SIGNAL PROCESSING APPARATUS AND METHOD
This invention relates to a signal processing apparatus and method. More particularly, but not exclusively, it relates to an apparatus and method for processing pulse coded signals.
The useful range of a radar system is determined in part by the length of a radiated pulse (T) with the resolution of the system being determined by the bandwidth of the radiated pulse (B).
Traditional non-pulse coded radar systems emit pulses (100) having a duration (t) that are short when compared to an inter-pulse interval (102) (f), see for example Figure 1. The utility of short pulses in range determination is limited by the fact that for these non-pulse coded systems B*T=1. Thus any increase in bandwidth in order to increase accuracy of range determination has a corresponding detrimental effect upon the useful range of the system.
Referring now to Figure 2, a time compensated adaptive beamformer apparatus (200), of the type used in non-pulse coded radar systems, comprises a plurality of input signals (202) which are sampled and fed to a multiplier (204), where they are multiplied by a signal f(k) derived from modulation signal generator 205. The signal f(k) is a function of time, and will typically be a linear signal which compensates for relative motion between a receiver array and a target during acquisition of input signals. This produces motion compensated output signals (206). This is detailed in the applicants UK Patent Application No. GB9814237.5'(Publication No. GB2339078A), the contents of which are incorporated herein by reference. Samples of both the input signals (202) and the compensated signals (206) are passed to a discrimination unit (208) which typically prevents those samples containing strong target
returns from affecting subsequent beamforming operations. Indeed, the effect of this discrimination unit is to apply a binary scale factor to the samples used in subsequent beamforming operations. It is also feasible to use a 'soft' sample selection scheme as a discriminator in which all samples are scaled according to the probability of their containing a non-homogeneity.
Discriminated signals (210) are passed to a weight computing unit (212) where time varying complex weighting coefficients are calculated for each input channel. These coefficients are applied to each of the input channels by a weighting unit (214) to steer a suppression null so as to maintain the null in the direction of an observed interference signal, such as a jamming signal.
This scheme is applicable when target returns have a limited temporal extent, which provides a discriminant between targets and noise sources, such as jamming signals.
Pulse coded radar, pulse compression, systems increase the bandwidth (B) of an output pulse by manipulating the phase or frequency within a pulse of duration t. Early pulse coded radar systems removed the constraint of the direct interdependence of the pulse duration with bandwidth by compressing the pulse of duration t into a narrow signal of duration 1/B using a pulse compression filter, typically a linear finite impulse response filter, matched to a transmitted, compressed, pulsed waveform.
However, the increased use of pulse coded, pulse compression, radar systems in which a coded pulse (300) has a duration (t) that is a significant fraction of the inter-pulse interval (302)(f), typically up to 33% to 50% of the pulse duration, present particular problems,
especially in the fields of jamming immunity and strong target discrimination, for scanning radars.
An adaptive beamforming apparatus (400) including pulse compression filtering is shown in Figure 4 and is similar to the adaptive beamforming apparatus (200) described hereinbefore with reference to Figure 2 and similar parts are accorded similar reference numerals in the four hundred series. The beamforming apparatus (400) differs from that of Figure 2 in that a pulse compression filter (401 ) acts upon input signals (402) prior to their being sampled and the samples being passed to motion compensation unit (404).
This arrangement has the effect of adding interfering signals arriving from multiple directions to form each sample and thus smearing out the null across a broad angular range. For example, consider a radar rotating at 1 revolution per second, and using a pulse of length 40μs, the effective width of the null would be around 0.015°. This represents a significant fraction of the 3.5° beamwidth of a typical radar.
In the case of a pseudo-random noise like interfering signal, a broadband thermal noise like signal is emitted by an interfering transmitter. This has the effect of raising the background level at a radar system's receiver antennas, reducing the target signal to noise ratio and effectively masking low amplitude returns from, for example, a target having a small radar cross-section.
In scanning pulse coded radar systems having a short pulse duration, typically 1 % of the inter-pulse period, receiver movement has the effect of smearing out a sharp, deep null (500), typically 0.004° wide at the -60dB level and 70dB attenuation to form a broad,
shallow null (502), typically 0.015° and having a mean attenuation of around 60dB, see for example Figure 5. This has been solved by generation of rapidly time-varying beamforming weights, weighting vectors, that are applied to an adaptive beamformer to effectively steer a sharp localised null corresponding to the interfering transmitter's location with the movement of the radar's receiver antenna over the short duration of the pulse.
However, in the case of pulse coded radar systems employing long pulse duration where the direction of arrival of an interfering signal is changing with time, for example, due to rapid motion of a platform upon which the receiving antenna is mounted, for example, a ship or rapidly moving airborne vehicle, this is not the case. In this instance, the effect of a long pulse compression filter is to add interfering signals arriving from each of the directions perceived by the array during it's motion to form each sample. In an adaptive beamforming architecture subsequent time-varying beamformer operations cannot remove this effect and interfering signal cancellation performance suffers. This is due to the null being very narrow, leading to poor cancellation as the interfering signal appears spread because of the finite time over which the spatial statistics of the location of the interfering transmitter are captured. Cancellation of a distributed source is less effective than a point source, because the formation of a deep wide null causes an increase in the background thermal noise. The adaptive beamformer compromises by reducing the width of the null in order to trade off thermal noise against jamming signal residual.
Also, a strong radar return in a given direction acts to reduce the sensitivity of a radar system to a weaker return. This is due to the relative scaling of target signals due to the high intensity of the strong radar return causing the weak return to merge with the
background noise level. Clearly, this problem is exacerbated where there is an interfering signal that causes the background noise level to be increased. This is because a strong target return will cause the adaptive beamformer to partially cancel the target signal as well as any jamming. Although undesirable this is not critical in respect of detecting the strong target return, which will still be above the noise level. However, in the case of a small target return being present the distortion to the beam pattern caused by the strong target return can cause the weak target return to be undetectable.
According to a first aspect of the present invention there is provided a signal processing apparatus, suitable for use with an adaptive beamforming array, comprising compensation means arranged to compensate a portion of each of a plurality of input signals received at a receiver array for variations therein due to relative motion between a target and the receiver array to produce compensated signals, characterised in that the apparatus further includes filter means arranged to receive a portion of each of the signals indicative of the compensated signals, and at least a portion of each of the input signals and effect thereupon to produce filtered signals.
Such an arrangement has the advantage that as filtering takes place after compensation for the time varying nature of received signals, in a long pulse duration pulse coded system, the addition of apparently spatially varying contributions of an interfering signal are removed as they are compensated for during the compensation process.
It will be appreciated that in radar applications the input signals prior to filtering by the filter means are usually termed expanded, or uncompressed, signals, or pulses, and the filtered signals are typically termed compressed signals, or pulses.
It will be appreciated that the signals indicative of the compensated signals may be the compensated signals themselves directly after motion compensation or may be signals that have undergone processing subsequent to motion compensation.
The filter means may be a finite impulse response (FIR) filter which may have an impulse response duration of at least the same duration as the signals indicative of the compensated signals, and the input signals. The use of an FIR filter with a long impulse response duration compared to the duration of the input signals enhances, any subsequent discrimination of strong signal components. Typically, an FIR filter with an impulse response duration less than the duration of the input signals will not be as effective as an FIR filter with a long impulse response duration compared to the duration for any subsequent discrimination of strong signal components, but will still have utility.
The filter means may comprise a pulse compression filter. The apparatus may comprise discrimination means arranged to receive signals indicative of filtered signals output from the filter means and produce discriminated signals therefrom.
The discrimination means may be an analogue filter, or a digital signal processor arranged to digitally process the filtered signals. The discrimination means may be arranged to threshold the filtered signals at a pre-determined level, remove portion of the filtered signals or attenuate a portion of the filtered signals.
The use of sample selection, discrimination, following pulse compression allows strong returns to be discriminated out from the received signals, for example by thresholding, and thus allows weak
returns to be observed more easily by reducing the scaling effect of the strong returns.
The apparatus may comprise computational means arranged to receive the discriminated signals and compute time varying beamforming weights therefrom and beamforming means arranged to apply the time varying beamforming weights to at least a portion of the input signals and/or the compensated signals.
The beamforming means may be arranged to output a weighting signal that applies the time varying beamforming weights to the input signals prior to the respective portions of the input signals passing to the compensation means.
Thus, a phased array can be steered using the compensated and filtered signals to maintain a narrow, deep, typically 50-60dB, null in the direction of an interfering transmitter even when the platform upon which it is mounted is rapidly moving.
The compensation means may be a signal processor that may be arranged to modulate the amplitude and/or phase of the portion of each of the input signals in a linear, sinusoidal, quadratic or other non-linear manner. This is as described in relation to modulation function f(k) on page 3 et seq, and in the description relating to the modulator 32 and multiplier 134 in Figures 1 , 2 and 4, of the published specification GB2339078A, the contents of which are hereby included by reference.
The filter means may be a linear finite response filter that may be matched to a transmitted pulse waveform, which is typically filtered. The filter means may be arranged to apply an amplitude taper to the portion of each of signals indicative of the compensated signals, and
at least the portion of each of the input signals. This controls pulse compression sidelobes.
The apparatus may be arranged to null an interfering signal by between 50-6OdB. The apparatus may further be arranged to null an interfering signal by up to, and possibly in excess of 70dB.
According to a second aspect of the present invention there is provided a method of signal processing for use with an adaptive beamformer comprising the steps of: i) sampling input signals received at a receiver array to produce sampled signals; ii) compensating the sampled signals for variations therein due to relative motion between a target and the receiver array to produce compensated signals; iii) applying a filtering operation technique to both the input signals and signals indicative of the compensated signals to produce filtered signals; and iv) applying the filtered signals to a beamformer to produce a beam.
The method may comprise discriminating filtered signals to produce discriminated signals therefrom.
The method may comprise thresholding the filtered signals at a predetermined level, removing a portion of the filtered signals or attenuating a portion of the filtered signal.
The method may comprise computing time varying beamforming weights from the discriminated signals and applying the time varying
beamforming weights to at least the sampled signals or the input signals and/or the compensated signals.
Compensating the sampled signals may comprise modulating the amplitude and/or the phase of the portion of each of the input signals in a linear, sinusoidal, quadratic or other non-linear manner.
According to a third aspect of the present invention there is provided an adaptive beamformer comprising signal processing apparatus according to the first aspect of the present invention.
According to a fourth aspect of the present invention there is provided a radar system comprising signal processing apparatus according to the first aspect of the present invention and/or an adaptive beamformer according to the third aspect of the present invention.
According to a fifth aspect of the present invention there is provided a wireless network receiver comprising signal processing apparatus according to the first aspect of the present invention.
The invention will now be described, by way of example only, with reference to the accompanying drawings, in which:
Figure 1 is a schematic representation of a series of short duration radar pulses of the prior art;
Figure 2 is a schematic representation of an adaptive beamforming apparatus of the prior art;
Figure 3 is a schematic representation of a series of long duration radar pulses of the prior art;
Figure 4 is a schematic representation of an adaptive beamforming apparatus of the prior art, incorporating pulse compression;
Figure 5 is a schematic representation of a profile of an ideal interference suppression null and an interference suppression null of the prior art;
Figure 6 is a plot of a simulation of a quiescent beam pattern
(gain (dB) vs sin θ) of a linear array comprising sixteen half- wavelength spaced array elements, with five hundred unambiguous range cells, a linear frequency modulation (LFM) pulse compression waveform three hundred samples long, and thermal noise power at each element set to unity prior to pulse compression;
Figure 7 is a plot of a simulation of a range profile (amplitude of return (arb. Units) vs range cell number) of the array simulated in Figure 6 with a target of power -20dB in range cell 400, broadside the array;
Figure 8 is a plot of a simulation of a beam pattern of the arrangement simulated in Figure 7 with a stationary 40dB jamming signal added;
Figure 9 is a plot of a simulation of a beam pattern of the arrangement simulated in Figure 7 with a moving 40dB jamming signal added and no motion compensation;
Figure 10 is a plot of a simulation of a beam pattern of the arrangement simulated in Figure 8 with a moving strong (9.5 dB)
target in range cell 250, slightly off broadside the array, with motion compensation after pulse compression;
Figure 11 is a plot of a simulation of a beam pattern of the arrangement simulated in Figure 8 with a stationary strong (9.5 dB) target in range cell 250, slightly off broadside the array, with pulse compression after beamforming;
Figure 12 is a plot of a simulation of a range profile of the arrangement simulated in Figure 11 ;
Figure 13 is a schematic representation of a first embodiment of an adaptive beamforming apparatus according to an aspect of the present invention;
Figure 14 is a schematic representation of a second embodiment of an adaptive beamforming apparatus according to an aspect of the present invention;
Figure 15 is a plot showing a simulated beam pattern of a linear array employing a signal processing apparatus according to an aspect of the present invention, the linear array comprising sixteen half-wavelength spaced array elements, with five hundred unambiguous range cells, a linear frequency modulation (LFM) pulse compression waveform three hundred samples long, thermal noise power at each element set to unity prior to pulse compression, with a moving 40dB jamming signal added and a moving strong (9.5 dB) target in range cell 250, slightly off broadside the array; and
Figure 16 is a flow diagram showing a method of signal processing in accordance with at least an aspect of the present invention.
Referring now to Figure 6, a linear array comprising sixteen half- wavelength spaced array elements, with five hundred unambiguous range cells, a linear frequency modulation (LFM) pulse compression waveform three hundred samples long, and thermal noise power at each element set to unity prior to pulse compression is simulated and a quiescent beam pattern (600) of the array is plotted. A typical thermal noise power in a practical array is about -140dBW. The maximum sensitivity, gain, (602) of the array is broadside the array with a number of sidelobes (604) having a gain of approximately - 30dB.
Referring now to Figure 7, the array modelled in respect of Figure 6 is modelled with a low strength target (700), having a -20dB return, relative to thermal noise at an array element, present in range cell 400, broadside the array. The target (700) can clearly be seen above the noise (702) present in the simulated array range profile.
Referring now to Figure 8, a beam pattern (800) of the array simulated in Figure 7 is plotted with an additional stationary 40dB jamming signal (802) added to the simulation. In this simulation the target (804) remains broadside and adaptive beamforming weights are computed. As can be seen the target signal to noise ratio (SNR) has fallen to 12.9 dB due to the presence of the pseudo-random thermal noise like jamming signal. Additionally, the adaptive beamforming introduces a sharp, typically 0.004° at the -60dB level, with better than 70dB attenuation, null in the direction from which the jamming signal originates.
Referring to Figure 9 a beam pattern (900) of the array simulated in Figure 7 with an additional moving 40dB jamming signal (902) simulated with no motion compensation. The jamming signal has a
direction of arrival that is swept through 0.1 beamwidths, a typical beamwidth being between 6°-30°, during the acquisition of five hundred data samples. The target (904) remains broadside the array in this simulation. This results in a poor target SNR of 2.9dB after pulse compression following beamforming. A broad null (906), greater than 0.9° wide, having on average less than -50dB attenuation, is observed in the beam pattern. This being broader than the null (806) shown in Figure 8 as the input signal has no motion compensation and therefore the pulse compression spreads information relating to the location of the jamming signal throughout the pulse. This is also the case in the absence of pulse compression, due to the effective motion of the jammer during the finite interval used to compute the beamforming weights.
Referring now to Figure 10, a simulated beam pattern (1000) of the array simulated in Figure 7, including a jamming signal (1002), as detailed in Figure 9 and a target (1004), and motion compensation following pulse compression shows a poor target SNR of 2.9dB, using the architecture shown in Figure 4. Additionally, a null (1006) is broadened compared to that of Figure 8. This is because the pulse compression filter spreads the effective jamming signal direction of arrival in a manner that is irrecoverable by motion compensation following the pulse compression.
Referring now to Figures 11 and 12, the effect of introducing a strong target (1 100) in range cell 250 into the arrangement simulated in Figure 7 are shown. As can be seen from the Figures, the weak target (1 102), broadside in range cell 400, has a very poor SNR of 2.5dB. Only the strong target (1 100) can be reliably detected in a simulated range profile (1 104). The corresponding beam pattern (1101 ) is shown in Figure 11.
Referring now to Figure 13, an adaptive beamforming apparatus (1300) comprises a plurality of input channels (1302), each carrying a received signal from a respective receiver element of a phased array antenna (not shown), a motion compensation unit comprising a multiplier (1303) driven by a modulation signal generator (1305), which is arranged to account for relative motion between a target and the antenna, a computational processor (1304) for calculating adaptive time varying beamforming weights for each input channel and a beamformer (1306) for applying the time varying beamforming weights to the input channels, a pulse compression unit (1308), typically operating using linear frequency modulation (LFM), and a discriminator unit (1310) which is arranged to threshold or otherwise remove undesired elements of the input signals.
Each received signal is sampled and a sample signal passed to the motion compensation unit (1303) where a compensatory algorithm is applied to the signal in order to compensate for motion of the antenna during the data acquisition period. This algorithm typically involves the application of a linearly varying function to the received signal to compensate for the temporal variation of the apparent received direction of a target due to relative motion between the target and the antenna during data acquisition.
The motion compensated samples and the received signal pass to the beamformer (1306) where previously calculated time varying beamforming weights are applied to both the samples and the received signal. These weighted signals now pass to the pulse compression unit (1308) where a finite response filter, typically an LFM dispersive filter, a non-linear frequency modulation (NLFM) filter or pseudo-random noise modulation (PN) filter, matched to an output waveform of a pulse coded transmitter, usually output from the phased array antenna, compresses the pulse. This allows range
information to be recovered from within the pulse coded return. The output of the pulse compression unit (1308) is split into two, with the output being passed both to the discriminator (1310) in a feedback path where the weights used in the beamformer (1306) are calculated, and to the next stage in the receiver processing where it may be processed in a standard manner.
The discriminator (1310) may remove strong returns from the range profile of the received pulse. I his improves the SNR of weak targets lying within the mainbeam of the phased array antenna by removing the effective scaling factor of the strong return. The removal of a strong target return from the input to the adaptive beamformer removes partial cancellation of a weak target signal, as well as any jamming, by the adaptive beamformer. Whilst not critical in terms of detecting the strong target return, which will still be well above the noise level, distortion to the beam pattern caused by the strong target return causes the weak target return to be undetectable.
The output of the beamformer (1306) may be used for display, for example on a cathode ray tube, or further signal processing, for example clutter filtering by extending the adaptive processing interval and cascading clutter and pulse compression filters.
A portion of this discriminated signal is sampled and used by the computational processor (1304) to calculate time varying beamforming weights to be applied to the next incoming pulse coded input signal by the beamformer (1306). The beamforming weights are typically calculated using gradient descent methods or by directly solving the least-squares problem using QR decomposition, for example.
It should be noted that as both the pulse compression process and the discrimination, (sample selection), process are linear processes and the beamforming weights are fixed for each adaptive processing interval the apparatus of Figure 13 can be rearranged to produce an adaptive beamforming apparatus (1400) of Figure 14 and similar parts will be accorded the same reference numerals in the fourteen hundred series.
The ordering of the components in Figure 14 differs from that of Figure 13 in that the adaptive beamforming apparatus (1400) comprises a plurality of input channels (1402) from which samples of input signals from a receiver array are taken and fed to a motion compensation unit comprising multiplier (1403) driven by modulation signal generator (1405). Both the input signals and sampled, motion compensated signals then undergo pulse compression in pulse compression unit (1408). The resulting compressed pulses are then fed to a discriminator unit (1410) where they are thresholded, selectively attenuated, selectively weighted or have portions selectively removed therefrom. Discriminated signals are then passed to a computational processor (1404) where they are used to calculate time varying beamforming weights. These time-varying beamforming weights are passed to the beamformer (1406) where they are applied to data from the incoming pulse that has been suitably delayed by a delay unit (1412).
In summary, the apparatus (1400) receives a block of data X(k), k=1 to n, comprising n samples of data each measured at m antennas, which enters the apparatus via channels and n samples of beamformed signals are output from the beamformer.
In the apparatus' of Figures 13 and 14 the pulse compression occurs after motion compensation. This has the effect of removing the
widening and reduction in depth of a null due to the constraint of carrying out the movement compensation upon a pulse compressed input signal because of the spreading of the spatial information relating to the location of an interfering signal within the compressed pulse.
Referring now to Figure 15, a simulated beam pattern (1500) of a linear array employing a signal processing apparatus according to an aspect of the present invention, the linear array comprising sixteen half-wavelength spaced array elements, with five hundred unambiguous range cells, a linear frequency modulation (LFM) pulse compression waveform 300 samples long, thermal noise power at each element set to unity prior to pulse compression, with a moving 40dB jamming signal added and a moving strong (9.5 dB) target in range cell 250, slightly off broadside the array, has a narrow, sharp null (1502) 0.004° wide at the -60dB level, and in excess of 70dB deep, and provides a weak, 20dB, target (1504) having a SNR 1 1 .4dB better than that of Figure 9. This simulation was carried out using an adaptive beamforming architecture as described in relation to Figure 14.
Referring now to Figure 16, a method of signal processing for an adaptive beamformer includes sampling inputs signals to produce sampled signals (Step 1600), compensating the sampled signals therein to produce compensated signals (Step 1602) and applying a pulse compression technique to both the input signals and the compensated signals to produce compressed signals (Step 1604).
Whilst described with reference to radar systems it is envisaged that the beamformer apparatus described hereinbefore can be applied to wireless networks, in particular to spread spectrum communications networks. For example, a network employing code division multiple
access (CDMA) long spreading codes in a non-stationary environment, for example a user moving whilst using a personal digital assistant (PDA), can work in a similar manner to pulse compression techniques in radar.