Foundations for a Multistatic Synthetic Aperture Radar (SAR) Imaging Capability

CAN UNCLASSIFIED

Foundations for a multistatic synthetic aperture radar (SAR) imaging capability

Development of a C‑Band prototype receiver and processor

Ryan A. English Christoph H. Gierull DRDC – Ottawa Research Centre

The body of this CAN UNCLASSIFIED document does not contain the required security banners according to DND security standards. However, it must be treated as CAN UNCLASSIFIED and protected appropriately based on the terms and conditions specified on the covering page.

Defence Research and Development Canada Scientific Report DRDC-RDDC-2021-R021 February 2021

CAN UNCLASSIFIED

IMPORTANT INFORMATIVE STATEMENTS

This document was reviewed for Controlled Goods by Defence Research and Development Canada (DRDC) using the Schedule to the Defence Production Act.

Disclaimer: This publication was prepared by Defence Research and Development Canada an agency of the Department of National Defence. The information contained in this publication has been derived and determined through best practice and adherence to the highest standards of responsible conduct of scientific research. This information is intended for the use of the Department of National Defence, the Canadian Armed Forces (“Canada”) and Public Safety partners and, as permitted, may be shared with academia, industry, Canada’s allies, and the public (“Third Parties”). Any use by, or any reliance on or decisions made based on this publication by Third Parties, are done at their own risk and responsibility. Canada does not assume any liability for any damages or losses which may arise from any use of, or reliance on, the publication.

Endorsement statement: This publication has been peer-reviewed and published by the Editorial Office of Defence Research and Development Canada, an agency of the Department of National Defence of Canada. Inquiries can be sent to: [email protected].

Template in use: EO Publishing App for SR-RD-EC Eng 2018-12-19_v1 (new disclaimer).dotm

CAN UNCLASSIFIED

Abstract

Analysis of the direct path received signal from the satellite at each receiver site was performed to extract relevant parameter values. The results were applied to range compression and back projection of the signal reflected from the scene to the receiver site, yielding good quality SAR imagery in all fifteen cases. Basic phenomenology of these bistatic scenes are under study, and the system is validated and baselined for expanding DRDC’s research into fully multistatic SAR imagery, polarimetric and interferometric bistatic phenomenology, and the opportunity to use other illuminators.

Significance to defence and security

DRDC has successfully validated and baselined an R&D ground-based bistatic SAR capability for C-band illuminators, allowing for the generation of near real-time SAR imagery at the receiver site independent of the space-based sensor system providing the illumination. Phenomenology specific to bistatic SAR imagery is under study and may be used on its own, or in combination with the traditional monostatic SAR imagery, to expand the imagery intelligence available about the scene contents.

DRDC-RDDC-2021-R021 i

Résumé

De plus, une analyse du signal du satellite reçu en trajet direct a été réalisée à chaque emplacement de récepteur pour en extraire les valeurs des paramètres pertinents. Les résultats ont été appliqués à la compression en portée et à la projection par transparence du signal réfléchi de la scène vers l’emplacement du récepteur, ce qui a produit des images SAR de bonne qualité dans les quinze cas. Par ailleurs, la phénoménologie de base des scènes bistatiques est à l’étude, et le système a été validé et établi comme base de référence pour étendre la recherche de RDDC à une phénoménologie bistatique polarimétrique et interférométrique de l’imagerie SAR entièrement multistatique, et créer une occasion d’utiliser d’autres illuminateurs.

Importance pour la défense et la sécurité

RDDC est parvenu à valider et à établir comme base de référence une capacité R & D au sol de SAR bistatique pour des illuminateurs en bande C et permette ainsi de générer des images SAR en temps quasi réel à l’emplacement du récepteur, indépendamment du système de capteur spatial qui produit l’illumination. La phénoménologie propre à l’imagerie SAR bistatique est à l’étude et peut être utilisée seule ou combinée à l’imagerie SAR monostatique traditionnelle pour accroître le renseignement par imagerie disponible au sujet des contenus de scènes.

ii DRDC-RDDC-2021-R021

Table of contents

Abstract ...... i Significance to defence and security ...... i Résumé ...... ii Importance pour la défense et la sécurité ...... ii Table of contents ...... iii List of figures ...... v List of tables ...... vii Acknowledgements ...... viii 1 Introduction ...... 1 2 The bistatic SAR paradigm ...... 3 2.1 Bistatic geometry ...... 3 2.2 Frames of reference ...... 3 2.3 Signal detection ...... 6 3 The prototype C-band receiver ...... 8 4 Bistatic image processing ...... 11 4.1 Direct path analysis ...... 11 4.1.1 Maximizing the direct path SNR ...... 12 4.1.2 Steps for direct path analysis ...... 12 4.1.3 Illumination period extents ...... 15 4.2 Reflected data processing ...... 16 4.2.1 Range-Doppler image ...... 16 4.2.2 Aligning the distance profile ...... 17 4.3 Image formation ...... 19 5 Field trials ...... 21 5.1 The RADARSAT-2 illumination source ...... 21 5.2 The experimental ground-based receiver system ...... 22 5.3 Field trial: DFL rooftop 2 ...... 22 5.4 Field trial: CRC boom lift ...... 23 5.5 Field trial: Carling Campus ...... 24 5.6 Field trial: 125’ boom lift ...... 26 5.7 Bistatic Active Radar Calibrator ...... 28 6 Analysis of collected data ...... 29 6.1 Direct path analysis ...... 29 6.2 Range resolution ...... 31 6.3 Azimuth resolution ...... 33 6.4 Bistatic imagery analysis ...... 35

DRDC-RDDC-2021-R021 iii

7 Conclusion ...... 38 References ...... 39 Annex A Step function detector ...... 41 A.1 Demarcation function ...... 41 A.2 Demarcation example...... 42 List of symbols/abbreviations/acronyms/initialisms ...... 44

iv DRDC-RDDC-2021-R021

List of figures

Figure 1: Bistatic geometry with moving Tx and elevated stationary Rx...... 3 Figure 2: Comparison between ECI (a–c) and ECEF (d–f) frames of reference for ground tracks of three consecutive RS-2 orbits. The ECI orbits (red) are a closed loop with the Earth’s surface moving eastward beneath it. In the ECEF reference, the orbits (blue) precess westward while the Earth’s surface remains fixed beneath it. Context imagery is the Blue Marble: Land Surface, Shallow Water, and Shaded Topography dataset from NASA’s Earth Observatory [18]...... 5 Figure 3: Comparison of an RS-2 orbit track in the ECI (red) and ECEF (blue) coordinate systems with the coordinate axes momentarily aligned near Ottawa, Canada. Context is provided using Landsat imagery from NOAA and distributed by the Google Image WMS Service. 6 Figure 4: Magnitude of sampled direct path receiver data (blue) with pulses present. The demarcation function (red) shows negative slope in the noise only regions and positive slope where signal is present...... 7 Figure 5: Receiver antennas for the DRDC bistatic prototype receiver system. The reflected signal from the ground is received via the fan antenna (a), while the direct path signal is obtained via a horn antenna (b)...... 8 Figure 6: Block diagram of clock components for the ground-based DRDC prototype bistatic SAR receiver hardware, showing dual local oscillators...... 9 Figure 7: Block diagram of ground-based DRDC prototype bistatic SAR receiver hardware components attached to the receiver antennas showing synchronized two channel processing...... 9 Figure 8: Block diagram of ground-based DRDC prototype bistatic SAR receiver hardware attached to the digitizer and storage showing synchronized two channel processing...... 10 Figure 9: Positive frequency spectrum (green) corresponding to direct path data fragment of 28.6 ms. The demarcation function (red, not to scale) contributes to the extraction of the spectrum edges to create a band pass filter (blue)...... 13 Figure 10: Samples from a series of direct path pulses around the signal region in (a) baseband and (b) after range compression...... 14 Figure 11: Magnitude in dB (blue), unwrapped phase (red, not to scale), and zero Doppler adjusted phase (green, not to scale) of range compressed direct path peaks. Inclusion of the pulses with magnitude above -3 dB are preferred, while those beyond a single phase cycle yield no computational benefit for this direct path analysis...... 15 Figure 12: (a) Range curvature for the pulse peak locations of the range compressed direct path data. (b) The black curve is the quadratic model of the range curvature. (c) The direct path range compressed data corrected for range curvature. (d) The reflected data from the main lobe pulses with Range and Azimuth compression. Data is from the 08 November 2015 bistatic collection...... 17

DRDC-RDDC-2021-R021 v

Figure 13: Quadratic range profile (blue) and scaled phase history (red) over central pulses of the main lobe for 02 March 2015 bistatic data collection...... 19 Figure 14: Backprojection processing for a pulse transmitted 2.5 km past broadside: (a) total path distance for pixels in a backprojection grid with 5m pixel spacing; (b) estimated main lobe illumination centred on white line (x = 2535 m) and terminating at the black line (x = -3295 m); and (c) distribution of response (in dB) from a single reflected pulse across the masked backprojection grid...... 20 Figure 15: Shirleys Bay Campus showing the location of the DFL rooftop deployments. The CRC and 125’ boom lifts are adjacent to DRDC’s T-86 building...... 23 Figure 16: Genie Z-45/25 RT articulating boom lift...... 24 Figure 17: Line-of-sight to the Carling Campus Building 5 tower...... 25 Figure 18: Genie S-125 telescopic boom lift...... 27 Figure 19: Forward-scattering bistatic VV imagery of 2015-10-05 having 20 incidence angle and 175 bistatic angle. (a) Response from a single pulse is distributed along iso-range ellipses. Iso-Doppler lines are vertical, intersecting the iso-range curves on either side of the mirror lines (white). (b) Scatterers above the mirror lines are real, those below are mirrored ambiguities...... 36 Figure 20: Side-looking bistatic HH imagery of 2015-10-16 with 48 incidence angle and 61.5 bistatic angle. Non-overlapping intervals of pulses used as a time series showing illumination moving from left to right. The mirror line is shown in white...... 37 Figure A.1: Magnitude of pulse interval (blue) showing signal region to the left and noise only region to the right of the demarcation point. The central tendacies of these regions are modelled by a discrete Heaviside function (red). The mean value (black) of the pulse interval along with the data itself is used to generate the demarcation function (green)...... 43

vi DRDC-RDDC-2021-R021

List of tables

Table 1: Deployment parameters for the Multistatic SAR field trial: DFL rooftop 2 acquisitions. 23 Table 2: Deployment parameters for the Multistatic SAR field trial: CRC boom lift acquisitions. 24 Table 3: Deployment parameters for the Multistatic SAR field trial: Carling Campus acquisitions.25 Table 4: Antenna positions for the Multistatic SAR field trial: Carling Campus acquisitions. . 26 Table 5: Deployment parameters for the Multistatic SAR field trial: 125’ boom lift...... 26 Table 6: Deployment parameters for the Bistatic Active Radar Calibrator...... 28 Table 7: Measured characteristic parameters of bistatic SAR data...... 29 Table 8: Measured characteristic parameters of bistatic SAR data...... 30 Table 9: Measured characteristic parameters of bistatic SAR data...... 30 Table 10: Range resolution in the slant plane down range of the receiver antenna, compared with the nominal impulse response width of the monostatic slant range image product given in [23]...... 32 Table 11: Nominal cross-range resolution for DFL rooftop and CRC boom lift collections. .. 33 Table 12: Nominal cross-range and BARC azimuth resolution derived from BARC measurement for Carling Campus and 125’ Boom Lift collections. The UltraFine, Wide MultiFine, and Spotlight modes are not subject to Equation (7)...... 34

DRDC-RDDC-2021-R021 vii

Acknowledgements

The authors would like to express their appreciation to the David Florida Laboratory and Jean-Pierre Charron for accommodating the deployment of the DRDC equipment on their site, and the Communications Research Centre Canada and Jerry Achtereekte for support and assistance in the deployment and use of the aerial lift platforms. The support received from personnel with the Vice Chief of Defence Staff, Public Services and Procurement Canada, and Brookfield Johnson Controls was essential to the success of the field trials conducted at the Carling Campus.

Extensive technical support for design and development of the receiver system hardware and for conducting the field trials was provided by Tony Laneve, Norm Reed, Pietro Reitano, and Heather Frampton of the Radar Technology and Engineering Group. Janice Lang of the Communcations Group provided photographic support.

Additional thanks to the Radar Electronic Warfare Section for the loan of equipment and their custom software to access the corresponding results. RAE expresses his appreciation to Chuck Livingstone for many useful conversations regarding the underlying operations of the RADARSAT-2 sensor.

viii DRDC-RDDC-2021-R021

1 Introduction

Under the Technology Investment Fund (TIF), Defence Research and Development Canada (DRDC) had embarked on a project to study the feasibility of bistatic SAR employing extremely large synthetic apertures times in which the transmitter and receiver do not communicate [1], [2]. The capability consisted of single stationary ground-based receivers deployed to capture the illumination of moving SAR transmitters, on both airborne and space-based radar (SBR) platforms, operating in either the C-band or X-band spectrum.

The work included collaboration with the US Air Force Research Laboratory (AFRL) under the auspices of the Trilateral Technical Research and Development Program (TTRDP), since rebranded the Technical Research and Development Program (TRDP). New collaborative bistatic work using RADARSAT-2 (RS-2) as the illuminator for AFRL’s Transportable Radar Receiver (TRR) occurred in late 2012. The TRR is a 12-channel phased array antenna, with a 13th channel for the direct path receiver, leading to new capabilities, such as beam-forming, that could be accomplished with the multi-channel data.

The turn of the millennium saw the resurrection of bistatic radar or, more precisely, of bistatic synthetic aperture imaging radar. Owing to the palpable advantages offered by bistatic SAR compared to conventional monostatic SAR, the world-wide research activity culminated in various proof-of-concept experiments including air-to-ground [2]–[4], air-to-air [5]–[7], and space-to-ground [1], [4]. By 2008, even the first space-to-air experiments had been successfully conducted and presented [8], [9]. Bistatic space-to-space applications had been originally discussed by Massonet in [10], and are currently operationally employed by TerraSAR-X/TanDEM-X to form a very high resolution topographic map of the entire Earth [11].

The main advantages of bistatic SAR are:  Low probability of intercept based on the fact that the receiver does not emit electromagnetic energy thereby making it much harder to be detected and intercepted either by electronic jamming or destruction;  Improved target recognition including stealthy targets based on the different aspect angle ranges resulting in much different reflection properties; and  Forward/backward looking capability for the receiving unit under specific favourable geometry conditions [12], a trait physically impossible for a classic monostatic side-looking SAR. This feature might be of operational use as a bad-weather landing aid in aircrafts, or as an obstruction warning and avoidance system in helicopters.

There is a trade-off, however, between these advantages and several technological challenges, including:  The two separate radar systems employ independent local oscillators so that their phase noise accumulates rather than cancelling out as in classic SAR [2];  The technological time, phase and antenna pointing (beam steering) synchronization challenges are much more complex [13];  The bistatic SAR image processing is much more difficult due to the truly 3D-geometry in contrast to the valid 2D-approximation for monostatic SAR. As a consequence, the bistatic range history cannot

DRDC-RDDC-2021-R021 1

be approximated as a parabola anymore and hence fast FFT-based focussing algorithms are not applicable [14]–[17]; and  Even the theoretical derivation and prediction of the achievable resolution, which predominantly depends on the geometry, is much more complicated and the mathematical framework to achieve this has only been rather recently developed [1].

A natural step beyond this is to physically separate the locations of the received channels so as to explore a truly multistatic SAR capability. To this end, DRDC has designed, deployed, and processed its own ground-based C-band receiver system with the intention performing a contemporaneous data collection of multiple receivers illuminated from a single transmitting source, namely RS-2. It is the development of this DRDC prototype C-band receiver system that is documented in this report.

2 DRDC-RDDC-2021-R021

2 The bistatic SAR paradigm

2.1 Bistatic geometry Bistatic SAR differs from monostatic SAR in that the transmitter (Tx) and receiver (Rx) do not use the same antenna, but are instead independent from each other. To form SAR imagery, at least one of the two must be in motion [1]. By using commercial SBR as the source of illumination, the simplest bistatic case involves the deployment of a stationary receiver. The bistatic SAR imagery is generated by measuring the radar signal scattered by objects in the scene toward the Rx antenna. Thus, line of sight must exist from the Tx antenna to the objects and from the objects to the receiver antenna. Elevation of the antennas is a primary method of achieving a maximal line of sight. While the ground based receiver does not need to lie within the footprint of the monostatic scene, there are advantages for having the receiver in the scene, which would allow direct path pulses generated within the main lobe of the transmitted signal to be collected to increase the signal-to-noise ratio (SNR). The geometry for such a bistatic SAR scenario is shown in Figure 1.

Figure 1: Bistatic geometry with moving Tx and elevated stationary Rx. 2.2 Frames of reference The evolution of SBR from RS-1 to RS-2 involves a change in the frame of reference used to describe the operation of these systems. In particular, the concept of the “broadside” direction has been redefined, allowing for a simplification in the calculations for RS-2 processing compared to RS-1.

When space-based radar illuminators are introduced, the frame of reference for the satellites is no longer synchronized to the rotating surface of the Earth. Typically, an earth centered inertial (ECI) coordinate system would be used, which is defined by the North vector, originating at Earth centre along the axis of Earth rotation passing to the North Pole and the vector from Earth centre to the ascending node of the orbit, which passes along the intersection of the orbit and equatorial planes. In the Mercator projection of this system, the nadir of the satellite orbit traces out a terminator like path while the Earth’s surface moves easternly beneath it. Note that the origin of this motion can still vary, by fixing a specific time at which the projection is viewed. Three consecutive orbits of RS-2 are shown in Figure 2(a–c). The Earth rotates eastward 25.25 in longitude, , during each 101 minute orbit, so that while the satellite returns to the same point in the frame of reference, the point on the ground beneath it is different each time, the dependency being that:

DRDC-RDDC-2021-R021 3

. () = (0) + , (1) where is time in seconds. Alternatively, when a frame of reference is fixed to the surface of the Earth, such as in the earth-centered, earth-fixed (ECEF) coordinate system, it is the orbit plane of the satellite that rotates with respect to the coordinate axes. ECEF is defined by the North vector and one that passes from the Earth centre and passes through the surface of the Earth at (L, F) = (0, 0), the intersection of the Prime Meridian and the Equator. In ECEF, after each 101 minute orbit, RS-2 returns to the same latitude, but has a ground track 25.25 west of that of the previous orbit, as shown in Figure 2(d–f).

(a) (d)

(b) (e)

(c) (f) Figure 2: Comparison between ECI (a–c) and ECEF (d–f) frames of reference for ground tracks of three consecutive RS-2 orbits. The ECI orbits (red) are a closed loop with the Earth’s surface moving eastward beneath it. In the ECEF reference, the orbits (blue) precess westward while the Earth’s surface remains fixed beneath it. Context imagery is the Blue Marble: Land Surface, Shallow Water, and Shaded Topography dataset from NASA’s Earth Observatory [18].

4 DRDC-RDDC-2021-R021

The implications of working in one frame of reference in contrast to the other is highlighted in Figure 3, which shows the divergence of the ECI nadir (red) and ECEF ground track (blue) from a momentarily aligned point. In the ECI satellite frame of reference, it is along the red line, and so first generation commercial SBRs, such as RADARSAT-1 (RS-1), would have the antenna pointed broadside to the orbital plane, requiring the Earth’s velocity to be accounted for in the image processing, e.g., [1], in the form of a Doppler centroid correction.

For second generation SBRs, such as RS-2, yaw steering has been introduced [19] causing the antenna direction to be altered by a yaw angle, Y, as a function of the orbit internal angle, a, namely,

Y = cos(a), (2) where is a constant coefficient fixed by the satellites orbit parameters. This yaw steering places the range direction perpendicular to the ground track line in the ECEF frame of reference, which yields a zero Doppler centroid at the image centre, and simplifies the image processing, at least for beam modes that adhere to the yaw steering formula in Equation (2).

Figure 3: Comparison of an RS-2 orbit track in the ECI (red) and ECEF (blue) coordinate systems with the coordinate axes momentarily aligned near Ottawa, Canada. Context is provided using Landsat imagery from NOAA and distributed by the Google Image WMS Service.

2.3 Signal detection

Monostatic SAR systems are able to record all the parameters used to create the transmission pulse, as well as dynamic information about the antenna location, geometry and timing information of the transmissions themselves. This information is not available for use by the bistatic receiver system unless known a priori or cooperatively communicated, e.g., access to the metadata of the monostatic image product. Otherwise, the characteristic parameters of the transmitted signal need to be estimated by measuring the signal itself, once an appropriate signal is detected within the received data.

For SAR systems, the principles of pulse-Doppler radar are employed, meaning that the radar signal is transmitted by a pulse followed by an interval where the radar signal is absent, and this pattern is repeated over a period of time for the duration during which the target location is illuminated. There are several characteristics which could be used to test for the presence of signal, but we shall first consider that of the mean value.

DRDC-RDDC-2021-R021 5

In a time interval containing solely noise, it is expected that the statistics of the received data will be constant whenever the time interval is long relative to the sampling interval. When signal pulses are present, it is expected that statistics over the pulse will be significantly different than the noise-only case.

Consider a time interval over which samples of data are recorded, where the first samples have signal data, while there is noise only for > . There are two distinct parts which we are able to characterize by the behaviour of the central tendency with the signal and noise regions, i.e., the mean value over each subinterval (see Figure 4). The central tendency of these regions can be modelled with a discrete Heaviside function. The demarcation between signal and noise regions can be located by maximizing the difference between the central tendencies on either side of the demarcation point, which is described in more detail in Annex A.

The demarcation function, (A.14), can be applied in the time domain to extract estimates for the pulse repetition interval (PRI), the pulse width, and the pulse replica from the direct path signal data to be used as a match filter for range compression. In the frequency domain, its application supports the derivation of the bandwidth and the intermediate frequency (IF) (or central frequency, if applicable).

Figure 4: Magnitude of sampled direct path receiver data (blue) with pulses present. The demarcation function (red) shows negative slope in the noise only regions and positive slope where signal is present.

6 DRDC-RDDC-2021-R021

3 The prototype C-band receiver

A C-band bistatic receiver system had previously been successfully deployed for the RS-1 SBR and Environment Canada CV-580 C-band airborne radar data collection under the previous TIF project. The prototype receiver system was modified to accommodate the characteristics of RS-2, specifically the 5.405 GHz centre frequency and dual polarized transmission, but otherwise leverages off the successes of the previous system.

Two fan antennaes were constructed, one H polarized, the other V polarized, to accommodate the polarimetric capabilities. For this prototype, only one of these is deployed at a time. The antenna pattern for this equipment has a vertical -3dB beamwidth of 12 and 45 in the horizontal direction. The antenna is mounted on a 2-axis swivel head allowing tilt and rotation, and is typically deployed on a tripod in an elevated location, such as a rooftop, or attached to the working platform of a tower, or boom lift (Figure 5a). The direct path signal is received using a horn antenna with a 20 beamwidth and similarly deployed (Figure 5b) so as to be pointing at the nominal point of closest approach by the illuminating platform.

(a) (b)

Figure 5: Receiver antennas for the DRDC bistatic prototype receiver system. The reflected signal from the ground is received via the fan antenna (a), while the direct path signal is obtained via a horn antenna (b).

Two local oscillators (LO) are used to convert the frequency down to an IF. The initial LO is at 4800 MHz, which combined with the RS-2 centre frequency yields an expected 605 MHz IF. A second LO, at either

DRDC-RDDC-2021-R021 7

660 MHz or 665 MHz is used to bring the IF to either 60 MHz or 55 MHz. The latter is used when the RS-2 beam mode is expected to have a bandwidth of 100 MHz, since the filter used for such wide-band acquisition has a band pass width of 115 MHz, which might introduce truncation effects into the recorded signal at an IF of 60 MHz.

Figure 6: Block diagram of clock components for the ground-based DRDC prototype bistatic SAR receiver hardware, showing dual local oscillators.

FIL1A AMP1A FIL2A MIX1A AMP3A FIL3A FAT1A AMP2A FIL4A IF1 (Output) RF1 (Input) R I L

ISO1A FIL5A

PD1 PLO1 FAT3 AMP5 1 REF IN 2 ISO1B FIL5B

FIL1B AMP1B FIL2B MIX1B AMP3B FIL3B FAT1B AMP2B FIL4B IF2 (Output) RF2 (Input) L R I

Figure 7: Block diagram of ground-based DRDC prototype bistatic SAR receiver hardware components attached to the receiver antennas showing synchronized two channel processing.

8 DRDC-RDDC-2021-R021

FAT2A MIX2A AMP4A FIL6A IF1-2 (Output) R I L

FAT4A AMP6A

PD2

2 FAT4B AMP6B

FAT2B MIX2B AMP4B FIL6B

L IF2-2 (Output) R I

Figure 8: Block diagram of ground-based DRDC prototype bistatic SAR receiver hardware attached to the digitizer and storage showing synchronized two channel processing.

DRDC-RDDC-2021-R021 9

4 Bistatic image processing

Receiver data are collected and stored in files containing signal data from each of the antenna for the system. Data files from the initial DRDC bistatic prototype have one direct path receive channel and one reflected signal receive channel, each polarized with either a horizontal or vertical orientation, depending on the antenna deployed. The direct path signal is used to extract and estimate characteristic parameters about the transmitted signal without any a priori requirements being imposed by the bistatic collection team. Much of this information is available from the metadata of the monostatic image product, once it is available, but minimizing the requirement for information about the receive signal other than what can be extracted from the received data is one of the goals of the system. Near real time image processing can be achieved when the dependency on the monostatic image product metadata is broken. 4.1 Direct path analysis

The direct path signal data serves several purposes in the bistatic image processing workflow. It provides an exact match filter for range compression; it establishes the timing for each and every transmitted pulse; and it provides the clean signal from which to establish the characteristics of the transmitted signal.

The raw direct path data can be partitioned in a few important ways. First, the recorded data may be divided into a pre-illumination period, followed by the illumination period, and then a post-illumination period. Normally, there is no signal from the transmitter collected before or after the illumination period, however it may occur that part of previous or subsequent image acquisitions may be captured as part of the data recording process. Thus, there is no guarantee that the pre- and post-illumination periods are entirely free of signal, however, any such signal can be ignored as if these periods were signal-free. As such, we shall henceforth consider the case where the pre- and post-illumination periods contain only noise without loss of generality.

The second division of data occurs in the illuminated period where each measured samples are expected to be either a combination of signal and noise or noise only. SAR transmitters generate signal for a set length of time, the pulse width, and then fall silent to allow the reflected signal to be received without any possible interference from the transmitter. During the latter period, the direct path channel records only noise, and this period also occurs for a set length time. The simplest way to construct a SAR transmission system is to have this pattern of alternating signal transmission and noise-only intervals repeated throughout the illumination period. Each pair of signal and noise-only intervals forms a pulse interval. Since each part is of fixed time, the PRI remains constant throughout the illumination period. Partitioning the received signal into these pulse intervals is key to correctly performing the image processing.

All RS-2 modes are stripmap with the exception of the SLA mode, which provides an electronically steered sliding spotlight mode, limited to 1 delta from broadside. As such, in most cases, the direct path signal shows a very high signal to noise ratio (SNR), allowing edge detection of the signal pulses, while the bistatic receiver lies within the main lobe of the transmitting antenna. But as the transmitting platform moves, nulls in the antenna pattern eventually fall upon the receiver and low SNR results, even though we are near the centre of the illumination period. Indeed, a robust bistatic image processing strategy should consider the case where the bistatic receiver is never in the main lobe of the transmitting antenna as this is effectively the case for monostatic receivers.

10 DRDC-RDDC-2021-R021

The algorithms used for analysis of the direct path require a minimum of three pulse intervals to be present (to fit a parabola) and five to ensure validity (permitting interpolation across three pulse intervals). For good measure, the number of pulses should span the 3dB width of the main lobe, when present. The maximum benefits are achieved once the number of pulse intervals span a full PRF shift of the range compressed pulse peak phase values, as will be seen in Section 4.1.2 (e.g., Figure 11), so any additional samples would only slow the processing. 4.1.1 Maximizing the direct path SNR

To maximize the SNR, range compression of the direct path signal is the goal, whereby each pulse interval is convolved with a matched filter. For a system where all the direct path pulses are identical, the same match filter can be used for each pulse interval. The optimal match filter, however, is generated from the pulse itself, leaving the dilemma of requiring one to already have a pulse interval in order to determine how to partition the data into pulse intervals. We therefore seek a strategy to first extract a single pulse without performing range compression, and then refine the partitioning following range compression.

SAR signal pulse waveforms are carefully constructed specifically to allow for the reflected signal to be processed into imagery. The signal is constructed from a chirp, a frequency modulation around a centre frequency. The interval in frequency space covered by this modulation is the SAR system bandwidth. Furthermore, any source of frequency external to this interval will be spurious, contributing only noise to the eventual image. One way to improve the SNR is to reduce the contribution of sources outside the frequency bandwidth. 4.1.2 Steps for direct path analysis To begin, a section of the illumination period, expected to contain all the pulses of the largest lobe of the transmitter antenna at the bistatic receiver location, is extracted. Upon this data, analysis is performed to estimate several key characteristic parameters, specifically, the following are performed:

1. Subtract the mean value, i.e., the DC bias;

2. Locate the sample with the peak magnitude;

3. Locate a sample with the next largest magnitude prior to the peak location;

4. Locate a sample with the next largest magnitude subsequent to the peak location;

5. Extract the subset of data in the fragment interval between the two secondary peaks;

6. Transform the fragment into frequency space;

7. Use a step function detector to locate the extents of the frequency spectrum;

8. Zero the frequency data external to the estimated step;

9. Use a histogram method to detect the edges of the frequency spectrum;

10. Estimate the bandwidth and centre frequency of the spectrum (see Figure 9);

DRDC-RDDC-2021-R021 11

11. Apply a band pass filter to modify the time domain data fragment;

12. Demodulate the data fragment into baseband;

13. Use a step function detector to locate the extents of all the signal pulse widths; and

14. Calculate initial PRI as the interval between leading edges of the signal pulse steps.

Figure 9: Positive frequency spectrum (green) corresponding to direct path data fragment of 28.6 ms. The demarcation function (red, not to scale) contributes to the extraction of the spectrum edges to create a band pass filter (blue).

The initial estimates of the signal parameters extracted from the fragment of data are then applied to the full section of data identified as covering the main lobe, as follows:

1. Trim the original direct path data section to contain an integer number of PRIs such that the peak location occurs ¼ into a pulse interval;

2. Apply band pass filtering, demodulation and frequency shift the subset data into baseband;

3. Partition the direct path baseband into pulse intervals;

4. Apply the step function detector to locate the edges of the signal data in the pulse interval of peak magnitude;

5. Estimate the pulse width of the signal data in the peak magnitude interval;

12 DRDC-RDDC-2021-R021

6. Create a match filter from the signal data of the peak magnitude pulse; and

7. Range compress each pulse interval in the direct path partition (see Figure 10) by convolving with the match filter.

(a) (b)

Figure 10: Samples from a series of direct path pulses around the signal region in (a) baseband and (b) after range compression.

Once range compressed, a refinement of parameter estimates is performed and the pulses synchronized. The latter is based on the expectation that the zero Doppler conditions occur when the transmitter is located at its minimum distance to the receiver, as described in Section 2.2. The steps performed are as follows:

1. Locate the range compression peak in each pulse interval;

2. Re-estimate the PRI using the interval between range compressed pulse peaks;

3. Repartition the range compressed data section using the re-estimated PRI;

4. Curve fit the range compressed peaks to determine optimal peak locations;

5. Refine PRI estimate by using the interval between optimal peak locations;

6. Extract the range compressed peaks;

7. Determine the zero Doppler pulse interval by autocorrelation of the range compression peaks; and

8. Estimate the frequency shift to the range compressed peak phase history that aligns the minimum range phase profile to the zero Doppler pulse (see Figure 11).

DRDC-RDDC-2021-R021 13

Figure 11: Magnitude in dB (blue), unwrapped phase (red, not to scale), and zero Doppler adjusted phase (green, not to scale) of range compressed direct path peaks. Inclusion of the pulses with magnitude above -3 dB are preferred, while those beyond a single phase cycle yield no computational benefit for this direct path analysis.

All the parameters required to perform range compression and synchronization with the direct path pulses can then be used to pre-process the signal data reflected from the illuminated scene. The extents for which valid imagery may be formed is not restricted to the section of data whereby the main lobe is illuminating the receiver. 4.1.3 Illumination period extents

The partition of the direct path section can be extended to the entire data file, trimmed so as to contain an integer number of pulse interval sized slots. The slots in the illumination period will contain pulses; those outside this period will not. Any slot may be tested for the presence of a signal by:

1. subtracting the DC bias,

2. applying the bandpass filter,

3. demodulating and applying the frequency shifts, and

4. convolving with the match filter.

14 DRDC-RDDC-2021-R021

Slots that have a signal present will generate a peak significantly larger than the mean value of the slot. Empirically, a tenfold ratio is indicative of signal containing slot while slots devoid of signal do not exhibit peak values more than a few times the mean value. To determine the extents of the illumination period, a binary search is conducted on the intervals from the first slot to the slot containing the zero Doppler pulse, and from the zero Doppler pulse to the last slot in the file. 4.2 Reflected data processing

The reflected data is band pass filtered, and shifted into baseband using the parameters extracted from the direct path analysis. The same direct path pulse is used for range compression of the reflected data, and interpolated based on the calculated optimal PRI into range lines of equal length, based on the estimated coverage distanced of the receive antenna. This range compressed reflected signal is the basis of image formation. 4.2.1 Range-Doppler image

Whereas additional information about the transmitter, in particular it’s state vectors, is required to fully implement back-projection image processing, as described in Section 4.3, the Range-Doppler image can be formed with the information currently in hand. Back-projection image formation is not dependent on any aspect of Range-Doppler processing in this subsection, but the Range-Doppler image does provide an earlier opportunity to identify anomalies in the data quality or parameter estimates performed up to this point.

The Range-Doppler image is formed by calculating the range curvature extracted from peak location of the direct path analysis. The range curvature is modelled as a quadratic function fitted along the direct path peaks, as shown in Figure 12(a–b). Straightening this curve provides the correction in timing that occurs due to the range difference between transmitter and receiver (Figure 12[c]). This line of direct path peaks is used as the match filter for azimuth compression through convolution with each corrected range line of the reflected data, which results in a Range-Doppler image (Figure 12[d]).

DRDC-RDDC-2021-R021 15

(a) (b)

Figure 12: (a) Range curvature for the pulse peak locations of the range compressed direct path data. (b) The black curve is the quadratic model of the range curvature. (c) The direct path range compressed data corrected for range curvature. (d) The reflected data from the main lobe pulses with Range and Azimuth compression. Data is from the 08 November 2015 bistatic collection.

4.2.2 Aligning the distance profile

For RS-2 data, the distance profile is extracted from the orbital state vectors of the satellite, either downloaded directly from the operator, MacDonald Dettwiler and Associates (MDA), under the terms of the GoC master license, of from the metadata provided with the monostatic image product from MDA. These state vectors are normally collected at regular intervals, e.g., every 4 seconds. For image product metadata, five state vectors are issued, the first and last corresponding to the transmission times of the first and last pulses, the order depending on the orbital direction of the collect, and the other three equally spaced between.

16 DRDC-RDDC-2021-R021

⃗ The state vector of the transmitter for each pulse received, G(), is calculated in the imagery coordinate system by aligning the minimum of the distance curve between the orbiting platform and the fixed ground-based receiver location to the minimum of the distance profile extracted from phase of the direct path peaks. The coordinate system is constructed from vectors in ECEF, ⃗, by normalizing the -coordinate running from the ground projection of the receiver antenna onto the WGS-84 ellipsoid to its 3-dimentional location at height, ℎ, above the ellipsoid. A nominal ′-direction is established along the direction of the vector from the first pulse state vector to the last pulse state vector of the satellite platform. The -coordinate is established by normalizing the cross product of ̂ and ′. Finally, the normalized cross product of and ̂ produce the definitive -direction.

C,F,⃗C(,F,)⃗ ̂ , ⃗ ⃗ C(,F,)C(,F,) ⃗ ⃗ ⃗ G G(), (3) ̂ ×⃗ , ̂ ×⃗ ×̂ . ‖ ×̂‖

This coordinate basis {, , ̂ } establishes the imagery coordinate system in which further processing may occur. As such, under this paradigm, each bistatic SAR data collection has its own unique coordinate system.

Using the state vectors of the satellite in this coordinate system, a range profile, (), of the direct path between the satellite and receiver is fitted to a quadratic function of time. It is expected that this profile will match the Doppler phase history profile, f(), used for zero Doppler adjustment in Section 4.1.2, at least for the pulses contained in the main lobe of the transmitter, according to:

l ( ) = f( ), (4) 2p where l is the wavelength of the signal being transmitted.

Aligning the minima of () and () establishes a global timestamp for the zero Doppler pulse to which all other pulses and data slots can be related. Figure 13 shows the quadratic curve derived using the state vectors of the monostatic RS-2 metadata from the standard image product, shown in blue, and the scaled phase history data (Equation (4)) of the central pulses for the direct path data collected 02 March 2015, shown in red, after the minima are aligned.

DRDC-RDDC-2021-R021 17

Figure 13: Quadratic range profile (blue) and scaled phase history (red) over central pulses of the main lobe for 02 March 2015 bistatic data collection.

4.3 Image formation

Imagery is formed using a time domain backprojection approach developed for [1], [20]. An earth-fixed grid of image pixels is laid out in the ground plane, at a constant height = above the ellipsoid, of the imagery coordinate system established in Section 4.2.2. For purposes of the initial bistatic SAR capability, the pixel spacing, and , is a selectable parameter for each image processing, but no elevation model is used for height. However, the processing algorithm is capable of using height values if made available for each pixel in the grid. This can be accomplished by including a matrix of elevation values from a Digital Elevation Model source oriented to the imagery coordinate system.

For each pulse, the total path length for each pixel is calculated from transmitter to pixel to receiver within the imagery coordinate system (Figure 14[a]). Pixels that lie within a mask estimated for the main beam lobe footprint for that pulse are processed (Figure 14[b]). The reflected rangeline associated with the pulse is then coherently added to the pixel values that share the same distance path length as each rangeline matrix element (range bin), interpolated for range bins that cross pixel edges (Figure 14[c]).

Once all pulses are distributed and coherently summed, the image is formed and is available for post processing, such as calibration, georeferencing, or the generation of one or more image products. These remain as future components of the bistatic SAR system.

18 DRDC-RDDC-2021-R021

(a)

(b)

(c)

Figure 14: Backprojection processing for a pulse transmitted 2.5 km past broadside: (a) total path distance for pixels in a backprojection grid with 5m pixel spacing; (b) estimated main lobe illumination centred on white line (x = 2535 m) and terminating at the black line (x = -3295 m); and (c) distribution of response (in dB) from a single reflected pulse across the masked backprojection grid.

DRDC-RDDC-2021-R021 19

5 Field trials

Led by the SBR group, the Radar Sensing and Exploitation section deployed, evaluated, and configured a prototype bistatic ground-based C-band receiver system in support of the development of a robust field deployable R&D capability that would see multiple ground-based receiver systems used to support multistatic SAR image generation and analysis. 5.1 The RADARSAT-2 illumination source

The Government of Canada (GoC) coordinates the prioritization of its requests for RS-2 imagery over Canadian territory using the Enhanced Management of Orders and Conflicts (EMOC) process [21] implemented by the Canadian Space Agency (CSA). Requests are submitted two full calendar months in advance to allow de-confliction to occur. Conflicting GoC requests are identified, resolutions proposed to the participants, and decisions implemented into an acquisition plan that is submitted to the RS-2 Order Desk against the GoC data allocation approximately one month ahead of the first acquisition in that month’s plan. Once submitted, the acquisition plan is compared against the RS-2 commercial requests by the RS-2 operator—namely MDA—and their de-confliction process is implemented.

Approximately 2–3 weeks before the start of a given month, a master list of the successful GoC RS-2 orders is disseminated to all registered GoC RS-2 Order Desk POCs, which includes all the order parameters, including nominal imagery footprints expected from the upcoming collections. In this manner, suitable bistatic/multistatic collection opportunities are identified on a non-interference basis, without any external request being made by DRDC. The EMOC process is not used for collections outside Canadian territory, so details about acquisitions over the rest of the globe are discoverable only when ordering conflicts occur.

A typical month has 15–20 potential imaging opportunities of the Ottawa area by RS-2. Of those, 4–8 slots include the Shirleys Bay campus in the image footprint. Although it is not strictly necessary to be in the footprint, it is advantageous to have the receiver system lie within the 3 dB antenna pattern of the main RS-2 lobe to ensure access to a clear example of the transmitted pulse signal for match filtering purposes and to correlate the pulse timing with the satellite state vectors.

MDA supplies predicted orbit model data six months in advance for use with its order desk software, the acquisition planning tool (APT). The relevant state vectors are included in the metadata (product.xml file) of each data product generated by MDA. CSA Order Desk participants also have access to post-acquisition orbit files based on the measured ephemeris of the satellite.

RS-2 is in a polar sun-synchronous orbit with an inclination of 98.6 at a mean altitude of 798 km with an orbital period of 100.7 minutes [22]. This results in a 24-day repeat pass cycle with respect to a specific point on the Earth’s surface. The RS-2 SAR operates with a centre frequency of 5405 MHz and has beam modes with nominal 11.6 MHz, 17.3 MHz, 30 MHz, 50 MHz and 100 MHz bandwidths [23]. Dual-polarized transmit modes, i.e., alternating H and V pulses, occur with up to 30 MHz bandwidths. Incidence angles within the footprint of commercially available modes can run from 17.5 up to 49.9. The antenna is pointed marginally away from the orbit broadside using yaw steering according to latitude to achieve zero Doppler orientation [19], most modes being stripmap. Azimuth steering is used in the spotlight modes to briefly stare at a central point for a limited spotlight capability.

20 DRDC-RDDC-2021-R021

5.2 The experimental ground-based receiver system

The experimental ground based receiver system was initially based on C-band requirements established in the 2002–2006 Bistatic SAR TIF project [1] when RADARSAT-1 was used as the illuminator, which operated at a center frequency of 5.3 GHz and did not have the higher bandwidth modes that RS-2 does. As such, the initial design of the receiver system did not include the RS-2 100 MHz bandwidth modes, which was addressed in mid-October when new filters were procured.

Two antennas are deployed as part of this system. A horn antenna for direct path acquisition is pointed skyward at the expected point of closest approach of the RS-2 satellite. For reception of the reflected energy, a fan-beam antenna with a 45 3bB beam width in azimuth is pointed downwards at terrain or targets within the expected imagery footprint. To achieve reasonable lines-of-sight, the requirement is to have this antenna elevated and pointed downward at the terrain. The fan-beam antenna has a 12 3dB beam in elevation and is typically pointed down 6 or more below the horizon.

Two local oscillators at 4800 MHz and 665 MHz are used to convert the signal to an IF of 60 MHz for narrow band modes. When working with 100 MHz bandwidth modes, the second oscillator is set to 660 MHz and yields a 55 MHz IF.

The two channels are digitized at a 250 MHz sampling frequency and recorded continuously for a predetermined number of samples, setting the final file size. To ensure capture of the entire illumination signal, and avoid prediction errors, a 3-minute continuous collect window centred on the expected product’s acquisition start time was implemented. Storing 2 bytes for each sample in 2 channels for 3 minutes at 250 MHz yields 180 x 109 bytes of data, or file sizes of approximately 167 GB. The duration of illumination for an entire RS-2 image acquisition is typically no more than 20 seconds. Additionally, the PRI of RS-2 is at least two orders of magnitude greater than the pulse width, indicating that a significant portion of inter-pulse samples contain no signal information. 5.3 Field trial: DFL rooftop 2

Three data collections acquired in March 2015 using the prototype system confirmed successful configuration of the system, other than the restriction on illuminator bandwidth. Prior Lab and Field Test results guided many aspects of the deployment. The receiver was deployed on the rooftop of the David Florida Lab (DFL), indicated in Figure 15, at Shirleys Bay (45 N 20.830ʹ, 75 W 53.012ʹ) at 92.0 m height above mean sea level (MSL). Ground level at the location is estimated to be 73 m MSL. At Shirleys Bay, the difference between MSL and the height above the ellipsoid (HAE) is 33.87  0.03 m. The deployment parameters are provided in Table 1.

RS-2 beam modes are given descriptive names. Incidence angles are quantized to pre-configured subsets of the available sensor swath and identified by number. Right or left looking modes are identified with an R or L, respectively [23]. The start date/time indicates the UTC time at which RS-2 illumination is intended to begin for that particular acquisition. The vector from the direct path receiver to the point of closest approach of the satellite orbit is identified by its azimuthal bearing measured clockwise from true north (TN) and the incidence angle measured from vertical at the direct path receiver. Similarly, the vector to a nominal “bistatic” scene centre is identified using the azimuthal bearing from TN and an elevation angle measured from the horizontal plane at the bistatic receiver.

DRDC-RDDC-2021-R021 21

Figure 15: Shirleys Bay Campus showing the location of the DFL rooftop deployments. The CRC and 125’ boom lifts are adjacent to DRDC’s T-86 building.

Table 1: Deployment parameters for the Multistatic SAR field trial: DFL rooftop 2 acquisitions.

RS-2 Tx Start Direct Direct Rx Bistatic Bistatic  beam Pol. Date/Time Path Path Pol. Azimuth Elevation mode Azimuth Incidence Angle Angle

Standard V 2015-03-02 262.5 51.5 V 85.0 -10.5 -2.5 8_R 23:11:53

Standard V 2015-03-09 261.5 47.5 V 85.0 -5.5 -3.5 7_R 23:07:41

Standard V 2015-03-30 261.0 35.0 V 85.0 -5.5 -4.0 3_R 22:55:00

5.4 Field trial: CRC boom lift

Initial investigations into collection requirements where the receiver antennas were elevated using a boom lift were conducted in August 2015 using a Genie Z45/25 RT articulating boom lift, shown in Figure 16

22 DRDC-RDDC-2021-R021

adjacent to Building T-86 at Shirleys Bay (45 N 20.833ʹ, 75 W 53.310ʹ). Ground level at the location is measured at 80.4 m MSL. The estimated elevation of the receiver antennas is 14.7 m above ground level (AGL). The deployment parameters are provided in Table 2.

Table 2: Deployment parameters for the Multistatic SAR field trial: CRC boom lift acquisitions.

RS-2 beam Tx Start Direct Direct Rx Bistatic Bistatic  mode Pol. Date/Time Path Path Pol. Azimuth Elevation Azimuth Incidence Angle Angle

Wide Fine H+V 2015-08-04 258.5 31.0 V 79.0 -6.0 -0.5 Quad 11_R 22:50:55

Fine Quad H+V 2015-08-05 98.5 47.5 V 278.0 -6.0 0.5 29_R 11:01:31

Figure 16: Genie Z-45/25 RT articulating boom lift. 5.5 Field trial: Carling Campus

Due to access limitations, rooftop and tower deployment locations have been difficult to arrange. An opportunity to deploy on the rooftop of the central tower at the Carling Campus, seen in Figure 17, recently acquired by DND was pursued to exercise and investigate the portability requirements of the receiver system. Ground level at the location is measured at 65.7 m MSL, with the rooftop at 90.3 m MSL. The deployed height above the rooftop is 1.4 m. The deployment parameters are provided in Table 3. To meet line-of-sight requirements, the antennas were deployed at various locations on the rooftop, as given in Table 4.

DRDC-RDDC-2021-R021 23

Table 3: Deployment parameters for the Multistatic SAR field trial: Carling Campus acquisitions.

RS-2 beam Tx Start Direct Direct Path Rx Bistatic Bistatic  mode Pol. Date/Time Path Incidence Pol. Azimuth Elevation Azimuth Angle Angle

Wide Fine H+V 2015-09-14 259.5 36.5 V 79.0 -6.0 0.5 Quad 16_R 22:54:38

Wide Fine H+V 2015-09-21 259.0 31.0 V 142.0 -6.0 -63.0 Quad 11_R 22:50:28

Wide 2_R V 2015-09-25 101.0 34.0 V 262.0 -6.0 19.0 11:13:26

Fine Quad H+V 2015-10-05 257.5 20.0 V 262.0 -6.0 175.5 1_R 22:42:12

Figure 17: Line-of-sight to the Carling Campus Building 5 tower.

24 DRDC-RDDC-2021-R021

Table 4: Antenna positions for the Multistatic SAR field trial: Carling Campus acquisitions. Date Antenna Position 2015-09-14 45 N 21.052ʹ 75 W 51.069ʹ 2015-09-21 45 N 21.046ʹ 75 W 51.060ʹ 2015-09-25 45 N 21.049ʹ 75 W 51.082ʹ 2015-10-05 45 N 21.049ʹ 75 W 51.082ʹ 5.6 Field trial: 125’ boom lift

Moving beyond the August trials with the 45’ boom lift, which appeared to produce stable results without motion compensation, a Genie S-125 telescopic boom lift, shown in Figure 18, was rented for a four week period and deployed next to the T-86 building at Shirleys Bay (45 N 20.813ʹ, 75 W 53.336ʹ). The boom was elevated to 83.5 m HAE. Ground elevation was measured at 46.0 m HAE at the base of the boom lift. The deployment parameters are provided in Table 5.

Table 5: Deployment parameters for the Multistatic SAR field trial: 125’ boom lift.

RS-2 beam Tx Start Direct Direct Rx Bistatic Bistatic  mode Pol. Date/Time Path Path Pol. Azimuth Elevation Azimuth Incidence Angle Angle

Wide UltraFine H 2015-10-16 98.5 48.0 H 339.0 -6.0 -60.5 24_R 11:01:07

Wide MultiFine V 2015-10-19 100.5 35.0 H 281.0 -6.0 -0.5 21_R 11:13:37

Spotlight A H 2015-10-22 258.0 25.5 H 332.0 -6.0 106.0 74_R 22:46:35

Spotligt A H 2015-10-25 260.0 40.5 H 80.0 -6.0 0.0 13_R 22:58:53

Wide Fine H+V 2015-11-08 259.0 30.5 V 79.0 -6.0 0.0 Quad 11_R 22:50:28

Fine Quad H+V 2015-11-09 98.5 48.0 V 80.0 -6.0 -161.5 29_R 11:01:09

DRDC-RDDC-2021-R021 25

Figure 18: Genie S-125 telescopic boom lift.

26 DRDC-RDDC-2021-R021

5.7 Bistatic Active Radar Calibrator

In addition to the C-band prototype ground-based bistatic receiver system, a Bistatic Active Radar Calibrator (BARC) has been deployed within the main acquisition footprint for the Carling Campus and 125’ boom lift acquisitions. The BARC was configured as a repeater, transmitting a copy of the received signal without delay. This effectively places an ideal point target at the BARC location in the imagery. The deployment parameters are provided in Table 6.

Table 6: Deployment parameters for the Bistatic Active Radar Calibrator.

MSL Bearing Elevation Bearing Elevation Date Location Altitude to RS-2 to RS-2 to Rx to Rx

2015-09-14 45 N 21.129ʹ 75 W 50.455ʹ 60 m 259.5 36.5 260.0 2.0

2015-09-21 45 N 20.540ʹ 75 W 50.584ʹ 65 m 259.0 31.0 325.5 1.5

2015-09-251 45 N 20.833ʹ 75 W 53.310ʹ 95 m 101.0 34.0 82.0 0

2015-10-051 45 N 20.833ʹ 75 W 53.310ʹ 95 m 257.5 20.0 82.0 0

2015-10-16 45 N 21.170ʹ 75 W 53.566ʹ 71 m 98.5 48.0 160.0 3.5

2015-10-19 45 N 20.897ʹ 75 W 53.549ʹ 80 m 100.5 35.0 119.5 7

2015-10-22 45 N 21.118ʹ 75 W 53.507ʹ 72 m 258.0 25.5 159.5 3.5

2015-10-25 45 N 20.880ʹ 75 W 53.103ʹ 73 m 260.0 40.5 249.0 6.5

2015-11-08 45 N 20.880ʹ 75 W 53.103ʹ 73 m 259.0 83.5 248.0 6.5

2015-11-09 45 N 20.880ʹ 75 W 53.103ʹ 73 m 98.5 48.0 247.5 6.5

1 The BARC was elevated using the CRC Boom Lift.

DRDC-RDDC-2021-R021 27

6 Analysis of collected data

The data are recorded in files that have the channels interleaved in blocks of 2 samples, each sample is stored in two bytes. For each data collect, direct path analysis is performed on a set of blocks that contain the peak pulse. 6.1 Direct path analysis

There are several characteristic parameters that are estimated by the processing algorithms described in Section 4.1.2. The measured characteristics for each of the fifteen data collections described in Sections 5.3–5.6 are presented in Table 7, Table 8, and Table 9.

The PRF is determined by dividing the sampling frequency by the PRI in samples. The bandwidth is determined by the difference between the frequency spectrum edges, while the spectrum centre is the mean value of those frequency spectrum edges (Figure 9). In baseband, the pulse interval having the largest peak magnitude is extracted as a replica source for match filter purposes. The pulse width is determined by the difference between the edges of the signal data region of the replica pulse interval (Figure 10). The phase shift is used for aligning the distance profile to the peak magnitude (Figure 11).

Additionally, Table 7–Table 9 list the number of pulses detected in the data file, and the block of samples in which the initial and final pulses can be found, as per Section 4.1.3.

Table 7: Measured characteristic parameters of bistatic SAR data.

Acquisition Date: 2015-03-02 2015-03-09 2015-03-30 2015-08-04 2015-08-05

PRF (Hz) 1353.42 1290.54 1356.90 1442.16 1383.93

BW (MHz) 12.34 12.48 12.66 29.20 30.73

Spectrum Centre (MHz) 60.51 60.52 60.50 59.90 59.85

Pulse Width (µs) 44.07 44.22 44.78 21.97 21.66

Phase Shift (Hz) 646 574 -676 654 -585

Available Polarizations 1 1 1 2 2

Number of Pulses 17087 16596 21611 11826 6930

Replica Pulse Number 4922 4725 4571 8618 2441

28 DRDC-RDDC-2021-R021

Acquisition Date: 2015-03-02 2015-03-09 2015-03-30 2015-08-04 2015-08-05

Start Block 8665 10368 5250 10559 11135

End Block 10170 11901 7148 11536 11732

Table 8: Measured characteristic parameters of bistatic SAR data.

Acquisition Date: 2015-09-14 2015-09-21 2015-09-25 2015-10-05 2015-10-16

PRF (Hz) 1300.08 1442.16 1336.28 1382.48 1580.81

BW (MHz) 29.60 29.28 12.08 29.75 100.54

Spectrum Centre (MHz) 59.85 59.88 60.48 59.88 53.60

Pulse Width (µs) 21.95 21.86 43.68 21.87 41.94

Phase Shift (Hz) 480 618 -529 571 -537

Available Polarizations 2 2 1 2 1

Number of Pulses 10627 11771 27683 6741 17688

Replica Pulse Number 7760 8581 18638 4698 9781

Start Block 18328 13811 13760 13801 13449

End Block 19302 14784 16230 16020 14783

Table 9: Measured characteristic parameters of bistatic SAR data.

Acquisition Date: 2015-10-19 2015-10-22 2015-10-25 2015-11-08 2015-11-09

PRF (Hz) 1142.96 1598.04 1598.04 1442.16 1383.93

BW (MHz) 53.40 100.86 100.52 30.07 29.05

DRDC-RDDC-2021-R021 29

Acquisition Date: 2015-10-19 2015-10-22 2015-10-25 2015-11-08 2015-11-09

Spectrum Centre (MHz) 61.54 53.59 53.60 59.80 59.88

Pulse Width (µs) 44.27 41.99 41.97 21.71 22.05

Phase Shift (Hz) -467 -480 -174 694 -474

Available Polarizations 1 1 1 2 2

Number of Pulses 7580 7448 8696 18789 6930

Replica Pulse Number 3748 741 4214 8551 2510

Start Block 13706 12294 13504 13539 13617

End Block 14497 12850 14152 15092 14214

6.2 Range resolution

In Equation 87 of [1], the nominal bistatic SAR range resolution is generated from a gradient method,

D = , (5) √ d

where is the speed of light, is the frequency bandwidth, and d is the difference in graze angles to the receiver and satellite from a point on the ground. For generality, this point is taken to be directly down range from the receiver in the scene centre of the receiver’s antenna.

The range resolution may also be estimated from the 3dB width in time, , of the range compressed direct path pulse, which measures the minimum time delay required for separate impulse responses to be differentiated. For a location down range of the receiver antenna, the additional distance travelled by the transmitted wave is + cos d, where is the separation distance in the slant plane of the receiver antenna. Thus, the range resolution can be calculated as

D = . (6) d

30 DRDC-RDDC-2021-R021

Table 10: Range resolution in the slant plane down range of the receiver antenna, compared with the nominal impulse response width of the monostatic slant range image product given in [23].

Acquisition Direct Path d Nominal Range Monostatic Slant Date T3dB Width Range Resolution Range Reference Resolution Resolution [23]

2015-03-02 72.0 ns 28.0 12.0 m 11.4 m 13.5 m

2015-03-09 71.2 ns 37.0 12.1 m 11.9 m 13.5 m

2015-03-30 71.2 ns 49.5 12.5 m 12.9 m 13.5 m

2015-08-04 28.8 ns 53.0 5.5 m 5.4 m 5.2 m

2015-08-05 28.8 ns 36.5 5.0 m 4.8 m 5.2 m

2015-09-14 28.8 ns 47.5 5.4 m 5.1 m 5.2 m

2015-09-21 28.8 ns 53.0 5.5 m 5.4 m 5.2 m

2015-09-25 70.4 ns 50.0 12.4 m 12.8 m 13.5 m

2015-10-05 28.8 ns 64.0 5.8 m 6.0 m 5.2 m

2015-10-16 9.6 ns 36.0 1.6 m 1.6 m 1.6 m

2015-10-19 18.0 ns 49.0 3.1 m 3.2 m 3.1 m

2015-10-22 9.2 ns 58.5 1.7 m 1.8 m 1.6 m

2015-10-25 9.2 ns 43.5 1.6 m 1.6 m 1.6 m

2015-11-08 28.8 ns 53.5 5.4 m 5.4 m 5.2 m

2015-11-09 28.8 ns 36.0 5.1 m 4.8 m 5.2 m

DRDC-RDDC-2021-R021 31

6.3 Azimuth resolution

The theoretical cross-range resolution for bistatic imaging geometries is based on the observation time of a location. The nominal bistatic cross-range resolution is derived in Section 4.3 of [1]. Specifically, Equation 86 of the reference provides that

l D = = , (7) q where is the antenna aperture length. For RS-2, the antenna aperture length is 15 m [22]. Furthermore, the bistatic cross-range resolution is twice the nominal cross-range resolution for a monostatic system. Indeed, for RS-2, [23] gives a resolution of 7.6 m or 7.7 m, except for Multi-look, Ultra-Fine, and Extra-Fine modes, which use half-apertures; the Spotlight modes, for which the main lobe direction is electronically steered; and the ScanSAR modes, where resolution is sacrificed for increased swath coverage.

To estimate the nominal cross-range resolution, the 3dB beamwidth, q, of the main lobe of the direct path pulse peaks is derived via

|⃗| q = , (8) where ⃗ is the velocity of the satellite and = min () is the minimum distance between transmitter and receiver. However, since the transmit antenna is electronically steered for Spotlight mode, there is no main lobe to measure and this method does not apply to Spotlight cases. The estimate for the 2015 bistatic data collections are tabulated in Table 11 and Table 12.

The azimuth resolution of the bistatic imagery may be confirmed using the presence of a calibration device. For data collections with a BARC in the scene, measurement of the 3dB width of the BARC is sufficient to estimate the resolution perpendicular to direction to the receiver antenna. For cross-range resolution, the 3dB width of the BARC is measured in the Range-Doppler image along the cross-range direction, using quadratic interpolation to obtain the -3 dB values. The cross-range pixel spacing was calculated as Rx^ = ⃗ ´ PRI. The results of these measurements for the 2015 scenes in which the BARC was deployed are tabulated in Table 12.

Table 11: Nominal cross-range resolution for DFL rooftop and CRC boom lift collections.

Acquisition Date Main lobe 3dB width l

2015-03-02 750.7 pulses 15.6 m

2015-03-09 681.3 pulses 15.5 m

32 DRDC-RDDC-2021-R021

Acquisition Date Main lobe 3dB width l

2015-03-30 607.2 pulses 15.6 m

2015-08-04 609.1 pulses 15.8 m

2015-08-05 732.3 pulses 15.5 m

Table 12: Nominal cross-range and BARC azimuth resolution derived from BARC measurement for Carling Campus and 125’ Boom Lift collections. The UltraFine, Wide MultiFine, and Spotlight modes are not subject to Equation (7).

Acquisition Main lobe l BARC b Pixel BARC ^ Date / Beam 3dB width q 3dB spacing Azimuth Mode width Resolution

2015-09-14 / 607.5 15.0 m 2.6 pixels 0.7 5.8 m 15.3 m Wide FineQuad pulses

2015-09-21 / 641.7 15.0 m 2.9 pixels 77.7 5.2 m 15.3 m Wide FineQuad pulses

2015-09-25 / 625.0 15.0 m 2.7 pixels 20.1 5.6 m 15.4 m Wide pulses

2015-10-05 / 567.8 15.1 m 2.7 pixels 175.5 5.5 m 14.9 m FineQuad pulses

2015-10-16 / 908.7 14.3 m 2.7 pixels 61.5 4.8 m 13.0 m UltraFine pulses

2015-10-19 / 551.8 14.5 m 2.4 pixels 19.0 6.6 m 15.9 m Wide MultiFine pulses

2015-10-22 / N/A N/A 2.9 pixels 98.5 4.7 m 13.5 m Spotlight

2015-10-25 / N/A N/A 3.1 pixels 11.0 4.7 m 14.4 m Spotlight

DRDC-RDDC-2021-R021 33

Acquisition Main lobe l BARC b Pixel BARC ^ Date / Beam 3dB width q 3dB spacing Azimuth Mode width Resolution

2015-11-08 / 641.6 15.0 m 2.9 pixels 11.0 5.2 m 15.1 m Wide FineQuad pulses

2015-11-09 / 761.3 14.9 m 3.4 pixels 149.0 5.5 m 18.6 m FineQuad pulses

6.4 Bistatic imagery analysis

The 2015 data collection constitutes examples of bistatic data covering a the variation of several significant aspects. Based on the bistatic angle, b, there are six quasi-monostatic geometries, two forward-scattering cases, and three cross-range looking examples. The remaining four collections, two with b = 11 and two with b = 19, are back-scattered, but the bistatic angles are too large to characterize as quasi-monostatic.

Figure 14 shows that range curves on the backprojection grid form closed ellipses around the receiver antenna location. Furthermore, the coordinate system is designed so that the Doppler profile is wholly vertical, i.e., each column of pixels has the same Doppler value for each pulse. As such, the pixels at which a vertical line intersects a given range ellipse will have identical Doppler and Range values (Figure 19[a]), and therefore be treated identically in the image formation process. In essence, the forward and back-scattered regions will mirror each other, at least they would if the transmitter were directly overhead and the range ellipses were circular. Because there is a non-zero incidence angle in the forward scattering direction, the mirror lines occur angled to the cross-range direction according the incidence angle, as seen in Figure 19(b).

When looking to the side, the illumination of the scene may occur before or after the receiver antenna is illuminated by the transmitter. Nevertheless, the scattered signal is received and processed. Indeed, since only pixels illuminated by the main lobe are populated for each pulse, we can observe this effect by partitioning the available pulses and generating a sequence of bistatic images generated from each interval of pulses. Figure 20 shows a times series of imagery for the 2015-10-16 collection constructed from a series of three nonoverlapping intervals containing 1273 pulses each. Illumination moves from left to right according to the motion of the satellite.

34 DRDC-RDDC-2021-R021

(a)

(b)

Figure 19: Forward-scattering bistatic VV imagery of 2015-10-05 having 20 incidence angle and 175 bistatic angle. (a) Response from a single pulse is distributed along iso-range ellipses. Iso-Doppler lines are vertical, intersecting the iso-range curves on either side of the mirror lines (white). (b) Scatterers above the mirror lines are real, those below are mirrored ambiguities.

DRDC-RDDC-2021-R021 35

(a)

(b)

(c)

Figure 20: Side-looking bistatic HH imagery of 2015-10-16 with 48 incidence angle and 61.5 bistatic angle. Non-overlapping intervals of pulses used as a time series showing illumination moving from left to right. The mirror line is shown in white.

36 DRDC-RDDC-2021-R021

7 Conclusion

The SBR group has leveraged its previous work proving the concepts of bistatic SAR imaging, leading to the development of a new R&D C-band bistatic SAR prototype system which has been successfully used to collect bistatic SAR data in fifteen field trials during 2015. The trials used RS-2 as the illuminator on a non-interference basis, i.e., there was no request or influence from the SBR group on the RS-2 operations that occurred.

The hardware for this initial system was largely leveraged from previous work, and used to give direction to the requirements, improvements, and design of future generations of DRDC bistatic receiver hardware. This has led to procurement of new equipment to construct the next generation R&D capability and increase the capability to include additional data channels. Concomitantly, the data analysis and image processing components of the system, also built upon algorithms leveraged from previous SBR work, has evolved in sophistication, analytical capability, and automation. This has well positioned the SBR group with a solid foundation of tools to expand the study of bistatic phenomenology under the Space-based Radar Exploitation project into fully multistatic SAR imagery, polarimetric and interferometric bistatic phenomenology, and the opportunity to use other illuminators.

Finally, the 2015 field trial experiments allowed for the investigation of deployment criteria and methods. In particular, the use of aerial lift platforms to elevate the receiver antennas showed great success in the ability to collect bistatic data without loss of image quality. The flexibility of using this equipment compared to fixed locations such as towers, rooftops, or topographical features of elevation opens up a far greater range of imaging geometries that can be easily explored.

DRDC-RDDC-2021-R021 37

References

[1] Gierull, C.H. (2004), Bistatic Synthetic Aperture Radar: TIF—Report (Phase I), DRDC Ottawa TR 2004-190, Defence R&D Canada – Ottawa, Technical Report.

[2] Gierull, C., Pike, C., and Paquet, F. (2006), Mitigation of Phase Noise in Bistatic SAR Systems with Extremely Large Synthetic Apertures, in Proceedings of EUSAR 2006, (Dresden, Germany, May, 2006).

[3] Goh, A.S., Preiss, M., Stacy, N.J.S., and Gray, D.A. (2008), Bistatic SAR experiment with the Ingara imaging radar: Preliminary results, in Proceedings of EUSAR 2008, pp. 49–52 (Friedrichshafen, Germany, June, 2008).

[4] Espeter, T., Walterscheid, I., Klare, J., Gierull, C., Brenner, A.R., Ender, J., and Loffeld, O. (2008), Progress of hybrid bistatic SAR: Synchronization experiments and first imaging results, in Proceedings of EUSAR 2008, pp. 217–220 (Friedrichshafen, Germany, June, 2008).

[5] Dubois-Fernandez, P., Cantalloube, H., Vaizan, B., Krieger, G., Horn, R., Wendler, M., and Giroux, V. (2006), ONERA-DLR bistatic SAR campaign: Planning, data acquisition, and first analysis of bistatic scattering behavior of natural and urban targets, IEE Proceedings–Radar, Sonar, and Navigation, 153 (3), pp. 214–223.

[6] Brenner, A.R., and Ender, J. (2006), Demonstration of advanced reconnaissance techniques with the airborne SAR/GMTI sensor PAMIR, IEE Proceedings–Radar, Sonar, and Navigation, 153 (2), pp. 152–162.

[7] Yates G., Horne, A.M., Blake, A.P., Middleton, R., and Andre, D.B. (2004), Bistatic SAR image formation, in Proceedings of EUSAR 2004, pp. 581–584 (Ulm, Germany, May, 2004).

[8] Baumgartner, S.V., Rodriguez-Cassola, M., Nottensteiner, A., Horn, R., Scheiber, R., Schwerdt, M., Steinbrecher, U., and Metzig, R. (2008), Bistatic experiment using TerraSAR-X and DLR’s new F-SAR system, in Proceedings of EUSAR 2008, pp. 57–60 (Friedrichshafen, Germany, June, 2008).

[9] Walterscheid, I., Espeter, T., Brenner, A.R., Klare, J., Ender, J.H.G., Nies, H., Wang, R., and Loffeld, O. (2010), Bistatic SAR Experiments With PAMIR and TerraSAR-X—Setup, Processing, and Image Results, IEEE Transactions on Geoscience and Remote Sensing, 48 (8), pp. 3268–3279.

[10] Massonet, D. (2001), Capabilities and limitations of the interferometric cart-wheel, IEEE Transactions on Geoscience and Remote Sensing, 39 (3), pp. 506–520.

[11] Krieger, G., Moreira, A., Fiedler, H., Hajnsek, I., Werner, M., Younis, M., and Zink, M. (2007), TanDEM-X: A Satellite Formation for High Resolution SAR Interferometry, IEEE Transactions on Geoscience and Remote Sensing, 45 (11), 3317–3341.

[12] Espeter, T., Walterscheid, I., Klare, J., Brenner, A.R., and Ender, J.H.G. (2011), Bistatic Forward-Looking SAR: Results of a Spaceborne-Airborne Experiment, IEEE Geoscience and Remote Sensing Letters, 8 (4), pp. 765–768.

38 DRDC-RDDC-2021-R021

[13] Espeter T., Walterscheid, I., Klare, J., and Ender, J. (2007), Synchronization techniques for the bistatic spaceborne/airborne SAR experiment with TerraSAR-X and PAMIR, in Proceedings of the IEEE IGARSS 2007, pp. 2160–2163 (Barcelona, Spain, July, 2007).

[14] Ender, J. (2003), Signal theoretical aspects of bistatic SAR, in Proceedings of the IEEE IGARSS 2003, pp. 1438–1441 (Toulouse, France, July, 2003).

[15] Bamler, R., Meyer, F., and Liebhart, W. (2007), Processing of bistatic SAR data from quasi-stationary configurations, IEEE Transactions on Geoscience and Remote Sensing, 45 (11), pp. 3350–3358.

[16] Neo, Y.L., Wong, F., and Cumming, I. (2007), A two-dimensional spectrum for bistatic SAR processing using series reversion, IEEE Geoscience and Remote Sensing Letters, 4 (1), pp. 93–96.

[17] Wong, F.H., Cumming, I.G., and Neo, Y.L. (2008), Focusing bistatic SAR data using the nonlinear chirp scaling algorithm, IEEE Transactions on Geoscience and Remote Sensing, 46 (9), pp. 2493–2505.

[18] Stockli, R. (2005), “Blue Marble: next generation,” NASA Goddard Space Flight Center. Online: modified 2005-10-13. URL: http://earthobservatory.nasa.gov/Features/BlueMarble/, accessed 2016-08-17.

[19] Livingstone, C., and Beaulne, P. (2012), Modelling RADARSAT-2 Doppler centroids for marine applications, DRDC Ottawa TM 2012-097, Defence R&D Canada – Ottawa, Technical Memorandum.

[20] Çotuk, N., and Gierull, C. (2004), Time-Domain Bistatic SAR Processor, DRDC Ottawa TM 2004-191, Defence R&D Canada – Ottawa, Technical Memorandum.

[21] CSA Order Desk (2013). Pre-Deconflicting your RADARSAT-2 APT acquisitions plans through EMOC: Overview of process and timelines, in Enhanced Management of Orders and Conflicts (EMOC) Workshop. Canadian Space Agency (Ottawa, April 24, 2013).

[22] Directorate of Communications (2011), “Satellite Characteristics,” Online: modified 2015-10-07, Canadian Space Agency. URL: http://www.asc-csa.gc.ca/eng/satellites/radarsat/radarsat- tableau.asp, accessed 2015-12-11.

[23] MacDonald Dettwiler and Associates Ltd. (2016), RADARSAT-2 Product Description, Technical report RN-SP-52-1238, issue 1/13, MacDonald Dettwiler and Associates Ltd., Richmond, Canada.

DRDC-RDDC-2021-R021 39

Annex A Step function detector

Section 2.3 introduces the concept of signal detection through the comparison of the central trendancy (mean value) within signal and noise only regions. Signals that have a nearly constant mean magnitude values are well-characterized using this method, however signals that exhibit a ramp to their central tendency may be poorly conditioned for this method to be successful. In particular, the magnitude values must support a central tendency that can be modelled by a Heaviside function, with each side remaining above or below the overall mean value across the entire interval.

A.1 Demarcation function

When a set of data meets this criteria, we are able to build a demarcation function to test if each point is a good demarcation point between signal and noise regions. Let { | ∈ [1, ]} be a set of consecutively sampled complex data. The mean value of the magnitude, , of this set is given by

1 = | |. (A.9)

We define the left and right central tendency for a given sample number, , to be

1 () = | |, (A.10) 1 () = | |. −

Note that () = (0) = . Using these, we define an initial demarcation function to be

()= () − () ( − ) ∑ | | − ∑ | | = ( − ) (A.11) ∑ | | − () = . ( − )

However, as → 0 and → , Equation (A.11) becomes unbounded, which is ill-conditioned at these edges. To address this, consider embedding the data set into a larger one of length = 3, whereby

40 DRDC-RDDC-2021-R021

, for ≤ , = , for < ≤ 2, (A.12) , for > 2.

Over this new superset, we define for = + , where ∈ [1, ] as before:

(′) ′(∑ | | − ′) = g ′( − ) 3( + ∑ | | − ( + )) = g (A.13) ( + )3 − ( + ) 3(∑ | | − ) = g , ( + )(2 − )

where g is a scaling factor to be determined. Observing that ( + ⁄2) = (g⁄3)(⁄2), we set g = 3, and define the demarcation function to be

9(∑ | | − ) () = ( + ) = , (A.14) ( + )(2 − ) which remains well-conditioned across the interval ∈ [1, ].

A.2 Demarcation example

In Figure A.1, the magnitude of recorded data from a direct path pulse interval is shown in blue. The interval is 13445 samples in length, with a clear demarcation between the signal region on the left for sample numbers ∈ [1,5378], and the noise only region for sample numbers above that. The mean value of the full interval is at 182.2 on this voltage scale, as shown by the black line. The central tendency of the signal region is 448.6, and 4.5 for the noise only region, shown in red. Equation (A.14) is used to generate the demarcation function over the interval, shown in green.

To the left of the demarcation point ( = 5378), the demarcation function has a positive slope despite the presence of sample values whose magnitude is below the mean value. The function peaks at the demarcation point, after which a negative slope exists across the noise only region. The location of this peak or, more correctly, the switch to the negative slope is a robust indicator of the demarcation point between the regions that have a measurable difference in central tendency values.

DRDC-RDDC-2021-R021 41

Figure A.1: Magnitude of pulse interval (blue) showing signal region to the left and noise only region to the right of the demarcation point. The central tendacies of these regions are modelled by a discrete Heaviside function (red). The mean value (black) of the pulse interval along with the data itself is used to generate the demarcation function (green).

42 DRDC-RDDC-2021-R021

List of symbols/abbreviations/acronyms/initialisms

AFRL Air Force Research Laboratory (U.S.) AGL above ground level APT acquisition planning tool BARC Bistatic Active Radar Calibrator CRC Communications Research Centre (Canada) CSA Canadian Space Agency DFL David Florida Laboratory (CSA) DND Department of National Defence DRDC Defence Research and Development Canada DSTKIM Director Science and Technology Knowledge and Information Management ECEF earth-centered, earth-fixed ECI earth-centered inertial EMOC Enhanced Management of Orders and Conflicts GoC Government of Canada H horizontal (polarization) HAE height above the ellipsoid IF intermediate frequency LO local oscillator MDA MacDonald Dettwiler and Associates (Ltd.) MSL mean sea level PRI pulse repetition interval RS-1 RADARSAT-1 RS-2 RADARSAT-2 Rx receiver SAR synthetic aperture radar SBR space based radar SNR signal to noise ratio TIF Technology Investment Fund TN true north TRDP Technical Research and Development Program

DRDC-RDDC-2021-R021 43

TRR Transportable Radar Receiver TTRDP Trilateral Technical Research and Development Program Tx transmitter V vertical (polarization)

44 DRDC-RDDC-2021-R021

DOCUMENT CONTROL DATA *Security markings for the title, authors, abstract and keywords must be entered when the document is sensitive 1. ORIGINATOR (Name and address of the organization preparing the document. 2a. SECURITY MARKING A DRDC Centre sponsoring a contractor's report, or tasking agency, is entered (Overall security marking of the document including in Section 8.) special supplemental markings if applicable.)

DRDC – Ottawa Research Centre CAN UNCLASSIFIED Defence Research and Development Canada, Shirley's Bay 3701 Carling Avenue 2b. CONTROLLED GOODS Ottawa, Ontario K1A 0Z4 NON-CONTROLLED GOODS Canada DMC A

3. TITLE (The document title and sub-title as indicated on the title page.)

Foundations for a multistatic synthetic aperture radar (SAR) imaging capability: Development of a C‑Band prototype receiver and processor

4. AUTHORS (Last name, followed by initials – ranks, titles, etc., not to be used)

English, R. A.; Gierull, C. H.

5. DATE OF PUBLICATION 6a. NO. OF PAGES 6b. NO. OF REFS (Month and year of publication of document.) (Total pages, including (Total references cited.) Annexes, excluding DCD, covering and verso pages.) February 2021 53 23

7. DOCUMENT CATEGORY (e.g., Scientific Report, Contract Report, Scientific Letter.)

Scientific Report

8. SPONSORING CENTRE (The name and address of the department project office or laboratory sponsoring the research and development.)

DRDC – Ottawa Research Centre Defence Research and Development Canada, Shirley's Bay 3701 Carling Avenue Ottawa, Ontario K1A 0Z4 Canada

9a. PROJECT OR GRANT NO. (If appropriate, the applicable 9b. CONTRACT NO. (If appropriate, the applicable number under research and development project or grant number under which which the document was written.) the document was written. Please specify whether project or grant.)

05ea

10a. DRDC PUBLICATION NUMBER (The official document number 10b. OTHER DOCUMENT NO(s). (Any other numbers which may be by which the document is identified by the originating assigned this document either by the originator or by the sponsor.) activity. This number must be unique to this document.)

DRDC-RDDC-2021-R021

11a. FUTURE DISTRIBUTION WITHIN CANADA (Approval for further dissemination of the document. Security classification must also be considered.)

Public release

11b. FUTURE DISTRIBUTION OUTSIDE CANADA (Approval for further dissemination of the document. Security classification must also be considered.)

12. KEYWORDS, DESCRIPTORS or IDENTIFIERS (Use semi-colon as a delimiter.)

synthetic aperture radar; SAR; bistatic SAR; RADARSAT-2; signal processing; image processing

13. ABSTRACT (When available in the document, the French version of the abstract must be included here.)

DRDC has developed a new R&D ground-based C-band bistatic SAR receiver system and collected fifteen bistatic data sets using RADARSAT-2 as the illuminator. The collections were made on a non-interference basis from existing acquisitions appearing in the Government of Canada’s monthly Enhanced Management of Orders and Conflicts plans. In addition to rooftop deployment of the receiver antennas, aerial boom lifts were used to experiment with more flexible deployment options. Several different bistatic angles were acquired during the campaign. Analysis of the direct path received signal from the satellite at each receiver site was performed to extract relevant parameter values. The results were applied to range compression and back projection of the signal reflected from the scene to the receiver site, yielding good quality SAR imagery in all fifteen cases. Basic phenomenology of these bistatic scenes are under study, and the system is validated and baselined for expanding DRDC’s research into fully multistatic SAR imagery, polarimetric and interferometric bistatic phenomenology, and the opportunity to use other illuminators.

RDDC a développé un nouveau système R et D de récepteur au sol de radar à synthèse d’ouverture (SAR) bistatique en bande C et a recueilli quinze ensembles de données bistatiques en utilisant RADARSAT-2 comme illuminateur. Les collectes ont été réalisées sans interférence à partir des acquisitions de données actuelles répertoriées dans les plans mensuels de gestion améliorée des commandes et des conflits du gouvernement du Canada. En plus du déploiement sur toiture des antennes du récepteur, des élévateurs à nacelle ont servi à explorer diverses options de déploiement plus souples. Divers angles bistatiques ont été acquis dans le cadre de la campagne. De plus, une analyse du signal du satellite reçu en trajet direct a été réalisée à chaque emplacement de récepteur pour en extraire les valeurs des paramètres pertinents. Les résultats ont été appliqués à la compression en portée et à la projection par transparence du signal réfléchi de la scène vers l’emplacement du récepteur, ce qui a produit des images SAR de bonne qualité dans les quinze cas. Par ailleurs, la phénoménologie de base des scènes bistatiques est à l’étude, et le système a été validé et établi comme base de référence pour étendre la recherche de RDDC à une phénoménologie bistatique polarimétrique et interférométrique de l’imagerie SAR entièrement multistatique, et créer une occasion d’utiliser d’autres illuminateurs.