A satellite antenna cannot be made arbitrarily large. Synthetic aperture radar solves a geometric problem by using the spacecraft's motion to collect a coherent sequence of echoes from each ground target, then computationally focusing those observations.
How a synthetic aperture is formed
As a platform moves along track, a target remains within the antenna's illuminated footprint across many pulses. Its range changes predictably; that phase history is recorded. Focusing algorithms align and combine the echoes so that their contributions reinforce at the target location and cancel elsewhere. The result is an effective along-track aperture much longer than the physical antenna. SAR therefore improves azimuth resolution while still depending on orbital geometry, antenna pattern, bandwidth, sampling and signal-to-noise ratio.
Two directions, two constraints
Range resolution depends primarily on the transmitted signal bandwidth: for a simple chirped pulse in slant range, the nominal resolution is approximately c / (2B), where c is the speed of light and B is bandwidth, before ground-projection and processing details. Azimuth resolution comes from synthetic aperture focusing and depends on the system geometry and processing. Pixel spacing is not necessarily the same as independent resolution; oversampling can place samples more densely without creating equivalent new detail.
Noise, antenna characteristics, quantisation, calibration error and focusing quality all influence what survives in the product. Incoherent multilooking can reduce speckle variance but also changes spatial detail; complex SLC data preserve phase needed for interferometric work, while detected GRD products serve different analyses. The appropriate product depends on the question.
Augmentation without invention
Research can test improved radiometric calibration, sub-pixel registration, speckle statistics, multitemporal integration, phase-preserving workflows and uncertainty-aware super-resolution. Each proposal should be compared to a documented baseline using independent reference data and relevant metrics. Computation can recover structure obscured by noise or exploit redundancy when the sampling supports it; it cannot guarantee arbitrary recovery of frequencies the instrument never sampled.
That boundary is productive. It distinguishes extracting more usable information from an existing measurement from manufacturing plausible-looking detail.