Find your next learning step
A focused SAR image combines echoes collected across an aperture. Splitting that aperture into different Doppler or squint looks creates repeated views with different phase sensitivity. Their interferometric differences can expose small azimuth misregistration and residual platform-motion error. The same sensitivity also responds to moving terrain or targets, so separation comes from modelling, redundancy and stable references—not from a single colourful phase map.
Open the chapter contents · Radar root · Gold Standard of Education · How X Works
Chapter contents
Choose the route closest to the decision, or read from Chapter 1. The chapters remain open; these grouped routes are a navigable index.
Route 1 · Define the decision
Route 2 · Work the ledgers
Route 3 · Transfer the method
Route 4 · Check limits and evidence
Route 5 · Ask and continue
CHAPTER 1 OF 12 · DEFINE THE DECISION
1. Begin with azimuth frequency, not image colour
Azimuth position and Doppler frequency are Fourier partners in SAR processing. A small relative time shift Δη between two images creates a phase ramp with azimuth frequency. Comparing separated Doppler looks converts that ramp into a measurable phase difference.
Multi-squint processing divides the azimuth spectrum into two or more sub-apertures. Each sub-aperture sacrifices some azimuth resolution but supplies an independent view of the same registration or motion error.
CHAPTER 2 OF 12 · DEFINE THE DECISION
2. Define the spectral-diversity observable
For two looks with Doppler centroids f₁ and f₂, an ideal constant azimuth time shift produces a differential phase approximately Δφ=2π(f₂−f₁)Δη, modulo 2π. Sign depends on the Fourier and interferogram conventions.
This relationship is local only when the shift varies slowly compared with the analysis window. Phase wrapping, decorrelation and different scatterer mixtures can make a direct pixel-by-pixel inversion unreliable.
CHAPTER 3 OF 12 · WORK THE LEDGERS
3. Work a two-squint phase ledger
Take looks centred at −250 Hz and +250 Hz, so their separation is 500 Hz. If the second image is shifted by 0.0004 s in azimuth, the expected differential phase is 2π×500×0.0004=1.257 rad, about 72 degrees.
| Look or difference | Doppler centre | Shift sensitivity | Expected phase |
|---|---|---|---|
| Backward squint | −250 Hz | 2πf₁Δη | −0.628 rad |
| Forward squint | +250 Hz | 2πf₂Δη | +0.628 rad |
| Forward−backward | 500 Hz | 2πΔfΔη | 1.257 rad |
| Recovered shift | — | Δφ/(2πΔf) | 0.400 ms |
CHAPTER 4 OF 12 · WORK THE LEDGERS
4. Connect a shift estimate to residual motion error
After ordinary motion compensation, navigation and trajectory errors can leave a baseline error varying along the aperture. That error perturbs phase history and causes azimuth misregistration between repeat-pass images. Multi-squint estimates the relative shift and can support a model of the residual baseline.
The inversion needs geometry: platform velocity, range, look angle and processing convention link an image-domain shift to a physical trajectory correction. A phase-to-shift conversion alone is not yet a motion solution.
CHAPTER 5 OF 12 · WORK THE LEDGERS
5. Change the look separation and expose ambiguity
If the centroid separation doubles to 1000 Hz while Δη stays 0.4 ms, the unwrapped phase doubles to 2.513 rad. Sensitivity improves, but the unambiguous shift interval shrinks because phase is observed modulo 2π.
| Change | Immediate effect | New risk | Check |
|---|---|---|---|
| Larger Δf | More phase per shift | Earlier 2π wrapping | Coarse offset prior |
| Narrower looks | Cleaner centroid distinction | Lower azimuth resolution | Point response |
| More looks | Overdetermined slope | Unequal coherence | Weighted fit |
| Larger window | Lower phase noise | Smears local changes | Residual map at two scales |
CHAPTER 6 OF 12 · TRANSFER THE METHOD
6. Separate platform error from scene motion
Platform residuals often vary smoothly along track and affect stable terrain coherently. Glacier flow, ocean current or moving targets can create genuine along-track displacement that also enters the azimuth offsets. Treating all measured shift as navigation error can erase the signal of interest.
Use stable control areas, multiple viewing geometries, external orbit information and spatial models. Separation becomes credible when one component follows platform geometry while another follows the moving scene and remains consistent in held-out looks.
CHAPTER 7 OF 12 · TRANSFER THE METHOD
7. Connect the established radar routes
The dispersive-shift guide combines range split-spectrum and azimuth evidence for ionospheric effects. The range-migration guide explains how motion couples into long coherent apertures.
Use the scintillation guide when coherence loss limits phase estimation, and return to the Radar root for Doppler and interferometry foundations.
CHAPTER 8 OF 12 · TRANSFER THE METHOD
8. Fit more than two looks when the data allow
With several squints, regress phase against Doppler centroid after unwrapping. The slope estimates Δη while the intercept absorbs a common phase term. Robust or coherence-weighted fitting can reduce the influence of damaged looks.
Inspect residual curvature. A straight line is the constant-shift model; systematic curvature can indicate higher-order motion error, defocus, ionospheric structure or an inconsistent look construction.
CHAPTER 9 OF 12 · CHECK LIMITS AND EVIDENCE
9. Know the limits and common misconceptions
Multi-squint is not a universal autofocus button. It relies on coherent common scene content and an interpretable phase model. Strong temporal decorrelation, azimuth ambiguities or rapidly moving targets can invalidate the estimator.
A smooth estimated correction is not proof of platform origin. Regularisation can manufacture smoothness, and broad scene motion can also be smooth. Preserve the raw look phases and independent controls.
CHAPTER 10 OF 12 · CHECK LIMITS AND EVIDENCE
10. Use an evidence and source routine
Record acquisition geometry, Doppler convention, aperture partition, look bandwidths and centres, common-band filtering, coregistration, coherence mask, unwrap method, phase-to-shift equation, ambiguity interval, stable reference mask, motion model, residuals and withheld validation.
Macedo and colleagues describe multi-squint spectral-diversity estimation for residual motion error. Prats and colleagues show why terrain motion can bias that estimate and develop extended multi-squint separation for glacier velocity. The DLR ionospheric study combines split-spectrum and multi-squint observables.
CHAPTER 11 OF 12 · ASK AND CONTINUE
11. Frequently asked questions
Why use more than two squints?
Extra looks test the expected linear phase–frequency relation and help identify outliers or curvature.
Does larger Doppler separation always help?
No. It improves sensitivity but narrows the unambiguous shift range and can reduce common coherent support.
Can a glacier shift be mistaken for aircraft motion?
Yes. Stable terrain, geometry and multiple observables are needed to separate them.
CHAPTER 12 OF 12 · ASK AND CONTINUE
12. Continue with an ambiguity transfer check
Repeat the ledger with a 1000 Hz look separation and a 0.7 ms shift. Compute the unwrapped phase, reduce it modulo 2π, then list the candidate shifts consistent with the wrapped result. Explain how a coarse image-offset estimate selects the correct ambiguity branch.
Continue through the Radar root, the Gold Standard of Education and the How X Works hub.
Top · Contents · Previous · How X Works