🛰 SAR · Satellite Imagery
Foundations — the physics of signal & image

Range & azimuth resolution

ΔR=c/2BTOPS trade-offMultilookwiki/sar-resolution
TL;DR

The two directions of a SAR pixel are governed by completely different principles. Range (front-back separation) is set by pulse bandwidth, expressed as ΔR=c/2B, while Azimuth (side separation) is set by the synthetic aperture and is effectively tied to the antenna length. On Sentinel-1 IW, Range (~5m) is far better than Azimuth (~20m), giving a rectangular 5m×20m pixel; this is because TOPS mode trades away Azimuth resolution to image a 250km-wide swath (SM mode is ~5m×5m). Because of this asymmetry, InSAR multilook for speckle reduction is usually applied as 3×1, averaging mainly in the Range direction.

Defining the two directions

  • Range resolution measures how close two objects can be along the radar line-of-sight while still being separated (front-back separation).
  • Azimuth resolution measures how well two objects are separated along the satellite's flight direction, that is, the sideways direction.
↕️AzimuthFlight direction, side separation
↔️RangeLine-of-sight, front-back separation
The two axes of a SAR pixel

Range resolution — set by bandwidth

  • Like a bat's ultrasound, if two echoes arrive almost simultaneously they sound as one, and if there is a time gap they sound as two. SAR separates objects by this time difference, where near objects return early and far objects return late.
  • How finely this time gap can be resolved depends on bandwidth, so wider bandwidth yields shorter pulses and lets closer objects be separated.
  • Sentinel-1's roughly 56MHz bandwidth corresponds to a theoretical ~2.7m resolution, while the actual method of producing short pulses is handled by Chirp and range compression.
Range-direction resolution
Speed of light
Pulse bandwidth
Range resolution is inversely proportional to bandwidth

Azimuth resolution — set by the synthetic aperture

  • A small antenna spreads its beam too widely to separate the side direction, so Sentinel-1's mere 12.3m antenna observes the same target hundreds to thousands of times while moving and phase-coherently sums them, building a virtual antenna several km long via the synthetic aperture.
  • As a result, Azimuth resolution is effectively tied to the real antenna length, approximated as Δ_az ≈ L/2 (L being the real antenna length).
  • L/2 ≈ 6m is only the theoretical approximation for Stripmap (SM) mode; IW uses TOPS mode, which electronically swings the beam in azimuth to image a 250km-wide swath and thereby sacrifices Azimuth resolution down to ~20m.
  • In other words, the actual resolution is not set by antenna length alone but by antenna, TOPS scanning, processing, and operating mode combined, yielding ~20m for IW.

Comparison across modes and multilook implications

  • On Sentinel-1 IW, Range is far better than Azimuth. Range is good thanks to chirp and bandwidth (56MHz allows an actual 2~3m, becoming ~5m in products after multilook, resampling, and GRD), while Azimuth is poor because the TOPS design prioritized a 250km-wide swath over high resolution.
  • Seeing a rectangular SLC pixel in SNAP is the normal result of Range 5m / Azimuth 20m, so there is no need to be alarmed.
  • Since Azimuth is already poor at 20m, smearing it further would drop resolution excessively, so InSAR multilook for speckle reduction is usually applied as 3×1, averaging mainly in Range.
RangeAzimuth
What it determinesFront-back separationSide separation
Governing factorPulse bandwidth (ΔR=c/2B)Synthetic aperture (Δ_az≈L/2)
S1 IW value~5m~20m
S1 SM value~5m~5m
Range vs Azimuth, and values by mode
Pitfalls & gotchas

Do not naively apply the L/2 ≈ 6m formula to the 12.3m antenna and expect IW Azimuth to be 6m. L/2 is only the theoretical approximation for Stripmap (SM) mode; IW uses TOPS mode, which swings the beam in azimuth to gain a 250km-wide swath and thereby sacrifices resolution to ~20m.

The topic cards on this page are compiled from the Brain Trinity wiki. The original wiki can be demoed live in an interview.Back to study log