Dual-Polarized FM Noise Radar

Transcription

1 Dual-Polarized FM Noise Radar Garrett Zook*, Patrick M. McCorick*, Shannon D. Blunt*, Chris Allen*, John Jakabosky* *Radar Systes Lab, University of Kansas, Lawrence, KS, USA Radar Division, Naval Research Laboratory, Washington, DC, USA Keywords: Polarietric radar, FM noise radar Abstract A technique denoted as pseudo-rando optiized (PRO) FM has recently been experientally deonstrated to facilitate a for of FM noise radar eission involving a unique, transitter-copatible wavefor for each pulse. Because the range sidelobes change each pulse, they are effectively suppressed when perforing coherent integration (i.e. Doppler processing) over ultiple pulses. Here, the PRO-FM schee is incorporated into a dual-polarized arrangeent, with independent wavefors eitted fro each orthogonally polarized channel, as a eans to suppress the crosscorrelation effects that otherwise arise fro siultaneous transission on both channels. Experiental easureents deonstrate the efficacy of this high-diensional eission. 1 Introduction Rando noise radar, where the wavefor possesses randoly generated aplitude and phase to appear noise-like (e.g. [1-3]), has been well studied in the literature. The notion of FM noise radar, in which the wavefor is constant aplitude with rando frequency odulation, is less well-known and has previously been liited to analytical study [4]. Traditional noise radar is known to be inherently low probability of intercept (LPI) while tending to be liited to low-power, short-range applications due to noise-like aplitude odulation. In contrast, FM noise radar is not LPI but the constant-aplitude, spectrally well-contained nature of FM is consistent with the use of a saturated power aplifier, thus enabling high-power, long-range applications. A physical realization of FM noise radar was recently deonstrated [5,6]. Denoted as pseudo-rando optiized (PRO) FM, this schee generates pulses (or CW segents) with wavefors that are initially rando FM signals and subsequently shaped spectrally via an alternating projection process to possess a power spectral density (PSD) conforing to a desired teplate (to the degree possible). The Gaussian PSD is an attractive choice because it corresponds to an associated wavefor autocorrelation that is likewise Gaussian and thus yields low range sidelobes relative to the tiebandwidth product [7]. When coherently cobined over a coherent processing interval (CPI) of such pulses (or CW segents) the varying sidelobe structures over the CPI do not cobine coherently, thereby providing further sidelobe suppression. The tie-varying, high-diensional nature of this physical eission schee has also recently been shown to 1) facilitate in-band spectral notches for interference avoidance while itigating uch of the degradation that otherwise occurs and 2) enable a for of tande-hopped radar/counication for spectru sharing [8-11]. Here the high diensionality of the FM noise radar structure is exploited to generate dual-polarized pulsed eissions of sufficient separability on receive (to within the liit of the antenna polarization isolation) that the orthogonally polarized channels can transit siultaneously. It is deonstrated using both siulation and experiental easureents that coherent integration over a CPI of these dual-polarized wavefors realizes a low enough cross-correlation to adequately separate the co-polarized and cross-polarized receive coponents. 2 PRO-FM Each PRO-FM pulse is first initialized with a rando FM wavefor. This initialization is accoplished by generating an N-length code of rando saples drawn fro a unifor distribution on [, ]. Representing noralized instantaneous frequency values, these N saples are used to produce an initial rando FM wavefor via the polyphasecoded FM (PCFM) ipleentation [12]. For s0, () t the initial rando FM wavefor for the th pulse, the desired PSD 2 G( f ), and a desired pulse aplitude shape ut (), the optiization is perfored by alternating between the application of [5,6] and 1 k1, k, r ( t) F G( f ) exp j F s ( t) (1) k1, k1, s ( t) u( t)exp j r ( t) (2) for K iterations. The operations F and F are the Fourier and inverse Fourier transfors, and () extracts the phase of the arguent. We shall use a Gaussian PSD and a pulse aplitude shape that is a constant over the pulsewidth T (on tie interval [0,T]) and zero elsewhere. Note that since the optiization in (1) and (2) requires use of discretized versions of G( f ), u(t), rk,(t), and pk,(t), it is necessary to oversaple these with respect to 3-dB bandwidth to ensure sufficient fidelity (i.e. to account for enough of the spectral roll-off region). If aplitude odulation (AM) is peritted, 1 1

2 this design approach has also been shown to facilitate a hybrid FM wavefor that achieves ultra-low sidelobes [13]. The set of M wavefors designed in this anner possess unique sidelobe structures. When M is relatively sall a range sidelobe odulation effect is known to occur [14,15] that can hinder clutter cancellation perforance. However, as M increases the cobination of the non-coherent sidelobes produces an incoherent averaging effect that yields reduced sidelobes for the CPI as a whole [6]. Appropriate design of isatched filters on a per-pulse basis can also iprove overall sidelobe suppression [6,16,17], though this option does not extend to dual-polarized operation due to the presence of the cross-polarized wavefor response. sidelobe levels (ah RMS and av RMS) reach a peak sidelobe level (PSL) of around 37 db. The sidelobe level is arkedly higher for the cross-correlation (ax RMS), which has a PSL of about 25 db. In other words, on average one can expect the cross-correlation to doinate the sidelobe response on a perpulse basis. 3 Dual-Polarized PRO-FM Traditional noise radar has previously been used in polarietric ipleentations [18-21] for low-power, shortrange applications. It was also recently shown [22] that a polarietric for of adaptive pulse copression (APC) can be applied on receive to separate the co-polarized and crosspolarized responses for arbitrary FM wavefors, though the coputational cost of adaptive processing can be soewhat high. Here the high-diensional benefits of noise radar are exploited via the PRO-FM construct to facilitate high-power, long-range operation of siultaneous dual-polarized eissions that, given a sufficiently high M to suppress the sidelobes, requires only atched filtering on receive. The extension of PRO-FM [5,6] to dual-polarized operation involves the generation of a second independent rando FM wavefor initialization and subsequent alternating projection optiization for each pulse. Note that the independent rando generation of each wavefor pair establishes an average cross-correlation that is inversely proportional to N (which approxiates the initial tie-bandwidth product BT, for B the 3-dB bandwidth). The subsequent spectral shaping to reduce autocorrelation sidelobes also serves to decrease B, and is thus expected to degrade (i.e. increase) the degree of cross-correlation between the pair of wavefors since no effort is ade to iniize cross-correlation aside fro relying on the rando initializations. Spectral shaping optiization that appropriately balances between autocorrelation and cross-correlation sidelobes reains a topic of ongoing investigation. As an exaple, consider the generation of dual-polarized PRO FM wavefors that have a 3-dB bandwidth of B = 200 MHz and a pulsewidth of T = 1 s. Denoting s,h(t) and s,v(t) as the = 1,, M wavefors to be eitted fro the horizontally and vertically polarized channels, respectively, the associated autocorrelations a,h() and a,v() and crosscorrelation a,x() can be coputed. For M = 5000 pulses, the root-ean-square (RMS) cobination over each set of auto/cross-correlations provides a sense of the average sidelobe level in a given auto/cross - correlation response. Figure 1 illustrates these RMS results, where it is found that, on average, the autocorrelation Fig. 1. RMS auto/cross-correlations for M = 5000 dualpolarized wavefor pairs (note ah RMS is virtually identical to av RMS) When the M = 5000 auto/cross-correlations are coherently integrated such as would occur when perforing Doppler processing, the responses in Figure 2 are obtained. Here it is observed that the autocorrelation responses (ah and av) yield PSL values of around 69 db and the cross-correlation (ax) PSL is roughly 57 db. These values are 32 db lower than the RMS values for a single pulse fro Fig. 1, which is 5 db short of the 37 db one would expect when cobining 5000 incoherent pulses. This discrepancy is likely due to the spectral shaping which serves to constrain the available degrees of freedo one would expect fro a spectrally white rando instantiation. Fig. 2. Coherently integrated auto/cross-correlations for M = 5000 dual-polarized wavefor pairs 2

3 If we alternatively consider M = 100 pulses, the auto/crosscorrelations in Fig. 3 are obtained. As expected for a factor of 50 reduction in the nuber of independent pulses, a roughly 17 db increase in the sidelobes is observed. In the next section we shall use this M = 5000 set of wavefor pairs, in cobination with pre-suing [23] as eployed in soe synthetic aperture radar (SAR) systes, to assess experientally the separation of the dual-polarized coponents. transit center frequency is 3.55 GHz. The wideband transit aplifiers and receive low-noise aplifiers (LNAs) have 27 db and 22 db gain, respectively. The horizontal and vertical wavefors for each pulse were produced by two channels of a Tektronix AWG70002 wavefor generator that has 10 bit resolution. The subsequent receive echoes were captured and digitized by a Tektronix DPO72304DX oscilloscope with 8 bit resolution. Fig. 3. Coherently integrated auto/cross-correlations for M = 100 dual-polarized wavefor pairs 4 Free-Space Experiental Measureents Using the dual-polarized wavefor pairs fro the previous section, free-space easureents were ade fro the roof of Nichols Hall on the University of Kansas capus. The target scene was the intersection of 23 rd and Iowa streets at a radial range interval between 1050 and 1250 and consisted of ultiple cars entering/leaving the intersection. Multiple capus buildings and treed areas were also within the field of view (see Figure 4). Fig. 4. Field of view for free-space easureents. The test setup is shown in Figure 5. The transit and receive chains use separate offset horn-fed dish antennas with 22.5 dbi gain. The antennas have cross-polarization rejection of at least 25 db and H/V port isolation of at least 20 db. The Fig. 5. Test setup for free-space easureents This test setup is used to eit the M = 5000 PRO FM wavefor pairs discussed in the previous section with a pulse repetition frequency (PRF) of 50 khz (5% duty cycle). Thus the CPI coprises 0.1 seconds. Because this PRF is far higher than is necessary to easure the expected velocities of ground vehicles, the effective PRF can be reduced by pre-suing [23] a nuber of pulses prior to Doppler processing (but after pulse copression) which serves to low-pass filter the data in the Doppler doain. While the pre-suing approach is well-known in SAR to reduce data handling/storage requireents [23], it has the additional benefit for these FM noise radar wavefors of providing greater diensionality for incoherent sidelobe cobining. For this case, the received echoes fro the 5000 dual-polarized wavefor pairs are first pulse copressed using their corresponding atched filters and then presued by 50. Therefore, the effective PRF of 1 khz, with an unabiguous velocity of /s. Doppler processing is then applied to the resulting 100 range profiles, thereby also reducing the coputational load of the processing (especially clutter cancellation). Figures 6-9 depict the HH, VV, and HV polarized responses fro the target scene after clutter cancellation, which was ipleented by siply projecting out the zero Doppler coponent. Multiple oving targets can be observed and the siultaneous illuination of both H and V channels enables direct coparison of the different responses without reducing the PRF to accoodate alternating illuination. It should 3

4 be noted that the vertically polarized receive channel was experiencing a sporadically faulty aplifier (hence the increased noise in the VV and HV responses) so these preliinary easureents will need to be repeated before fir conclusions can be drawn. That said, it is clear that usual inforation can be gleaned fro the co-pol and cross-pol channels. 5 Conclusions Leveraging the pseudo-rando optiized (PRO) FM noise radar eission, preliinary experiental easureents have deonstrated that dual-polarized FM noise radar can facilitate the eission of the horizontal and vertical channels siultaneous, thereby avoiding the need for alternating illuination. Such capability ay also enhance perforance for subsequent polarietric processing [24,25]. Pre-suing, known to reduce data handling requireents for SAR, is also found to further enhance the benefit of incoherent sidelobe cobining facilitated by the PRO FM schee. Noting that the polarization isolation is a liiting factor on how well dualpolarized separation can be achieved, further easureents with greater isolation are needed to deterine just how well FM noise radar can perfor relative to the very good separation suggested by siulation. Acknowledgeents Fig. 7. HH range-doppler response after clutter cancellation Fig. 8. VV range-doppler response after clutter cancellation Fig. 9. HV range-doppler response after clutter cancellation This work was sponsored in part by the US Office of Naval Research under Contract #N C-2029 and the US Naval Research Laboratory. References [1] B.M. Horton. Noise-odulated distance easuring systes, Proc. IRE, vol. 47, no. 5, pp , May [2] X. Xu, R.M. Narayanan. Range sidelobe suppression technique for coherent ultra wide-band rando noise radar iaging, IEEE Trans. Antennas & Propagation, vol. 49, no. 12, pp , Dec [3] M. Malanowski, K. Kulpa. Detection of oving targets with continuous-wave noise radar: theory and easureents, IEEE Trans. Geoscience & Reote Science, vol. 50, no. 9, pp , Sept [4] S.R.J. Axelsson. Noise radar using rando phase and frequency odulation, IEEE Trans. Geoscience & Reote Sensing, vol. 42, no. 11, pp , Nov [5] J. Jakabosky, S.D. Blunt, B. Hied. "Wavefor design and receive processing for nonrecurrent nonlinear FMCW radar," IEEE Intl. Radar Conf., Washington, DC, USA, May [6] J. Jakabosky, S.D. Blunt, B. Hied. "Spectral-shape optiized FM noise radar for pulse agility," IEEE Radar Conf., Philadelphia, PA, USA, May [7] J.A. Johnston and A.C. Fairhead, Wavefor design and Doppler sensitivity analysis for nonlinear FM chirp pulses, IEE Proc. Counications, Radar & Signal Processing, vol. 133, no. 2, pp , Apr [8] J. Jakabosky, S.D. Blunt, A. Martone. "Incorporating hopped spectral gaps into nonrecurrent nonlinear FMCW radar eissions," IEEE Intl. Workshop on Coputational Advances in Multi-Sensor Adaptive Processing, Cancun, Mexico, Dec [9] J. Jakabosky, B. Ravenscroft, S.D. Blunt, A. Martone. "Gapped spectru shaping for tande-hopped 4

5 radar/counications & cognitive sensing," IEEE Radar Conf., Philadelphia, PA, USA, May [10] B. Ravenscroft, P.M. McCorick, S.D. Blunt, J. Jakabosky, J.G. Metcalf. "Tande-hopped OFDM counications in spectral gaps of FM noise radar," IEEE Radar Conference, Seattle, WA, USA, May [11] B. Ravenscroft, S.D. Blunt, C. Allen, A. Martone. Analysis of spectral notching in FM noise radar using easured interference, IET Intl. Radar Conf., Belfast, UK, Oct [12] S.D. Blunt, M. Cook, J. Jakabosky, J. de Graaf, E. Perrins. "Polyphase-coded FM (PCFM) radar wavefors, part I: ipleentation," IEEE Trans. Aerospace & Electronic Systes, vol. 50, no. 3, pp , July [13] J. Jakabosky, S.D. Blunt, and T. Higgins. Ultra-low sidelobe wavefor design via spectral shaping and LINC transit architecture, IEEE Intl. Radar Conf., Washington, DC, USA, May [14] S. Blunt, M. Cook, J. Stiles. Ebedding inforation into radar eissions via wavefor ipleentation, Intl. Wavefor Diversity & Design Conf., Niagara Falls, Canada, Aug [15] C. Sahin, J. Metcalf, and S. Blunt. Matheatical characterization of range sidelobe odulation, IET Intl. Radar Conf., Belfast, UK, Oct [16] A. O Connor, J. Kantor, and J. Jakabosky. Joint equalization filters that itigate wavefor-diversity odulation of clutter, IEEE Radar Conf., Philadelphia, PA, USA, May [17] C. Sahin, J. Metcalf, and S. Blunt. Filter design to address range sidelobe odulation in transit-encoded radar-ebedded counications, IEEE Radar Conf., Seattle, WA, USA, May [18] Y. Xu, R.M. Narayanan, X. Xu, J.O. Curtis. Polarietric processing of coherent rando noise radar for buried object detection, IEEE Trans. Geoscience & Reote Sensing, vol. 39, no. 3, pp , Mar [19] R.M. Narayanan, C. Kuru. Ipleentation of fully polarietric rando noise radar, IEEE Antennas & Propagation Letters, vol. 4, no. 1, pp , June [20] L. Maslikowski, K. Kulpa, D. Glushko, F. Yanovsky. Atospheric precipitation sensing with a short-range C-band noise radar, Intl. Radar Syp., Dresden, Gerany, June [21] A. Stove, G. Galati, G. Pavan, F. De Palo, K. Lukin, K. Kulpa, J.S. Kulpa, L. Maslikowski. The NATO SET- 184 noise radar trials, Intl. Radar Syp., Krakow, Poland, May [22] P. McCorick, J. Jakabosky, S.D. Blunt, C. Allen, B. Hied. "Joint polarization/wavefor design and adaptive receive processing," IEEE Intl. Radar Conf., Washington, DC, USA, May [23] W.M. Brown, G.G. Houser, R.E. Jenkins. "Synthetic aperture processing with liited storage and presuing," IEEE Trans. Aerospace and Electronic Systes, vol. AES-9, no. 2, pp , Mar [24] D.L. Evans, T.G. Farr, J.J. van Zyl, and H.A. Zebker. Radar polarietry: analysis tools and applications, IEEE Trans. Geoscience & Reote Sensing, vol. 26, no. 6, pp , Nov [25] R. Touzi, S. Goze, T. Le Toan, A. Lopes, and E. Mougin. Polarietric discriinators for SAR iages, IEEE Trans. Geoscience & Reote Sensing, vol. 30, no. 5, pp , Sept