIGMA.) 450 for producing the desired filtered or range sidelobe
suppressed signals. The summed signals are applied from the output of
summer 450 to detector 218b of FIG. 3. The number of taps, and the weights
to be applied.
FIG. 5 is a simplified block diagram of another arrangement described in
the above-mentioned Urkowitz application, which is better suited to larger
Doppler frequency shifts and/or larger duration-bandwidth products than
the structure of FIG. 3. Elements of FIG. 5 corresponding to those of FIG.
3 are designated by like reference numerals. In FIG. 5, the I+jQ signal,
representing the complex envelope of the radar echo, plus whatever
receiver noise is combined with the echo, is applied by way of port 210 to
Doppler filter bank 216, without being pulse-compressed. Filter bank 216
separates the signal into frequency bins, and applies the signal in each
bin to a separate processor 28, which performs the functions of both pulse
compression and range sidelobe suppression (PC & SS). As with the
arrangement of FIG. 3, the output from the lowest-frequency bin, namely
the f.sub.0 bin, is applied directly to its associated processor 428a,
without a multiplication or frequency conversion. The output signals from
filter elements f.sub.1 though f.sub.M-1 are individually applied to a
corresponding multiplier 320. For example, the output port of filter
element f.sub.2 of filter bank 216 is applied to an input of a multiplier
320c. Multiplier 1420c also receives from a source (not illustrated in
FIG. 15a) an oscillation signal exp(-j2.pi.f.sub.2 k.tau..sub.0) which is
the negative of the center frequency of filter element f.sub.2. As
described above, this has the effect of converting the signal output of
filter element f.sub.2 to baseband. The baseband signal at the output of
multiplier 320c is applied over a data path 321c to PC & SS 428c. The
output signals of each of the other filter elements of filter bank 216
(except filter element f.sub.0) are similarly processed, with the result
that all the filter element output signals are converted to baseband
signals with a bandwidth corresponding to that of the filter element. As
mentioned, the bandwidth is small, on the order of a few Hertz or less.
FIG. 6 is a simplified block diagram of a signal processor 428 which may be
used in FIG. 5. For definiteness, FIG. 6 represents pulse compression and
range sidelobe suppressor processor 428c of FIG. 5. In FIG. 6, processor
428c includes a cascade of two FIR filters 630, 660. Downconverted signals
from multiplier 420c of FIG. 5 are applied to the input port 640 of a
delay line (analog) or shift register (digital) 642, which allows the
signals to propagate to the right. A set of taps 644a, 644b . . . 644n
spaced by .tau..sub.0, the range sample interval, samples the propagating
signal and applies the samples to a set of multipliers 646 which weight
the samples. A combinatorial summing (.SIGMA.) circuit 650 sums the
weighted signal samples to produce an intermediate filtered signal on a
data path 652. The intermediate filtered signal is applied by way of data
path 652 to a second FIR filter 660, which is structurally similar to
filter 630, but may have different delay, number of taps and tap weights.
Filter 660 produces a pulse compressed, range sidelobe suppressed signal
on a data path 662c for application to corresponding magnitude detector
218c of FIG. 5. Since filters 630 and 660 of FIG. 6 are linear, they may
be cascaded in either sequence: filter 630 may provide pulse compression
and filter 660 may provide sidelobe reduction, or vice versa. Also, as is
well known in the art, the functions of filters 630 and 660 may be
combined into a single filter. The salient requirement of the
abovementioned Urkowitz application is that the range sidelobe reduction
function be provided individually for the signal component in each
frequency bin. When this requirement is met, the range sidelobe
suppression can be optimized for each frequency increment, and the
suppression can be maintained.
The general scheme of matched filtering (i.e., pulse compression) and range
sidelobe suppression is described in conjunction with FIG. 5. The
combination of pulse compressor and range sidelobe suppression follows
each of the complex multipliers. Since each complex multiplication removes
the residual Doppler phase shift across the uncompressed pulse, no
residual Doppler phase shift remains on the uncompressed pulse. Each pulse
compressor and range sidelobe suppressor is a zero Doppler design. All of
the pulse compressor and range sidelobe suppressors are therefore
identical in the arrangement described in the aforementioned Urkowitz
application.
The above discussion is general in the sense that the range sidelobes need
not have shapes and structure that are related to the shape of the main
lobe. However, as mentioned earlier, in some instances the sidelobes are
displaced and reduced versions of the main lobe. This is particularly true
in the case of polyphase sequences and binary phase sequences in which the
dwell at each phase defines a subpulse or "chip" interval. In such cases,
the sidelobe suppression filter taps and some of the pulse compression
filter taps need not be as densely spaced as the range sampling period
.tau..sub.0. The tap spacing need only be equal to the chip or subpulse
duration .tau. when the sidelobes are displaced, reduced-amplitude
versions of the main lobe. Signals having this property consist of
subpulses or chips, each of which is a simple single-frequency subpulse.
The subpulses are distinguished from one another by their phase, which
changes according to a phase sequence pattern or law. For such waveforms,
the matched filters may take on the form illustrated in FIG. 7. In FIG. 7,
elements corresponding to those of FIG. 6 are designated by the same
reference numerals.
In FIG. 7, pulse compression filter 630 is seen to consist of the cascade
of two separate transversal filter portions 630a and 630b. Filter portion
630a is matched to the form of the subpulse, and filter portion 630b is
matched to the subpulse-to-subpulse pattern or phase sequence of the set
of subpulses. The spacing between taps on subpulse-matched filter portion
630a is the range sampling period or interval .tau..sub.0. The spacing
between taps on pattern-matched filter portion 630b is the subpulse
spacing .tau., which is larger than the range sample spacing .tau..sub.0.
Range sidelobe suppression filter 660 of FIG. 7 also has its tap spacing
equal to the subpulse spacing .tau..
In FIG. 7, the number of taps associated with subpulse-matched filter
portion 630a is N.sub.2, i.e. N.sub.2 -1 plus the tap numbered zero, and
those taps are spaced in time by range sampling interval .tau..sub.0.
Similarly, pattern-matched filter portion 630b has N.sub.3 taps separated
by subpulse spacing .tau., where .tau. is an integer multiple of
.tau..sub.0. Range sidelobe suppression filter 660 of FIG. 7 has M.sub.2
taps, also spaced .tau..
The output of pulse compression filter 630 on data path 652 is the time
sampled version of the signal time autocorrelation function. That is, the
signal component is the time sampled version of the compressed pulse. This
signal component is the input to range sidelobe suppression filter 660.
The weights or weighting functions associated with pattern matched filter
630b are the conjugate time reverses of the pattern of cos .theta..sub.n,
where .theta..sub.n is the pattern of phase changes in the transmitted
waveform.
As described above, range sidelobe suppression filter 660 has taps that are
separated by a subpulse duration when the transmitted signal waveform is a
binary phase or polyphase sequence. Particular classes of binary phase
sequence are the Barker sequences and the pseudorandom sequences. The
pseudorandom sequences permit much freedom in making a choice of sequence
length, while it is frequently necessary to concatenate Barker sequences
to get long sequence lengths. Barker sequences are restricted to lengths
2, 3, 4, 5, 7, 11 and 13. To get, for example, a sequence of length 65,
one could concatenate 5 sequences of length 13 arranged in a particular
pattern. Although the sidelobe structure is not as simple as that of a
single Barker sequence, better suppression of sidelobes is believed to be
obtainable with concatenated Barker sequences than with other forms of
binary phase sequences, such as pseudorandom sequences of similar length,
when processing is performed as described above.
As shown in FIG. 7, and described above, the pulse compression for biphase
and polyphase sequences may be considered as the cascade of a filter
matched to a single subpulse and a filter matched to the pattern of phase
changes. In most circumstances, the Doppler phase shift across a subpulse
is very small and is negligible. In such circumstances, the subpulse
matched filter may be placed before the Doppler filter bank as illustrated
in FIG. 8. Elements of FIG. 8 corresponding to those of FIG. 3 are
designated by the reference numerals. In FIG. 8, a subpulse-matched filter
corresponding to filter 630a of FIG. 7 receives I+jQ signals from port
210. The subpulse-filtered signals are applied to Doppler filter bank 216
for separation into narrow frequency bands. The filtered signals from
filter element f.sub.0 are at baseband, and they are applied over a data
path 301 to a pattern-matched filter 830a, corresponding to filter 630b of
FIG. 7. The filtered signals produced at the outputs of filter elements
f.sub.1 . . . f.sub.M-1 are each applied to a mixer 320 for conversion to
baseband, as described above, and the resulting baseband signals are each
applied to a further pattern-matched filter 830. For example, the output
of filter element f.sub.2 is converted to baseband by a multiplier 320b,
and the resulting baseband signal is applied to pattern-matched filter
830b. The signals filtered by pattern-matched filters 830a-830m are then
applied to further filters 860a-860m, respectively, for range sidelobe
filtering. The resulting signals are individually applied to corresponding
detectors 218 for detection. The detected signals, representing the energy
found in each Doppler filter band, are coupled for thresholding or other
further processing.
Thus, one subpulse matched filter will serve for all Doppler frequency
shifts. Since only one subpulse matched filter is needed, it may be placed
anywhere before the Doppler filter bank, including in the analog portion
of the receiver, as an analog filter. As an analog filter, it may take
many forms, including that of an surface acoustic wave (SAW) device.
The arrangements described above have relatively complex filtering in each
Doppler channel following the Doppler filter. It is possible to simplify
the filtering andor improve the range sidelobe rejection according to the
invention by selection of the transmitted pulse phase sequence to include
complementary phase sequences, together with provision of matched
filtering for each sequence, followed by summing of the two
matched-filtered sequences. This is effective because, in short, the
selection of complementary-phase-sequence pairs causes the range sidelobes
of the two filtered sequences to be of mutually opposite amplitude or
polarity so that, when summed, the range sidelobes cancel while the main
range lobes add. This eliminates the need for separate range sidelobe
suppression filters.
In order to perform the invention, transmitter controller 3 of FIG. 1 must
cause each transmitted pulse (each sequence of phase-modulated subpulses
or chips) to be matched or accompanied by a corresponding transmitted
pulse in which the phase sequence of the subpulses is complementary to the
first phase sequence. For this purpose, the term complementary means that
the sum of the time autocorrelation functions of the two pulse sets or
sequences ideally has no sidelobes outside of the main lobe. Waveform 1000
of FIG. 10a represents a pulse formed from four subpulses or chips 1001,
1002, 1003 and 1004, having amplitudes of 1, -1, 1, 1, respectively, which
may be viewed as unit vectors with relative phases of 0, .tau., 0, 0,
respectively. FIGS. 10a-10i (where the hyphen represents the word
"through") represent steps in forming an autocorrelation function, and
FIGS. 10j-10s represent the result of the autocorrelation. As is well
understood by those skilled in the art, autocorrelation "scans" the time
function across a corresponding time function "moving" in the negative
time direction, multiplying together the "overlapping" portions and
summing the products. For example, an autocorrelation is performed on
waveform 1000 of FIG. 10a by allowing it to stand still (or move to the
right), while causing a similar waveform 1000', including subpulses 1001',
1002', 1003' and 1004' to move to the left, as indicated by the direction
arrows in FIG. 10a. In FIG. 10a, waveforms 1000 and 1000' do not overlap,
so their product is zero, and no output signal is produced, as illustrated
in FIG. 10j. While the amplitudes of the positive and negative excursions
of both pulses 1000 and 1000' are unity, pulse 1000' is illustrated as
slightly larger than pulse 1000 to allow them to be visually
distinguished. In FIG. 10b, corresponding to time interval 0-1 (where one
time interval corresponds to the duration of a subpulse or chip),
subpulses 1004 and 1001' overlap, both are positive so their product is
positive, and the overlap region is increasing in area, so the resulting
autocorrelation 1010 is a positive-going ramp increasing from zero
amplitude, as illustrated between times 0 and 1 in FIG. 10k.
At the end of time interval 0 to 1, the overlap of subpulses 1004 and 1001'
is complete, and ramp 1010 of FIG. 10k reaches a maximum value of 1.
Immediately thereafter, negative subpulse 1002' begins to overlap positive
subpulse 1004, while positive subpulse 1001' moves to the left, to overlap
portions of subpulse 1003, as illustrated in FIG. 10c. The product of
subpulse 1001' multiplied by portions of subpulses 1004 and 1003 remains
constant in the time interval 1-2, while the product of negative subpulse
1002' multiplied by portions of positive subpulse 1004 increases in
magnitude, with a negative | | |
