Closing the Loop: Engineering Better CCD into the Lynx SAR

 

Defence Technology Review · Radar Systems Engineering · April 2026

Airborne SAR Coherent Change Detection Navigation Systems Signal Processing UAS Sensors ISR

Lynx SAR · CCD Performance · Systems Engineering

A systematic examination of the navigation, processing, and flight management improvements that would transform the Lynx multimode radar from a capable but inconsistent coherent change detector into a reliable operational tool — and what the current state of the art makes possible that was not available when Lynx was first fielded.

Technical Analysis · Based on Declassified Sandia, DSTO, NASA/JPL Research · April 2, 2026

Bottom Line Up Front: The Lynx SAR/GMTI radar has fielded coherent change detection (CCD) capability since its original Sandia design, but operational experience consistently produced CCD pairs that showed nothing useful. The root causes form a well-understood hierarchy: squint angle inconsistency between passes driven by the LN-200 IMU's attitude accuracy limitations; registration errors that persist even with 3 cm position accuracy; autofocus inconsistencies between passes; and temporal environmental decorrelation that cannot be controlled but can be managed operationally. All of the hardware-controllable failure modes now have engineering solutions available that did not exist when Lynx was designed. A phased upgrade — navigation-grade IMU, multi-antenna GPS attitude, precision autopilot track-keeping, improved post-processing registration, and operational protocol discipline — could elevate Lynx CCD from occasional capability to reliable operational mode.

The Lynx multimode SAR/GMTI radar is, by almost any measure, a remarkable engineering achievement. Weighing less than 120 pounds, capable of 0.1 meter spotlight resolution and coherent change detection from a UAV platform, it was the first system of its kind to demonstrate operational CCD from an unmanned aircraft and accumulate thousands of mission hours in deployed combat operations. Yet engineers who worked closely with the system during its development and operational deployment will confirm a frustrating truth that never appeared in press releases: a significant fraction of CCD image pairs produced nothing interpretable. The coherence background was flat, the contrast between changed and unchanged regions was absent, and the images were indistinguishable from noise.

Understanding why this happened — and what can now be done about it — requires working through the physics of what coherent change detection actually demands from a sensor system, and then mapping those demands honestly against what the Lynx navigation subsystem, flight management, and processing chain were capable of delivering. That mapping exercise, informed by foundational work at Sandia National Laboratories, the Australian Defence Science and Technology Organisation (DSTO), and NASA's Jet Propulsion Laboratory, reveals a hierarchy of fixable problems and a clear engineering path forward.

Why CCD Is So Much Harder Than SAR Imaging

The gap between "good SAR image quality" and "good CCD quality" is wider than most people outside the field appreciate, and understanding it is the starting point for any upgrade analysis. SAR image formation requires that the motion compensation system track the radar's position accurately enough during a single synthetic aperture dwell — typically a few seconds — to focus the image correctly. The LN-200 IMU with carrier-phase GPS integration achieves this admirably: position accuracy to 3 cm, accelerometer quality sufficient to compensate vibration and maneuver-induced phase errors, and short-term attitude accuracy adequate for beam pointing. These are the requirements the LN-200 was designed and selected to meet, and it meets them.

CCD imposes an entirely different and much more stringent set of constraints. Rather than requiring accurate navigation during a single collection, it requires that two separate collections — potentially hours or days apart, in different wind conditions, at slightly different altitudes, with different atmospheric states — produce images whose complex-valued pixels are co-registered to sub-pixel accuracy and whose imaging geometries match to a fraction of a degree. The coherence calculation then compares these two images pixel by pixel, and any source of decorrelation that is not caused by an actual physical change in the scene manifests as false change. There are many such sources, and they compound multiplicatively: the total coherence of the background is the product of every individual coherence term, so a single term that is poor drives the whole product toward zero regardless of how well the others are controlled.

Armin Doerry of Sandia, who was centrally involved in the original Lynx development, has written what remains the most comprehensive public treatment of what it actually takes to get good CCD results. His core observation deserves to be quoted directly and taken seriously: the coherence calculation itself is the easy part. The hard part is ensuring the two input images have the underlying characteristics to yield a quality result. Everything that follows in a CCD upgrade analysis flows from that observation.

"The coherence calculation is the easy part. The hard part is making sure the two input images have the underlying characteristics to yield a quality result." — A. W. Doerry, Sandia National Laboratories, "SAR Data Collection and Processing Requirements for High Quality Coherent Change Detection," SAND2008-0911C

Improvement 1: Replace the LN-200 with a Navigation-Grade IMU

The LN-200 fiber-optic gyro IMU has been Lynx's motion compensation sensor since the original Sandia design. It is available in 0.5 and 1.0 deg/hr gyro bias classes, and it has performed its primary function — SAR image formation motion compensation — reliably across more than 40,000 units produced. For that function, it remains a sound choice. For CCD, it is the single most improvable hardware element in the system.

The limitation is not the position accuracy achievable with carrier-phase GPS integration, which can be driven to 3 cm with a well-designed EKF. The limitation is attitude accuracy, specifically heading, which determines the squint angle of the synthetic aperture relative to the aircraft's ground track. Squint angle consistency between passes is the most operationally sensitive geometry parameter for CCD. Doerry's framework establishes that even 2 degrees of squint angle difference between passes causes noticeable CCD degradation for any scene with topographic relief, and anecdotal evidence from operational experience confirms this threshold is frequently violated when flying in variable crosswinds without explicit squint angle control.

The LN-200's heading accuracy during straight and level flight — the collection geometry most relevant to CCD — is poor relative to its position accuracy for a fundamental reason: heading is weakly observable from a single GPS antenna during non-maneuvering flight, and the EKF's ability to correct heading errors depends on aircraft dynamics that are absent during the steady passes where heading accuracy matters most. A 0.5 deg/hr gyro bias accumulates meaningful heading error over a typical synthetic aperture dwell and inter-pass transit interval even with frequent GPS position updates, because heading is corrected indirectly through platform dynamics rather than directly from the measurements.

The current market offers two complementary solutions to this problem that were not available when Lynx was designed. The first is navigation-grade miniaturized FOG IMUs. The iXblue/Exail UmiX series, for example, achieves 0.03 deg/hr in-run gyro bias and 0.008 deg/√hr angular random walk in an 89 mm diameter, 75 mm tall package drawing 4 watts — dimensions and power budget compatible with integration into the Lynx electronics assembly. That 0.03 deg/hr figure represents roughly a 15-fold improvement in gyro bias performance over the LN-200's best class, and it directly improves heading accuracy during straight and level collection by the same factor. Northrop Grumman's LR-500 Quad Mass Gyro IMU is another candidate, offering FOG/RLG-class performance in a MEMS package with the same digital interface as the LN-200, potentially enabling a near drop-in mechanical replacement.

The second solution, which addresses the heading observability problem directly rather than just improving the IMU's open-loop accuracy, is multi-antenna GPS attitude determination. By placing two or more GPS antennas at known separations on the Predator B or Gray Eagle airframe and using moving-baseline carrier-phase RTK techniques, heading can be observed directly from the GPS geometry at accuracies of a few hundredths of a degree, continuously, regardless of aircraft dynamics. This is the approach JPL developed for the UAVSAR program, where it proved essential for achieving the track repeatability required for repeat-pass interferometry. The heading accuracy achievable from a dual-antenna baseline of 2 meters at carrier-phase RTK is roughly 0.03 degrees RMS — sufficient to close the squint angle budget for CCD at 0.1 m Lynx spotlight resolution across essentially all terrain types.

The combination — navigation-grade FOG IMU for short-term dynamics and attitude propagation, plus dual-antenna GPS for continuous absolute heading aiding — addresses both the open-loop IMU drift problem and the observability problem simultaneously. The two solutions are also architecturally synergistic: the navigation-grade IMU provides the high-rate, low-noise attitude propagation that the dual-antenna GPS heading needs between update cycles, while the GPS heading prevents the IMU attitude errors from accumulating during the long straight collection passes where single-antenna GPS provides no heading correction.

LN-200 vs. Current Navigation-Grade Alternatives

• LN-200 (Northrop Grumman): 0.5–1.0 deg/hr gyro bias, 0.05–0.1 deg/√hr ARW. Fielded on Lynx since original design. Excellent for SAR motion compensation; limiting for CCD squint angle consistency.
• iXblue/Exail UmiX U9 (military grade): 0.03 deg/hr in-run bias, 0.008 deg/√hr ARW. Navigation-grade FOG in 89 × 75 mm package, 4W, ITAR-free. Approximately 15× gyro bias improvement over LN-200. Suitable for direct integration.
• Northrop Grumman LR-500 QMG: FOG/RLG-class MEMS performance, LN-200-compatible digital interface. Potential near-drop-in replacement at the circuit card level.
• Honeywell HG3900: All-silicon MEMS near-navigation-grade performance meeting or exceeding FOG/RLG systems. Design verification 2026, production 2027. Future option for next-generation Lynx variant.
• Dual-antenna GPS attitude (e.g. NovAtel OEM7720 + moving baseline RTK): Direct heading observation to 0.02–0.05 deg RMS from 2 m baseline. Addresses observability limitation of single-antenna EKF. Complements rather than replaces IMU upgrade.

Improvement 2: Precision Autopilot Track-Keeping

Navigation accuracy addresses the sensor's knowledge of where the aircraft was during each collection pass. Track-keeping accuracy addresses whether the aircraft actually flew the desired path in the first place. These are related but distinct requirements, and the distinction matters operationally because an excellent navigation system that accurately records a poor flight path produces exactly the same CCD degradation as a poor navigation system recording a good flight path.

The most instructive precedent for solving the track-keeping problem for airborne SAR CCD is NASA's UAVSAR program. UAVSAR was designed from the outset for repeat-pass interferometry — functionally identical to CCD in its track requirements — and its team discovered early that manual piloting, even with real-time GPS position display, could not reliably hold an aircraft within the required track tube. Their solution was the Platform Precision Autopilot (PPA), a GPS-guided flight management overlay that provided direct closed-loop control of cross-track and altitude deviations relative to a stored reference track. In operational science missions, the PPA controlled the Gulfstream III to within ±2.5 meters of the desired track more than 90 percent of the time over 342 data collection runs — substantially better than its 5-meter specification.

The Predator B and Gray Eagle platforms that host Lynx already have highly capable autopilots — far more capable than the Gulfstream III's baseline flight control. The architecture needed for CCD-quality track-keeping is not a new autopilot but rather a CCD mission mode in the flight management software: a mode that, when activated, retrieves the stored GPS track from the reference pass, computes real-time deviations in cross-track position and altitude, and feeds correction commands to the existing autopilot to drive those deviations below the CCD geometry tolerance. This is primarily a software and integration task, not a hardware replacement, and it builds on the precision navigation solution described above — the better the navigation accuracy, the finer the achievable track control.

The squint angle component of track-keeping is more subtle and requires explicit attention. Even if the GPS ground track is reproduced precisely, a different crosswind between passes produces a different crab angle and therefore a different squint angle for the synthetic aperture. There are two mitigation approaches. The first is operational: schedule CCD pairs for similar wind conditions, accept squint angle variation as an environmental constraint, and rely on post-processing aperture trimming to compensate residual squint differences. The second is hardware: use the antenna gimbal's existing azimuth drive capability — Lynx's gimbal already steers in azimuth for GMTI and spotlight modes — to actively compensate squint angle during collection by rotating the antenna to maintain a consistent bearing to the scene center regardless of aircraft crab angle. This converts a wind-dependent squint variation into a controlled, compensated geometry that is reproducible across passes. It would require a software mode that computes the required squint compensation from real-time wind measurement or estimated crab angle and commands the gimbal accordingly during the collection window.

NASA UAVSAR Precision Autopilot: The Relevant Precedent

The UAVSAR Platform Precision Autopilot (PPA) was developed by NASA Dryden Flight Research Center specifically to enable repeat-pass interferometric SAR. Its design principle — real-time GPS-guided closed-loop correction of deviations from a stored reference track — is directly applicable to a Lynx CCD mission mode. Key validated performance figures:

• Requirement: within a 5-meter radius tube over a 200 km course, calm to light turbulence, 90% of time
• Achieved: ±2.5 meters in altitude and cross-track, more than 90% of time, across 342 data runs over 20 flights
• Implementation: autopilot interface computer + inertial navigation system overlay; no new primary flight control hardware
• Key finding: "Pilots are unable to fly the aircraft with the desired accuracy; an augmented autopilot will be required to meet these objectives"

The Predator B and Gray Eagle autopilot quality substantially exceeds the Gulfstream III baseline used for UAVSAR testing, suggesting that equivalent or better track-keeping is achievable with a GPS-guided CCD mission mode overlay.

Improvement 3: Post-Processing Registration and Autofocus

Even with improved navigation and track-keeping, image-domain registration remains a required and non-trivial processing step. Doerry is emphatic on this point: virtually no SAR data collected with even high-quality motion measurement is good enough to avoid the need for registration corrections. The navigation and track-keeping improvements reduce the magnitude of the residual misregistration that registration must correct, but they do not eliminate the need for it.

The Lynx CCD processing chain performs registration as part of its standard CCD workflow, but operational experience suggests several improvements would materially improve the success rate of the registration step. The most impactful is iterative autofocus prior to registration. Both images in a CCD pair should be individually autofocused to the tightest achievable focus quality before the registration comparison is attempted. Doerry notes anecdotally that many CCD results can be improved significantly simply by reiterating autofocus operations on both images as the very first processing step. The reason is fundamental: registration cross-correlation works best when the image pair has the sharpest and most consistent point spread functions, because registration errors appear as sub-pixel offsets in the cross-correlation peak. Any focus mismatch between the two images — which can occur when the autofocus algorithm converges differently on the two passes due to different scene content or signal conditions — manifests as a registration error that is indistinguishable from a true geometric offset.

Beyond autofocus, several registration algorithm improvements are relevant. Aperture trimming — restricting both images to the common overlapping Fourier support defined by the actual imaging geometry of each pass — can be computed analytically from the navigation logs when navigation quality is high. With the improved navigation solution described above, this computation becomes reliable and should be a standard pre-processing step rather than an optional one. Spatially variant warping using dense tie-point grids, rather than simple global polynomial corrections, is important for scenes with topographic relief, where the layover geometry differences between passes are spatially non-uniform. Modern GPU-accelerated processing makes dense tie-point registration computationally practical for real-time or near-real-time ground processing that was infeasible when Lynx was first fielded.

A further processing improvement that draws on the better navigation data is automatic CCD quality prediction before the coherence calculation is even performed. Given two sets of navigation logs, the expected baseline decorrelation, squint angle mismatch, and registration residual can be computed analytically and used to predict whether a given pair is likely to yield useful CCD output. Pairs that fall outside the decorrelation budget can be flagged for the analyst before processing, avoiding the interpretation effort wasted on inherently unproductive pairs and providing diagnostic feedback to improve future collection planning. This prediction capability is straightforward to implement once reliable attitude logs are available — a capability that the improved navigation system directly enables.

Improvement 4: Operational Protocol and Mission Planning

Not all CCD improvement requires new hardware or software. A significant fraction of blank CCD pairs in operational Lynx employment resulted from collection conditions that no sensor improvement can overcome: excessive wind between passes decorrelating vegetation; inter-pass intervals too long for the terrain type; and collection geometries that deviated too far from the reference pass due to airspace constraints or mission priorities. Addressing these requires operational discipline and mission planning protocols that treat CCD as a precision measurement mode rather than a standard imaging mode.

The most actionable protocol changes flow directly from the physics. Inter-pass time interval should be specified as a function of terrain type and expected wind: minutes for vegetated terrain in calm conditions, hours acceptable for bare soil or gravel, essentially unlimited for paved roads or concrete. This is not a hard engineering requirement but an operational guideline that dramatically improves the probability of a productive CCD pair. CCD collection passes should be scheduled for similar wind conditions where possible — examining forecast wind profiles at collection altitude and planned collection time — because wind change between passes is the dominant uncontrollable decorrelation source for vegetated terrain at X-band.

Collection geometry should be treated as a mission parameter to be planned and recorded, not inferred after the fact. The reference pass geometry — altitude, ground speed, heading, squint angle, range to scene center — should be stored in the mission planning system and used to generate precise track and geometry guidance for the repeat pass. Currently this information is often reconstructed from navigation logs after the fact; making it available in the planning and execution chain before the repeat pass is flown directly improves track-keeping quality by giving the autopilot a precise target rather than a general waypoint route.

Finally, terrain classification should inform CCD mode selection. A scene over flat, low-reflectivity bare soil in dry conditions may support CCD with relatively relaxed geometry control; the same scene over urban terrain with buildings, or over forested hills with significant relief, requires tighter geometry matching and carries higher risk of registration failure due to layover geometry differences. A terrain-aware CCD mode selector in the ground control station, informed by a digital elevation model of the target area, could automatically compute the expected squint angle tolerance and flag collection geometries that fall outside it before the pass is flown.

Improvement 5: SNR Margin for Difficult Terrain

One decorrelation source that is independent of navigation quality or processing sophistication is SNR. CCD measures coherence between two images, and in regions where the scene backscatter is close to the noise floor, the independent noise contributions to the two images register as apparent change. The Sandia framework specifies that the scene clutter SNR should be at least 10 dB above the noise floor for useful CCD; below this threshold, coherence of the unchanged background degrades independently of any physical scene change, reducing the contrast available for change detection.

For Lynx operating at maximum specified standoff ranges against low-reflectivity terrain — dry sandy soil, calm water surfaces at low grazing angle, smooth paved roads — the 10 dB clutter SNR margin is not always available. This is less an argument for increasing transmit power (Lynx's 320 W TWTA is already near the practical limit for its size class) than for carefully managing collection geometry to favor higher-SNR viewing angles and closer standoff ranges for CCD-priority missions. It is also an argument for the CLAW software to automatically compute expected clutter SNR as part of CCD mission planning and flag scenarios where the margin is likely to be inadequate.

A longer-term hardware path to improved SNR is the transition to active electronically scanned array (AESA) antenna architecture. Lynx uses a mechanically steered antenna. An AESA would offer improved transmit power efficiency through distributed T/R modules, better sidelobe control for clutter suppression, and the possibility of simultaneous multi-beam operation for GMTI/CCD interleaving without the mechanical slew time penalty. This is not a near-term upgrade — AESA integration into the Lynx form factor would be a substantial development program — but it represents the logical direction for a Block 40 or next-generation Lynx to take the CCD capability to its physical limits.

A Phased Upgrade Roadmap

The improvements described above span a range of cost, risk, and development effort. Not all are equally urgent or equally achievable within Lynx's current form factor and program constraints. A phased approach, ordered by impact per unit of integration cost, would look approximately as follows.

Phase 1 — Navigation and processing (highest impact, moderate cost): Replace the LN-200 with a navigation-grade FOG IMU such as the iXblue/Exail UmiX U9 or Northrop Grumman LR-500. Add a dual-antenna GPS attitude solution using a moving-baseline RTK receiver integrated into the existing navigation EKF. Upgrade ground processing to implement iterative autofocus prior to registration, analytical aperture trimming from navigation logs, and automatic CCD quality prediction from navigation data. These changes require no modification to the antenna, gimbal, transmitter, or flight vehicle, and they directly address the dominant failure modes identified in operational experience.

Phase 2 — Flight management (moderate impact, integration effort): Implement a CCD precision track-keeping mode in the Predator B and Gray Eagle flight management software that retrieves the stored reference pass GPS track and provides real-time cross-track and altitude correction guidance. Add active gimbal-based squint angle compensation using real-time crab angle estimation from the navigation solution. Develop terrain-aware CCD collection planning tools in CLAW that predict baseline decorrelation and flag out-of-tolerance collection geometries before the pass is flown. These changes involve software development and integration testing but no new hardware beyond the navigation improvements in Phase 1.

Phase 3 — Longer-term architecture (transformational impact, significant development): Investigate AESA antenna architecture for a next-generation Lynx variant, enabling distributed transmit power, improved SNR margin, and simultaneous beam operations. Integrate AI-based change detection algorithms — such as the fusion-based approaches now emerging in the academic literature — directly into the onboard processor to supplement and eventually replace threshold-based amplitude change detection for small object detection. Explore multi-pass CCD stack processing, which uses more than two passes to improve coherence estimation and reduce false alarm rates, enabled by the improved track repeatability of Phases 1 and 2.

Requirement	Parameter	Current Lynx	Improved Target	Type
Nav — Gyro bias	In-run gyro bias stability	0.5 deg/hr (LN-200)	≤ 0.03 deg/hr	Hard
Nav — Heading	Squint angle repeatability, flat terrain	~1–3° (EKF limited)	≤ 0.5° RMS	Hard
Nav — Heading	Squint angle repeatability, terrain with relief	Not specified	≤ 0.2° RMS	Hard
Nav — Position	Cross-track position accuracy	~3 cm (current EKF)	≤ 3 cm (maintain)	Hard
Track-keeping	Cross-track deviation from reference pass	Not controlled	≤ 5 m RMS, 90% of time	Hard
Registration	Residual pixel misregistration after correction	Not specified	≤ 0.1 pixel (range & az)	Hard
SNR	Clutter SNR above noise floor	≥ 10 dB (at specified range)	≥ 10 dB (maintain)	Hard
Temporal	Max inter-pass interval, vegetated terrain	Not controlled	≤ 30 min (guideline)	Probabilistic
Temporal	Max inter-pass interval, bare soil / gravel	Not controlled	≤ 4 hr (guideline)	Probabilistic
Environmental	Max wind change between passes, vegetated	Not monitored	≤ 5 kn change (guideline)	Probabilistic
Processing	Autofocus iterations before registration	Single pass	Iterative to convergence	Hard
Processing	Aperture trimming from nav logs	Not implemented	Standard pre-processing	Hard

What Success Looks Like

The question of what constitutes a meaningful improvement in CCD performance deserves a concrete answer. In the current Lynx operational context, a CCD pair is considered successful if it produces a coherence image with sufficient background coherence and changed-region contrast that an analyst can identify true changes with acceptable false alarm rates. The blank-pair failure rate — pairs that produce no interpretable output — is the primary metric of failure, and reducing it is the primary metric of improvement.

A reasonable target for an upgraded Lynx CCD capability, informed by the DSTO Ingara results and the NASA UAVSAR precedent, is a blank-pair rate below 20% in conditions of calm to light wind over terrain with moderate relief, with CCD-quality output achievable on vegetated terrain when the inter-pass interval is below 30 minutes. Under these conditions, the DSTO Ingara team was able to detect fresh footprints on a grassy field — a result that represents the physical lower bound of what is achievable with X-band CCD under optimal conditions. That is the benchmark toward which an improved Lynx should aspire.

The Australian work also demonstrated something operationally important: when the sensor system and collection conditions are right, CCD at X-band on a UAV platform is not merely adequate — it is extraordinary. The ability to detect a person's footsteps in a field from an aircraft kilometers away is, for counter-IED operations, border surveillance, and force protection applications, a capability with no substitute at any price in the optical imaging domain. That is the capability the original Lynx designers were reaching for, and the one that a phased upgrade program could reliably deliver.

"The ability to detect a person's footsteps in a field from an aircraft kilometers away is a capability with no substitute at any price in the optical domain. That is the capability the original Lynx designers were reaching for."

The Broader Lesson for UAS SAR Design

The Lynx CCD story carries a lesson that extends well beyond this specific system. When SAR change detection capability is included in a sensor specification, the requirements are almost always written in terms of image formation quality — resolution, dynamic range, SNR at specified range — rather than in terms of the inter-pass geometry consistency that determines whether CCD will actually work. Resolution and SNR are measurable on a single pass. Geometry consistency requires two passes and is difficult to test in a factory acceptance environment. The result, consistently, is that CCD capability is fielded with navigation and flight management systems optimized for single-pass imaging, and the operational community discovers the gap only after deployment.

For any future UAS SAR design that includes CCD as a required mode, the requirements derivation process should explicitly flow down from the desired CCD performance — in terms of minimum detectable change magnitude and terrain type — to navigation accuracy requirements that distinguish position accuracy from attitude accuracy, to track-keeping requirements that specify cross-track deviation tolerance separately from waypoint navigation accuracy, and to processing requirements that specify autofocus convergence criteria and registration accuracy independently of image quality metrics. These are different requirements from those used for SAR imaging, they flow from different physics, and they demand different engineering solutions.

The good news is that the engineering solutions are now available, better understood, and more affordable than they were when Lynx was designed. Navigation-grade miniaturized FOG IMUs, multi-antenna GPS attitude systems, GPS-guided precision autopilots, and GPU-accelerated dense registration algorithms are all commercially available at cost and size points compatible with UAV payload integration. The window to bring Lynx CCD performance to the level its designers envisioned — and the DSTO team demonstrated was physically achievable — has never been more open.

· · ✦ · ·

Lynx SAR Coherent Change Detection LN-200 Navigation-Grade IMU iXblue UmiX Dual-Antenna GPS UAVSAR PPA Doerry Sandia DSTO Ingara UAS ISR GA-ASI Squint Angle Image Registration Aperture Trimming

Verified Sources & Formal Citations

[1]

Primary Technical Authority on CCD Requirements: A. W. Doerry, "SAR Data Collection and Processing Requirements for High Quality Coherent Change Detection," Sandia National Laboratories, SAND2008-0911C, 2008. The definitive public treatment of grazing angle, bearing angle, squint, registration, and SNR requirements for CCD. https://www.osti.gov/servlets/purl/1146354

[2]

Lynx Original Technical Description: S. I. Tsunoda et al., "Lynx: A High-Resolution Synthetic Aperture Radar," Sandia National Laboratories / General Atomics. Documents original Lynx specifications including 0.1 m spotlight resolution, CCD mode, LN-200 IMU, and Carrier-Phase GPS navigation. https://www.osti.gov/servlets/purl/4263

[3]

Lynx Multi-Mode Radar Current Product Specification: General Atomics Aeronautical Systems, Inc. (GA-ASI). SAR/GMTI, DMTI, MWAS, ACD, AMMOD modes described. https://www.ga-asi.com/radars/lynx-multi-mode-radar

[4]

Lynx Block 30 CCD and 1,000 Mission Hours: GA-ASI Press Release, May 7, 2010. Confirms CCD algorithms on Block 30, 1,000 collective mission hours on Sky Warrior in Iraq. https://www.ga.com/lynx-block-30-radar-surpasses-1000-mission-hours-on-sky-warrior-uas

[5]

DSTO Ingara CCD — Australian X-Band Experimental Work: M. Preiss and N. J. S. Stacy, "Coherent Change Detection: Theoretical Description and Experimental Results," Defence Science and Technology Organisation, Technical Report DSTO-TR-1851, Edinburgh, South Australia, 2006. Footprint detection results and statistical change detection framework. Available via DTIC: https://apps.dtic.mil/sti/tr/pdf/ADA458753.pdf

[6]

Preiss, Gray & Stacy — Log-Likelihood Change Statistic: M. Preiss, D. A. Gray, and N. J. S. Stacy, "Detecting Scene Changes Using Synthetic Aperture Radar Interferometry," IEEE Transactions on Geoscience and Remote Sensing, vol. 44, no. 8, pp. 2041–2054, 2006. Bayesian framework for the LLCS, precursor to the Berger NCB statistic used in the Lee & Kim fusion paper.

[7]

UAVSAR Precision Autopilot: NASA Dryden Flight Research Center / JPL, "Platform Precision Autopilot Overview and Mission Performance," NASA Technical Report, 2009. Validates ±2.5 m track-keeping over 342 data runs; establishes that manual piloting cannot meet repeat-pass interferometry track requirements. https://ntrs.nasa.gov/api/citations/20090029224/downloads/20090029224.pdf

[8]

UAVSAR System Description: JPL Earth Science / eoPortal. UAVSAR 10 m tube track specification, GPS-guided FMS, L-band polarimetric repeat-pass interferometry design. https://earth.jpl.nasa.gov/estd-missions/airborne/uavsar/

[9]

LN-200 IMU Specifications: Northrop Grumman. 0.5–1.0 deg/hr gyro bias classes, MEMS accelerometers, production since 1994, >40,000 units produced. https://www.northropgrumman.com/what-we-do/mission-solutions/assured-navigation/ln-200-fog-family-advanced-airborne-imu-ahrs

[10]

iXblue/Exail UmiX Series IMU: iXblue (now Exail). UmiX U5/U9 specifications: 0.03 deg/hr in-run bias, 0.008 deg/√hr ARW, Ø89 × 75 mm, 4W, ITAR-free, military qualified. https://www.ixblue.com/north-america/defense/inertial-measurement-units/

[11]

FOG Performance Classification: iXblue v. Safran, Partial Award, December 14, 2023, ICC Arbitration. Establishes that FOGs with 0.01 deg/hr bias stability are navigation/strategic grade; 0.5–1.0 deg/hr is tactical grade. https://jusmundi.com/en/document/decision/en-ixblue-sas-v-safran-electronics-defense-sas

[12]

Northrop Grumman LR-500 Quad Mass Gyro: Northrop Grumman. MEMS QMG near-FOG/RLG performance, LN-200 compatible digital interface, vibration-isolated. https://www.northropgrumman.com/what-we-do/mission-solutions/assured-navigation/lr-500-quad-mass-gyro-qmg

[13]

Honeywell HG3900 MEMS IMU: Honeywell Aerospace. All-silicon MEMS near-navigation-grade; meets or exceeds FOG/RLG performance; design verification 2026, production 2027. https://aerospace.honeywell.com/us/en/about-us/blogs/mems-the-word-meet-our-near-navigation-grade-tactical-imu

[14]

Dual-Antenna GPS Heading Theory and Application: Trimble GNSS/OEM. Moving Baseline RTK principle: centimeter-accurate baseline vector gives heading to 0.02–0.05 deg RMS from 1–2 m baseline; independent of aircraft dynamics. https://oemgnss.trimble.com/en/technologies/gnss-heading-systems

[15]

SAR CCD Processing Requirements — SNR and Registration: A. W. Doerry, Sandia National Laboratories. Clutter SNR ≥ 10 dB requirement; registration as the hardest processing step; autofocus iteration benefit. Same reference as [1] above.

[16]

Lynx Block 20A VideoSAR: GA-ASI Press Release, June 19, 2013. Identifies Block 20A as most advanced Lynx variant; VideoSAR mode demonstrated on King Air 200; confirms continued Lynx development. https://www.ga-asi.com/ga-asi-revolutionizes-sar-full-motion-video

[17]

Lee & Kim Fusion CCD Paper (Companion Article Subject): G. Lee and K.-T. Kim, "Unsupervised SAR Change Detection of Small Objects via Fusion of Difference Images," IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 19, pp. 10394–10409, 2026. DOI: 10.1109/JSTARS.2026.3674537.

[18]

3D SAR CCD Under Forest Canopy — Ingara Research Continuation: P. B. Pincus, M. Preiss, and N. J. S. Stacy, "3D SAR Coherent Change Detection for Monitoring the Ground Under a Forest Canopy," IET Radar, Sonar & Navigation, 2019. Extends Ingara CCD work to foliage-penetrating scenarios with vertical beamforming; also cites Lynx as related system. https://ietresearch.onlinelibrary.wiley.com/doi/abs/10.1049/iet-rsn.2018.5641